Role of MEK partner-1 in cancer stemness through MEK/ERK pathway in cancerous neural stem cells, expressing EGFRviii

Background Glioma stem cells (GSCs) are a major cause of the frequent relapse observed in glioma, due to their high drug resistance and their differentiation potential. Therefore, understanding the molecular mechanisms governing the ‘cancer stemness’ of GSCs will be particularly important for improving the prognosis of glioma patients. Methods We previously established cancerous neural stem cells (CNSCs) from immortalized human neural stem cells (F3 cells), using the H-Ras oncogene. In this study, we utilized the EGFRviii mutation, which frequently occurs in brain cancers, to establish another CNSC line (F3.EGFRviii), and characterized its stemness under spheroid culture. Results The F3.EGFRviii cell line was highly tumorigenic in vitro and showed high ERK1/2 activity as well as expression of a variety of genes associated with cancer stemness, such as SOX2 and NANOG, under spheroid culture conditions. Through meta-analysis, PCR super-array, and subsequent biochemical assays, the induction of MEK partner-1 (MP1, encoded by the LAMTOR3 gene) was shown to play an important role in maintaining ERK1/2 activity during the acquisition of cancer stemness under spheroid culture conditions. High expression of this gene was also closely associated with poor prognosis in brain cancer. Conclusion These data suggest that MP1 contributes to cancer stemness in EGFRviii-expressing glioma cells by driving ERK activity. Electronic supplementary material The online version of this article (doi:10.1186/s12943-017-0703-y) contains supplementary material, which is available to authorized users.


Background
A subset of cell populations showing specific biological properties associated with high tumorigenicity and cellular plasticity has been designated cancer stem cells (CSCs) or tumor-initiating cells (TICs). There is emerging evidence that CSCs in glioma [glioma stem cells (GSCs), or glioma stem-like cells (GSLCs)] may be the main cause of poor clinical outcomes due to their resistance to chemo-and radiotherapy [1][2][3]. Therefore, GSCs are considered an important target for glioma therapy [4], and studies to understand the biology of GSCs are being actively undertaken [5,6].
Recently, it has been demonstrated that neural stem cells (NSCs) undergoing neoplastic transformation are responsible for glioma formation [7][8][9]. NSCs are more susceptible to oncogenic transformation than differentiated glial cells [7,10]. For example, immortalized human fetal NSCs established by the introduction of v-myc (HB.F3 cells) [11], which are widely used as a model system to examine the possible therapeutic effects of genetically modified NSCs or glial cells [12][13][14], undergo oncogenic transformation by H-Ras (forming cancerous neural stem cells (CNSCS): F3.Ras cells) [10] but not Akt [15]. In particular, oligodendrocytes derived from F3 cells are resistant to oncogenic transformation by H-Ras [10], implying that NSCs are more susceptible to oncogenic transformation than glial cells (e.g., astrocytes and oligodendrocytes). Importantly, F3.Ras cells have similar molecular properties as GSCs [6], implying that common mechanisms control neural cancer stemness in both GSCs and CNSCs.
Amplification and gain-of-function mutations of epidermal growth factor receptor (EGFR) are the most common genetic alteration in the brain cancers, with a frequency of approximately 40% in glioma [16]. In particular, the type III EGFR mutation (also called EGFRviii, or del2-7 EGFR) results in an in-frame deletion of 267 amino acids from the extracellular domain of EGFR, resulting in constitutive activation [17]. This is the most frequent genetic mutation in glioblastoma multiforme (GBM), with an overall prevalence of 20-30%. Although RAS, a downstream effector of EGFR, is one of the most frequently mutated oncogenes in many types of cancer [18], mutations of RAS in glioma are relatively rare [19]. Instead of RAS mutations, loss-offunction mutations are observed in neurofibromin 1 (NF1), a negative regulator of Ras signaling toward MEK/ ERK1/2 [19]. Consistent with this model, mice with mutations in Nf1 and Trp53 develop astrocytoma [20] with stem cell characteristics [21]. Similarly, dual knockout of phosphatase and tensin homolog (Pten) and Trp53 results in a high-grade malignant glioma that resembles primary human GBM and shows increased NSC self-renewal capacity [22]. Notably, Akt activation due to PTEN loss of function [23] and MEK/ERK1/2 activation are both important for the self-renewal and tumorigenicity of GSCs [24]. However, the differences between Akt and MEK/ ERK1/2 downstream of EGFR activation have remained less clear in glioma and GSCs.
Late endosomal/lysosomal adaptor, MAPK and MTOR activator 3 (LAMTOR3), which encodes MEK partner-1 (MP1), was initially identified as a scaffolding protein for MEK1 and ERK1 that enables ERK1 activation [25]. It has been well characterized as a key regulator of endosomal signaling [26], and a role for MP1 in cancer, via MEK and ERK hyperactivation, has recently been demonstrated in pancreatic tumorigenesis [27].
In this study, another type of CNSCs was established by expression of EGFRviii in F3 NSCs (F3.EGFRviii cells). Although EGFRviii signaled through Akt under adherent culture conditions, EGFRviii signaling through MEK/ERK1/2 became predominant when the cells were cultured under spheroid-inducing conditions, which promote neural cancer stemness. In addition, MP1 was shown to mediate this switch in signaling from Akt to ERK1/2, and therefore promoted the phenotype of neural cancer stemness in F3.EGFRviii cells. Finally, MP1 expression was strongly associated with poor survival in human glioma patients.

Quantitative real-time PCR
Total RNA was extracted from cells using Total RNA Extraction Kit (Intron, cat# 17061), and then converted to cDNA using PrimeScript RT Master Mix (Takara, cat# RR036) in accordance with the manufacturer's instruction. The synthesized cDNAs were used as templates to perform the real-time PCR with Light Cycler 480 Instrument II (Roche) using SYBR Premix Ex Taq (Takara, cat# RR420) under the following conditions: denaturation at 95°C for 30 s, followed by 40 cycles of 95°C for 5 s, 58°C for 15 s, and 72°C for 20 s. The average threshold cycle for each gene was determined from triplicate reactions and then levels of gene expression relative to TATA-binding protein (TBP) or Ribosomal protein L13a (RPL13A) were determined. Primer pairs were listed on the following table (Additional file 1: Table S1). ERK dependent gene expression was demonstrated by RT 2 profiler PCR array human MAP kinase Signaling Pathway as described by the manufacturer. The reactions were carried out in an Applied Biosystems 7900HT Fast-Real Time PCR System.

TCGA analysis
The DNA copy number, mRNA expression and clinical data obtained from about 500 GBM patients were downloaded from the TCGA data portal (https://tcga-data.nci.nih.gov/). Gene expression data were generated by the Agilent microarray chips, and multiple probes were averaged to get a single expression value per gene. DNA copy number data were generated by the Affymetrix SNP6.0 chips, and the segmented copy numbers were averaged by gene. The samples with EGFR amplification were defined by both upregulated mRNA expression levels (≥2folds) and high DNA copy numbers (≥ 8) of EGFR gene. To compare mRNA expression levels of LAMTOR3 between EGFR-amplified and EGFR-normal samples, t-test was used. Kaplan-Meier survival analysis and log-rank test were performed to estimate and compare survivals of glioblastoma patients by LAMTOR3 mRNA expression levels. The glioblastoma patients were grouped by the LAMTOR3 expression levels, which were divided into 3 equal intervals; high, med and low.

Statistical analysis
Graphical data were presented as mean ± S.D. Statistical significance among three groups and between groups were determined using one-or two-way analysis of variance (ANOVA) following Bonferroni multiple comparisons post-test and Student's t-test, respectively. Significance was assumed for P < 0.05(*), P < 0.01(**), and P < 0.001(***).

Establishment of EGFRviii-expressing human neural stem cells
Doxycycline (Dox)-inducible EGFRviii was stably established in F3 cells to generate Dox-inducible F3.EGFRviii cells, as described previously [10]. As shown in Fig. 1, even without Dox treatment, F3.EGFRviii cells showed morphological changes compared with F3 cells (Fig. 1a), due to slight leakage of EGFRviii expression in the absence of Dox (Additional file 2: Fig. S1). However, the morphology, typical of transformed cells (e.g., condensed cytoplasm, focus formation, and loss of contact inhibition) was clearly observed in F3.EGFRviii cells in response to Dox treatment, as previously reported (Fig. 1a, red line) [10]. That these alterations in cellular morphology were only observed after constant Dox treatment indicates that the low level of EGFRviii leakage in the absence of Dox (Additional file 2: Fig. S1) was insufficient to induce oncogenic transformation of F3.EGFRviii cells.
As expected, Dox treatment dramatically induced EGFRviii mRNA (Fig. 1b) and protein (Fig. 1c). When EGFRviii expression was induced by Dox treatment, active phosphorylation of EGFR (pY1068) was observed at the plasma membrane (white arrows, Fig. 1d). Therefore, we further examined the two crucial downstream pathways of EGFRviii, the PI3K/Akt and MEK/ERK1/2 pathways, by measuring the levels of phosphorylated Akt and ERK1/2, respectively. Akt was activated regardless of Dox treatment, which might result from leakage of EGFRviii expression (Additional file 2: Fig. S1). MEK/ ERK1/2 activation became apparent after Dox treatment, along with EGFRviii expression (Fig. 1e).

High expression of LAMTOR3 in F3.EGFRviii spheres for ERK1/2 activation
To examine the molecular mechanism underlying ERK1/2 activation in F3.EGFRviii spheres, the activity of MEK1/2, the sole upstream kinase for ERK1/2 [47], was examined. Surprisingly, the level of phosphorylation of MEK1/2 was similar between F3.EGFRviii spheres and adherent cells, although ERK1/2 activation was clearly enhanced in F3.EGFRviii spheres (Fig. 4a). We hypothesized that an increase in the expression of genes involved in the MEK/ ERK1/2 signaling axis might be responsible for the increased ERK1/2 signaling in F3.EGFRviii spheres. To address this possibility, F3.EGFRviii cells were analyzed using a PCR array for the human MAPK signaling pathway to identify gene(s) involved in regulating the MEK/ERK1/2 signaling axis that was specifically altered in F3.EGFRviii spheres [48]. Among the 84 genes evaluated, 14 genes showed a significant change in expression in F3.EGFRviii spheres compared with adherent cells (Fig. 4b and Additional file 5: Fig. S4A). Next, to narrow down the possible candidates among those 14 genes, we searched the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) and chose three independent GSE studies, in primary brain cancer (GSE23806), primary glioblastoma (GSE15824), and neuroblastoma (GSE44537), that compared the gene expression signatures of glioma or GSCs with those of normal cells or glial tumors, respectively (Additional file 5: Fig. S4B). Among 578 genes with significantly altered expression [p < 0.05, 2-fold higher (shown in red) or lower (shown in blue)], putative gene candidates that may be responsible for ERK1/2 activation in F3.EGFRviii spheres were further refined based on the gene ontology term 'protein complex scaffold.' Four putative gene candidates [two downregulated genes (blue) and two upregulated genes (red)] were identified (Fig. 4c), and the relative expression levels of the four candidates in F3.EGFRviii spheres were confirmed in parallel with SOX2, which serves as a positive control for sphere formation (Additional file 5: Fig. S4C). Importantly, significant induction of LAMTOR3, the gene that encodes MEK partner-1 (MP1), which activates ERK1 [25] and ERK1/2 [27], was upregulated in F3.EGFRviii spheres in our PCR array (Fig. 4b), and was also upregulated in GSCs (or spheres) (Fig. 4c). In an independent experiment, the upregulation of LAMTOR3 mRNA (Fig. 4d) and MP1 protein was confirmed in F3.EGFRviii spheres in parallel with high ERK1/2 phosphorylation and SIRT1 expression (Fig. 4e). Intriguingly, a positive correlation between LAMTOR3 and SIRT1 was also confirmed in a clinicogenomics database of GBM (http://betastasis.com/ glioma/rembrandt) (Additional file 5: Fig. S4D). Additionally, analysis of a dataset of gene expression in GSCs compared with normal glioma cell lines (GSE23806) revealed that LAMTOR3 expression was significantly higher in glioma cells with stemness (i.e., both glioma cells established by neurosphere culture and GSCs) in parallel with high FABP7 and SOX2 expression than in normal glioma cell lines (Additional file 5: Fig. S4E).

LAMTOR3 expression and ERK activation in F3.EGFRviii cells
Considering the high expression of MP1 in F3.EGFRviii spheres, which showed high ERK1/2 activity, we hypothesized that MP1 expression may contribute to the ERK1/2 activation. To test this hypothesis, ERK1/2 activity was evaluated after depletion of MP1 from F3.EGFRviii cells. Among two different shRNA sequences for LAMTOR3, shLAM-TOR#2 resulted in the clearest downregulation of the MP1 protein level (Fig. 5a and Additional file 6: Fig. S5A). As predicted, loss of MP1 expression was sufficient to significantly decrease the level of ERK1/2 phosphorylation in F3.EGFRviii cells (Fig. 5b). Importantly, the loss of MP1 resulted in only marginal effects on clonogenic growth (Additional file 6: Fig. S5B), anchorage-independent growth potential (Additional file 6: Fig. S5C), and the expression of GLI1, OLIG2, and SIRT1 when F3.EGFRviii cells were grown in adherent cultures (Additional file 6: Fig. S5D). Next, cells were treated with PMA to stimulate Akt and MEK/ERK1/2 simultaneously, similar to the experiments shown in Fig. 3f. Interestingly, while MEK1/2 signaling remained intact after PMA treatment, activation of ERK1/2 and its downstream target S6 kinase (RSK) was clearly diminished (Fig. 5c). Activation of other MAPKs, such as p38 and JNK, and Akt signaling was only marginally affected by the loss of MP1 expression (Fig. 5d). These data suggest that MP1 expression may be critical for the shift in signaling preference toward ERK1/2 during F3.EGFRviii sphere formation, as shown in Fig. 3f.

Role of LAMTOR3 in F3.EGFRviii spheres
Considering the clear ERK1/2 activation during sphere formation in F3.EGFRviii cells and the role of MP1 in ERK activation in those cells, we speculated that the Fig. 4 High expression of LAMTOR3 in F3.EGFRviii spheres is associated with ERK1/2 activation. a The levels of pAKT, pMEK1/2, and pERK1/2 in adherent and spheroid F3.EGFRviii cells were determined by immunoblot analysis, using α-tubulin as a loading control. b Expression profile analysis of a MAPK signaling-associated gene set through MAPK pathway superarray in F3.EGFRviii cells and F3.EGFRviii spheres (red, genes upregulated >2-fold; blue, genes downregulated >2-fold; gray, no significant change). The black arrow indicates LAMTOR3. c GEO analysis to deduce candidate genes that distinguish signaling in F3.EGFRviii spheres. Three independent GSE studies, two using primary brain tumor cells (GSE23806, GSE15824) and one using induced cancer stem cells from neuroblastoma samples (GSE44537), were used to select commonly altered genes. Candidate genes were narrowed down using gene ontology for 'protein complex scaffold' genes. Red and blue indicate upregulated or downregulated genes, respectively. d The expression of LAMTOR3 in adherent F3.EGFRviii cells and F3.EGFRviii spheres was determined by real-time PCR (n = 3). e The levels of MP1, SIRT1, and pERK1/2 were determined by immunoblotting in adherent and spheroid F3.EGFRviii cells using α-tubulin as a loading control decrease in ERK1/2 activation caused by the depletion of MP1 might impair their sphere formation. To test this hypothesis, the sphere-forming capacity of F3.EGFRviii cells was evaluated after depletion of MP1. As predicted, LAMTOR3 knockdown by either siRNA (Additional file 6:   Fig. S6B). Furthermore glial tumor markers such as OLIG2 and GLI1 and GSC markers such as MSI1 and SOX2 were also significantly reduced in F3.EGFRviii spheres (Fig. 6b). SIRT1, the expression of which is important for neural stemness as well as neural cancer stemness and survival [6], was also markedly reduced at both the mRNA (Fig. 6c) and protein (Fig. 6d) level by depletion of MP1.

Prognostic significance of LAMTOR3 in brain tumors
To investigate the prognostic significance of MP1, a survival analysis was performed in high-grade glioma with data from a publicly available clinicogenomics study (GSE4271) [49] using DRUGSURV (http://www.bioprofiling.de/cgi-bin/GEO/DRUGSURV/start_GENE.pl). As shown in Fig. 7a, high LAMTOR3 expression was largely correlated with poor clinical outcome in 77 patients (p = 0.122). The clinical significance of LAMTOR3 in relation to EGFR amplification was further investigated using a large GBM dataset from The Cancer Genome Atlas (TCGA). The glioblastoma samples were categorized into two subgroups based on the DNA copy number and mRNA expression level of EGFR. Among 500 glioblastoma samples in TCGA, 173 samples were categorized as EGFR-amplified samples (red circle), and 246 samples (green circle) with normal or low DNA copy numbers and EGFR expression were categorized as EGFRnormal samples (Fig. 7a). LAMTOR3 expression was significantly higher in the EGFR-amplified group (p = 0.0002) (Fig. 7b). The effects of LAMTOR3 expression on survival also differed according to EGFR status. In the EGFR-amplified group, the probability of survival was significantly better in samples with low LAMTOR3 expression, but there were no significant differences in survival according to LAMTOR3 expression among the samples without EGFR amplification (Fig. 7c).

Discussion
Previously, we established CNSCs from F3 cells using the H-RasV12 oncogene [10], and demonstrated that SIRT1 expression is important for maintaining neural cancer stemness in CNSCs and GSCs [6]. Considering the frequent gain-of-function mutations of EGFR in glioma, characterization of CNSCs expressing EGFRviii is important for understanding the molecular properties of neural cancer stemness.
F3.EGFRviii cells showed the typical transformed cell phenotypes, such as clear morphological changes (Fig. 1a), accelerated growth (Fig. 2a), and anchorage-independent growth (Fig. 2c), and readily formed spheres with high induction of neural cancer stemness markers including SOX2, SIRT1, and OLIG2. Interestingly, while activation of Akt rather than MEK/ERK1/2 following EGFRviii expression appeared to be dominant in adherent cells (Fig. 1e), ERK1/2 was highly activated when F3.EGFRviii cells were subjected to sphere formation (Fig. 3e), suggesting that neural cancer stemness is associated with a switch in preference from Akt signaling to ERK1/2 signaling after EGFR activation. Induction of LAMTOR3 was found in GSCs or GSLCs, which have been described as cells with neural cancer stemness (Fig. 4c and Additional file 5: Fig. S4E), and was also identified as an important mediator for ERK1/2 activation following EGFR stimulation (Fig. 5). Thus, depletion of LAMTOR3 in F3.EGFRviii cells decoupled EGFR signaling from ERK1/2, thereby impairing sphere formation and attenuating the expression of genes associated with neural cancer stemness (Fig. 6), despite only marginal effects on signaling to MEK1/2 and other MAPKs (Fig. 5c and d).
Considering that the PI3K/Akt and MEK/ERK1/2 pathways are two major downstream signaling pathways of activated EGFR in glioma [50], therapeutic approaches for the inhibition of these signaling pathways are being actively studied [24]. Indeed, ERK activity has been found to be highly elevated in glioma under EGFR amplification [51], while Akt activation is also frequent [52] and has been shown to be important for neural cancer stemness [53]. However, it is still unclear how ERK1/2 activity remained elevated despite clear downregulation of EGFRviii in F3.EGFRviii sphere (Fig. 3e). To resolve this discrepancy, MAP kinase signaling pathway PCR array was applied. Of interest, V-Mos Moloney Murine Sarcoma Viral Oncogene Homolog (MOS), which is responsible for ERK activation as a MAPK kinase kinase (MAPK3K) level in meiosis [54], was upregulated in F3.EGFRviii sphere (Additional file 5: Fig. S4A) as consistent as previous reports the high expression of MOS in ependymal glioma [55] and astrocytic tumors [56]. Alternatively, LAMTOR3, enabling ERK1 activation [25] as a scaffolding protein was highly upregulated in F3.EGFRviii sphere (Additional file 5: Fig. S4A). Notably, recent studies reveal that LAMTOR3 expression mediates ERK1/2 activation independently of mutations in RAS, thereby contributing to pancreatic tumorigenesis [27]. This finding implies that LAMTOR3 controls oncogenic signals in addition to mediating endosomal signaling [57]. In addition, recent studies show that ERK1/2 activity promotes chemo-and radio-resistance in GBM [51]. Consistently, poor prognosis in patients with high LAMTOR3 was even more prominent after chemotherapy (http:// betastasis.com/glioma/rembrandt, data not shown).
The finding that neural cancer stemness correlates with a clear signaling preference for either Akt or ERK following EGFRviii activation suggests that the dependency of GBM for either Akt or ERK may depend on the expression of LAMTOR3. TCGA analysis revealed that LAMTOR3 expression in GBM patients was significantly higher when accompanied by gain-of-function mutations in EGFR, and was correlated with poor prognosis (Fig. 7). These findings suggest that high LAMTOR3 expression in GBM may promote EGFR downstream signaling through ERK.
Collectively, these studies used a model of EGFRviiiexpressing CNSCs to reveal novel insights into the molecular mechanisms involved in maintaining neural cancer stemness.

Conclusions
Using F3.EGFRviii CNSCs model, we demonstrated LAMTOR3 in GBM served as an important signaling mediator to control MEK1/ERK1/2 pathway, of which activation contributed to maintaining 'neural cancer stemness'.

Additional files
Additional file 1: Table S1.