| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Articles: Therapeutics, Targets, and Development
MAP-ing glioma invasion: Mitogen-activated protein kinase kinase 3 and p38 drive glioma invasion and progression and predict patient survival
Cancer and Cell Biology Division, Translational Genomics Research Institute, Phoenix, Arizona
Requests for reprints: Michael E. Berens, Translational Genomics Research Institute, 445 North Fifth Street, Phoenix, AZ 85004. Phone: 602-343-8760; Fax: 602-343-8844. E-mail: mberens{at}tgen.org
Abstract
Although astrocytic brain tumors do not metastasize systemically, during tumorigenesis glioma cells adopt an invasive phenotype that is poorly targeted by conventional therapies; hence, glioma patients die of recurrence from the locally invasive tumor population. Our work is aimed at identifying and validating novel therapeutic targets and biomarkers in invasive human gliomas. Transcriptomes of invasive glioma cells relative to stationary cognates were produced from a three-dimensional spheroid in vitro invasion assay by laser capture microdissection and whole human genome expression microarrays. Qualitative differential expression of candidate invasion genes was confirmed by quantitative reverse transcription-PCR, clinically by immunohistochemistry on tissue microarray, by immunoblotting on surgical specimens, and on two independent gene expression data sets of glial tumors. Cell-based assays and ex vivo brain slice invasion studies were used for functional validation. We identify mitogen-activated protein kinase (MAPK) kinase 3 (MKK3) as a key activator of p38 MAPK in glioma; MKK3 activation is strongly correlated with p38 activation in vitro and in vivo. We further report that these members of the MAPK family are strong promoters of tumor invasion, progression, and poor patient survival. Inhibition of either candidate leads to significantly reduced glioma invasiveness in vitro. Consistent with the concept of synthetic lethality, we show that inhibition of invasion by interference with these genes greatly sensitizes arrested glioma cells to cytotoxic therapies. Our findings therefore argue that interference with MKK3 signaling through a novel treatment combination of p38 inhibitor plus temozolomide heightens the vulnerability of glioma to chemotherapy. [Mol Cancer Ther 2007;6(4):1212–22]
Introduction
Gliomas are the most common primary intracranial tumors, causing 15,000 deaths per year in the United States (1). Despite modern diagnostics and treatment, there have not been major improvements in survival over the last 30 years, resulting in a static median survival time of 
15 months. Young patient age and high Karnofsky performance status contribute to slightly better prognosis (2). It is thought that the invasive behavior of glioblastoma multiforme cells that are left behind after debulking surgery is one of the most important causes for poor clinical outcome, enabling tumor cells to actively egress from the main mass and invade into the surrounding normal brain where they are out of reach of surgical resection, radiation, and chemotherapy (3). Interactions between tumor cells, extracellular matrix, and adjacent stromal cells in conjunction with mechanical and biochemical events support active cell movement and drive glioma cell invasion into normal tissue (4). The operational fluctuation between invasion and proliferation as a working hypothesis in malignant glioma pathobiology is supported by the finding that tumor recurrence often occurs within 3 cm of the resection cavity, substantiating the dispersing glioma cells as the source of recurrent tumors (5).
In an effort to discover and validate gene products driving malignant invasion, we used an in vitro system capable of supporting active cell movement and invasive behavior similar to that reported for glioblastoma multiforme (6). A three-dimensional invasion assay was used where multicellular glioma spheroids are suspended in self-assembling collagen I gel (7), which provides a three-dimensional meshwork allowing for invasion under biochemically well-controlled conditions. Laser capture microdissection (LCM) and whole genome expression profiling were used to identify genes differentially expressed between invasive and stationary glioma cells. For the first time, we report mitogen-activated protein kinase (MAPK) kinase 3 (MKK3) to be up-regulated with glioma invasion in vitro and in vivo. The MKK3/6 pathway is reported as the major activator of p38
and has been described to be autoactivated through interaction with TAB1 in mouse embryonic fibroblasts (8). In breast cancer, the MKK3/6 pathway, activated through H-Ras, induces the invasive phenotype. In contrast to reports that depict the activation of p38 as a proapoptotic event induced by chemotherapy in a variety of solid tumors (9), specifically pancreatic carcinomas (10), we found that small-molecule inhibition of p38 enhances glioma cell susceptibility to cytotoxicity mediated by methylating agent temozolomide (11). p38 was recently reported to be involved in lysophosphatidic acid–induced migration of glioma cells (12).
Functional studies underscore significant engagement of MKK3 and p38 in invasion; small interfering RNA (siRNA) knockdown of MKK3 and small-molecule inhibition of p38 result in decreased in vitro and ex vivo invasion coincident with increased susceptibility to apoptotic stimuli, rendering MKK3 and p38 promising targets for anti-invasive therapies. Moreover, we suggest that phosphorylated MKK3 (pMKK3) and phosphorylated p38 (pp38) may be biomarkers of glioma progression as their activation status correlates significantly with tumor grade; MKK3 gene expression was found to be a negative predictor for patient survival in two independent glioma expression data sets.
Materials and Methods
Cell Culture Conditions and Invasion Assay
Human astrocytoma cell lines U87, U118, SNB19, and U251 (American Type Culture Collection, Manassas, VA) and a subline of U87 stably transfected with the truncated form of epidermal growth factor receptor (U87
EGFR; a kind gift of Dr. Antonio Chiocca (Department of Neurological Surgery, Ohio State University, Columbus, OH); refs. 13, 14) were maintained in MEM (Invitrogen, Carlsbad, CA) supplemented with 10% heat-inactivated fetal bovine serum (Hyclone Laboratories, Inc., Logan, UT).
Three-Dimensional Invasion Assay
Spontaneous multicellular spheroids, U87 and U87EGFR, were suspended in 360-µL collagen I gels (Vitrogen, Angiotech, Palo Alto, CA) supplemented with MEM, 2% fetal bovine serum in eight-well chamber slides (Nunc, Rochester, NY) and overlain with 240 µL MEM, 10% fetal bovine serum. After 7 days, invasion assays were snap-frozen, cryosectioned at 15-µm thickness or fixed in 10% neutral buffered formalin, paraffin embedded, and sectioned at 5-µm thickness for immunofluorescence. Hanging drop culture (15) was used for all subsequent spheroid production where 2 x 105 cells/mL for U87 and SNB19 lines and 2 x 106 cells/mL for U118 and U251 cells were seeded in 20-µL droplets and cultured over a water-filled dish for 4 days. Viability was confirmed with the live/dead assays as described (16).
A sparse/dense paradigm, as previously described (17, 18), was adopted to three-dimensional cell culture conditions. To mimic the stationary phenotype, U87, U118, SNB19, and U251 glioma spheroids were propagated in 1,000-µL collagen I gel for 2 h after gelation. In separate cultures, the invasive phenotype was approximated by inclusion of single-cell suspensions of 2 x 106 cells in 1,000 µL of collagen I gel for 24 h. Cells were isolated from gels by collagenase digestion (Sigma-Aldrich, St. Louis, MO) followed by two washes with PBS; whole-cell lysates underwent immunoblotting (see below).
Reagents
LY479754 is an ATP-competitive small-molecule inhibitor of p38/ß and was a kind gift of Eli Lilly & Co. (Indianapolis, IN; ref. 19). Temozolomide was purchased from Sequoia Research Products (Oxford, United Kingdom).
Antibodies and dilutions used for immunohistochemistry were as follows: MKK3, 1:200 (Abgent, San Diego, CA); p38, 1:50 (R&D Systems, Minneapolis, MN); pMKK3/6, 1:200 and pp38, 1:50 (Cell Signaling Danvers, MA). For immunofluorescence detection, secondary Cy3-coupled antibody was used at 1:2,000 (Promega, Madison, WI). For immunoblotting, the following antibodies were used: MKK3, pMKK3/6, p38, pp38 (Cell Signaling), and -tubulin (Upstate, Charlottesville, VA).
Laser Capture Microdissection
LCM was done using the PixCell II LCM microscope (Arcturus, Mountain View, CA) as previously described (20). Modified H&E staining of cryosectioned U87 and U87EGFR spheroid invasion assays was carried out before microdissecting 400 invasive and 2,500 stationary core cells onto individual LCM caps (CapSure Macro LCM Caps, Arcturus) in three biological replicates, representing each cell line.
RNA Isolation, Amplification, Labeling, and Microarray Hybridization
LCM captured cells were lysed from caps and RNA was extracted with the PicoPure system (Arcturus) and analyzed qualitatively on a Bioanalyzer (RNA 6000 Pico Assay, Agilent, Palo Alto, CA). Linear amplification (1.5 rounds; RiboAmp HS, Arcturus) and Cy5 dye (Perkin-Elmer, Boston, MA) incorporation in the second round of amplification were done using the Agilent low input kit; amplified RNA was quantified photometrically (Gene Quant II, Amersham, Buckinghamshire, United Kingdom). Universal human reference RNA (Stratagene, La Jolla, CA) underwent one round of linear amplification with incorporation of Cy3 dye (Perkin-Elmer). Cy5-labeled amplified sample RNA (1.5 µg) was hybridized against 1 µg of Cy3-labeled amplified universal reference RNA on two-color whole human genome oligonucleotide microarray (Agilent). Gene chips were scanned and feature extracted using Agilent image analysis and statistical software.
Microarray Data Processing
Signals from sample and universal reference were background subtracted and Lowess corrected. One-way ANOVA filtering to remove genes not differentially expressed across U87 and U87EGFR samples was followed by two-sample t test (P < 0.05) to identify significantly differentially expressed genes between stationary (core) and invasive (rim) populations. This list was subsequently filtered for genes that showed ±1.5-fold differential expression in the rim population of either cell line as compared with core. Genes were categorized according to reported subcellular localization and function (Supplementary Table S1);1 from each category, genes with the highest degree of differential expression between core and rim were selected for technical validation by quantitative reverse transcription-PCR (RT-PCR).
Quantitative RT-PCR
Unamplified RNA isolated from three separate biological replicates of three-dimensional invasion assays and 100 ng of amplified and labeled RNA (analytes from the microarray experiments) were reverse transcribed (SuperScriptIII, Invitrogen). cDNA was amplified using the LightCycler instrument and FastStart DNA master SYBR Green reagent (Roche, Penzberg, Germany) with gene-specific primers (Operon, Huntsville, AL) and histone 3B as a housekeeping gene (Supplementary Table S2).1 The ratio of mRNA expression in invasive cells/stationary cells was calculated for each gene using the respective crossing points applied in the following formula: F = 2(IH-IG)-(SH-SG), where F is fold difference, S is stationary cells, I is invasive cells, G is gene of interest, and H is histone 3B (6).
Functional Studies
Two independent sets of siRNAs specific for human MKK3 gene were used: Hs_MAP2K3_5 HP validated siRNA (M1; from Qiagen) and a mixture of two siRNAs (M2) against GenBank accession no. BC032478 (CATGCGCACGGTCGACTGTTT and GAAGCCCTCCAATGTCCTTAT). siRNA against GL2 luciferase was used as a control (AACGTACGCGGAATACTTCGATT). Transient transfection of U87 and SNB19 cells was done as previously described (21) and gene knockdown was confirmed by immunoblotting.
LY479754 was tested for anti-invasive effects and potentiation of cytotoxicity induced by temozolomide in three-dimensional invasion assay. Photomicrographs were taken daily (x2.5 magnification; Zeiss Axiovert 130, Echingen, Germany) and the rate of radial dispersion (i.e., invasion rate) from serial measured radii encompassing the stationary core and invading cells was assessed and plotted over time (22). To measure cell kill, invasion assays were incubated with live/dead reagents (Molecular Probes, Eugene, OR) on day 5 as described elsewhere (16). Optical cross sections were acquired at the equator of spheroids with a Pascal 5 confocal microscope (Zeiss, Newthorn, NY) separately for the red and green channels. For each treatment condition, the ratio of area of red pixels/area of green pixels was calculated; the ratio of treated over ratio of untreated condition was then used to estimate relative cell kill.
Ex vivo Invasion Assay on Rat Brain Slices
A rat brain slice model system was used as previously described (23) and modified according to the organotypic culture methods previously reported (24). Experiments were conducted in compliance with Institutional Animal Care and Use Committee approved protocols. The cerebrum from 4-week-old male Wistar rats [Crl:(WI)BR; Charles River Lab, Wilmington, MA] was immediately removed and cut vertically to the base in 400-µm-thick sections using a vibratome (1000 Plus, The Vibratome Co., St. Louis, MO). One hundred thousand U87 or SNB19 glioma cells transfected with green fluorescent protein and siRNA against MKK3, luciferase, or untransfected were micropipetted onto brain slices (typically, six brain slices were used in each experiment). Glioma cell invasion into the brain slice was quantitated with a confocal microscope, capturing serial optical sections every 5 µm downward from the surface plane to the bottom of the slice from which invasion rates were calculated as previously described (25).
Immunofluorescence and Immunohistochemistry
Antigens were unmasked with 1x Reveal for pMMK3 and pp38 in the Decloacking Chamber (Biocare Medical, Concord, CA) at 115°C for 5 min; for MKK3, slides were microwaved in 10 mmol/L sodium citrate solution at 50% power for 5 min. For immunohistochemistry, endogenous peroxidase activity was quenched using 3% hydrogen peroxide in TBS-Tween 20 for 15 min. Slides were blocked with 5% goat serum, incubated with primary antibody overnight, followed by secondary antibody for 1 h. For immunohistochemistry, Vectastain mouse and rabbit kits (Vector Labs, Burlingame, CA) were used. Immunofluorescence was observed by confocal microscopy.
Tissue Microarray
Sections from a glioma invasion tissue microarray consisting of 24 WHO grade 4 glioma specimens (similarly to the one previously described; ref. 6) were subjected to the described staining methods using MKK3, pMKK3, p38, and pp38 antibodies. A pathologist (S.N.) assessed percentage of cells exhibiting negative (0), weak (+), moderate (++), and strong (+++) immunoreactivity (6, 26).
Immunoblot Analysis
Glioma cells or glioma tissue specimens were lysed in sample buffer as previously described (27), separated by 10% SDS-PAGE, transferred to nitrocellulose membrane (Invitrogen), and probed with specific antibodies. Horseradish peroxidase–conjugated secondary antibodies were detected by a chemiluminescence system (NEN, Boston, MA). Following stripping, membranes were reprobed with -tubulin antibody. Signals were quantified by densitometry using ImageJ and relative production levels of pp38, total p38, pMKK3, and total MKK3 (each normalized as the target protein:-tubulin protein ratios) were calculated (23). Positive controls included on each membrane were used to normalize and compare signal intensities across multiple membranes.
Clinical Samples
For immunoblot analysis, 41 frozen human glioma biopsy specimens (7 low-grade astrocytomas, 4 anaplastic astrocytomas, and 30 glioblastoma multiforme) were collected at the time of primary resection under an Institutional Review Board–approved protocol; none of the patients had been subjected to chemotherapy or radiation before surgery. Thirteen nonneoplastic control specimens were identified from the periphery of tumors. All specimens were evaluated by a pathologist and classified based on the revised WHO criteria for tumors of the central nervous system (28).
Gene Expression Profiling of Human Brain Tumor Specimens
For microarray analysis, snap-frozen nontumor brain tissues (epilepsy resection; n = 24) and astrocytoma specimens (8 low-grade astrocytomas, 21 anaplastic astrocytomas, and 82 glioblastoma multiforme) were collected with clinical annotation at Hermelin Brain Tumor Center, Henry Ford Hospital (Detroit, MI) under an Institutional Review Board–approved protocol (GEO accession no. GSE4290). After exclusion of genes whose expression levels did not show variance across all samples by at least 30%, 7,000 genes of high-quality signal intensity remained.
Statistical Analysis
Comparison between tumor samples, normal brain, and between treatment groups in invasion assays were done by unpaired, two-sided t tests under assumption of unequal variance. Median staining intensities of core and rim samples on the tissue microarray were compared by Pearson 
2 test. Kaplan-Meier methods were used to compare median survival rates between high (
mean expression) and low expressors (<mean expression) of target gene MKK3 in malignant astrocytomas. Prognostic significance was assessed by Cox proportional hazard regression. A previously published data set (26) was used for independent validation. Differences were considered significant at P < 0.05.
Results
Invasive Glioma Cells In vitro Show a Unique Gene Expression Profile
We applied a three-dimensional invasion assay to model in situ glioma cell dispersion using an in vitro model system wherein glioma cells actively invade into a three-dimensional matrix (Supplementary video). Glioma spheroids (220–250 µm in diameter) propagated in collagen I gel were found to be viable after up to 5 days in culture as assessed by confocal microscopy of live/dead stained invasion assays wherein no significant number of dead cells was found (data not shown). To identify genes driving invasion away from glioma spheroids, we used LCM to harvest populations of invasive (400 cells) and spheroid core cells (2,500) from glioma cell lines U87 and U87EGFR. The Bioanalyzer (Agilent) revealed good RNA integrity (robust 18S and 28S ribosomal peaks) and an RNA yield of 0.9 to 2 ng per sample. Whole human genome oligonucleotide microarray analysis identified 212 genes significantly differentially expressed between invasive and stationary cells in U87 and 382 genes in U87EGFR. Invasive cells from both cell lines shared 90 differentially expressed transcripts (Supplementary Table S1).1 These genes were classified into categories considered important in glioma biology and selected genes were validated by quantitative RT-PCR. Out of the group of survival/apoptosis–related genes, MKK3 was selected for further validation because it showed the highest degree of differential regulation between invasive and stationary cells in this bin (5-fold up-regulation in U87 and 3-fold up-regulation in U87EGFR). Quantitative RT-PCR validation using amplified and labeled RNA and unamplified RNA isolated from separately microdissected invasive and stationary U87 and U87EGFR cells confirmed transcriptional up-regulation of this gene with in vitro invasion (Fig. 1A
).
|
MKK3 Is Phosphorylated in Sparsely Seeded Glioma Cells in Three-Dimensional Collagen I Gel
To quantitatively assess activation status of MKK3 in cells propagated in three-dimensional collagen I gel, a sparse/dense culture model was used. Time-lapse microscopy confirmed adoption of invasive phenotype by sparsely seeded cells within 45 min (data not shown). pMKK3 was found to be elevated in U87, U118, SNB19, and U251 glioma cells propagated under sparse conditions relative to densely packed cognates (Fig. 1C); coincidentally, total MKK3 was found to be elevated under sparse conditions. Data from the sparse/dense culture model recapitulate findings from immunofluorescence experiments for U87 wherein invasive cells exhibited stronger staining for MKK3 and pMKK3 and support the notion that overexpression and phosphorylation of MKK3 are linked to glioma invasion.
MKK3 Is Overexpressed in Invasive Glioma Cells In situ and Correlates with Active p38
To seek clinical validation of these findings, levels of MKK3, pMKK3, p38, and pp38 were assessed on a human glioma tissue microarray, which contains tissue punches that capture dispersion of infiltrative glioblastoma multiforme. MKK3 was found to be significantly overexpressed in tumor cells residing at the invasive edge of glioblastoma multiforme compared with their stationary counterparts from the tumor core (Fig. 2A, B, and J
; P = 0.002); normal brain astrocytes exhibited moderate staining whereas neurons showed weak or no staining (Fig. 2C). Similarly, pMKK3 showed moderate or weak staining in 30% and 60% of invasive cells compared with 0% and 30% in stationary cells (P = 0.014; Fig. 2D, E, and J) whereas normal brain showed only weak staining (Fig. 2F). More than 80% of invasive glioma cells showed moderate to strong staining for pp38 compared with only 33% of stationary cells in tumor core (P = 0.016; Fig. 2G, H, and J); normal brain exhibited only weak staining for pp38 (Fig. 2I).
|
-tubulin as loading reference. Levels of both pMKK3 and total MKK3 protein were significantly higher in glioblastoma multiforme samples (P = 0.003; n = 30) and anaplastic astrocytomas (P = 0.016; n = 4) than in normal brain tissues (n = 13; Fig. 3B). The levels of pp38 protein were significantly higher in glioblastoma multiforme (n = 30; P < 0.001), anaplastic astrocytoma (n = 4; P = 0.012), and low-grade astrocytoma (n = 7; P = 0.004) samples than in normal brain tissues (n = 13). Glioblastoma multiforme (n = 30; P < 0.01), anaplastic astrocytoma (n = 4; P < 0.05), and low-grade astrocytoma tissues (n = 7; P < 0.01) had significantly higher levels of total p38 than normal brain samples (n = 13); significant correlations between tumor grade, pp38 levels (r = 0.518; P < 0.001), and total p38 (r = 0.509; P < 0.001) were identified (Fig. 3B). These data indicate that protein levels and phosphorylation of MKK3 and p38 were increased in glioblastoma multiforme tissue, which is the most malignant and invasive form of astrocytic tumors.
|
MKK3 Is a Prognostic Marker in Malignant Astrocytomas
Whole genome expression profiling of a series of human brain tumor specimens derived from Henry Ford Hospital (Detroit, MI) was carried out at the National Cancer Institute and revealed MKK3 expression to be significantly higher in glioblastoma multiforme (P < 0.001) and anaplastic astrocytoma (P = 0.046) than in normal brain specimens (Fig. 3D). In malignant astrocytomas, high expression of MKK3 (median expression) predicted significantly shorter median survival [359 days; 95% confidence interval (95% CI), 276–442 days] than low expression [<median expression; 598 days (95% CI, 577–619 days); P = 0.04; Fig. 3E]. Similar results were obtained analyzing a previously published data set (26) in which high MKK3 expression predicted significantly shorter median survival of malignant glioma patients (77 weeks; 95% CI, 43.3–110.7 weeks) than did low expression of MKK3 (102 weeks; 95% CI, 54.3–149.7 weeks; P = 0.04).
Cox proportional hazard analysis revealed MKK3 as a significant predictor for short-term survival (Cox coefficient, 0.165; P < 0.001; 95% CI, 0.07–0.26) in our data set as well as in a similar patient data set previously published (ref. 26; Cox coefficient, 0.97; P = 0.006; 95% CI, 0.28–1.66).
Inhibition of p38 Results in Decreased In vitro Invasiveness and Sensitizes Cells to Apoptotic Stimuli
To assess effects of p38 inhibition on invasion and cell survival, U87, U118, and SNB19 glioma spheroids were treated with LY479754 and temozolomide. After exposure to 10 µmol/L LY479754, U87 spheroids revealed 34 ± 9% (P < 0.001) reduction of invasion rate, U118 showed 28 ± 17% (P < 0.001), and SNB19 showed 17 ± 4% reduction of invasion rate (P = 0.003) as compared with vehicle control (Fig. 4A
). Five days after initiation of treatment with 0.1 µmol/L LY479754 + 200 µmol/L temozolomide, U87 exhibited 4.1 ± 1.56-fold (P = 0.003) increase in relative cell kill and SNB19 exhibited 5.1 ± 1.5-fold (P < 0.001) increase in relative cell kill when cotreated with 10 µmol/L LY479754 + 200 µmol/L temozolomide, as compared with temozolomide-only treatment (U87, 1 ± 0.78; SNB19, 1 ± 0.5; Fig. 4B).
|
Discussion
We previously reported analysis of the invasive transcriptome of in situ human gliomas, which identified and validated target gene products implicated in the malignant dispersion process (6). In an effort to further discover and validate mediators of invasion, we used a well-established three-dimensional in vitro model system capable of reporting events inherent to the active egress of glioma cells from a multicellular tumor spheroid (7, 29). The importance of studying cell behavior in an environment providing spatial complexity is increasingly recognized (30), and the three-dimensional invasion assay allows for examination of cell signaling and phenotypic events under culture conditions more similar to an in situ situation as compared with the conventional two-dimensional cell culture (31, 32). Collagen I is not abundantly present in the human brain; it was selected as a model system that creates a three-dimensional fiber structure through which glioma cells can invade (33). Unlike Matrigel, which has been widely used to study invasion of multiple tumor types and which contains qualitatively and quantitatively varying growth factors or artificial, allogenic extracellular matrix proteins (34), collagen I is a chemically well-defined self-polymerizing substrate. This allows glioma cells to adopt an in situ–like shape and a gene expression profile that more closely resembles that of primary tumor explants (35).
Gene expression profiling was done on stationary and invasive glioma cells isolated by LCM from this in vitro model system (7, 36). Invasive cells from U87 and U87EGFR shared 90 differentially expressed genes that, concordantly expressed, can be interpreted as common participants in three-dimensional invasion. U87 and U87EGFR cells were chosen for gene expression profiling experiments as they might be considered a first approximation of the two major subgroups in glioblastoma multiforme based on EGFR status (37): primary glioblastoma multiforme (U87EGFR, expressing constitutively active EGFRvIII; ref. 13), which are characterized by amplification or overexpression of EGFR, and secondary glioblastoma multiforme (U87), which are more prominent in a younger population and do not present with EGFR modification (38). p53 mutation, an important hallmark of progression from low-grade astrocytomas, is not found in U87.
We recently reported that migrating glioma cells show reduced sensitivity to apoptotic stimuli and therefore theorized an inverse link between migration and vulnerability to apoptosis (18). To follow this apparent association, we investigated a candidate from the class of apoptosis/survival genes: MKK3 was identified to have the highest degree of differential regulation between invasive and stationary cells in this category and was therefore selected for further validation.
Whereas p38 has been extensively studied in the context of survival and apoptosis of different cancers, we show for the first time that MKK3 is a key upstream activator of p38 in glioma. MKK3, a gene not previously implicated in glioma pathology, is a dual-specificity protein kinase that preferentially activates p38 (39). Stimulation of human glioma cells with epidermal growth factor resulted in dose-dependent activation of MKK3 and p38 whereas siRNA-mediated knockout of MKK3 prevented p38 activation (data not shown). MKK3 is activated (phosphorylated) through environmental stress–induced pathways (40) and transforming growth factor ß, a driver of glioma proliferation and migration (41, 42). MKK3, activated through H-Ras, has been recently reported to induce an invasive phenotype in breast epithelial cells (43) and, together with downstream effector MAPK-activated protein kinase 2, increase the stability of urokinase-type plasminogen activator mRNA (44). MKK3 downstream target p38 is phosphorylated, leading to translocation to the nucleus, enabling it to activate transcription factors such as c-Myc, myocyte enhancer factor 2C (45), and signal transducers and activators of transcription 1 (46).
Up-regulation of MKK3 during in vitro invasion was technically confirmed by quantitative RT-PCR of stationary and invasive glioma cells isolated by LCM from spheroid invasion assays. Immunofluorescent staining of formalin-fixed, paraffin-embedded sections of three-dimensional invasion assay confirmed significant overexpression and activation of MKK3 protein in invasive cells compared with stationary cognates. Inconsistent distribution of antibody in this application was ruled out by staining for glial fibrillary acidic protein and vimentin, two important markers of glial tumors (47), which were found to be unchanged between invasive and stationary cells (data not shown). A sparse/dense paradigm was used to analyze levels of protein expression between cells adopting the invasive (i.e., sparse) or stationary (i.e., dense) phenotype. Four human glioma cell lines displayed increased levels of total and pMKK3 when adopting an invasive phenotype, corroborating quantitative RT-PCR and immunofluorescence results and suggesting that MKK3 overexpression and activation are important contributors to in vitro glioma invasion. -Tubulin was selected as a loading control in experiments as we recently showed that its levels are not altered between stationary and migratory glioma cells in vitro (23).
To assess the clinical relevance of MKK3 and p38 overexpression and activation for glioma invasion, we used a human glioblastoma multiforme invasion tissue microarray. MKK3, pMKK3, and pp38 were detected predominantly in invasive glioma cells in the tumor rim, whereas tumor core exhibited only weak staining. Whereas pp38 staining was predominantly nuclear, faint cytoplasmic staining in invasive glioma cells is interpreted as a subcellular localization where p38 is activated (i.e., phosphorylated) by cytoplasmic protein pMKK3, leading to pp38 translocation and accumulation in the nucleus. Normal brain exhibited no staining for these proteins. Total p38 staining was found to be unchanged between stationary and invasive glioma cells (data not shown). Semiquantitative analysis of tissue microarrays revealed significantly higher median staining intensity for MKK3, pMKK3, and pp38 in a higher percentage of invasive malignant glioma cells than in noninvasive cognates. This observation confirms findings from in vitro assays that MKK3 and p38 are overexpressed and phosphorylated in invading glioma cells.
Immunoblot analysis on an independent panel of 41 human glial tumor samples revealed significant up-regulation of MKK3 and pMKK3 in glioblastoma multiforme and anaplastic astrocytoma compared with normal brain specimens, suggesting an important role for MKK3 in glioma progression. Likewise, levels of total and activated p38 were found to correlate with tumor grade, also emphasizing their role in tumor progression. These findings are in line with our interpretation of MKK3 and p38 as drivers of glioma invasion because high-grade gliomas are increasingly invasive and would be anticipated to express higher levels of these proteins. The biochemical activity and causal connectivity of the MKK3-p38 signaling axis in glioblastoma multiforme invasion were affirmed by the significant correlation in their respective phosphorylation ratios. Furthermore, our findings underscore MKK3 as an important activator of p38 in glioblastoma multiforme as shown by siRNA-mediated knockdown of MKK3 that yields reduced levels of activated p38 coincident with reduction of pMKK3, underlining the functional relationship of these two proteins.
Based on these findings, we suggest that levels of activated MKK3 and p38 serve as biomarkers for glioma progression. We therefore analyzed gene expression data derived from 111 human astrocytoma specimens and 24 normal controls in which MKK3 was found to be significantly up-regulated in glioblastoma multiforme samples compared with normal brain tissue. These data are concordant with immunoblot and immunohistochemistry results, showing that MKK3 is transcriptionally and translationally up-regulated during glioma progression and invasion. In two independent glioma expression data sets, Cox proportional hazard analysis revealed MKK3 to have significant predictive value for shorter patient survival, and Kaplan-Meier analysis showed that patients with high MKK3 expression have significantly shorter median survival than those with low expression. This finding might be interpreted as a consequence of the increased invasiveness of tumors with higher MKK3 expression, leading to a more unfavorable prognosis. Complementary results derived from the analysis of independent clinical sample sets queried by different methods such as immunohistochemistry (tissue microarray), immunoblotting (surgical specimen), and gene expression profiling support the importance of MKK3 and p38 for glial tumor progression, tumor invasion, and patient survival.
In vitro and ex vivo studies confirmed the biological significance of MKK3 and p38 in glioma invasion. Pharmacologic inhibition of p38 activity yielded significant reduction of invasion in both three-dimensional invasion assay and organotypic brain slice assay and these results were echoed by treatment with siRNA against MKK3.
Support for p38 as an important driver of invasion is derived from the observation that high levels of activated p38 coincide with greater invasion response to treatment with p38 inhibitor. Whereas siRNA knockdown of MKK3 yields similar reduction of invasion in U87 and SNB19, it causes less reduction in pp38 levels in SNB19 (low pp38 levels) compared with U87 (high pp38 levels). When investigating a cell line with intermediate levels of pp38 (U118), we observed an invasion response in-between U87 and SNB19 on treatment with LY479754. A strong correlation (r = 0.937) between the phosphorylation ratio of MKK3/p38 and reduction of invasion in U87, SNB19, and U118 glioma cells further underscores the relevance of these molecules for glioma invasion (data not shown). In conjunction with the differences in invasiveness on treatment with p38 inhibitor, we suggest that pp38 levels are predictors for a response to anti-invasive therapies with LY479754.
LY479754 in combination with cytotoxic agent temozolomide led to 4- to 5-fold increased cytotoxicity in our three-dimensional in vitro assay, rendering this a synthetic lethal event with promising clinical applications in glioma. Comparable with results from the invasion assays, sensitivity to temozolomide was enhanced at much lower concentrations of the p38 inhibitor in the cell line with the highest pp38 levels (U87), whereas SNB19 required higher doses of LY479754.
LY479754 might inhibit kinases other than p38; at the concentrations used, it is nontoxic as a single agent while greatly enhancing temzolomide-mediated cytotoxicity. The role of p38 in cancer is tissue and microenvironment specific; in a number of solid tumors, p38 is believed to transmit proapoptotic signals induced by chemotherapeutic agents (48, 49). In multiple myeloma and leukemia, p38 acts as a prosurvival factor and its inhibition leads to G2-M arrest, which, in combination with bortezomib or etoposide treatment, enhances cytotoxic effects (50, 51). Consistent with this observation, we hypothesize a similar mechanism of action for glioma cells treated with p38 inhibitor and alkylating agent temozolomide (52).
In this report, gene expression profiling of glioma invasion in a three-dimensional model system identified MKK3 as a previously unrecognized candidate involved in the invasive behavior of human glioma. We show that MKK3 is not only transcriptionally up-regulated but also biologically active, phosphorylating p38 in invasive glioma cells in vitro and in vivo. MKK3 was found to be a predictor of short-term survival and activated MKK3 and p38 were found to correlate with tumor progression, rendering these molecules as biomarkers for patient survival and tumor progression. Functional studies confirmed the biological significance of these molecules in glioma invasion. Clinical utility of interference with MKK3 signaling through p38 inhibition in malignant glioma might be twofold: inhibiting tumor cell invasion and sensitizing glioma cells to temozolomide-induced cytotoxicity in the sense of a synthetic lethal event. These promising results mandate testing of LY479754 ± temozolomide in a xenograft model to assess efficacy against invasive glioma in vivo. On successful completion, we envision clinical testing of p38 inhibitors alone or in combination with temozolomide as innovative anti-invasive strategies and anticipate that tumors with high pp38 levels are more susceptible to this treatment than tumors with low levels of p38.
Acknowledgments
We thank T. Mikkelsen (Henry Ford Hospital, Detroit, MI), H.A. Fine, and J.C. Zenklusen (National Cancer Institute, Bethesda, MD) for providing gene expression data of human glioma samples; A.R. Chaudhury (Ohio State University, Columbus, OH) for providing the tissue microarray; P. Stafford for insightful discussions about data analysis (ASU Biodesign Institute, Tempe, AZ); J. Wu (Barrow Neurological Institute, Phoenix, AZ) for assisting in organotypic brain slice culture; and R.M. Campbell (Eli Lilly & Co., Indianapolis, IN) for providing LY479754.
Footnotes
Grant support: NIH grants NS042262 and CA085139 (M.E. Berens).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 Supplementary material for this article is available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/). 
Received 11/16/06; revised 1/29/07; accepted 2/23/07.
References
-trypsin inhibitor heavy chain 2. Cancer Res 2006;66:1464–72. MAP kinase. J Biol Chem 2006;281:26225–34. MAP kinase inhibitors with excellent in vivo efficacy. J Med Chem 2005;48:2270–3.[CrossRef][Medline] Mariani L, Beaudry C, McDonough WS, et al. Death-associated protein 3 (Dap-3) is overexpressed in invasive glioblastoma cells in vivo and in glioma cell lines with induced motility phenotype in vitro. Clin Cancer Res 2001;7:2480–9.
B pathway activation and BCL-XL/BCL-W expression. J Biol Chem 2005;280:3483–92. and p38
MAPK isoforms by TGF-ß1 in murine mesangial cells. J Biol Chem 2002;277:47257–62. MAPK enhances proteasome inhibitor-induced apoptosis of myeloma cells by modulating Hsp27, Bcl-X(L), Mcl-1 and p53 levels in vitro and inhibits tumor growth in vivo. Leukemia 2006;20:1017–27.[CrossRef][Medline] Kurosu T, Takahashi Y, Fukuda T, Koyama T, Miki T, Miura O. p38 MAP kinase plays a role in G2 checkpoint activation and inhibits apoptosis of human B cell lymphoma cells treated with etoposide. Apoptosis 2005;10:1111–20.[CrossRef][Medline] Hirose Y, Katayama M, Berger MS, Pieper RO. Cooperative function of Chk1 and p38 pathways in activating G2 arrest following exposure to temozolomide. J Neurosurg 2004;100:1060–5.[CrossRef][Medline]This article has been cited by other articles:
|
|
 |
|
  E. Suswam, Y. Li, X. Zhang, G. Y. Gillespie, X. Li, J. J. Shacka, L. Lu, L. Zheng, and P. H. King Tristetraprolin Down-regulates Interleukin-8 and Vascular Endothelial Growth Factor in Malignant Glioma Cells Cancer Res., February 1, 2008; 68(3): 674 - 682. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
