Axl and Growth Arrest–Specific Gene 6 Are Frequently Overexpressed in Human Gliomas and Predict Poor Prognosis in Patients with Glioblastoma Multiforme

Materials and Methods

Fresh-frozen tissue. This study was done according to the ethical regulative of the institutional review board. During open neurosurgery, tissue samples from brain tumor patients were snap frozen in liquid nitrogen within 3 min of harvesting and stored at −80°C. mRNA expression levels of Axl, Gas6, and tubulin (housekeeping gene) were investigated using reverse transcription-PCR.

Total RNA isolation and cDNA synthesis. Total RNA was isolated from 42 fresh-frozen glioma specimens using the acid guanidium thiocyanate-phenol-chloroform extraction method (11). After NaCl acetate treatment to reduce genomic DNA, total RNA was extracted with phenol-chloroform, precipitated with isopropanol, resuspended with lysis buffer, reprecipitated with isopropanol, washed with 80% ethanol, and finally solved in TES/diethyl pyrocarbonate solution. Afterwards, the total RNA concentration was determined by absorbance measurement (260 and 280 nm) and the quality of total RNA was verified by a 1% agarose gel electrophoresis. Only samples with 28S/18S ratios >2 and with no evidence of DNA contamination and RNA degradation were used for continuative cDNA synthesis. cDNA synthesis was done using 5 μg of total RNA in an oligo(dT) primer mix (1 pmol, total volume 9 μL; Chipco), which was heated at 70°C for 3 min and then cooled on ice. A mastermix containing 5× reverse transcriptase buffer, 10 mmol/L deoxynucleotide triphosphates, 100 mmol/L DTT, 1 unit (0.5 μL) RNase inhibitor, and 50 units (2 μL) avian myeloblastosis virus reverse transcriptase (Molecular Diagnostics, Roche) for each cell line and frozen tumor sample was added into a tube and subsequently heated at 42°C for 2 h. A stop reaction was done with 80 μL Tris-EDTA 10/0.1 followed by heating at 72°C for 7 min. The quality of the cDNA was verified using a 1.5% agarose gel electrophoresis. Afterwards, the cDNA probe was denatured with 10 μL 1 N NaOH, pH neutralized with 5 μL 2 N HCl and 5 μL 2 mol/L Tris-HCl (pH 7.5), and purified using the QIAquick PCR Purification kit (Qiagen) according to the manufacturer's instructions.

PCR. The PCR method was used to determine mRNA expression levels of Axl, Gas6, and tubulin (housekeeping gene) in fresh-frozen glioma samples. The following primer sequences were applied: Axl, 5′-GGTGGCTGTGAAGACGATGA-3′ (forward) and 5′-CTCAGATACTCCATGCCACT-3′ (reverse); Gas6, 5′-ACATCTTGCCGTGCGTGCCCTTCA-3′ (forward) and 5′-ATTCCGCGCCAGCTCCTCAACAGA-3′ (reverse); and tubulin, 5′-AAGTGACAAGACCATTGGGGGAGG-3′ (forward) and 5′-GGGCATAGTTATTGGCAGCATC-3′ (reverse). All primers were synthesized by MWG Biotech. PCR was done using 1 μL cDNA as template in a mastermix containing 10× PCR buffer without MgCl₂, 25 mmol/L MgCl₂, 10 mmol/L deoxynucleotide triphosphates, 10 pmol/μL of each specific primer, and 0.5 IU Taq DNA polymerase (Fermentas Taq Polymerase kit, Roche) for each sample and carried out in a DNA thermal cycler (Eppendorf). PCR cycling conditions began with an initial DNA denaturation step at 94°C for 2 min followed by 32 cycles at 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s. Amplified PCR products were applied to a 2% agarose gel for electrophoresis done at 80 V for 30 min. To analyze the mRNA expression frequency, reverse transcription-PCR results were evaluated semiquantitatively (no, low, moderate, or high).

Whole tissue sections and tissue microarray population. To investigate the immunohistochemical staining patterns of Axl and Gas6 protein in neoplastic and nonneoplastic tissue samples, we applied representative whole sections of paraffin-embedded glioma tissue WHO grades 2 to 4 (30 samples) and nonneoplastic hippocampal tissue derived from epilepsy surgery (10 samples). For clinicopathologic correlation studies, a population-based tissue microarray (TMA) was constructed containing paraffin-embedded tumor samples from 24 adult patients with newly diagnosed glioma WHO grades 2 to 3 and 55 patients with newly diagnosed GBM WHO grade 4 (Tables 1 and 2 ).

TMA construction. Tissue from brain tumor patients (1996-2005) treated by open surgery with partial, subtotal, or total tumor resection was routinely fixed in 4% formaldehyde solution and embedded in paraffin. For TMA construction, paraffin blocks were retrieved from the files of the Institute of Pathology, Innsbruck Medical University. The histologic diagnosis of the gliomas follows the WHO classification and grading system (1). The neuropathologic findings were confirmed by a board-certified neuropathologist, and representative tumor areas for TMA cores were identified on H&E-stained slides. Using a tissue microarrayer (MTA-1, Beecher Instruments), the paraffin blocks were punched with a diameter of 0.6 mm and the core was transferred to a recipient block. From each brain tumor, triplicate cores of different sites of the most representative tumor regions were collected. Finally, five recipient blocks were heated at 60°C for 30 min and subsequently cooled for 24 h. TMA blocks were cut into 4-μm sections and placed onto coated slides (PSA-4x, Instrumedics, Inc.) using an adhesive tape transfer system (Paraffin Tape Transfer System, Instrumedics).

Immunohistochemistry. Whole sections of formalin-fixed, paraffin-embedded glioma specimen derived from tumor surgery and nonneoplastic hippocampal brain samples derived from epilepsy surgery were studied for Axl and Gas6 staining patterns. The sections were deparaffinized in xylene (2 × 20 min) and rehydrated through a descending concentration sequence of ethanol. Heat-induced epitope retrieval was done by boiling the slides in NaCl citrate buffer (pH 6.0) using a microwave (2 × 8 min, 750 W). Afterwards, the slides were cooled for 30 min at room temperature within the buffer. Nonspecific staining was blocked by 5% H₂O₂ in 0.3% PBS-Triton X-100 (15 min) and 10% normal horse serum in 0.3% PBS-Triton X-100 (60 min). Incubation with primary Axl and Gas6 antibodies (1:100 dilution in 1% normal horse serum/0.3% PBS-Triton X-100) was carried out overnight at 4°C. Antibodies were purchased from the following suppliers: mouse monoclonal anti-human Axl (MAB 423-36; Max Planck Institute of Biochemistry, Department of Molecular Biology, Martinsried/Munich, Germany; antibody specificity was proved by Western blot and fluorescence-activated cell sorting) and goat polyclonal anti-human Gas6 (C-20, Santa Cruz Biotechnology). The bound primary antibody was detected by a biotinylated anti-mouse (Axl) or biotinylated anti-goat (Gas6) IgG secondary antibody (both made in horse; 60 min; Vector Laboratories, Inc.). The staining sensitivity was increased using the Vectastain Elite ABC Kit PK-6100 Standard as described in the supplier's manual, and the reaction was visualized through incubation with liquid substrate 3,3′-diaminobenzidine tetrahydrochloride (Sigma) for 5 min. Afterwards, the slides were counterstained with Meyer's hemalaun (Merck), mounted with Entellan (Merck), and coverslipped. Negative controls were included in each batch replacing the primary or secondary antibody and staining with an isotype-matched IgG2 monoclonal antibody.

Immunohistochemical staining of glial fibrillary acidic protein (GFAP), neurofilament proteins, and Ki-67 was done according to the manufacturer's standard protocols in a pathologic routine laboratory (Institute of Pathology, Medical University of Innsbruck). Antibodies were purchased from the following suppliers: rabbit polyclonal anti-human GFAP (Dako), mouse monoclonal anti-human Ki-67 (MIB-1, Dako), and mouse monoclonal anti-human neurofilament proteins (Biogenex). Immunohistochemical pictures were taken using a Nikon Eclipse E800 microscope.

Neuropathologic evaluation. Staining pattern and intensity were evaluated by a board-certified neuropathologist. At least two of three TMA spots had to be representative. A semiquantitative scoring system was applied following the German immunohistochemical score (GIS; ref. 12). Percentage of positive cells was graded as 0 (negative), 1 (up to 10%), 2 (11-50%), 3 (51-80%), or 4 (>80% positive cells) and staining intensity as 0 (no staining), 1 (weak), 2 (moderate), or 3 (strong). The final immunoreactive GIS was defined as the multiplication of both grading results (percentage of positive cells × staining intensity). Protein expression frequencies were analyzed by dividing the GIS into no/low (GIS < 5), moderate (GIS ≥ 5 to GIS ≤ 8), and high (GIS > 8) levels. For evaluation of the prognostic effect of Axl and Gas6, high or low protein expression levels were considered through GIS dichotomization at the median (GIS ≤ 6 versus GIS > 6). Axl/Gas6 coexpression was determined by a GIS >6 of both Axl and Gas6.

Statistical analysis. Statistical analysis was done using the variables time to tumor progression (TTP), progression-free survival at 6 months (PFS6), OS, and OS at 12 months (OS12). TTP and OS were determined in months from the date of surgery of a primary or secondary GBM. TTP end points corresponded with the detection of a recurrent tumor by magnetic resonance imaging or computed tomography. Survival end points matched to dates of death or follow-up end points. Patients who were alive at the follow-up date were censored for survival analyses. Kaplan-Meier curves were plotted to estimate TTP and OS as a function of Axl and/or Gas6 expression and group differences were assessed by the log-rank method. Adjusted analysis of the effect of Axl and Gas6 protein expression on TTP and OS in glioblastoma patients was carried out by multivariate data analysis using the Cox proportional hazards regression model. The following clinically relevant variables were considered for statistical analysis: age (<60 versus ≥60 years), Karnofsky performance scale (KPS; <80 versus ≥80), tumor location (frontal versus nonfrontal), extend of resection (partial versus total/subtotal resection), chemotherapy (+ versus −), and radiation (+ versus −). The forward stepwise procedure using 95% confidence intervals was adopted to determine the independent prognostic value of Axl and Axl/Gas6 expression in multivariate analysis (Table 3 ; patients with a KPS <70 and patients with stereotactic or open biopsy were not included in this study). Multivariable binary logistic regression analysis, including age, KPS, tumor location, extend of resection, chemotherapy, and radiation as covariables, was done to evaluate the association of Axl and Gas6 expression with PFS6 and OS12. Differences of expression frequencies between glioma subgroups (WHO grades 2/3/4) and primary and secondary GBM were evaluated by the Kruskal-Wallis test. Statistical analysis was done using Statistical Package for the Social Sciences 12.0 software (SPSS, Inc.) and P values of <0.05 were considered statistically significant.

Results

Axl and Gas6 are frequently overexpressed in human gliomas. Reverse transcription-PCR studies of fresh-frozen tumor samples (Fig. 1A ) and immunohistochemical studies of paraffin-embedded tumor tissue (Fig. 1B-G) showed that Axl and Gas6 are frequently overexpressed in gliomas WHO grades 2 to 4, including astrocytic, oligodendroglial, and ependymal origin (Table 2). In GBM WHO grade 4, moderate to high expression levels were found for Axl mRNA in 61% (n = 19), for Axl protein in 55% (n = 30), for Gas6 mRNA in 81% (n = 25), and for Gas6 protein in 75% (n = 41), respectively. Axl/Gas6 coexpression was detectable in 61% (n = 19) at the mRNA and in 43% (n = 22) at the protein level. The Kruskal-Wallis test showed no significant differences of Axl and/or Gas6 expression frequencies between the glioma subgroups WHO grades 2/3/4 as well as primary and secondary GBM (Table 2). In contrast to tumor tissue (Fig. 2A ), Axl protein was not detectable in nonneoplastic brain tissue distant to the tumor (Fig. 2B) or hippocampal brain tissue derived from epilepsy surgery (data not shown). However, Gas6 showed a pronounced neuronal membrane and cytoplasmatic granular staining in nonneoplastic brain tissue (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Reverse transcription-PCR and immunohistochemical staining of Axl and Gas6 in human gliomas. A, mRNA expression of Axl, Gas6, and tubulin in human fresh-frozen glioma samples. mRNA expression frequency was evaluated semiquantitatively (no, low, moderate, or high). B to G, immunohistochemical staining pattern of Axl and Gas6 in gliomas WHO grades 2 to 3 of astrocytic and oligodendroglial origin and GBM WHO grade 4 (TMA spots).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Immunohistochemical staining pattern of Axl and Gas6 in tumor (GBM), vascular cells, and reactive astrocytes. A, Axl expression of GBM cells. B, neoplastic (GBM) and nonneoplastic tissue in direct proximity (as an artifact of tissue preparation) show that Axl is only detectable in GBM tissue. C to E, Axl expression of pseudopalisading tumor cells (P), which surround a central necrotic (hypoxic) area (N). Axl and GFAP are coexpressed in tumor cells. F (left), Gas6 expression in the hypoxic border zone (BZ) of a necrotic area, but not in pseudopalisades. (F, right) adjacent (not migrating) GBM cells show a substantial Ki-67 proliferation rate, whereas pseudopalisading (migrating) tumor cells reveal no cell proliferation (G and H). Microvascular hyperplasia (*) shows a strong Axl expression of endothelial cells (EC) and is surrounded by Axl-positive tumor cells (TC). I, tumor vessels reveal a strong Axl expression of endothelial cells, vascular smooth muscle cells (VSMC), and pericytes (PC). J and K, in the peripheral infiltrative zone of a GBM, Axl and GFAP are coexpressed in reactive astrocytes (RA). L, reactive astrocytes bordering to the infiltrating boarder zone are Axl and GFAP positive and their astrocytic end feet are in close contact with capillaries.

Axl and Gas6 expression shows a distinct cellular distribution pattern in GBM. A characteristic immunohistochemical finding was a membrane-accentuated Axl staining of glioma cells (Fig. 2A-D), particularly of pseudopalisades (Fig. 2C and D) and opposing tumor cells (Fig. 2D). In addition, vessels within a necrotic area were typically surrounded by Axl- and GFAP-positive tumor cells (Fig. 2E), representing necrotic areas and pseudopalisades caused by intravascular thrombosis. In corresponding tissue sections, pseudopalisades showed a weak Gas6 expression (Fig. 2F) and staining of the proliferation marker Ki-67 revealed less proliferative potential of glioma cells within pseudopalisades and necrotic areas compared with adjacent tumor cells (Fig. 2F). In nonpseudopalisading GBM cells, Axl and Gas6 were frequently found to be coexpressed (Fig. 1F and G).

A second distinct neuropathologic finding was that both Axl (Fig. 2G and H) and Gas6 (data not shown) were strongly expressed in endothelial cells of microvascular hyperplasia and tumor cells adjacent to microvascular hyperplasia showed a pronounced Axl staining (Fig. 2G). Furthermore, arterioles and venules in the peripheral infiltrating zone were characterized by Axl (Fig. 2I) and Gas6 (data not shown) expression of endothelial cells, vascular smooth muscle cells, and pericytes. In contrast, vessels of nonneoplastic brain tissue showed neither Axl nor Gas6 staining (data not shown).

A third characteristic feature was a prominent Axl staining of reactive astroglial cells, which were identified by their astrocytic morphology, positive GFAP expression, and negative neurofilament protein staining (Fig. 2J-L). Such reactive astrocytes were observed in the peripheral GBM infiltrating zone enchased by tumor cells (Fig. 2J and K). Typically, reactive astroglial cells showed Axl/GFAP coexpression, a pronounced Axl staining of the astrocytic end feet, and a weak Gas6 staining (Fig. 2L). However, Axl, Gas6, and GFAP expression was absent in vessels and astrocytes more distant to the GBM (data not shown).

Axl and Gas6 expression correlates with poor prognosis in GBM patients. Using a population-based TMA, we assessed the prognostic value of Axl and Gas6 expression in GBM patients. Evaluation of Axl expression revealed a median TTP of 8.9 months (range, 6.1-17.0) for patients with low (n = 19, GIS ≤ 6) and 3.9 months (range, 1.6-10.0) for patients with high Axl expression values (n = 16, GIS > 6; P < 0.001 by the log-rank test, Fig. 3A ). The adjusted hazard ratio for TTP was 3.2 (95% confidence interval, 1.5-6.8) in patients with high compared with those with low Axl expression and shows a significant relative increase in the risk for early tumor relapse (P = 0.001; Table 3). Evaluation of Axl/Gas6 coexpression appeared that, in patients with low coexpression levels (n = 25), the median TTP was 7.9 months (range, 1.6-17.0) compared with 3.9 months (range, 2.0-9.1) in those with high levels (n = 10; P = 0.0015 by the log-rank test, Fig. 3B). The adjusted hazard ratio for TTP in the group of high Axl/Gas6 coexpression was 3.6 (95% confidence interval, 1.5-8.4) compared with the group of low expression levels and again showed a significant relative increase in the risk of early tumor relapse (P = 0.003; Table 3). In addition, multivariate binary logistic regression analysis revealed that high Axl expression (P < 0.001) and high Axl/Gas6 coexpression (P = 0.001) were significantly associated with a PFS <6 months (Table 4 ).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Prognostic value of Axl and Gas6 expression in GBM patients. A to C, Kaplan-Meier curves of Axl expression and Axl/Gas6 coexpression in GBM patients related to TTP and OS. $, log-rank test.

The median OS was 13.1 months (range, 0.8-27.5) in patients with low (n = 31) and 9.4 months (range, 2.4-22.3) in patients with high Axl expression (n = 18; P = 0.057 by the log-rank test, Fig. 3B; P = 0.09 by multivariate Cox regression model, Table 3). In patients with low Axl/Gas6 coexpression (n = 39), the median OS was 13.2 months (range, 0.8-27.5) and 9.2 months in those with high levels (range, 3.3-21.2; n = 9; P = 0.115 by the log-rank test; P = 0.142 by multivariate Cox regression model; Table 3). However, 92% (n = 23) of GBM patients with OS >12 months had a low Axl/Gas6 coexpression compared with only 67% (n = 18) of those with OS <12 months (P = 0.045 by multivariate binary logistic regression analysis; Table 4).

Discussion

Due to the dismal prognosis of GBM with currently available therapies, there is an urgent need for new treatments based on a better understanding of the pathophysiologic and molecular properties of gliomas. The present study shows that the receptor tyrosine kinase Axl and its ligand Gas6 are biologically relevant not only to glioma cells in vitro and animal models in vivo (10) but also for human gliomas. We show that in human glioma tissue Axl and Gas6 are frequently expressed at the mRNA and protein level, and in GBM patients, high Axl expression and Axl/Gas6 coexpression are significantly associated with shorter TTP and PFS <6 months. In addition, Axl and Axl/Gas6 are related to poorer OS and Axl/Gas6 coexpression is significantly associated with OS <12 months. However, in evaluation of the prognostic significance of Axl and Gas6 expression, TTP and PFS6 are the more reliable study end points compared with OS and OS12 because newly diagnosed GBMs are treated according to a standard therapy protocol until tumor progression, whereas various subsequent therapies are used after tumor progression (13).

Recently, Vajkoczy et al. (10) have shown that inhibition of Axl signaling almost completely suppresses glioma cell migration into a fetal rat brain spheroid and into brain tissue of a mouse i.c. xenograft model. In our study, we show that Axl is predominantly expressed in pseudopalisading glioma cells, characterized by an accumulation of tumor cells around necrotic areas. Such cell formations are composed by a wave of actively migrating tumor cells moving away from a central hypoxic/necrotic area, resulting from an increased metabolic demand and/or intravascular thrombosis (14, 15). In pseudopalisading glioma cells, the subcellular distribution of Axl expression shows a membrane-accentuated staining pattern, particularly of opposing tumor cells. Due to the cell adhesion properties of the Axl extracellular domain, homophilic Axl-Axl interactions (9, 16) between tumor cells and heterophilic Axl-extracellular matrix binding (17, 18) seem to be important for Axl-mediated glioma cell migration. We also observed an accumulation of Axl-positive tumor cells adjacent to microvascular neoformations. This finding corresponds to a characteristic feature of glioma invasiveness (i.e., invading glial tumor cells spreading along perivascular regions; refs. 19, 20). GFAP, an astrocyte-specific intermediate filament controlling cell shape and cell movement (21, 22), is commonly detectable in glioma cells (23). Our comparative immunohistochemical studies showed that Axl and GFAP are coexpressed in glioma cells of pseudopalisades. This observation suggests a functional association of Axl and GFAP during glioma cell migration. Invading glioma cells are the most difficult to treat and represent the source for local tumor recurrences. Thus, the significantly shorter TTP of GBM patients with high Axl expression may reflect the biological contribution of Axl to a more invasive glioma phenotype.

TTP was also significantly reduced in GBM patients with high Axl/Gas6 coexpression. Such a coexpression was particularly found in nonpseudopalisading tumor cells. As previously shown for other receptor tyrosine kinases (24), Axl and Gas6 may also form autocrine and paracrine stimulation loops. Gas6, originally identified to be induced under growth arrest conditions (7, 8), protects various neuroepithelial cell types from apoptosis in vitro (25–27). Cell death due to apoptosis and coagulative necrosis is fundamental in GBM (14, 15). Therefore, we suggest a second biological role of the Axl/Gas6 signaling pathway in glioma cell survival. Whereas Axl was only detectable in gliomas, Gas6 showed a strong expression in neurons of nontumoral tissue. A comparable neuronal expression of Gas6 mRNA has already been reported in the adult rat brain (26) and in primary cultures of rat hippocampal neurons (28, 29). In vitro Gas6/Axl signaling attenuated neuronal cell death caused by serum starvation (26), secretory phospholipase A2 (28), and amyloid β protein (29).

Furthermore, Axl and Gas6 were frequently coexpressed in vascular cells of tumor vessels. This included endothelial cells of microvascular hyperplasia, a form of tumor neoangiogenesis related to and induced by hypoxic pseudopalisading cells (14), but also endothelial cells, vascular smooth muscle cells, and pericytes of arterioles and venules adjacent to the GBM tissue. In contrast, such an expression pattern was not observed in vessels of nonneoplastic brain distant to the tumor and nontumoral tissue derived from epilepsy surgery, respectively. In previous in vitro and experimental in vivo studies, both proteins were found to be expressed in several vascular cell types regulating cell survival, proliferation, and migration (30–36). Axl has also been identified as a key modulator of tube and neointima formation (36) and vascular remodeling (37). In this study, we describe that Axl and Gas6 are also coexpressed in several vascular cell types of tumor vessels in humans in vivo, suggesting that Axl and Gas6 may be critically involved in neovascularization of GBMs.

Finally, we observed a strong Axl/GFAP coexpression in reactive astrocytes within and adjacent to the tumor. Reactive astrocytes have crucial functions in modulating the pathology and regenerative response to various brain lesions (38) and are able to secrete a range of astrocytic factors that may protect neuronal integrity and reduce microglial activation (38). In addition, astrocytic end feet, which are in close proximity to the microvessel wall to modulate the blood-brain barrier function (39), show a pronounced Axl staining. In many glial tumors, astroglial-endothelial signaling breaks down and leads to a leakage of capillaries (40) followed by tumor edema. Antiangiogenic therapy targeting receptor tyrosine kinases, such as vascular endothelial growth factor and platelet-derived growth factor receptors, has the potential to reduce tumor angiogenesis as well as capillary leakage of structurally and functionally abnormal tumor vessels (41). Therefore, we suggest that Axl may also have comparable functions in modulating tumor angiogenesis and blood-brain barrier similar to vascular endothelial growth factor and platelet-derived growth factor receptors.

The aggressiveness of GBM is reflected by a diffuse local infiltration into the brain parenchyma and a high vascularization supporting peripheral tumor growth (15). The present study shows that in glial tumors Axl is frequently expressed in both glioma and vascular cells and its biological effect is represented by the association with poor clinical outcome. At present, targeted therapies inhibiting receptor tyrosine kinases critically involved in oncogenesis and tumor angiogenesis seem to be one of the most promising therapeutic approaches for a variety of neoplastic diseases, including malignant gliomas (42). Therefore, the Axl/Gas6 signaling pathway represents a novel target for glioma treatment directed against tumor cell migration and survival as well as tumor neoangiogenesis. This could become clinically applicable by designing either small molecular inhibitors or gene expression interference approaches (42).