EGR‐1 is Regulated by N‐Methyl‐D‐Aspartate‐Receptor Stimulation and Associated with Patient Survival in Human High Grade Astrocytomas

Abstract

Early growth response‐1 (EGR‐1) is considered a central regulator in tumor cell proliferation, migration and angiogenesis and a promising candidate for gene therapy in human astrocytomas. However, conflicting data have been reported suggesting that both tumor promoting and anti‐tumor activity of EGR‐1 and its regulation, expression and prognostic significance still remain enigmatic. Our study explored EGR‐1 expression and regulation in astrocytomas and its association with patient survival. As we detected two EGR‐1 mRNA variants, one containing a N‐methyl‐D‐aspartate‐receptor (NMDA‐R) responsive cytoplasmic polyadenylation element (CPE), further experiments were performed to determine the functional role of this pathway. After NMDA stimulation of SV‐FHAS and neoplastic astrocytes, real‐time polymerase chain reaction showed an increase of the CPE, containing EGR‐1 splice variant only in astrocytoma cells. The surface expression and functionality of NMDA‐R were demonstrated by flow cytometric analysis and measurement of increased intracellular Ca²⁺. EGR‐1 was mainly restricted to tumor cells expressing NMDA‐R and significantly up‐regulated in astrocytic tumors compared with normal brain. Further, EGR‐1 expression was significantly (P < 0.007) associated with enhanced patient survival and was an independent prognostic factor in multivariate analysis in high grade astrocytomas. The NMDA‐R‐mediated EGR‐1 expression, therefore, seems to be a promising target for novel clinical approaches to astrocytoma treatment.

INTRODUCTION

Despite improvement of surgical treatment and progress in radio and chemotherapeutical strategies, median survival times in diffuse astrocytomas remain as low as 1.6 years in anaplastic astrocytoma and only 0.4 years in glioblastoma (23). The migratory and proliferative capacity of astrocytomas which is accused to be at least partly responsible for the poor prognosis might be associated with Ca²⁺ permeable glutamate receptors 13, 20, 26. Along those lines, glutamate levels in human gliomas correlate with the degree of cell proliferation (2). Cellular mechanisms linking glutamate receptors to stimulation of migration and proliferation so far remain elusive but may include the immediate early gene—early growth response‐1 (EGR‐1) (also known as Krox24, Zif268, NGFI‐A, Zenk, tis8, etr103 or d2) which is regulated upon glutamate receptor activation 5, 8, 30. However, in tumor cells EGR‐1 may both, stimulate and suppress malignant progression. In glioblastoma cell lines, EGR‐1 binds to GC‐rich sites of the human transforming growth factor‐beta1 promotor leading to decreased proliferation and suppressed transformation. At the same time, tumor cell migration is stimulated via activation of fibronectin 4, 18. In carcinoma cell lines, EGR‐1 protein expression seems to be responsible for various features of tumor progression, such as proliferation, migration, chemoinvasion and angiogenesis 7, 22, 23. In contrast, sustained expression of EGR‐1 in endothelial and fibrosarcoma cells blocks angiogenesis and tumor growth (19). In glioma cell lines, EGR‐1 may down‐regulate cyclin D1 and may, thus, exert a tumor suppressing function (27). In prostatic cancer, EGR‐1 expression increases with the grade of malignancy (6). In nephroblastomas enhanced EGR‐1 expression is associated with an unfavorable clinical prognosis (9). In contrast, in human breast and lung cancer, EGR‐1 expression decreases with the grade of malignancy 11, 16. Tumor suppressor capacity of EGR‐1 is further suggested by (i) direct transactivation of the phosphate and tensin homologue deleted on chromosome 10 (PTEN) gene; (ii) induction of p53; and (iii) binding of transcription factor c‐Jun leading to increased proapoptotic activity 17, 32. EGR‐1 activation could be mediated by N‐methyl‐D‐aspartate‐receptor (NMDA‐R) stimulation as the EGR‐1 gene possesses two different potential NMDA‐R‐responsive cytoplasmic polyadenylation elements (CPE1 and CPE2) which are located close to the polyadenylation signal (HEX) (Figure 1) (29). These CPEs are responsible for the post‐transcriptional translation of mRNAs that probably belongs to a dormant cytoplasmic mRNA pool 21, 33. Signal transduction upon NMDA‐R stimulation is mediated by the activation of the Aurora kinase which, in turn, phosphorylates the CPE binding protein which is required for a polyadenylation‐dependent translation (12). Until now, it is not clear how EGR‐1 expression is regulated and why it exerts a dual role in astrocytomas. Further, it remains to be established how the prognosis of astrocytomas may be related to the EGR‐1 expression. In the present study, we, therefore, address the NMDA‐R dependent regulation and prognostic impact of EGR‐1 expression in human astrocytomas.

Figure 1.

Schematic illustration of the early growth response‐1 (EGR‐1) gene and mRNAs with two full‐sequence entries found in GenBank. At the 3′UTR (untranslated region), two potential N‐methyl‐D‐aspartate‐receptor‐responsive cytoplasmic polyadenylation elements (CPE1 and CPE2) in close relation to polyadenylation signal (HEX) are depicted. These CPEs are only found in one EGR‐1 mRNA variant. Phylogenetic conservation values are indicated.

MATERIALS AND METHODS

Cell culture and stimulation

The human glioblastoma cell lines—LN229, LN428, U158, U251, U373 (kindly provided by N. de Tribolet, Lausanne, Switzerland) and the immortalized astrocytic cell line—SV‐FHAS (kindly provided by A. Muruganandam, Ottawa, Canada) were grown in DMEM supplemented with 10% fetal bovine serum and antibiotics (penicillin/streptomycin) under standard cell culture conditions. LN229 or SV‐FHAS were either treated with 10 µM NMDA or with 50 mM KCl and assessed for viability and EGR‐1 expression at 0, 10 and 30 minutes after exposure. KCl was applied as an unspecific stimulating agent. Additionally, cellular viability was tested by cell counter and analyzer systems (CASY). For each time point, 1 million cells were collected and frozen for further analysis.

Primary glioblastoma cell culture

Ex vivo glioma cell cultures, passages 4–9 of five glioblastoma patients were generated after tumor resection at the University Hospital of Zürich as previously described (24). All procedures were conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of the Canton Zürich. The diagnoses were confirmed at the Institute of Neuropathology, University Hospital of Zürich.

Real‐time reverse transcription–polymerase chain reaction

Total RNA was harvested from the tissue culture cells by using Trizol Reagent^® (Invitrogen, Carlsbad, California, USA). Consecutively, 1 µg of every RNA sample was reverse‐transcribed using Superscript II^® (Invitrogen, Carlsbad, California, USA) in the same run. Polymerase chain reaction (PCR) was performed with standard protocols using SYBRGreen^® (Molecular Probes, Eugene Oregon, USA) as fluorescent detection dye in a Biorad iCycler^® (Munich, Germany). Standard curves for five twofold dilutions of a reference RNA sample were generated to determine the PCR efficiency under the experimental conditions for the different transcripts and the housekeeping gene (GAPDH). All PCR reactions for a given sample were pipetted in triplicates, in order to control the variability of the PCR amplification. The two EGR‐1 variants were detected by GenBank research. The primer pairs used were the followings (sv—short variant; lv—long variant):

EGR‐1‐sv_for.:	5′‐AAAGTTTCACGTCTTGGTGCC‐3′,
EGR‐1‐sv_rev.:	5′‐GCTCAGCTCAGCCCTCTTCC‐3′,
EGR‐l‐v_for.:	5′GGCTTATAAACACATTGAATGCG‐3′,
EGR‐l‐v_rev.:	5′ACACCACATATCCCATGGGC‐3′.
Gapdh_for:	5′‐TCA ACA GCG ACA CCC ACT CC‐3′,
Gapdh_rev:	5′‐TGA GGT CCA CCA CCC TGT TG‐3′.

Flow cytometry

Glioma cell lines and primary glioblastoma cells were analyzed for the expression of NMDA‐R 1, the key receptor subunit of the heteromeric NMDA‐R using flow cytometry. The cells were washed twice with flow cytometry buffer (PBS without Ca, Mg, 2 mM EDTA, 0.5% BSA) and stained with either phycoerythrin‐labelled mouse immunoglobulin G_2a (IgG_2a) (BD Pharmingen, Heidelberg, Germany) for isotype control or with mouse monoclonal IgG_2a anti‐NMDA‐R 1 antibody (Chemicon International, Temecula, CA, USA). Cells were incubated on ice for 30 minutes in the dark and washed twice with cold flow cytometry buffer (3.5 mL, 350 g, 7 minutes) before they were analyzed on a FACS Calibur flow cytometer. The cytometer was equipped with a 488 nm argon‐ion laser and 635 nm red diode laser and Cell Quest 3.3 Software (BD Biosciences). The specific fluorescence index (SFI) was calculated as the ratio of the mean fluorescence values obtained with the specific NMDA‐R antibody and the isotype control antibody.

Measurement of intracellular Ca²⁺

Fura‐2 fluorescence was utilized for cytosolic Ca²⁺ determinations (31). LN229 cells were loaded with Fura‐2 (2.5 µM, Molecular Probes, Goettingen, Germany) for 30 minutes at 37°C. Fluorescence measurements were carried out with an inverted phase‐contrast microscope (Axiovert 100, Zeiss, Oberkochen, Germany). Cells were excited alternatively at 340–380 nm and the light was deflected by a dichroic mirror into the objective (Fluar 40×/1.30 oil, Zeiss, Oberkochen, Germany). Emitted fluorescence intensity was recorded at 505 nm and data acquisition was performed by Axon Imaging Workbench (Axon Instruments, Foster City, CA, USA). Experiments were made prior to, during and following exposure to Ca²⁺ free solution (5 mM ethylene glycol tetraacetic acid (EDTA) added).

Patients and tissues

We investigated a total of 193 brain tumor samples obtained from the brain tumor bank of the Institute of Brain Research (Neuropathology), University Tuübingen, from patients who underwent surgical treatment from 1993 to 2003. The specimens consisted of 26 World Health Organization (WHO) grade I (pilocytic), 37 WHO grade II and 52 grade III astrocytomas and 78 primary glioblastomas (WHO grade IV astrocytoma). 23 autopsy cases from an established normal brain bank were also included. The pathological diagnoses were confirmed by at least two neuropathologists. The histopathological typing and grading was done according to the WHO criteria for tumors of the nervous system (14). Glioma patients were followed for up to 12.6 years. Within this time frame, 102 patients died after a mean survival of 22.2 months (range 1–151); however, 78 were still alive with a mean follow‐up of 44.4 months (range 1–136). Thirteen patients were lost to follow‐up immediately after operation, so this population could not be integrated into the Kaplan–Meier survival analysis. For age‐stratified groups, patients were divided by median split. In the group of pilocytic astrocytomas, only one patient died during the time of observation; therefore, WHO grade I astrocytomas were not taken into account for survival analysis.

Tissue microarray

Tissue microarray (TMA) composite paraffin blocks were constructed by extracting cylindrical tissue core biopsies from different paraffin donor blocks and re‐embedding these into a pre‐punched hole on a single recipient (microarray) paraffin block at defined array coordinates (15). Using this technique, representative tissue sections were extracted and prepared as TMA (Beecher Instruments, Inc., Sun Prairie, WI, USA) with a core diameter of 600 µm. Prior to this TMA construction a hematoxylin & eosin stained slide of each block was analyzed to avoid inappropriate regions (eg, necrosis, hemorrhage) on the TMA slides. The TMA blocks were then cut with a microtome (3 µm thickness) and placed on SuperFrost Plus slides (Microm International, Walldorf, Germany).

Immunohistochemistry

All specimens were fixed in 4% formalin (pH 7.4), embedded in paraffin followed by preparation as TMA. Immunohistochemical stainings were performed, according to standard protocols. The following primary antibodies were used: rabbit polyclonal anti‐EGR‐1 against the carboxy terminus (dilution: 1:200; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) followed by a biotinylated secondary anti‐rabbit F(ab′)2 fragments (DAKO Cytomation, Hamburg, Germany). Exemplarily, double‐labeling was performed with Dako EnVision using 3,3′‐diaminobenzidine (DAB) (1:500; glial fibrillary acidic protein (GFAP); dilution: 1:500; Chemicon, Temecula, CA, USA) and Fast Red (EGR‐1) as chromogens. For controls, the primary antibodies were omitted.

Light microscopy and counting

Evaluation of the immunohistochemical stainings and photographic documentation was performed, using an Olympus Vanox AH‐3 light microscope. To quantify the reaction, 200 cells per sample were counted. In the tumor, we excluded endothelial, neuronal, parenchymal inflammatory and intra‐luminal cells by cellular morphology.

Immunofluorescence

Tumor tissue sections were double stained with mouse monoclonal IgG_2a anti‐NMDA‐R 1 antibody (dilution: 1:250; Chemicon) and EGR‐1 (dilution: 1:200; Santa Cruz Biotechnology). After deparaffinization, slides were incubated in citrate buffer (pH 6.0), for 30 minutes using microwave heat treatment. For detection of antibody binding with fluorochrome‐conjugated secondary antibody, the sections were blocked by incubation for 30 minutes with 5% (w/v) skimmed milk, 0.3% (w/v) Triton X‐100 (Serva, Heidelberg, Germany) and 0.4% (w/v) NaN₃ in TBS. Antibodies were diluted in the same solution and sections were incubated overnight at 4°C. After three washes in TBS for 10 minutes, sections were incubated for 45 minutes with the secondary antibody at room temperature. The secondary antibodies labeled with cyanin‐derivative dye Cy2 (dilution: 1:50) and Cy3 (1:100) were purchased from Dianova (Hamburg, Germany). Following washes in TBS, nuclei of the glioma cells were stained with SYTOX green (dilution 1:1000; Invitrogen, Karlsruhe, Germany). After further washes in TBS, glass slides were mounted in glycerol. For controls, the primary antibodies were omitted. Fluorescence was visualized with a confocal laser scanning microscope (Axiovert 135M, Zeiss, Oberkochen, Germany) and images were processed using Adobe Photoshop (version 6.0, Adobe, Mountain View, CA, USA).

Statistical analysis

For the analysis of PCR data, we processed the raw data CT (cycle threshold) values by calculating mean normalized expression values which reflect the relative expression of a target to the housekeeping gene (28). Statistical analysis was performed using a pair wise‐fixed allocation randomization test^® which has particularly been designed for the evaluation of quantitative real‐time PCR results. The average coefficient of variation of the triplicates was 12%. Accordingly, a difference of more than 55% between two samples had a statistical power of 95%. We, therefore, only assigned significance to differences which were greater than 55%. The significance levels were the following: P < 0.05 (*); P < 0.01 (**); P < 0.001 (***).

For the analysis of immunohistochemical EGR‐1 stainings, we used an analysis of variance (ANOVA) with subsequent Tukey HSD (honestly significant difference) test with a global significance level of 5% after arcsine transformation of the EGR‐1 positive fractions. The results were backtransformed in order to obtain the means and the 95% confidence intervals (CI) of the means. For the adjustment of the P‐values because of multiple testing, we used the method of Bonferroni‐Holm. To analyze patient survival in high grade astrocytomas (WHO grade III and IV), we performed an exponential parametric survival fit. To describe patient survival starting from the day of operation, a univariate survival analysis using product limit (Kaplan–Meier) life table was used and tested by the log‐rank method. Samples were stratified by protein expression into two groups of approximately equal size (high and low expression). Survival data were analyzed controlling for EGR‐1 expression, patient age, WHO grade, gender, Karnofsky performance score (KPS) localization, extent of resection and edema using the Cox proportional hazard model. As there were only few deaths among patients with WHO grade I and II astrocytomas, these grades were excluded from the survival analysis. Age was included as a continuous variable. The KPS was dichotomized such that values less or equal to 80 and more than 80 were combined. The extent of resection (full or partial resection), gender (female/male) and localization (supratentorial/infratentorial) were included as dichotomous variables. The edema level was analyzed in pre‐operative T2 magnetic resonance imaging (MRI) scans in most cases or—if no MRI was available—in cranial computed tomography (CCT) scans. Edema levels were classified as follows: no = 0; low = 1; high = 2. Edema scores were analyzed in cooperation with the Department of Neuroradiology (University of Tübingen). JMP 7 software (Cary, USA) was used for statistical analysis.

RESULTS

Predominantly CPE‐containing EGR‐1 variants increase upon NMDA stimulation

Under non‐activating conditions, the expression of EGR‐1‐sv is about fivefold higher than the expression of EGR‐1‐lv (1, 2) in both the malignant glioma cell line LN229 and the immortalized astrocytic cell line SV‐FHAS. After 10 minutes treatment with NMDA, only the glioma cell line LN229 showed a significant 8.6‐fold induction of EGR‐1‐lv and a less pronounced 2.2‐fold induction of EGR‐1‐sv (Figure 2). While EGR‐1‐lv still showed a significant 3.3‐fold increase at 30 minutes, EGR‐1‐sv was insignificantly decreased 1.7‐fold at this point. Depolarization with KCl induced similar but less pronounced effects in LN229 (Figure 2A and B). In contrast, SV‐FHAS cells only responded to depolarization with KCl whereas no significant increase was detected after NMDA stimulation. Cell viability was not influenced in a relevant manner by the experimental conditions at each time point (data not shown).

Figure 2.

Expression of EGR‐1‐sv (A,C) and EGR‐1‐lv (B,D) in both the malignant glioma cell line LN229 (A,B) and the immortalized astrocytic cell line SV‐FHAS (C,D) under resting conditions and after stimulation with N‐methyl‐D‐aspartate or KCl. Cell viability was not influenced in a relevant manner by the experimental conditions at each time point (data not shown).

Glioblastoma cells express the NMDA receptor

To address the question to what extent astrocytic tumor cells express NMDA receptors at the cell surface, we investigated LN229 by flow cytometry. About 3% of all measured tumor cells strongly expressed NMDA‐R on the cell surface. The other 97% of LN229 cells showed a slight to moderate shift toward positive (SFI: 1.6) (Figure 3). In addition, four different cell lines and five patient‐derived ex vivo glioblastoma cell cultures were assessed for their NMDA receptor expression by flow cytometry. The positively gated NMDA‐R fractions reached 15.93% with a SFI of 1.83 at maximum (Table 1).

Figure 3.

Flow cytometry of N‐methyl‐D‐aspartate‐receptor (NMDA‐R) expression in LN229 glioma cells. The panels show flow cytometric analysis of fluorescence/side scatter (FL1/SSC) for isotype control (upper panel) and analysis of LN229 cells after incubation with the mouse monoclonal immunoglobulin G_2a (IgG_2a) anti‐NMDA‐R 1 antibody (middle panel). 2.91% of LN229 cells stain positive for NMDA‐R 1 and can be grouped in high, medium, low positive cells for FL1. FL1 is illustrated as histograms (lower panel) comparing isotype control (open bold line) and NMDA‐R antibody (filled). 96.96% of the cells show a slight shift toward positive, but remain negative for FL1 (x‐axis: number of events; y‐axis: fluorescence intensity). Shown are data from one representative experiment (n = 3). Abbreviation: PE = phycoerythrin.

Table 1.

NMDA‐R expression in established glioma cell lines and ex vivo glioblastoma cells assessed by flow cytometry. Abbreviations: NMDA‐R = N‐methyl‐D‐aspartate‐receptor; SFI = specific fluorescence index; Ig = immunoglobulin.

Cell line/primary cells	SFI*	NMDA‐R positive cells (%)
U158 (cell line)	1.18	6.16
LN428 (cell line)	1.30	2.76
U373 (cell line)	1.39	5.02
U251 (cell line)	1.83	15.93
LN229 (cell line)	1.60	2.91
Glioblastoma 1 (primary)	1.16	2.58
Glioblastoma 2 (primary)	<1	0.24
Glioblastoma 3 (primary)	1.30	3.52
Glioblastoma 4 (primary)	1.01	4.33
Glioblastoma 5 (primary)	1.01	0.06

^* Fluorescence value obtained with NMDAR‐I antibody staining/fluorescence value obtained with IgG isotype control antibody.

NMDA receptors are functionally active in LN229 glioma cells

In order to determine the effect of NMDA treatment on the regulation of intracellular Ca²⁺ activity, we treated LN229 cells with 10 µM NMDA. This treatment is known to accelerate Ca²⁺ entry. The Fura‐2 fluorescence ratio (340/380 nm) in the presence of extracellular Ca²⁺ was significantly lower before NMDA treatment (4.04 ± 0.37, n = 4) than after NMDA treatment (6.56 ± 0.32, n = 4) (Figure 4). These results show that the astrocytic tumor cells possess functionally active NMDA receptors.

Figure 4.

Increase of Ca²⁺ entry into LN229 cells. A. Representative original tracings showing the Fura‐2 fluorescence ratio (340/380 nm) in Fura 2 loaded LN cells 229 exposed initially to Ca²⁺ free Ringer (0 Ca²⁺, 5 mM EGTA), to Ringer containing 1 mM extracellular Ca²⁺. During exposure to Ca²⁺ solution N‐methyl‐D‐aspartate (NMDA) (10 µM) was added to stimulate the NMDA receptors. B. Mean (±standard error of the mean; n = 4) of the Fura‐2 fluorescence ratio after incubation as in (A) in Ca²⁺ free Ringer, in Ca²⁺ containing extracellular fluid, and in Ca²⁺ and NMDA containing Ringer. * indicates significant (P ≤ 0.05, ANOVA) difference to respective value prior to removal of Ca²⁺.

The nuclear expression of EGR‐1 in human astrocytic tumors in vivo is strongly associated with NMDA receptor expression

Immunofluorescent staining (n = 5 glioblastomas) showed that EGR‐1 expression in human glioblastomas was restricted to the nucleus (Figure 5A). In double staining experiments, nuclear EGR‐1 protein was mainly seen in cells strongly co‐expressing the NMDA‐R on the cell surface (Figure 5B).

Figure 5.

Immunofluorescence of early growth response‐1 (EGR‐1) and N‐methyl‐D‐aspartate (NMDA) receptor in human glioblastomas. A. A strict nuclear localization of EGR‐1 in a subset of glioblastoma cells was detected (EGR‐1: red; nuclear staining: Sytox green). Yellow staining reflects the co‐expression of EGR‐1 with the nuclear Sytox green staining. B. Nuclear EGR‐1 protein (green) is strongly associated with NMDA receptor (red) expression in human glioblastomas.

EGR‐1 is significantly up‐regulated in human astrocytomas compared with normal central nervous system tissue

In normal central nervous system tissue, classical immunohistochemistry revealed that about 9.4% (CI: 6.7–12.5%) of all glial cells expressed EGR‐1. In comparison with normal brain, all grades of astrocytic tumor displayed significantly higher EGR‐1 levels. EGR‐1 expression was mainly found within the nuclei but occasionally also in the cytoplasm (Figure 6A and B). Double staining experiments showed that EGR‐1‐positive cells co‐expressed GFAP, corroborating their astrocytic origin (data not shown). The EGR‐1 expression levels were in a similar range throughout all WHO grades with a mean expression rate of 16.9% (CI: 13.2%–20.9%) in pilocytic astrocytomas, 15.5% (CI: 12.8%–18.4%) in diffuse and 17.8% (CI: 15.3%–20.4%) in anaplastic astrocytomas and 19.4% (CI: 17.3%–21.6%) in glioblastomas (Figure 6A and B).

Figure 6.

Early growth response‐1 (EGR‐1) immunohistochemistry. Two glioblastoma samples are depicted showing increased nuclear EGR‐1 immunoreactivity (arrows in A and B) in areas with higher cellularity whereas tumor areas exhibiting lower cell density (asterisk in A) or even resembling mainly normal central nervous system tissue (asterisk in B) show less EGR‐1 positivity.

EGR‐1 expression is significantly associated with enhanced patient survival in high grade astrocytomas and an independent prognostic factor in multivariate analysis

As WHO grade III and IV astrocytomas showed no significant differences of EGR‐1 expression, a combined survival analysis was performed. The exponential parametric survival fit revealed a highly significant positive association of EGR‐1 expression and patient survival in high grade astrocytic tumors (P = 0.007) (Figure 7A). In a Kaplan–Meier analysis where EGR‐1 expression levels were split by median, the log rank test also demonstrated significantly longer survival rates for the group exhibiting higher EGR‐1 expression levels (Figure 7B). In WHO grade II astrocytomas, the exponential parametric fit did not show a significant association of EGR‐1 levels and patient survival (P =  0.43). In the group of WHO grade I astrocytoma, only one patient died; therefore, pilocytic astrocytomas were not included in survival analysis. A multivariate analysis performed in high grade astrocytomas included EGR‐1 expression, patient age and WHO grade, which are both known to be strong predictors for patient survival in human astrocytomas, as well as gender, Karnofsky score, tumor localization, extent of resection and edema score (Table 2). As previously reported, patient age had the strongest association with survival, and was a stronger predictor of survival that WHO grade (3). Interestingly, for EGR‐1, a significant association with survival remained in multivariate analysis (P = 0.0224) with a P‐value in the range of the WHO grade (0.0201). In contrast, gender, Karnofsky score, tumor localization, extent of resection and edema could not withstand the multivariate analysis.

Figure 7.

Association of early growth response‐1 (EGR‐1) expression with patient survival in high grade astrocytomas (World Health Organization grade III and IV). A. Exponential parametric survival fit analyzing survival time (note log scale) in relation to EGR‐1 expression (arcsine transformed % values) showed a highly significant positive association of patient survival with higher EGR‐1 expression rates (P = 0.007) (x‐axis: EGR‐1 expression; y‐axis: survival time). B. Kaplan–Meier analysis of the same data dichotomized by a median EGR‐1 expression split revealed a significant prognostic benefit of patients with high EGR‐1 levels (log rank: P = 0.035; green line: high EGR‐1 expression level; red: low EGR‐1 expression level; n = 130).

Table 2.

Multivariate analysis: survival data controlled for EGR‐1 expression, patient age, WHO grade, gender, Karnofsky performance score, localization, extent of resection and edema. Hazard ratios for continuous variables refer to one unit of the corresponding variable. Abbreviations: EGR‐1 = early growth response‐1; WHO = World Health Organization; CI = confidence intervals.

Variable	Results of multivariate Cox regression
	P‐value	Hazard Ratio (95% CI)
EGR‐1 expression	0.0224	0.06 (0.01–0.67)
Patient age	0.0001	1.05 (1.02–1.07)
WHO grade	0.0201
WHO grade IV vs. III		2.21 (1.12–4.47)
Gender	0.327	0.88 (0.68–1.13)
Karnofsky score	0.87	1.00 (0.97–1.02)
Localization	0.05	4.47 (1.00–11.59)
Extent of resection	0.14	1.22 (0.94–1.63)
Edema	0.19
Edema: low vs. no		1.21 (0.48–3.58)
Edema: high vs. low		1.60 (0.90–2.96)

DISCUSSION

EGR‐1 is known as a central regulator of expression because of its capacity to control a large number of different genes. However, the functional role of EGR‐1 is still a matter of controversy. Being considered as a tumor suppressor gene in some tumors like breast or lung carcinoma, EGR‐1 is perceived as a tumorigenic factor in other tumors such as prostate cancer 6, 11, 16, 22. Further, EGR‐1 is of interest in diffusely infiltrating astrocytomas, as it has been shown to almost completely abolish tumor cell growth and migration, features that contribute to the incurability of these tumors (4). Therefore, EGR‐1 was proposed as a most promising candidate for future gene therapy in human astrocytomas (4). Hence, the aim of the present study was to analyze the expression, regulation and prognostic power of EGR‐1 in human astrocytomas. By GenBank research, we detected two potential NMDA‐R‐responsive cytoplasmic elements in the EGR‐1 gene (Figure 1). Further, we recovered two different EGR‐1 mRNA variants, of which only the elongated variant (EGR‐1lv) possessed the potential NMDA‐R‐responsive CPE. EGR‐1lv showed early (after 10 minutes) and highly significant increase in response to NMDA in LN229 glioma cells (Figure 2). In contrast, the shorter mRNA (EGR‐1sv) only exhibited a very late increase. This effect was restricted to neoplastic astrocytes whereas SV‐FHAS cells did not react to NMDA stimulation. These findings suggest that EGR‐1 could be regulated in a sequence‐dependent manner in human astrocytomas. Flow cytometric analysis revealed that several astrocytomas cell lines and primary cultures express NMDA‐R (Table 1, Figure 3). In additional experiments, we could demonstrate that these NMDA‐Rs are functionally active in the same cell line (Figure 4). The NMDA‐R pathway may, thus, be a functional regulatory element for EGR‐1 expression in human astrocytomas. Localization of EGR‐1 protein revealed a predominantly nuclear pattern which was strongly associated with membranous NMDA‐R expression (Figure 5). It can be concluded that (i) NMDA‐R seems to influence EGR‐1 protein expression; and (ii) the rapid up‐regulation and nuclear localization reflect the role of EGR‐1 as fast and central regulator of gene expression. Along those lines, EGR‐1 was recently shown to directly suppress different oncogenes and therefore the development of malignancies (10). As previously reported, we could demonstrate EGR‐1 expression in normal brain glial cells (1). A surprising result was the up‐regulation of EGR‐1 in all astrocytic WHO grades compared with normal brain as one would assume a decrease or even a loss of a tumor suppressor gene during malignant transformation (25). The constant EGR‐1 levels among different WHO grades are difficult to explain. As EGR‐1 is a transcription factor which can be induced by various stimuli and is involved in many different pathways, the meaning of similar EGR‐1 levels in both low and high grade astrocytomas might be due to a changing tumor micro‐environment in neoplasms of different grades of malignancy. There are also other, to date unexplained molecular and histological features such as microvascular proliferations that are present in both WHO grade I and IV astrocytomas but do not carry the same predictive value. The highly significant association of EGR‐1 expression with prolonged patient survival in high grade astrocytomas supports the hypothesis of a tumor suppressive capacity of EGR‐1 that is still functional after malignant transformation of astrocytic cells. The positive association of EGR‐1 with patient survival in human high grade astrocytic tumors is in line with the results of esophageal squamous cell carcinomas (34). In high grade astrocytomas, EGR‐1 remained a significant independent predictor of patient survival in multivariate analysis reaching a similar significance as the WHO grade. Both factors were inferior only to patient age which is known as a very powerful factor influencing patient survival in human astrocytomas (3). Even though EGR‐1 DNAzymes suppressed tumor proliferation, migration and angiogenesis in breast cancer cell lines, these promising results probably can not without reservations be translated to human astrocytic tumors (7). Other tumor suppressive regulatory mechanisms of EGR‐1 seem to be stronger than the suggested tumorigenic effects. The blockage of glutamate receptors like alpha‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic acid (AMPA) receptors have been suggested as a possible therapeutic strategy for the prevention of astrocytoma invasion (13). Our results indicate that there is no basis for such attempts concerning the NMDA‐R EGR‐1 axis. Therefore, further studies are needed to elucidate the dichotomic gene regulatory function of EGR‐1 in human astrocytomas.