Background
GBM is the highest grade of astrocytoma and is also the most common primary brain tumor in adults. Approximately 50% of patients with GBM die within a year of diagnosis, despite the use of many aggressive treatment approaches [
1]. Lack of reliable prognostic markers for these patients is a hindrance to improving therapy and individualizing therapeutic interventions. Amplification and/or overexpression of the
EGFR gene, mutation of the
p53 gene, and proliferation indices have all been proposed to predict survival of patients with GBM and to play a role in the pathophysiology of their tumors [
2,
3]; however, other studies have shown no such association with outcome [
4‐
6]. One reason for this discrepancy is that strong clinical factors such as patient age need to be included [
7,
8]. Although clinical parameters such as age, Karnofsky performance status at diagnosis, and extent of resection are routinely used in clinical practice to predict the outcome of patients with GBM, none of these variables have a direct connection with tumor pathogenesis.
In a previous study, gene expression profiling of a group of GBM specimens identified a cluster of about 50 named genes whose expression was inversely associated with survival [
9]. In examining the annotations of "biological process" in the Gene Ontology terms for each gene [
10], the annotation "neurogenesis" appeared most frequently, suggesting a common role for these genes in central nervous system development. In contrast, a number of other annotations for biological process such as "cell proliferation," "inflammatory response," and "immune response" were underrepresented in these genes. Because several of these genes are involved in cell-cell and cell-matrix interactions and cell migration, we hypothesized that their increased expression might be related to more infiltrative and aggressive tumor behavior. Based on the results of the preceding analyses and the availability of antibodies, we chose to investigate the prognostic value of one gene,
FABP7, in greater detail [
9].
Although FABP7 is a cytoplasmic protein, its varying subcellular localization between nucleus and cytoplasm has been reported in developing brain [
11], glioma cell lines [
12], and GBM specimens [
9]. Increased FABP7 expression was also found in glia following nerve injury [
13,
14]. We separately scored FABP7 immunoreactivity in nucleus and cytoplasm, and found that nuclear FABP7 immunoreactivity was inversely correlated with survival of patients with GBM, particularly in younger cases [
9]. This result is consistent with other reports that emphasize the effect of age upon various prognostic factors [
7,
8], and such a pattern is similar to a recent finding of an association between EGFR overexpression and poor prognosis in younger GBM patients [
7].
FABP7 is a member of the multi-gene fatty acid-binding protein (FABP) family and binds to very-long-chain polyunsaturated fatty acids (C > 16) such as docosahexaenoic acid (DHA) with very high affinity
in vitro [
15]. FABP7 appears to have different roles in different tissue types. It is highly expressed in radial and Bergmann glial cells throughout the developing central nervous system and gradually declines in the adult [
16,
17]. FABP7 is required for neuron-induced glial differentiation and subsequent migration of neurons along the glial processes, but has no effect on cell proliferation and adhesion [
11]. In Schwann cells, FABP7 expression is downstream of the Ras-independent EGFR signaling pathway, and it regulates interactions between Schwann cells and axons in normal peripheral nerves and peripheral nerve tumors without affecting cell proliferation and migration [
13]. Differential expression of different types of FABPs is found in several types of tumors and their normal-cell counterparts, and FABPs have been shown to modulate growth and differentiation of normal and neoplastic cells [
18]. FABP7 was also shown to induce mammary differentiation and to mediate growth inhibition of breast cancer cells [
19,
20].
It has been suggested that FABPs increase the solubility of fatty acids in the cytoplasm when transporting fatty acids between membrane compartments, and bring fatty acids to their nuclear targets [
21]. In addition, subcellular localization of FABPs appears to be tissue specific and closely associated with gene expression and function. Liver-type FABP targets fatty acids to the nucleus and interacts with peroxisome proliferator-activated receptors α and γ to regulate gene expression [
22‐
24]. Soluble FABP7 was found to induce differentiation of the mammary gland in mice [
20]. FABP7 is present in both the nucleus and cytoplasm in glial cells, as well as in the conditioned media of glial cells in culture [
11], but is found only in the cytoplasm of Kupffer cells in the liver [
25].
In this study, we examined the expression patterns and subcellular localization of FABP7 in specimens from normal individuals, from individuals with gliosis only, and from patients with gliomas differing in grade and histology. We also used independent sets of GBM specimens to analyze the relationship of subcellular localization of FABP7 with patient outcome and EGFR expression, and to seek possible mechanisms underlying the functions of FABP7 in GBM.
Methods
Cell culture
Glioma cell lines were obtained from the Neurosurgery Tissue Bank at the University of California, San Francisco. Immortalized human astrocytes were provided by Dr. Russ Pieper (University of California, San Francisco) [
26]. All cells were maintained in Eagle's minimal essential medium with 10% FBS and 5% CO
2.
Tissue specimens
Frozen and paraffin-embedded specimens were obtained from the Neurosurgery Tissue Bank at the University of California, San Francisco, and University of Texas, M. D. Anderson Cancer Center after approval from the Committee on Human Research. Tissue sections for immunohistochemistry were of 5 μm in thickness. Gliotic and normal brain tissues were obtained from epileptic patients and postmortem specimens, respectively. Clinical data of patients with primary GBM that were used for analyzing the correlation between FABP7 and EGFR immunoreactivity are summarized in Tables
2 and
4. Two cohorts of GBM patients (61 and 44 cases, respectively) for EGFR expression analysis are the same sets used in a previous study [
9].
Antibodies
FABP7-specific polyclonal antibodies were gifts from Drs. N. Heintz (Rockefeller University, New York, NY) and R. Godbout (University of Alberta, Alberta, Canada). Antibodies from both sources produced similar staining patterns and specificity on Western blots and immunohistochemistry using GBM specimens ([
12] and data not shown). A dilution of 1 to 400 was used for both immunostaining and migration assays, and a dilution of 1 to 500 was used for immunoblotting; secondary antibodies alone did not show detectable signal. Dilution of antibodies against glial fibrillary acidic protein (GFAP) (ICN; Costa Mesa, CA) and EGFR (clone F4; Sigma, St. Louis, MO) for immunohistochemistry was 1:1000 and 1:400, respectively. Peroxidase-conjugated and biotinylated secondary antibodies were obtained from Vector Laboratories (Burlingame, CA). Fluorescine-conjugated and Rhodamine-conjugated secondary antibodies and normal rabbit serum were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA).
Western blot analysis
Total RNA was extracted from frozen tissues specimens using Trizol (Invitrogen; Carlsbad, CA) as described previously [
9], and genomic DNA was removed from the interphase and organic phase by ethanol precipitation. The protein fraction was then purified by isopropanol precipitation, washed several times in 0.3 M guanidine hydrochloride in 95% ethanol, and resuspended in 1% SDS. The protein concentration of each sample was quantitated by using a D
c Protein Assay Kit (Bio-Rad; Hercules, CA), and equal amounts of protein for each sample were separated by SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad), blocked with 10% skim milk, incubated with specific antibodies, and visualized using a Super Signal West Pico Chemiluminescent kit (Pierce; Rockford, IL).
Immunohistochemistry
All frozen tissue sections used for immunohistochemistry were fixed in 4% formaldehyde, treated with H
2O
2, blocked with normal serum, incubated with primary antibodies at 4°C overnight or room temperature (RT) 2 hours, incubated with biotinylated secondary antibody and peroxidase-labeled streptavidin at RT for 30 min, to visualize the immunoreactivity with the DAB Reagent kit (KPL; Gaithersburg, MA). Staining of paraffin-embedded sections followed the same protocol, except for prior de-waxing and antigen retrieval by microwave heating. Immunostaining and semi-quantitative scoring of p53 and EGFR expression were performed as previously described [
7].
SF763 glioma cells were plated and incubated overnight in Lab-Tek chamber slides (Nalge Nunc International; Rochester, NY) followed by 24 hours of 0.5% serum starvation and 2 days of 50 ng/ml EGF (Invitrogen) treatment. After fixation in 4% formaldehyde, cells were blocked with normal serum, followed by 2 hours of RT incubation with the primary antibody and 1 hour of RT incubation with the secondary antibody. Cells were then covered with Vectashield (Vector Laboratories) to prevent fading of fluorescence. The fluorescence intensity of 200 control or EGF-treated cells was digitally recorded and the ratios were calculated.
Antisense inhibition and migration assay
Antisense oligodeoxynucleotides (ODNs) used were complementary to the position -13 to 7 of the FABP7 cDNA, and sense ODNs complementary to the same region were used as a control. The first 3 and the last 3 phosphodiester bonds on the ODNs were modified to phosphorothioate bonds to prevent degradation. SF763 glioma cells were serum-starved in 0.5% serum-containing medium for 24 hours, followed by 2 days of 50 ng/ml EGF treatment in the same low-serum media. Control cells were maintained in the low-serum medium for 2 days. ODNs were incubated with FuGene (Roche, Basel, Switzerland) at RT for 30 min, and then added to SF763 cells on the second day of EGF treatment to a final concentration of 100 nM.
The inserts of TransWell chambers (Corning, Corning, NY) with 5μm pores were incubated with 100 μg/ml of rat-tail type 1 collagen (BD Biosciences, San Jose, CA) overnight at room temperature, and washed with phosphate-buffered saline (PBS). At the end of the 2-day EGF treatment, cells were dislodged using 2 mM EDTA in PBS and then resuspended in the same treatment media as before (control or EGF, sense or antisense ODNs). Low-serum medium was placed in the bottom well, and 1 × 104 cells were plated into each insert. After 4 hours, un-migrated cells were removed with cotton swabs and migrated cells were fixed and stained using a HEMA 3 stain set (Fisher Diagnostics, Middletown, VA). For each insert, cell numbers were counted from five randomly chosen fields under 200× magnification.
Data analysis
All statistical analyses used SPSS for Windows (Release 11.5.0). The fluorescence intensity recorded using the Openlab software (Improvision, Lexington, MA) and migration data were analyzed using Student's t test. Correlation of nuclear localization of FABP7 with patient survival was analyzed using the Cox proportional hazards regression. Hazard ratios provide information about the direction of an association (a numeral over 1 indicates an increased risk with the positive variable, and a numeral under 1 indicates a decreased risk) as well as the magnitude of the risk. To evaluate the relationship between nuclear FABP7 and EGFR upon patient survival, new variables were used to divide patients into four groups based on the immunoreactivity of nuclear FABP7 and EGFR of their tumors: dual negative as "0", nuclear FABP7-negative/EGFR-positive as "1", nuclear FABP7-positive/EGFR-negative as "2", and dual positive as "3". Bivariate correlations were evaluated by the Spearman test. A p value < 0.05 was considered statistically significant for all tests.
Discussion
In previous work, we identified nuclear FABP7 immunoreactivity as a prognostic marker for patients with GBM [
9]. In a separate report, increased expression of FABP7 was also found in GBM patients surviving less than 2 years compared to those who survived longer [
29]. Two cellular functions for FABP7 have been identified: cell migration and differentiation in the developing central nervous system [
11], and Schwann cell-axonal interactions in the peripheral nervous system [
13]. Although suggestive, neither offers a clear explanation of the action of FABP7 in gliomas. The goals of this study were to determine a plausible biological role for FABP7 in glioma pathogenesis. We characterize the expression of FABP7 in normal brain, gliotic tissues, and glial tumors. We show that increased FABP7 expression occurs in a subset of reactive astrocytes, that FABP7 expression is restricted to cells of astrocytic lineage in glioma, and that there is almost no nuclear FABP7 immunoreactivity in well-circumscribed pilocytic astrocytoma. In addition, we establish an association between nuclear FABP7 and EGFR expression both in human GBM tumors and in a glioma cell line. Our data suggest that FABP7 might play a role in GBM pathogenesis through its participation in the EGFR signaling pathways. Most importantly, we identify nuclear FABP7 immunoreactivity as a marker to predict the outcome of patients with EGFR-positive GBM.
Transcription of both
FABP7 and
EGFR genes showed marginal association based upon the microarray datasets, but only nuclear FABP7 immunoreactivity correlated with EGFR expression. Age is one of the most important prognostic factors for GBM patients, and it has been shown in separate studies that the status of EGFR expression is a better prognostic factor in younger patients [
7,
33]. Our data indicate that the prognostic value of nuclear FABP7 is also influenced by patient age [
9]. However, the correlation between nuclear FABP7 and EGFR expression did not have preference to patient age in our study. Collectively, it appears that other factors (age is probably one of them, see below) yet to be identified participate in regulating the transcriptional and translational mechanisms shared by both
FABP7 and
EGFR genes, as well as in regulating the subcellular localizations of FABP7. This is compatible with the fact that EGFR expression and nuclear FABP7 immunoreactivity are mutually exclusive in a significant portion our clinical specimens. It will be pressing to define in the future why nuclear FABP7 is not present in some EGFR-positive GBM and whether other EGFR-downstream pathways are active in EGFR-negative/nuclear FABP7-positive tumors.
It has been previously reported that EGFR signaling induces FABP7 expression in a Ras-independent pathway in normal and tumor Schwann cells [
13], providing evidence of possible interaction between these two proteins. Our demonstration in GBM specimens that expression of both FABP7 and EGFR correlated with each other at both protein and mRNA levels, and in SF763 glioma cells that EGFR activation induced nuclear translocation of FABP7, warrants further investigation of whether FABP7 is a direct downstream target of EGFR activation in GBM and which EGFR pathway FABP7 is associated with. It has been shown that liver-type FABP carries fatty acids to interact with nuclear receptors that in turn regulate gene expression [
22‐
24]. If FABP7 utilizes a similar mechanism to regulate gene expression in response to EGFR activation, this would expand the scope of EGFR signaling pathways in GBM tumors. Future studies will examine the specific molecular pathways linking EGFR and FABP7, and nuclear functions of FABP7.
In normal cerebral cortex, we identify a unique population of glia positive for FABP7 but negative for GFAP that we designated as Type 1 cells. The origin and exact role of these cells remains unclear. NG2, a chondroitin sulfate proteoglycan, is expressed by oligodendrocyte progenitors, and identification of one type of GFAP-negative/NG2-positive astrocytes in adult normal brain and a subset of gliomas led to a hypothesis that certain gliomas arise from the NG2-positive progenitor cells [
34]. According to the morphology and frequency of appearance [
35], Type 1 cells identified in this study clearly do not belong to this category. However, based upon the expression patterns of FABP7 during the development of central nervous system, in adult brain, and in gliomas, transformation of FABP7-positive cells may contribute to the histogenesis of a subgroup of gliomas.
In gliotic brain tissue, FABP7 expression is increased in a subset of reactive astrocytes and demonstrates variable subcellular localization in the cytoplasm and nucleus. Such differential patterns in both expression and subcellular localization of FABP7 are also seen in cells with astrocytic features in various types of glioma. As noted above, grade I pilocytic astrocytoma is the only type of glioma examined that does not show nuclear FABP7 staining in our studies. Although our size of sampling was limited, the statistical analyses clearly demonstrate that the chance of detecting nuclear FABP7 in pilocytic astrocytoma is small. Because pilocytic astrocytomas are well-demarcated lesions whose pattern of growth is clearly distinctive from diffusely infiltrative higher-grade (grades II to IV) astrocytoma and oligodendroglial tumors (both ODG and OAC), considering the association of nuclear FABP7 with poor prognosis of GBM patients, cytoplasmic localization of FABP7 may be associated with less infiltrative phenotype of neoplastic astrocytes. Interestingly, we did find positive correlation between cytoplasmic FABP7 immunoreactivity and patient survival in cases older than the medium age among the first set of 61 GBM specimens (data not shown), although this observation has not been validated in an independent set of samples. It would be important in the future to investigate the downstream signaling pathways for FABP7 in the nucleus and cytoplasm and their association with cell motility, and how the translocation of FABP7 is regulated, especially under the context of patient age.
Although FABP7 expression does not predict the outcome of patients with ODG and OAC [
36], FABP7 may have prognostic value for grade II and grade III astrocytomas due to its heterogeneous patterns of expression and subcellular localization in these two tumor types. In other experiments, we noted that immortalized non-tumorigenic astrocytes express similar amounts of FABP7 compared to glioma cell lines (data not shown). FABP7 overexpression in glioma cells does not affect cell cycle progression and activation of apoptosis [
9]. In gliotic brain tissue increased FABP7 expression coincides with GFAP expression in a subset of reactive astrocytes. Increased expression of FABP7 was seen in brain tissues after systematic administration of a neurotoxin, kainic acid [
14]. FABP7 was released into patients' serum after acute ischemic stroke [
37]. For these reasons, FABP7 expression or the presence of nuclear FABP7 alone is unlikely a factor unique to glioma oncogenesis and progression. On a practical level, however, anecdotal experience suggests that occasional cases exist where the main differential diagnosis is pilocytic astrocytoma versus GBM (both of these astrocytic tumors exhibit microvascular proliferation), and the presence of nuclear FABP7 would support the diagnosis of GBM.
In addition to this report, elevated levels of FABP7 mRNA and protein in GBM specimens compared to those in normal adult brain have been previously demonstrated [
9,
12]. Opposite results were reported in breast and prostate cancer, where decreased expression of FABP7 was found in tumor specimens when compared to normal tissues [
19,
32]. Greater FABP7 expression was seen in rarely metastasized melanoma cell lines compared to their frequently metastasizing counterparts [
38]. Poorly-differentiated prostate tumors lose FABP7 expression, but more FABP7 is expressed in well-differentiated prostate cancer specimens than in primary normal prostate cells [
32]. Ectopic expression of FABP7 induces cell differentiation and suppresses tumor growth [
19,
20] and was shown to mediate the cytotoxicity of DHA to breast cancer cells [
20]. One explanation for these divergent results in patterns of expression and effects of forced expression is that functions of FABP7 in neoplastic and normal cells are tissue-specific. A strong candidate factor involved in such tissue-specific functions is EGFR, since we observed in clinical specimens the correlation of gene expression between
FABP7 and
EGFR, and the correlation of nuclear FABP7 immunoreactivity with EGFR expression; however, this association appears to be only in GBM, but not in normal brain, gliotic tissues, or other types of glioma examined.
Our immunohistochemical results show the change of subcellular localization of FABP7 in normal, gliotic, and neoplastic brain tissues and suggest several interesting avenues for future studies. Based on its amino acid sequence and protein structure, the primary activity of FABP7 appears to be binding of fatty acids. Although the binding affinity of FABP7 for various types of long-chain fatty acids has been studied
in vitro, and a likely
in vivo ligand for FABP7 has been proposed [
15], the actual fatty acids bound to FABP7 are yet to be determined. Because FABP7 does not possess an obvious nuclear localization signal, nuclear translocation of FABP7 might require other carrier proteins. Defining these interacting proteins and ligands should clarify the biological roles of FABP7.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
YL designed the study, performed the experiments, analyzed the data, and wrote the manuscript. AWB and KDA participated in evaluating the histopathology and immunohistochemistry of the specimens. NG participated in analyzing the data and critically editing the manuscript. All authors read and approved the final version of the manuscript.