Abstract
We carried out a gene expression profiling study using cDNA array technology with 24 primary glioma tissues of low-grade (oligodendroglioma), intermediate-grade (anaplastic oligodendroglioma and anaplastic astrocytoma), and high-grade (glioblastomas multiforme) tumors and found that insulin-like growth factor binding protein 2 (IGFBP2) was consistently overexpressed only in glioblastoma multiforme. The cDNA array results were confirmed by Northern and Western blotting. The fact that the IGFBP2 gene, which is normally expressed in fetal cells and turned off in adult cells, becomes reactivated in the most advanced stage of glioma suggests that glioma progression is a result of dedifferentiation or results from a block of differentiation. Identification of IGFBP2 as a gene associated with glioma progression demonstrates the power and utility of high-throughput gene expression profiling in cancer gene discovery.
Introduction
Cancer development is a highly complex process involving multiple genetic and epigenetic changes. Conventional approaches investigating one or several candidate genes at a time result in necessarily limited information. Recently, rapid progress in the human genome project and development of high throughput parallel gene expression profiling technologies have permitted the study of numerous genes simultaneously in a single experiment (1, 2, 3, 4, 5) . Furthermore, this approach permits investigation of multiple genes without prior knowledge of their functions, expressions, or, in the case of expressed sequence tags, identities, thus opening a new era of molecular diagnosis and therapeutics. In this study, we systematically investigated the expression of 588 known cellular genes in primary tissue samples from patients with O, 5 AO, AA, and GBM. We found that IGFBP2 is expressed at high levels in GBM tissues but not in the other patients’ samples, suggesting that IGFBP2 expression may be associated with the formation of or progression to glioblastoma.
The IGFs stimulate cellular proliferation in an autocrine fashion in many tumors. It has been reported that human gliomas produce IGFs and express elevated levels of IGF receptors compared with normal brain tissue (6, 7, 8) . IGF also enhances three-dimensional growth of glioblastoma spheroids in vitro (9) . The effect of IGFs on cells is regulated by a family of IGFBPs (10, 11, 12) , which can both attenuate and stimulate the mitogenic effect of IGFs (12, 13, 14, 15, 16) . Studies of the temporal and spatial expression of IGFBPs during development reveal that members of the IGFBP family are expressed in a tissue-specific and developmental stage-specific manner (17 , 18) . Unlike some of the other members, IGFBP2 is predominantly expressed in fetal tissues that are highly proliferative, such as the early postimplantation epiblast, the apical ectodermal ridge, and the progenitors of spleen and liver cells. In the nervous system, IGFBP2 is expressed in fetal astroglial cells. After birth, IGFBP2 expression significantly decreases in glial cells (19 , 20) .
An association of IGFBP2 with several different malignancies has been noted. Studies have shown that patients with prostate carcinoma have elevated serum IGFBP2 compared with patients with benign prostatic hyperplasia (21, 22, 23) . Increased expression of IGFBP2 was also found in malignant ovarian cyst fluid (24 , 25) . Recently, it was shown that introduction of IGFBP2 expression increased tumorigenicity of recipient cells in a nude mouse xenograph model (26) .
Materials and Methods
Primary Glioma Tissues.
All samples were acquired from the Brain Tumor Center tissue bank of the University of Texas M. D. Anderson Cancer Center. Tissue bank specimens were quick-frozen shortly after surgical removal and stored at −80°C. Although it is not known whether and to what the extent the time delay between tumor removal and tumor freezing affects gene expression, all of the tumor tissue samples experienced a similar length of delay. Thus, the tumor harvesting procedure would affect all samples in a similar manner and would not be expected to contribute to the difference in gene expression patterns among samples. H&E-stained frozen tissue sections are routinely prepared from all tissue bank specimens for screening purposes. All tissue specimens for cDNA array analysis were screened by a neuropathologist (G. N. F.), and diagnoses were confirmed by a second neuropathologist (J. M. B.). Specimens were specifically selected for dense tumor cellularity, and the amount of normal brain tissue present was <10% in all cases. Grading of astrocytic neoplasms and oligodendrogliomas was performed according to the St. Anne/Mayo criteria (27 , 28) . Mixed oligoastrocytic tumors and AOs with necrosis (“oligodendroglial glioblastomas’) were excluded from this study. Two tumor-free brain tissue samples taken surgically were also selected for analysis. (We should point out that because these two tissues were taken from glioma patients, some invasive tumor cells may be present, although not detected by histological screening.)
Isolation of Total RNA and mRNA from Tissues.
The tissues were ground to powder under frozen conditions and lysed in the lysis buffer TRI Reagent (Molecular Research Center, Cincinnati, OH). After isolation of total RNA, an aliquot was run on a denaturing formaldehyde agarose gel to check quality. Good RNA is indicated by the lack of a smear on the lower part of the gel (a smear indicates RNA degradation) and by the presence of 28S ribosomal RNA twice as intense as that of 18S rRNA. After good-quality total RNA was obtained, the mRNA was isolated using the poly(dT) mRNA isolation column provided by Qiagen, Inc.(Chatsworth, CA). Using this column, 1–2 μg of mRNA can be isolated from 200 μg of total RNA. For cDNA array studies, 0.5–1.0 μg mRNA is sufficient. We obtained high-quality RNA from >90% of biopsy tissues.
Hybridization to the Human Atlas cDNA Expression Array I Blots.
The cDNA fragments representing 588 human genes with known functions and known tight transcriptional controls were immobilized in duplicate onto a nylon membrane (Clontech Laboratories, Inc.). Each cDNA fragment is 200–500 bp long. To minimize cross-hybridization and nonspecific binding of cDNA probes, each fragment was selected as a unique sequence without a poly(A) tail, repetitive elements, or highly homologous sequences. 32P-labeled cDNA probes were generated by reverse transcription of 0.5–1.0 μg of each analyzed poly(A)+ RNA sample in the presence of [α-32P]dATP. Each cDNA probe was then hybridized to an array. After a high-stringency wash, the hybridization pattern was analyzed by autoradiography and quantified by phosphorimaging using ImageQuant software with the Storm 840 Phos-phorImager. Because the amount of each cDNA fragment on the membrane is in excess (10 ng), binding of cDNAs to the probes is linear.
To normalize the relative gene expression, we selected the GAPDH gene as an internal reference. Our preliminary studies showed that GAPDH is a reliable reference housekeeping gene, the expression level of which was very stable among different samples. A ranking analysis showed that the amount of GAPDH is among the top 2% of all 588 transcripts in all samples. GAPDH was preferred also because it has been conventionally used as an internal reference for measuring gene expression levels in Northern blotting, RNA protection, and quantitative PCR assays.
Our previous studies using cell lines demonstrate that the technology is reproducible when membranes from the same lot of production are used. During the course of this study, membranes from the same lot were used.
Northern and Western Blotting.
Northern and Western blotting were performed as described previously (29) . The IGFBP2 cDNA probe (446 bp) was obtained by PCR using primer pairs AGCCCCTCAAGTCGGGTA and TGCGGTCTACTGCATCCG. The monoclonal antibody for IGFBP2 was obtained from Upstate Biochemicals (Lake Placid, NY).
Statistics.
We compared the gene expression distributions between histology groups using a Kruskal-Wallis test.
Results
High-quality total RNAs were isolated from frozen tissue of 13 GBMs, 11 AAs, 10 AOs, and 8 Os. Among these RNA samples, 24 had sufficient amounts (>150 μg) for isolating mRNA (>0.5 μg) for cDNA array assay. The mRNAs were reverse-transcribed to 32P-labeled cDNAs and hybridized to Atlas cDNA array blots to profile the expression of 588 cellular genes. The hybridized blots were imaged and quantified by ImageQuant software. Our previous profiling studies using cell lines showed that mRNAs isolated from one cell line at two different times resulted in very similar gene expression profiles, whereas mRNAs from three different glioma cell lines produced very different profiling patterns (30) . Anticipating heterogeneity among tumors, we were initially surprised to observe that the overall expression patterns among different patients were rather similar. Typical examples from each of the subcategories and from two normal brain tissues are shown in Fig. 1 ⇓ . Nevertheless, differences in gene expression were detected among tissues. One that showed consistent disease stage-dependent differences is the IGFBP2 gene, which was overexpressed in GBMs but not in other gliomas tested (Fig. 2 ⇓ and 3 ⇓ ). Although IGFBP2 expression was found previously to be elevated in prostate carcinomas and ovarian cancers (21 , 22 , 24 , 25) , significant elevation of IGFBP2 in GBM has not been reported previously. Previous studies have suggested that the VEGF gene is overexpressed in GBM (31) , and our analyses demonstrated a consistent increase, validating the cDNA array approach (Fig. 3) ⇓ .
cDNA array images of normal brain tissues and tissues from GBM, AA, AO, and O. mRNAs were isolated from these tissues, reverse-transcribed into 32P-labeled cDNAs, and hybridized to Atlas cDNA Array I membranes. On the membranes, 588 known cellular genes of different functional groups (printed in six squares) were included. A complete list of gene names and their locations can be found in the instruction manual and webpage from Clontech. After stringent washes, the membranes were exposed to PhosphorImager screen and analyzed by ImageQuant program.
The IGFBP2 gene is elevated only in GBMs, as detected by cDNA array gene expression profiling. The first gene groups of all of the tissues shown in Fig. 1 ⇓ , where IGFBP2 is located, are assembled to highlight the IGFBP2 gene as indicated by arrowheads for GBM. Weak signals can be seen in other tissues.
Correlation of IGFBP2 and VEGF expression with histology in the four types of tissues. The expression levels of IGFBP2 and VEGF are quantified and normalized against that of GAPDH by ImageQuant program. The relative expression levels were correlated with histology. A, IGFBP2; B, VEGF.
To further confirm our finding regarding IGFBP2 in a larger sample size and also by a conventional approach, we used Northern blotting assay. Some of the tissues that did not provide enough mRNA for cDNA array assay had enough RNA for Northern blotting. Examples of Northern blotting are shown in Fig. 4A ⇓ , which confirmed the results from cDNA array, indicating that IGFBP2 was elevated only in GBM. Several tissue samples from each category were also subjected to protein isolation and Western blotting using an antibody specific for IGFBP2 protein. The results in Fig. 4B ⇓ show that the IGFBP2 protein was also elevated in GBM tissues. Combining the results from cDNA array and Northern blotting, we accumulated comparable numbers of cases for each category. Statistical analyses showed that IGFBP2 expression was significantly elevated in GBMs (Fig. 5) ⇓ by either cDNA array or Northern assay. Elevation of IGFBP2 in GBM but not in lower-grade tumors implies that the IGF pathway is involved in glioma progression. Because insulin-like growth factors (I and II) or IGFBP1 and IGFBP3, which are represented in the cDNA arrays, were expressed at similar levels in all grades of gliomas (not shown), the molecular change in GBM may be specific to IGFBP2.
Elevation of IGFBP2 mRNA and protein expression in GBM as detected by Northern and Western blotting. RNA and protein were isolated from tissues as described in “Materials and Methods.” The levels of IGFBP2 mRNA were detected by Northern blotting using a IGFBP2 cDNA probe (A). IGFBP2 protein in the tissues was analyzed by Western blotting using an antibody specific for IGFBP2 protein (B). The membrane was probed with anti-actin antibody for protein loading control.
Association of IGFBP2 expression with GBM. The relative IGFBP2 expression levels after normalizing to that of GAPDH from either cDNA array (•) or Northern blotting (○) were correlated with histology type. Overexpression of IGFBP2 is significantly correlated with GBM (P < 0.001) using the Kruskal-Wallis test.
Discussion
Extensive studies of the genetic changes that accompany cancer development and progression have made it increasingly clear that cancer development is a highly complex process involving multiple factors or genes. Multiple genetic changes linked to glioma malignancy have been identified. The tumor suppressor gene p53 is mutated frequently in all grades of gliomas (32) . Several growth factors and growth factor receptors are frequently overexpressed in gliomas (33, 34, 35) . Among these, VEGF, an important angiogenesis factor, is highly expressed in glioblastomas (31) . On the basis of this and other evidence, it is believed that genetic and molecular alterations fundamentally change the signal transduction pathways through which cells interact with their surroundings and respond to environmental cues.
These genetic alterations do not necessarily happen simultaneously in each cancer patient, and different genetic changes may have different impacts on the cells. Some of the genetic changes may have a direct correlation with specific response to therapy. The most effective therapy may eventually be based on individualized blueprints or holistic molecular diagnoses, rather than patterns of expression or deletion of only one or a few genes.
One of the goals of the human genome project is to reveal the genetic lexicon of human genes, which will allow a complete genotyping of individual patient tumor samples. In this study, we have tested the utility of gene expression profiling by using a parallel gene expression array that includes 588 genes. This prototype study has yielded several key findings from which additional inferences can be drawn:
(a) Although heterogeneity among tumor samples was observed, consistent and recurring patterns, such as overexpression of IGFBP2 in GBM, could be easily identified. The validity of this approach was demonstrated by analyzing genes whose expression patterns in gliomas have been studied previously. For example, the VEGF gene is known to be overexpressed in GBM (31) , and the results from our cDNA array analysis confirmed this conclusion. An additional example is provided by p16 gene expression, which was found to be absent in most of the GBMs by our cDNA array analyses (data not shown). This finding also is consistent with literature reports (36) .
(b) This approach may provide important data and potential insights with regard to tumor etiology. The discovery of IGFBP2 overexpression confined to GBM suggests that the IGF pathway may contribute to glioma progression. Furthermore, the observation that IGFBP2 is normally expressed in less-differentiated fetal astroglial cells and that IGFBP2 expression is absent in mature cells suggests that GBM development may represent a defect in astrocyte differentiation.
(c) The segregation of specific gene expression patterns with a specific disease stage points to the potential utility of gene expression profiling in identification of tumor markers for molecular diagnosis. This is an important issue because classification by conventional histopathological criteria can be difficult and subjective for many tumors with mixed morphological phenotypes. Gene expression profiling may provide a molecular approach to assist diagnosis. This, in turn, may facilitate the individualization of patient management because patients with different histological tumor types respond differently to specific therapies. For example, Os with chromosomes 1p and 19q abnormalities respond better to PCV (Procarbazine, cytoxan, and vincristine) chemotherapy compared with Os with other types of chromosomal changes, although the responsible genes involved on the two chromosomes are presently unknown (37) .
Acknowledgments
We thank Drs. Eric Holland, Gordon Mills, Robert Bast, and Stanley Hamilton for critical review of the manuscript and Qiang Zhang for participation in data quantitation.
Footnotes
-
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 This work was partially supported by NIH Grants CA67987 and CA55164 and a grant from University of Texas M. D. Anderson Brain Tumor Center. C. H. R. was a visiting fellow partially supported by the Korean Cancer Center Hospital.
-
↵2 These two authors contributed equally to this study.
-
↵3 Present address: Department of Neurosurgery, Korea Cancer Center Hospital, 215-4, Gongneung-Dong, Nowon-Ku, Seoul 139-240, Korea.
-
↵4 To whom requests for reprints should be addressed, at Department of Pathology, The University of Texas M. D. Anderson Cancer Center, Box 85, 1515 Holcombe Boulevard, Houston, TX 77030. Phone: (713) 792-3778; Fax: (713) 745-1183; E-mail: 12507vl{at}mdanderson.org
-
↵5 The abbreviations used are: O, oligodendroglioma; AO, anaplastic oligodendroglioma; AA, anaplastic astrocytoma; GBM, glioblastoma multiforme; IGFBP2, insulin-like growth factor binding protein 2; IGF, insulin-like growth factor; VEGF, vascular endothelial growth factor; GAPDH, glyceraldehyde-3-phosphate dehydrogenase.
- Received May 24, 1999.
- Accepted July 14, 1999.
- ©1999 American Association for Cancer Research.
Volume 59, Issue 17
