Release of an inhibitor of angiogenesis upon induction of wild type p53 expression in glioblastoma cells.

The earliest genetic alteration in human astrocytoma progression is mutation of the p53 tumour suppressor gene, while one of the earliest phenotypic changes is the stimulation of neovascularization. Here, we tested the role of p53 in the angiogenic process by introducing a tetracycline-regulated wild type p53 gene into null glioblastoma cells. The parental cells expressed strong angiogenic activity while upon induction of wild type, but not mutant, p53 expression, the cells secreted a factor able to neutralize the angiogenicity of the factors produced by the parental cells as well as of basic fibroblast growth factor.

Aberrant CpG-island methylation has non-random and tumour-type-specific patterns.

CpG islands frequently contain gene promoters or exons and are usually unmethylated in normal cells. Methylation of CpG islands is associated with delayed replication, condensed chromatin and inhibition of transcription initiation. The investigation of aberrant CpG-island methylation in human cancer has primarily taken a candidate gene approach, and has focused on less than 15 of the estimated 45,000 CpG islands in the genome. Here we report a global analysis of the methylation status of 1,184 unselected CpG islands in each of 98 primary human tumours using restriction landmark genomic scanning (RLGS). We estimate that an average of 600 CpG islands (range of 0 to 4,500) of the 45,000 in the genome were aberrantly methylated in the tumours, including early stage tumours. We identified patterns of CpG-island methylation that were shared within each tumour type, together with patterns and targets that displayed distinct tumour-type specificity. The expression of many of these genes was reactivated by experimental demethylation in cultured tumour cells. Thus, the methylation of particular subsets of CpG islands may have consequences for specific tumour types.

p53 activates expression of HIC-1, a new candidate tumour suppressor gene on 17p13.3.

For several human tumour types, allelic loss data suggest that one or more tumour suppressor genes reside telomeric to the p53 gene at chromosome 17p13.1. In the present study we have used a new strategy, involving molecular analysis of a DNA site hypermethylated in tumour DNA, to identify a candidate gene in this region (17p13.3). Our approach has led to identification of HIC-1 (hypermethylated in cancer), a new zinc-finger transcription factor gene which is ubiquitously expressed in normal tissues, but underexpressed in different tumour cells where it is hypermethylated. Multiple characteristics of this gene, including the presence of a p53 binding site in the 5' flanking region, activation of the gene by expression of a wild-type p53 gene and suppression of G418 selectability of cultured brain, breast and colon cancer cells following insertion of the gene, make HIC-1 gene a strong candidate for a tumour suppressor gene in region 17p13.3.

Genetic origin of mutations predisposing to retinoblastoma.

Retinoblastoma is one of several human tumors to which predisposition can be inherited. Molecular genetic analysis of several nonheritable cases has led to the hypothesis that this tumor develops after the occurrence of specific mitotic events involving human chromosome 13. These events reveal initial predisposing recessive mutations. Evidence is presented that similar chromosomal events occur in tumors from heritable cases. The chromosome 13 found in the tumors was the one carrying the predisposing germline mutation and not the homolog containing the wild-type allele at the Rb-1 locus. These results suggest a new approach for identifying recessive mutant genes that lead to cancer and a conceptual basis for accurate prenatal predictions of cancer predisposition.

Molecular detection of chromosomal translocations that disrupt the putative retinoblastoma susceptibility locus.

A candidate DNA sequence with many of the properties predicted for the retinoblastoma susceptibility (RB1) locus has been cloned (S. H. Friend, R. Bernards, S. Rogelj, R. A. Weinberg, J. M. Rapaport, D. M. Albert, and T. P. Dryja, Nature [London] 323:643-645, 1986). The large size of this gene (ca. 200 kilobases [kb]) and its multiple dispersed exons (Wiggs et al., N. Engl. J. Med. 318:151-157, 1988) complicate molecular screening strategies important in prenatal and presymptomatic diagnosis and in carrier detection. Here we used field inversion gel electrophoresis (FIGE) to construct a restriction map of approximately 1,000 kb of DNA surrounding the RB1 locus and to detect the translocation breakpoints in three retinoblastoma patients. DNA probes from either the 5' or 3' end of the gene were used to detect a 250-kb EagI restriction fragment in DNA from unaffected individuals. Both probes identified an additional hybridizing fragment in the DNA from each patient, permitting the breakpoints in all three to be mapped within the cloned RB1 gene. Analysis of the breakpoint in one translocation cell line allowed the RB1 gene to be oriented with its 5' end toward the centromere. The 5' end of the gene also appeared to be associated with a clustering of sites for several infrequently cleaving restriction enzymes, indicating the presence of an HpaII tiny fragment island. The detection and mapping of the translocation breakpoints of all three retinoblastoma patients to within the putative RB1 gene substantiated the authenticity of this candidate sequence and demonstrated the utility of FIGE in detecting chromosomal rearrangements affecting this locus.

Loss of wild-type p53 bestows a growth advantage on primary cortical astrocytes and facilitates their in vitro transformation.

Primary cortical astrocytes were isolated from normal (+/+), heterozygous (+/-), or homozygous (-/-) p53-knockout mice. The normal astrocytes grew slowly and underwent crisis after limited division, while the homozygously defective cells grew rapidly and without contact inhibition. These -/- cells could not initially form colonies in soft agarose but acquired this capability after 10 passages in FCS or basic fibroblast growth factor but not epidermal growth factor. Almost all -/- astrocytes weakly expressed glial fibrillary acidic protein at passage 10 and were also A2B5+ when cultured in basic fibroblast growth factor. Most heterozygous cells resembled normal ones; however, some survived crisis, grew rapidly, and formed colonies. Outgrowing cells had all lost the wild-type p53 allele. These molecular and cellular events mimic the early stages of human brain tumors, suggest a role for p53 in the earliest stages of disease progression, and provide an experimental system to analyze the effects of other tumor-specific mutations in the disease process.

p53 gene mutation and mdm2 gene amplification are uncommon in medulloblastoma.

Loss of heterozygosity is common for the short arm of chromosome 17 in medulloblastomas, and putative medulloblastoma suppressor loci have been localized to 17p13. The colocalization of the p53 tumor suppressor gene to 17p13 raises the possibility that its mutant alleles may play a role in the malignant transformation of "medulloblasts." Mutations and deletions of the p53 gene have been described in many tumor types and in the germline of some individuals with the Li-Fraumeni syndrome, but reports on the status of the p53 and mdm2 (a gene coding for a p53-associated protein reportedly amplified in human sarcomas) genes in medulloblastomas are few and an indication of their roles, if any, in the etiology of this important childhood tumor has yet to emerge. Here we have analyzed polymerase chain reaction-amplified products of exons 4-9 (95% of reported p53 mutations occur within this region) of the p53 gene in 9 medulloblastomas for potential mutations using the technique of single strand conformation polymorphism analysis and DNA sequencing. We found only one mutation, an A-T to T-A transversion involving the second base of codon 285 and resulting in the substitution of valine for glutamic acid, amplification of the mdm2 gene could be detected in zero of eight of these tumors. These findings suggest that genetic events associated with the inactivation of p53 gene occur in only a minor subset of medulloblastomas.

D-type cyclins complex with the androgen receptor and inhibit its transcriptional transactivation ability.

D-type cyclins regulate distinct cellular processes, such as mitotic cell cycle control, differentiation, and transcription. We have previously shown that the D-type cyclins are critical for the androgen-dependent proliferation of prostate cells. Here, we sought to determine whether cyclin D1 directly influences the transactivation potential of the androgen receptor, a transcription factor that strongly influences androgen-dependent proliferation. We found that ligand-mediated transcriptional activation of a physiological target, prostate-specific antigen, by the androgen receptor was inhibited by cyclins D1 and D3. The ability of D-type cyclins to inhibit androgen receptor transactivation was not shared with other cyclins, and cyclin D1 was as effective as dominant negative mutants of the androgen receptor in inhibiting transactivation. This function of cyclin D1 was independent of its role in cell cycle progression and is likely elicited through its ability to form a specific complex with the androgen receptor. These data underscore the various mechanisms through which the androgen receptor is regulated and also point to a negative feedback role for cyclin D1 in controlling androgen-dependent growth.

The phosphoinositol phosphatase activity of PTEN mediates a serum-sensitive G1 growth arrest in glioma cells.

The PTEN gene (also called MMAC1 and TEP1) at chromosome 10q23 is mutated in a variety of predominantly late-stage tumors and has been shown to suppress glioma cell growth in vitro and in vivo. Here we sought to determine the mechanism by which PTEN mediates growth inhibition. Using the mutant PTEN glioma cell line, U87MG, as a transfection recipient for a series of PTEN alleles, we provide direct evidence that this capacity requires phosphatase activity. Mutations mapping upstream, within, and downstream of the catalytic domain ablated activity toward a 3' phosphorylated phosphoinositide substrate of PTEN, whereas alleles with mutations flanking the catalytic domain retained activity toward the acidic protein polymer substrate, Glu4Tyr1. Thus, catalytic activity toward phosphoinositide substrates was required for growth suppression, whereas activity toward the protein substrate was dispensable for growth suppression. Finally, we used apoptotic and cell proliferation analyses to show that PTEN-mediated growth inhibition under reduced serum conditions was due to a G1 cell cycle block rather than to an induction of apoptosis.

Silencing of p16/CDKN2 expression in human gliomas by methylation and chromatin condensation.

The product of the p16/CDKN2 locus, p16ink4, negatively regulates the cell cycle through binding and inactivation of cyclin-dependent kinases (CDKs) 4 and 6. This locus is frequently targeted for deletion in cell lines and primary tumor tissues. In gliomas, although up to 50% do not have detectable expression of p16/CDKN2 protein or mRNA, often the gene is wild type in sequence. Here, we tested the hypothesis that transcriptional repression of p16/CDKN2 in gliomas may be mediated by aberrant methylation of the CpG island, which is in the 5' region of the locus. Partial rather than complete p16/CDKN2 methylation was detected in 24% (10 of 42) of the gliomas, regardless of tumor grade, but was not observed in normal brain (0 of 10). We tested whether this partial methylation could inhibit expression in a human tumor cell line in which suppressed p16/CDKN2 expression was associated with both methylation and tightly compacted chromatin around the p16/CDKN2 promoter. Exposure of these cells to 5-aza-2-deoxycytidine resulted in a dramatic increase in promoter accessibility and induction of p16/CDKN2 expression, indicating that chromatin structure, CpG island methylation, and p16/CDKN2 expression are intimately associated. Taken together, these data suggest that methylation occurs in only a subset of cells within gliomas and that the methylation-associated inactivation of p16/CDKN2 expression observed in many common human cancers may mechanistically result from structural changes in the chromatin containing the p16/CDKN2 locus.

Muscle-specific gene expression in rhabdomyosarcomas and stages of human fetal skeletal muscle development.

Rhabdomyosarcomas (RMS) bear a morphological resemblance to developing striated muscle. It has been reported that two histologically distinct subtypes of RMS, embryonal and alveolar, behave differently in many clinical aspects, such as age distribution, primary site, and prognosis. We have investigated the expression of various genes, which are preferentially expressed in normal muscle tissue or cell culture (actins, myosins, and creatine kinases, and myogenic regulatory genes MyoD, myogenin, MRF4, and Myf5), in embryonal and alveolar subtypes and compared the results to the stages of developing human fetal limb muscle. The data showed that each of the RMS tumors tested, regardless of histological features, expressed MyoD1 and MRF4 transcripts. Expression of the myogenin gene was detectable in all alveolar RMS (n = 8), whereas only 5 of 8 embryonal RMS expressed myogenin transcripts. Trace levels of Myf5 transcripts were visible in all alveolar RMS and 7 of 8 embryonal RMS. The alpha-skeletal, alpha-cardiac, and beta- and gamma-cytoplasmic actin transcripts were detectable in all alveolar RMS. While the beta- and gamma-cytoplasmic actin transcripts were evident in all embryonal RMS, only 3 of 8 and 6 of 8 embryonal RMS expressed detectable levels of alpha-skeletal and alpha-cardiac actin transcripts, respectively. The embryonic form of myosin heavy chain was detectable in 1 of 8 of each type of tumor. Myosin light chain-1/3 transcripts were detectable in 4 of 8 alveolar RMS and 5 of 8 embryonal RMS. Brain creatine kinase transcripts were detectable in all alveolar RMS and 4 of 8 embryonal RMS, whereas none of the RMS samples contained detectable levels of the muscle form of creatine kinase. A comparison of the expression profiles with those of normal developing human fetal limb muscle (from 7.5 to 24 weeks' gestation) suggested that RMS resembled a relatively restricted segment of fetal muscle development. Furthermore, the data also showed a great deal of overlap in the differentiation state achieved by the embryonal and alveolar subtypes of RMS, suggesting that the clinicopathological difference between these two may not be due to malignant transformation of the cells from different positions in the normal pathway of myogenesis.

Induction of vascular endothelial growth factor expression in endothelial cells by platelet-derived growth factor through the activation of phosphatidylinositol 3-kinase.

Increased numbers of platelet-derived growth factor beta receptors betaPPDGFRs) on neovascular endothelial cells is a common occurrence in several pathological conditions including wound healing, inflammation, and glioma tumorigenesis. Here we sought to test the biological significance of this by determining whether expression of wild-type betaPDGFR by normal aortic endothelial cells affected the expression of the vascular endothelial growth factor (VEGF), a critical angiogenesis regulator and mitogen for such cells. The results showed that PDGF could increase transcription and secretion of VEGF by betaPDGFR-expressing endothelial cells. Moreover, we further demonstrated a requirement for the activation of phosphatidylinositol 3-kinase (PI3K) in this response by using chemical inhibitors of PI3K, mutant PDGFR, and dominant-negative PI3K. These studies suggest a novel mechanism by which PDGF induces VEGF expression in endothelial cells, define VEGF as a downstream target for PI3K, and invoke a role for PI3K in angiogenesis.

Loss of heterozygosity for loci on chromosome 10 is associated with morphologically malignant meningioma progression.

Meningioma is a common tumor of the central nervous system which displays morphological heterogeneity. In order to determine whether this phenotypic variability is associated with distinct or overlapping genetic lesions, we compared genotypes at several loci defined by allele length polymorphism in tumor and normal tissues from patients with meningioma. In particular, we concentrated on loci on chromosomes 22 and 10 because these genomic regions have previously been shown to be altered in the former in sporadic and familial meningiomas and in the latter as a late stage event in progression of another common brain tumor, astrocytoma. We examined 38 tumors which were classified as benign, atypical, or malignant by morphological criteria, invasive characteristics, or both. We found that loss of heterozygosity (LOH) for loci on chromosome 22 occurred in 5 of 15 benign, 2 of 2 atypical, and 5 of 10 malignant meningiomas. Similar alterations of chromosome 10 were found in 0 of 20 benign, 1 of 2 atypical, and 4 of 13 malignant meningiomas. Among the malignant tumors, LOH for loci on chromosome 10 occurred in 2 of 4 morphologically malignant tumors and in 2 of 4 morphologically and invasively malignant tumors. In contrast, LOH was not observed for any of the 5 informative tumors classified as malignant by invasive characteristics only. LOH for loci on chromosome 22 accompanied (but was not restricted to) allelic loss of loci on chromosome 10. These data suggest that the progression of meningiomas from arachnoidal cells to the morphologically malignant phenotype may, in part, entail the loss of a tumor suppressor gene(s) on chromosome 22 early in the process and that this may be compounded by alterations of chromosome 10, the LOH of which is associated with morphological signs of malignancy.

Replacement of the p16/CDKN2 gene suppresses human glioma cell growth.

The p16/CDKN2 gene has many features of a growth suppressor gene: it maps to 9p21, a frequent region of loss of heterozygozity in a variety of tumor types; it encodes an inhibitor of cyclin-dependent kinase 4; and its homozygous deletion is common in tumor-derived cell lines. However, the lower frequency of alteration of the gene in primary tumor tissue as compared to the cognate tumor cell lines has brought this interpretation into question. We have assessed the growth suppressive function of p16/CDKN2 by gene transfer. The introduction of full-length p16/CDKN2 cDNA caused marked growth suppression in p16/CDKN2-null human glioma cells, but was without significant effect in those cells with endogenous wild-type p16/CDKN2 alleles. These results provide functional evidence in support of the hypothesis that the p16/CDKN2 gene is a functional growth suppressor gene, at least in gliomas.

Loss of heterozygosity for 10q22-10qter in malignant melanoma progression.

Karyotypic and molecular data indicate that genetic events involving the chromosome region 10q22-10qter may be related to tumorigenesis in malignant melanoma. To test this we analyzed 10 polymorphic microsatellite repeats in the region 10q22-qter, using a polymerase chain reaction-based assay for loss of heterozygosity and DNA isolated from normal and tumor tissue from 26 individuals with malignant melanoma. The samples included 19 paired normal and malignant tissues representing various stages of melanoma as well as 7 cases in which samples from at least 2 different points in time during tumor progression were available. Our findings indicate that loss of heterozygosity of 10q22-10qter is a frequent event, that the observed loss of heterozygosity does not result from whole chromosome loss, and that it is associated with tumor progression. Finally, the appearance of new alleles in two of the tumors may indicate the involvement of DNA replication errors in melanoma analogous to such events in other tumor types.

Chromosome 9 deletion mapping reveals interferon alpha and interferon beta-1 gene deletions in human glial tumors.

We have applied restriction fragment length polymorphism analysis to a 30-member panel of primary glioma DNAs, which had been previously examined for loss of genetic information (C. D. James, E. Carlbom, J. P. Dumanski, M. Hansen, M. Nordenskjold, V. P. Collins, and W. K. Cavenee, Cancer Res., 48:5546-5551, 1988), to determine the frequency and sublocalization of loss of genetic information from chromosome 9. We have also utilized scanning densitometry for dosage determination of the 9p-localized interferon alpha and interferon beta-1 genes among these same tumors. Our results reveal the following: (a) for those cases in which loss has occurred, the region of common loss lies on the short (p) arm of the chromosome; (b) loss of genetic information from the short arm of chromosome 9 occurs frequently in glial tumors of intermediate (anaplastic, grade III) and high (glioblastoma, grade IV) histological malignancy (10 of 20 cases) but not in tumors of low (grade II) histological malignancy (0 of 10 cases); (c) tumors with 9p deletions are hemi- or nullizygous for interferon beta-1 and the interferon alpha gene cluster; (d) cases of interferon nullizygosity occur exclusively among tumors of highest histological malignancy (glioblastoma). These data, especially the determination of a region of nullizygosity, suggest proximity to or residence within a gene(s) whose function(s) is (are) critical to the suppression of the malignant evolution of glial tumors.

Genes for epidermal growth factor receptor, transforming growth factor alpha, and epidermal growth factor and their expression in human gliomas in vivo.

Anomalies of the epidermal growth factor receptor (EGFR) gene, including amplification, rearrangement, and overexpression, have been reported in malignant human gliomas in vivo. In vitro glioma cell lines coexpress EGFR and at least one of its ligands, transforming growth factor alpha, suggesting the existence of an autocrine growth stimulatory loop. We have studied the tumor tissue from 62 human glioma patients and examined the structure and quantity of the EGFR gene and its transcripts, as well as the quantity of the receptor protein. In addition we have examined the genes and transcripts coding for the pre-pro forms of epidermal growth factor and transforming growth factor alpha, the two endogenous EGFR ligands. EGFR gene amplification was detected in 16 of the 32 malignancy grade IV gliomas (glioblastoma) studied (50%), but only in 1 of 30 gliomas of lesser malignancy grade (I-III). All tumors with an amplified gene overexpressed EGFR mRNA. More than one-half (62.5%) of the glioblastomas with amplified EGFR genes also showed coamplification of rearranged EGFR genes and concomitant expression of aberrant mRNA species. Overexpression, without gene amplification, was observed in some of the low grade gliomas, and aberrant EGFR transcripts were also seen in some cases without gene amplification or detected gene rearrangements. mRNA expression for one or both of the pre-pro forms of the ligands was detected in every tumor studied. Thus, several mechanisms for the activation of the EGFR-mediated growth stimulating pathway are possible in human gliomas in vivo: expression of a structurally altered receptor that may have escaped normal control mechanisms; and/or auto-, juxta-, or paracrine stimulating mechanisms involving coexpression of receptor and ligands, with or without overexpression of the receptor.

DNA hypermethylation is associated with 17p allelic loss in neural tumors.

It has long been debated whether the accumulation of allelic losses in tumors involves the selection of cells which have stochastically lost chromosomal regions or whether there is, inherent to the neoplastic state, a process which predisposes to genetic instability. Changes in DNA methylation are commonly seen in human tumors and can alter chromosome structure. We now have examined specific types of primary neural tumors which allow us to determine relationships between abnormal DNA hypermethylation and allelic loss. In primary brain tumors which frequently lose chromosome 17p (30-50%) even in the earliest stages, we now show that 84% (21 of 25) exhibit hypermethylation at locus D17S5, on 17p. However, in primary neuroblastomas, a tumor type which does not lose chromosome 17p, no regional hypermethylation is observed. These data suggest that on chromosome 17p, regional D17S5 hypermethylation constitutes a molecular change which is associated with genetic instability.

A defined region of loss of heterozygosity at 11q23 in cutaneous malignant melanoma.

Karyotypic and molecular data indicate that genetic alterations of the long arm of chromosome 11 (11q) may be involved in malignant melanoma. To test this we analyzed 5 polymorphic microsatellite repeats on 11q using a PCR-based assay for loss of heterozygosity in normal and tumor tissues from 24 individuals with cutaneous malignant melanoma of various stages. Our findings indicate that a tumor suppressor gene that plays a role in malignant melanoma is located on the long arm of chromosome 11, likely within a 51 cM region at 11q23. Its loss appears to be a late event in tumor progression and may serve as an indicator for a less favorable clinical outcome.

A common mutant epidermal growth factor receptor confers enhanced tumorigenicity on human glioblastoma cells by increasing proliferation and reducing apoptosis.

Alterations of the EGFR gene occur frequently in human gliomas where the most common is an in-frame deletion of exons 2-7 from the extracellular domain, resulting in a truncated mutant receptor (deltaEGFR or de 2-7 EGFR). We previously demonstrated that introduction of deltaEGFR into human U87MG glioblastoma cells (U87MG.deltaEGFR) conferred remarkably enhanced tumorigenicity in vivo. Here, we show by cell-mixing experiments that the enhanced tumorigenicity conferred by deltaEGFR is attributable to a growth advantage intrinsic to cells expressing the mutant receptor. We analyzed the labeling index of the proliferation markers Ki-67 and bromodeoxyuridine and found that tumors derived from U87MG.deltaEGFR cells had significantly higher labeling indexes than those of tumors derived from U87MG cells that were either naive, expressed kinase-deficient mutants of deltaEGFR, or overexpressed exogenous wild-type EGFR. We also utilized terminal deoxynucleotidyl transferase-mediated nick end-labeling assays and showed that the apoptotic index of U87MG.deltaEGFR tumors was more than 4-fold lower than that of parental U87MG tumors. This decrease in cell death was inversely correlated with the expression level of Bcl-X(L), a negative regulator of apoptosis, which was more than 3-fold higher in U87MG.deltaEGFR-derived tumors than in those derived from parental cells. Similar observations were obtained in vitro in serum-free conditions. These results suggest that deltaEGFR exerts its pronounced enhancement of glioblastoma tumorigenicity by stimulating proliferation and inhibiting apoptosis and that the effects are directly attributable to its constitutively active signal.