Polymerase Chain Reaction

PCR is one technique that has revolutionized the applicability of gene-based technologies to clinical medicine. It permits the identification and amplification of DNA sequences from nearly any source, as long as part of the sequence is known. Therefore, any known DNA sequence can be detected in any sample containing genetic material, including tumors.

Initially, two oligonucleotide primers (short pieces of synthetic DNA with sequences designed to stick to regions flanking the DNA sequences to be amplified) are designed and synthesized chemically. DNA containing the target sequence to be amplified is mixed together with oligonucleotide primers (often referred to as oligos), free nucleotides, and heat-stable DNA polymerases. Each cycle of PCR requires heating the mixture to melt double-stranded DNA into two single strands, cooling to anneal or “stick” primers to the template DNA to be amplified, synthesis of a new complementary strand from free nucleotides, and finally, melting and annealing to separate new from old strands of DNA. Each cycle results in doubling of the initial template quantity. After 10 cycles of PCR, the original template DNA is amplified approximately 100 times, and after 20 cycles, approximately 1 million times. Once a PCR product is separated on a gel, DNA fragments of expected size are collected, and the nucleotide sequence is checked to verify amplification of the correct target.

PCR can be used to detect or quantify particular DNA sequences in a sample, for example, to detect the presence of viral sequences or to determine whether a tumor sample contains known sequences or a known genetic alteration. PCR products can also be used to generate purified DNA of known sequences for use in the construction of novel DNA molecules. Perhaps most important, PCR is now used routinely to detect new mutations in amplified or deleted regions of tumor DNA and new mutations in known or putative tumor suppressor genes and oncogenes.

DNA Construction and Gene Expression

Once DNA products are isolated from PCR experiments, their function can be studied. One simple way to study the function of a gene is to insert its encoding DNA into a cell. The transcriptional machinery of the cell can then transcribe the DNA to RNA, which is then translated to protein. If the product is an oncogene, for example, introduction of its DNA into cells may induce unrestrained proliferation of these cells.

Transfection

Polynucleotides may be introduced into cells in a number of ways. The most primitive and least effective way involves simply incubating “naked” DNA (or other types of nucleotides) in cell culture, where they are eventually taken up by cells and transcribed. Other techniques require DNA construction and insertion of the DNA into a cloning vector through the use of restriction enzymes and ligation techniques. Plasmids gain easy access into most types of cells, and the DNA carried within a plasmid can be replicated and transcribed, its RNA translated, and eventually the protein product of the gene expressed within a given cell. The process of incorporating DNA into cells growing in tissue culture is called transfection. By using transfection techniques, the function of a gene can be gleaned by observing the phenotypic changes that occur in a cell after transfection with a certain type of nucleotide construct.

Viral transfection or, more appropriately, viral infection of cells is particularly useful with animal cells and is a very efficient means of delivering genes to the cells in question. Viral vectors contain powerful promoters that drive expression of the genes incorporated into them. The three most commonly used viral vectors are retroviruses, herpesviruses, and adenoviruses. Retroviruses integrate into the host genome and become permanent in all progeny. The other vectors do not.

Gene Expression in Mice

To verify that a gene defect is the cause of a disease, it is necessary to evaluate the in vivo biologic consequences of genetic alterations. This can be accomplished by using transgenic or knockout mice. Such genetic manipulation of disease-associated genes results in genetically engineered models (GEMs) of disease.⁴ Genetic manipulation in mice is now fairly sophisticated, and gene expression can be induced or halted in animals with remarkable temporal and cell-specific specificity.

Transgenic Mice

Transgenic mice are created by introducing a desired DNA sequence, referred to as a transgene, into fertilized mouse eggs, which are subsequently placed in the oviducts of pseudopregnant female mice. Progeny conceived through this process are derived from the altered oocyte and contain the transgene in all cells in their body. Transgenes contain promoter and regulatory sequences that direct gene expression in one or more tissue types. Transgenes may also include inducible promoter regions so that expression of the transgene can be temporally and spatially controlled by administration of an inducing agent. In most cases, the foreign DNA does not disrupt cellular function, and expression of the transgene occurs only in tissues that use the transgene promoter sequence. Progeny are screened by analyzing the mouse tissue for expression of the desired transgene.

With this technology, human genes can be transferred to animals and their ability to cause disease ascertained. By linking the gene of interest with a tissue-specific promoter, expression can be restricted to a particular organ or cell type. Transgenic mice can also be used to measure the activity of signal transduction pathways or other molecular functions of a cell. For example, p53 activity can be measured by introducing the consensus sequence for p53 binding upstream of the firefly luciferase gene, thereby inducing transcription of the luciferase gene upon p53 binding. The luciferase gene converts a substrate called luciferin into light, and the photons emitted from the reaction can be measured directly by a camera. Emission of light from living organisms is known as bioluminescence. In the p53/luciferase scenario, mice that produce luciferase after p53 binding will convert luciferin into a bioluminescence signal, and this signal serves as an indirect measurement of the DNA binding activity of p53.⁵ The potential applications of transgene technology in mice are virtually unlimited, and glioma mouse models are beginning to use transgenic animals to measure the proliferation rates of tumors, measure the status of tumor signal transduction pathways, and study a host of other events that occur during the genesis of glioma.

Knockout Mice

Creation of knockout mice, in which genes are mutated or disrupted, results in a loss of function. The technology is similar to but more elaborate than transgenic protocols because the gene to be ablated, along with its flanking regions, must be characterized. First, a DNA construct is designed that has sequence homology with the flanking regions of one or more exons of the target gene. In addition to the homologous sequences, the new construct usually includes a gene encoding drug resistance or sensitivity. Once a construct is generated, it is transfected into murine embryonic stem cells (derived from mouse embryos) and then selected for expression of the foreign DNA by using the appropriate drug-based selection strategy. Because of the homology between the vector and the gene being targeted, some cells incorporate the foreign DNA construct via homologous recombination, with the complementary sequences either disrupting or completely replacing the target gene.

Transfected embryonic stem cells are propagated in culture, analyzed to verify the presence of the desired DNA construct, and then injected into developing mouse blastocysts. Once implanted into a pseudopregnant female, the blastocysts continue to develop normally and give rise to chimeric progeny in which some cells and tissues are derived from the original blastocyst and others are derived from the genetically altered embryonic stem cells. All cells derived from embryonic stem cells, including some gametes, have the gene of interest disrupted or replaced with a nonfunctional sequence. Matings between these genetically manipulated animals and wild-type mice result in a few offspring heterozygous for the disrupted gene. Subsequent mating of two heterozygotes results in a percentage of progeny homozygous for the mutation and therefore deficient in expression of the targeted gene.

Knockout mice provide a definitive method for establishing in vivo gene function; loss of a function after deletion of a gene is direct evidence that the gene is responsible for that function. Unlike the transgenic situation, in which the DNA construct is present in every cell but expressed in only certain cell types, in a knockout mouse the gene of interest is missing from every cell, but only cells or tissues in which the gene in question would normally be expressed are phenotypically affected.

Conditional Knockouts

Conditional knockout mice are created by introducing agents that selectively remove genetically engineered sequences from mouse chromosomes.⁴ In one such system, known as Cre-lox, the knockout mice are created as described earlier, except that palindromic loxP sites flank the target gene DNA construct. The Cre recombinase protein identifies loxP sites, and the DNA sequence between flanking loxP sites can be excised by the Cre protein to yield a functional knockout. The advantage of this system is that Cre expression can be spatially restricted to certain cell types through the use of tissue-specific promoters. Moreover, Cre constructs may be engineered so that Cre expression is induced only when certain drugs are administered to an animal expressing the Cre transgene. Such temporal and spatial control of knockout genes enables realistic modeling of carcinogenesis because this system can re-create a multistep biologic situation in which specific genes are deleted or overexpressed at specific junctures during tumor progression.

Somatic Cell Gene Transfer

Somatic cell gene transfer involves the transfer of DNA to nongermline cells, often using retroviruses engineered to contain specific genes.⁴ The concept of somatic cell gene transfer dates back to the discovery of Rous sarcoma virus (RSV) by Rous. Injection of RSV into the organs of several different types of animals results in the formation of tumors because of transfer of the v-src oncogene to somatic cells. However, RSV-based somatic cell gene transfer has some serious disadvantages: (1) there is no control over the type of cells infected by the virus and (2) the use of RSV somatic cell gene transfer allows the examination of only one oncogene.

One somatic cell gene transfer system that obviates these problems is the RCAS/tv-a system.⁴ This system combines transgene technology with viral infection to achieve cell-specific, gene-specific somatic cell transfer of a number of different oncogenes. Replication-competent avian leukemia virus (ALV) family splice acceptor (RCAS) vectors can be engineered to contain specific genes. Moreover, the RCAS virus specifically infects cells that express the ALV type a receptor tv-a. Thus, in transgenic animals engineered to express tv-a only in certain cell types, RCAS vectors can be used to deliver oncogenes in a cell type–specific manner.

Fluorescence In Situ Hybridization

Cancer, especially in its most malignant forms, progresses by accumulating multiple errors in the genome of affected organisms. Many of the genetic errors occurring in cancer specifically affect genes that normally function to repair damaged DNA. As a result of errors in DNA damage response mechanisms, the entire genome of an affected organism becomes prone to error, thereby leading to a phenomenon called genomic instability. One consequence of genomic instability is that damaged genomic DNA is prone to single- and double-strand breaks. At a chromosomal level, these breaks can result in very large losses, gains, or translocations of chromosomal regions. These chromosomal areas encode genes or regulatory elements of genes involved in cancer initiation and progression, and identification of these abnormal chromosomal regions is important for a general understanding of cancer biology. As a consequence, cytogenetics, which encompasses the study of chromosomal structure, has become a central part of cancer biology.⁶

Before the advent of molecular cytogenetic techniques, the primary method for studying chromosomal aberrations was direct visualization of chromosomes during the metaphase period of the cell cycle. Chromosome condensation during this time allows the highly compact metaphase chromosomes to be directly visualized with light microscopy. Chromosomal spreads—metaphase or interphase chromosomes placed on glass slides—can distinguish unique properties of individual chromosomes and allows chromosome regions to be identified. Aberrant chromosome structures, such as those found in Down’s syndrome, can therefore be readily identified by looking at chromosome spreads. This technique allowed proper enumeration of diploid human chromosomes and also identification of the chromosome numerical aberrations observed in Kleinfelter’s and Down’s syndromes. However, gross cytogenetics fails to provide the resolution required to identify the small regions of DNA amplification or deletion that are frequently encountered in neoplastic diseases.

Molecular cytogenetics is an amalgamation of molecular biology and the conventional cytogenetic methods described earlier. The prototypical molecular cytogenetic technique, fluorescence in situ hybridization (FISH), is now widely used for the identification of chromosomal aberrations in cancer genomes. FISH techniques can easily identify large chromosomal abnormalities such as rearrangements and deletions; by using modern FISH techniques investigators can occasionally identify single nucleotide abnormalities in chromosomes. FISH requires the use of a probe, a synthetic oligonucleotide sequence that corresponds to a known region of the genome. The probe is then hybridized to the target DNA (usually genomic DNA of the researcher’s interest). During synthesis of the probe, modified nucleotides containing fluorophores or antigenic sites are used to create a modified oligonucleotide. Oligonucleotide probes modified with fluorophores can be visualized directly because of the fluorescence emitted by the probes; alternatively, antigen-modified oligonucleotides can be visualized by immunologic or enzymatic reactions. After synthesis of the probe, both the probe and target DNA are denatured and then mixed to allow binding of the probe and target DNA. Structurally normal target DNA will exhibit a signal at the expected site in the chromosome, whereas rearranged or deleted target DNA will exhibit a signal in an aberrant location or no signal at all, respectively.

Array Platform Technology

Modern molecular genetics is aided by the availability of high-throughput systems that permit concurrent analysis of thousands of genes or gene products. A DNA microarray is one such tool and consists of solid platforms that contain thousands of spots of immobilized complementary DNA (cDNA) or genomic DNA oligonucleotides. The identity of each spot on the array is known and the DNA is labeled with fluorescent tags. In a typical experiment, genomic DNA or cDNA from experimental samples, called “probes,” are labeled with fluorescent tags that differ from the immobilized “target” DNA on the microarray. The sum of the fluorescent signal obtained from hybridizing the probe and target represents the relative abundance of either the probe or target. Because a laser scanner can analyze the fluorescent signals from array experiments, each experiment can be performed in a relatively short time. DNA microarrays can provide enormously useful information about the expression levels of DNA.

Comparative Genomic Hybridization Arrays

Although FISH has significantly advanced our progress in the discovery of abnormal target genes and chromosomes, the technique is imperfect for solid tumors in that it requires preparation of tumor metaphase chromosomal spreads that do not differ markedly from normal tissue. Because advanced solid tumors acquire many chromosomal abnormalities, it is often impossible to interpret tumor chromosomal spreads with accuracy inasmuch as chromosomes in solid tumors are often altered beyond recognition. To combat this problem, a variant of FISH called comparative genomic hybridization (CGH) was invented in the early 1990s.⁶ CGH obviates the need for chromosome spreads by using differentially labeled test and reference genomic DNA oligonucleotide samples, which can then be applied onto genomic DNA microarrays. CGH arrays permit unprecedented ability to evaluate chromosomal abnormalities in cancer. In particular, this technique is especially sensitive for the identification of changes in DNA copy number, such as the increases in gene copy number that occur at the epidermal growth factor receptor (EGFR) locus in a significant number of glioblastomas.

Single Nucleotide Polymorphism Genotyping Arrays

FISH and CGH techniques improve the analysis of gross chromosomal abnormalities (especially changes in copy number) that occur in cancer. However, comprehensive analysis of cancer genetics must address other specific events such as deleterious point mutations and the sequential loss of tumor suppressor alleles (loss of heterozygosity [LOH]) because these events are key contributors to the initiation and progression of cancer. Large-scale identification of point mutations and LOH requires the ability to scan large swaths of tumor genomes with high resolution.

Single nucleotide polymorphism (SNP) analysis is a powerful molecular genetic tool capable of identifying changes in DNA copy number, point mutations, and LOH.⁶ SNPs are nonfunctional point mutations in the genome that occur at a frequency of about 1% in the human genome. They occur at a frequency of one SNP every few hundred base pairs of genomic DNA. Because many SNPs have been sequenced, they can be used to compare the haplotypes of cancer genomes with other nontumor DNA from the same individual. Moreover, because the genomic location of SNPs is known, changes in the SNP genotype of tumor DNA identify genomic regions that can be further analyzed by PCR to detect precise DNA aberrations (point mutations, deletions, insertions, and so on) that underlie the changes in SNP haplotype. Furthermore, because of the availability of microarray platforms, up to 500,000 SNPs can be analyzed in a single experiment, thus allowing excellent genomic resolution.

Gene Expression Arrays

CGH, FISH, and SNP techniques identify changes in genomic DNA. Although these platforms can be used to make inferences about gene expression, they do not provide direct data about the transcription of these genes from DNA to messenger RNA (mRNA). The importance of studying changes in gene expression is aptly represented by the expression pattern of vascular endothelial growth factor (VEGF) during the genesis of glioma. Although the VEGF gene is rarely mutated in glioblastoma multiforme (GBM), VEGF mRNA and protein levels are often markedly elevated, which leads to an increase in tumor angiogenesis and virulence.⁷

Gene expression arrays are now widely available. In practice, gene expression array techniques are identical to the CGH and SNP arrays, except that cDNA (transcribed from mRNA) is most frequently used as a substrate for hybridization. Currently, up to 19,000 unique genes can be analyzed for their expression levels in a single array experiment.

Applying Molecular Genetic Tools to Glioma Analysis

Despite the many technologic advances in the molecular characterization of gliomas, there was, until recently, a dearth of data linking molecular genetic changes to clinical outcomes for gliomas. These shortcomings partially resulted from studies that relied on single molecular genetic techniques and a small number of glioma samples. Recently, however, a number of studies have used a combination of molecular genetic and other techniques to analyze large samples of GBM, and the results from these studies are promising.

Phillips and colleagues recently used both gene expression and CGH profiling of high-grade gliomas to explore genes that correlate with patient survival.⁸ Using gene expression microarrays, they identified three gene clusters that correlate with patient survival. Consistent with these findings, CGH analysis of patients with shorter survival showed frequent loss of the phosphatase and tensin homologue from chromosome 10 (PTEN) tumor suppressor locus and gains at the EGFR and phosphatidylinositol-3′-kinase (PI3K) oncogene loci in comparison to patients with longer survival.

Perhaps the most comprehensive single genetic analysis study of human gliomas to date comes from the efforts of The Cancer Genome Atlas (TCGA), a national initiative whose goal is to identify and catalogue the wide array of genetic aberrations that cause cancer.⁹ TCGA uses a number of genetic tools, including CGH, SNP genotyping, PCR genotyping, gene expression arrays, microRNA profiling, and DNA methylation arrays, to characterize a cancer specimen. Using 206 GBM samples in a pilot study, the group has identified hitherto unknown genetic aberrations that probably contribute to glioma genesis. In particular, they identified novel deletions of NF1 and PARK2 in glioma samples and characterized NF1 (neurofibromin 1) as a frequent glioma tumor suppressor gene. Beyond discovering individual genetic aberrations in glioma, TCGA characterized a number of systematic, nonrandom patterns that typify genetic aberrations in glioma. For example, the study revealed a tendency for GBM to acquire single mutations in core signaling pathways rather than multiple mutations in a single pathway (Fig. 100-1). One prediction of this model is that aberrations in the receptor tyrosine kinase (RTK) pathway, to use one example, appear to relieve the pressure for accumulating additional RTK pathway aberrations but do not relieve the pressure to acquire errors in the p53 tumor suppressor pathway or in cell cycle progression pathways. This important finding supports the idea that adequate treatment of GBM may require combinations of drugs that simultaneously target components of the three core GBM signaling pathways: growth factor receptor signaling, cell death/apoptosis, and cell cycle progression. In addition to these core signaling pathways, gliomas often use additional proangiogenic and drug efflux mechanisms. Pharmacologic manipulation of these pathways, in addition to the core glioma pathways, might be a necessary part of the treatment regimen for high-grade gliomas.

FIGURE 100-1 Overview of core glioma signaling pathways and potential therapeutic compounds that target different components of the pathways. CDK, cyclin-dependent kinase; FTI, farnesyltransferase inhibitor; GF, growth factor; HIF-1α, hypoxia inducible factor-1α; mTOR, mammalian target of rapamycin; NF1, neurofibromin 1; PI3K, phosphatidylinositol-3′-kinase; PTEN, phosphatase and tensin homologue deleted from chromosome 10; RB, retinoblastoma; RTK, receptor tyrosine kinase; TSC1/2, tuberous sclerosis complex 1/2; VEGF, vascular endothelial growth factor. Orange, tumor suppressors; green, oncogenes; yellow, pharmacologic inhibitors.

Other studies, using molecular genetic techniques similar to those used in TCGA, corroborate many of the findings identified in TCGA and a number of other studies of glioma genetics. Parsons and colleagues used direct sequencing, DNA copy number analysis, and gene expression studies to identify mutations in the core pathways just discussed.¹⁰ In addition to identifying mutations in these core glioma pathways, they discovered mutations in isocitrate dehydrogenase 1 (IDH1), a gene previously not associated with gliomas, that predicted better overall survival for patients harboring these mutations. Moreover, mutations in IDH1 tend to segregate with favorable GBM characteristics such as younger age and secondary GBM. IDH1 is an enzyme that selectively catalyzes the oxidative decarboxylation of isocitrate to 2-oxoglutarate. The fact that IDH1 is an enzyme raises the intriguing possibility that pharmacologic inhibition could re-create the effects of IDH1 mutations. However, it should be noted that the relationship of IDH1 mutations to the catalytic function of the protein is currently unknown. Because IDH1 is selectively mutated in less aggressive GBMs, elucidation of the biologic function of IDH1 could provide novel therapeutic targets that convert aggressive high-grade gliomas into more differentiated glioma types, such as that found in younger patients. IDH1 mutation and methylguanine methyltransferase (MGMT) promoter methylation are the only individual genetic alterations that predict improved survival in GBM. Overall, these studies highlight recent progress in the discovery of new glioma-associated genes and lay the groundwork for drug discovery.

Potential Therapeutic Intervention: RTKs, PI3K, AKT, and Ras

Two EGFR inhibitors, erlotinib (Tarceva, Genentech, San Francisco, CA) and gefitinib (Iressa, AstraZeneca, Wilmington, DE), have been used in glioma clinical trials. Unfortunately, these clinical trials did not show a significant reduction in glioma progression.¹⁵ Although these results are disappointing, it is possible that the molecular genetic heterogeneity found in gliomas account for some of the response patterns in individual tumors. Highlighting this point are recent studies suggesting that EGFR and PTEN status predict responses to EGFR inhibitors. In one study, gliomas responding to an EGFR inhibitor were EGFRvIII positive and had wild-type PTEN, whereas nonresponding tumors often displayed PTEN deletions.¹⁶ Mechanistically, this pattern of response implies that PTEN tumor suppressor loss activates downstream RTK signaling cascades, regardless of the RTK status. Several other promising RTK inhibitors are currently being evaluated for gliomas.

PI3K pathway inhibitors are now available for the treatment of gliomas. In particular, AKT and mTOR, the downstream effectors of PI3K activation, are receiving considerable attention in glioma treatment strategies. Perifosine (KRX-0401, Keryx Biopharmaceuticals, New York, NY), an oral AKT inhibitor, has shown great promise in mouse GEMs and is now in clinical trials for the treatment of malignant gliomas.¹⁷ Rapamycin (sirolimus, Wyeth Pharmaceuticals, Collegeville, PA) and CCI-779 (temsirolimus, Wyeth Pharmaceuticals, Collegeville, PA), both mTOR inhibitors, have shown antiglioma activity in vitro and are now being used in clinical trials.

Overview of the Cell Cycle

Normal cells cycle through several phases, each of which serves a specific function in either cell growth or cell division. During the G₁ phase, genes are expressed and protein synthesis occurs, but cells still contain the normal complement of chromosomes consisting of two copies each (2n; diploid). Some of the proteins synthesized in G₁ contribute to cell function and growth, whereas others are necessary prerequisites for DNA synthesis, which occurs in the next phase, S (for synthesis). During the S phase, the cell replicates its DNA so that it contains two copies of genetic material, thereby resulting in a doubled chromosome number (4n). The subsequent G₂ phase is characterized by growth and additional protein synthesis, which results in one cell with sufficient protein and DNA content to support two daughter cells. The cell then proceeds into the M (for mitosis) phase, where the two sets of identical chromosomes migrate to opposite poles and two separate cells with identical genetic information are created. The resulting daughter cells each contain 2n DNA. At this point the cell either begins the cycle anew at G₁ or enters a quiescent phase called G₀, in which the cells are essentially at rest. Normal cells spend most of their time in G₀ and G₁. The two regulatory pathways that normally maintain cells in G₁, called G₁ arrest pathways, have been extensively studied and are tightly interwoven.

One pathway, the tumor suppressor Rb pathway, is initiated by the protein p16^INK4 and another protein, p15^INK4. The indirect function of p16^INK4 and p15^INK4 is to maintain cells in G₁ by preventing phosphorylation of the protein product of the retinoblastoma (RB1) gene. The normal function of nonphosphorylated or hypophosphorylated Rb protein is to halt the cell cycle at G₁. As the cell cycle progresses, Rb is continually phosphorylated by cyclin-D/cyclin-dependent kinase 2/4/6 (CDK2/4/6) complexes. Phosphorylation of Rb inactivates the protein, which allows cells to progress past G₁ and into the latter stages of the cell cycle.

The other pathway, the p53 tumor suppressor pathway, begins with a protein called p19^ARF in mice and p14^ARF in humans. The Arf proteins inhibit another protein called Mouse Double Minute 2 homolog (Mdm2), whose normal function is to promote degradation of the p53 tumor suppressor protein. The p53 protein has a variety of regulatory roles in the cell cycle and during programmed cell death; its major role during the cell cycle is to recognize damaged DNA and halt the cell cycle in G₁ until appropriate cell fate decisions can be made. p53 accomplishes this goal by increasing the production of another protein, p21, a cyclin-dependent kinase inhibitor (CDKI) that represses the function of CDK2 and CDK4. Thus, p53 indirectly maintains Rb in a hypophosphorylated state by inhibiting the CDK complexes that normally phosphorylate Rb and alleviate G₁ arrest. Cell cycle dysregulation can lead to cancer through the repression of tumor suppressor genes (i.e., inactivation of p53, Rb, p16^INK(4)a, p15^INK4, p14^ARF, and CDKIs) or the promotion of cell cycle oncogene activity (i.e., activation of CDK4/6 and Mdm2/4). Both these scenarios are seen in gliomas.

Cell Cycle Abnormalities in Gliomas

Cell cycle aberrations feature prominently in the genesis of glioma. The most frequent genomic aberration in GBM cell cycle control is the simultaneous deletion of p14^ARF and p16^INK4. p16^INK4, like p21, functions as a CDKI that represses the activity of CDK4. Remarkably, both the p14^ARF and p16^INK4 genes are encoded by the same genetic locus, CDKN2A, alternative splicing of which gives rise to either p14^ARF or p16^INK4. Thus, deletion of CDKN2A results in the absence of its two gene products. CDKN2A is deleted in approximately 50% to 60% of all GBMs and is apparently more frequently deleted in primary than in secondary GBMs.^9,10 Another genetic locus, CDKN2B, which resides near CDKN2A on chromosome 9p21, encodes for p15^INK4 and is almost always deleted along with CDKN2A. The p15 protein inhibits CDK4 and CDK6 function. Inactivation of CDKN2A/CDKN2B or p53 is an important step in glioma genesis because nearly all GBMs feature inactivating mutations at either of these two loci.^9,10 In the case of p53, mutations or homozygous deletions occur in approximately 30% to 60% of GBMs. Although previous evidence suggested that p53 alterations were overwhelmingly associated with secondary GBMs, recent data point to an important role for wild-type p53 in the prevention of primary GBMs as well. In addition to these three major tumor suppressors, RB1, the gene for the Rb protein, is homozygously deleted in about 10% of all GBMs.⁹ Thus, it is clear that loss of cell cycle tumor suppressor genes is a major mechanism for GBM pathogenesis.

Although cell cycle dysregulation in gliomas occurs mainly as a result of tumor suppressor inactivation, genomic aberrations in cell cycle oncogenes are also found. For example, amplification of CDK4 is present in approximately 10% to 20% of GBMs, and a smaller number of GBMs feature amplification of the CDK6 gene.^9,10 In addition to the CDKs, many amplifications at the genomic loci of the MDM2 and MDM4 oncogenes have been reported in GBM (Fig. 100-2).

FIGURE 100-2 Pathways controlling the cell cycle as targets for glioma therapy. A, Pathways leading to G₁ cell cycle arrest and diagram of the cell cycle. Potential targets are indicated by open arrows. B, The gene structure of INK4a-ARF, which encodes both the gene products regulating the G₁ cell cycle arrest pathways.

Programmed Cell Death

Damaged cells reach a type of crossroad after DNA damage. These cells often undergo cell cycle arrest as described earlier, but the cells must then assess the severity of the damage and decide whether to continue with cell division or to self-terminate because of catastrophic DNA damage. Cell death mechanisms are, by definition, defective in cancer.

Regulation of cell death is partly carried out by the familiar p53 protein. p53 is a multifaceted protein. In addition to its role during cell cycle progression, p53 is also a central participant in determining whether damaged cells live or whether they are terminated in a process known as programmed cell death, or apoptosis. Therefore, p53 inactivation can facilitate inappropriately speedy progression through the cell cycle on the one hand and inappropriate survival of damaged cells on the other. Because of its central role in regulating cell fate decisions, p53 is referred to as the “guardian of the genome.”

The balance of cellular proapoptotic and antiapoptotic signals determines cell death decisions. The Bcl family of proteins includes positive and negative regulators of the apoptosis process, with Bcl-2, Bcl-x, and Bcl-xl serving as antiapoptotic members. The proapoptotic function of p53 is carried out by its transcriptional target genes, many of which belong to the proapoptotic arm of the Bcl family. In particular, Bax, p53 upregulated modulator of apoptosis (PUMA), Bak, and NOXA are four Bcl family proteins whose overexpression, induced by p53, causes termination of cells by apoptosis.¹⁸ The most frequent p53 mutations in GBM, the “hotspot” mutations, can diminish the ability of p53 to transactivate both p21 and Bax. However, in some GBMs there is emerging evidence that certain p53 mutations specifically incapacitate its proapoptotic functions while sparing its cell cycle arrest functions. In one case of grade II astrocytoma, a germline point mutation in one allele of p53 led to decreased induction of Bax but maintained the ability to induce p21. However, after progression to GBM, the tumor acquired additional point mutations in p53 that rendered the tumor cells incapable of inducing either p21 or Bax. Hence, gliomas gain a selective growth advantage by incapacitating the proapoptotic functions of p53 in addition to its cell cycle arrest properties, and it appears that both functions can hasten the progression of glioma.¹⁹

Potential Therapeutic Intervention: Cell Cycle and Apoptosis

Given the importance of the cell cycle and apoptosis pathways in tumor behavior, is it possible to intervene to correct the problem? Therapeutic efforts are complicated because dysregulation of G₁ arrest and apoptosis in glioma is usually induced by loss-of-function mutations; for example, loss of CDKN2A, p53, and RB is common in glioma. Although gain-of-function mutations (such as RTK hyperactivity) can be combated by pharmacologic inhibition, loss of CDKN2A and TP53 cannot be corrected by simple pharmacologic means. The initial thought would be to attempt gene therapy to transfer the CDKN2A or p53 genes back into tumor cells. The problem with this is immediately apparent: the efficiency of gene transfer is not 100%, and the small percentage of cells not incorporating the gene might continue to express the faulty pathway and therefore maintain tumor behavior.

Conditional expression of wild-type TP53 in tumor cells has shown remarkable decreases in tumor growth in animal models, thus suggesting that p53 gene therapy can diminish cancer growth. In addition, TP53 gene therapy has been applied successfully to glioma cells in culture. As such, the idea of gene therapy held tremendous promise for the treatment of gliomas in the 1990s; however, clinical trials conducted earlier this decade involving adenoviral transfer of the p53 gene have been disappointing.²⁰ The failure of tumor suppressor gene therapies may be purely technical, and better success may be attained if higher rates of viral infection are achieved in tumor cells. One alternative approach for targeting tumor cells would be to eradicate them by using replicative viruses that selectively replicate in cells lacking certain tumor suppressors.

Another way to intervene in the cell cycle/apoptosis pathway is to selectively inhibit the oncoproteins in the pathway (i.e., CDK4/6 and MDM2/4) with small-molecule inhibitors. Blocking the function of proteins such as CDK4 and Mdm2, which provide oncogenic signals when hyperactive, would allow normal control downstream of disruption of the pathway and promote the return of normal cell cycle regulation. Several pharmacologic CDKIs exist, and some of them are currently in clinical trials. In addition to pharmacologic CDKIs, the recent discovery of an Mdm2 inhibitor family, the nutlins, bodes well for the pharmacologic targeting of cell cycle progression.

Overview

Until recently, the prevailing view was that cancer cells are simply terminally differentiated cells that acquire mutations and the capacity for dysregulated growth. A more recent model of carcinogenesis, the CSC model, has begun to challenge this older paradigm. The CSC model argues that the most tumorigenic of all cancer cells are those that resemble normal stem cells. Thus, the CSC is a type of multipotent tumor cell capable of giving rise to the more differentiated populations in the bulk of the tumor. In human gliomas, some evidence suggests that certain glioma CSCs can be identified by their concurrent expression of the stem cell marker nestin and the transmembrane glycoprotein CD133. CD133⁺ cells isolated from GBMs are capable of generating GBMs in xenograft models.²¹ Moreover, although 100,000 CD133⁻ cells were incapable of generating xenograft GBMs, only 100 CD133⁺ cells were required to produce GBMs that phenocopy the original human GBM. This evidence suggests that at least in a subset of GBMs, CD133⁺ cells may be tumor-initiating CSCs.

Whether glioma CSCs originate from normal neural stem cells or whether they represent the acquisition of an undifferentiated character by mature brain cells is currently unknown. What is known from genetic analysis of GBMs is that they display a significant increase in gene products usually associated with normal neural progenitor and stem cells. Proteins such as Olig2, nestin, and CD133, which are usually associated with immature stem cells, are enriched in GBM. Indeed, a subset of GBM, the proneural group, can be classified primarily by its ability to produce genes normally associated with neuronal progenitor cells.⁸ The similarities between glioma CSCs and normal stem cells create a quandary: how do we develop therapies that target the stem-like properties of GBM without destroying our own normal neural stem cells? The answer to this question might be to target GBM CSCs where they live, in the perivascular niche of aggressive brain tumors.

Cancer Stem Cells, the Perivascular Niche, and Radioresistance

Normal neural stem cells reside in a stem cell niche in the subventricular zone (SVZ) of the lateral ventricles. Here, information from the microenvironment and surrounding vasculature provides survival and other cues to the resident stem cells. Perhaps the most intriguing property of SVZ stem cells is their close proximity to a vascular plexus enveloping the entire SVZ. Not surprisingly, GBM and medulloblastoma CSCs also reside next to blood vessels (Fig. 100-3).¹¹ This perivascular property of CSCs implies that tumor vascular proliferation provides critical cues to CSCs and probably helps them maintain their multipotent status.

FIGURE 100-3 The glioma tumor microenvironment. Chemoresistant and radioresistant cancer stem cells (CSCs) exist in a perivascular niche with astrocytes (ASTs), macrophages (MΦ), and oligodendrocytes (OGDs). CSCs give rise to transit-amplifying cells (TACs) in the tumor, which then give rise to the tumor bulk. Please see Gilbertson and Rich.¹¹

Stem-like cells, in addition to possessing high tumorigenic potential, are also emerging as the population of tumor cells with the highest resistance to current GBM treatment regimens. Recent evidence suggests that CD133⁺ xenograft GBMs are more resistant to radiation than are their CD133⁻ counterparts. According to one study, the radiation resistance mechanism of CD133⁺ cells involves activation of the DNA damage checkpoint kinases Chk1 and Chk2. These kinases are involved in postreplication DNA damage repair. During glioma formation, Chk1- and Chk2-induced DNA damage repair might obviate fatal DNA damage in tumor cells and promote survival despite the genotoxic stress induced by radiation.²²

Another radiation resistance mechanism used by stem-like cells is to temporarily arrest proliferation after radiotherapy because rapidly proliferating cells are the primary targets of radiation-induced cell death. In medulloblastomas, there is evidence that CSCs, residing in the tumor perivascular niche, briefly stop proliferating. These nestin-expressing cells respond to radiation by inducing cell cycle arrest, but unlike cells from the bulk of the tumor, these cells do not die and instead reenter the cell cycle 72 hours after irradiation.²³ Whatever the mechanisms for their radiation resistance, it is likely that certain undifferentiated cells survive radiation and contribute to repopulation of the tumor bulk, thus highlighting the importance of a therapeutic approach that targets them.

Cancer Stem Cells, the Blood-Brain Barrier, and Chemoresistance

Resistance to chemotherapy is virtually universal in GBM. One explanation for chemotherapy resistance is that the architecture of the brain vasculature forbids entry of many drugs into the brain parenchyma. The blood-brain barrier (BBB), equipped with tight junctions, pericytes, and astrocytic end feet, significantly diminishes the amount of paracellular transport from the luminal (blood-facing) side of the vasculature to the abluminal (brain-facing) side. The reduction in paracellular transport of drugs and solutes means that most drugs and solutes must depend on transmembrane transport, transcellular migration, and finally efflux into cerebrospinal fluid or interstitial fluid. Two large families of transmembrane proteins govern the transcellular transport of drugs across the BBB: the adenosine triphosphate (ATP)-binding cassette (ABC) family of transporters and the solute carrier (SLC) superfamily of transporters. These transporters are capable of effluxing a number of substrates, from RTK inhibitors such as gefitinib to cytotoxic agents such as camptothecin and methotrexate. Each individual transporter has a defined spectrum of substrate specificities, but there are often multiple different transporters and transporter families in a given cell. This suggests that many CSCs have the capacity to remove several different drugs from their intracellular compartments. Many of these transporters are highly expressed on the luminal side of the normal and tumor vasculature, where they are responsible for efflux of drugs back into the bloodstream. Endothelial cell transporters at the BBB might explain one aspect of the chemoresistance observed in glial brain tumors. However, even when solutes make it across the vasculature and into the brain parenchyma, they face a series of antichemotherapy resistance mechanisms mediated by CSCs, the tumor bulk, and their support cells.

CSCs might be responsible for a portion of the resistance to chemotherapeutic agents that make it across the BBB. Analogous to their role during radiation resistance, CSCs display marked resistance to chemotherapy. Normal stem cells and CSCs from a variety of organs produce a number of drug transporters. Many ABC family proteins, including ABCB1 and ABCC1, are highly expressed in hematopoietic stem cells but not in their more differentiated progeny. One ABC family member, ABCG2, is an intriguing candidate for chemoresistance in some GBMs. ABCG2 substrates include imatinib, methotrexate, irinotecan, and doxorubicin, but not temozolomide. Recent reports suggest that ABCG2 is significantly upregulated in the CSC-like population of cells derived from the C6 glioma cell line. In mouse models of GBM, ABCG2 is also highly enriched in the CSC population, and this ABCG2-expressing CSC population, called the side population, gives rise to aggressive glial tumors in mice. The hope is that inhibition of ABC and other transporter families can sensitize tumor cells to the effects of chemotherapy. Unfortunately, clinical trials with the ABC inhibitors tested thus far have not shown improved sensitivity to chemotherapy, although no specific inhibitors for ABCG2 have been used. Nonetheless, chemotherapeutic sensitization of GBM with single or combined drug transport inhibitors is a potential option for the future treatment of GBM.

Angiogenesis

Without a blood supply to carry oxygen and nutrients to rapidly proliferating cells, tumors would essentially be self-limited. During early neoplastic growth, the same capillaries that serve surrounding normal tissues feed tumor cells. As tumors grow and require additional blood supply for survival, they send out signals to neighboring vessels that instruct endothelial cells to proliferate and form new vessels. The endothelial cells that make up these blood vessels are not neoplastic themselves; they simply respond to recruitment messages from the tumor cells. Angiogenesis is the process by which tumor cells and endothelial cells communicate to give rise to new blood vessels. Not surprisingly, the language of this communication requires growth factor–growth factor receptor interactions and signal transduction. Both tumor cells and endothelial cells produce a number of positive regulators of angiogenesis, including VEGF, a powerful angiogenic growth factor that was briefly discussed earlier. In addition to VEGF, other proangiogenic factors such as basic fibroblast growth factor (bFGF), interleukin-6 (IL-6) and IL-8, hypoxia inducible factor-1α (HIF-1α), and members of the angiopoietin family are produced by tumors. At the same time, tumor cells often decrease the production of negative regulators of angiogenesis, especially thromobospondins, certain interferons, and endostatins.¹²

Malignant gliomas frequently exhibit a so-called angiogenic switch that occurs during the transition from low-grade to high-grade tumors and is marked by intense proliferation of blood vessels in the tumors. One major regulator of angiogenic switch is HIF-1α, which is induced by hypoxic regions of the tumor and serves as a transcription factor for many angiogenic targets, especially VEGF and VEGF receptors (VEGFRs).¹² VEGFRs are RTKs and, on binding VEGF, induce a series of signal transduction pathways leading to endothelial cell proliferation, increases in vascular permeability, and promotion of tumor cell invasion. Other events occurring during the transition from low-grade to high-grade tumor, such as loss of the PTEN tumor suppressor, activation of the PI3K pathway, activation of Ras, and activation of other RTKs, promote vascular proliferation via VEGF-VEGFR–independent mechanisms. The importance of tumor angiogenesis is now being reviewed anew, especially given the relationship of CSCs to perivascular regions in most GBMs.

Invasion

Pioneering neurosurgeons appreciated the highly infiltrative nature of malignant gliomas. Desperate to cure these invasive tumors surgically, they performed hemispherectomies in patients with apparent unilateral gliomas to attain good surgical resection margins. However, even after such dramatic interventions, GBMs would arise from the contralateral hemisphere. These findings highlight one devastating reality about glioma cells: they are highly invasive and easily spread to regions of the brain where there is no radiographic evidence of disease. The invasive properties of GBMs are established mainly by the interaction of tumor cells with their microenvironment. The microenvironment, which consists of nontumor cells such as macrophages, endothelial cells, nontransformed glial cells, and the extracellular matrix (ECM), acquires specific features that permit tumor cell proliferation and invasion.

At a molecular level, gliomas produce a large number of proteins that function to aid glioma cell propagation from the tumor bulk to more distant sites within the brain. GBMs use cell adhesion molecules whose function is to interact with the ECM, extracellular proteases that break down the ECM, and promigratory/cytoskeletal proteins that help cells migrate from one region of the brain to the next. Chief among these proinvasion molecules is the cell adhesion receptor CD44. CD44 is a transmembrane glycoprotein for the ECM components osteopontin (OP) and hyaluronic acid (HA).¹² OP and HA binding of CD44 causes cells to become more motile by rearranging the actin cytoskeleton and forming cell protrusions known as lamellipodia. Other transmembrane proteins in glioma cells, particularly those of the integrin family, induce cytoskeletal rearrangements and increase cell motility on binding ECM components. CD44, OP, HA, and members of the integrin family are significantly upregulated in gene expression studies of GBM.

Invasive GBM cells must overcome an ECM that is normally restrictive to cell migration. To achieve significant tumor cell migration, GBMs engage in an active ECM remodeling program that uses ECM proteases. Four protein families are particularly important for glioma-induced ECM remodeling: the zinc-dependent matrix metalloproteinase (MMP) family, a family of cysteine proteases called cathepsins, the α disintegrin and metalloproteinase (ADAM) family, and the urokinase-type plasminogen activator receptor (uPAR). These proteins all play important roles in degrading protein components of the ECM, which allows the ECM to loosen and thereby enhances glioma cell migration through the ECM. Glioma gene expression studies confirm the hyperactivity of these pathways in GBMs, and the levels of these genes show positive correlation with glioma grade.¹²

Potential Therapeutic Intervention: Angiogenesis and Invasion

Data from preclinical studies suggest that tumor angiogenesis and invasion are attractive targets for glioma treatment. Data from clinical trials are just now emerging. Many antiangiogenesis therapies focus on VEGF, including anti-VEGF antibodies, anti-VEGFR antibodies, small-molecule tyrosine kinase inhibitors with high VEGFR specificity, and soluble VEGFRs (VEGF-Trap) that bind free VEGF. Clinical trials with bevacizumab (Avastin, Genentech), a monoclonal anti-VEGF antibody, and the topoisomerase inhibitor irinotecan have shown a modest survival benefit in patients with recurrent malignant gliomas. VEGF-Trap is currently in clinical trials for malignant gliomas. Small-molecule VEGFR inhibitors have been less successful as monotherapy or in combination with standard agents for the treatment of malignant gliomas.

Given the close physiologic relationship between angiogenesis and invasion, it is hard to interrupt one of these processes without affecting the other. For example, because integrins are expressed on both endothelial cells and tumor cells, treatments aimed at integrins will probably disrupt the interaction of both cell types with the ECM. In the case of endothelial cells, angiogenesis would be inhibited, whereas tumor cells treated with anti-integrin therapies should have limited invasive and proliferative properties. Cilengitide (Merck, Whitehouse Station, NJ), an α_vβ_3/5 integrin inhibitor, has shown promising results in early clinical trials of malignant glioma.¹² Because α_vβ₃ and α_vβ₅ integrin are expressed on both tumor and endothelial cells, it is possible that the antineoplastic mechanism of the molecule is related to inhibition of tumor cell migration, endothelial cell migration, and endothelial cell proliferation. Many other agents that specifically target microenvironment factors in malignant gliomas are being developed, and these agents will be an important part of integrated approaches to the treatment of glioma.

Astrocytoma Models

GEMs confirm the importance of many genetic alterations seen in human astrocytoma. Central nervous system–specific overexpression of oncogenes such as RTKs and their downstream signal transduction partners can result in astrocytoma. Fourteen percent of transgenic mice expressing the RSV oncogene v-src under control of the glial fibrillary acidic protein (GFAP) promoter eventually form astrocytomas of the brain and spinal cord. Similarly, in mice that express a constitutively active form of Ras from the GFAP promoter, malignant astrocytomas develop by 9 months.⁴

GEMs also capture the effects of tumor suppressor loss in astrocytoma. Compound heterozygous loss of Nf1 and TP53 causes high-grade astrocytomas whose penetrance is partially dependent on the mouse strain. Similarly, conditional knockout of Nf1 in p53-null mice produces malignant astrocytomas that appear to arise from the SVZ.²⁴ Loss of cell cycle pathway tumor suppression can also result in high-grade astrocytoma. For example, expression of the pRb-inactivating SV40 T antigen T₁₂₁ in GFAP-positive cells leads to high-grade astrocytomas.

The RCAS/tv-a system is capable of inducing astrocytomas with histologic fidelity to human astrocytoma and relatively short latency. Astrocytomas can be formed in mice with activated Akt and Ras (RCAS-Akt Myr Δ 11-60 and RCAS-Ras^G(12)D) delivery to nestin-positive, tv-a–positive (Ntv-a) mice.²⁵ Combined RCAS-Ras^G12 with loss of the PTEN tumor suppressor also results in astrocytoma in mice. Interestingly, these same RCAS-oncogene combinations in GFAP-positive, tv-a–positive (Gtv-a) mice did not result in astrocytoma formation, thus suggesting that specific cell types require specific genetic alterations for astrocytoma formation.

Oligodendroglioma Models

Human oligodendrogliomas express high levels of PDGF ligand and PDGFRα. Therefore, it is not surprising that overexpression of PDGF ligand results in oligodendrogliomas in a number of mouse models. One well-studied GEM of PDGF overexpression in gliomas is the RCAS/tv-a system, where RCAS-PDGFB–injected Ntv-a mice acquire oligodendrogliomas with high penetrance and a latency of 3 weeks to a few months, depending on the tumor suppressor loss status of the mice.⁴ Notably, RCAS-PDGFB–infected Gtv-a mice result in mixed oligoastrocytomas, again reflecting cell-type and gene-alteration specificities in the formation of different glial tumors.

Another somatic cell gene transfer system using the replication-competent Moloney murine leukemia virus (MMLV) engineered to express PDGFB also results in oligodendrogliomas of varying grade. This system is somewhat limited because of the lack of cell-type specificity in transfer of the PDGF oncogene.

EGFR activation also has a role in the formation of oligodendrogliomas. GEMs of oligodendroglioma can be created in mice that have activated viral EGFR homologue v-erbB expression driven by the glial promoter S100β. These mice usually form low-grade oligodendrogliomas. However, like RCAS/tv-a oligodendrogliomas, tumor grade can be increased by heterozygous deletion of p53 or Ink4a/Arf.⁴