Rogier Versteeg
The challenge
The past 20 years of clinical cancer management have brought great success: cure rates for many paediatric tumours have reached 80 per cent or higher. This remarkable progress in treatment has been achieved without a significant contribution from the insights obtained in molecular biologic cancer research, although major contributions to prognostic classification have come from cytogenetic identification of chromosomal translocations, particularly in leukaemias and lymphomas. Treatment has been mainly improved by new variations on the classical treatments: cytostatic drugs, radiation, and surgery. However, since the molecular identification of the first oncogene in the late 1970s, basic cancer research has seen an extraordinary development in the understanding of mechanisms which control normal and aberrant cell division. The challenge for the next decennium will be to use these insights to improve cancer treatment further.
Many tumours are still refractory to current therapies. Some tumour types have seen relatively little progress in improving outcome (e.g. neuroblastoma), while other diagnoses still include a small percentage of aggressive and incurable tumours (e.g. Wilms tumour). Diagnostic difficulties persist and current classifications are often insufficiently sensitive to identify high-risk tumours at diagnosis. In addition, it is necessary to spare patients with favourable diagnoses from aggressive therapies which may cause severe side effects during treatment, and from the late effects of the therapy which are now recognized as an increasing problem in survivors of childhood cancer. Can the fundamental insight into the molecular machinery that drives the cancer cell be used to solve these problems?
Several major contributions can be expected in the next few years. These promise significant breakthroughs in treatment and are now moving beyond the stage of speculation. Two particular developments can be highlighted. First, many proteins that cause cancer have been identified. They are mutated forms of normal cellular proteins and there is an opportunity to develop drugs that specifically block these oncogenic proteins. Secondly, new technologies can now analyse the level of expression of all genes in a tumour. This permits the construction of a blueprint for the genetic constitution of a large series of tumours and allows the identification of gene expression patterns that may mark the different clinical and biologic subtypes of a tumour. This approach should facilitate tumour stratification and ultimately offer opportunities for more individualized therapy. Despite such optimism, some caution is also justified. Cancer cells are renowned for their plasticity and their ability to escape from treatment strategies; this may limit the realization of opportunities offered by new technology.
Cancer proteins: general principles
Hundreds of genes and proteins play an essential role in cancer pathogenesis. They contribute to cancer when they are either too active or too inactive. Oncogenes contribute to cancer when they are overactive. This arises, for example, by activating mutations or by mechanisms of overexpression. Tumour suppressor genes normally function to control cell division and homeostasis. When these genes are inactivated, for example by chromosomal defects or mutations, cells can become cancerous. The pathogenesis of cancer is generally considered as a multistep process in which defects in several oncogenes and/or tumour suppressor genes combine to result in uncontrolled cell growth. The oncogenes and tumour suppressor genes so far identified can be categorized in to several main groups according to their function in the cell. Most cancer cells have defects in genes in several functional categories, which include the following: loss of cell cycle control; signal transduction defects; DNA stability control defects; apoptosis defects; metastatic capacity.
Loss of cell cycle control
The most central part of a cancer cell is the machinery that drives cell division. During the G1 phase, the cell grows and prepares for division. The G1 phase is followed by the S phase, in which replication of the DNA occurs. The next phase of the cell cycle is the G2 phase, in which DNA replication is completed and the cell prepares for division, which proceeds in the M phase. After the M phase, cells can either continue cycling by starting a new G1 phase or enter the resting phase (G0 phase). The cell cycle is a tightly controlled process, driven by a series of proteins called cyclins and cyclin-dependent kinases (CdKs). When complexed with the appropriate cyclin, CdKs can phosphorylate the Rb protein, which results in the functional activation of proteins of the E2F family and the subsequent activation of the gene transcription program necessary for cell division. The cyclins and CdKs that promote cell division are controlled by a series of small inhibitory proteins called cyclin-dependent kinase inhibitors (CKIs).
Most cancer cells have defective cell cycle machinery. For instance, retinoblastoma is characterized by homozygous defects in the Rb gene (the first tumour suppressor gene that was identified) and the Rb gene is mutated in patients with familial retinoblastoma. Other examples of defects in cell cycle genes include gene amplification of cyclins (e.g. cyclin D1 in breast cancer and, sporadically, in neuroblastoma).
Signal transduction defects
Hundreds of genes in the cell play a role in signal transduction. This is the process by which extracellular signals are received by receptors on the cell membrane and transmitted to the nucleus where they induce transcription programs. Signal transduction plays a role in hundreds of physiologic processes, including regulation of differentiation and cell division. Examples of components of the signal transduction cascade that can play a role in cancer are membrane receptors for extracellular growth factors. These can be mutated in a way which permanently activates the downstream cascade even when no growth factor is bound to the membrane receptors. A well-studied signal transduction pathway that is activated in many tumours is the Wnt–APC pathway (Figure 3.1). This is best understood for its role in colon cancer, but the pathway is also activated in hepatoblastoma. In normal colon, the mucosal crypts harbour rapidly dividing epithelial cells that migrate apically to the villi. Epithelial cells of the villi do not divide, and ultimately die. The Wnt–APC pathway controls the transition of the rapidly dividing epithelial cells in the crypts to the differentiated non-dividing cells on the villi. The Wnt growth factor is locally produced in the crypts and can activate receptors on the surface of the epithelial cells. As the cells migrate away from the crypts towards the villi, they lose the Wnt signal and start to express p21, resulting in blockage of the cell cycle. Most colon tumours have inactivating mutations in the APC gene which result in the inappropriate release of β-catenin, activation of c-myc and promotion of uncontrolled cell division in the absence of the Wnt signal.
|
|
|
Fig. 3.1 The Wnt–APC pathway controls the transition of dividing cells to non-dividing cells in several tissue types (e.g. colon epithelium). Wnt is an extracellular protein, and when it binds to membrane receptors a signal is transmitted to a complex consisting of the proteins GSK3-β and APC, inactivating the complex and releasing the protein β-catenin (β-CAT). β-Catenin enters the nucleus where it forms a complex with the TCF transcription factor and induces transcription of the c-myc gene. In turn, c-myc transcription silences the gene for the protein p21. As p21 normally blocks the cell cycle, the final result of Wnt induction is activation of the cell cycle. |
This illustration is an oversimplification as the Wnt–APC pathway involves many more genes, and mutations can arise elsewhere in the pathway. For instance, β-catenin is frequently mutated in hepatoblastoma and, sporadically, in Wilms tumour and medulloblastoma. These mutations prevent the normal breakdown of the β-catenin protein, and accumulation of the protein results in the activation of the cell cycle. Many other signal transduction routes can ultimately activate the cell cycle, but an understanding of the Wnt pathway illustrates how defects in signal transduction can promote uncontrolled cellular growth.
DNA stability control defects
Hundreds of genes in a cell are dedicated to the faithful preservation of the genetic information stored in the DNA. The billions of cells in the body are under the permanent influence of mutagenic events (e.g. sunlight, endogenously produced oxygen radicals, or genotoxic compounds from the environment). Mutations in DNA are detected by a series of specialized proteins which have a direct link to several of the fundamental processes in the cell. When DNA damage is limited, these proteins can stop cell division so that the cell has time to repair defects before they are replicated and irreversibly fixed. In parallel with the signal sent to the cell cycle machinery, this process recruits a series of repair proteins that restore the original DNA sequence. However, when the DNA damage is beyond a certain threshold, the system considers the cell to be beyond repair and drives it into apoptosis (see below). Many cancer cells have defects in the genes that mediate the repair of damaged DNA. Examples are the MSH1 and MSH2 genes that are defective in many colon cancers. When such genes become mutated in a single somatic cell, the cell can accumulate hundreds of other mutations. Some of these mutations may affect oncogenes and/or tumour suppressor genes, which can lead to cancer. Another gene with a central role in repair regulation is p53. The p53 protein is the intermediate between the proteins that identify the DNA defects and the switch that decides whether the affected cells should stop dividing or go into apoptosis. Inactivating mutations of the p53 gene are among the most prevalent defects in tumours and, when present, damage is no longer adequately dealt with, as the cell cycle cannot be stopped.
Apoptosis defects
Each cell has an intrinsic machinery to allow it to commit suicide (‘programmed cell death’). The machinery can be triggered by many different stimuli. These stimuli can be physiologic signals, for example in neuronal cells during embryogenesis, or by an attack by T cells that activates the intrinsic apoptotic machinery in virally infected cells. Cells with serious DNA damage or which have aberrant expression of oncogenes are also driven into apoptosis. In general, it is believed that inappropriate combinations of growth-stimulating signals trigger apoptosis as a surveillance mechanism against cancer. Therefore it is not surprising that many tumour cells display defects in their apoptotic machinery. If these are defects that prevent the death of aberrant cells, this will result in their continued proliferation and outgrowth to tumours.
Two main routes to apoptosis have been identified. There is an extrinsic route, activated by extracellular molecules like interferon-γ or TRAIL. These molecules bind to so-called ‘death-receptors’ which transduce the apopotic signal to the caspase 8 protein. Activated caspase 8 can subsequently activate the so-called ‘executioner caspases’ which damage cellular proteins and ultimately destroy the chromosomes by fragmenting the DNA. The cell decomposes in a controlled way and is cleaned up by macrophages. There is also an intrinsic route to apoptosis. This route is activated by, for example, extensive DNA damage or inappropriate oncogene activation. Mitochondria play a key role in the activation of the intrinsic route. Pro-apoptotic signals trigger the release of the mitochondrial protein cytochrome c. Outside the mitochondrium, cytochrome c forms a complex with other proteins that triggers the activation of caspase 9. Caspase 9 also activates the executioner caspases, promoting a final process similar to that achieved by the extrinsic route.
Several genes functioning in the apoptosis pathway can be defective in tumours. The bcl-2 gene controls the release of cytochrome c from mitochondria in the intrinsic apoptotic pathway. In B-cell lymphoma, the bcl-2 gene is permanently activated by a chromosomal translocation (t8;14) and thereby inhibits the release of cytochrome c, blocking the subsequent path to apoptosis. The bcl-2 gene is a member of a larger g family of genes that all control cytochrome c release, and many tumours show aberrant expression of this family. The extrinsic route to apoptosis can also be inactivated in tumours. For instance, aggressive neuroblastomas may lack expression of caspase 8 with an inability to progress the apoptotic pathway.
Metastatic capacity
One of the hallmarks of malignant tumours is their metastatic capacity. Until recently, it was assumed that clonal evolution of primary tumours led to subclones with a metastatic capacity. In other words, metastasis was considered as a discrete step in malignant development, resulting from altered gene expression or new mutations. Many genes have been identified that affect the metastatic capacity of tumour cells in experimental settings. However, in human tumours, hardly any mutations have been identified in genes that primarily function to confer a metastatic phenotype. Recent experiments have challenged the concept of clonal selection in the primary tumour and progressive acquisition of metastatic capacity by subclones. This implies that the combination of genes that causes the primary tumour also determines whether the tumour is metastatic or not, and that metastatic capacity may not represent a discrete step in malignant progression.
Is there a fundamental difference between paediatric and adult cancer?
It is an open question whether paediatric tumours have an essentially different mechanism of pathogenesis and progression from adult tumours. The emerging picture is that one or more characteristic genetic aberrations are found in each tumour type, but that they represent variations on the same theme. For instance, although neuroblastoma and medulloblastoma can have amplification of the N-myc oncogene, adult small-cell lung tumours can also show the same aberration. The N-myc gene is member of a small gene family, which also includes the c-myc and L-myc genes. The N-myc and c-myc genes have very similar functions. The c-myc gene can also be amplified in small–cell lung cancer, and is also amplified or rearranged in numerous other tumours, including leukaemia. Therefore activation of the myc genes cannot be seen as an exclusive property of paediatric tumours. The same holds for activation of cell cycle genes. The cyclin D1 oncogene is amplified or overexpressed in neuroblastoma, and also in several adult tumours. There are certainly aberrations that are specific for some paediatric tumours, like the EWS–FLI translocations in Ewing sarcoma and the PAX–forkhead translocations in rhabdomyosarcoma. However, certain adult tumours also show specific genetic aberrations, and the existence of specific abnormalities in paediatric tumours does not indicate a special biologic characteristic of childhood cancer.
A clear-cut biologic characteristic of paediatric tumours is their peak incidence at an early age and their disappearance after this age. With some exceptions, the adult tumours are of other tissue or cell types, and show an age-related increase in incidence. However, this does not necessarily imply a different biology: both types of tumour stem from rapidly dividing tissues, and the rapid cell division during embryogenesis and infancy gives a window of opportunity for tumours, just as the cumulative number of cell divisions in tissues with a rapid turnover may do in adults (e.g. affecting the epithelial cells of colon, breast, and lung).
Another potential difference between paediatric and adult tumours could lie in aberrations of control genes for embryonal development. Studies of Drosophila embryogenesis have identified many genes controlling the differentiation of tissues and organs. These genes appear to be faithfully conserved in humans and many appear to be mutated in human tumours. So far, there is no specific evidence for a prevalence of such genes to be mutated in paediatric cancer, although suggesting that these genes play a role not only in early embryogenesis, but also in the control of cell division and differentiation in adult tissues. For instance, the Wnt–APC pathway is involved not only in the differentiation of early embryonal cell lineages but also in colon epithelium in adults. Accordingly, mutations in the Wnt–APC pathway are found in some paediatric tumours and in some adult tumours.
In conclusion, it is difficult to describe any specific molecular biology of paediatric tumours. All tumour types, adult and paediatric, clearly show genetic aberrations that cause disturbances in a series of major cellular pathways, such as those controlling the cell cycle, apoptosis, signal transduction, and DNA stability. The encouraging consequence is that new drugs developed to specifically inhibit proteins in these pathways may work just as well in tumours at all ages.
Towards protein-specific drugs
The identification of many of the proteins that cause, or contribute to, the malignant transformation of cells has started a huge effort to identify molecules that might specifically inhibit these proteins. Most promising are the so called ‘small molecule’ drugs. While other approaches like gene therapy, immune therapy or antisense therapy are still experimental, the first small molecules that specifically inhibit the protein products of oncogenes have already successfully entered the therapeutic arena.1 The best example is imatinib mesylate (STI571, Gleevec™), a small molecule that specifically blocks the activity of the Abelson protein, a tyrosine kinase. The abl gene is translocated and activated by the translocation of chromosomes 9 and 22 in chronic myeloid leukaemia (CML) and also in a subset of childhood acute lymphoblastic leukaemia (ALL). Gleevec was developed as a specific inhibitor of the tyrosine kinase activity of the Abelson protein. It is clinically highly successful in the treatment of CML, and trials on high-risk childhood Ph+ ALL are ongoing (see Chapter 5). Unfortunately, a small percentage of patients develop tumour resistance to Gleevec; the Abelson protein is mutated in some of these resistant tumours, and Gleevec is no longer able to inhibit the mutated form. Another form of resistance may result from de novo amplification of the abl gene in relapsed tumours. Although Gleevec inhibits the tyrosine kinase activity of the Abelson protein, it has also been found to inhibit some related tyrosine kinases, for example the c-Kit tyrosine kinase that plays a central role in gastrointestinal stromal cell tumour (GIST). As tyrosine kinases are members of a large gene family and function in many different cellular processes, Gleevec is currently being tested in a range of tumours, and other drugs are being developed to act against other cell cycle proteins and growth factor receptors.
The development of these new types of cancer drugs is very expensive and it is inevitable that pharmaceutical companies will focus their efforts on the development of drugs for the major types of adult cancer. However, some of these drugs may also hold promise for paediatric tumours, as many of the pathways initially identified as playing a role in one specific cancer are later found to be involved in other tumour types as well. Examples include the Wnt–APC pathway, originally identified in colon cancers but also activated in hepatoblastoma, medullo-blastoma and Wilms tumour, and the N-myconcogene, which was originally identified as an amplified oncogene in neuroblastoma but is also amplified in small-cell lung carcinoma. Therefore it will be very important to test drugs developed for adult use in preclinical models of paediatric tumours and in subsequent clinical studies.
It is also possible that some oncogenes are activated only in one paediatric tumour type, but although this might represent a perfect target for a drug, the relatively small number of patients with a specific paediatric tumour is unlikely to encourage efforts to develop drugs for such targets. However, the lesson of Gleevec is that a drug originally developed against one specific protein in one specific tumour may ultimately have a much wider therapeutic spectrum, and on this basis it would be economically more interesting to develop novel drugs directed against proteins known only to cause cancer in children.
How complete is our understanding of oncogenesis?
The functional categories of genes involved in cancer, described above, are relatively imprecise. Hundreds of oncogenes and tumour suppressor genes have been identified and, although most of them can be tentatively placed in one of these functional categories, the complexity of the mechanisms involved precludes the precise characterization of many of them. Each cell type has its own developmental program which is modulated by cues from surrounding tissues. Therefore each tissue can have its specific receptors for these cues and a specific signal transduction cascade that activates or blocks the cell cycle. Thus it is likely that many different signal transduction routes converge on the cell cycle. As a consequence, defects may exist in many different routes, all with the same effect—uncontrolled cell division.
Although many of the basic principles of oncogenesis have been identified in the past 25 years, major elements have probably escaped detection. For example, in neuroblastoma, only the N-myc gene (and, to a lesser extent, caspase 8) has been identified as a major mechanism, yet amplification and overexpression of N-myc is only found in 20 per cent of neuroblastomas. In the remaining 80 per cent of neuroblastomas, not a single gene has been identified that causes tumorigenesis. However, many structural chromosomal defects have been identified in neuroblastoma, for example extra copies of the long arm of chromosome 17 (17q+) and deletions of long stretches of one copy of chromosome 1(1p del) or chromosomes 4, 11, and 14. This suggests that these chromosomal regions could harbour important genes which might contribute to pathogenesis. Although several candidates for such genes have been proposed, clear identification is awaited. The picture described for neuroblastoma holds for many other childhood tumours. Often only one or two major abnormalities are identified, complemented with impressive series of chromosomal defects with an unknown role in oncogenesis. Much effort is still needed to gain sufficient information to understand the pathogenesis of individual tumours at the molecular level.
Microarrays: a tool for molecular classification of tumours
One of the most far-reaching investigative tools (brought about by work on the Human Genome Project) is genetic profiling by microarray technology. Microarrays can measure the mRNA expression levels of all genes in a tissue or cell line in a single experiment. A microarray consists of thousands of DNA sequences representing human genes which are arranged in a grid on a small solid surface (e.g. a microscope slide). Each array can contain up to 20 000 spots, with each spot representing one gene. Messenger RNA is isolated from a tumour, labelled with a fluorescent dye, and incubated with the array. Homologous mRNA sequences associate with their complementary DNA on the array. Therefore highly expressed mRNAs produce a fluorescent marker at the site of their corresponding gene. The array is scanned with a laser to measure and map the level and distribution of fluorescence, indicating the activity of genes within the tumour. In this way, one experiment, taking only a few days, can assess the expression of 20 000–40 000 human genes.
The application of microarray technology to large series of human tumours describes the gene expression profile of each tumour type in great detail. Sophisticated software can search for patterns in the expression profiles and relate them to the biologic or clinical characteristics of the tumours. For example, studies of a large series of leukaemias have been able to identify expression profiles that discriminate between T-ALL, B-ALL and acute myeloid leukaemia (AML).2 In other examples, a series of ‘small round cell blue tumours’ reveal expression profiles which discriminate between neuroblastoma, rhabdomyosarcoma, Ewing sarcoma, and Burkitt lymphoma.3 These experiments reflect the old clinical insight that tumours of a specific tissue lineage share certain essential characteristics. More important for the clinician is the possibility that this technology could identify prognostic subgroups within a single tumour type. When the clinical course of a series of, for example, 100 tumours analysed by microarray is known, statistical analyses can identify gene sets that may together predict the prognosis of individual patients. As hundreds of genes can be included in these prognostic profiles, the results may be more discriminating than prognosis based on classical analyses. For example, microarray evaluation of a series of medulloblastomas can reliably discriminate between the desmoplastic and classical subtypes and, more importantly, can also identify a gene expression pattern that predicts survival.4 A similar analysis of a series of leukaemias showed that ALLs with a translocation of the MLL gene, which is known to be a poor prognostic marker, have a gene expression pattern that discriminates them from other ALLs and AMLs and suggests a resemblance to an early haematopoietic progenitor.5 These tumours were found to have a high expression of the FLT3 gene, a receptor tyrosine kinase. This finding has also opened therapeutic opportunities, as a novel small molecule drug (PKC412) that inhibits the FLT3 kinase activity and could inhibit cell growth was active in a mouse model of MLL-activated tumours.6
Microarrays have also been used to identify a set of genes that predict the metastatic potential of tumours. The results of such experiments imply that the gene expression profile of a primary tumour may predict its later metastatic spread (see above), suggesting that metastatic cells do not result from a progressive selection of more aggressive subclones but are similar to the primary tumour cells. Metastatic potential may then be an intrinsic property of the primary tumour. Clinically, this implies that it may become possible to assess the risk of metastasis at first diagnosis.7
The predictive power of many microarray analyses should improve with further enhancement at both a technical level and the level of statistical analysis of the complex data generated.
Towards a tailored therapy for each patient: how far in the future?
Microarray technology shows great promise in the prediction of response to therapy. By analysing the expression profiles of a large tumour series, it should be possible to establish whether tumours that respond to a specific therapy have a different expression profile from that seen in non-responding tumours. This could ultimately improve the selection of treatment for individual patients. Array technology will be even more promising when combined with treatment protocols which include innovative target drugs that inhibit specific proteins, such as Gleevec. As these drugs are directed against only one or a few proteins, it should be possible to use the gene expression profile of the tumour from an individual patient to predict, on the basis of gene expression, whether the tumour is likely to respond to treatment with the specific drug. Therefore the availability of rapid screening systems for gene expression profiles of individual tumours, combined with the development of series of novel drugs that can specifically inhibit oncoproteins, raises the hope that individualized therapies may become available in the not too distant future.
References
1. Smith JK, Mamoon NM, Duhe RJ (2004). Emerging roles of targeted small molecule protein-tyrosine kinase inhibitors in cancer therapy. Oncol Res 14, 174–225.
2. Golub TR, Slonim DK, Tamayo P, et al. (1999). Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 286, 531–7.
3. Khan J, Wei JS, Ringner M, et al. (2001). Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks. Nat Med 7, 673–9.
4. Pomeroy SL, Tamayo P, Gaasenbeek M, et al. (2002). Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 415, 436–42.
5. Armstrong SA, Staunton JE, Silverman LB, et al. (2002). MLL translocations specify a distinct gene expression profile that distinguishes a unique leukaemia. Nat Genet 30, 41–7.
6. Armstrong SA, Kung AL, Mabon ME, et al. (2003). Inhibition of FLT3 in MLL. Validation of a therapeutic target identified by gene expression based classification. Cancer Cell 3, 173–83.
7. van't Veer LJ, Dai H, van de Vijver MJ, et al. (2002). Gene expression profiling predicts clinical outcome of breast cancer. Nature 415, 530–6.
