(415) 502-0229 (asst)
(415) 502-4192 (office)
(415) 502-6791 (lab)
(415) 502-3179 (fax)
1450 3rd St., MC 0128; PO Box 589001
San Francisco, CA 94158-9001
deliveries: 1450 3rd Street, HD-320; San Francisco, CA 94158
The group's main research interests have been the elucidation of the molecular mechanisms of multistage carcinogenesis, with particular emphasis on mouse models of chemically induced skin tumor development. Our strategy has been to identify the sequence of somatic genetic alterations which are associated with discrete stages of tumorigenesis: initiation, promotion, progression to locally invasive lesions and development of metastases. In each case where a genetic change has been identified, we have attempted to address the question of causality by making extensive use of transgenic or knock-out mice, and to investigate the biological consequences of this genetic change for cell behaviour during transformation. We have developed cell lines from tumors representing each of these stages of carcinogenesis, and in some cases have characterized cells from successive stages of development from the same tumor. These have proved invaluable for studies of the causal genetic and biological changes associated with tumor progression. The results have shown that the kinds of genetic and biological alterations seen in mouse tumors are very similar to what is observed in human malignancies, thus providing strong evidence that the mouse is a uniquely appropriate model for the development of cancer in humans. More recently, a major emphasis has been on the study of genetic predisposition and the relationships between germline predisposition and tumor suppressor genes. We are presently extending our studies on skin carcinogenesis to other mouse model systems for tumor development in the lung, prostate and lymphoid system. The goal for the next few years is to exploit mouse model systems for the identification of genes that are important for cancer susceptibility or cancer progression in human populations, as a route to the discovery of novel diagnostics, new possibilities for prevention, or therapeutics.
This summary will focus on: 1) Genetic modifiers of cancer susceptibility; 2) Development of transgenic and knockout models for multistage tumor development; 3) Genetic alterations and biological changes during tumor invasion and metastasis
1. Identification of Genetic Modifiers of Cancer Susceptibility
Epidemiological evidence has shown that the major causes of cancer include environmental agents such as radiation, or chemicals of natural or synthetic origin The particular genetic background of each individual is also extremely important in determining his or her overall cancer susceptibility. One person in three in an average lifespan develops cancer, but two out of three do not. Certain individuals may therefore be predisposed to develop tumors, by virtue of their particular genetic background, while others appear to be more resistant. These susceptibility genes may control intrinsic cellular growth functions, or may influence the ways in which environmental carcinogens interact with and cause mutations in target cells. The isolation of such genes is one of the major goals of human cancer research worldwide, since they have the potential not only for use in identification of high risk individuals, but for development of novel strategies for prevention or therapy.
One of the major challenges facing the pharmaceutical industry is the identification of appropriate genetic "targets" for drug discovery. The human genome project will in the foreseeable future provide us with a database of many thousands of genes, but the validation of specific genes as candidates for drug screening is an enormous task, particularly in the case of multigenic diseases such as cancer. Our research program is concentrated on the identification of germline susceptibility genes for cancer, which are by definition pre-validated targets since we know, even without having cloned these genes, that they affect the incidence or progression of human cancers.
Such genes will be difficult to find from studies on human families, since they do not segregate as single mendelian traits which can be easily mapped. However, the mouse is ideally suited to studies of polygenic traits such as diabetes, obesity, osteoporosis, and in particular cancer. In addition, since these tumor modifier genes presumably control multiple facets of tumor development such as tumor incidence, growth rate, capacity for angiogenesis or ability to metastasize, their identification could provide us with new opportunities to intervene at specific stages of tumor formation in humans. Our goal is therefore to carry out a comprehensive survey of the numbers and locations of tumor modifier genes using several of the best characterized models of carcinogenesis in the skin, lung, prostate, and lymphoid system.
In previous work on skin carcinogenesis, we have mapped at least ten independent tumor resistance loci in the spretus genome. Some of these affect the early development of tumors, while others influence primarily tumor progression. A specific subset of three of these modifier loci dramatically prolongs the survival of tumor-bearing mice, suggesting that they modify either tumor growth rate or the ability to invade and metastasize. Candidate genes have been identified for several of these loci and further studies on their identification are in progress. Other groups have mapped susceptibility genes for lung, colon or liver cancer to the same genomic loci, thus raising the possibility that some of these modifier genes may have pleiotropic effects on multiple tumor systems. At present, a total of around 80 tumor susceptibility loci have been mapped in the mouse genome, and it is expected that further loci will be uncovered as the mapping proceeds.
At present, it is not feasible to carry out whole genome scans to find polymorphisms in linkage disequilibrium with multiple cancer phenotypes in human populations. Candidate regions of the genome, and candidate genes therein will therefore chosen on the basis of results of mouse modifier studies for testing in human cancer patients and control groups. Candidate genes will be sequenced to find polymorphisms in the human population, and the frequencies of these polymorphic variants in affected sibs vs unaffected control will give an indication their relevance to susceptibility to the different cancer types. Particular attention will be paid to modifiers that appear on the basis of mouse studies to be common among different mouse strain combinations, and/or to be important for several tumor types. Depending on the accuracy with which the modifiers are mapped, it may be possible to carry out these preliminary studies using polymorphisms in the syntenic regions of the human genome, without necessarily identifying the critical gene. We have therefore initiated collaborations with groups that have access to considerable numbers of samples of both constitutional and tumor DNA from human breast, ovarian, and other cancers, together with a strong interest in (and experience of) the analysis of the effects of low penetrance genes on human cancer susceptibility.
In conclusion, it is anticipated that a large number of prevalidated cancer targets will be identified over the next few years from studies of cancer susceptibility in mouse and human populations. Although positional cloning remains a difficult task, the total sequence of the human and mouse genomes that will be available in the near future, together with microarray technology for screening of differentially expressed genes, will obviate the necessity for much of this work. Accurate mapping of tumor modifier genes is possible now, and is a necessary prerequisite for the eventual identification of these important cancer genes.
2. Transgenic and Knock-out Models for Multistage Cancer Development
Tumours develop in both mice and humans as a consequence of sequential genetic alterations at critical genomic loci. These changes involve mutations in oncogenes such as ras, and in tumour suppressor loci, including the p53 and p16 genes. Our strategy has been to investigate the genetic alterations which take place during mouse skin tumour development, and subsequently to use both transgenic and knock-out mice to test the functions of candidate genes in vivo. Genes implicated in multistage carcinogenesis include H-ras, p53 and Transforming Growth Factor Beta (TGFbeta).
The steps involved in sequential induction of benign and malignant mouse skin tumors using initiators and promoters of carcinogenesis have been well characterized. Initiation is accomplished by a single treatment with an initiating dose of the carcinogen Dimethylbenzanthracene (DMBA) followed by multiple treatments with the tumor promoter 12-O-tetradecanoyl-phorbol-13-acetate (TPA). Initiation results in activation of the mouse cellular H-ras gene by mutation at codon 61, and promotion stimulates the outgrowth of benign papillomas. A proportion of these progress to carcinomas over a period of 6-18 months, and in some cases highly invasive spindle tumors are formed that have lost many of the characteristics of their epithelial cell precursors. Many of the genetic and biological changes that occur during this well defined model system have been identified, and are summarized in Figure 1. We have also made a substantial investment in the isolation and characterization, both genetically and biologically, of cell lines representing each of the morphologically distinguishable stages of carcinogenesis. These have proved invaluable for the dissection of the causal events involved in progression to the invasive spindle stage. In some cases (indicated by arrows on Figure 1), cell lines from different stages were derived from the same original target cell and share common genetic alterations.
Our early studies showed that the same proto-oncogene (H-ras) was consistently activated in multiple independent tumors derived from the mouse epidermis (1,2). This study also showed that quantitative changes in ras gene dosage are important during tumor progression, since the mutant gene was present at relatively low copy number in premalignant lesions, but became duplicated or amplified during tumor progression Amplification of the mutant allele and/or loss of the normal ras allele was specifically associated with a particular morphological change seen during the acquisition of metastatic properties (3,4).
Our attempts to demonstrate the causal nature of these genetic changes in tumorigenesis led us to develop transgenic models in which keratin gene promoters were used to target the expression of specific oncogenes or growth factors to subpopulations of cells within the skin. We have used promoters which target expression of transgenes to the differentiating suprabasal cells either constitutively or inducibly, or to subpopulations of basal cells. These studies have given us some important insights into the biological consequences of mutant ras expression in vivo. The use of a keratin 10 gene promoter, which targets predominantly suprabasal cells, led to the surprising result that the main consequence of ras expression in vivo was induction of a terminal differentiation phenotype. The transgenic animals showed evidence of increased keratinisation without hyperplasia, but subsequently developed focal areas of hyperplasia which progressed to form benign papillomas (5). These lesions however were terminally benign, and failed to give rise to malignancies in the lifetime of the animals. Expression of the same mutant ras transgene in a sub-population of hair follicle cells, which contain the putative stem cells, led to the development of tumors which progressed to malignancy, indicating the importance of the target cell as a major determinant of malignant potential (6).
We have also extensively studied the functions of the p53 tumor suppressor gene in responses to DNA damage and in carcinogenesis in vivo (7,8). This gene is mutated in about 30% of malignant skin tumors, and experiments with knock-out mice have demonstrated that animals deficient in p53 function show increased progression of benign tumors to malignancy. However, the same animals had a decreased incidence of benign tumors, contrary to expectations (7). Hence, although these results illustrated the causal nature of p53 loss in tumor progression, they also highlighted another surprising function for p53 in allowing the outgrowth of early benign tumor cells. Progression to the undifferentiated spindle stage of carcinogenesis is associated with further changes involving amplification of mutant ras together with deletions at the Ink4a locus, suggesting that concomitant alterations of p53 and Rb pathways are involved in progression to invasion and metastasis (9).
3. Genetic Alterations and Biological Changes During Tumor Invasion and Metastasis
Our studies on the biology of mouse skin tumor invasion have demonstrated the existence of at least two pathways leading to the acquisition of invasive properties: a) genetic alterations involving the p16/Ink4a locus and b) overexpression and activation of TGFbeta signaling.
a) A Genetic Basis for Tumor Invasion and Metastasis
The isolation of cell lines representing discrete stages of tumor progression from the same primary tumor has enabled us to dissect some of the critical events required for invasion and metastasis (1). Squamous carcinomas of the skin are relatively poorly invasive and metastasize infrequently. They can however progress to spindle cell carcinomas, a high risk tumor type with dramatically increased invasiveness, and greater metastatic potential (1,2,3). This squamous to spindle cell conversion is the result of EMT of the squamous tumor cells . Some spindle cell lines have acquired these properties through genetic alterations, the most frequent of which involve deletions at the p16INK4 locus together with amplification of the mutant H-ras gene (3,4,5). Others have reached the same end point through alternative genetic changes involving overexpression of the p16 gene, suggesting the presence of other mutation in the Rb pathway (but not in Rb itself, which seems to be intact in these cells). Both of these classes of genetic changes are recessive, as shown by fusion with normal cells, and indicate loss of differentiation-controlling genes during tumor progression (4). Our results suggest a novel role for p16/Ink4a in control of tumor invasion and cell adhesion rather than (or in addition to) its known functions in cell cycle control.
b) The Role of Transforming Growth Factor beta in Tumor Suppression and Tumor Progression
Mice expressing TGFbeta develop fewer papillomas than non-transgenic littermates, but have elevated rates of tumour progression. This indicates that TGFbeta can act either positively or negatively at different stages of carcinogenesis. We have proposed that TGFbeta can act as a negative regulator of cell cycle progression during the early phase of tumorigenesis, but positively induces invasion and angiogenesis during progression to malignancy. This possibility is supported by studies involving transfection of dominant negative TGFbeta type II receptor constructs into invasive carcinoma cells. This results in suppression of the invasive phenotype and the restoration of expression of some adhesion proteins such as E-cadherin. We have shown that TGFbeta signalling can also be upregulated by genetic alterations in highly malignant spindle carcinoma cells. Many of the features characteristic of the invasive, metastatic phenotype of these cells can be reversed by the introduction of dominant negative interfering mutants that disrupt Smad signalling. Thresholds of Smad activity are important at distinct stages of tumor progression, cooperating with activated ras to induce invasion or dissemination to distant sites.