Over the past several years, we have been studying cancer and cancer-related processes with a focus on in vivo approaches. Using increasingly precise methods of genetic engineering in the mouse, our laboratory has generated a series of novel strains containing germline mutations in several genes implicated in human cancer. We are using these strains (as well as strains from other laboratories) to direct single mutations or combinations of mutations to tissues and cells of interest, and we have developed a number of powerful new models of human cancer. These models resemble the human disease at both the genetic and phenotypic levels.
The development of tumor cells from normal cells requires the sequential acquisition of mutations in several cellular genes. In general, two classes of genes are affected in tumor progression: genes that normally act to promote cell division (oncogenes) and genes that function to arrest or inhibit cell division (tumor-suppressor genes). Human familial cancer syndromes (in which affected individuals have a greatly increased risk of developing particular types of cancer) are often caused by the inheritance of a mutant allele of a tumor-suppressor gene or an activated allele of an oncogene. Tumor suppressors are thought to regulate cell growth negatively, and they contribute to carcinogenesis when mutated or lost. Thus, individuals who carry only one functional copy of a given tumor-suppressor gene are predisposed to cancer, because all of their cells are just one mutational event from lacking an important negative growth regulator.
Many tumors of epithelial origin in humans acquire mutations in the K-ras oncogene, including approximately 30 percent of non-small-cell lung cancer (NSCLC), 50 percent of colon cancers, and 90 percent of pancreatic cancers. Our laboratory has been investigating the effects of K-ras mutation in the mouse. To control timing and tissue-specific activation of K-ras, we have developed a conditional, activatable allele of the K-ras oncogene and have used it to generate various mouse models of human cancer. In the lung, activation of oncogenic K-ras leads to the development of atypical adenomatous hyperplasia, adenomas, and adenocarcinomas. Combining mutations in K-ras and p53 in the lung led to the development of more advanced tumors, which exhibited desmoplastic stroma, increased invasiveness, and metastatic potential.
With advanced gene-targeting methods, generating mouse models of cancer that accurately reproduce the genetic alterations present in human tumors is now relatively straightforward. The challenge is to determine to what extent such models mimic human disease with respect to the underlying molecular mechanisms that accompany tumor progression. Working with the Todd Golub's lab (HHMI, Dana-Farber Cancer Institute), we have developed a method that uses gene-expression profiling to compare mouse models of cancer with human tumors. We applied this method to the analysis of our model of K-ras–mediated lung cancer and found a good relationship to human lung adenocarcinoma, thereby validating the model. Furthermore, we found that although a gene-expression signature of KRAS activation was not identifiable when analyzing human tumors with known KRAS mutation status alone, integrating mouse and human data uncovered a gene-expression signature of KRAS mutation in human lung cancer. The K-ras lung cancer model has also been used in collaboration with the Golub lab to perform microRNA-profiling experiments. The data from the model are consistent with human data, demonstrating a general down regulation of miRNA expression in tumors compared to normal tissue.
Through analysis of the expression-profiling data, we identified cathepsin cysteine proteases as being highly up-regulated in lung cancer specimens. In collaboration with Ralph Weissleder's group (Massachusetts General Hospital), we demonstrated that an optical probe activated by cathepsin proteases could detect murine lung tumors in vivo as small as 1 mm in diameter. By serially imaging the same mouse, we were able to use optical imaging to follow tumor progression. This cathepsin-imaging strategy is now being evaluated for use in intra-operative imaging during surgical resection of mouse tumors, with the hope that it could improve local control of cancer in humans.
Injury models have suggested that the lung contains anatomically and functionally distinct epithelial stem cell populations. We have isolated such a regional pulmonary stem cell population, termed bronchioalveolar stem cells (BASCs). These stem cells were enriched, propagated, and differentiated in vitro and found to be activated by the oncogenic protein K-ras. Our studies suggest that BASCs are a stem cell population that maintains the bronchiolar Clara cells and alveolar cells of the distal lung and that their transformed counterparts give rise to adenocarcinoma.
For several years, we have been attempting to develop mouse models of familial retinoblastoma (caused by a germline mutation in the RB gene) and of Li-Fraumeni syndrome (LFS; caused by a germline p53 tumor-suppressor gene mutation). Although mice heterozygous for null mutations in Rb or p53 are cancer prone, they do not exhibit the spectrum of tumors observed in the relevant human syndromes. Using more sophisticated methods in gene-targeting technology as well as compound mutant analysis, we have now created strains that do develop retinoblastoma and the tumor types seen in LFS. Rb mutation in the retina does not result in retinoblastoma formation, even in the absence of p53. However, on either a p107- or p130-deficient background, Rb mutation in the retina causes retinal dysplasia and retinoblastoma. This represents the first breedable model of retinoblastoma. We have recently performed DNA copy-number analysis on these retinoblastomas and observed genomic amplification of the N-myc oncogene. N-myc amplification also occurs in a subset of human retinoblastomas. Finally, we have also studied the role of pRb, p107, and p130 in the regulation of cell cycle progression and terminal differentiation in various tissues of the mouse.
The majority of human tumors have mutations in the p53 gene. The development of a more accurate model of LFS required the construction of knockin strains carrying missense mutations of p53 (as opposed to the original null mutations). These p53 point mutants are thought to act as dominant-negative proteins and have been proposed to have dominant gain-of-function effects as well. We demonstrated that mice heterozygous for either of two p53 point mutations developed a broader spectrum of tumors than mice heterozygous for a p53-null allele. This included the development of carcinomas, which are frequent in LFS and rare in the p53+/– mouse. By comparing tumor phenotype in p53R172H/– or p53R270H/– mice with p53–/– mice, we also addressed possible gain-of-function effects. The point mutant alleles promoted the development of carcinomas, which do not occur in p53–/– mice. Thus, we provided the first support in vivo for a gain-of-function of tumor-associated point mutations in p53.
We have also been studying whether the reactivation of p53 could have therapeutic effect in established tumors using a Cre-lox–based approach. We have observed dramatic response to p53 restoration in nearly all tumors tested to date. The response to the reactivation of p53 differed between tumor types, with lymphomas undergoing apoptosis and sarcomas undergoing cell cycle arrest with features of senescence. This work indicates that the signaling pathways that impinge on p53 remain active in established tumors and provides support for efforts to activate this pathway in human cancer therapy.
Finally, we have continued to develop additional mouse tumor models, including several representing major human cancer types for which good preclinical models have been lacking. These include pancreatic cancer, ovarian cancer, soft tissue sarcoma, and invasive colon cancer.