Mouse Models of Leukemia: Pros and Cons

Model systems aim to provide robust platforms for investigating the basic genetic and biochemical components of malignant behavior. Furthermore, mice potentially can be used in preclinical evaluation of novel therapies.^1,2  A common theme is the ability to perform controlled experiments that are difficult or impossible in humans. Unlike patients, mice can be designed to have both a defined genotype and congenic siblings that serve as controls.

Based on the extensive molecular understanding of human leukemias, many of these diseases have now been modeled in mice. Several distinct methods can be used to introduce oncogenic mutations into the murine hematopoietic system.³  In conventional transgenic models, an oncogene is integrated at a random site in the genome. While these systems have proven quite valuable, they suffer from poor control over oncogene copy number and expression pattern due to integration effects. Conditional gene targeting addresses these concerns by modifying the endogenous locus of a proto-oncogene or tumor suppressor gene and allows a mutation be induced at a specific time and/or lineage. To date, this approach provides the most accurate genetic model of oncogenic mutations. Finally, retroviral transduction is a rapid method for generating series of genetically related leukemias. In some cases, proviral insertion can be deliberately exploited to generate leukemias by insertion mutagenesis.⁴ 

Genetic diversity may be regulated: Acute leukemia in humans or mice involves multiple cooperating mutations, but most mouse models are designed with only one mutation in the germline. Therefore, additional genetic changes take place during leukemic transformation. These secondary events occur at random, leading to some genetic diversity among the tumors. Retroviral insertion screens have been a highly successful approach for systematically defining such cooperating lesions. Alternatively, multiple oncogenic mutations can be designed into the model by crossbreeding or other strategies. Leukemia development in this setting will require fewer random mutations, and the resulting tumors will be more similar to one another.⁵  In essence, by varying the number of engineered oncogenic mutations, the investigator may control the degree of genetic diversity in the system.

Well-established cell biology assays: Decades of research into basic mechanisms of hematopoiesis have revealed extensive similarities between human and murine hematopoiesis and have yielded a large set of techniques to assess cell biology in normal and diseased states. As a result, the effects of an oncogenic mutation on cell fate can be determined with some clarity. For example, traditional colony-formation assays readily demonstrate the enhanced self-renewal imparted by some oncogenic transcription factors and the hypersensitivity of myeloid progenitors to GM-CSF in models of human myeloproliferative disease.^6,7  In one example of applying a classic cell biology assay to leukemia therapeutics, the therapeutic index of imatinib mesylate was predicted by its differential effect on myeloid progenitor colonies grown from CML but not normal bone marrow.⁸ 

Primary tumor cells are accessible for analysis: Once established, murine malignancies can be harvested and subjected to biochemical or genomic analysis. The high proportion of malignant cells in target organs results in nearly pure populations for study, although subfractionation can be performed if desired. Tumors may be queried for secondary genetic mutations, proliferative activity, and, importantly, the inhibition of the molecular target of therapeutic agents in the target cell population.^9,10 

Transplantability: Hematopoietic malignancies are almost always transplantable into naïve hosts. The reproducibility of engrafted tumors facilitates performing controlled, replicated experiments to examine responses to therapeutic intervention in vivo. The system is easily scaled up, making complex comparisons of multiple strains and treatment regimens feasible.

Genetic manipulation of primary tumor cells: Retroviral transduction of primary leukemia cells can produce a series of tumors with related genetic alterations. By varying both transduced genes and the genetic background of transduced cells, genetic contributions to disease and response to therapy can be analyzed efficiently. The ability to simultaneously produce large numbers of recipients with defined mutations in the hematopoietic compartment allows such hypotheses to be tested more quickly at a fraction of the cost and complexity involved in mating mouse strains.^11,12  Recent advances in RNA interference technology have extended this approach to inhibiting target gene expression.

Clear disease endpoints: Besides survival, intermediate endpoints of clinical appearance, peripheral blood counts, and/or lymphadenopathy are robust measures of efficacy in preclinical therapeutic studies. Noninvasive imaging systems can also quantify tumor burden over time if cells are marked with appropriate reporter genes. These measures collectively establish a basis for comparing disease progression in treatment studies. The clarity of these endpoints lends considerable power to statistical analysis, allowing small trials to yield meaningful results.

The ease of quickly generating large cohorts of diseased animals, combined with the statistical power of small but uniform samples, is especially valuable when evaluating a large number of potential therapies. Therefore, mice may prove particularly helpful when novel agents are tested in combination.

Inaccuracies: There are three fundamental ways in which mice can model human disease poorly. First, the engineered genetic lesions can only approximate those found in patient samples. Some conditional models come very close, but most are imperfect in some way. A second type of inaccuracy results from the inability to ensure that a single cell of the proper type undergoes the initiating oncogenic mutation. In most systems, the mutation occurs in a large "field" of genetically identical cells, failing to model the important process of clonal selection. Finally, whereas many mouse cancer models rely upon engineering a specific mutation into the germline, it is not always certain that these mutations represent bona fide initiating events in human hematologic cancers. Discordant results between mouse preclinical studies and human trials may reflect the fact that a specific mutation that initiates leukemogenesis in the mouse may not play the same role in human disease.

Pharmacology: A second concern relates to the mouse host, regardless of how accurately the tumor is modeled. Potential differences in pharmacology or toxicology can affect the interpretation of therapy trials, which depend on finding anti-tumor effect at a nontoxic dose. A notorious example involves investigations of camptothecins, which have vastly different pharmacology in mice and humans.¹³ 

Alternatives: Some translational researchers prefer xenograft models, in which human tumors are engrafted into mice. These have one potential advantage — they use actual human cancer cells. However, these systems lack the power of genetic systems applicable to mice and suffer from the same potential pharmacologic problems. Lastly, immortalized cell lines have revealed important facets of biochemistry and molecular biology, but are less useful for discovering more subtle and complex effects that are restricted to primary cells. Studies of drug sensitivity in cell lines are relatively fast, cheap, and easy, but experience has shown that they do not necessarily provide lasting value.

Will the increasing sophistication of mouse cancer models yield systems with enough predictive value and throughput to identify effective treatments for human diseases? Experience with models of acute promyelocytic leukemia (APL) is particularly encouraging. First, leukemias in PML-RARA transgenic mice were found to respond to all-trans-retinoic acid (ATRA) and arsenic trioxide (As2O3) with differentiation and apoptosis, as had already been seen in pioneering human clinical trials.^14,15  In this sense, the patients correctly predicted outcomes in the mice. Later, the model systems predicted a synergistic interaction between ATRA and As2O3, an effect only recently validated in the clinic.¹⁶ New predictions that cAMP agonists also cooperate with ATRA will soon be tested in patients.¹⁷  Perhaps the mouse models will soon replace traditional Chinese medicine as the inspiration for novel therapeutic approaches in APL and other leukemias.