The Mouse Tumor Biology Database: A Public Resource for Cancer Genetics and Pathology of the Mouse¹

Mice have been used in cancer research for nearly a century (1, 2). Many of the first lines of genetically homogeneous mice were produced in the 1920s for cancer research specifically. For instance, MacDowell and Taylor (3) used a selective inbreeding strategy to establish a strain of mice, C58, that displayed a 90% incidence of leukemia. Today, there are hundreds of inbred strains that carry distinct combinations of alleles in a fixed state and vary greatly in the types of tumors they develop. For instance, AKR/J mice characteristically present with lymphocytic leukemia (4), whereas a related strain, A/J (5), shows a high frequency of lung tumors (6, 7). Inbred mouse strains represent unique genetic resources for the identification of alleles or quantitative trait loci that predispose to oncogenesis. For example, nine colon cancer susceptibility loci, Scc1 through Scc9, that are distributed throughout the mouse genome have been mapped in crosses between tumor-resistant BALB/cHeA mice and the susceptible strain, STS/A (8, 9). Inbred strains are also the raw material for the generation of genetically engineered mice (transgenics and gene-targeted mutants) that have become indispensable tools for cancer research. They have already added much to our knowledge of the effects of specific genetic anomalies on the cellular processes related to cancer (10).

Knowing about the inherent cancer phenotype of a particular mouse strain is of utmost importance for the development of new mouse models for cancer research. However, finding this information can be difficult because of the dispersed nature of the literature and inconsistencies in the nomenclature used for describing tumors in mice and designating mouse strains. In response to an NCI initiative, MTB^{3, 4}was developed at The Jackson Laboratory as an electronic resource focusing on the genetics, epidemiology, and pathology of cancer in genetically defined strains of mice. The primary objective of MTB is to collect information regarding the effect of genetic background on tumorigenesis in mice. To accomplish this goal, MTB curators integrate published and unpublished data on tumor frequency or incidence, genetics, and pathology for different strains of laboratory mice. MTB is designed to aid researchers in determining the intrinsic cancer-related phenotypes of different mouse strains, reviewing patterns of mutations in specific tumors, and identifying genes that are commonly mutated across a spectrum of cancers in the mouse (11, 12, 13).

The main data source for MTB is the peer-reviewed scientific literature. The curation staff of the Mouse Genome Database (14) ⁵ and MTB collaboratively review >350 journals each month. In addition, electronic searches of PubMed at the National Library of Medicine (Bethesda, MD) are performed periodically to identify recent articles describing mouse cancer models. MTB also contains unpublished information obtained from extramural contributors and routine health surveillance studies of The Jackson Laboratory mouse colonies (15).

MTB focuses on providing the most comprehensive coverage possible of tumor pathobiology data associated with inbred strains of mice commonly used in cancer research and recently developed genetic mouse models of cancer. The latter category includes genetically engineered mice, chemically induced mutants (e.g., strains emerging from N-ethyl-N-nitrosourea mutagenesis projects), and spontaneous mutants. MTB does not contain information on tumor cell lines or preclinical trials using tumor xenograft models, because these types of data are available from other web-based resources (Table 1).

Priority for data entry in MTB is given to the types of cancers that are listed by the NCI as being the most prevalent in the United States population (16). As of December 2001, MTB contained 9429 entries of tumors observed in 2393 strains or cohorts⁶ of mice, representing a body of 679 publications (Table 2). Since its initial public release in 1998, use of MTB by the scientific community has increased steadily, reaching a current average of ∼450 queries/day. Users of the database include scientists from academic and private research institutions, as well as biotech and pharmaceutical companies.

The major data types in MTB include mouse strains, genes and mutations, tumor types, organs/tissues or cell types of tumor origin, pathology images, and literature citations. Each of the primary data types is represented on the MTB home page by an icon that is linked to a query form (Fig. 1). MTB supports several different query mechanisms, including browsing using controlled vocabularies and customizable query forms.

Researchers who are primarily interested in tumors of a particular organ or tissue can use the “Quick Organ/Tissue Search” option on the MTB home page (Fig. 1). This tool consists of a scrollable list of tissues or cell types for which database entries are available. A database user simply selects the name of the organ/tissue of interest to retrieve database records that match the selected term. Similarly, the “Genetics Search” option can be used to query MTB for tumors associated with mutations in a particular gene of interest. For instance, a search for Trp53 (p53) will retrieve all of the records of tumors that arose in Trp53-mutant mice and records for other mice in which somatic Trp53 mutations were detected in the tumor tissue.

An overview of the cancer spectrum characteristic of inbred mouse strains can be obtained from a graphical matrix, the Tumor Frequency⁷ Grid (Fig. 1). For ease of comparison, strain families are grouped according to their genealogy (5). Color-coded fields indicate the highest reported frequency of spontaneous tumors in a particular organ or tissue. The Tumor Frequency Grid doubles as a query tool; a mouse-click on a colored cell returns a summary table of all of the tumor records pertinent to that particular organ or tissue in the selected strain.

The “Tumor Search” form can be used to perform custom queries. As shown in Fig. 2, the user determines the specificity of the query depending on the number of fields selected from pull-down menus or parameters entered into text fields. This allows database users to cast a broad net (e.g., “show me all records of lymphoblastic leukemias”) or to ask very specific questions (see legend to Fig. 2).

To facilitate database searches, MTB uses standardized nomenclature and controlled vocabularies. The diagnostic terminology comes from standard pathology texts (17, 18, 19). Anatomical terms are based on the Mouse Anatomical Dictionary (20) that was developed for the Mouse Gene Expression Database⁸ (21). Nomenclature for mouse strains, genes, and germ-line alleles adheres to the guidelines of the International Committee on Standardized Genetic Nomenclature for Mice.⁹ MTB also supports the use of nonstandard genetic nomenclature for database queries. As shown in Fig. 2, a user interested in tumors of p19^ARF knockout mice may type the acronym “ARF” into the Strain Name field of the Strain Search form and select “targeted mutation (knockout)” from the Strain Type list. MTB will retrieve all of the records pertinent to mice carrying targeted mutations in the Cdkn2a locus that affect expression of p19^ARF.

In response to a query by a user, MTB initially returns a tabular summary of all records that match the selected query terms and constraints. The number of records displayed depends on the specificity of the query. The user can then select an individual record from the summary table and retrieve a Detail Information page (Fig. 3).

MTB also contains “negative” data. These include records of tumor-free mice that served as negative controls in a particular study and cases where animals carrying a mutation that is associated with human cancer remained disease-free. For example, loss-of-function mutations in the human FANCC gene result in the familial cancer susceptibility disorder, Fanconi anemia, whereas homozygous Fancc knockout mice do not develop malignancies (22, 23).

The high degree of data integration allows the user to retrieve a maximum amount of information from one query. Each tumor record is linked to frequency data, information about the genetics of the mouse strain, age of the animals at diagnosis, and a literature citation. Where applicable, MTB reports the experimental regimen (referred to as “treatment”) used to promote or inhibit tumor formation. Tumor pathology data are represented by annotated histology photomicrographs that are associated with detailed captions authored by veterinary pathologists (Fig. 3). MTB currently contains ∼100 images of mouse tumors, some of which represent previously unpublished data obtained from The Jackson Laboratory mouse colonies.

Hypertext links connect MTB to other databases providing supplementary information. For instance, gene and allele symbols are hyperlinked to the Mouse Genome Database, which contains gene mapping and expression data, genomic linkage maps, phenotype descriptions, and links to orthologous genes in other mammals. If a mouse strain is available from The Jackson Laboratory or from the MMHCC,¹⁰ hyperlinks lead to strain and availability information from the JAX Mice¹¹ database or the MMHCC Repository,¹² respectively. Where appropriate, MTB provides links to tumor type-specific or strain-specific databases, such as the Biology of the Mammary Gland¹³ web site at the NIH, and the Transgenic Adenocarcinoma of the Mouse Prostate¹⁴ web site at Baylor College of Medicine (Houston, TX).

Although expert curation of the literature is the primary data acquisition strategy, MTB also invites direct submissions of tumor pathology photomicrographs from the research community. Because most print journals impose restrictions on the acceptable number of figures per article, researchers are often forced to select only a few representative images. Consequently, much valuable pathology data remains unpublished. By providing scientists with the opportunity to contribute supplementary photomicrographs, MTB seeks to fill this gap. To facilitate the data submission process, we have implemented JaxPath, a web-based editorial tool that allows researchers to review their images and edit annotations and captions on-line. JaxPath records are password-protected and are merged into the publicly accessible database only with explicit permission from the submitting investigators. Information on the pathology data submission process and sample records can be obtained by following the JaxPath link from the MTB home page (Fig. 1).

MTB is the most extensive resource of mouse tumor pathobiology data currently available on the web (24). Maintaining and expanding the database through literature data curation is an integral part of The Jackson Laboratory’s Mouse Genome Informatics program. The MTB information content is updated weekly. New tools are being designed for MTB to enhance the representation of quantitative trait loci mapping data, to accommodate large-scale tumor genotype and gene expression profiling data, and to include images produced with in vivo tumor imaging technologies that are being adapted to mice. Hyperlinks will connect MTB with the mouse cancer phenotype database being developed by the MMHCC. To ensure database interoperability, MTB will adopt the controlled vocabularies of diagnostic disease terms that are being evaluated by the MMHCC. Finally, MTB welcomes critical comments and suggestions from the cancer research community to improve data accuracy and user-friendliness.

We thank Dr. Jerrold M. Ward (NCI, Bethesda, MD) for helpful advice. For critical reading of the manuscript we thank Dr. Vivienne I. Rebel (Dana-Farber Cancer Institute, Boston, MA), and Drs. Louise M. McKenzie and Antonio J. Planchart (The Jackson Laboratory, Bar Harbor, ME).