Pathology of Genetically Engineered and Other Mutant Mice. Группа авторов

Figure 4.5 Fitting mouse models to the concepts of human disease. The mouse beige mutation (Lyst^bg) was found early on to be a good model for Chediak‐Higashi syndrome in humans. By contrast, the mild form of the hairless gene (Hr^hr) mutation was considered 30 years ago to be a model for normal human skin and was used for sun screen testing.

Some monogenic mouse mutations have features that are essentially identical to or at least very similar to those seen in humans with the same disease. An excellent example is the beige mouse [19] (one of many mammalian species [20, 21] that have mutations in the lysosomal trafficking regulator [Lyst^bg] gene) as a model for Chediak‐Higashi syndrome in humans (Figure 4.5), which can be called a phylogenetically valid model.

One of the big problems with working up (phenotyping) new mutants is that investigators often focus only on the organ or disease process, or molecular pathway of interest rather than do complete analysis of the animal. This results in incomplete definition of the mouse making accurate comparisons to the proposed human disease. Likewise, this is often the problem when defining the human disease, where clinical specialists focus on their area of interest and ignore other organ systems. Combined, this becomes the proverbial blind men describing the elephant [22].

By contrast, before the gene responsible for the hairless mouse was identified, many investigators used it as a model for “normal human skin” because the mice lacked hair as adults, making it useful for topical drug or ultraviolet light carcinogenesis studies. The more severe allelic mutation, rhino (Hr^rh) was used for testing drugs, such as retinoic acid analogs, for treating wrinkles [23–26]. Other than being without hair neither were accurate models for normal human skin, but lesions were later shown to be a model for human papular atrichia [27, 28]. The orthologous human disease was eventually identified [29] as was the gene, which allowed direct comparisons between the correct human disease with the correct mouse model [29–34]. This has been the case for numerous single mutant gene mouse models where finding the underlying gene allowed for accurate comparisons with human patients with mutations in the orthologous gene (Figure 4.6).

      The one gene–one disease concept remains an oversimplification of disease. Maintaining a single gene mutation as a colony on an inbred strain minimizes variability (a relatively reproducible model) but does not eliminate it within that strain as the mice are not absolutely identical genetically [35]. Many diseases have a sexual dichotomy, and environmental effects, such as diet, can result in different phenotypes at different institutions [36]. Moving the mutation from one inbred strain onto another strain (creating congenic strains) can result in loss to exaggeration of the phenotype, or even to a different phenotype due to the effects of modifier genes or gene redundancy. For example, mouse models of human cancer genes often result in a mouse model with a similar but not identical phenotype [37], where mouse genetic background may play a role in the phenotype. Trp53 null mutant mice (allele not designated) developed high rates of mammary tumors when bred on the BALB/cMed background but not C57BL/6 or 129/Sv congenic backgrounds, indicating a more complicated genetic predisposition toward mammary tumorigenesis than only mutations in this one gene [38]. These concepts are discussed in detail in Chapter 3, Genetics. However, these observations explain the inability of one inbred mouse strain carrying one or more mutated genes to model all variations seen in human patients. This is because humans are, for the most part, outbred and live in a relatively “dirty” environment. One concern raised about mouse models has been that mice cannot be good models of human disease because they live in environmentally controlled housing (boxes) under pathogen‐free conditions with abundant food, water, and access to good veterinary care. When one steps back and looks at how humans exist in a modern society, they too live in a box with many of the same benefits as mice [39].

Figure 4.6 Validation of mouse models over time. In 1989 few mouse models accurately reflected human diseases. This changed just a few years later, in 2002, as genetic mutations in specific genes were identified in both mice and humans that allowed more accurate comparisons.

      Genocopies, defined as mutations in differing genes that result in a similar phenotype in humans or mice, further complicate matching mouse genetic‐based diseases with those found in human populations. Such is the case for genes that are receptors, ligands, or that interfere with the kinetics of these interactions, exemplified by anhidrotic ectodermal dysplasia models (the tabby–crinkled–downless syndrome or mimics, see Chapter 10 on Skin for more details) [40, 41]. Similarly, ectopic mineralization of soft tissues is a complicated process involving numerous genes that, when mutated, can result in overlapping phenotypes because they are part of a complex molecular network [42]. As molecular pathways or networks are further refined, not only do protein–protein interactions become better defined but also candidate genes for similar diseases are identified that can then be mutated in mice to determine biologically if this is actually the case.

Large‐Scale Mutagenesis Projects to Discover New Mouse Models

      The traditional approach to determining gene function has been to work up spontaneous mutations in mice. An example was the allelic series of desmoglein 4 (Dsg4) mutations found first in mice [43, 44]. These were initially compared to human Netherton's syndrome because occasionally abnormal hairs (trichorrhexis invaginata), pathognomonic for Netherton's syndrome, were found in these mice. However, the mutated gene in human Netherton's syndrome was later found to be serine peptidase inhibitor, Kazal type 5 (SPINK5) [45], which did not map to the location of the mouse mutation, making it obvious that another mechanism was involved. Desmoglein 4 was later identified as the genetic basis of the mouse lanceolate hair phenotype. Soon thereafter, the orthologous human disease was found to be the result of mutations in the human DSG4 gene [46]. Not only was the mouse disease identified first, but by finding more than one disease had a so‐called “pathognomonic lesion,” it demonstrated how that concept, of a lesion specific for a disease, can be misleading.

Another example where the concept of “pathognomonic lesion” is an over‐simplification is with hair disease characterized by decreased levels of sulfur containing proteins (trichothiodystrophy, illustrated in Chapter 10, Skin). Trichothiodystrophy was a term used for a specific human disease [47, 48]. Hair shaft structural defects are quite common in mice with a number of mutations and most have low sulfur levels, making that a phenotype not, a pathognomic lesion. Trichothiodystrophy is now considered to be a heterogeneous group of autosomal recessive disorders that share sulfur‐deficient brittle hairs in humans [49]. Many of these defects are due to mutations in specific keratin genes or genes that regulate keratin expression (such as the forkhead box N1 gene causing the nude mouse [Foxn1^nu] and human ortholog [FOXN1]), especially the hard keratins that are high in sulfur containing amino acids [50].

      Two main strategies to generate mouse models at scale have been developed based on forward and reverse genetics. The aim of both being to make mutations experimentally, and then characterize the resulting phenotypes.

      The forward genetic approaches define the phenotype and aim to identify mouse orthologs of genes responsible for human disorders through cross‐species phenotypic similarity. These can utilize mice carrying spontaneous mutations [51], those that are the result of chemical mutagenesis [52–56] or other means, such as radiation induced mutation. In addition to identifying null gene mutations, forward genetics approaches often identify hypomorphic, gain‐of function, and dominant‐negative