animal models (regardless of species) of human diseases were identified and proposed empirically. Namely the best fit was based largely on clinical similarities between the human and animal disease (Figure 4.1). This concept of “best fit,” or the “Cinderella Effect,” [1] has evolved considerably with the advance of technology to include direct correlations at the level of the genetic variation found in humans with specific diseases.
However, access to advanced technology does not guarantee the identification and development of valid animal models. In fact, without critical observation and proper experimental design, misinterpretation of experiment results can derail the valuable contribution that might be obtained from the use of animal models. In a relatively recent example, an investigation on inflammation compared eight‐week‐old C57BL/6J male mice (one inbred strain, one sex, one age) with a heterogeneous group of outbred human patients up to 55 years of age with blunt trauma, burns, and bacterial endotoxin responses. The results showed consistency in the overall transcriptional response in humans, but major differences when compared to those seen in a mouse model, which led the authors to conclude that mouse models were without value [2]. This paper caused a sensation in lay media (http://www.nytimes.com/2013/02/12/science/testing‐of‐some‐deadly‐diseases‐on‐mice‐mislead‐report‐says.html?smid=pl‐share) as well as scientific circles. However, re‐evaluation of the data revealed important areas of similarity [3]. Because the transcriptome data used in that study were available in publicly accessible databases, other groups were able to reanalyze the data and showed that mouse and human comparative studies were valuable even in this model [3, 4]. This change in interpretation, still debated, reflects a fundamental experimental design flaw in the study that led to the misinterpretations, namely “comparing apples to oranges” [4] and whether mouse and human diseases can be compared or not. One human, or even one group of humans, is unlikely to accurately model all humans, which is the lesson learned from approaches to personalized medicine [5]. One should not be surprised to find that young laboratory mice of a single strain, a single sex (the genetic equivalent of a single individual or a group of identical twins), and exposed to experimental insults, does not mirror the inflammatory response of a heterogeneous human adult population of mixed ages and both sexes. It is a general point that multiple inbred strains or outbred strains need to be used when attempting to model human disease [6]. Therefore, careful comparison at the most basic levels are necessary to accurately create and validate animal models for human diseases. This applies to genetic‐based diseases as well. Definitions of human diseases include common, often characteristic lesions, as well as less frequent to rare lesions (subtypes of the disease). These can be conceptually presented as a Gaussian curve with each bar representing a subgroup of any particular disease (Figure 4.2). Inbred mouse strains essentially present as an infinite number of one individual with both sexes due to their genetic homogeneity. As such, each inbred strain carrying a defined mutation may represent a model of one subtype of the human disease not all subtypes. Some strains may exhibit the common or characteristic lesion while others do not.
Real examples of this were exemplified by creating a null mutation in the epidermal growth factor receptor gene (Egfrtm1Dwt). When this allele was moved by breeding onto other strains to create hybrid stocks, a variety of phenotypes were observed, ranging from embryonic lethality at 4–11.5 days post conception (DPC) to live births with runting (Figure 4.3) [7, 8]. While this is an interesting observation, the question is really how can these observations help us to better understand human diseases. An example of how this phenomenon can explain the human disease subtypes came via a spontaneous hypomorphic allele of the laminin gamma 2 gene (Lamc2jeb) in mice. When the gene was mapped by crossing the parental strain with five other inbred strains, in the process of creating five new congenic strains, a large variation in phenotypes was evident [9]. Skin fragility (blister formation) was a feature on all backgrounds, characteristic of junctional epidermolysis bullosa. However, on one background, nail abnormalities were observed. Mapping for these quantitative differences revealed the collagen 17A1 (Col17a1) was a major modifier gene that resulted in the mice having nail dystrophy, a feature of one subtype of the human disease [10]. A second gene was later identified that explained the musculoskeletal weakness in another subtype of this human disease (Sproule and Sundberg, unpublished work). Figure 4.4 illustrates how using websites that define specific features of subtypes of a human disease can facilitate phenotyping mouse models to understand the underlying genetics behind these variations.
Figure 4.1 Nonhuman (including mouse) models were historically compared empirically to human diseases. Diseases present a variety of phenotypes and severity levels in a mixed population, as it does with a human population as illustrated by the Gaussian curve. Where the lesions (phenotypes) overlapped was where they were considered to be potentially useful as models. Many did not hold up over time.
Figure 4.2 Human population response to disease is often highly variable between individuals. By contrast, inbred mice carrying the same genetic mutation are more monomorphic and can present as a specific subtype of the human disease.
Figure 4.3 Effect of strain on phenotype for a single gene mutation. A classic example of the effect of an inbred strain background on phenotype was when the epidermal growth factor receptor null mutation (Egfrtm1Dwt) was transferred by breeding onto different hybrid genetic backgrounds.
Sources: Based on Threadgill et al. [7] and Sibilia and Wagner [8].
Basic Concepts of Defining a Mouse Model for a Human Disease
There are numerous criteria needed to make an animal model for a human disease, many of which were standardized in a fascicle series edited by George Migaki during the 1970s–1990s (Table 4.1) [11]. Models can be built on spontaneous or induced mutations, or those experimentally generated by various methods (including chemicals, infectious and physical agents, surgery, etc.). Simply finding a case or case series is a start but, ideally, the model should be reproducible either by breeding or by using some type of manipulation [12]. Sometimes a model may combine an underlying genetic lesion with an additional challenge such as infection, aging, or high fat diet to manifest its full range of phenotypes. An example of this would be the feeding of high‐fat diets to mice to understand the genetics of atherosclerosis [13]. Furthermore, the model should be readily available for other scientists to validate and utilize. Single gene mutations in mice, which result in a genetically reproducible disease, have been the cornerstone of mouse genetics and models for more than a century. However, the locus or allele heterogeneity in human genetic disorders also makes a specific gene mutation mouse model useful for
