Genetically Engineered Animal Models in Toxicologic Research

1.1

Methods for Genetic Modification

In 1981, Gordon and Ruddle coined the term “transgenic” to describe an animal in which an exogenous gene has been introduced into its genome ( ). The practices of introducing genes with known function and deliberately inducing genetic mutations into the genome of animals to illuminate novel phenotypes (a strategy termed “reverse genetics”) were in part driven by the challenges of identifying spontaneous mutations in endogenous genes and interpreting the associated downstream events (an approach followed in “forward genetics”). Reverse genetics required less time and expense during the early days of this technological revolution. However, with the constant advances in DNA, RNA, and protein methods, forward genetics has become more feasible as a routine strategy for genetic investigations. This shift in approach must always be considered in the context of the animal model. In the late 1980s, the term “transgenic” was extended to gene-targeting experimentation and the production of chimeric or “knockout” mice in which a gene (or genes) was selectively removed from the host genome in some (for chimeras [i.e., animals with two or more genetically distinct cell populations]) or all cells ( ). Today, a transgenic animal can be defined as one having any specific genetic modification induced by transfer of genetic material from one organism to another. Transgenic animals are most commonly produced through one of three general approaches: (i) germline modifications of gametes; (ii) microinjection of DNA or gene constructs into the pronucleus of zygotes; or (iii) incorporating genetically modified cells, including embryonic stem cells (ES cells), into later-stage embryos ( ; ). Rodents (especially mice and rats) are the most commonly engineered species for biomedical research (see Animal Models in Toxicologic Research: Rodents , Vol 1, Chap 17 ). These technologies are employed in many species beyond mice and rats, including bacteria, fungi, plants, nematodes, insects, amphibians, fish, poultry, rabbits, ruminants, swine, and nonhuman primates. For more information on these species, refer to other chapters in this volume on various animal models employed for product discovery and development (see Animal Models in Toxicologic Research: Nonmammalian , Vol 1, Chap 22 ; Animal Models in Toxicologic Research: Rabbit , Vol 1, Chap 18; Animal Models in Toxicologic Research: Dog , Vol 1, Chap 19 ; Animal Models in Toxicologic Research: Pig , Vol 1, Chap 20 ; Animal Models in Toxicologic Research: Nonhuman Primate , Vol 1, Chap 21 ).

Pronuclear/nuclear microinjection of DNA is a basic physical technique for transgenic technology today ( ). Microinjection of DNA into zygotes (fertilized one-cell embryos) typically involves the use of micromanipulators and a microinjection apparatus to introduce a solution containing recombinant DNA into the nuclear genome on a permanent basis ( Figure 23.1 ). Injections usually are made into the larger (derived from the male) of the two pronuclei, originally for the simple reason that it is a bigger target; thus, the FVB strain of mouse with its very large pronuclei has been a favored platform for engineering the nuclear genome by direct DNA microinjection ( ). After injection, zygotes then are implanted into the oviduct or uterus of hormone-primed pseudopregnant recipients. The resulting progeny are tested for their transgene status by genotyping a tissue biopsy (usually an ear punch or tail snip), typically at or shortly after weaning ( ). “Transgene-positive” founder animals (F ₀ ) are retained for later phenotypic analysis and/or breeding to generate F ₁ and other generations of progeny that bear the desired genetic modification. Most transgene-negative littermates of the F ₀ animals are culled, but some individuals of both sexes usually are kept to serve as age-matched “wild-type” controls for the underlying genetic background.

The “constructs” (i.e., DNA vectors, usually a linearized viral plasmid) that are used to carry the transgene have been engineered to contain several different components that promote efficient transgene expression. The most important components are (i) a promoter, a specific DNA sequence that binds the transcription factors needed to recruit RNA polymerase and drive transcription of the engineered gene; (ii) the DNA sequence for the transgene of interest; and (iii) a polyadenylation (polyA) sequence that signals when to terminate transgene transcription ( Figure 23.1 ). It is essential that the spatial (tissue location) and temporal (timing) effects of the promoter be well characterized in advance, often through utilizing reporter genes in place of the transgene, prior to introducing the desired transgene in order to investigate the potential for “off-target” effects arising from unexpected expression of the transgene in time or tissues. The transgene sequence may be obtained by isolating specific sequences from an organism's endogenous gene complement (i.e., genomic DNA) or by reverse-engineering RNA transcripts to recover the gene sequence from which it was generated (i.e., complementary DNA [cDNA]). Transgenes may be designed to increase their expression in response to xenobiotics by engineering them to contain a xenobiotic response element in correct association with the promoter. Alternatively, cell ablation may be achieved using a toxicogenic approach, in which the promoter-directed expression of a toxin gene results in intracellular toxin production and cell death.

Following microinjection, DNA typically integrates randomly into the nuclear DNA (nDNA) of the zygote, generally in a single chromosomal location. The exact site of integration will differ for each F ₀ animal, indicating that each line will have to be analyzed individually for the extent of transgene expression and the existence of a reliably transmitted, transgene-mediated phenotype. Transgene integration may occur as a single copy or as concatemers (many copies connected in series) containing from two or three up to several hundred copies of the transgene. When multiple copies are introduced, they are organized predominantly in a head-to-tail fashion. Several different engineered transgenes may be assembled in a single construct to link their expression in vivo. Similarly, the tendency for transgenes to integrate at a single site may be used to link the functions of two or more independent transgenes carried on different constructs. In such cases, after simultaneous coinjection, the constructs integrate randomly but in the same site, yielding coordinated expression of several transgenes.

Since DNA microinjection is usually undertaken in individual zygotes (one-celled embryos), the transgene is anticipated to be present in all cells arising from the zygote. However, occasionally, the integration of the microinjected DNA into the host genome is delayed, as when incorporation of the transgene does not occur until after blastomeres (i.e., the totipotent cells that comprise early embryos) have begun dividing. With delayed integration, some, but not all, embryonic cells will contain the transgene; such embryos, in which transgene-positive and transgene-negative cells are intermingled, are termed “mosaic.” These mosaic F ₀ animals will have transgene expression when genotyped, and thus are technically transgenic, but they will only be able to produce transgenic progeny if the engineered gene is present in their germ cell progenitors. Therefore, the offspring (F ₁ generation) of transgenic F ₀ must be tested again to ensure that expression of the transgene is heritable.

Expression of a randomly inserted transgene may induce a phenotype in several fashions. First, the transgene may elicit an effect of its own that represents the inherent function of the protein that is specified by the newly introduced gene. Second, the transgene may function in a “dominant-negative” manner to inactivate or regulate the actions of endogenous genes. Third, the transgene may lead to “insertional mutagenesis” if its site of random insertion disrupts the sequence and function of an existing gene ( ); in essence, this outcome leads to an accidental gene deletion, or “knockout,” the unintended phenotype of which will need to be differentiated from any true phenotype resulting from expression of the transgene. For all three of these mechanisms, host DNA near sites of random integration often undergoes sequence duplication, deletion, or rearrangement as a consequence of transgene incorporation. Furthermore, the regulatory elements in the host DNA near the site of insertion, and the general availability of that chromosomal region for transcription, appear to have major roles in affecting the location and expression level of the incorporated transgene. This “positional effect” may partially explain why expression of the same transgene can vary dramatically among individual F ₀ animals and their progeny. In particular, the positional effect may differentially affect the function of the promoter sequences used to direct transgene expression to different cell types and organs. Thus, when characterizing transgenic animals that were generated with any method where random integration of the transgene occurs, it is essential to keep in mind that cell type- or tissue-specific expression of the transgene depends not only on the specificity of the promoter sequence but also on the positional influence of the integration site. For this reason, any phenotype identified using transgenic animals must be similar in at least two lines derived from different F ₀ founders before it can be concluded that the engineered gene is responsible for producing the effect ( Figure 23.1 ).

Viral-mediated gene transfer represents an alternative means of inserting transgenes permanently into the nuclear genome ( ; ). This technique may be utilized to explore molecular pathways or as a therapeutic modality (see Gene Therapy / Gene Editing, Vol 2, Chap 8). In this approach, a retroviral or lentiviral vector is incubated with mammalian cells, commonly an ovum or zygote so that the genetic material will be present in essentially all adult cells. This technique is efficient and effective across a number of species, although species- and strain-specific differences in the efficiency of transgenesis have been reported. Infection with retroviruses commonly leads to insertion of a single transgene copy at a single integration site. Potential problems with the use of retroviral vectors include mutagenicity, oncogenicity, the limited size of the transgene payload, and the common occurrence of gene instability and recombination ( ). Lentiviral vectors provide significantly better integration efficiency and transgenic animal yields relative to both retroviruses and microinjection methods for transgenesis. Furthermore, lentiviruses permit a larger payload, promote stable integration of multiple gene copies, and do not induce mutagenic and oncogenic events. The main drawbacks to their use are the potential for transgene silencing, genetic mosaicism, and vector-related toxicity resulting in a variable percentage of the F ₁ generation with transgene expression. Viral vectors must be handled using appropriate biocontainment procedures and work areas, particularly if replication competent, to avoid any potential for infecting the scientific staff. In practice, both the vector construction and in vivo experiments are conducted using conditions suitable for Biosafety Level (BSL) 2.

Systemic gene delivery (“gene therapy”) into adult animals using viral vectors is a commonly used method to overexpress transgenes after an individual's development is entirely or nearly complete. The main drawback to this strategy is that persistence of the transgene is impacted by the nature of the vector. For example, adenovirus (AdV) is more immunogenic than are both adenoviral-associated viral (AAV) and retrovirus, so transgene levels are likely to be higher for a longer period if the transgene is introduced using one of the latter two options. Furthermore, the inflammatory reaction to AdV vectors may induce confounding functional and/or structural changes in the infected cells which in some circumstances may overlap with or obscure the phenotype exerted by the transgene. These viral vectors (especially AAVs and retroviruses) are common test articles in human clinical trials for gene therapy products, and now are being approved for use ex vivo and in vivo to correct diseases caused by single-gene mutations in human patients (see Gene Therapy /Gene Editing, Vol 2, Chap 8).

RNA interference (RNAi) is a useful means for transiently repressing the expression and function of targeted nuclear genes ( ). The most common means for producing this effect is to introduce antisense oligonucleotides (ASOs) or small interfering RNAs (siRNAs). Delivery of these inhibitors may be done by pronuclear/nuclear microinjection (typically employed for embryos) or by parenteral injection (usually employed in adults). This approach typically elicits a partial decrease (i.e., “knockdown”) rather than complete loss (“knockout”) of gene function. In this regard, RNAi is gaining popularity as a means for investigating the functional consequences of innovative drug candidates, as pharmacological intervention reliably reduces but seldom completely halts activity of the targeted molecular pathway.

When using the RNAi approach, it must be remembered that variations in complimentary oligonucleotide length and composition may affect the degree and duration of suppressed gene expression. Furthermore, it has been shown that saturation of endogenous RNAi pathways can overwhelm the balance of mRNA degradation and production, thereby creating the potential for cytotoxicity. Strategies to manage this balance during therapeutic interventions are continuing to evolve (see Nucleic Acid Pharmaceutical Agents, Vol 2, Chap 7).

Gene targeting by homologous recombination is another routine technique used to produce genetically engineered animals ( ). For this procedure, the permanent insertion of an engineered gene of interest may inactivate expression of the normal gene, thereby creating a null mutation (i.e., “knockout”). Alternatively, the insertion may replace the endogenous gene with a different functional gene (i.e., “knockin”) to evaluate its role in a given disease or examine the function of a similar gene from another species (including human). In general, gene targeting by homologous recombination is undertaken in vitro by introducing the modified gene sequence into ES cells, confirming that recombination has produced the correctly engineered genotype, and then inserting the modified ES cells either by microinjection into a blastocyst (a multicelled, cavitated embryo) or by coaggregation with a morula (a multicelled, noncavitated embryo) ( Figure 23.2 ). For most mouse studies, the ES cells of choice are derived from one of two strains, C57BL/6 or 129 (which includes a variety of substrains with minor but nonetheless essential differences in their genetic backgrounds) ( ; ). The introduction of engineered ES cells into the multicelled embryos results in the production of chimeric F ₀ animals in which the engineered gene is expressed in only a portion of the embryonic tissues. Crossing of F ₀ in which the mutated gene is present within the germ cell precursors will produce one or more lines of offspring (F ₁ ) possessing the modified gene in all cells on a permanent and heritable basis.

Typical constructs used to target genes by homologous recombination ( Figure 23.2 ) are more complex than those designed to deliver transgenes that will be randomly inserted ( Figure 23.1 ). The most important components in gene-targeting constructs are (i) the mutated gene of interest, (ii) DNA sequences for genes encoding selection markers, and (iii) identification and accurate construction of the 5′ and 3′ end regions of homology that will allow the correct insertion and replacement of the mutated gene in place of the endogenous gene. After introduction of the assembled construct into ES cells, homologous sequences between the specific endogenous gene of interest and its engineered counterpart carried by the construct can bind to each other as the nuclear DNA (nDNA) is being replicated. A small fraction of the time, binding of the construct to the gene of interest will result in reciprocal recombination so that the engineered mutation is exchanged into the host genome at its proper site and the endogenous gene is transfered into the plasmid. Subsequent treatment of ES cell cultures with the xenobiotic selection agents will spare only those ES cells in which homologous recombination introduced the appropriate pattern of xenobiotic response and resistance genes ( Figure 23.2 ).

Since ES cells with gene-targeting constructs are deployed in multicelled embryos, the mutant gene always will be integrated in only a portion of embryonic cells. Confirmation of the gene integration can be performed in ES cells prior to their introduction into embryos. Ideally, most members of the founding generation (F ₀ ) will be chimeras without needing to exploit host embryo engineering (i.e., manipulations to ablate the host blastomeres in the developing embryos). However, only some will carry the engineered gene in the germ cell precursors because the altered ES cells may or may not form germ cells in a given animal. If a founder's gametes bear the mutant gene, cross-mating with wild-type animals will produce a heterozygous (F ₁ ) generation in which animals derived from gametes arising from the engineered ES cells will bear one copy of the mutant gene and one copy of the wild-type gene. Breeding F ₁ animals will yield an F ₂ generation in which approximately 25% of the progeny will be anticipated to carry two mutant genes at the targeted site (i.e., homozygous knockouts or knockins), 50% will be heterozygotes with one mutant copy and one wild-type copy, and 25% will have two wild-type genes. However, this predicted 1:2:1 Mendelian ratio may be absent when the F ₂ progeny are genotyped (typically at weaning, or shortly thereafter) if the double dose of mutant gene results in death during gestation (i.e., “embryonic lethality”), at or soon after birth (i.e., “perinatal lethality”), or at any juvenile stage prior to weaning. In general, some age-matched wild-type and heterozygous littermates of both sexes should be retained as control animals for subsequent studies of knockout phenotypes ( ).

Induced pluripotent stem (iPS) cells (or iPSCs) represent an alternative source of stem cell-like elements into which modified genetic material may be inserted ( ; ; ). In this approach, differentiated cells (commonly fibroblasts in mice and rats) can be reprogrammed to enter a more embryonic-like (pluripotent) functional state when cultured by using a cocktail of embryonically expressed factors. The morphology and growth characteristics of iPS cells resemble those of totipotent ES cells. Importantly, like ES cells, following injection into blastocysts iPS cells can contribute to embryonic development for multiple cell lineages and also support the generation of genetically modified founder animals.

Nuclease-based genome editing is an enzyme-mediated method of generating nuclear gene modifications that can be adapted to many different species including plants, invertebrates, and vertebrates (both small laboratory animal species and livestock) that lack readily available, validated ES cell sources ( ; ). Genome editing may be undertaken to explore basic biological mechanisms or to produce therapeutic alterations to mutant genes (see Gene Therapy / Gene Editing, Vol 2, Chap 8). Four options have been identified and evaluated: homing endonucleases (“meganucleases”), transcription activator-like effector nucleases (TALENs), zinc finger nucleases (ZFNs), and the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated nuclease (Cas) system. Large recognition sites make meganucleases perfect for genome engineering; however, the number of naturally occurring meganucleases is finite, limiting their overall utility ( ; ). Accordingly, they will not be discussed further here. TALENs are nonspecific DNA-cleaving nucleases composed of a FokI nuclease domain (that generates a double strand break with cohesive, staggered ends) fused to a custom-made DNA-binding domain that is composed of highly conserved repeats derived from transcription activator-like effectors (TALEs) present naturally in Xanthomonas proteobacteria. Most of the engineered TALE repeat arrays use four domains with hypervariable regions (designated HD, NG, NI, NN) that specifically recognize C, T, A, and G bases, respectively. A single TALEN pair is commonly used to generate nonhomologous end joining (NHEJ)–induced knockout mutations, and the use of two TALEN pairs can generate deletions and/or inversions of large chromosomal segments. The TALENs can target any DNA sequence and have been employed to alter genes in many eukaryotic cells, including yeast, plants, invertebrate, and vertebrate animals as well as human somatic and pluripotent stem cells. In contrast, ZFNs are artificial restriction enzymes formed by fusion of the FokI nuclease domain to a zinc finger DNA-binding domain. In most projects, ZFNs are transfected as either DNA plasmids or mRNA into zygotes by microinjection. Upon ZFN cleavage of their target site, endogenous cell processes are harnessed to produce targeted mutations (e.g., point mutations, deletions, insertions, inversions, duplications, and translocations) that result in gene knockouts.

The CRISPR/Cas system is a ribonucleoprotein complex developed by reverse engineering of elements involved in the prokaryotic adaptive immune response ( ; ; ; ; ; ; ). The CRISPR sequences are a family of DNA direct repeats that are present naturally in many prokaryotic (archaeal and bacterial) genomes. The CRISPR repeats represent residual viral sequences that serve as the “memory” sequences for prior episodes of bacteriophage invasion. Subsequent bacteriophage infections lead to transcription of the CRISPR repeats into RNA transcripts, and the small RNAs then interact with a Cas endonuclease to form a complex that binds to and degrades viral gene sequences. When employed to modify the genome in eukaryotic cells, the CRISPR/Cas system requires both an engineered guide RNA (gRNA) to target the desired gene sequence and a Cas (or equivalent) endonuclease to produce targeted DNA cleavage, thereby producing a site wherein modified genetic material may be introduced ( Figure 23.3 ).

The CRISPR/Cas system has become the genome editing method of choice in the last half-decade because it has several key advantages over both the other nuclease-based genome editing approaches (specifically TALENs and ZFNs) and conventional gene targeting by homologous recombination. The first and foremost improvement is target design simplicity. Both TALENs and ZFNs require complex protein engineering steps to produce molecules that recognize the target DNA, which is both costly and relatively slow. In contrast, the gRNAs utilized in CRISPR/Cas editing are based on simple complementary RNA–DNA interactions, and synthesis of gRNA oligonucleotides (oligomers) is inexpensive and fast. Second, using the CRISPR/Cas system is undertaken easily by introducing (by microinjection or viral transfection) genetic material encoding the gRNA and Cas protein into embryos. This direct approach bypasses the processes of transfecting and selecting mouse ES cells that are needed for the homologous recombination gene targeting technique described above, thereby greatly expediting the speed of animal model generation (typically taking 6–12 months for homologous recombination but reduced to 2–4 weeks for CRISPR/Cas). Third, CRISPR/Cas readily supports the introduction of multiplexed (multiple gene) mutations simply by injecting several gRNAs simultaneously to target several genes at once. Fourth, CRISPR/Cas can be used in any species with essentially no technical modifications except for defining a new gRNA sequence as microbes, plants, and animals all use DNA as their basis for transmitting genetic information. Finally, the CRISPR/Cas system provides comparable on-target gene editing efficiency and a reasonable degree of off-target gene modification compared to other currently available methods of genome modification.

“Conditional” gene modification can allow for spatial and/or temporal control of gene expression, and is most often accomplished via a combination of both gene targeting (i.e., site-specific homologous recombination) and DNA microinjection (i.e., random insertion) techniques. One example is site-specific recombinase technology, such as the use of the bacteriophage-derived Cre (cyclic recombinase) enzyme in conjunction with the small loxP sequences on DNA that serve as recognition sites for the Cre enzyme ( Figure 23.4 ) ( ; ). When the targeted gene of interest is flanked by two separate loxP sites (i.e., “floxed”) on each end of the gene, coexpression of Cre in the same cell will alter the target gene between the loxP sites. Depending on the orientation of the loxP recognition sites with respect to one another, the action of Cre will excise, exchange, integrate, or invert the intervening targeted gene sequence; the gene alteration produced by Cre is effectively irreversible. Conditional mutagenesis in this setting requires two separate parent lines, one bearing the floxed gene of interest and the other harboring a transgene encoding Cre. The outcome of breeding these parental lines depends in large part on the promoter driving Cre expression, as the nature of the promoters defines when and where Cre will be produced.

To achieve spatially defined conditional gene modification, constitutive expression of a tissue-specific promoter is used to direct Cre expression to a particular cell population or tissue of interest. The activity of Cre will alter the targeted gene of interest only in the elements in which the promoter is active, leaving the targeted gene intact in the other tissues where Cre is not expressed. The function of a widely expressed gene in different tissues may be investigated by breeding a parental line with a floxed target gene with various established Cre lines in which Cre expression in each line is controlled by a different tissue-specific promoter.

Some genetic mutations can have deleterious effects when they occur in development (e.g., embryonic lethality) or over the lifetime of an animal (e.g., early-onset degenerative or neoplastic diseases). These effects can be abrogated by temporally defined conditional gene modification. For this purpose, the parental line bearing the recombinase is constructed so that an inducible xenobiotic response element capable of either enhancing or repressing recombinase expression is linked to the promoter ( Figure 23.4 ). This design provides control over gene expression by allowing exposure to the xenobiotic to direct the onset or inhibition of expression for the targeted mutation. Tamoxifen (given in oil by parenteral injection) or tetracycline (added to the water) frequently is used for this purpose in product discovery studies ( ; ; ). Literature reports suggest that the response of engineered targeted genes does not depend on the dose of xenobiotic.

Spatial and temporal control of gene expression can be conducted simultaneously for a gene of interest by combining inducible elements on promoters directing expression only in specific cell types, tissues, or organs at particular developmental time points. In the parental line carrying the recombinase, the gene for this enzyme is engineered so that its expression is controlled by both time- and tissue-specific promoters ( Figure 23.5 ).

Although conditional targeting of gene expression using current methods is powerful for answering many questions, studies must be strategically designed for accurate interpretation ( ). Appropriate design of safety assessment studies with GEM must include multiple control groups; animals with and without promoters that have both received and not received the xenobiotic are necessary to differentiate whether or not the phenotype represents an effect of the engineered gene or a consequence of xenobiotic exposure. In addition to the potential for compound-related effects of xenobiotics on the animal system, consideration should also be given to the possible presence of endogenous recombinases and the potential effects on adjacent cells of expressing exogenous gene elements (e.g., reporter genes) within the engineered gene.

Nuclear transfer (i.e., cloning) is an option employed in generating transgenic livestock. The use of this methodology in mammals captured the imagination of scientists and the general public, and received widespread press, including the successful cloning of a sheep in the mid-1990s. Importantly, nuclear transfer during that decade was particularly important for gene targeting experiments in mammalian species other than mice because pluripotent, germline-competent ES cells had not been identified and validated in nonmurine species. Thus, nuclear transfer was essential for knockout experimentation in several biologically relevant species, in particular livestock ( ).

In this procedure ( Figure 23.6 ), the intact nucleus of a somatic cell from a transgenic donor animal is introduced into the cytoplasm of an enucleated zygote. The most common technique involves culturing the cells from the transgenic donor animal and then either physically inserting the nucleus into the zygote by microinjection or fusing a nucleated cell from the transgenic donor with the recipient embryo. The donor cells may be either stem cells or differentiated adult cells. The reconstructed oocytes then are transferred into a pseudopregnant surrogate dam, and the resulting progeny tested for the presence of the transgene.

Insertional mutagenesis has developed into a powerful set of technologies for gene identification, even though it started out as a confounding artifact encountered in characterizing transgenic founders generated by microinjection. Techniques commonly employed today seek to alter endogenous gene activity by inserting new genetic material (DNA or DNA-targeting constructs) to directly modify endogenous gene sequences, thereby leading to disruption of their expression or control mechanisms. The inserted genetic material may be engineered to include regulatory sequences (e.g., promoter and enhancer traps) and reporter genes (e.g., promoter-less fluorescent protein genes) that produce phenotypic features tied to expression of the inserted construct. For example, gene trapping is a variant of insertional mutagenesis designed to disrupt gene function by randomly introducing a reporter gene construct into enhancer, promoter, or gene sequences ( ). A “productive integration” event brings the reporter gene under the transcriptional regulation of the machinery that drives expression of an endogenous gene.

Chemical mutagenesis is a common approach to large-scale genomic investigations in mice. While other techniques for generating knockout animals often induce inactivating null mutations (i.e., total loss of an affected gene's function), mutagenesis with N -ethyl- N -nitrosourea (ENU) or similar agents is more likely to generate hypomorphic mutations (i.e., partial loss of function) ( ). This difference is important as the genetic mutations found in human disease are often related to hypomorphic rather than full gene deletions. Although most mutations induced by ENU are recessive loss-of-function mutations, approximately 25% of ENU-induced events result from haploinsufficiency (i.e., presence of a single functional gene that cannot produce enough of the gene product to sustain full function) or dominant gain-of-function or dominant-negative mutations.

Transposons (transposable elements ) are mobile DNA sequences that not only can integrate into the host genome but also can transfer (“jump”) genetic material between the genomes of different hosts. Gene transposition can occur as a “copy and paste” event (for retrotransposons) or a “cut and paste” event (for DNA transposons); retroviruses are a variant of retrotransposon. At present, this technology seems destined to serve as a tool for phenotype-driven research (“reverse genetics”) rather than a means of exploring genotype-driven processes (“forward genetics”). In comparison with ENU mutagenesis, some transposon systems have the distinct advantage that the site of mutagenesis can be rapidly identified.

Sperm-mediated DNA transfer has been demonstrated in several species as an alternative to standard transgenesis methods that modify the oocyte or zygote ( ). This approach uses the intrinsic ability of sperm cells to bind and internalize exogenous DNA molecules and then afterward enter the oocyte at fertilization by means of artificial insemination. Sperm-mediated DNA transfer has the advantages that neither embryo handling nor expensive equipment is required. This method has been used successfully to produce pigs bearing multiple transgenes. However, the methodology appears to work with a limited number of males and requires very large amounts of DNA vectors. Ongoing studies have attempted to identify the underlying mechanisms needed to make this technology successful on a routine basis. Interestingly, since this work was first reported, further development of spermatogonial stem cell–based technology has proven effective at generating founder transgenics and rescuing lineages through the male germline.

Any of the methods reviewed above may be used to manipulate a single nuclear gene per line or to produce a line that simultaneously expresses several modified nuclear genes. Until recently, the most common way of producing models with multiple modified genes has been to generate rodent lines having single-gene alterations and then undertake appropriate backcrossing. However, CRISPR/Cas genome editing has been demonstrated to reliably generate by a single engineering event genetically modified animals that bear multiple inserted genes; the reproducibility, rapidly, and relatively low cost of this technology for producing multi-transgenic animals now are facilitating a revolution in the speed with which new animal models of disease are being developed, especially in nonrodent species (e.g., dogs, pigs, nonhuman primates). Analysis of animal models with several engineered genes—especially “humanized” GEM and GER lines (i.e., those expressing human homologs of mouse genes or bearing human cells)—is an important translational science strategy for assessing complex metabolic interactions and mechanisms of polygenic diseases (see below).

Mitochondrial transgenesis may be employed to alter the extranuclear genome, which is harbored in the mitochondrial DNA (mtDNA). The techniques for genetic engineering of mtDNA are fairly primitive in comparison to the well-defined procedures used for traditional genetic engineering of nDNA ( ; ; ; ). Mitochondrial disorders encompass diverse disease states, affecting processes from control of cellular energy and lipid metabolism to the regulation of cell death. Although some mitochondrial disorders arise from mutations in nDNA, many instances of mitochondrial dysfunctions arise instead from abnormalities of mtDNA. The composition of mtDNA is highly conserved across vertebrate species, being 16.5 kb in length with a unique genetic code, genome structure, transcriptional and translational apparatus, and tRNAs. Relative to nDNA, mtDNA is markedly more susceptible to environmental perturbations (e.g., mutagens) and metabolic conditions (e.g., factors inducing oxidative stress), perhaps because of an absence of protective histones and/or a limited capacity for repairing mtDNA. Therefore, effects on mtDNA and mitochondrial function are of utmost importance not only for understanding the pathophysiology of heritable mitochondrial diseases but also in terms of assessing pharmaceutical activity and safety.

Early mammalian embryos may carry from tens of thousands to hundreds of thousands of copies of the mitochondrial genome. Therefore, to produce phenotypic changes, a threshold level of heteroplasmy (more than one mitochondrial haplotype) must be achieved. Both mitochondrial microinjection and ES cell transfer have been employed to modify the mitochondrial genome. Experience has shown that ES cell transfer technologies, starting with ρ° (rho zero) stem cells (i.e., cells that have been depleted of their own mtDNA by incubation in ethidium bromide, an inhibitor of mtDNA replication), are the most effective means for generating founder chimeras with high levels of mitochondrial heteroplasmy or in producing homoplasmy of the engineered mitochondrial gene (i.e., 100% of the mtDNA is derived from the ES cells) in second-generation mice. Mice engineered so that cells possess a baseline level of mitochondrial stress in the absence of other phenotypic effects have been used to model the impact of toxicant exposure on otherwise clinically silent mitochondrial disease ( ).

1.2

Nomenclature Conventions

A solid understanding of nomenclature conventions for gene mutations is an essential skill for scientists who work with genetically engineered animals. The nomenclature conventions have been standardized for mouse and rat strains, substrains, stocks, crosses, and genetic traits, and will be adapted in time for engineered animals of other species.

The rodent nomenclature standards are designed to succinctly and specifically identify the exact nature and origin of the model, including those that have been genetically modified. When comparing data from different publications describing mice with genetic manipulations of the same gene or genes, it is essential to understand whether or not the investigators are using the same model—including details regarding the altered gene and the genetic background—or different models. The reason for this requirement is that even seemingly subtle variations in the modified gene and/or genetic background may induce significant divergence in the ultimate phenotypes. Information on background strains, the exact genes that have been manipulated and the modifications that have been made, the method(s) of genetic manipulation, the source of the animal, and the amount of genetic uniformity based on the degree of backcrossing or inbreeding are summarized concisely when official nomenclature is used. Differences in any of these parameters may explain in part variations in experimental results among laboratories, although differences in environment, handling, and technical procedures cannot be excluded as other contributing factors.

Toxicologic pathologists and toxicologists who are not intimately involved with genetically modified rodent models often are not aware of the importance of nomenclature or how to decipher it. Glossaries and web-based resources now make this information readily available with a few computer keystrokes or clicks of a mouse. A guide to nomenclature conventions for mice is available from The Jackson Laboratory ( http://www.informatics.jax.org/mgihome/nomen/ ). A brief glossary of essential terms in the field is presented at the end of this chapter.

1.3

Model Selection

GEM, GER, and increasingly nonrodent models are important in product discovery and development for many reasons. Key functions include to facilitate target validation, to study predicted and unexpected efficacy and toxicity, to examine biological processes leading to resistance, to evaluate pharmacokinetic/pharmacodynamic (PK/PD) characteristics, and to discover biomarkers that will predict the onset and peak of disease, monitor its progression, and/or detect subclinical disease ( ; ; ; ; ; ; ; ; ; ). In practice, many studies conducted with GEM and GER models of disease are designed to simultaneously address multiple endpoints, especially evaluation of both product efficacy and potential toxicity.

In general, genetically engineered animal models are best used to test specific hypotheses that illuminate the relevance and/or pathogenesis of toxicities that have been observed previously in conventional (“wild type”) animals or humans. The most serviceable GEMs/GERs have routine husbandry needs, which simplify colony management and handling during experimental manipulations; have been engineered to faithfully recapitulate multiple aspects (though usually not all) of a human disease; exhibit moderate or greater penetrance (i.e., a large proportion of animals bearing a mutant gene develop the expected phenotype); and show evidence of the disease following a short latent period. Model and species selection are critically important for the success of studies in terms of both efficacy and/or toxicity evaluations as well as resource management (e.g., minimizing the cost, labor, and time required to complete a study). Key biological factors in model and test species selection for biomedical product development include target specificity, potency, off-target potential, cross-species similarities in PK/PD attributes, and organ-specific anatomy and physiology. In general, if a disease model is available in multiple animal species, the best choice for routine efficacy and toxicity studies is the one lowest on the phylogenetic scale (e.g., GEM or GER). However, for some products, the most suitable model for predicting safety liabilities may be a nonrodent, whether spontaneous or genetically engineered, due to the greater similarity in anatomy and physiology between the target organs/systems in test animals and humans (for example, nonhuman primate vs. human brains).

Therapeutic class is another key consideration during model/species selection. Small molecules generally interact with multiple cells or molecular pathways in many species, may be metabolized to toxic intermediates (although sometimes differently across species), and are not usually immunogenic. In contrast, biologics (nucleic acid- or protein-based products) generally are highly targeted to specific cellular receptors with high species specificity, are degraded to nontoxic peptides or amino acids, and can be highly immunogenic. Thus, two main issues in attempting to study biotherapeutics in GEMs/GERs are that the test article (a human-derived molecule or cell) either (1) may not bind with and/or activate rodent molecular pathways, and thus will not be pharmacologically active, or (2) may be so immunogenic in rodents that longer-term studies are not feasible due to the formation of neutralizing antibodies (see Protein Pharmaceutical Agents, Vol 2, Chap 6 ) . Surrogate molecules, such as a homologous rodent protein, may be used in GEMs/GERs to estimate potential efficacy of the human test article. However, in many cases, regulators tasked with evaluating human risk will prefer evaluating the real test article using in vitro assays in human cells (see In Silico, In Vitro, Ex Vivo , and Non-Traditional In Vivo Approaches in Toxicologic Research , Vol 1, Chap 24 ) coupled with in vivo testing in nonrodents.

Animal characteristics are important during species selection (see Issues in Laboratory Animal Science that Impact Toxicologic Pathology , Vol 1, Chap 29 ). Smaller animals require less test material per subject, though significantly more animals overall may be needed on study to meet blood volume requirements for PK and clinical pathology evaluations. Compared to rodent models, large animal models of disease often are less desirable because of their relatively long gestation periods, smaller litter sizes, older age of puberty, and higher husbandry costs.

To generate GEM and GER models, several mouse and rat inbred strains are preferentially used. For mice, the most common strains are 129, C57BL/6, FVB, and BALB/c. Each strain has advantages and disadvantages. For example, 129 mice are preferentially used for gene targeting because their ES cells are relatively easy to harvest. However, 129 mice exhibit great variation among various substrains due to deliberate and accidental outcrossing ( ; ), have poor breeding efficiency, are susceptible to autoimmune diseases, and have a moderate overall tumor incidence (7%–21%) with a high incidence of spontaneous lung tumors (up to 46%) and testicular teratomas (up to 30%) depending on the 129 substrain. Similarly, C57BL/6 mice are desirable for gene targeting studies since their ES cells are relatively easy to manipulate, and this genetic background is considered to be a standard, disease-resistant inbred strain for basic biomedical research. Furthermore, C57BL/6 mice have good breeding efficiency, low antibody affinity, and low overall tumor incidence (1%–7%) with a low incidence of spontaneous leukemia (7%). For transgenic experiments (especially for retinal degeneration studies), FVB mice are preferred due to their large zygote pronuclei (which makes microinjection of engineered DNA simple), fully inbred background, and vigorous breeding efficiency with large litters. However, FVB mice are aggressive to cage mates, susceptible to noise-induced seizures leading to brain necrosis, have high incidences of tumors (50%–60%, especially of the mammary gland), and are highly sensitive to the B strain of Friend murine leukemia virus. The BALB/c strain is used for genetic engineering on occasion based on their routine use for other purposes such as generating monoclonal antibodies and studying infectious diseases. BALB/c mice are resistant to experimental allergic encephalomyelitis and have good breeding efficiency, but they also exhibit a high overall tumor incidence (43%) with moderate lung tumor incidence (21%). In rats, ES cells have been derived from Sprague–Dawley, Fisher 344, and other strains. In contrast to mouse ES cell development, the literature suggests that in rats the strain genetic background is not a major barrier to establishing ES cell lines; furthermore, in rats there is no evidence showing strain incompatibility between host embryo and rat ES cells for chimera generation. The advent of CRISPR/Cas technology has greatly expanded the production of GER models ( ), and also genetically engineered nonrodent models ( ; ; ; ).

Study designs using genetically engineered animal models differ according to the purpose. For discovery (basic biology) studies to investigate the effects of one or multiple modified genes, all genes need to be studied alone and in combination to determine their individual and collective (e.g., additive, synergistic, or competing) significance in the onset and progression of a disease phenotype. Similarly, the different genes and gene combinations may need to be studied with and without posttranslational modification. In contrast, for development (e.g., efficacy and safety) studies, only the ideal combination of genes that specifies the desired disease phenotype needs to be studied, usually both with and without the presence of the test article. Extrapolation of data from these studies must be undertaken with caution, in full consideration of the model limitations relative to the human disease. However, the basis for most studies using genetically engineered animals is that the induced genotype and phenotype provide a biological system with an enhanced degree of similarity to the proposed pathogenesis of the human condition.

In the biomedical setting, recent reports call into question the translatability of efficacy findings obtained in animal studies given the number of drug candidates that have failed in human clinical trials. For instance, xenograft studies using human tissues implanted in immunodeficient (either spontaneous or genetically engineered) rodents have several shortcomings including defects in DNA repair in severe combined immunodeficiency ( scid ) mice (which limits testing of some cytotoxicants), a profoundly narrowed capacity for mounting an immune response (which prevents testing of immunomodulatory agents), and the overall frailty of nude ( Foxn1 ^nu ) mice (which limits their capacity to tolerate novel treatments). Other disadvantages inherent in many genetically engineered rodents include differences in their telomere biology (which alters the type and frequency of secondary cytogenetic events leading to neoplastic transformation), tumor suppressor mechanisms (e.g., p53 vs. p16), cytokine biology (e.g., the lack of an interleukin-8 homolog in mice), metabolism, and chromosome expression (e.g., the tumor suppressor genes p53 and Brca1 are on the same chromosome in mice, so codeletion is common). Overall, many disease models in GEMs, GERs, and genetically modified nonrodents are incompletely characterized, and studies may be undertaken in strains or breeds with very little historical information on spontaneous background findings over time. Use of these models in hazard identification and characterization as well as risk and safety assessment often is plagued by a lack of concordance with the human population of interest (e.g., single inbred rodent species/strain vs. many heterogeneous populations among humans). These factors may have a negative impact on risk assessment for new products due to the presence of poorly understood findings with unknown or no (i.e., GEM/GER-specific) relevance to humans, thus leading to increased development timelines and cost.

2.1

Genotyping

Genotyping , the first essential datum required to validate a new engineered model, is done to confirm the presence and magnitude of expression of an engineered gene. The genotype is commonly expressed in relation to the presence or absence of the modified gene since the remainder of the genetic background is presumed to remain essentially identical across all animals in the highly inbred rodent strains favored for genetic engineering. Appropriate expression of a modified gene or absence of a native gene may be established by assaying genomic DNA, typically by real-time polymerase chain reaction and electrophoresis or Southern blot. This assessment usually is performed before or at weaning as designated by an institutionally approved animal protocol, typically through convenient tissue sampling such as a biopsy (tail snip, ear punch, digit amputation) or skin swab ( ; ). In mice generated by backcrossing two or more strains, relative contributions of the strains can be elucidated via single-nucleotide polymorphism (SNP) panels, where analysis of over 94 inbred mouse strains has demonstrated the existence of 8 million SNPs ( http://mouse.cs.ucla.edu/mousehapmap/full.html ) ( ; ). Microsatellite (i.e., repetitive DNA sequences) and SNP profiles have become widely available for mice and rats. As few as 29 SNPs have proven to be sufficient to monitor genotypes of over 300 mouse strains ( ). Despite successes in using genome-wide association studies (GWAS) to link specific genes to various disease traits in humans, animals models such as inbred laboratory mice have a more limited genomic diversity due to their restricted genetic origin and constant selection pressure ( ). In fact, the majority of genetic variations observed in classical inbred mouse strains are derived from only three Mus musculus subspecies: Mus musculus domesticus , Mus musculus musculus , and Mus musculus castaneus . A large GWAS investigation of transcript abundance in 16 classical and 3 wild-derived inbred mouse strains has demonstrated the low statistical power of such studies in this species, with allelic association of various transcripts giving rise to spurious associations approaching those arising by random chance. Moreover, the genetic markers with the most significant p-values frequently fail to map to the actual expressed locations of the quantitative trait loci (QTLs) that regulate RNA expression levels. Thus, in the laboratory mouse, it appears to be necessary to combine GWAS with chromosomal linkage analysis to provide sufficient statistical power to attain higher mapping resolution, as less accuracy is achieved with either approach alone in inbred mice relative to their success as stand-alone methods in the more genetically heterogeneous human population. As an intraspecies methodology, however, GWAS has demonstrated some limited utility. For instance, when investigating genetic drivers of fibro-osseous lesions in 28 strains of inbred mice, 11 candidate genes have been identified; the resulting bone changes produced by these monogenic mutations may be compared to known phenotypes in the Mouse Genome Informatics (MGI) database ( ). The synergistic power of combining GWAS with large, multistrain compendiums of mouse data may lie in the ability to identify candidate genes that are putatively involved in complex polygenic diseases ( ). Synthesis of GWAS data with phenotypic data from the MGI and International Mouse Phenotyping Consortium (IMPC) databases recently enabled the identification of SNPs in 23 orthologous human genes involved in metabolic disease states and 51 new metabolism-associated genes ( ). Similarly, in a study of osteoporosis and bone mass regulation, GWAS performed in a panel of inbred mouse strains elucidated a QTL (quantitative trait locus, a portion of DNA that correlates with variation in a measurable phenotypic trait) for bone mineral density that mapped to a region including the lipoma HMGIC (high-mobility-group, isoform C) fusion partner ( Lhfp ) gene. Knockout mice created to study this gene also have increased cortical bone mass, identifying Lhfp as a negative regulator of bone progenitor cells and osteoblast activity ( ).

Confirming a genetic alteration by genotyping (i.e., defining a DNA change) does not necessarily guarantee its magnitude of expression. An intermediary step bridging genotyping and phenotyping is transcriptomic (mRNA) evaluation of an engineered gene. Evaluation of mRNA levels by northern blot assay, reverse transcription–polymerase chain reaction (RT-PCR), ribonuclease protection assay, and/or in situ hybridization (ISH) is required to confirm that an engineered construct is transcriptionally active. Sample selection is guided based on the anticipated distribution of gene expression (e.g., whole body vs. tissue-specific). Special care may be needed in analyzing RNA expression in certain tissues where the mRNA transcripts accrue in the cell body but the ultimate protein product is transported to a distant cell process.

Genotyping culminates with evaluation of the level of protein expression associated with the gene of interest, usually via Western blot analysis, enzyme histochemistry (EHC), and/or immunohistochemistry (IHC) (see Special Techniques in Toxicologic Pathology , Vol 1, Chap 11 ). An array-based approach to assess multiple tissue types in parallel often is useful. The apparent absence of protein detection may not reflect a lack of gene expression. Instead, protein expression levels might be sufficient to cause effects and yet not reach the sensitivity of the detection method. An example of this phenomenon is disruption of cell function or cell death that accompanies low-level expression of a biotoxin gene. Therefore, in some instances where transgene effects occur without detectable protein expression, it is necessary to use even more sensitive methods like proteomics to demonstrate very low levels of protein or RT-PCR or other transcriptomic methods to detect transgene-specific transcripts as a proxy for the extent of protein generation. As noted above, the distributions of mRNA and protein do not overlap in all tissues; for example, synaptic proteins at the neuromuscular junctions are specified by mRNA produced in neuronal cell bodies located in the central nervous system. Therefore, protein analysis is a preferred means of confirming appropriate gene expression if reliable reagents are available. Where feasible, assays that measure the levels of functionally active protein (e.g., EHC) may be preferable to tests that only establish the presence of the molecule (e.g., western blots, IHC). Expression of targeted genes may be evaluated by demonstrating the presence of an exogenous marker protein (e.g., bacterial β-galactosidase (βGal) [lacZ] or green fluorescent protein) inserted in the construct to confirm successful gene replacement ( ).

Maintenance of genetic fidelity in a breeding colony of genetically engineered animals is referred to as genetic quality control or genetic quality assurance. Genotyping is only part of genetic quality control, which also encompasses pedigree management and refreshing breeding stock. Genetic drift will generate one phenotypic mutation every 1.8 generations which, if breeding schemes are not properly managed, can become fixed in a colony after six to nine generations. When inbreeding mice, one impactful mutation is expected to emerge every 10 generations, and after 20 generations, a genetically distinct substrain is created. Poor genetic quality control can result in apparent loss or amplification of a desired phenotype over time that culminates in irreproducible data and degradation of the model. Further information about proper breeding practices in laboratory animals can be found in texts listed in the References section.

2.2

Phenotyping

Phenotyping is the characterization of any or all anatomic and physiological aspects of an animal model. While this often is considered in the limited sense of characterizing the effect of a specific genetic modification, a much broader approach is required to fully understand any animal model. The nature and method of genetic manipulation, the background strain (stock) and substrain, the source of the animals, husbandry practices (see Issues in Laboratory Animal Science that Impact Toxicologic Pathology , Vol 1, Chap 29 ), and microbial flora all can alter characteristics of the final animal model and should be considered when designing specific experiments. Though in the context of genotyping a transgenic animal is presumed to be genetically identical to its parental strain except at the altered gene loci, in phenotyping a genetically engineered animal—particularly in more complex models—is presumed to be different from its originating strain(s) until proven otherwise. Methods, resources, and study design considerations for phenotyping GEM and GER models are detailed elsewhere ( ; ; ; ; ; ; ).

To be useful for biomedical research, engineered animal models typically must develop a phenotype that is well correlated to the presence of the genotype. Expression of a modified gene may produce an anatomic change, functional alteration, both of these effects, or neither. However, caution must be used in deciding whether or not the engineered construct has defined the entire phenotype for a given gene. Multiple instances have been reported wherein subtle differences in the engineered gene have produced divergent phenotypes. For example, mice lacking cytochrome P450 (CYP) 1A2—a liver enzyme in humans and mice that metabolizes many exogenous chemicals and drugs—develop fatal pulmonary deficits in many but not all pups if a single exon is deleted, while the surviving single-exon knockout animals as well as mice in which four CYP1A2 exons have been removed exhibit no phenotype without the administration of a pharmacologic challenge ( ; ). In a similar manner, transgenic GEM models of Alzheimer's disease, which overexpress various mutant isoforms of amyloid precursor proteins (APPs), have been reported to develop amyloid plaques in the brain and cognitive deficits, although the extent of the anatomic and behavioral changes varies among models. For translational research purposes, the favored GEM models of Alzheimer's disease exhibit both robust functional and substantial structural abnormalities.

2.3

Large-Scale Phenotyping

Since 2002, large-scale, multicenter phenotypic screening efforts have sought to elucidate disease mechanisms by systematically generating and characterizing single-gene mouse mutant lines. An alternate name commonly applied to this analytical approach is “high-throughput” phenotyping. Foundational programs include the International Knockout Mouse Consortium, the European Union Mouse Genetics Research for Public Health and Industrial Applications, and the European Mouse Disease Clinic. Now, with advancing GWAS and PheWAS (phenome-wide association study) resources, the Mouse Genetics Project and IMPC carry on using the model of network pleiotropy (i.e., the ability of a single gene to produce two or more apparently unrelated effects) to understand the genotypic–phenotypic relationships ( ). To date, the IMPC has phenotyped 5861 genes in 6255 mutant mouse lines on the C57BL/6N background; this inbred strain was chosen as the test system due to its common use in biomedical research. The large-scale phenotyping endeavor is human resource intensive, requiring tremendous parallel research efforts in numerous fields of biology, and the need to collate data for multiple different organs and endpoints from measurements undertaken in many laboratories confounded early efforts to maintain the high-throughput nature of such studies. Multidimensional datasets demand involvement of bioinformaticians in these efforts. Prediction systems that link disease genes and phenotypic attributes in GEMs have been constructed (e.g., http://combio.snu.ac.kr/targo/ ). Large-scale phenotyping projects often adopt strategies similar to those described in the International Mouse Phenotyping Resource of Standardised Screens (IMPReSS) guidance document, which can be accessed through the IMPC ( http://www.mousephenotype.org/ ).

The objective of initial large-scale phenotypic screening is to rapidly identify the existence of functional and/or structural phenotypes using a defined number of tests, after which promising founders may be subjected to a more comprehensive characterization to better define the nature and mechanisms responsible for any phenotype of interest. These large-scale screening methods emphasize routine tests such as in-life functional evaluations (e.g., open field test, grip strength, electrocardiography, radiography), followed by anatomic pathology (macroscopic and microscopic analysis) and clinical pathology (e.g., hematology and serum/plasma chemistry) evaluations. Additional methods may be incorporated based on the research indication and availability of appropriate innovative technologies. Examples of such supplemental techniques include noninvasive imaging (see In Vivo Small Animal Imaging : A Comparison to Gross and Histopathologic Observations in Animal Models , Vol 1 , Chap 13 ) and various “omics” platforms (see Toxicogenomics: A Primer for Toxicologic Pathologists , Vol 1, Chap 15 ).

The goal of such large-scale phenotyping studies is to find one or more easily measured biomarkers that can be used in vivo to monitor the progression of disease and candidate therapies in both animal models and human patients. This approach also advances the ideal of precision medicine by identifying ideal biomarkers that are specific for a disease, or certain subpopulations of patients with that disease (see Biomarkers: Discovery, Qualification and Application, Vol 1, Chap 14 ). In these high-throughput screening research programs, new GEM or GER models often are phenotyped as young adults (6–16 weeks of age). An obvious pitfall is that age-related phenotypes and altered survival rates are missed in this paradigm. Initial phenotyping studies may begin with two to five animals/sex/genotype, while studies designed to probe subtle phenotypes and elucidate mechanisms of disease often require higher numbers per group (often 10 or more, especially to provide adequate power for clinical pathology assessments). For transgenic animals generated through microinjection, phenotypic analysis will include the F ₁ hemizygous (+) engineered animals along with age-matched controls (ideally nontransgenic littermates since they share the identical genetic background); inappropriate selection of control animals from a different substrain may confound interpretation ( ). For gene targeting projects, individuals selected for initial screening are homozygous knockout (−/−) animals and wild-type [+/+, or “normal”] littermates. Heterozygous (+/−) littermates or other GEM lines with differing levels of expression of the modified gene also may be included as the intermediate gene dosage(s) may yield a diluted phenotype ( ). In conditional models, additional genetic controls include animals carrying the floxed allele without recombinase and animals expressing the recombinase only. Animal numbers should be balanced in terms not only of age but also sex. In general, animals of both sexes are evaluated to be sure that sexually dimorphic phenotypes are found. Where time permits, phenotyping of gene-targeted animals may be delayed until the engineered model has been backcrossed for 10 or more generations to obtain a homogenous genetic background as phenotypes in GEM and GER lines may be magnified or masked in founders or earlier generations of progeny. Reviews that describe the common spontaneous lesions in mouse strains typically used in generating GEM models are an essential source of historical data ( ; ; ; ; ; ).

2.4