Animal Models in Toxicologic Research: Rodents

2.1

Overview of Species Selection for Toxicity Studies

Rodent species are an integral part of safety evaluation in toxicology programs to support product development and licensing. Rodent use for this purpose is detailed by various regulatory guidelines for biomolecule and small molecule therapeutics ( ; ), medical devices ( ), and agricultural and industrial chemicals (EFSA, multiple; U.S. Environmental Protection Agency, multiple). For safety assessment of small molecule and other xenobiotic therapies, additional data from nonclinical toxicity studies in nonrodent mammalian species are usually recommended for both oncology ( ) and non-oncology ( ) indications, but there are circumstances in which nonclinical data from a single pharmacologically relevant rodent species alone may be acceptable (e.g., for genotoxic drugs targeting rapidly dividing cells ( )). Pharmacologic relevance of the test species is required for regulatory toxicity studies conducted with biotechnology-derived pharmaceuticals like nucleic acids and proteins ( ) since toxicity is generally driven by exaggerated pharmacology and/or immunogenicity to the foreign (humanized or fully human) biomolecule ( ; ). Rodents should be evaluated and a rationale for their use considered if they are selected as a species for nonclinical testing of human-origin biotherapeutic products. Regulatory guidance on rodent/nonrodent species selection is summarized in Table 17.1 .

Different types of rodent toxicity studies are summarized in Table 17.2 . General toxicity studies in rodents for both therapeutic xenobiotics as well as agricultural and industrial chemicals are conducted in healthy, wild-type adult animals that are housed in autoclaved containers equipped with autoclaved bedding, free access to food and water, and often with environmental enrichment materials (e.g., exercise devices, nesting materials) designed to minimize stress. The test article is given according to the route by which humans might be exposed (commonly the oral, inhalational, or intravenous routes for therapeutic agents or the oral route for agrochemicals). Study designs of increasing duration are intended to identify the maximum tolerated dose, with intermediate (“mid”) and low doses to examine dose/exposure responsiveness and establish threshold (e.g., effect and no effect) levels that can be related to estimated efficacious and/or toxic doses.

For rodent carcinogenicity testing, there are now several options. Historically, test articles were evaluated in life-time bioassays in both mice (80 weeks) and rats (104 weeks), and less frequently in hamsters (104 weeks). These tests are characterized by large group sizes (usually 50 per sex per dose) to ensure sufficient survival for valid statistical analysis. High rates of spontaneous tumors occur in such studies for some rodent strains ( ; ); the kinds of tumors, their frequencies, and the target organs vary by the species and among strains, but significant amounts of historical control data are available to aid data interpretation. In contrast to the traditional use of life-time mouse and rat carcinogenicity studies, assessment of carcinogenicity now often involves abbreviated (26-week) bioassays using genetically engineered mice (GEM) with intrinsic sensitivity to potential carcinogens. The two mouse strains most commonly used for this purpose are the p53 ^+/− (formal strain designation: B6;129S7- Trp53 ^tm1Brd N5) and Tg-rasH2 (formal strain designation: CByB6F1-Tg(HRAS)2Jic) models ( Genetically Engineered Animal Models in Toxicologic Research , Vol 1, Chap 23 and Carcinogenesis: Mechanisms and Evaluation , Vol 1, Chap 8 ). The p53 ^+/− mouse model (in which cells lack one copy of the p53 tumor suppressor gene [involved in DNA repair]) is carried on a mainly C57BL/6 background and is generally used to detect genotoxic agents ( ). The Tg-rasH2 mouse model (which carries the c-Ha-ras oncogene plus its promotor/enhancer on a mixed BALB/c and C57BL/6 F ₁ background) is widely used to identify both nongenotoxic or genotoxic agents ( ; ; ). The key advantages of GEM models for carcinogenicity testing are that they require substantially less time for the in-life phase and the rate of spontaneous background tumors therefore is very low. While these features make interpretation of the final pathology data set easier, less historical control data are available for GEM models of carcinogenicity.

Other genetically engineered rodent models are employed on occasion for toxicity testing. These systems are used for materials with particular properties (e.g., the Tg.Ac mouse is sensitive to nongenotoxic materials applied to the skin ( )) or to investigate potential mechanisms of cell damage (e.g., Xpa and Xpa/p53 ^+/− mice, which have deficient DNA repair capabilities and so are more sensitive to genotoxic chemicals ( )). Engineered rodents also are growing in popularity as translational models for in vivo biomedical and toxicological research since substitution of mouse elements by insertion of human genes (“knock in”) or engraftment of human tissues permits evaluation of human cells and pathways when challenged by xenobiotic exposures ( ; ; ; ).

2.2

Rodent Species Used for Special Studies

In addition to general toxicity studies, a number of specialized studies are performed using rodents to address particular research questions. Key studies in this regard include various developmental bioassays—mainly developmental and reproductive toxicity, developmental neurotoxicity ( ; ), juvenile toxicity ( ), and extended one-generation reproductive toxicity ( ) studies (also see The Role of Pathology in Evaluation of Reproductive, Developmental, and Juvenile Toxicity , Vol 1, Chap 7 )—and target organ/systemic toxicity studies (especially immunotoxicity and neurotoxicity) that require integrated functional and structural assessments. Large treatment groups are necessary in such bioassays because functional and structural testing for a given dose may need to be performed in different cohorts to avoid any impact of functional testing (i.e., environmental enrichment) on the anatomic differentiation of the organ mediating that function. For this reason, rodents (and rabbits, which are not rodents but lagomorphs, see Animal Models in Toxicologic Research: Rabbit , Vol 1, Chap 18 ) are a default test system for these specialized studies ( ; ; ). Historical control data are useful when selecting and assessing the suitability of an animal model ( ; ; ).

Such specialized studies typically are done using outbred mice (e.g., CD-1) or rats (e.g., SD and Wistar Han). These species and stocks are utilized to mimic the heterogeneous genetic backgrounds in most human populations. Hamsters may be selected as the rodent species in some cases if the ADME profile for the test article in mice or rats differs substantially from that of humans. In general, guinea pigs are not employed for such specialized studies because the precocial offspring (i.e., born with well-developed locomotor and sensory capabilities as well as the ability to consume solid food at birth) are not good models for the developmental status of altricial (i.e., underdeveloped at birth) human infants and toddlers and because the pattern of neurotoxicity in guinea pigs for classic neurotoxicants can differ from that of other rodent species ( ). However, guinea pigs may be used as a system for specialized studies if the test article is not pharmacologically active in other rodent species ( ).

Human populations are often subject to comorbidities (e.g., cardiac and/or metabolic diseases such as hypertension, diabetes mellitus, obesity), conditions which may increase the manifestation of toxic effects. Various RMD (e.g., leptin-deficient [ob/ob] and nonobese diabetic mice for diabetes and obesity), whether spontaneous or genetically engineered, may be used to recapitulate such comorbidities in the face of toxic challenges (see Section 2.3 below). A more specific, controlled option is inclusion of a challenge (i.e., stress testing) to increase the functional activity and raise the likelihood of detecting a deficit or toxicity that is subthreshold in healthy, unstressed animals. For example, the heart may be stressed by exercise, mechanical, or pharmacologic approaches as a means to more sensitively detect cardiotoxicity. Dobutamine infusion causes beta-adrenergic stimulation, increasing heart rate, blood pressure, stroke volume, and cardiac output, with measurable echocardiographic changes ( ). This approach has shown to have utility in the early detection of functional cardiac effects of diesel exhaust and the functional and structural cardiotoxicity of anthracycline ( ; ). In general, however, toxicity testing in rodents is performed in normal animals that lack comorbidities (except for spontaneous/age-related changes) as the primary study objective is to define potential toxicity of a test article to the mainstream population rather than particularly susceptible subpopulations.

An important consideration for rodent-based testing in terms of best modeling the human population as a whole is the feeding regimen. A substantial proportion of adults in developed countries—often 40% or more—are overweight or obese. The ranges of background findings as well as the lifespans of mice and rats differ substantially depending on whether they are fed ad libitum or are given a caloric-restricted diet ( ). The responses of ad libitum versus caloric-restricted rats to toxicants also diverge ( ). Thus, the choice of feeding schedule is one factor that must be considered with care when seeking to separate the potential toxicity of test articles from confounding husbandry effects.

2.3

Rodent Models of Disease and Genetically Modified Animals

In some cases, a toxicity assessment may represent only one element of the study design. A common multipurpose study design is the combined efficacy/toxicity study, where the experiment is conducted in a RMD and the data set for the test article will address simultaneously such questions as a test article's efficacy (“Did it work?”), toxicity (“Was it safe?”), and the cellular and molecular effects underlying these responses. Such experimental designs are common for investigating monogenetic congenital diseases, many of which have reproducible and somewhat consistent disease phenotypes that may be examined using conventional anatomic pathology and/or clinical pathology methods.

The choice of RMD for use in toxicity testing depends on the experimental question(s) to be addressed. In most cases, RMD are used to test a hypothesis related to the disease pathogenesis (e.g., infection, induced surgical, metabolic, or immune-mediated pathology) and/or efficacy of a potential new therapy, while wild-type rodents are employed for standard toxicity testing to identify hazards and assess safety. This dichotomy reflects several practical factors. Many RMD exist on inbred (i.e., genetically homogeneous) backgrounds, which do not provide a good representation of more genetically diverse human populations. Second, RMD with their underlying disease manifestations may have truncated lifespans or exhibit substantial inherent anatomic, functional, and/or molecular abnormalities which have nothing to do with test article exposure and yet might be difficult to differentiate from any effects actually induced by the test article, particularly if data on spontaneous background lesions are lacking. Third, a fair percentage of RMD have limited fecundity, which will require an extended breeding program or enrollment period to produce enough healthy animals to populate multiple test groups. Fourth, for some RMD, each individual animal—especially for new spontaneous or genetically engineered lines—may be expensive to acquire and maintain. Finally, in many cases, the RMD will need to be characterized in detail prior to its use for toxicologic research due to the absence of historical control data.

Genetically engineered RMD are especially popular systems for product discovery and development, whether such mutations arise spontaneously or by deliberate engineering ( Genetically Engineered Animal Models in Toxicologic Research , Vol 1, Chap 23 ). Genetically abnormal RMD are used most often to investigate a particular hypothesis, such as exploring molecular mechanisms that contribute to disease initiation and progression; to validate therapeutic targets; or to confirm the efficacy of novel drug candidates. Historically, most genetically engineered RMD have been produced in mice, for two reasons. First, transgenesis (i.e., insertion of novel DNA carrying a gene of interest) is relatively easy in mice due to the very large pronuclei in zygotes (one-celled embryos) of some strains (e.g., FVB). Similarly, mice traditionally have been preferred for gene targeting because reliable embryonic stem (ES) cell lines have been harvested from several strains (e.g., 129, C57BL/6) that permit fairly reliable albeit low-yield homologous recombination to remove (“knock out”) an existing mouse gene or to replace it (“knock in”) with a homologous gene from a different species (including humans). Rats have been used for transgenesis, but the absence of a reliable rat ES cell line has limited production of knockout rat models. The recent advent of nuclease-based genome engineering technology (e.g., the clustered regularly interspaced short palindromic repeats [CRISPR]/CRISPR-associated protein 9 [Cas9] platform) will revolutionize the production of RMD since this procedure can be performed to edit genes in any species ( Genetically Engineered Animal Models in Toxicologic Research , Vol 1, Chap 23 )—and already has been so employed in many rodent (and nonrodent) species.

Various RMD are utilized for more development-oriented tasks in which the main point of the study is to describe a test article's properties (i.e., investigational toxicology) rather than to investigate a specific hypothesis. As noted above, RMD may be employed to simultaneously evaluate a test article for both predicted efficacy and undesirable toxicity, generating a therapeutic index in that model. Such models have been used to support safety assessment in a variety of disease contexts, from inborn errors of metabolism (e.g., Niemann–Pick Disease) to osteoporosis and mouse tumor models for oncolytic virus therapies, and to investigate human toxicities (e.g., Zucker diabetic fatty rats for exocrine pancreatic injury with glucagon-like peptide-1 agonists) ( ). Immunodeficient RMD, whether spontaneous or purposefully engineered, are also critical systems for evaluating human tissue responses in vivo. Such RMD typically are produced in immunodeficient mice either by introduction of viable human cells (e.g., Pdx [patient-derived xenograft] or “avatar” mice) or by chemically induced ablation of normal mouse cells bearing a toxicant-responsive gene followed by their replacement with human cells from the same tissue (e.g., “chimeric” mice with humanized bone marrow, liver, and thymus) (see Genetically Engineered Animal Models in Toxicologic Research, Vol 1, Chap 23 ). The utility of such RMD is to evaluate questions including the degree to which chemotherapeutic regimens selectively target human-origin tumor cells (Pdx model) and the ability of human metabolic enzymes to affect the pharmacokinetics and tissue manifestations of toxicity of test articles in vivo (chimeric model).

3.1

Pharmacologic Translational Relevance and Interspecies Pathophysiologic Concordance

Pharmacologic relevance of a species (i.e., whether or not the potential therapeutic being tested is active at the rodent [or nonrodent] receptor) is a major determinant (i) when selecting rodents (vs. other species; see Section 2.1 ) for pharmacodynamics, efficacy, and toxicity studies and (ii) in the evaluation of on-target but exaggerated pharmacologic effects in safety studies. This concept is applicable to all xenobiotics when interpreting the potential human relevance of findings in animals and is particularly important when testing the safety of biomolecule therapeutics (i.e., xenobiotics containing human-derived protein or nucleic acid sequences that are highly target-specific, metabolized differently to chemical entities, and likely immunogenic in animals ). Once pathophysiological changes have been identified and characterized in rodent models, their relevance, or translatability, to humans must be verified.

In the efficacy context, RMD (mainly in mouse and rat) are a key part of the data package to support the hypothesis that administration of a novel test article will benefit the patient population with a particular disease. It is vital to understand the comparative pathophysiology and the extent to which any pharmacologic effect in rodents may be predictive of human-relevant mechanisms and ultimately a clinically beneficial outcome. This process ideally starts by engaging the pathologist from the earliest stages of target validation and is applied through animal model selection, model characterization, and interpretation of experimental data. Relevant RMD may be engineered (e.g., via genetic manipulation), induced (e.g., by surgery), or spontaneous. In general, any given RMD only incompletely recapitulates the disease as seen in humans; this tendency is especially clear for many degenerative RMD such as engineered mouse models of Alzheimer's disease (e.g., bearing human transgenes encoding mutant forms of amyloid-β precursor protein, apolipoprotein E4, and/or presenilin 1) or osteoarthritis (e.g., null mutations in collagen 2a1 [ Col2a1 ] or osteoprotegerin [ Tnfrsf11b ] genes). Key questions can help the pathologist in assessing the translational value of an RMD, and its limitations, in the context of testing mechanistic hypotheses regarding efficacy. How much is known about the cause, pathophysiological mechanisms, and progression of the human disease? How closely does the RMD recapitulate most of the major aspects of the human condition? What are the similarities and differences in the pharmacologic target and pathway biology in rodents versus people? How reproducible are the pathobiological endpoints in rodents and humans? Has a given RMD been used historically, and how well has it predicted human pharmacologic responses? The pathologist plays an essential role in both understanding comparative pathophysiological mechanisms and compiling evidence of efficacy.

In the toxicology context, evaluation of translational relevance occurs stepwise through hazard identification and characterization to subsequent risk assessment for human (and sometimes animal) populations. Hazard identification and characterization is the phase of development where effects resulting from tissue, organ, or systemic injury caused by a test article are detected (identification) and then explored with respect to their progression, impact (harmful [“adverse”] or not harmful), relationship to dose (or more accurately exposure, based on toxicokinetic data), and ability to be reversed or repaired (which collectively represent characterization). In terms of toxicologic pathology endpoints, macroscopic and histopathologic evaluation, organ weights, and clinical pathology findings (hematology, serum chemistry, and sometimes urinalysis and coagulation assays) are the main contributors to the study data set. These pathology data typically are correlated to clinical signs, physical examinations, or other allied evaluations of functional perturbation.

Biological responses in RMD to novel xenobiotics have been demonstrated over decades to be generally reasonable predictors of human responses, though translational deficiencies are recognized in their ability to predict clinical trial outcomes ( ). Standard rodent toxicity studies typically evaluate test articles administered to achieve continuous exposure during the dosing phase, ranging up to relatively high doses (up to lifetime exposure) in normal animals. In contrast, human treatment/exposure events typically occur intermittently at lower doses over a much longer absolute period, and in the case of novel therapeutic candidates are administered to patients with a disease and sometimes one or more comorbidities. Anatomic or physiologic differences may limit the translational relevance of findings in rodents to predicting human responses. For example, the presence of lesions in the nonglandular region of the stomach of the rodent is considered to be of minimal relevance for human risk assessment. This structure in rodents may be directly affected by application of test article via gavage dosing, but the nonglandular region does not have a counterpart in humans. Similarly, thyroid follicular hypertrophy/hyperplasia associated with induction of hepatocellular enzymes (specifically uridine diphosphate–glucuronyl transferase [UGT]), leading in rodents to increased metabolism and inactivation of thyroid hormones T ₃ and T ₄ and upregulation of the feedback loop to increase thyroid stimulating hormone secretion, is considered to translate very poorly to humans. This divergence is due to the presence of thyroxine-binding globulin in humans, which increases the circulating half-life of T ₃ and T ₄ . That said, substantive interspecies differences do not always minimize the utility of rodent-derived data for human risk assessment. In the brain, the numbers and organization of cerebrocortical neuron populations in the lissencephalic rodent brain differ significantly from those of gyrencephalic species like dogs and primates; similarly, the usual range of rodent behaviors is substantially less complex than the behavioral and psychological make-up of humans. Despite this, rodent neurotoxicity data acquired by toxicologic pathology analysis tend to be a suitable screen for human risk assessment. Toxicologic pathologists have a central role in determining the significance of these structural and functional differences with respect to the predictivity of toxicologic findings in rodents to responses that might occur in nonrodents and humans.

Evaluating concordance between findings in animal and human studies (and also between rodent and nonrodent species) to establish translation is therefore vital to the risk assessment enterprise. Primary goals of rodent studies are (1) to establish evidence of efficacy with greater confidence and to minimize human (or animal) risk, particularly in the context of specific modalities for treating disease, and (2) to improve the mechanistic understanding of targets/pathways responsible for toxic outcomes following exposure to potential therapeutic agents or environmental chemicals. In toxicology, undesirable findings (i.e., “toxicity”) may be target-mediated (“on target”) exaggerated pharmacology, side (“off target”) effects due to test article activation of other biological pathways, or nonspecific (“chemical”) effects associated with certain biological or physico-chemical properties of a test article. Additional caveats apply when assessing the translatability of rodent data to humans, notably that true-positive results are identified infrequently in both animals and humans since

• (i)

  most xenobiotic test articles do not progress to human exposure because development of those with significant hazards/with small safety margins is terminated as they have an unfavorable risk assessment, and

• (ii)

  effects seen at high doses/exposures in rodent toxicity studies may not manifest at or fully recapitulate the toxicity profile observed in humans faced with much lower exposures.

The differences in exposures in rodent studies versus humans also mitigate against the establishment of true-negative outcomes for a given finding, since the presence of another/different toxicity may ethically limit human dose escalation in clinical studies. Practically, this inherently imbalanced program design limits informative data in humans to toxicities which have been shown during nonclinical safety testing in rodents to be both readily monitored using sensitive, predictive biomarkers and that are reversible, with suitably large safety margins.

Despite these limitations, various evaluations have shown the utility of rodent-derived data as a means of predicting human responses. For example, physiological responses in genetically engineered rodents where null mutations (knockouts) ablate a particular molecular pathway are good predictors of therapeutic efficacy for test articles designed to reduce the activity of that pathway in human patients ( ). In contrast, rodent (primarily rat) toxicity data show a positive human toxicity concordance rate of only 43% compared to 71% for rodent and nonrodent species together and 63% for nonrodent species alone ( ). The concordance and/or predictivity of rodent data for human responses varies with the organ system being assessed and the data set undergoing analysis, though there is general agreement among different published analyses ( ; ; ). For rodent studies, the types of human toxicity with the highest concordance—biochemical (e.g., clinical chemistry findings), cutaneous, gastrointestinal, hematological, and hepatic effects—are those best identified across all species used in regulatory toxicity studies. There are some exceptions noted among species: rodents have the highest predictivity of any test species for biochemical endpoints of toxicity, whereas cardiovascular toxicity shows better concordance in nonrodent species ( ; ).

Looking at other comparative indices, rodent toxicity findings provide a suitable screen for human risk assessment. Across all organ systems, toxicity findings in rodents have (i) a similar low sensitivity to that of nonhuman primates (NHPs), approximately 26% in rodent, versus 42% in dogs; (ii) a similar high specificity to those of both dogs and NHPs (approximately 88%); but (iii) the lowest positive predictive value for toxicity (29%) compared to approximately 39% and 47% in dogs and NHPs, respectively ( ). Conversely, the absence of target organ toxicities in rodent or nonrodent test species strongly predicts a similar outcome in humans (a negative predictive value of 86%), an important factor in risk assessment ( ). A “big data” approach to a much larger dataset (3290 approved drugs) showed that the highest proportion of “true positive” concordance between toxic responses in humans and rats for given organ systems was in endocrine, hematological and lymphatic, hepatobiliary, metabolic, renal and urinary, respiratory, and thoracic and mediastinal systems ( ). For mice, the highest proportion of “true positives” with respect to predicting human toxic responses also included endocrine, hematological and lymphatic, hepatobiliary, respiratory, and thoracic and mediastinal organ systems; however, nervous system as well as skin and subcutaneous organ responses also could be extrapolated in this species ( ).

Another recent review explored whether both rodent and nonrodent species are necessary for general toxicity testing as currently described by regulatory guidance seeking toxicity data from two species. Reduction to a single species for longer-term toxicity studies was only applied for 8/133 drug candidates but might have been possible for more, regardless of drug modality, as the same or similar target organ toxicity profiles (rodent vs. nonrodent, across all organ systems) were identified in the short-term studies for an average of approximately 39% of candidates ( ). These data suggest that where there is concordance between rodent and nonrodent findings in shorter term toxicity studies, a single species might be sufficient to investigate chronic toxicity; at this point, it is unclear whether this single-species approach could be based on rodent testing (which would be preferable given the larger possible group sizes, lower cost, and reduced regulatory burden associated with rodent toxicity studies). Extensive further analysis of larger datasets is required to more fully understand the context-dependent benefits/risks of such an approach.

In conclusion, the translational relevance of safety assessment via rodent toxicity studies alone is variable across different organ systems and is generally less predictive than nonrodent data when each species is interpreted in isolation. However, when combined, an integrated analysis of rodent and nonrodent toxicity data improves the strength of the risk assessment in predicting human responses. Importantly, an absence of findings in rodents has a high negative predictive value for the risk of toxicity developing in subjects exposed during human clinical studies. Taken together, these data indicate that the utility of rodents as test systems depends on the nature of the scientific question, but that rodent data are vital factors in safety assessment.

3.2

Biological and Cell Therapies

Biopharmaceutical test articles such as monoclonal antibodies and peptides ( Protein Pharmaceutical Agents , Vol 2, Chap 6) as well as antisense oligonucleotides ( Nucleic Acid Pharmaceutical Agents , Vol 2, Chap 7) are developed for efficacy in human-specific biologic systems. Toxicity of human-derived biomolecules in animals manifests essentially as either (1) exaggerated pharmacologic activity, which is likely to be a translationally relevant effect in a pharmacologically relevant test species ( ), or (2) immunologic reactions (e.g., antidrug antibodies [ADA] and immune complex disease) as the animal immune system recognizes the human molecules as antigenic, which are less translationally relevant. Even if ADAs do not cause pathologic changes, neutralizing antibodies can limit exposure to, and pharmacologic activity of, the test article, thereby restricting assessment of toxicity. Ideally, any rodent species used in toxicity studies for biomolecules will be pharmacologically relevant; if so, then testing in a rodent species (as well as a nonrodent species) would be expected to meet current regulatory guidance ( ). If the toxicity profile in short-term studies supporting Phase I clinical trials is similar in both animal species, then longer-term toxicity studies in a single species, preferably rodent, may be sufficient to support Phase II/III clinical trials for biotherapeutic candidates ( ).

However, due to the high selectivity of biopharmaceuticals, even if there is activity of a human-derived molecule at the rodent receptor, differences in activity and knowledge of downstream [patho]biology in the rodent and human orthologs are important to account for when undertaking risk assessment. Often, only one pharmacologically relevant species, generally an NHP (due to higher DNA/amino acid sequence homology with the human protein), can be identified. In such cases, a surrogate, rodent-specific version of the human therapeutic (e.g., an antibody against the murine target) may be generated since it should have a similar impact (pharmacologic and toxic) in the rodent test system as the human-specific therapeutic candidate would in patients. However, since the intended human therapeutic is not actually being tested in rodents, it may be preferable to only use the nonrodent species to produce animal data suitable for human risk assessment.

Evaluation of toxicity for cell therapies (e.g., engineered stem cells to replace a depleted cell population or modified T-cells for tumor immunotherapy) is very challenging in rodent species ( Stem Cells and Regenerative Medicine , Vol 2, Chap 10). Immunodeficient mouse models, where genetically altered strains with multilineage defects in leukocyte function are “humanized” by transplanting functional human immune cells (e.g., by infusing human peripheral blood mononuclear cells or seeding the bone marrow with human hematopoietic progenitor cells), can be used to explore human-specific responses to cell therapies, with the understanding that the human cells are acting in the context of a murine host system ( Genetically Engineered Animal Models in Toxicologic Research , Vol 1, Chap 23 ). In efficacy models involving xenografted cells, species differences are less of an issue since the human tumor cell lines or Pdx grafts are hosted by the rodent but the pharmacologic target against which efficacy being tested is of human origin.

Testing of antibody–drug conjugate commonly involves a two-species approach for safety assessment irrespective of pharmacological relevance in the rodent ( ). The reason for this testing paradigm is that evaluation for off-target toxicity is feasible in rodents via pathologic evaluation of tissues for toxic effects associated with the conjugated drug. Use of a pharmacologically relevant nonrodent species (usually NHPs) in which the antibody backbone is active permits assessment of both on-target and off-target toxicities ( ).

3.3