Alternative Strategies for Toxicity Testing of Species-Specific Biopharmaceuticals

Abstract

Although toxicology studies should always be conducted in pharmacologically relevant species, the specificity of many biopharmaceuticals can present challenges in identification of a relevant species. In certain cases, that is, when the clinical product is active only in humans or chimpanzees, or if the clinical candidate is active in other species but immunogenicity limits the ability to conduct a thorough safety assessment, alternative approaches to evaluating the safety of a biopharmaceutical must be considered. Alternative approaches, including animal models of disease, genetically modified mice, or use of surrogate molecules, may improve the predictive value of preclinical safety assessments of species-specific biopharmaceuticals, although many caveats associated with these models must be considered. Because of the many caveats that are discussed in this article, alternative approaches should only be used to evaluate safety when the clinical candidate cannot be readily tested in at least one relevant species to identify potential hazards.

Although toxicology studies should always be conducted in pharmacologically relevant species, the high degree of species specificity of many biopharmaceuticals can present certain challenges in identification of a relevant species. This specificity also implies that the toxicity is generally due to on-target activity (based on the intended pharmacology) rather than off-target effects (nonspecific binding) as is often the case for traditional small molecule pharmaceuticals. ICH S6 ¹ defines a pharmacologically relevant species as “one in which the test material is pharmacologically active due to expression of the receptor or an epitope (in the case of monoclonal antibodies).” In some cases, it may not be possible to evaluate the toxicity of a clinical product in animals (for example, when the product is active only in humans or chimpanzees, or if immunogenicity limits the ability to conduct a thorough safety assessment). Alternative approaches to evaluating the safety of a biopharmaceutical should be considered in these cases. ICH S6 has identified the following as potentially viable alternative approaches: animal models of disease, transgenic/knock-out models, and surrogate molecules (Table 1). Each of these has unique advantages and disadvantages that must be considered in the development of preclinical safety assessment strategies (Table 2). Multiple approaches to understanding the toxicities of a development compound may be equally valid, provided that they can be scientifically justified as relevant to understanding the potential safety of the molecule. This article will discuss some of the benefits and limitations of each approach and will provide a series of examples from approved products as well as current development programs on how these alternative approaches have been used. Although most of the discussion will focus on protein biopharmaceuticals (eg, recombinant proteins, monoclonal antibodies, and fusion proteins), the principles for alternative testing approaches are also relevant for nonprotein biopharmaceuticals (eg, peptides, antisense oligonucleotides, small interfering RNAs, and aptamers) and could potentially be applied to some new chemical entities (eg, small molecules showing restricted species specificity).

Animal Models of Human Disease

One challenge in identifying a pharmacologically relevant species for biopharmaceutical toxicity testing is that, although the target may be expressed in different animal species and may even have comparable function as in humans, the target may not be recognized (pharmacologically modulated) by the clinical candidate. Alternatively, the therapeutic target may be expressed or upregulated only in certain disease settings. Examples include nonhost proteins such as viral or bacterial antigens that are not normally present in healthy animals, or host proteins that are expressed only in certain disease states such as unique cancer antigens, Rh factor, or altered (mutated) host proteins. Because such targets are unlikely to be present in normal animals, these represent perhaps the most challenging targets for which to assess the pharmacologic and toxicologic effects of the clinical candidate.

Animal models of human disease have been used not only to facilitate compound selection and allow for an early understanding of mechanism of action but also to assess safety. The use of animal models of disease to assess in vivo activity as well as toxicity can provide a better understanding of the therapeutic index and therefore improve clinical dose selection. ² Animal models of human disease may exist spontaneously (eg, dogs with factor VIII deficiency) or may be experimentally developed for preclinical testing. Common examples of the latter include the use of virally challenged animals to test the efficacy of a vaccine or a therapeutic directed against viral proteins, mice inoculated with xenogenic (human) tumors expressing the target antigen, mice challenged with infectious agents, or genetically modified animals that develop spontaneous disease. As with any model, there are both advantages and disadvantages to this approach (Table 1). One key consideration is that the immune system may be very different between rodents and humans. ³ For example, animal models of many inflammatory diseases such as multiple sclerosis, where complex multicomponent processes are involved, may differ substantially between mice and humans. ⁴ In addition, animal models of disease are generally not well characterized to understand what the “background” lesions are relative to what may be a compound-related effect.

Genetically Modified Animals

Genetically modified (knock-out, knock-in, transgenic) mice are rapidly gaining acceptance as tools for mechanistic research and validation of target biology and have considerable potential as specific models of toxicological importance. Gene-targeted or knock-out (KO) animals have been created using molecular and cellular genetic engineering techniques to produce animals that specifically lack an endogenous gene and therefore fail to express the related protein(s), whereas transgenic (Tg) mice are engineered to overexpress a target protein. ⁵ Many KO and Tg mice appear morphologically normal although functional abnormalities may be apparent; in other cases, these mice may exhibit a normal phenotype. Pharmacological challenges or other physiological stressors may also unmask subtle phenotypes (functional and/or morphologic changes resulting from the genetic engineering event). ^6,7 In addition, alteration of a target that has a significant role in prenatal growth and development may have significant adverse effects (eg, may be embryolethal) that are not relevant when inhibiting the target in an adult. Embryolethal effects can be avoided through development of conditional KO animals in which target disruption is limited to a specific developmental stage and/or tissues of interest. ⁸

Knock-out mice have been used to assess drug specificity, investigate mechanisms of toxicity, and screen for mutagenic and carcinogenic activities of therapeutic candidates. Similarly, the effect of target blockade by novel therapeutic candidates can be estimated in KO mice; for example, generation of viable and fertile animals with null mutations for a potential target protein may suggest that pharmacological inhibition of the molecule in vivo is unlikely to elicit major adverse effects on normal physiological functions. ⁹ As such, an apparent lack of a deleterious phenotype could be used as supportive evidence for the safety of inhibiting the intended target following extensive characterization of the KO mouse.

However, because KO mice may have uncharacterized compensatory mechanisms or redundant pathways that are not readily apparent to replace the function of the absent protein(s) or target, ^10–12 their use in assessing safety will likely be supportive rather than providing definitive safety data. One possible alternative to avoid this issue could be directed toward utilizing conditional KO mice, which permit evaluation of effects resulting from inhibition of a particular gene product during only the relevant stage of life. Finally, when considering animal models of disease using a complete or partial gene KO, it is important to consider the genetic structure and function of the gene, because genetic mutations underlying human disease may have a different physiological effect in the rodent because of differences in gene structure and/or function between species. ¹³

“Humanized” knock-in (KI) animals, in which the human gene is inserted into the mouse genome (either independently or in conjunction with a knock-out of the endogenous mouse gene), are of particular utility in evaluating the efficacy and toxicity of human biopharmaceuticals that are not pharmacologically active in normal rodents. ^14,15 One criticism of this approach is that humanized mice may express one or a few human proteins of interest, but other proteins that interact with the human molecules are still of mouse origin. The physiological effects of human-mouse protein interactions may differ slightly—or substantially—from that of the normal human-human association (eg, Toll-like receptors ^16,17 ). It is important to note that in several cases, KI mice have been used to support appropriate function of the human KI gene. ^13,15 Prior to use, comprehensive studies must be conducted to define the biology of mouse-human protein interactions to allow validation of the human gene KI mice as appropriate models. As transgenic insertions may not be targeted to a specific site in the genome, some of the regulatory sequences around them may function differently than those of the endogenous gene, which may affect both spatiotemporal patterns as well as overall levels of protein expression. In addition, administration of the clinical candidate in these animals may still be immunogenic thus limiting long-term studies in KI mice.

With the increasing number of biopharmaceuticals on the market, data from studies in genetically engineered mice will become important to demonstrate that they have utility as a viable alternative to testing in nonhuman primates and that findings of efficacy and toxicity obtained using these models are relevant to humans. This will take time, because the ultimate validation will not occur until there are clinical data to compare with relevant preclinical models.

Despite their potential utility, the use of KO/Tg mice for toxicity studies may be technically challenging because of the activity of the target (ie, may be different in mice and humans) and/or early embryonic lethality. In cases of embryonic lethality, development of alternative models, such as the BRCA1 and BRCA2 conditional KO animals, may enable a more appropriate evaluation of target inhibition in the adult animal. ⁸

The technical and feasibility challenges associated with using KO/Tg mice may significantly affect program time lines, and consequently, the potential use of KO/Tg mice in a safety assessment program must be considered very early in development. As previously discussed, standard KO mice that lack a target throughout embryogenesis and development may not accurately reflect disease or pharmacological interventions in the clinical scenario, where protein function may be nullified only during adulthood. Another important consideration when using a KO/Tg mouse is that they are often generated on different background strains (which may produce different phenotypes) or on strains that have not typically been used in toxicity assessments (eg, C57Bl/6 and S129), and therefore less historical data may be available. Multigenerational studies or increased numbers of KO/Tg control animals may therefore be required for interpretation of safety results in the absence of adequate historical data. This may be particularly pertinent in reproductive toxicity assessments, where fetal variations or abnormalities may be strain dependent. For less well-characterized strains of mice, it may be necessary to rely on the characterization of the wild-type mice as well as the genetically modified mice. A similar situation exists for pharmacology or disease models, which are notoriously specific for certain strains of mice. Immune function in these strains is generally not well characterized and specific strain differences are known to exist, especially in the immune response. ¹⁸ Therefore, the pharmacologic activity of the biotherapeutic must be evaluated in the KO/Tg animal prior to committing to its use for safety assessment. Last, genetically modified mice are often developed in discovery or academic research settings, where the incidence of latent or opportunistic pathogenic infections tends to be higher, and may necessitate rederivation of the strain prior to scaling up their production. This could potentially result in a 1- to 2-year delay before the model would be ready to use in a safety assessment setting. During the rederivation process and throughout the use of a KO/Tg model, it may be necessary to genotype each animal to ensure that hetero- or homozygosity and/or copy number are appropriate for the desired genetic modification. There are also certain KO/Tg mice that are sterile or not viable as homozygotes, so the animals must be generated from heterozygotes and therefore must be genotyped each time. Finally, as all of these models have strengths and limitations, the scientific rationale for their use must be justified and thoroughly discussed (generation of the genetically modified animal, backcrossing stage, strain, etc).

Surrogate Molecules

Considerations for Use of a Surrogate Molecule for Safety Assessment

For the purpose of this discussion, a surrogate molecule is one that is selected for use in place of the development compound to evaluate pharmacology and/or toxicology parameters in a species that is believed to be biologically relevant to humans. These molecules have also been referred to as analogous or homologous compounds. However, until the clinical candidate has been evaluated in humans, the extent to which the surrogate molecule is truly homologous or analogous to the clinical candidate cannot be completely understood. Of course, the same may be said with respect to the biological relevance of the animal species selected for testing of the clinical candidate. Although all analogous or homologous compounds are surrogates for the clinical candidate, not all surrogates will be proven to be analogous or homologous (ie, a monoclonal antibody [mAb] to the animal target may be a completely different molecule). These caveats should be kept in mind when interpreting the results of any alternative testing strategy.

Surrogate molecules may enable an evaluation of the consequences of modulating the target-mediated pharmacologic activity of the drug in cases when an evaluation of the clinical candidate is limited by species specificity. The surrogate molecule should resemble the clinical candidate as much as possible with regard to pharmacologic activity (ie, target epitope recognized, binding affinity, potency, etc) if it is to provide relevant and useful safety information. Because the formulation, production process, range of impurities/contaminants, glycosylation patterns, and many other factors could also influence the findings on a study, one should also understand the essential quality characteristics of the molecule. This may be challenging and will require significant resources to characterize a molecule that will never be a drug candidate.

Although in some cases surrogate molecules have emerged as scientifically accepted tools for the safety assessment of biopharmaceuticals, surrogate molecules inherently differ from the clinical candidates they are designed to represent. As such, several questions regarding the validity of a surrogate molecule for toxicity assessments, as well as the overall advantages and disadvantages of using a surrogate molecule over the clinical candidate, should be taken into consideration. Both in vivo and in vitro assessments should be used to characterize the relative similarity of the surrogate molecule to the clinical candidate. An important consideration for certain monoclonal antibodies is use of the appropriate isotype for the surrogate, particularly for those with anticipated ADCC or CDC activity as part of the mechanism of toxicity or pharmacologic action.

In vitro assessments should include pharmacologic activity, potency, affinity/avidity, as well as general physicochemical characterization. In addition, consideration must be given to the appropriate isotype to understand the similarity of ADCC or CDC activity of the Fc region of the antibody. Similarities in tissue binding and specificity may also be useful in assessing the appropriateness of the surrogate molecule. In vivo evaluations of the pharmacokinetics and pharmacodynamics of the surrogate molecule should also be conducted, with comparisons drawn to the clinical candidate where possible. Investigations into biomarkers of the surrogate molecule’s activity and/or application of the surrogate molecule in models of efficacy may also aid in further characterization.

Limitations in the utility of surrogate molecules are dictated by the nature of the molecule itself. The surrogate molecule should be comparable in pharmacologic activity as well as structure, to the clinical candidate. However, in some cases, the surrogate molecule may be structurally distinct from the clinical candidate but have similar pharmacodynamic effects in vivo. This is often a requirement if significant interspecies variability exists in the target molecule structure, but may also occur when the development of the surrogate molecule is conducted either separately from the clinical candidate or using different manufacturing methods. In cases where an animal model lacks the human target, a surrogate molecule may be developed to mimic biologically equivalent pharmacologic effects rather than specific target recognition. An example of this might be a molecule targeting murine keratinocyte (KC) protein as a surrogate for a clinical candidate directed against human IL-8. ^19,20

Additional limitations of the surrogate molecule may exist with respect to the magnitude or specificity of its biological effect, if the intended pharmacological effect is either amplified or reduced relative to the clinical candidate. It is important to be aware that surrogate molecules may elicit different downstream signaling cascades that could yield potentially significant, yet irrelevant, effects. Differences in either pharmacologic effect or molecular structure can also affect the pharmacokinetics of the surrogate molecule. Thus, when the pharmacokinetic profile of the clinical candidate is known in a relevant animal species, relative comparisons between the surrogate molecule and the clinical candidate should be made, taking into consideration species differences in affinity, target distribution, or differences in other clearance processes (ie, Fc-mediated clearance) that may impact pharmacokinetics. When the pharmacokinetic profile of the clinical candidate in an animal model is unknown due to species specificity, the similarity in pharmacologic effects between molecules (assessed using in vitro or in vivo data) can be used to justify comparability.

When a surrogate molecule believed to be representative of the clinical candidate is identified, and a decision to use it to support development of the clinical candidate is made, the sponsor has effectively assumed the responsibility of codeveloping 2 molecules, only one of which will be tested in the clinic. Several factors should be considered when assessing the feasibility of surrogate molecule codevelopment with the clinical candidate, including the requirements for development, manufacture, and characterization of the surrogate molecule. The extent to which the surrogate molecule will be manufactured to Good Manufacturing Practice (GMP) guidelines or tested in safety studies compliant with Good Laboratory Practice (GLP) guidelines will need to be decided upon, with appropriate justification(s). Development, including validation (or qualification), of different assays to detect the surrogate molecule and to monitor the effects of immunogenicity (ie, anti-product antibodies) is also necessary. As production processes associated with manufacturing a surrogate molecule can be different from the clinical product, attention should be given to any potential impact these changes may have on the pharmacologic activity and/or the range of impurities. Manufacture of the surrogate molecule will require additional resources to develop the process, formulation, and packaging of the material. In cases of limited manufacturing capabilities or when clinical processes are utilized in the manufacturing of the surrogate molecule, additional pressure may be placed on manufacturing infrastructure. Additional assays are needed for characterization (aggregates, clips of the intact protein and impurities such as host cell protein, endotoxin, etc) and stability testing to support the duration of the toxicology studies.

Similar to the use of KO/Tg animals in preclinical safety evaluations, the use of surrogate molecules should be undertaken with an understanding of the time, cost, and effort of surrogate molecule development, in context with the applicability and relative benefits toward assessing clinical safety. Ultimately, a clear understanding and characterization of the pharmacological and biological behavior of the surrogate molecule is critical not only to the design of a relevant toxicity program, but also to the determination of the overall risk-to-benefit ratio for humans. As long-term repeated dosing of a human biopharmaceutical in rodent species is often not feasible due to immunogenicity and development of neutralizing antibodies that reduce exposure, a murine surrogate molecule may be developed on the assumption that it would be less immunogenic in mice. However, just as a human protein can be immunogenic in humans, the homologous protein can also be immunogenic in rodents. Therefore, the potential for immunogenicity from a surrogate molecule should be evaluated as early as feasible. The intended application of a surrogate molecule in a nonclinical safety program should also be considered (ie, for developmental and reproductive toxicology [DART], for chronic toxicology, etc), as it will define the extent of in vitro and in vivo studies required for adequate surrogate molecule characterization. For example, if a surrogate molecule is intended for use in DART assessments, in vitro placental transfer studies should be considered at an early stage of surrogate molecule development to characterize the relevance of the surrogate molecule to the clinical candidate in this respect and demonstrate that the fetus will be exposed to the drug. ²¹ Finally, determination of clinically relevant starting doses from preclinical toxicity evaluations using a surrogate molecule should involve careful consideration of relevant biological and pharmacokinetic differences and their potential impact on extrapolation of a safe dose.

Potential Advantages/Disadvantages of Using a Surrogate Molecule

In situations where the biopharmaceutical is pharmacologically active only in humans and/or chimpanzees, the use of a surrogate molecule could be helpful for determining potential hazards associated with the use of the product. Although limited nonterminal safety studies can be conducted in chimpanzees, the small group sizes, the heterogeneity of the animals that are available, previous drug exposure, and lack of histopathological endpoints limit the use of chimpanzees to an assessment of acute tolerability rather than an assessment of toxicity. Because of this, many companies now screen candidates and select lead molecules that have pharmacologic activity in at least one toxicologically relevant species. However, when a suitable species cannot be identified, a full single-species toxicology assessment can be conducted using a surrogate molecule. To date, the molecular classes that have shown the greatest degree of species specificity are the monoclonal antibodies and the interferons. Surrogate monoclonal antibodies and homologous interferons have been generated for some species-restricted products and have been used in nonclinical safety assessments. ^21–24 For the monoclonal antibodies, surrogate antibodies could potentially be developed for any species, but the most common surrogate molecules are rat anti-mouse antibodies. These rat antibodies can then be chimerized (rat Fab-mouse Fc) or engineered to replace certain rat sequences with murine sequences. One consideration for surrogate antibodies or Fc-fusion proteins is that the Fc region in the murine surrogate molecule must match the appropriate effector function of the Fc in humans (ie, the IgG2a isotype in mice to match the effector function of an IgG1 isotype in humans). ^25,26

One of the advantages of developing a murine surrogate for toxicology evaluation is that the outbred CD-1 mouse is a common toxicological species and therefore standard, well-validated toxicology protocols and endpoints can be applied to these studies. For immunomodulatory biopharmaceuticals, many of the standard immunotoxicity endpoints have been well established in mice, and many reagents are available for evaluating immunological endpoints. An added advantage of using the mouse is that many well-established disease models have been created in mice, allowing both pharmacology and toxicology to be conducted in the same species and providing a potential margin of safety in that species. A disadvantage of using mice is the limited blood volume for various endpoints and thus the relatively large number of animals needed for each study.

With the availability of an appropriate murine surrogate molecule, acute, subchronic, and chronic toxicity studies can potentially be conducted, depending on the immunogenicity of the surrogate molecule. Additional endpoints can be incorporated into the studies based on the pharmacology of the molecule, such as immunotoxicity endpoints for immune modulating biopharmaceuticals as mentioned previously. It is important to evaluate toxicokinetic parameters to demonstrate that exposure to the surrogate molecule is maintained throughout the dosing period and that antibodies that may develop against the surrogate molecule do not limit the exposure. However, if mice are used for toxicology evaluations, the limited blood volume does not allow the toxicokinetic and immune response to be measured in the main toxicology animals. Consequently, satellite animals must be included for the toxicokinetic and immune response evaluations, making direct correlations between exposure, immune response, and toxicity difficult.

For biopharmaceuticals that require an evaluation of potential effects on reproduction and/or development, DART studies in rodents using a surrogate molecule can be an acceptable option. This approach for DART studies can be used for molecules that do not cross-react with relevant toxicology species, or can be used to reduce the use of nonhuman primates for those molecules that show cross-reactivity only to humans and nonhuman primates. However, because of the caveats discussed previously, it may be more appropriate for developmental studies to evaluate the clinical candidate in nonhuman primates than to use a surrogate molecule that may not be truly representative. With the use of a surrogate molecule in rodents, all aspects of the DART evaluation as outlined in ICH S5 (R2) ²⁷ can be evaluated using standard well-established protocols. Numerous testing facilities have the experience and capabilities of conducting DART studies in rodents, and large historical databases are available for comparison with normal background findings.

Rodent studies may be particularly useful for the evaluation of fertility (ie, the number of successful pregnancies). Mating behavior and number of successful pregnancies cannot be evaluated practically in nonhuman primates because they have a naturally low fertility rate and high spontaneous abortion rate. Reproductive potential can, however, be evaluated in primates by evaluation of hormones, menstrual cycles, semen parameters, and other indicators of reproductive capability. In contrast, fertility measured by pregnancy success is a well-established endpoint in rodents.

Embryofetal development studies can also be conducted in rodents using a surrogate molecule. There are, however, certain caveats concerning the use of a surrogate molecule for embryofetal development studies that must be taken into consideration. The mechanisms that regulate large molecular weight protein transport and diffusion across the placenta are species dependent and differ between rodent and primate placenta. ²⁸ For antibodies that are known to be transported across the placenta, the timing of placental transfer and the efficiency of transfer differs between rodents and primates. ^29,30 In rodents fetal exposure to antibodies occurs earlier in development than in primates. Therefore, adverse effects may be observed in the rodents that are not relevant to humans and may therefore overpredict the potential risk. For large molecular weight proteins that do not cross the placenta, embryofetal development studies are likely to be restricted to maternal effects rather than direct teratogenic effects and a study in rodents with the surrogate molecule may model these effects similarly as a study in primates.

Pre- and postnatal development studies can also be conducted in rodents using the rodent surrogate molecule. In this case, endpoints for establishing effects of treatment on postnatal development are well established in rodents. The rodent postnatal development studies allow for an evaluation of sexual maturity and second-generation reproductive performance that cannot be evaluated in nonhuman primates because of the long period of time between birth and sexual maturity. However, because the toxicity of biopharmaceuticals is generally related to exaggerated pharmacology, unless there is a scientific reason why the molecule might affect sexual maturation, the absence of this information is not considered to be critical to assessing human safety.

In addition to the discussed limitations of using primates for reproductive toxicity studies, the other considerations for reproductive and developmental toxicity testing include the ethical use of primates versus rodents, especially for fertility studies, and the duration of the studies. A primate embryofetal development study can take up to 1 year to complete and a primate pre- and postnatal development study can take 2 years or more to complete. Reproductive toxicity studies can be conducted in rodents using a surrogate molecule in a fraction of this time, and could provide safety information earlier in the development process than is possible for primate studies. However, this advantage must be balanced with the previously mentioned disadvantages associated with developing a surrogate molecule and the possibility that the surrogate molecule in the rodent may not be pharmacologically identical to the clinical candidate in the primate. Also the resources required to develop, characterize, and test a surrogate molecule can exceed that of conducting nonhuman primate DART studies, so there may not be any advantage to the sponsor in developing a surrogate molecule. The most relevant model to assess risk of reproductive effects in humans should be utilized and justified to the regulatory agencies. Therefore, because of their limitations, the surrogate molecule approach should be considered as an alternate to nonhuman primate DART studies, not as an additional species.

There are several examples of approved products on the market in the United States for which the safety assessment included surrogate molecules in the regulatory filing. These products include Actimmune (interferon-γ; InterMune, Brisbane, CA), Remicade (infliximab; Centocor, Horsham, PA), Raptiva (efalizumab; Genentech, Inc, South San Francisco, CA, and Xoma Ltd, Berkeley, CA), Cimzia (Certolizumab pegol; UCB Inc, Smyrna, GA), and Solaris (eculizumab; Alexion Pharmaceuticals, Cheshire, CT). For example, to conduct a thorough safety assessment of interferon (IFN)-γ, which lacks activity in rodents, the sponsor elected to develop a recombinant murine IFN-γ and used that product to conduct toxicology studies in mice. ^22,23 Efalizumab and infliximab, monoclonal antibodies recognizing human CD11a and tumor necrosis factor-α (TNF-α), respectively, are active in humans and chimpanzees only. For both products, initial toxicology studies that supported the safety of clinical trials, were conducted in chimpanzees. To conduct a more thorough safety evaluation, which was necessary for product approval, the sponsors for these products elected to develop antibodies that recognized mouse CD11a and TNF-α. ^21,24 Ecalizumab, an anti-C5 antibody, showed cross-reactivity only to humans. Therefore, a surrogate mouse antibody was developed and chronic and DART studies were conducted with this surrogate molecule (Solaris approval information, http://www.fda.gov/). Certolizumab pegol is an anti-TNF antibody Fab′ fragment conjugated with PEG to extend the terminal plasma elimination half-life. Inasmuch as certolizumab pegol does not cross-react with mouse or rat TNF-α, reproduction studies were performed in rats using a rodent anti-murine TNF-α pegylated Fab fragment (cTNF PF), similar to certolizumab pegol. A few of these examples will be highlighted in more detail in the following section, along with cases of molecules still in development.

Case Study Examples Using Alternate Approaches to Safety Assessment

Approved Products

Efalizumab

Efalizumab is a recombinant humanized monoclonal IgG1 antibody specific for the α subunit (CD11a) of leukocyte function associated antigen-1 (LFA-1) approved for treatment of chronic plaque psoriasis. Efalizumab binds specifically to human and chimpanzee CD11a. As a result of the limited species binding, muM17, a murinized rat anti-mouse chimeric IgG2a surrogate antibody specific for murine CD11a was developed for preclinical safety evaluation. The murine surrogate antibody was of the IgG2a isotype to match the potential effector functions of an IgG1 isotype in humans. The general preclinical safety program conducted with muM17 to support registration of efalizumab included a tissue cross-reactivity study, 1- and 6-month repeat dose toxicity studies, male and female fertility, embryofetal development, and pre- and postnatal development studies. In addition, special immunotoxicology evaluations were conducted in a 1-month subcutaneous immunotoxicity study and as part of the pre-and post-developmental toxicity study to assess the effect of muM17 on immune function.

Prior to initiation of the preclinical studies, the pharmacology of muM17 was characterized and the activity of muM17 was demonstrated to be comparable with efalizumab in binding and pharmacodynamic assays. Characterization experiments included the measurement of binding affinity of muM17 to murine CD11a, in vitro activity of muM17 using the murine mixed lymphocyte reaction, and in vivo activity in a mouse model of delayed type hypersensitivity. In addition, muM17 was shown to have a comparable pharmacodynamic effect as efalizumab in inducing downmodulation of CD11a on peripheral blood T cells following ligand binding.

Results from preclinical studies using muM17 were consistent with distribution of the target antigen and pharmacology of blocking CD11a/ICAM-1 (intracellular adhesion molecule-1) interactions. In the tissue cross-reactivity study, the pattern of staining using a panel of murine tissues was similar to results observed with efalizumab on a panel of human tissues. Results from the male and female fertility, and embryofetal development studies, demonstrated no effects on either fertility or fetal development from administration of muM17 to CD-1 mice. Consistent with these findings, no toxicities were observed in either the 1- or 6-month toxicity studies.

The results from the 1-month subcutaneous immunotoxicology study demonstrated administration of muM17 to young adult mice as 4 weekly doses resulted in decreased IgM and IgG responses to sheep red blood cells (SRBC), a T-cell-dependent antigen, immediately following the dosing phase. This result is consistent with the biology of blocking LFA-1/ICAM-1 interactions. After serum levels of muM17 decreased below the level of detection following a 4-week recovery period, humoral immune responses to SRBC were comparable with controls. Similar decreases in humoral immune responses to SRBC were demonstrated in F1 generation mice exposed to muM17 via the dam during gestation and lactation in the pre- and postnatal development study. However, in contrast to the normalization of humoral immune responses observed in the young adult mice following clearance of muM17, humoral immune responses to SRBC in the F1 mice remained decreased relative to age-matched vehicle control animals following a 22-week recovery period.

Additional data regarding placental transfer and antigenicity of the surrogate antibody were obtained from the toxicology studies. Placental transfer of muM17 was confirmed in the reproductive toxicity studies. Fetal and maternal serum concentrations were approximately proportional whereas muM17 concentrations in amniotic fluid were considerably lower than maternal serum. Antigenicity of muM17 was shown to be low in all studies with an incidence of less than 1% for the entire surrogate molecule program. ²¹ The adverse safety events or toxicities observed in the clinic with efalizumab are reported to include infection, malignancy, immune thrombocytopenia, and hemolytic anemia. ³¹ Approximately 2 years after marketing approval a Dear Health Care Professional letter was authored notifying prescribing physicians of potential toxicities. ³² Most of the adverse events observed in the clinic were consistent with the drug’s mechanism of action but were not predicted based on the safety profile of the murine surrogate antibody.

The surrogate antibody program utilized the consistent biology of LFA-1/ICAM-1 between species to build a robust preclinical program that was successfully accepted by regulatory agencies in the United States, European Union, and Japan. It is important to recognize that, because the use of chimpanzees is greatly restricted and therefore of limited use for assessing safety, the mouse surrogate molecule effectively represented the primary safety evaluation; it was not developed to allow evaluation of a second, rodent species. At first pass, use of the clinical candidate in chimpanzees may appear to be the more relevant approach to safety assessment compared with the use of a surrogate in rodents. However, as access to chimpanzees becomes further restricted in the future due to both supply (decreased breeding in United States ³³ ) and ethical considerations (supported by proposal of the 2008 Great Ape Protection Act to the US Congress and adoption of similar legislation in several EU countries), it will be vital to provide sufficient scientific evidence regarding relevance of the surrogate molecule in regulatory submissions, as well as complimentary data (ie, in vitro assays) necessary to adequately bridge surrogate data and relevance to humans.

Infliximab

Infliximab is an anti-human TNF-α (TNF) mAb that was first approved in the United States in 1998 for the treatment of Crohn’s disease. Infliximab is a chimeric IgG1 mAb that binds to human TNF and is highly species specific (cross-reacts only with human and chimpanzee TNF). Therefore, an analogous anti-TNF mAb (cV1q) that selectively inhibits the functional activity of mouse TNF was developed to assess chronic and reproductive toxicity. ²⁴ Similar to efalizumab, the murine surrogate antibody was of the IgG2a isotype to match the potential effector functions of an IgG1 isotype in humans. This surrogate molecule inhibited disease activity in murine models of Crohn’s disease, and was thus demonstrated to be pharmacologically similar to infliximab. For registration, an embryofetal development toxicity study, a combined male and female fertility study, and a chronic, 6-month toxicity study in CD-1 mice was performed with the murine surrogate. An immune response to the anti-murine mAb did develop in most animals, but this immune response did not affect exposure of the animals to the mAb (ie, no change in PK parameters). A nonterminal acute tolerability study was also conducted in chimpanzees with the anti-human TNF mAb.

Pre- and postnatal development studies in mice using the murine surrogate antibody were also conducted postmarketing. ³⁴ In addition to all of the standard evaluations of postnatal development, an evaluation of immune function was conducted in the F1 mice. When the results from the studies conducted with mice using the surrogate are compared with published information on TNF-deficient mice, ³⁵ the results are generally similar but not identical. Although the genetically deficient mice and surrogate-targeted mice show a general concordance with regard to effects on fertility and embryofetal development, the genetically deficient mice show a lack of splenic germinal centers and reduced functional immune responses that are not observed in the surrogate molecule-treated mice. Therefore, the genetically deficient mice are useful models for understanding the biology of TNF, but are imperfect models for evaluating the safety of mAb treatment.

This example illustrates an overall approach for an mAb that showed cross-reactivity only to humans and chimpanzees. This approach allowed for the safe dosing of infliximab in clinical trials and was also acceptable to the regulatory agencies in the United States, European Union, and Japan. The adverse effects of infliximab that have been observed in the clinic have been mostly related to selective downmodulation of immune responses leading to an increase in some opportunistic infections, which is directly related to the pharmacology of the molecule. As expected, infections were not seen in the normal healthy animals used in the toxicology studies, but have been seen in TNF-α-deficient animals that are challenged with specific pathogens. Therefore, taking the entire weight of evidence into consideration, the human toxicities were predictable based upon the nonclinical information and the pharmacology of the molecule.

Although the toxicity studies conducted in mice with the surrogate molecule were acceptable to support the clinical use of a human mAb with cross-reactivity only in humans and chimpanzees, the mouse studies were not acceptable to the regulatory agencies to support the clinical use of another human anti-TNF mAb that was pharmacologically active in humans, chimpanzees, and macaques. For the latter mAb, regulatory agencies required embryofetal and postnatal development studies in cynomolgus monkeys with the humanized mAb, ³⁶ despite DART studies previously conducted with a surrogate mAb, and significant chronic-use clinical experience with the original chimeric mAb, as well as several other anti-TNF therapies. Therefore, although it has been proposed that surrogate molecules could be used to reduce the use of nonhuman primates, ³⁷ to ensure the regulatory acceptance of a surrogate molecule-only approach for a clinical candidate that shows cross-reactivity to nonhuman primates, a dialogue must be conducted between the company and the agencies to ensure that all parties are in agreement with the scientific justifications for the approach.

Interferon-γ

IFN-γ is an immunomodulatory cytokine that is highly species specific, with pharmacologic activity only in humans and nonhuman primates. In subchronic toxicity studies in cynomolgus monkeys, the clinical side-effects that were observed (fever, lethargy, anorexia, and changes in hematology and chemistry parameters) were comparable with clinical effects seen in humans. ²³ However, a neutralizing antibody response was seen that attenuated the response in a 13-week study compared with the 4-week monkey study. A murine version of IFN-γ was utilized in a 4-week mouse study, where the nature of the treatment-related findings and organ systems affected were similar to the observations in cynomolgus monkeys treated with the human protein, and yet no neutralizing antibodies developed. ³⁸ Thus, in this case the murine surrogate was not developed to allow safety evaluation in a second, rodent species, but rather to better characterize the response without the confounding factor of neutralizing immunogenicity.

In developmental toxicity studies, both the human IFN-γ and the murine surrogate molecule were abortifacients in cynomolgus monkeys and mice, respectively. The murine surrogate molecule was also used to evaluate potential developmental and reproductive capacity of juvenile animals associated with chronic treatment, because juvenile patients with chronic granulomatous disease are the main patient population (Actimmune; http://www.fda.gov/cder/foi/label/2007/103836s5098LBL.pdf). In this study, mice were treated with daily doses (0, 0.02, 0.2, or 2 mg/kg/d) from postnatal day 8 through day 60 to determine the effects on maturation, behavioral/functional development, and reproductive capacity. ²² Male mice in the high-dose group had delayed sexual maturation, reduced epididymal and testes weights, reduced sperm count and concentration, and sperm abnormalities, and showed reduced mating performance and fertility despite the absence of altered histopathology of the testes. Motor activity was also decreased in all mice in the high-dose group. Although it is unknown whether these findings would be found in humans treated chronically with IFN-γ, this information does appear in the label for Actimmune (along with the caveat of unknown significance). Prior to conducting these studies with the surrogate molecule, a careful review and comparison of data regarding biochemical properties, biological activity, and disposition profiles of both proteins in similar test systems was performed. The characterization of the mouse surrogate molecule in this case allowed for further exploration of the reproductive and behavioral effects of IFN-γ that would have been difficult to evaluate in the nonhuman primate.

Products in Development

Keliximab, Clenoliximab

Keliximab is a primatized IgG1 mAb directed toward domain 1 of human CD4. ^15,39 Clenoliximab is an IgG4 version, developed to reduce Fc interactions with Fc receptors and thus mitigate T-cell depletion and cytokine release. Keliximab and clenoliximab show cross-reactivity only to human and chimpanzee CD4. To evaluate the preclinical safety of these molecules, the sponsors developed a transgenic mouse that expressed human CD4 in place of mouse CD4. Both monoclonal antibodies were shown to be pharmacologically active in these KO/KI mice. ¹⁵ With this KI mouse, the human therapeutic antibodies could be tested for nonclinical safety. The KI mouse was characterized by demonstration of appropriate expression of human CD4 and by evaluation of immune system function following challenge with infection or tumor cells. The preclinical safety studies conducted with keliximab in the KI mice included single and repeated dose toxicology studies, male and female fertility, embryofetal development, pre- and postnatal development studies with functional immune response evaluation in the F1 generation, and host defense assays. ^15,39 A nonterminal acute tolerability study was also conducted in chimpanzees. Toxicity studies conducted in the mice showed the expected reduction in CD4 cells. The mice did develop an immune response toward the human mAb and in some instances anaphylactic reactions occurred. However, sufficient numbers of animals survived and exposure levels were sufficiently maintained to adequately evaluate the toxicity.

This example demonstrates a novel approach in which KI mice expressing the human target antigen were developed so that the human protein could be tested in preclinical studies. These methods have the disadvantage that human proteins can be highly immunogenic in animals and thereby limit the duration of the studies, impact the exposure of the animal to the therapeutic, or lead to adverse effects. Also this approach requires development and extensive characterization of the animal model, including genotyping of all animals used for the studies, to ensure appropriate expression of human CD4. In addition, the model tests only specific inhibition of the intended target and may not expose any secondary pharmacology attributable to closely related or downstream targets associated with administration of the human protein. However, when the clinical candidate is active only in humans and chimpanzees, data in KI mice treated with the clinical candidate can be useful in assessing safety.

Fully Human mAb (Anti-Cytokine Receptor)

A fully human mAb targeting a cytokine receptor was developed for treatment of allergic inflammatory responses (data on file, Amgen Inc). Because the activity of the clinical molecule was limited to humans and chimpanzees, a surrogate molecule approach was taken for the safety program to support clinical development, utilizing 2 well-characterized chimeric anti-mouse and anti-monkey surrogate antibodies. These molecules have similar respective activity for murine and cynomolgus monkey cytokine inhibition as does the clinical molecule for human cytokine inhibition. The characterization of these molecules included binding and functional activity in cell-based assays in vitro (murine and monkey) and in vivo pharmacologic activity (murine only). The safety program along with various noteworthy rationales behind the necessity for many of the studies is described below.

Prior to the first in human (FIH) clinical trial, the molecules that were available to conduct preclinical studies included the clinical molecule, which only cross-reacted with the target in humans and chimpanzees, and the anti-murine surrogate antibody. A single-dose chimpanzee study was conducted primarily to model the pharmacokinetics of the clinical molecule in a relevant species, and a 4-week mouse toxicology study with the anti-murine surrogate was conducted as the definitive safety study to support the FIH trial. Because the clinical candidate molecule was not evaluated in the toxicology studies where histopathology evaluation was available, a rabbit local tolerance study was conducted with the clinical molecule to assess the irritation potential of the formulation to support the FIH trial.

Information from the literature indicated that the cytokine of interest is important in the maintenance of pregnancy. In a previous embryo-fetal development study conducted with a human soluble cytokine receptor that had the same pharmacologic activity (antagonism of the cytokine) in cynomolgus monkeys, an increased frequency of spontaneous abortions and stillbirths was observed. ⁴⁰ To determine if a better model could be established to understand the mechanism of this toxicity, a mouse study was conducted with a murine surrogate molecule. The cynomolgus monkey reproductive findings were not reproducible in mice with a murine surrogate of the soluble cytokine receptor or with an anti-murine cytokine receptor antibody, suggesting that the mouse was not an appropriate species for evaluation of the reproductive effects following inhibition of this pathway. Thus, the cynomolgus monkey surrogate molecule was developed primarily to evaluate reproductive toxicity. The murine surrogate antibody was utilized for fertility evaluation, because an effect on fertility was not expected based on information in the literature. The complete package for reproductive toxicology evaluation thus consisted of a fertility study with the murine surrogate molecule in mice and an embryofetal and prenatal development study with the cynomolgus monkey surrogate molecule in the cynomolgus monkey.

Results from the anti-monkey surrogate molecule for the reproductive toxicology evaluations as well as availability of results from a 1-month repeated-dose cynomolgus monkey toxicology study with the monkey surrogate molecule were available at the time point for initiation of the subchronic study. Although both surrogate molecules were similar to the clinical candidate, to limit resources originally dedicated to 2 surrogate molecules, the cynomolgus monkey surrogate molecule was the only one utilized to conduct the subchronic (3–month) and chronic (6-month) repeated-dose toxicology studies.

Fully Human mAb (Anti-Cytokine)

A fully human mAb targeting a cytokine was developed for the treatment of inflammatory disease (data on file, Amgen Inc). The clinical molecule bound to recombinant human cytokine with high affinity, but had lower affinity (approximately 30-fold) for the cynomolgus monkey cytokine. In addition, the clinical molecule could neutralize the ability of the human cytokine to stimulate human cells in a cell-based assay, but was not able to efficiently neutralize the cynomolgus monkey cytokine. This demonstrates that binding alone may not be sufficient to demonstrate species relevance. A murine surrogate molecule was not available at the time of initiation of the preclinical safety program. Thus, to enable preclinical studies in cynomolgus monkeys, a surrogate antibody was developed by fusing the F(ab) portion of a mouse anti-human cytokine, known to neutralize the cynomolgus monkey cytokine, with human Fc. The chimeric surrogate molecule had nearly as high an affinity for the human cytokine as did the clinical molecule, and higher binding affinity for the cynomolgus monkey cytokine compared with the clinical molecule. The surrogate molecule also neutralized the ability of cynomolgus monkey cytokine to stimulate cynomolgus monkey cells in a manner similar to the neutralization of human cytokine activity on human cells by the clinical molecule. Thus, the surrogate molecule was used in the monkey for the nonclinical safety program.

Toxicology studies were conducted in cynomolgus monkeys with the surrogate molecule to support the early stage clinical development plan. Prior to the FIH clinical trial, studies conducted included a repeated-dose study of 1-month duration and a safety pharmacology study (cardiovascular, respiratory, and central nervous system). Results from the 1-month study indicated a pharmacodynamic effect with the surrogate molecule in the monkey that would be predicted based on one of the known activities for the cytokine of interest, as noted in the literature. Interestingly, this pharmacodynamic effect was not evident in the FIH trial, indicating that there may be a species difference in response to inhibition of the cytokine of interest or that there is a difference between the monkey surrogate molecule and the clinical candidate. This example highlights the challenges when using a surrogate molecule and conflicting data are generated. Further work is then needed to understand whether the difference is due to the use of a different molecule, or whether the activity of that target is different in the animal species compared with humans, or between different animal species (as seen in the previous example).

Fusion Protein

A soluble lymphotoxin β receptor consisting of the extracellular domain of human lymphotoxin β fused to the Fc region of human IgG1 (LTβR-Fc) is currently in development for the treatment of rheumatoid arthritis. The pharmacologic activity is limited to humans and nonhuman primates. During development, an anti-murine lymphotoxin β receptor Fc IgG fusion surrogate molecule was generated to evaluate the pharmacology of the soluble receptor in rodents. The Fc region in the murine surrogate molecule was of the IgG2a isotype to match the effector function of an IgG1 isotype in humans. Adult mice treated with the surrogate LTβR-Fc showed reduced immune responses and reduced disease activity in a number of murine disease models. ⁴¹ Mice that are genetically deficient in LTβ have also been generated and shown reduced immune responses and an absence of lymph nodes. ⁴² Pregnant mice treated with the surrogate LTβR-Fc molecule had offspring that lacked lymph nodes. ⁴³ However, administration of the human LTβR-Fc to pregnant monkeys (a pharmacologically relevant species) was not associated with an absence of lymph nodes. ⁴⁴ The differences observed between the surrogate in the mouse and the human therapeutic in the monkeys may be due to differences in the timing of lymph node development relative to placental transport of antibodies in mice versus primates, although monkey fetal exposures were not reported in this study.

This example illustrates how results obtained with a surrogate molecule in mice or extrapolated from KO mice may not necessarily be representative of the results that are likely to occur in nonhuman primates or humans, particularly in developmental toxicity studies in which species differences are known to exist in the stages of embryofetal development. In this case, the surrogate and the genetically deficient animals identified the potential hazard, but may have overestimated the human risk.

Questions Regarding the Use of Surrogate Molecules

If a Surrogate Molecule Is Used to Assess Any Aspect of Safety, Is It a Requirement to Conduct All Toxicology Studies in That Species?

The nonclinical safety programs for small molecule drugs routinely require assessment of general toxicity in 2 species, including a rodent (usually the rat) and a nonrodent (usually the dog) (ICH M3 (R1) ⁴⁵ ). The ICH S6 guidance document for biopharmaceuticals states that “Safety evaluation programs should normally include 2 relevant species. However, in certain justified cases, one relevant species may suffice (e.g., when only one relevant species can be identified or where the biological activity of the biopharmaceutical is well understood).” Furthermore, the guidance document goes on to say that “even where two species may be necessary to characterize toxicity in short term studies, it may be possible to justify the use of only one species for subsequent long-term studies (e.g., if the toxicity profile in the 2 species is comparable in the short term).” Therefore, for biopharmaceuticals that have a very specific mechanism of action that is well understood, such as monoclonal antibodies and receptor fusion proteins, a 2-species assessment may not always be necessary.

For species-restricted biopharmaceuticals, the most common species used for nonclinical safety testing is a nonhuman primate, usually a macaque. The toxicities that have been observed in macaques for biopharmaceuticals have generally been directly related to the pharmacology of the molecule and off-target toxicities are rarely observed. Because the surrogate molecule has many limitations that have been mentioned previously, the surrogate molecule provides supportive information only in those situations where there is no other option for nonclinical safety testing or where information can be obtained in the rodent that cannot be adequately evaluated in the nonhuman primate (eg, fertility). In cases in which a surrogate molecule is available and is considered optimal for use in developmental and reproductive toxicity evaluations, inclusion of standard toxicity endpoints (ie, clinical pathology endpoints) in preliminary dose-ranging studies may compliment the overall weight of evidence regarding drug safety. These may provide a useful comparison to the findings seen in the general toxicology studies with the clinical candidate.

Should a Surrogate Molecule Be Developed to Allow for Assessment of Carcinogenicity?

The question of carcinogenicity is not so much one of whether a surrogate molecule should be used for carcinogenity testing, but whether 2-year bioassays in rodents are relevant models for evaluating carcinogenic potential for biopharmaceuticals. ⁴⁶ Carcinogens generally fall into 3 major categories: genotoxic carcinogens, cellular proliferators, and immune suppressants. ⁴⁷ Biopharmaceuticals have a large molecular weight that precludes them from diffusing into cells and interacting with DNA. Therefore, biopharmaceuticals are unlikely to be genotoxic carcinogens.

Biopharmaceuticals that induce cellular proliferation such as insulin-like growth factor and growth hormone are assumed to be associated with a greater risk of tumor development. Increased cellular proliferation can be detected with in vitro studies and may be evident in repeated-dose toxicology studies of sufficient duration. The absence of hyperplasia in a repeated-dose toxicity study may be indicative that the biopharmaceutical is unlikely to be carcinogenic because of increased proliferation, when considered in context with the duration of dosing, level of analysis, and known target biology. A recently published study has described the development of mouse- and rat-specific growth hormones for the evaluation of carcinogenic potential in 2-year bioassays as surrogate molecules for human growth hormone. ⁴⁸ The studies showed no increase in tumors in the treated animals. However, it is not clear that these negative results in rodents will change either the perception or labeling of the product, even in the context of 40 years of clinical experience using human growth hormones.

Finally, immune suppressive agents are assumed to increase a patient’s susceptibility to certain tumor types (especially lymphomas and skin cancer) because of decreased host defense. ⁴⁹ This hypothesis has been based upon clinical experience, not upon 2-year bioassays, which are frequently negative for small molecule nongenotoxic immunosuppressive agents. Immunosuppressive drugs and immunomodulating biopharmaceuticals carry warnings on their product labels for a potential increased susceptibility to tumors. In a 2-year bioassay, mice infected with both mouse leukemia virus and mammary tumor virus and then treated with abatacept, which inhibits T-cell activation, were found to develop lymphomas and mammary tumors, respectively. ⁵⁰ Therefore, this study reinforces the hypothesis that immunosuppression can increase susceptibility to oncogenic viruses but did not provide new information and did not change the warnings in the product label.

The carcinogenic risk assessment for biopharmaceuticals can best be made based upon an understanding of the biology of the molecule. An example of a rational, scientific-based assessment of carcinogenicity potential in the absence of animal carcinogenicity testing has been published for interleukin-10. ⁵¹ Overall, for the majority of biopharmaceuticals, there is little rationale for conducting 2-year bioassays for biopharmaceuticals with either the human therapeutic (if it shows rodent cross-reactivity) or with a surrogate molecule, because the assessment of carcinogenic potential and communication of risk could be achieved by other means (ie, in vitro data, pharmacology data, appropriate wording in the product label, etc). In the end, assessment of carcinogenic potential should be made on a case-by-case basis considering the relevance of carcinogenicity testing in the context of target biology, to adequately communicate carcinogenic risk to patients.

How Should the Relevance of the Surrogate Molecule to the Clinical Candidate Be Determined? What Is the Importance of Understanding Target-Mediated Pharmacologic Effects Versus Compound-Related Off-Target Effects?

It is important that any compound selected for use as a surrogate molecule be shown to be pharmacologically similar to the development compound, because, if not, it is possible that erroneous conclusions about safety could be reached. A few examples are described below.

In the first example, a sponsor was developing an IgG1 Fc-mutated humanized mAb against a chemokine receptor present on T cells and monocytes that was believed to be important in leukocyte trafficking and activation in inflammatory and autoimmune diseases (C. Horvath, personal communication, 2009). The antibody had been tested extensively in vitro with human cell systems and demonstrated to be a ligand-binding antagonist with no signaling through the receptor and no target cell-depleting effects. These properties were confirmed in early clinical trials with the humanized mAb. To potentially address reproductive toxicology and host resistance concerns related to target blockade, an anti-rodent mAb was desired for use as a surrogate molecule for toxicity testing. One such anti-rodent IgG2 mAb had been published by an academic laboratory to be effective in several animal models of autoimmune disease and to induce no target cell depletion. This mAb was subsequently in-licensed by the sponsor for use as a development surrogate molecule and small quantities were synthesized for evaluation. The anti-rodent mAb was administered to normal mice and target cells were immunophenotyped and evaluated for receptor expression and blockade by flow cytometry. Unlike the development mAb, the surrogate mAb resulted in depletion of all target cells within the circulating blood, as well as spleen tissue, within 15 min of dosing. T cells and monocytes did not begin to return to the circulation until approximately 3 days later. Upon reviewing the publications that demonstrated efficacy with this rodent mAb, it was found that the cell-depleting properties were not known because early time points after dosing had not been evaluated. In light of this information (cell depletion with the surrogate molecule vs no target cell depletion with the clinical molecule), the sponsor questioned whether the results with the surrogate mAb in models of autoimmune disease would be reliable indicators of the results that might be expected for the development mAb in human autoimmune diseases. An attempt to alter the cell-depleting properties of the surrogate molecule by antibody engineering was then undertaken. Despite almost 2 years of dedicated efforts, the cell-depleting properties of this mAb could not be eliminated by isotype switching or Fc mutations. The sponsor concluded that the anti-rodent mAb could not be used as a surrogate molecule because either the surrogate mAb was not pharmacologically similar to the antihuman mAb (eg, different Fc-Fc receptor interactions) or the rodent species was not biologically similar to humans (eg, different target expression and/or function). In this example, a safety profile generated with the surrogate mAb may have overestimated potential concerns (eg, extent of immunomodulation) because the surrogate molecule resulted in destruction of the target cells, rather than blockade of a specific cellular function.

The second example is a recent one in which reliance on results of a surrogate mAb may have contributed to an underestimation of the potential safety concerns for the development mAb. TGN1412 was a humanized IgG4 mAb directed against CD28 on T cells being developed for oncology and autoimmune diseases. In vitro, the TGN1412 “superagonist” mAb directly activated human T cells in the absence of a second costimulatory signal and resulted in cytokine release and T-cell proliferation, with a bias toward a regulatory immunophenotype (T_reg cells). However, when dosed to normal volunteers, TGN1412 resulted in massive cytokine release (a “cytokine storm”), accompanied by rapid, severe T-cell depletion (within 1 hour, the earliest time point at which T-cell counts were measured) associated with headache, rigors, myalgia, hypotension, tachycardia, fever, and multiorgan failure. ⁵² This response of T-cell depletion and cytokine release had not been described in preclinical studies. The pharmacologic effect of CD28 stimulation had been evaluated extensively in rats with a surrogate anti-rat CD28 mAb, JJ316. ⁵³ This mAb was shown to be effective in several animal models of autoimmune disease, such as adjuvant-induced arthritis (AA) and experimental allergic encephalomyelitis (EAE), but not collagen-induced arthritis (CIA). In normal rats, the surrogate mAb was associated with marked increases in blood and spleen T-cell counts (up to ~ 20-fold) and marked expansion (up to ~ 6-fold) of lymphoid tissues, such as spleen and lymph nodes, within 3 days. ^54,55 No adverse effects related to cytokine release were reported in this study. The results generated with the surrogate mAb therefore suggested that administration of an anti-CD28 mAb could be well tolerated.

Because TGN1412 (or an IgG1 version, TGN1112) recognized CD28 in rhesus and cynomolgus monkeys, it was evaluated in these species and was reported to be efficacious in a monkey model of rheumatoid arthritis (CIA). In toxicology studies in monkeys, TGN1412 was associated with minimal cytokine release and, at approximately 2 weeks after dosing, with only mild increases in blood T-cell counts (~ 2-fold) and minimal evidence of expansion of lymphoid tissues. ^56–58 These results for TGN1412 in monkeys were in contrast to those obtained with JJ316 in rats, where profound lymphocytosis occurred. Although a cytokine storm did not occur in monkeys, close examination of the reported lymphocyte counts suggests that T cells may have been depleted in monkeys after dosing (as they were in humans). It is not possible to confirm this because early postdose time points were not evaluated. T-cell depletion was, however, a prominent finding when TGN1412 was administered to another species expressing human CD28+ T cells. When H2d Rag2 –/–gc−/− KO mice are irradiated and their bone marrow reconstituted with human CD34+ fetal liver (stem) cells, they develop human immune systems (HIS) with all major human myeloid and lymphoid cellular compartments, including CD28+ human T cells. When the original mouse precursor of TGN1412 (mAb 5.11A1) was administered to these mice, they developed rapid, profound T-cell depletion that persisted through 60 days. ⁵⁹ These results are in contrast to those obtained with the anti-rodent surrogate mAb. Thus, perhaps testing of the development compound in this “surrogate species” expressing human T cells was more representative of the potential safety concerns for the development mAb. These examples highlight some of the caveats that must be considered when electing to support development products with surrogate compounds. Further, because the generation of surrogate molecules for toxicology studies is a time- and resource-consuming effort, use of a surrogate molecule should be warranted only in special cases. Surrogates studies should not be conducted based on “no effect” in nonhuman primate studies if no toxicity was predicted based on super pharmacology or the absence of the target in a normal animal.

Discussion/Conclusions

This review illustrates that alternative approaches, including animal models of disease, KO/Tg or humanized mice, or surrogate molecules, can be appropriate to improve the predictive value of preclinical safety assessments for species-specific biopharmaceuticals, although many caveats must be considered (Tables 3 and 4). Surrogate molecules may be particularly useful for repeated-dose toxicity studies (in cases where the molecule cross-reacts only with the human and chimpanzee target), or for specialty studies such as fertility toxicity testing, immunotoxicity testing, host resistance models, and so forth (in cases where the molecule cross-reacts with primates, but not rodents). However, having a surrogate molecule available does not obligate evaluation of a “second species” or the conduct of carcinogenicity studies. Rather, alternative approaches should only be used to evaluate safety when the clinical candidate cannot be readily used to identify the hazard in an appropriate nonclinical species.

There are many issues to be considered when pursuing alternative approaches with a surrogate molecule. Indeed, it is critical to sufficiently understand the similarity and relevance of the surrogate molecule to the clinical candidate (eg, based on characteristics of the molecule, formulation, and pharmacologic activity). Examples have been given in which surrogates have both under- or overpredicted the effects seen in humans, and produced different results compared with the clinical candidate in nonhuman primates. However, the potential for over-and underprediction of human effects is a challenge common to both traditional and alternative toxicity assessments in any nonhuman species. When a difference between results or lack of predictivity occurs with a surrogate molecule, additional questions arise regarding whether the differences between nonclinical and human studies are due to pharmacological or molecular differences between the surrogate and the biopharmaceutical, or simply due to differences in species biology. Thus, it is critical to fully understand and extensively characterize the pharmacology of the surrogate molecule in the appropriate species both in vitro and in vivo, and ensure that its pharmacology is as similar to that of the clinical candidate in humans as possible. Because the pharmacology is often poorly understood, the use of alternative approaches in the safety assessment of biopharmaceuticals should be supplemental and evaluated as part of a case-by-case, weight-of-evidence approach. However, any of these alternatives can be valuable support for a biopharmaceutical safety assessment program in which unique challenges of species specificity often make a standard toxicology program inappropriate to conduct.