A Risk-Based Approach to Immunogenicity Concerns of Therapeutic Protein Products, Part 3: Effects of Manufacturing Changes in Immunogenicity and the Utility of Animal Immunogenicity Studies

January 1, 2005
Amy S. Rosenberg, MD

Alexandra Worobec, M.D.

BioPharm International, BioPharm International-01-01-2005, Volume 18, Issue 1

Manufacturing changes - such as changes in formulation or source material - can impact a product’s immunogenicity.

FDA utilizes a risk assessment strategy to evaluate the potential immunogenicity of novel protein therapeutics and licensed products following manufacturing changes. The first two parts of this article discussed FDA's assessment of the critical factors influencing the immunogenicity of such products and the consequences of immune responses to treated patients. The first part focused on the clinical consequences of immune responses to protein therapeutics, while the second part examined product- and patient-related factors that influence immunogenicity. This third and final part focuses on the effects of changes in manufacturing and the utility of animal models in assessing a product's immunogenicity.


The following changes in the manufacture of protein products have a substantial potential to affect product attributes associated with immunogenicity:

  • formulation

  • virus or adventitious agent clearance and inactivation

  • source material or cell line

  • container closure.

Changes in formulation and container closure as they impinge on immunogenicity have already been discussed in Part 2 of this article. Changes in virus and adventitious-agent clearance may involve addition of filtration steps, low pH steps, or heat inactivation.67-69 Filtration-generated shear forces as well as hydrophobic filter surfaces may cause protein denaturation and aggregation. Heat is well known to denature proteins and facilitate their aggregation, and protein conformation can be exquisitely sensitive to pH effects. Therefore, such changes in manufacturing require assessment of protein degradation and aggregation.

Changes in source material or cell line may adversely affect proteins through changes in local concentration, as increased molecular collisions facilitate aggregation. Moreover, products made in bacterial systems are normally aggregated in inclusion bodies and require refolding and renaturation to make them soluble and functional.


Animal immunogenicity studies can be useful in assessing aspects of immune responses to therapeutic proteins. However, except in rare cases, animal models do not predict accurately the incidence of human immune responses to therapeutic-protein products. For proteins conserved in evolution, particularly those with a biologically unique function, animal models may illuminate the adverse effects of neutralizing antibody on the endogenous protein. This is particularly important where animal knockout (KO) models have not been generated or where the KO is lethal to the embryo or fetus. Since xenodeterminants certainly contribute to induction of immune responses to human proteins in animals, the incidence of immune responses to human proteins in animals reflects responses to a foreign protein, and thus will often overestimate the product's ability to induce such responses in humans. However, animal models do allow examination of the consequences of immune responses to therapeutic proteins.

Assessment of the inherent immunogenicity of a biological therapeutic requires treatment of animals with the species-specific recombinant animal homolog. This was indeed done in assessing the immunogenicity of thrombopoietin (TPO). When recombinant pegylated human megakaryocyte growth and development factor (PEG-MGDF) or TPO was tested in animal models, animals mounted immune responses that neutralized the product. Because TPO is highly conserved in evolution, such antibodies also cross-reactively neutralized the endogenous animal TPO, rendering animals thrombocytopenic.74,75 This point illustrates the utility of animal models in assessing the consequence of an immune response. However, to test whether such responses were generated only because of responses to the xenodeterminants on the human TPO or PEG-MGDF, species-specific TPO was tested in several animal models. Stunningly, animals generated neutralizing-antibody responses to species-specific TPO which rendered them thrombocytopenic, reflecting the underlying lack of tolerance to this protein.76 Therefore, the high sequence conservation of TPO in evolution, coupled with the findings of studies employing species-specific factor, markedly enhanced the predictive power of the animal studies for human immune responses. However, the species-specific protein product is likely to differ significantly from the recombinant human therapeutic in ways that affect immunogenicity (see Part 2). Further, species differences in major histocompatibility complex (MHC) proteins, which present peptides of the therapeutic protein product to CD4+ T cells, eliciting their help for antibody formation, also make it extremely difficult to model human immune responses in animals.

One way to eliminate the xenogeneic response to a human protein therapeutic in animals is to use human-protein transgenic mice. For example, human IFN-α transgenic mice very nicely illustrated the effects of IFN-IFN and IFN-serum-albumin aggregates on induction of immune responses.58 Such models may be helpful in determining the level and kinds of aggregates or other product degradation forms triggering immune responses. However, there are some major limitations to the utility of this technology. First, protein expressed via a transgene may neither reflect the protein's normal physiology nor the immune system's normal tolerance level to the protein in humans. Indeed, the transgene expression level and pattern depends on copy number, promoter and enhancer specificity and activity, and insertion-site modulation of expression. Moreover, if the protein is inactive in the mouse because its receptor specificity is species specific, the influence of its immunomodulatory activity on the immune response to self and other proteins will be lost. Second, as previously mentioned, mouse MHC proteins differ from human leukocyte antigen (HLA) proteins, and peptides of therapeutic proteins that might form T-cell immunogens with human HLAs may not do so in mouse MHC and vice-versa.

An easier and potentially useful approach to generating a mouse that is tolerant to a human protein utilizes the phenomenon of neonatal tolerance. By injecting mice with foreign proteins within the first 24 hours of life, it is frequently possible to render them tolerant to that protein.77 Such mice could then be used to test the ability of a human therapeutic protein product to induce immune responses. As with human-protein transgenics, caveats apply to this model regarding differences in the mechanism of tolerance in the neonatal model and in the expression of the protein in humans. Indeed, in neonatally tolerized mice, tolerance may wane over time in the absence of sustained expression of the human protein.35,78 Therefore, a more potentially useful model may involve a gene delivery approach with a retroviral or lentiviral vector expressing the human protein of interest.79 Studies of immune responses to human protein therapeutics in normal animals may have some utility in determining whether differences have emerged in a product or whether differences exist between similar products that cause them to be more or less immunogenic. For such studies to be useful, there must either be a low baseline of immune response or a slow temporal pattern of immune-response development in the animals, thereby providing a clear window to observe differences. The same strain, age, and sex of rodents, which originate from the same facility, should be utilized for such studies.

Animal studies are not helpful for determining the propensity of humans for hypersensitivity responses to protein therapeutics.


Immunogenicity of biological therapeutics is an important concern that can be examined using a risk-assessment strategy. Sensitive and qualified assays to assess immune responses, particularly to high-risk products, should be in place from the earliest stages of product development and should be performed in real time during the clinical trial. Consultation with FDA regarding an appropriate testing program is highly recommended.



References 1-73 can be found in Parts 1 and 2 of this article. See the November and December 2004 issues of

BioPharm International.

74. Abina M-A, et al. Thrombopoietin (TPO) knockout phenotype induced by cross-reactive antibodies against TPO following injection of mice with recombinant adenovirus encoding human TPO. J. Immunol. 1998; 160:4481-89.

75. Dale D, Nichol J, Rich D, Best D, Slichter S, Sheridan W, et al. Chronic thrombocytopenia is induced in dogs by development of cross reacting antibodies to the MpL ligand. Blood 1997; 90:3456-61.

76. Koren E. From characterization of antibodies to prediction of immunogenicity. Dev. Bio. (Basel). 2002; 109:87-95.

77. Marshall-Clarke S, Reen D, Tasker L, Hassan J. Neonatal immunity: how well has it grown up? Immunology Today 2000; 21:35-41.

78. Garza K, Agersborg S, Baker E, Tung KS. Persistence of physiological self antigen is required for the regulation of self tolerance. J. Immunol. 2000; 164:3982-89.

79. Zhang J, Xu L, Haskins M, Parker Ponder K. Neonatal gene transfer with a retroviral vector results in tolerance to human factor IX in mice and dogs. Blood 2004; 103:143-151.


The authors thank Drs. Karen Weiss, Keith Webber, Kathleen Clouse, Jay Lozier, and Andrew Chang for critical reading of the manuscript.

Amy S. Rosenberg, M.D., is director of the Division of Therapeutic Proteins, CDER, FDA, Bldg 29A, Room 2D-16, 5600 Fishers Lane, Rockville, MD 20857, fax 301.480.3256, amy.rosenberg@fda.gov.

Alexandra S. Worobec, M.D., is a Medical Officer in the Division of Therapeutic Biological Oncology Products, CDER, FDA, 1451 Rockville Pike, Rockville, MD 20852, alexandra.worobec@fda.gov.