More MS news articles for Nov 2001

Lessons from animal models for human autoimmune diseases

Nature Immunology 2, 781 - 784 (2001) © Nature America, Inc.
Veena Taneja & Chella S. David
Department of Immunology, Mayo Clinic, Rochester, MN 55905, USA. (

A variety of different animal models are used to study autoimmune disease. What have we learned?

Scientific literature contains a plethora of mouse models touted to be valuable in studies of autoimmune states with unknown etiology and pathogenesis. Although no single animal model perfectly mimics a human disease, experimental animal models afford opportunities to study the role of individual genes in order to investigate the genetic, environmental and pathogenic aspects of an autoimmune disease. Different animal models of the same autoimmune disease have unique values: some models highlight genetic factors, whereas others emphasize effector mechanisms. What is important, however, is the integration of the diverse data the models generate into a coherent framework for understanding autoimmunity. The parameters of autoimmunity are the same: genetic predisposition, the "trigger" that causes disease onset and the affected target tissue, which determines the specificity of the disease (Fig. 1). Here we will discuss some of the major spontaneous and induced animal models for human autoimmune diseases, their contributions and inadequacies, what future animal models we need and how animal models can be used in the management of human diseases.

Figure 1. The basic parameters of autoimmunity are the same.

In the thymus, a self-peptide with high binding affinity for HLA molecules leads to negative selection of T cells. In contrast, a self-peptide with low binding affinity can lead to positive selection of T cells, some of which could be potentially autoreactive. Some of the HLA molecules are able to positively select more autoreactive T cells and may predispose an individual to autoimmunity. Viral or bacterial antigens that mimic self-antigens are presented in the context of HLA molecules and can activate these autoreactive T cells in the periphery, thus triggering an autoimmune response. The response may be directed to an organ that overexpresses costimulatory molecules or cytokines and expresses an antigen that matches the specificity of autoreactive T cells. This in turn leads to inflammation, expansion of B cells, pathogenesis and destruction of the organ.

Multigene models

Among the multigenic spontaneous models for human autoimmunity, two models—NOD (nonobese diabetic, for diabetes) and NZB/W (lupus)—are particularly significant. A thoroughly studied model of spontaneous autoimmunity is autoimmune diabetes in the NOD mouse, which closely resembles human type 1 diabetes mellitus (also called insulin-dependent diabetes mellitus, or IDDM) (1). They share a similar pathology and immunological basis, and islet-reactive CD4+ T cells and CD8+ T cells are implicated in both diseases. Although the autoantigen is not known, both NOD mice and humans show responses to glutamic acid decarboxylase (GAD), insulin and insulinoma-associated protein 2 (Ia2). The T cell response is predominantly restricted by the NOD major histocompatibility complex (MHC) class II molecule, I-Ag7, which shares homology with the human IDDM-predisposing molecule HLA-DQ8. Both I-Ag7 and DQ8 carry non–aspartic acid residues at position 57 on the b chain. In germ-free colonies of NOD mice, the onset of disease occurs spontaneously and within a defined time frame. However, in conventional facilities, the onset of the disease is variable and the contributions of other non-MHC genes critical. Thus, this model is strain-specific and requires the involvement of multiple genes. Heterozygosity in general leads to abrogation of disease, which suggests that most of the genes critical for onset may be recessive. In human type 1 diabetes, class II genes play a dominant role and, in some cases, there is actually a higher relative risk of disease for heterozygous individuals, which involves two predisposing haplotypes: HLA-DQ8–HLA-DR4 and HLA-DQ2–HLA-DR3, for example. Thus, although several of the non-MHC genes implicated in human type 1 diabetes may be important within a population, in an individual, the presence of the predisposing class II gene combined with one or two environmental or genetic influences can cause the onset of disease. Although the NOD mouse has provided considerable information about the multiplicity of genes that contribute to type 1 diabetes, an animal model in which individual genes and environmental influences can be identified is warranted.

Lupus is a complex disease that is characterized by autoantibody production and involves many organs and many antigens. Several murine models of spontaneous lupus-like disease have been described, with NZBWF1 being the best characterized model. (2) NZB/WF1 hybrids develop lupus-like autoimmunity characterized by immunologlobulin G autoantibody production and progressive severe glomerulonephritis that closely resembles human systemic lupus erythematosus (SLE). As in human lupus, associations with MHC alleles correlate with nephritis susceptibility in certain mouse strains. Genetic analysis in the NZB/WF1 mice and their derivatives have provided a wealth of information on how various genes contribute to specific pathological events in the disease process. Similar to human lupus, data from these studies indicate that the disease expression requires interactions between several non-MHC and MHC genes. Studies with the "speed congenics" have identified several genomic intervals that contribute to different degrees of SLE pathogenesis (3). The identification of murine susceptibility genes should provide insights into the genetic mechanisms involved in the development of this complex disease in predisposed individuals and an opportunity to identify syntenic genes in humans.

MHC-transgenic models

Among the MHC transgenic (Tg) mice that develop spontaneous diseases, two animal models—the HLA-B27–expressing rat and mouse models (for human spondyloarthropathies) and the HLA-DQ8 rat insulin promoter (RIP).B7-1 C57BL/6 (for type 1 diabetes)–Tg mice—are noteworthy. The gene encoding HLA-DQ8 is the predominant predisposing gene in human type 1 diabetes. Mice expressing the HLA-DQ8 transgene are not susceptible to spontaneous diabetes. However, they do lose their tolerance to self-GAD and potential autoreactive T cells in the periphery cause insulitis in the pancreas, although further pathogenesis is absent. The same phenomena could occur in humans with predisposing type 1 diabetes MHC genes. The onset of diabetes may require a second insult in the pancreas, perhaps in the form of a viral or bacterial infection, overproduction of a cytokine or overexpression of an accessory molecule. To simulate such conditions, HLA-DQ8 (type 1 diabetes–predisposing)- and the HLA-DQ6 (diabetes protective)-Tg mice were crossed with RIP.B7-1–Tg mice, which—in the context of RIP—express the costimulatory molecule B7-1 in the b cells of islets (3). HLA-DQ8 RIP.B7-1–Tg mice develop spontaneous diabetes, whereas HLA-DQ6 RIP.B7-1–Tg mice do not (4). Thus, the HLA-DQ8 RIP.B7-1–Tg mice represent an excellent model for human type 1 diabetes. The disease is dominant in these mice so that the contribution of various genetic and environmental elements to the onset, modulation and protection of the disease can be investigated. Interestingly, one class II molecule (HLA-DR4) can down-regulate diabetes in the DQ8 RIP.B7-1–Tg mice, which suggests interactions between various class II molecules in the disease process (5).

The first spontaneous disease models that expressed a human HLA were HLA-B27–Tg rats and mice (6, 7). When animals are moved from pathogen-free environments to conventional colonies, animals from both species develop spondylitis-like features, which indicates that the onset of the disease requires an environmental trigger. The disease primarily affects male rodents, as with the human disease. These models confirmed that the gene encoding HLA-B27 itself, and not other closely linked genes, is responsible for spondylitis. But the model also generated controversies and additional questions. Why did the B27-Tg rats require more than 100 copies of the genes encoding HLA-B27 and human b2-microglobulin (b2M) for the onset of disease and why did HLA-B27 mice develop disease only in the absence of the mouse b2M? Studies in mice have suggested that the B27 molecule may reach the cell surface as an empty or free heavy chain in the absence of b2M and present an exogenous antigen to activate autoreactive T cells. Data from studies on human cells have shown that b2M-free B27 homodimers are transported to the cell surface and have the capacity to load exogenous peptides (8). These homodimers resemble a class II molecule that can present peptides in a classical class II antigen-processing pathway. Thus, the studies in rodent models and humans are converging in delineating spondylitis.

TCR-transgenic models

There are two well studied T cell receptor (TCR)-Tg mice that develop spontaneous autoimmune disease. BDC2.5-Tg mice express rearranged genes encoding a and b TCR chains that are derived from the islet-reactive diabetogenic CD4+ T cell clone BDC2.5. Although these mice provide valuable information about what processes make T cells diabetogenic, disease cannot be transferred, and only 20–25% of the TCR-Tg NOD mice develop disease when housed in specific pathogen-free conditions (9). Other studies have shown differences in the T cell clones generated from T cells of NOD mice and TCR-Tg mice (10), limiting the use of TCR-Tg mice in disease pathogenesis. This Tg mouse is an excellent model with which to study the role of the TCR, the specificity of T cells and the peptides involved in disease. But caution needs to be exercised in extrapolating this information to human diseases because diabetic patients do not have T cells expressing a predominant TCR.

KRN mice express rearranged TCR genes from a T cell hybridoma that recognizes amino acids 41–61 of bovine pancreas ribonuclease in the context of H-2Ak. Crossing KRN mice with a NOD strain (H-2Ag7) leads to the development of spontaneous systemic arthritis, which shares many features with rheumatoid arthritis (RA), in the transgene-positive offspring (11). The disease is triggered by recognition of self-MHC–peptide complexes by Tg T cells, which leads to a breakdown in tolerance and self-reactivity. The molecular target of both T and B cells is an enzyme of the glycolytic pathway, glycosyl-phosphatidyl inositol (GPI), which is essentially expressed in all the tissues including joint cartilage. However, it differs from the human disease by the absence of rheumatic factor (RF), the presence of inflammation of the spine and absence of germinal centers in the lesions. Also, the disease is more aggressive in mice, with an excess of myeloid cells infiltrating the synovial membrane. The disease is restricted by H-2Ag7, a molecule very similar to human HLA-DQ8 which is linked to RA. GPI is expressed on the surface of synovial lining and anti–GPI IgG can form immune complexes with GPI with subsequent immune response leading to development of joint pathology (12). This model may be important in dissecting human disease even though the requirement of oligoclonal T cells may not be true for humans.

Knockout models

Mice deficient in interleukin 2 (IL-2), IL-2 receptor-a, IL-10 and with mutated genes encoding TCRa develop spontaneous inflammatory bowel disease (IBD), which has similarities to human Crohn's disease. IL-10–deficient mice develop chronic enterocolitis, due to aberrant immune responses to normal enteric antigens, that progresses with age (13). Chronic enterocolitis shares immunological, pathological and physiological similarities with human IBD in its pattern of inflammation, increased colonic aerobic bacteria with accompanying decreased lactobacilli, increased gut and intestinal permeability and endotoxemia. Studies in some models have led to clinical trials that use various cytokines to treat the human disease. As human IBD seems to result from a genetically determined defect, it would be useful to have a mouse model of this disease to identify triggering antigens, characterize negative and positive TCR thymic selection and dissect the complex network of cytokines and their regulation.

Induced models

Experimental studies in induced models have the advantage over those in spontaneous models in that the onset and progression of disease can be controlled. Viral and other microbial infections are implicated in triggering immune responses to host autoantigens that are cross-reactive. Histories of viral infection preceding the onset of disease support this theory. In the context of autoimmune diseases, attention has focused on the search for the autoantigen that initiates the immune response. Most of the induced animal models use viral antigens or the self-antigen of the disease-specific tissue target. The induced models also have the advantage that they can focus on single gene effects with the use of transgene or gene-deletion models.

Although it has been proposed that some autoimmune diseases may have a viral etiology, virus-induced autoimmunity remains a controversial topic. It is possible that many viruses can set the stage for generating autoimmune processes. Evolution has led to selection of efficient MHC molecules, which can present multiple epitopes of infectious agents to activate T cell populations and clear infection. Although these molecules can effectively present viral antigens, they may predispose an individual to autoimmunity. Infectious agents may have a role in the onset of some human autoimmune diseases. Several virus-induced models for autoimmune diseases have provided important information.

Epidemiological studies of multiple sclerosis (MS) provide the strongest evidence for the involvement of viral etiology in the onset of disease. Theiler virus–induced demyelination, a model for human MS, has several features that are similar to the human disease: an immune-mediated demyelination, involvement of CD4+ T helper cells, delayed type hypersensitivity response to antigens and pathology. The Theiler virus was isolated from mice with demyelinating disease. In the presence of some class I molecule alleles, the virus is immediately cleared and the mice remain healthy. In the context of other class I molecules, the virus is not cleared and replicates in the brain. This leads to three possible scenarios. In the first, the immune system is incapable of clearing the infection and the massive viral load kills the animal. In the second, the immune system eventually resolves the viral load and the animal lives a normal life. In the third, a stand-off leads to a chronic infection in the brain and activation of self-myelin–reactive T cells and autoantibodies, which causes demyelination and disease. This mouse model may provide a scenario that closely resembles the human disease. In HLA-Tg mice, class II molecules altered the severity of demyelination without influencing viral load (14).

Several animal models have been developed in which a potential or known autoantigen is used to induce disease. In humans, such autoantigens are seldom involved in the onset of disease, but are potential targets. Loss of tolerance to autoantigens would result in the positive selection of self-reactive T cells. In most cases, induction of disease requires the use of complete Freund's adjuvant, thus indirectly involving a microbial agent. Although induced models may be artificial, the resultant clinical signs and pathology resemble human diseases of unknown etiology. Over the past decade, numerous studies and approaches have led to identification of candidate autoantigen for many diseases: myelin basic protein (MBP), proteolipid protein (PLP) and myelin oligodendrocyte glycoprotein (MOG) in experimental autoimmune encephalomyelitis (EAE); thyroglobulin in thyroiditis; acetylcholine receptor in myasthenia gravis; interphotoreceptor retinoid-binding protein in uveitis and type II collagen in RA.

RA is characterized by synovial inflammation and the destruction of joints. Immunization of some mouse strains with heterologous type II collagen leads to development of collagen-induced arthritis (CIA) with joint histopathology similar to that seen in RA. In both mice and humans, pathology is mediated by T and B cells. The disease susceptibility in mice is restricted to H-2A loci (q,r) (the human analog of DQ), whereas protection is mediated by a polymorphism in the I-Eb loci (the human analog of DR). RA has been explained on the basis of the "shared epitope": the third hypervariable region (HV3) (amino acids 67–74 of DRB1*0401), and other alleles sharing this region, present a putative arthritogenic epitope that results in initiation of autoimmunity. On the basis of extensive studies in CIA, we propose an alternative hypothesis: the DRB1 "shared epitope" shapes the T cell repertoire in the thymus by serving as a self-peptide presented by DQ molecules. Thus, a low-affinity binding DRB1 HV3 peptide would positively select potential autoreactive T cells, whereas high-affinity binding peptide would lead to negative selection (15). Our model implies a role for both the DR and DQ haplotype in RA predisposition. In humans, it is difficult to study the function of individual class II genes, as they are expressed in haplotypes and strong linkage disequilibrium occurs between DR and DQ alleles. Thus, HLA class II DR and HLA class II DQ-transgenic mice that lack endogenous class II molecules were generated to study the role of MHC in arthritis. These human-mouse models have helped to define the role of HLA in arthritis (16). An important role of HLA-DQ in arthritis was evident when Ao.DQ8-Tg mice developed severe CIA. Studies with Tg mice that express both DR and DQ alleles, show that polymorphism in DRB1 loci could modulate the DQ-restricted disease by being protective, permissive or neutral, thus supporting our hypothesis (17). The limitations of this model are that it bypasses the role of class I molecules and type II collagen may not be the only target in the human disease.

The rodent model of EAE is the most commonly studied model for MS. Despite intensive study and knowledge of the target cells, genetic factors that influence the severity of MS and the etiology of the disease are not understood. The autoantigens used to induce EAE are components of myelin, either crude myelin extract or purified basic MBP, PLP or MOG. Although these models imitate some histopathological aspects of MS, the extent of demyelination and distribution of lesions are different. Like MS, susceptibility to EAE is restricted by class II genes. Studies with HLA-Tg and human TCR–Tg mice show that HLA class II molecules can mediate disease by presenting an MBP self-peptide to T cells (18). However, the variations of the MHC alleles associated with MS in different ethnic populations make it difficult to assess the role of these alleles in the severity of or predisposition to disease. EAE induced with MOG is characterized by chronic (relapsing and progressive) disease with central nervous system demyelination. This model suggests an important role for autoantibodies in the disease. As in human disease, genetic background and gender of the animals also have an impact. The initial event that leads to T cell activation in MS remains elusive, although cross-recognition between infectious microbes and myelin antigens have been implicated.

Future models

We need to improve the existing models to make them more closely resemble human disease. Models with human MHC, without endogenous mouse alleles should be developed so that immune-mediated processes are restricted by human molecules. Mice expressing two or more HLA genes that simulate human haplotypes might reflect variations in human disease. New spontaneous disease models that combine HLA-Tg mice with cytokines, costimulatory molecules, viral Tg and knockout mice would further refine human disease models. Use of these models and the availability of the mouse and human genome maps should help decipher the role of non-MHC genes in human autoimmune disease.


  1. Leiter, E. H., Gerlingof, I. C. & Flynn, J. C. in Experimental Models of Diabetes (ed. McNeill, J. H.) 257-294 (CRC Press, Boca Raton, FL, 1999).
  2. Drake, C. G., Rozzo, S. J., Vyse, T. J., Palmer, E. & Kotzin, B. L. Immunol. Rev. 144, 51-74 (1995). | PubMed | ISI |
  3. Wakeland, E. K., Wandstrat, A. E., Liu, K. & Morel, L. Curr. Opin. Immunol. 11, 701-707 (1999). | Article | PubMed | ISI |
  4. Wen, L. et al.. J. Exp. Med. 191, 97-104 (2000). | PubMed | ISI |
  5. Wen, L. Chen, N. Y., Tang, J. Sherwin, R. & Wong, F. S. J. Clin. Invest. 107, 871-880 (2001).
  6. Hammer, R. E., Maika, S. D., Richardson, J. A., Tang, J. P. & Taurong, J. D. Cell 63, 1099-1112 (1990). | PubMed | ISI |
  7. Khare, S. D., Bull, M. J., Hanson, J., Luthra, H. S. & David C. S. J. Immunol. 160, 101-106 (1998). | PubMed | ISI |
  8. Allen, R. L., O'Callaghan C. A., McMicheal, A. J. & Bowness, P. J. Immunol. 162, 5045-5048 (1999). | PubMed | ISI |
  9. Suri, A. & Katz, J. D. Immunol. Rev. 169, 55-66 (1999). | PubMed | ISI |
  10. Dobbs, C. M. & Haskins, K. J. Immunol. 166, 2495-2504 (2001). | PubMed | ISI |
  11. Matsumoto, I., Staub, A., Benoist, C. & Mathis, D. Science 286, 1732-1734 (1999). | Article | PubMed | ISI |
  12. Schaller, M., Burton, D. R. & Ditzel, H. J. Nature Immunol. 8, 746-753 (2001).
  13. Kuhn, R., Lohler, J., Rennick, D., Rajewsky, W. & Muller, W. Cell 75, 263-274 (1993). | PubMed | ISI |
  14. Drescher, K. M. et al. J. Clin. Invest. 101, 1765-1774 (1998). | PubMed | ISI |
  15. Zanelli, E., Gonzale-Gay, M. & David C. S. Immunol. Today 16, 274-278 (1995). | PubMed | ISI |
  16. Taneja, V. & David C. S. Immunol. Rev. 169, 67-80 (1999). | PubMed | ISI |
  17. Taneja. V., Griffiths, M. M., Luthra H. S. & David C. S. Int. Immunol. 10, 1449-1457 (1998). | Article | PubMed | ISI |
  18. Madsen, L. S. et al. Nature Genet. 23, 343-347 (1999). | Article | PubMed | ISI |

© 2001 Nature Publishing Group