Autoimmune diseases in humans arise from the chance accumulation of a group of genes in the host that together heighten susceptibility to autoimmune responses [1]. The most prominent genes are members of the major histocompatibility complex (MHC), but a number of additional genes normally involved in regulation of the immune response have now been identified as contributing to the so-called autoimmune diathesis, that is, a genetic predisposition to develop autoimmune disease. In addition, stochastic variation is introduced by the specialized mechanisms that determine the broad diversity of immune system recognition, such as V(D)J recombination, receptor editing, and somatic mutation. Thus, the immune system has the potential to recognize the universe of pathogens, present or future. At the same time, antigens of the host will also be recognized. If appropriately stimulated, self-reactive lymphocytes can produce clones of lymphocytes capable of generating autoantibodies. Multiple control systems usually operate to limit and regulate such clones, but when regulation fails, autoimmune disease (or malignant disease) may result.

In addition to the genetic predisposition, it seems that an environmental factor is required to trigger the onset of the autoimmune process in humans. Among the many environmental triggers that have been implicated in the initiation of autoimmune disease, infection has remained the most prominent (Fig. 1). Infectious agents such as viruses can provide the autoantigen-specific stimulus required for activation of the T cell receptor (TCR) or B cell receptor. A number of mechanisms have been described to account for autoantigen-specific responses during or after infection including molecular mimicry, T cell degeneracy, epitope spreading, spillage of normally inaccessible antigens, and alteration of autologous antigens. Some experimental evidence for each of these potential mechanisms has been described [2]. Less attention has been directed to the requirement for additional nonantigen-specific stimuli to initiate the autoimmune response. It has long been realized that proliferation of T cells requires costimulation by a second signal in addition to engagement of the TCR. In naïve T cells, TCR engagement alone is usually ineffective in activating the pathways that are needed to induce production of proinflammatory cytokines and other growth factors. Insufficient input by secondary signals renders the cell refractory to TCR-stimulated growth, a phenomena referred to as anergy [3]. When TCR engagement alone initiates proliferation, it eventually results in T cell apoptosis because of insufficient induction of antiapoptotic factors [4]. These mechanisms prevent the unremitting proliferation of peripheral T cells that recognize self-antigens through their TCR alone. The second signal determines whether an encounter with a potential self-antigen is aborted at an early stage or proceeds to a pathogenic autoimmune outcome. Similar mechanisms prevail in the B cell. Collectively, the second signals necessary for induction of autoimmune disease have been referred to as bystander activation or the adjuvant effect. Information has accumulated rapidly in recent years that allows us to analyze the adjuvant effect at a molecular level.

Fig. 1
figure 1

Infection in the initiation of autoimmune disease

Full size image

Adjuvants do not have to arise from an external source. In fact, many studies have recently shown that endogenous adjuvants, especially in the form of small ribonucleoproteins, stress proteins, or products of damaged cells, can promote autoimmune responses [5–7]. Often, however, the responses are transient and do not progress to clinical disease.

Multiple sclerosis is an autoimmune disease that is associated with environmental exposure during early childhood. Several different viruses have been implicated. Some may act by molecular mimicry, whereas others provide the adjuvant effect. However, adjuvants can enhance or reduce immune responses. For example, in a model of multiple sclerosis induced by vaccinia virus expressing myelin protein, infection with the lymphocytic choriomeningitis virus can induce type 1 interferons (IFN) and reduce diseases, whereas the murine cytomegalovirus can activate antigen-presenting cells, increasing interleukin (IL)-12 and type II IFN and enhancing disease [8]. Instead of cytomegalovirus, complete Freund adjuvant can activate the immune system to induce disease.

Other autoimmune diseases can behave quite differently. For example, patients receiving type I IFN for treatment of malignancy show a significant increase in autoimmune thyroid disease [9]. In contrast, treatment of NOD mice with complete Freund adjuvant prevents the development of insulitis and autoimmune diabetes [10].

We first became aware of the importance of adjuvant in the induction of autoimmune disease during our early studies on the initiation of experimental autoimmune thyroiditis by immunization with thyroglobulin [11] (Table 1). We found that injection of purified thyroglobulin plus complete Freund adjuvant induced production of both thyroglobulin-specific autoantibodies and inflammatory lesions in the thyroid gland. In contrast, injection of thyroglobulin with incomplete Freund adjuvant, lacking the mycobacterial component, gave rise to autoantibody production but no lesions in the thyroid gland itself. We were able to follow up these observations in collaboration with Kong et al. [12] (Table 2). We found first that the injection of thyroglobulin with incomplete Freund adjuvant plus muramyl dipeptide, a component of a mycobacterial cell wall, induced characteristic lesions of thyroiditis. Similarly, injection of thyroglobulin and concurrent intravenous administration of bacterial lipopolysaccharide (LPS) induced the full picture of autoimmune disease with the production of both autoantibodies and lesions in the thyroid. On the other hand, most of the other adjuvants we tested at the time, including poly(A–U), alum, bentonite, silica, and silicone oil, induced only autoantibody production but not thyroid lesions. These findings closely associated the adjuvant effect with concurrent administration of products of infection.

Table 1 Immunization with thyroglobulin with complete Freund adjuvant and incomplete Freund adjuvant
Full size table
Table 2 Injection of thyroglobulin with complete Freund adjuvant, incomplete Freund adjuvant, muramyl dipeptide, lipopolysaccharide, and a polymer of adenine and uridine
Full size table

Several years later, we investigated the adjuvant effect from viral infection [13]. We found that in genetically susceptible strains of mice, we could induce autoimmune myocarditis by infecting the mice with a cardiotropic strain of coxsackievirus B3 (CB3). The autoimmune disease was characterized by production of autoantibody specific for the cardiac isoform of myosin. Subsequently, we were able to induce the same autoimmune disease by direct immunization of the susceptible mice with cardiac myosin or myosin-derived peptides in the absence of the virus (Fig. 2). For the production of myocardial lesions, all of these injections of cardiac myosin or myosin peptide, however, required coadministration of complete Freund adjuvant. Sometimes, we even included pertussis toxin as a second adjuvant to increase the severity of the disease. Because we were curious about the mechanisms by which infection alters the immune process, we tested the effect of CB3 infection in mice that were genetically resistant to the induction of autoimmune myocarditis. Under usual conditions, infection of some strains of mice, such as B10.A, with CB3 induced an acute limited viral myocarditis that was not followed by the autoimmune disease. If, however, we coadministered LPS with the virus, autoimmune myocarditis developed after the viral myocarditis subsided [14] (Table 3). Because we assumed that the role of LPS as a coadjuvant was to initiate a cascade of proinflammatory cytokines, we tested the coadministration of recombinant IL-1β and found that CB3 infection plus IL-1β also led to autoimmune myocarditis even in genetically resistant strains of mice [15].

Fig. 2
figure 2

Multiple methods induce EAM in mice

Full size image
Table 3 Infection in B10.A mice
Full size table

The same experiments were performed using cardiac myosin as an antigen. Immunization of B10.A in mice with cardiac myosin even with complete Freund adjuvant lead to no or only mild lesions. If, however, we included LPS with the immunization regime, then severe myocarditis was found in most of the mice [16], showing that with resistant animals, we require an even more potent adjuvant effect.

These experiments focused our attention on the very early response to the adjuvant, the response associated with natural innate immunity. With Fairweather [17], we found that the difference between strains of mice destined to develop autoimmune myocarditis and those resistant to it was apparent as early as 6 h after viral infection, taking the form of significant differences in the production of IL-1β (Table 4). To account for such a prompt early distinction between responder and nonresponder animals, we tested a number of potential antigen-presenting cells and found that there was a marked difference in the numbers of mast cells in the spleens of responder vs nonresponder mouse strains. The finding encouraged us to scrutinize the mast cell as a potential antigen-presenting cell in the initiation of the autoimmune form of myocarditis that may follow previous viral infection. We found, in fact, that mast cells not only upregulated MHC class II, a sign of activation but also the costimulatory molecules B7-1/2 and toll-like receptor 4 (TLR4) (Table 5). The experiments have led us to propose that the mast cell is a key early player in the generation of the adjuvant effect required for the initiation of an autoimmune disease after viral infection.

Table 4 Early cytokine response to CB3 (6 h)
Full size table
Table 5 Mast cells
Full size table

A final series of experiments was carried out using a surrogate marker of the adjuvant effect, where a sudden drop in the thyroid hormone, thyroxine, after immunization using complete Freund adjuvant occurred [18]. Referring back to our earlier experiments, we found that this drop in thyroxine occurred when mice were immunized with thyroglobulin and complete Freund adjuvant or thyroglobulin and LPS but not when animals were injected with thyroglobulin and incomplete Freund adjuvant or thyroglobulin alone. Looking further for a mechanism, we found that the adjuvant effect of LPS was demonstrable in wild-type mice but not in animals deficient in MyD88. This adaptor molecule is a cofactor in activating several members of the TLR family including TLR4 and TLR2. These experiments allow us to assign an obligatory role to the TLRs in the adjuvant effect of LPS. MyD88 is also essential for the induction of experimental autoimmune myocarditis by immunization with cardiac myosin and complete Freund adjuvant [19].

These experiments suggest that infection plays a major role in triggering autoimmunity. Accumulating data have shown that microbial products modulate the immune response by interacting with TLRs expressed on antigen-presenting cells. The resulting profile of cytokines, Th1, Th2, or Th17, in turn determines the pathogenicity of the autoimmune response. Returning to the mast cell, we then tested mice genetically deficient in mast cells and found that they also failed to develop the adjuvant effect when injected with complete Freud adjuvant or LPS.

In summary, we conclude that active infection or products of infection can provide the adjuvant effect necessary for the induction of many autoimmune disorders. The adjuvant effect depends upon giving early nonspecific signals to the initial innate immune response that later shape the adaptive immunity responsible for the pathogenic manifestations of autoimmunity.