Molecular Mimicry
Molecular mimicry, bystander activation, and viral persistence with or without epitope spreading are three mechanisms that can initiate immunoreactivity leading to autoimmune disease. It is relatively easy to envisage how molecular mimicry could induce autoimmunity. Molecular mimicry represents a shared immunologic epitope with a microbe and the host (
33). For example, individuals with rheumatic fever can develop an autoimmune disease due to infections with group A beta-hemolytic streptococci. Sera from infected individuals can have antibodies reactive with heart, joints, brain, and skin (
158). Heart reactive autoantibodies can be removed by absorption with whole group A streptococci or cell wall preparations. Monoclonal antibodies derived from rheumatic fever patients cross-react with streptococcal antigens such as the group A carbohydrate antigen and the M protein (a virulence factor associated with streptococci) and myosin. Cross-reactive peptides from M protein and cardiac myosin can induce autoimmune disease in mouse models of rheumatic heart disease (reviewed in reference
21). This is one of the best examples of molecular mimicry in autoimmune disease (
21).
In a viral system, viruses have been shown to have cross-reactive epitopes with host self proteins (
33). One of us (R.S.F.), with colleagues, produced various monoclonal antibodies to measles virus and herpesviruses (
33). As expected, most of the monoclonal antibodies reacted with cellular proteins from uninfected cells; some of the antibodies were viral specific (reacting with only viral antigens) and a few monoclonal antibodies reacted with both viral and cellular proteins. An extension of this observation was published in a study by Srinivasappa et al. (
125) showing that almost 4% of antiviral monoclonal antibodies also reacted with self proteins.
Mimicry can also take place at the level of the T-cell. We had previously shown that the hepatitis B virus polymerase shared an immunologic epitope with myelin basic protein (MBP) (
32). When the viral peptide was injected into rabbits, some of the animals developed an experimental autoimmune (allergic) encephalomyelitis (EAE)-like disease, had T-cell reactivity, and developed antibodies to MBP. In subsequent years, Wucherpfennig and Strominger (
155) showed that viral peptides could activate autoreactive T cells against MBP. Similarly, Hemmer et al. (
51,
52), using combinatorial libraries, found that MBP-specific T cells reacted to a variety of viral and bacterial proteins. Therefore, cross-reactive immune responses between viruses and host are relatively common; but, in order for autoimmune disease to occur, we predict that the cross-reaction takes place between the virus and host at a “disease-related” epitope. If this does not occur, autoimmunity may arise but no disease transpires.
Disease-inducing epitopes are those peptides of autoantigens that can be presented by major histocompatibility complex (MHC) class II molecules on antigen-presenting cells (APCs) to autoreactive CD4
+ T cells. Some of the MBP peptides that can induce EAE in different animal species are reviewed by Alvord (
5). These epitopes, when injected with complete Freund's adjuvant (CFA) into the appropriate species and strains, can induce EAE. Use of peptides with slightly different amino acid compositions can result in protection or a downmodulation of disease. This is often known as the altered peptide ligand strategy to modulate disease. This approach was validated in EAE models of multiple sclerosis (MS) (reviewed by Martin et al. [
89]); but when used in MS patients, it met with mixed success (
12,
65).
In most if not all the models where molecular mimicry has been used to induce an autoimmune disease, an adjuvant such as CFA or an actual infection is required. This suggests that, in addition to having a cross-reacting disease, inducing epitope-sufficient activation of APCs is required.
Antigen-Processing Pathways
There are two basic pathways used to “present” viral antigens to host T cells. Endogenous proteins are processed and presented through the MHC class I pathway. Entry into the class I pathway begins when intracellular self or viral proteins are ubiquitinated via lysine amino acids within the protein. Additional ubiquitins are then attached to the original ubiquitin, and the resulting polyubiquitinated protein is targeted to the proteasome, where it is cleaved into short peptides. The peptides proceed into the endoplasmic reticulum via a transporter system where the peptides encounter MHC class I molecules (class I protein and β2 microglobulin). Depending on the conformation, charge, and sequence of the peptide, it will associate with the MHC class I molecule with different affinities. These peptide-MHC class I complexes proceed through the Golgi apparatus and are subsequently transported to the surface of the cell. These peptide-MHC class I complexes are then recognized by T-cell receptors (TCRs) on CD8+ T cells.
CD8
+ T cells could induce immunopathology that has the potential to initiate autoimmune disease by two nonmutually exclusive mechanisms. The first would be that the virus and the host contain a cross-reactive CD8 epitope. In the rat insulin promoter (RIP)-transgenic models for diabetes, the LCMV glycoprotein (GP) and nucleoprotein (NP) epitopes have been inserted into the genome of certain strains of mice. The expression of the viral epitopes is driven by the RIP such that the epitopes are found in the pancreas and regarded as self. Depending on the levels of expression in the pancreas and in the thymus, diabetes is induced in an acute or slow time frame. If no expression of the epitope is observed in the thymus, epitope-specific CD8
+ T cells are not negatively selected against and are present in the periphery. Upon infection with LCMV (encoding the cross-reactive epitope), an acute inflammatory response is mounted with diabetes appearing in a matter of days or weeks postinfection. If the epitope is found in the thymus, high-affinity CD8
+ T cells are mostly deleted. Upon infection with LCMV, these mice develop diabetes in weeks to months following infection. The generation of sufficient numbers of CD8
+ T cells requires CD4
+ T-cell help for expansion. This is reviewed in references
102,
104, and
142.
The second mechanism would be due to CD8
+ T cells killing virus-infected cells such as in the CNS (Fig.
1). Self antigens contained in dead or dying cells could then be presented by APCs to CD4
+ or CD8
+ T cells, leading to the autoimmune disease. Cytokines, particularly gamma interferon (IFN-γ), would be required to activate the APCs with upregulation of MHC class II molecules for efficient presentation to CD4
+ T cells. IFN-γ is essential for the development of diabetes in the RIP models (
144).
In both instances above, self proteins would be released into the environment and engulfed by additional dendritic cells and/or macrophages. These self antigens could then, with the appropriate stimulatory signals, such as costimulatory molecules and Toll-like receptor (TLR) signaling. on the APCs, stimulate autoreactive T cells, leading to further damage at the original site of infection or within the organ or tissue containing the self antigen (Fig.
1).
Generally, antigenic epitopes in exogenous proteins are presented via the MHC class II pathway. Exogenous autoantigens are taken up by specialized APCs by endocytosis and ultimately enter endosomes where they are degraded into peptides. Here, the peptides associate with MHC class II molecules. These peptide-MHC class II complexes are then transported to the surface of the cell. These complexes can then be recognized by TCRs on CD4
+ T cells. Class II molecules are only found on selected cell-types within the body, whereas class I molecules are found on most cells with the possible exception of neurons, but this is still somewhat controversial. Viral antigens could be phagocytosed by macrophages and dendritic cells and processed through the MHC class II pathway leading to activation of viral or self CD4
+ T cells (Fig.
1). There is some “cross talk” between the two pathways: occasionally, endogenous proteins can be presented via class II molecules and exogenous antigens can be presented by class I molecules on APCs (
1,
16,
48,
81,
150).
Generation of autoreactive T cells.
Depending on the experimental model or disease, CD8
+ T cells can be the effector cells, cytotoxic T lymphocytes (CTLs), that can cause cell destruction (
138), whereas CD4
+ T cells can also be the effector cell (CTL); but, conventionally the CD4
+ T cells activate macrophages, and macrophages are the effector cells (delayed type hypersensitivity response). In the diabetes model discussed below, effector cells are mainly CD8
+ T cells. Here, CD8
+ Tcells can directly kill islet cells and secrete proinflammatory cytokines. In the model of CNS autoimmune disease, EAE, the cells that can transfer disease are conventionally thought of as CD4
+ T cells (
95,
108,
113,
114,
130). These cells can secrete myelinotoxic cytokines that damage the oligodendrocyte and generate an inflammatory focus to which macrophages are recruited that in turn cause demyelination. Thus, in these instances, the actual mechanism of killing or tissue damage can be the CD8
+ T cells, where these cells can kill target cells directly, while CD4
+ T cells can initiate damage more by a bystander mechanism.
There is some debate whether oligodendrocytes or myelin can express MHC class II molecules in vivo. If they do not, then autoreactive CD4+ T cells would recognize self peptide on microglial or other class II-positive cells in the CNS and produce cytokines and chemokines resulting in the recruitment and activation of macrophages (cells of the innate immune system). Macrophages would release interleukin (IL)-1 and TNF and start engulfing myelin. Some of the targeting of macrophages to the myelin could also be due to myelin-specific antibodies. Macrophages would recognize myelin-antibody immune complexes via Fc receptors and begin engulfment of myelin.
Myocarditis: Autoimmune or Immune-Mediated Pathology?
Several forms of cardiac insult can result in myocarditis; but, we shall focus on virus induced myocarditis, and on whether the myocarditis is caused by (i) the infection itself; (ii) the immune response to the infection; or (iii) autoimmunity. Myocarditis is surprisingly common, as revealed by a necropsy study of more than 12,000 victims of violent or accidental deaths (that is, deaths which were, presumably, unrelated to heart disease); myocarditis was present in approximately 1% of these individuals (
41) indicating that, at any given time, ∼2 million Americans have inflammatory infiltrates in the heart. However, myocarditis is often asymptomatic; only a subset of cases, probably around 10%, exhibit clinical disease, developing symptoms such as chest pains, palpitations, or signs of heart failure. Individuals in the larger, symptom-free, group usually recover without obvious sequelae, but are by no means free of risk; acute myocarditis, even when asymptomatic, predisposes to catastrophic dysfunction of the electrical pathways in the heart and can lead to the collapse and death of young and vigorous individuals, especially during exertion (
11,
145).
Although the majority of symptomatic patients recover well from acute myocarditis, the disease can have serious long-term sequelae; some 10 to 20% of people with symptoms (i.e., ∼20,000 to 40,000 patients per year in the United States) will develop chronic disease, and a substantial proportion of these individuals progress over time to dilated cardiomyopathy (DCM) (
101,
124), which is thought to have an incidence (new cases per year) of 3.5 to 8.5 cases per 100,000 population (∼9,000 to 20,000 new cases annually in the United States) (
39). DCM is a serious condition in which one or both ventricles dilate and decompensate, with resulting cardiac failure. There is a 50% mortality in the 2 years following diagnosis (
40), and the most effective treatment is heart transplantation; indeed, DCM is the condition underlying almost half of all heart transplants (
49). In many cases, histological examination reveals extensive cardiac fibrosis suggestive of prior myocardiocyte destruction (
87).
Here, we shall focus on myocarditis induced by an enterovirus, type B coxsackievirus (CVB), which, as discussed below, is known to replicate in the heart tissue and to induce strong inflammatory responses therein. Therefore, the damage to heart muscle may be most simply explained by direct microbial cytolysis and/or by the immunopathological consequences of the antimicrobial immune responses. However, in addition to these straightforward explanations, autoimmunity has been invoked to explain the acute and chronic diseases mentioned above.
Coxsackievirus myocarditis.
Several viruses cause myocarditis, but the role of enteroviruses is very well established. Cardiovascular signs and symptoms are present in 1.5% of all enteroviral infections, and CVB is the commonest cause of infectious myocarditis; the incidence of cardiovascular symptoms is 3.5% for CVB and 0.7% for type A coxsackievirus and for another enterovirus, echovirus (
44). CVB has been isolated from the hearts of patients with myocarditis, CVB-related nucleic acid signals have been found (by PCR and in situ hybridization) in the myocardium, and serologic studies implicate CVB in the acute disease. Furthermore, CVBs isolated from stool or pharyngeal specimens of patients with acute myocarditis have been administered to mice and have infected the heart (
38,
153).
Demonstration of infectious CVB in the human myocardium has been more difficult, since myocardial biopsy remains unusual, but necropsy specimens have yielded infectious CVB (
38,
128,
129), which is cardiotropic in mice (
128). Slot blot hybridization studies have shown positive signal for CVB RNA in myocardial biopsy specimens of approximately 45% of patients with myocarditis or DCM compared with none of the controls (
90), and ∼43% of patients with healed myocarditis or DCM remained positive for CVB signal (
7). High levels of neutralizing antibodies are found in about 50% of patients, and serial antibody studies show a fourfold or greater change in paired sera in approximately half of patients (
90). As further evidence that enteroviruses may cause DCM, this chronic disease occurs in 10 to 20% of patients with proven prior enteroviral myocarditis, while its incidence in the total population is approximately 0.005%, and a large study confirmed this strong correlation (
P < 0.001) between prior coxsackievirus infection and DCM (
115). Acute myocarditis and DCM are, therefore, significant contributors to human morbidity and mortality, and the role of CVB has been clearly demonstrated. Several CVB3 isolates, when inoculated into normal mice, causes myocarditis (
37,
53,
70), pancreatitis (
92,
122), and neonatal CNS infections (
31) and thus faithfully recapitulate many aspects of CVB infection and disease in humans.
What Mechanisms Might Underlie CVB Myocarditis?
While there is no doubt that CVB3 can cause myocarditis in mice, the precise mechanism underlying this pathogenic outcome remains controversial. Five possible pathogenic mechanisms are outlined in Table
2. From this table, it is clear that, although both the acute and chronic diseases induced by CVB almost certainly have a large immunopathological component, this does not necessarily imply autoimmunity; mechanisms 1 and 2 are sufficient to explain the observed clinical phenomena, as long as the virus (or, at least, some viral materials) can persist in the host animal. So, what is the evidence for viral persistence?
In tissue culture, CVB can establish long-term persistent infection in a variety of cell types, including human myocardial cells (
50,
64) and human and murine lymphoid cells (
91,
157); infectious virus can be recovered over a period of weeks to months. The in vivo situation is less well understood. CVB RNA can persist for many months in skeletal muscle, apparently as double-stranded RNA, and RNA persistence correlates with the degree of myositis observed (
133-
135). However, in these and other in vivo studies, infectious virus could not be isolated at the later stages, despite the presence of CVB-related RNA sequences. It is important to draw a clear distinction between viral RNA and infectious virus; the two are not necessarily equivalent, and terminology such as “CVB persistence,” often used to describe the presence of CVB-related nucleic acid signal, should be employed only if infectious virus can be identified within, or reactivated from, the tissues.
Do CVB materials also persist in heart muscle? In vivo, CVB has been detected by in situ hybridization in biopsy specimens of human DCM patients; one could argue that this represented an acute infection, present by coincidence at the time of biopsy, but the failure to detect infectious virus suggests that an acute infection was not present. Recent studies in several mouse strains have shown long-term persistence of CVB-related nucleic acid signal in the heart, associated with chronic myocarditis and fibrosis (
68); the signal is found in several organs, including heart and is often highly localized, being found near regions of inflammation (
68,
69). The identification of CVB RNA long after the primary infection provides several potential explanations for chronic myocarditis: first, it remains possible that infectious virus may be sporadically reactivated; second, viral protein expression alone can be toxic to cells (
148,
149); and third, the upregulation of viral protein expression could lead to a recrudescent immunopathology. Thus, in principle, chronic myocarditis and DCM may be explained by persistent CVB materials and, as in the acute phase, there may be immunopathology, but there is no need to invoke autoimmunity.
But how might CVB materials persist, especially if infectious virus is not detectable beyond ∼14 days postinfection? Recent findings from several laboratories indicate that there are interactions between CVB and the infected cell; in particular, CVB may respond to, and may regulate, the cell cycle. A cell cycle effect on picornaviral replication was suggested by studies carried out some two to three decades ago (
27,
78,
86,
126), but it has not been clearly delineated, and, judging from its omission from recent reviews on virus-cell cycle interactions (
106,
131), appears not to be widely appreciated. These studies are described in several recent publications (
8,
30,
35,
82,
83,
93,
105,
134) and will be summarized only briefly here.
We have found that the outcome of infection of tissue culture cells depends on their cell cycle status; infection of quiescent cells (G
0) or cells blocked at the G
2/M phase leads to low levels of viral protein synthesis and inefficient production of infectious virus; but “release” of the cell, allowing it to pass through G
1, results in increased viral gene expression and infectious virus production. Thus, the virus appears to respond to the cell cycle status. Others have shown the reciprocal; the virus can affect the cell cycle, arresting cells at the G
1/S boundary, by increasing the degradation of cyclin D1 (
85). Therefore, the virus seems to have evolved (i) to arrest the cell at the stage most beneficial to the virus's replication and (ii) to remain quiescent in cells that fail to enter the G
1 stage.
What viral component might allow the virus to “sense” the cell status and to respond appropriately? Picornaviruses contain, in their 5′ untranslated region, an internal ribosome entry site (IRES), to which cell cycle-regulated proteins may bind, regulating picornaviral protein expression (
109), and some viral IRESs appear to respond to the cell cycle status in tissue culture (
140). IRES elements have been identified in cellular mRNAs, and many of the encoded cellular gene products are associated with the cell cycle (
111), although these cellular IRESs are most active in G
2/M, when CVB gene expression is low. Perhaps CVB has incorporated a cellular IRES, but subsequent modifications have allowed it to operate best when host translation is almost entirely cap dependent. In that way, the virus can kill two birds with one stone: it can shut down cap-dependent translation at a time when the host most relies on it and at the same time can very efficiently translate its own proteins in the absence of competing host IRESs.
Thus, many of the requirements are in place to explain CVB-induced myocarditis, without invoking autoimmunity. However, this merely shows that autoimmunity may not be required; it does not directly address whether or not it actually is responsible for the disease. We believe that three key questions must be asked. First, are autoreactive responses induced by CVB infection? If not, then autoimmunity can be dismissed as a cause of myocarditis. However, even if autoreactive responses are found, their mere presence does not prove that they are pathogenic; a second question must be asked, do the autoreactive responses contribute to disease? Only if the answer is affirmative should we approach the third question, what is the underlying mechanism of autoimmune disease? One might expect that myocarditis would result from a single autoimmune mechanism; but, over the past three decades, at least four distinct mechanisms have been proposed: autoantibodies, autoreactive MHC class I-restricted CD8
+ T lymphocytes, autoreactive MHC class II-restricted CD4
+ T lymphocytes, and most recently, T cells carrying γδ T-cell receptors. Further complicating the issue, it has been suggested that the mechanism of autoimmune postviral myocarditis may be dependent on the mouse strain; for example, that autoimmune T cells may be responsible in BALB/c mice and autoantibodies in the DBA/2 strain (
72).
Cardiac autoantibodies induced by CVB were first described in 1 of 55 sera that were screened for antimyosin antibodies; the serum that scored positive was from an individual who had coxsackievirus-caused pericarditis (
29). Since then, a large number of autoantibodies have been described in the sera of patients with myocarditis (summarized in reference
36), but the clinical relevance for many is unclear, because many of the target proteins are intracellular (
107). Autoreactive antibodies against cardiac myosin were identified in mouse models of CVB infection (
152), and an association was found between susceptibility to chronic myocarditis and the presence of autoreactive antibodies (
151). One possible explanation of these data was molecular mimicry; perhaps CVB infection induced antiviral antibodies that cross-reacted with myosin, but this was shown not to be the case (
98). In studies of chronic myositis, both CVB-specific antibodies and autoantibodies were found, but there was no statistically significant association with the extent of myopathy; rather, the autoantibodies appeared to be an independent reflection of the damage done by the virus infection (
132). Taken together, these data indicate that CVB myocarditis favors the induction of autoantibodies, but these may be the consequence of disease rather than its cause.
T lymphocytes have long been implicated in CVB-induced myocarditis (
154), and adoptive transfer studies have identified cytolytic CD8
+ T cells (at that time, known as Lyt-2
+ cells) as major players (
46). Subsequent analyses have confirmed and extended these findings; there is no doubt that CD8
+ T cells contribute substantially to the myocarditis that is induced by CVB3. But are the cells autoreactive (potentially causing autoimmune disease), or are they specific for viral materials? Adoptive transfer data identified CD8
+ T cells that appeared to recognize uninfected myocardiocytes, but the antigen target of these autoreactive cells was not identified (
57,
58), and subsequent analyses revealed two types of cytolytic T cells induced by CVB infection; autoreactive CD8
+ T cells and virus-specific CD4
+ T cells (
28).
Ongoing studies of the nonviral (and, almost certainly, autoimmune) myocarditis induced by inoculation of cardiac myosin had now progressed to a point at which autoantibodies were no longer considered a likely mediator of disease, and suspicion focused on T cells (
99,
100), in this case, and in contrast to the earlier report regarding CVB, the major autoreactive population of T cells were CD4
+. MHC class II-restricted peptides from cardiac α-myosin have been identified that, when inoculated with adjuvant, induce myocarditis in susceptible mice (
24,
110). However, a link to CVB-induced disease remains tenuous, because the T cells induced by those peptides do not cross-react with CVB, and these peptide-specific CD4
+ T cells have not been identified as a factor in CVB-induced myocarditis.
The most recent T-cell family to be implicated contains γδ T cells. Despite their having been discovered some time ago, the biological function of these cells remains unclear. They play a protective role in various noninfectious models of chronic stimulation (
73), wound healing (
61), and tumor immunity (
62), and they can be activated by nonspecific stimuli (
96), suggesting that foreign (e.g., viral) antigens may not be required for activation of many of these cells; their TCRs are, presumably, activated by unidentified endogenous materials and, as such, they may be categorized as autoreactive. Several functions have been ascribed to γδ T cells during CVB3 infection. The cells appear to directly interact with myocardiocytes and have been proposed as the main effector population responsible for myocardial injury associated with DCM-like signs during CVB3-induced myocarditis (
55). One population of γδ T cells appears to suppress CVB-induced myocarditis (
56), while another (expressing a different Vg receptor) exacerbates disease by secreting IFN-γ (
56), thereby activating CD4
+ T cells, which in turn are required to activate autoreactive CD8
+ T cells (
59). These data represent one of the few cases in which the biological role(s) of γδ T cells has been investigated during microbial infection, and it will be interesting to identify the antigen(s) recognized by these cell populations.
In summary, there is no doubt whatever that autoreactive antibodies and T cells can be induced during CVB infection. However, the evidence that these virus-induced autoreactive responses are themselves pathogenic is relatively scant, and the invocation of different mechanisms of autoimmunity in different hosts may not be necessary. Furthermore, it may be significant that immunosuppression is not a recommended treatment for myocarditis. If the chronic disease were autoimmune in nature, one would predict that immunosuppression might have been an effective treatment; that this treatment is not recommended indicates that an autoimmune mechanism is unlikely.
Using Occam's razor gives a simpler explanation: that the long-term disease results from reactivation of viral materials that have persisted in host cells, with consequent viral cytolysis and/or immunopathology. This concept is exemplified by the lifelong infection established by herpes simplex virus in dorsal root ganglia. Contrary to the prevailing wisdom, which holds that herpes simplex virus is truly latent for much of the time, it appears that herpes simplex virus does not remain silent within the ganglia; rather, it is constantly “trying” to reactivate, and this recrudescence is actively suppressed by CD8
+ T cells that recognize viral antigens (
66). This explains why immunosuppression leads to more frequent herpes simplex virus eruptions, because the immune system is unable to hold the virus in check. Perhaps CVB myocarditis should be viewed in the same light, and further studies should be focused on the mechanism by which CVB establishes persistence or latency and the circumstances that may lead to viral reactivation.