Any explanation of how microbial
infections might set off autoimmune diseases must take into account the
observation that all individuals appear to harbor potentially autoreactive
lymphocytes, but that these cells remain innocuous unless somehow activated.
Mechanisms by which infectious agents might activate these cells fall into
two major classes: antigen-specific and antigen-nonspecific. The cornerstone
of the antigen-specific theory is epitope mimicry: an antigenic determinant
on one of the microbe's proteins is structurally similar to a determinant
on one of the proteins made by the host, although different enough to be
recognized as foreign by the host's immune system (Fig. 1). For T cells,
the focus of this Review, the determinants involved would be linear peptide
stretches of about 8–15 amino acids (aa) long. The immune response to the
microbial determinant would then cross-react with host tissue and eventually
result in autoimmune destruction. The antigen-nonspecific theory has several
variants: they are grouped loosely under the term "bystander activation".
For all these mechanisms, no particular microbial determinant is implicated.
For example, infection might cause host cell destruction, which results
in the release of large quantities of normally sequestered proteins. These
could then be trafficked to the draining lymph nodes or presented at the
invasion site. Alternatively, or in addition, microbial insult could alter
the phenotype of professional or nonprofessional antigen-presenting cells
(APCs), rendering them more effective by enhancing antigen-processing machinery,
display of major histocompatibility complex (MHC) molecules at the cell
surface or expression of costimulatory molecules. Such an insult could
also induce the synthesis of inflammatory cytokines such as tumor necrosis
factor-a (TNF-a)
and interleukin 1 (IL-1), whose activities include activation of APCs and
modification of lymphocyte migration patterns. Finally, infection might
provoke polyclonal lymphocyte activation via either a mitogen or a superantigen
effect.
Figure 1. Alternative
explanations for autoimmunity
Of late, epitope mimicry has been
the favored explanation for the proclivity of microbial infection to precipitate
an autoimmune reaction. This is partly due to increasing awareness that
T cell receptor (TCR) recognition of MHC-peptide complexes is extremely
degenerate, and does not even require primary structure homology between
two peptides presented by a given MHC molecule (3, 4). Yet, careful review
of the evidence for epitope mimicry leaves one with the impression that
the case is far from proven. Before discussing some of this evidence, it
may be worthwhile to state what criteria an "air-tight" case should satisfy.
We propose the following five criteria, which are similar to those presented
previously (1, 5).
(ii) The responsible microbial and
self-proteins should be identified: more specifically, the culprit epitopes.
The microbial epitope (and protein) should be able to elicit a T cell response
that is cross-reactive to the self-epitope (and protein). As discussed
below, difficulties in satisfying this criterion may arise because of the
increasingly appreciated degeneracy of the T cell repertoire.
(iii) The relevance of the microbial
and self-epitopes, and the cross-reactive response to them, needs to be
established because they are involved in the evolution of both the infection
and the autoimmune disease. Can APCs that display these epitopes be found
within the infectious-autoimmune lesion or the draining lymph nodes? Are
cross-reactive T cells a significant component of the immune response elicited
by the infection? Are they expanded during the autoimmune disease? Answers
to the last two questions should be more forthcoming as the use of MHC-peptide
tetramer technology becomes more widespread.
(iv) A requirement for both the microbial
and self-epitopes in the development of the autoimmune disease should be
demonstrated. This can be done by assessing the effect of deleting or altering
the epitopes of the microbial and self-proteins. Of course, care needs
to be taken to rule out the possibility that mutation of the microbial
protein in question does not change other relevant properties of the microbe,
for example, infectivity or replication.
(v) It must be established that T
cells elicited by the microbe and cross-reactive to the microbial and self-epitopes
can, and are necessary to, provoke the autoimmune disease. This can be
achieved with adoptive-transfer experiments, a TCR-transgenic mouse system
or a combination of the two. OspA–LFA-1 and antibiotic-resistant
Lyme arthritis
Lyme disease is a multisystem illness
that results from infection by the tick-borne spirochete Borrelia burgdorferi
(13, 14). A prominent late manifestation, particularly in North America
where the species B. burgdorferi sensu stricto predominates, is an inflammatory
joint disorder that resembles rheumatoid arthritis. In about 10% of patients
with Lyme arthritis, joint inflammation resists extended antibiotic therapy.
After drug treatment, no spirochetal DNA has ever been detected in the
synovial tissue or fluid of such patients, although it is easily found
before administration. This has provoked the hypothesis that antibiotic-resistant
borrelial arthritis is an autoimmune disease.
Support for the theory that autoimmunity
underlies Lyme arthritis has come from observations of an MHC association
and an anti-borrelia immune response associated in time and severity with
the arthritis symptoms. The majority of individuals with the treatment-resistant
disease have the HLA-DRB1*0401 or HLA-DRB1*0101 alleles which, interestingly,
are also more frequent in rheumatoid arthritis patients (15, 16). Also
characteristic of many individuals, and appearing coincidentally with the
onset of a prolonged arthritis episode, is a high titer of immunoglobulin
G (IgG) antibodies that recognize the outer surface protein A (OspA) of
B. burgdorferi (16, 17). T helper 1 (TH1) cells reactive to OspA are often
found as well (18-20). Finally, immunization with recombinant OspA has
been effective in preventing Lyme disease in two clinical trials (21, 22)
and is now available as a vaccine. These findings suggest that an HLA-DRB1*0401-
or HLA-DRB1*0101-restricted immune response that is OspA-specific somehow
precipitates joint-specific autoimmunity.
A breakthrough came when the immunodominant
HLA-DRB1*0401-restricted peptide of OspA was identified (23). With the
use of a computer alogarithm (24), the nine-residue peptide OspA(165–173)
was predicted to be the peptide most effectively bound by HLA-DRB1*0401;
this was confirmed experimentally in competitive binding assays. In addition,
when injected with OspA protein, mice that were transgenic for HLA-DRB1*0401
responded primarily to the OspA(165–173) peptide, as did T cells from an
HLA-DRB1*0401+ antibiotic-resistant patient with Lyme arthritis, which
were challenged in vitro. A search of the Genbank Database identified one
human protein, leukocyte function–associated antigen 1a
(hLFA-1a), which
contains the peptide hLFA-1a(L332–340).
hLFA-1a(L332–340)
has homology to the dominant epitope of OspA and was predicted to bind
strongly to HLA-DRB1*0401 (which was eventually confirmed experimentally)
(23). Most importantly, synovial fluid T cells from patients with antibiotic-resistant
arthritis, but not antibiotic-sensitive or other inflammatory arthritides,
could respond to both OspA(165–173) and hLFA-1a
(25). Tetramer technology has now been applied to this field and has permitted
direct enumeration of HLA-DRB1*0401-restricted T cells that are OspA(164–175)-specific.
It has identified increased numbers of these cells in the synovial fluid
compared to in the blood of treatment-resistant arthritis patients (26).
It has also allowed the demonstration of OspA(165–184) and hLFA-1a(L326–345)
cross-reactivity at the single cell level, although for only 10% of OspA(165–184)-reactive
clones and with a markedly lower hLFA-1a(L326–345)
response (27).
These observations suggested the
following epitope mimicry scenario. B. burgdorferi sensu stricto infects
the host and disseminates to multiple tissues, including the joints. Some
time later, often after several months, an inflammatory immune reaction
begins in the joints; it is characterized in HLA-DR4B1*0401 individuals
by an anti-OspA IgG response and TH1 cell reactivity to the OspA(165–173)
peptide. Interferon-g
(IFN-g) produced
by the TH1 cells up-regulates expression of LFA-1a
on synoviocytes and invading leukocytes and HLA-DR4 molecules on APCs.
As a result, there is enhanced presentation of self-peptides derived from
LFA-1, which are either endogenously synthesized or phagocytozed, that
augments and propagates the inflammatory response, even after borrelial
antigens have been cleared. Support for this scenario comes from the finding
that LFA-1 is, indeed, highly expressed on cells infiltrating the synovia
of treatment-resistant Lyme arthritis patients, but not those with other
arthritides (28).
Thus, the case for OspA–LFA-1a
epitope mimicry being responsible for the initiation of antibiotic-resistant
Lyme arthritis fulfills two, perhaps three, of the criteria we have proposed.
First, the association between infection by B. burgdorferi sensu stricto
and the eventual development of chronic, treatment-insensitive arthritis
is well documented. Inflammation occurs once the microbe has been cleared,
but the immune response to the OspA protein and the time of onset and severity
of joint inflammation correlate well. Second, the culprit epitopes were
defined as OspA(165–173) and hLFA-1a(L332–340);
in addition, T cells that were capable of responding to both were elicited
in HLA-DRB1*0401-transgenic mice immunized with OspA. Third, dual-responsive
T cells were found specifically in patients with antibiotic-resistant Lyme
arthritis, especially in the inflammatory lesions.
However interesting and suggestive
this putative example of epitope mimicry appears to be, it cannot yet be
considered definitive. A major problem has been the lack of an adequate
rodent model. Thus, it has not been feasible to evaluate the effect of
engineered mutations of the OspA and LFA-1a
epitopes on arthritis development; nor has it been possible to provoke
arthritis in normal recipients by transferring dual OspA–LFA-1a–reactive
T cells. HLA-DRB1*0401-transgenic mice injected with OspA responded primarily
to the relevant OspA(165–173), but developed no signs of arthritis (or
any other autoimmune disease (23) ). This was probably because the corresponding
region of LFA-1a
differs between mice and humans (29). Another problem is the extensive
cross-reactivity that has been shown for T cells reactive to the OspA(164–173)
epitope: "supertopes" have been identified via amino acid substitution
analysis and used to screen protein databases. Many (475) supertope-matching
peptides were found in human or murine proteins and 16 of these could stimulate
at least one of seven OspA(164–175)-reactive T cell hybridomas (8). Thus,
one must consider the chance of finding some self-peptide epitope that
cross-stimulates with OspA purely due to "multiple sampling". A similar
point has been made concerning candidate T cell epitopes and epitope mimics
in a chronic borrelial disease of the central nervous system (9). A final
problem worth mentioning is that certain features of the proposed epitope
mimicry scenario remain unsatisfying: for example, there are no clues as
to what precipitates an inflammatory response several months after the
borrelial infection and no evidence for whether LFA-1a
peptides are actually being presented in the joints or, if so, by which
APCs.
A simplistic (almost certainly oversimplistic)
alternative scenario that needs to be ruled out is as follows. An OspA(165–173)-directed
immune response is made systemically (perhaps, but not necessarily, including
the joint). This leads to the production of anti-OspA and overproduction
of inflammatory cytokines, in particular TNF-a
and IL-1. Antibodies, cytokine effectors or both provoke a self-propagating
arthritis that is similar to those reported for different mouse models
(30, 31). This could explain the inflammatory joint response in the absence
of any evidence of the inciting microbe. In this scenario, dual reactivity
of synovial T cells for OspA and LFA-1a
is a chance event that merely reflects the impressive degeneracy of TCR
recognition of MHC-peptide complexes.
UL6-corneal antigen and herpetic
stromal keratitis
Infection of the eye with herpes
simplex virus 1 (HSV-1) can provoke a chronic inflammation of the corneal
stroma that is called herpetic stromal keratitis (HSK); HSK is a leading
cause of human blindness (32). Some, though not all, strains of mice develop
HSK when ocularly infected with HSV-1 isolated from infected human tissue,
and these strains provide a very useful animal model (33). In mice, the
disorder is thought to be mediated primarily by CD4+ TH1 cells, as they
will transfer the disease into ocularly infected immunodeficient recipients.
However, roles for CD8+ T cells, CD4+ TH2 cells and antibodies have also
been proposed (32). Somewhat paradoxically, stromal opacity in mice peaks
1–2 weeks after HSV-1 infection, when viral titers have plummeted and virus-derived
transcripts are no longer detectable (34). Therefore, it was inferred that
the perpetuation of inflammation that results in HSK is the manifestation
of an autoimmune response to a corneal antigen.
A more convincing, though still quite
indirect, argument for autoimmunity came when the basis of murine strain
variations in HSK susceptibility was determined. Development of HSK after
HSV-1 infection is controlled in a monogenic dominant manner by genes linked
to the Igh locus35, in particular by the genetic segment encoding the Ig
heavy chain constant region (36). Inbred strains carrying the Igha, Ighd
or Ighe alleles are susceptible to HSK, whereas those harboring the Ighb
allele are resistant (35, 37, 38). When injected in a tolerogenic mode
into Ighd animals shortly before corneal HSV-1 infection, purified Igs
from resistant Ighb, but not susceptible Ighe, mice protected Ighd animals
from HSK (39). Because transfer of T cells from HSV-1–infected Ighb mice
that had been injected with purified Igs did not provoke HSK in ocularly
infected immunodeficient recipients, whereas transfer of cells from uninjected
infected animals did, it was suggested that T cell tolerance had been affected.
This apparent tolerization led to
the hypothesis that HSK is mediated by T cells that recognize Ighb-derived
peptides. In concordance with this theory, injection of Ighd mice with
Ighb, but not Ighe, antibodies in an immunization mode elicited T cells
capable of provoking HSK in cornealy infected immunodeficient recipients.
Immunization with Ighb antibodies, specifically of the IgG2a isotype, was
as effective; indeed, two TH1 clones specific for IgG2ab, but not TH1 clones
of other specificities, could induce HSK under these conditions. In addition,
the two IgG2ab-specific clones responded to murine corneal extract in vitro,
but not to extracts from other tissues. The IgG2ab peptide responsible
for stimulating the two clones encompassed aa 292–308. This peptide could
block HSV-1–induced HSK when pre-injected under tolerizing conditions,
and immunization with this peptide also elicited T cells capable of provoking
HSK in ocularly infected immunodeficient hosts. On the basis of these results,
it was argued that HSK is an autoimmune disease induced by CD4+ TH1 cells
that are elicited by HSV-1 infection and reactive to both a corneal antigen
and IgG2ab.
The critical next step was to more
directly link the dual T cell reactivity to IgG2ab and a corneal protein
with reactivity to an HSV-1–encoded protein (40). The two HSK-inducing
TH1 clones mentioned above, C1-6 and C1-15, responded to extracts of HSV-1–infected,
but not uninfected, Vero cells. A search of the Genbank Database for HSV-1
proteins with sequence homology to the peptide IgG2ab(292–308) revealed
the best match with UL6(299–314), which had identical or chemically similar
amino acids at seven of eight sequential positions. A 15 residue peptide
that includes this sequence specifically stimulated both clones and also
prevented HSK when pre-injected under tolerizing conditions into HSV-1–infected
mice. In addition, when injected into animals under immunizing conditions,
it elicited T cells that could provoke HSK in cornealy infected immunodeficient
recipients. Strikingly, HSV-1 variants that lacked UL6, and which were
also replication-defective, did not make proteins that stimulated the two
keratogenic TH1 clones. In addition, they did not provoke HSK in susceptible
strains, whereas another replication-defective isolate still could. These
results led to the proposition that HSV-1 infection elicits TH1 reactivity
to the UL6(299–314) peptide and cross-reactivity to an as yet unidentified
corneal antigen; the latter precipitates a corneal inflammation that culminates
in HSK. A second cross-reactivity to the IgG2ab(292–308) peptide results
in tolerization of UL6(299–314)-reactive T cells in Ighb strains and resistance
to HSK.
This body of data seems to add up
to a strong case for T cell epitope mimicry: it is arguably the strongest
to date. There is at least partial satisfaction of four of the five criteria
listed above. First, there is a clear association between HSV-1 infection
and HSK, although it must be said that the correlation has not yet been
extended to the response to a particular HSV-1 protein. Second, the inciting
microbial epitope has been identified as UL6(299–314); the culprit corneal
self-epitope has not yet been defined, in fact, direct evidence of cross-reactivity
to a corneal antigen is limited to two T cell clones. Third, the requisite
microbial epitope deletion experiment has been done but, obviously, not
the corresponding self-epitope deletion analysis. Fourth, the two UL6(299–314)-reactive
T cell clones could provoke HSK in cornealy infected immunodeficient hosts,
as could T cells from mice immunized with the UL6 peptide, although ocular
insult was always required.
Given this impressive body of supporting
data, and the elegance of certain of the experiments, reports that question
a role for epitope mimicry in this context—at least mimicry involving the
UL6 sequence—have been provocative. Two ovalbumin-specific TCR-transgenic
mouse lines on a recombination-activating gene–deficient background developed
severe HSK upon HSV-1 infection. This was despite the fact that their monoclonal
TCRs were not reactive to an HSV-1–encoded protein and they did not develop
anti–HSV-1 T cell reactivity (41, 42). In addition, no cross-reactivity
in the responses of Ighdmice to injection of the UL6(299–314) and IgG2ab(292–308)
peptides could be detected. Surprisingly, the anti-UL6(299–314) response
did not even cross-react with extracts of HSV-1–infected cells (42). In
addition, when B cell–deficient Ighbmice—which were now capable of responding
to IgG2ab(292–308)—were infected with HSV-1, they developed HSK, but no
IgG2ab(292–308)-responsive T cells could be found, nor even any UL6(299–314)-reactive
cells.
These findings clearly bring into
question a role for molecular mimicry that involves the UL6 and IgG2b peptides.
They suggest alternative hypotheses (Fig. 1) that invoke bystander activation
and note that HSK in the animal model always requires corneal insult. According
to one scenario, ocular HSV-1 infection results in local injury. This promotes
a proinflammatory environment in the cornea, permits CD4+ T cells of any
specificity to enter when they normally could not and provokes their activation
and further differentiation in an antigen-nonspecific manner that is probably
mediated by cytokines. Although it fits the data nicely, this scenario
is not entirely satisfying because the supporting data rely too heavily
on systems—both in the TCR-transgenic and B cell–deficient mouse lines—that
do not permit effective clearance of HSV-1 after the infection. Because
of this they are not precisely reflective of standard infected mice. A
related scenario would be that a local anti–HSV-1 response is involved,
but that HSK results from intermolecular epitope spreading or simply virus-induced
immunopathology.
There is currently no clear explanation
for the discrepancies between these two sets of results, although they
might be related to the different strains of mice or viruses employed and/or
to other aspects of the analysis systems used. The further complexity of
T cell precursor frequency was additionally highlighted when it was found
that epitope mimicry is essential for HSK development after low-level HSV-1
infection of animals harboring a limited number of autoreactive T cells,
whereas innate immune mechanisms sufficed with stronger infection and higher
T cell numbers (43).
Judgment
We have weighed what we consider
to be the two best-argued examples of T cell epitope mimicry between microbial
and self-peptides that participate in autoimmune disease. Our conclusion
is that the case is not yet convincing enough to espouse, either for these
two examples or for the many others that have been reported but are based
on sparser substantiating data. The increasingly more appreciated degeneracy
of the T cell repertoire (8-12) implies that the potential for such a role
clearly exists, and new computational and experimental screening tools
make identification of candidate epitopes almost too easy. Indeed, those
attracted by the concept of epitope mimicry must now be wondering less
about how autoimmunity is provoked and more about why it does not happen
more often. What we need at this point are new approaches for subjecting
the candidates to experimental validation, in particular in the difficult
human system.
References
(i) As a prelude, an association
between the particular microbial infection and the particular inflammatory
state should be sought. This might be correlative, showing a temporal relationship
between the two or demonstrating that the severity of the inflammation
is influenced by the strength of the infection. Or it might be causal,
directly showing, usually in an animal model, that the specified infection
precipitates the specified inflammatory state. To establish the autoimmune
nature of the inflammation, it is important to show that it persists in
the absence of the inciting microbe. This criterion may be challenging
to satisfy in some cases: the microbe may have been cleared too long before
disease manifestation to have been noted; it may be a common infecting
agent, provoking disease only in combination with more rare genetic or
environmental elements; or it may just prime the immune system, the immediate
disease-provoking stimulus being a second virus or some nonspecific "adjuvant".
These criteria have been designed
to stringently distinguish between a mechanism that involves epitope mimicry
and the diverse mechanisms that rely on the different variants of bystander
activation. Within this framework, we will next examine two of the most
often cited examples of epitope mimicry.
© 2001 Nature Publishing
Group