Peptide-based immunotherapy of autoimmunity: a path of puzzles, paradoxes and possibilities

Introduction

Our understanding of the basic mechanisms controlling the generation of productive immunity continues to evolve at a great pace. Models of how the immune response may be diverted from tackling invading pathogens into mounting an inappropriate response towards self tissue antigens also continue to develop. Armed with this knowledge, however, we are yet to develop effective and broadly applicable strategies to prevent or treat autoimmune disorders. This review assesses the prospects for developing antigen-specific therapies through the use of synthetic peptide antigens, and specifically altered peptide ligands (APL), to target pathogenic T-cell autoreactivity.

Identification of T-cell epitopes within autoantigens has allowed the development of a minimalist approach to certain experimental models. Thus single peptides can replace purified autoantigen or homogenized tissue as the agents that elicit T-cell autoreactivity and therefore pathology. If we are using peptides to provoke disease, we should also be able to use peptides to prevent disease, and many reports over the last 15 years have confirmed this. But is this applicable to human autoimmune disorders and what is the best approach to use?

Altered peptide ligands: definitions and activities

By using analogue peptides with defined substitutions at individual residues, we can determine residues that interact either with the T-cell receptor (TCR) [analogues do not stimulate antigen-specific T cells but retain the ability to bind major histocompatibility complex (MHC)] or with the MHC (loss of both T-cell stimulation and MHC binding). The term ‘altered peptide ligand’ was first coined a decade ago by Evavold et al. to define peptides with alterations at TCR binding residues.¹ The definition of APL has since become broadened to include analogue peptides containing one or more substitution(s) at any residue. APL can be divided based on their ability to stimulate antigen-specific T cells (Fig. 1). The native peptide or APL that give equivalent response patterns are referred to as agonists. APL requiring increased doses to induce qualitatively normal responses are weak or subagonists, whilst those able to stimulate at greatly reduced doses are superagonists. The original work by Allen and colleagues identified APL with ‘partial agonist’ activity, being able to stimulate effector function (cytokine production, target cell lysis) in T-cell clones in the absence of the concomitant proliferation provoked by agonist peptides.^2,3 Soon after, Sette's group reported APL with alterations at TCR contact residues that did not overtly stimulate T-cell clones, but inhibited activation induced by agonist when presented on the same antigen-presenting cell (APC).^4,5 This inhibition was not simply due to MHC blockade and was antigen-specific (i.e. an APL of antigen x would not inhibit activation of T cells specific for antigen y). APL with these inhibitory properties are termed TCR antagonists. APL have proved useful tools in the dissection of early signalling events proximal to TCR ligation. Debate continues over the precise mechanisms underlying partial agonist and TCR antagonist effects and is documented elsewhere.^6,7

Figure 1.

Altered peptide ligands: subtypes and activities. APL can be divided according to their effects on antigen-specific T-cell populations in vitro.

APL and immunotherapy: Early experiments

Pioneering work in the field of peptide-induced modulation of autoimmunity used murine experimental autoimmune encephalomyelitis (EAE)⁸ which serves as a model for multiple sclerosis (MS). EAE is induced by immunization with complete Freund's adjuvant (CFA) mixed with antigens of central nervous system (CNS) myelin, notably myelin basic protein (MBP), proteolipid protein (PLP) and myelin oligodendrocyte glycoprotein (MOG). Central to the disease is activation of myelin-reactive CD4⁺ T helper type 1 (Th1) cells which infiltrate the CNS and establish an inflammatory lesion. As there are several well-defined T-cell epitopes in these myelin antigens^9–12, the disease can be induced using single peptides. For this reason it is the EAE model that has seen the most extensive investigations of peptide therapy. Three immunodominant regions have been of particular interest: PLP(139–151) and two regions within MBP, the acetylated N-terminal peptide (Ac1–9) and peptides contained within 80–105.

The first epitope to be identified,⁹ and consequently tested as a therapy,¹³ was MBP Ac1–9. The molecular basis for the interaction of Ac1–9 with the A^u MHC class II molecule and TCR has been characterized extensively^13–15 and is summarized in Fig. 2. The residues responsible for binding to A^u are 4Lys and 5Arg.¹³ However, the natural peptide binds extremely poorly,^9,16,17 due to 4Lys interacting unfavourably with a hydrophobic pocket within the A^u peptide binding cleft, as highlighted by mutational studies and predictive computational analysis.^18,19 Hence we can produce APL with greatly increased binding affinities for A^u by substitution of residue 4Lys, most notably with Ala, Val and Tyr^13,15,16 (Fig. 2). As we would predict, increased binding for class II translates into increased capacity to stimulate Ac1–9-specific T cells in vitro,¹³ such that femtomolar doses of the 4Tyr APL will stimulate responses that require nanomolar concentrations of wild-type Ac1–9.²⁰ Thus these position 4 APL behave as superagonists.

Figure 2.

How the Ac1–9 peptide interacts with TCR and MHC class II. The Ac1–9 peptide of MBP binds to the A^u class II molecule through interactions involving residues 4Lys and 5Arg. Residue 4 has a dominant effect on binding, with the wild-type Lys residue interacting unfavourably with the A^u peptide-binding cleft. APL with alterations at position 4 can greatly increase binding to A^u. These APL consequently act as superagonists in vitro, stimulating T cells raised against wild-type Ac1–9 at correspondingly low concentrations. Residues 3 and 6 interact with TCR. This scheme represents interactions with the Tg4 TCR for which residues 6 and 3 act as primary and secondary TCR contact residues, respectively. This preference in TCR contacts is reversed on the analysis of polyclonal Ac1–9-specific populations. This figure summarizes data from refs ¹³–¹⁵ and ²⁰.

The initial peptide therapy experiments of Wraith et al. focused on the 4Ala APL and uncovered a conundrum.¹³ Whilst the Ac1–9 4Ala peptide was a superagonist in vitro, it induced little or no EAE in vivo. Moreover, mice that were co-immunized with a mixture of 4Ala and wild-type 4Lys also showed little EAE. Preimmunization with 4Ala in incomplete Freund's adjuvant (IFA) also prevented EAE development on immunization with 4Lys in CFA.²¹ This was not due to the expansion of a regulatory population as T cells from 4Ala-primed mice could not transfer protection from EAE to syngeneic recipients. The basis for these paradoxical in vitro and in vivo effects has remained unclear for a decade but our recent data explain these findings (see below). Was this phenomenon peculiar to the Ac1–9 model? Later studies from van Eden's laboratory using EAE and adjuvant arthritis in rats suggested that it may be generally applicable.²² APL of MBP(72–85) and the arthritis-related peptide 180–188 of mycobacterial heat-shock protein 65 (hsp 65) were generated that showed increased binding affinities for the RT1B¹ rat class II molecule. In co-immunization experiments it was found that the MBP APL specifically inhibited EAE but not arthritis, indicating direct effects on antigen-specific T cells. These early experiments therefore pointed to applications for APL in antigen-specific therapy of autoimmune disorders.

TCR antagonism and autoimmunity

The idea of TCR antagonist peptides as therapeutics was first applied to the EAE model induced with the immunodominant PLP[139–151] epitope in SJL mice.²³ Residue 144Trp was identified as the dominant TCR contact for this epitope. Position 144 APL were identified that inhibited the in vitro activation of encephalitogenic 139–151-specific T-cell clones. When pools of these antagonist APL were added in equimolar amounts to the wild-type 139–151 prior to immunization in CFA, they were found to reduce significantly the incidence and severity of resulting EAE. Subsequently the approach was modified to design a single APL with substitutions at both 144 and 147.²⁴ This L144/R147 APL inhibited in vitro activation of a panel of encephalitogenic T-cell clones showing distinct TCR gene usage. The analogue prevented EAE when co-administered with native 139–151 and, furthermore, could limit progression of EAE if given early after the onset of disease.

Human autoreactive T cells can also be modulated by APL. T-cell reactivity to a mitochondrial 38000 MW islet antigen has been described early after onset of type 1 diabetes.²⁵ APL based on this antigen were found to act as TCR antagonists when presented with wild-type antigen.²⁶ T cells derived from myasthenia gravis (MG) patients were found to respond to two epitopes within the human acetylcholine receptor (AChR) α subunit. APL of these two peptides, or a hybrid APL combining both analogues in a single peptide, showed effective antagonist activity on MG T-cell responses to wild-type AchR peptides.²⁷ These findings were reproduced in an experimental model of MG in which APL prevented the development of clinical signs.^28–30 Several studies have also reported APL-induced modulation of T cells derived from MS patients and specific for region 80–100 of MBP.^31–33

There is a conceptual problem, however, with the use of TCR antagonist APL as therapeutic tools. Whilst antagonists are clearly capable of inducing some early signalling events, these do not appear to have long-lasting profound effects on T-cell reactivity (although partial agonist APL may anergize T-cell clones in vitro³⁴). Antagonist peptides inhibit when presented on the same APC as the agonist peptide (and usually when presented in molar excess). How then are we to achieve effective treatment with antagonists? They should have no effect on potentially autoreactive T cells in the absence of the native self antigen and therefore would not be a good option for prophylaxis. As treatments for active disease it is not inconceivable that antagonist and self antigen could be presented on the same APC. This would most likely happen in peripheral lymphoid organs however, and not in the affected organ where pathogenic self-reactive T cells would be active. Perhaps a more efficacious approach therefore is to design APL that convert autoreactive T cells to a benign or protective functional phenotype without the requirement for co-presentation of self antigen.

From TCR antagonism to immune deviation

Central to the aetiology of most experimental autoimmune models is the activation of CD4⁺ T cells of the Th1 functional phenotype producing pro-inflammatory cytokines such as interferon-γ (IFN-γ) and tumour necrosis factor-α (TNF-α).³⁵ The development of antigen-specific Th1 responses can be counteracted by signals that drive the immune response towards the Th2 pathway.³⁶ Whilst efforts to control an established autoreactive Th1 response by introduction of Th2 cells have failed,^37,38 this approach has proved successful if the Th2 population is present during the initiation of the Th1 response.^39,40 Further analysis of the effects of the PLP L144/R147 APL revealed more than just TCR antagonism. Priming of 139–151-reactive T cells was not prevented by co-immunization with L144/R147. T-cell clones specific for the APL cross-reacted with native 139–151 but secreted Th2 rather than Th1 cytokines. These Th2 clones were able to suppress EAE when transferred to syngeneic mice and to inhibit encephalitogenic activity of 139–151-specific T cells in vitro prior to adoptive transfer to naïve recipients.⁴¹ Moreover, preimmunization with L144/R147 suppressed EAE on subsequent immunization with different epitopes from PLP, MOG and MBP.⁴² This ‘bystander suppression’ is a powerful tool for treatment of autoimmunity as discussed below.

Why should alteration of the dominant TCR contact residue(s) of a peptide lead to preferential expansion of Th2 cells when using an immunization regime (with CFA) that normally induces Th1 expansion? Studies comparing clones derived from mice immunized with either wild-type 139–151 (Th1) or a Q144 APL (Th2) revealed that Th1 cells primarily recognized residues 143, 144 and 147, whereas Th2 clones recognized residues 141 and 142.⁴³ Therefore, immunization with APL changed at Th1-binding residues (such as L144/R147) would expand Th2 cells. There have been several reports that the use of APL influence the in vitro Th1/Th2 differentiation.^44–46 These presumably reflect different strengths of antigenic signal (different doses of agonist peptide have been reported to effect differentiation; low:Th2; high:Th1; very high:Th2).^47,48 The Q144 studies, however, reveal a unique and intriguing phenomenon: the differentiation of a T-cell population being determined by its fine specificity for antigen. Why this should be in this system remains a puzzle.

Therapeutic use of the ability to convert a pathogenic Th1 response to a benign Th2 response has also been reported in EAE induced with APL derived from the MBP(87–99) sequence in both mice and rats.^49,50 In these studies, exposure of MBP-specific T-cell lines to APL induced a shift in cytokine production with reduction in TNF-α accompanied by increases in interleukin-4 (IL-4). This region of MBP is of particular interest because it is also implicated in the pathogenesis of human MS, being the major MBP epitope recognized by T cells from DR2⁺ MS patients.⁵¹ Experiments using established human T-cell clones specific for this region identified APL that behaved as TCR antagonists of proliferation and the production of IL-2, IL-4, IL-10 and IFN-γ, but specifically induced de novo production of transforming growth factor-β₁ (TGF-β₁).³¹ These APL were therefore switching off production of both Th1 and Th2 cytokines in preference for TGF-β₁. Subsequent studies deriving APL-reactive T cells directly from MS peripheral blood revealed APL that inhibited IFN-γ production, allowing selective expansion of IL-4-producing cells.³²

Co-polymer 1 (also known as Cop1, Copaxone, or glatiramer acetate) is a random polymer of alanine, glutamate, lysine and tyrosine with protective effects in several EAE models.⁵² Cop1 is an approved drug for the treatment of MS, reducing the progression of disability and rate of relapse.⁵³ Investigations of the protective effects of Cop1 treatment have shown the expansion of myelin-specific Th2 populations in both EAE and MS, indicating immune deviation as a mode of action.^54–56 Cop1 has also been shown to protect against experimental uveoretinitis⁵⁷ and a similar co-polymer inhibited activation of DR4-restricted human T-cell clones specific for type II collagen, a putative autoantigen in rheumatoid arthritis.⁵⁸ Although their precise modes of action remain to be defined, random co-polymers may therefore provide an approach to more general therapeutic applications.

Taken together these studies provide evidence that: (a) APL can be used to modulate experimental autoimmune models; (b) this may be due to TCR antagonism, immune deviation, or both acting synergistically; and (c) similar phenomena of TCR antagonism and immune deviation can be elicited in vitro using human autoreactive T-cell populations.

Caution, read the small print: the complexities of the autoimmune repertoire

The findings outlined above would argue for the use of APL as TCR antagonists or immune deviators in human autoimmune disorders. Based on our own work, however, it is important to add a note of caution. We have assessed the potential for TCR antagonism using the MBP(Ac1–9) model of EAE.^15,59 The pathogenic Ac1–9-reactive T-cell repertoire is dominated cells by expressing the Vβ8.2 gene.⁶⁰ This model seemed very suitable therefore to test TCR antagonists whose effects may be limited to closely related TCRs. For these studies we used T cells from the Tg4 mouse that expresses a transgenic Ac1–9-specific TCR.⁶¹ As with previously tested Ac1–9-specific cells,^13,14 Tg4 cells recognized residues 6Pro and 3Gln of Ac1–9 as primary and secondary TCR contact residues, respectively. Thus no substitution was permitted at position 6, whereas limited changes (Pro→Met, His, Phe and Tyr) at position 3 could be made without loss of stimulatory activity.¹⁵ Having confirmed position 6 as the primary TCR contact residue, we identified three position 6 APL with potent TCR antagonist activity on Tg4 cells in vitro. We next tested these three antagonist APL for encephalitogenic activity in vivo (i.e. immunization with individual APL + CFA in the absence of wild-type Ac1–9). As predicted, Tg4 mice developed EAE after immunization with wild-type Ac1–9, but not after immunization with the three antagonist APL. Surprisingly, however, these APL were fully capable of inducing EAE in normal H-2^u mice.¹⁵

Thus we had the paradoxical findings that APL that inhibited encephalitogenic T cells in vivo actively induced disease in vivo. The reason for this became apparent when we analysed the fine-specificity of polyclonal Ac1–9-reactive T-cell populations derived from normal H-2^u mice. These cells recognized residues 3 and 6 as primary and secondary TCR contacts, respectively, i.e. the opposite of Tg4 T cells. Importantly, these polyclonal populations were activated by the three position 6 APL that induced EAE in non-transgenic mice. This diverse T-cell cross-reactivity against Ac1–9 APL pointed to considerable complexity in the Ac1–9-reactive T-cell repertoire, in contrast to previous assumptions of an almost monoclonal restriction in TCR usage. This complexity clearly represents a major hurdle to the development of EAE blocking antagonist APL. However, we subsequently identified a position 3 APL that inhibited activation of not only Tg4 cells, but also polyclonal T-cell lines.⁵⁹ This APL (with a substitution at the dominant TCR contact residue for the polyclonal response) was an effective inhibitor of EAE when co-immunized with wild-type Ac1–9. Another position 3 APL that only acted as an antagonist for Tg4 cells was a markedly less effective EAE blocker in vivo. These results confirm the complexity of the Ac1–9-reactive T-cell repertoire and show that inhibition of the entire population is required for effective prevention of disease.

Our findings in the Ac1–9 EAE model cast doubt on the effectiveness of antagonist APL as therapies in autoimmunity. To identify TCR antagonists we need to generate T-cell clones for in vitro analysis. This often results in the dominance of T cells robust enough to withstand the selective pressures of cloning, but that are not representative of the entire in vivo repertoire. The Ac1–9 system is analogous to this as the Tg4 TCR was derived from one of a panel of clones showing almost identical TCR usage.^60,61 Designing our antagonist APL on such clones can give a distorted impression of the most suitable peptide to use. Similarly, this may well be the case with APL that induce immune deviation. Thus an APL that converts a Th1 clone to produce IL-4, IL-10, or TGF-β in vitro may have the reverse effect on a distinct T-cell clone when given in vivo. This unpredictability led us to argue against the use of antagonist or immune deviating APL in human autoimmune disorders.¹⁵ Such an approach in an outbred human population might aggravate rather than reduce pathology.

Reports on two phase II clinical trials in MS patients include results suggesting that this may indeed be the case.^62,63 The effects of APL based on MBP(83–99) were assessed. In one study, although there was no general positive or negative effect of treatment, two of eight patients developed exacerbations associated with APL administration.⁶² The APL immunized for T cells producing Th1 cytokines, that cross-reacted with the wild-type MBP sequence and that could be isolated from the cerebrospinal fluid. The second study reported the induction of Th2 responses (in patients receiving the highest dose of APL), as assessed by in vitro challenge with either APL or wild-type MBP(83–99).⁶³ Whilst no overall improvement in clinical parameters was apparent, there was a suggestion of a reduction in development of CNS lesions in patients receiving the lowest dose of APL. This dose of APL did not, however, induce Th2 immunity in those individuals tested. This trial was suspended due to hypersensitivity reactions in a significant proportion of patients, again implying an (over)active Th2 response to APL.

APL-based therapy (at least for TCR antagonism and Th2 immune responses) is therefore complicated by: (a) the complexity of the T-cell repertoire; (b) the unpredictability of the effects of APL; and (c) harmful hyper-reactivity to the APL. It is clear that potential therapy will not be based on a single APL and may even require bespoke drugs for each patient after a rigorous investigation of immune reactivity to modifications of the chosen T-cell epitope.

Superagonists and apoptosis

If the use of antagonist or immune-deviating APL is problematic, is there a better approach? What do we want to do to the autoreactive T cells driving the pathology? Perhaps the safest outcome is to kill them. It is well established that CD4⁺ T cells, once activated are highly sensitive to deletion via activation-induced cell death. High doses of antigen can lead to apoptosis of antigen-specific T cells in vivo.^64,65 This ‘propriocidal’ cell death has been proposed as a major homeostatic mechanism that shapes the immune repertoire by removal of ‘overstimulated’ T cells.⁶⁶ If the apoptotic signal is driven by the strength of signalling through the TCR, can a high dose of wild-type antigen be replaced by using lower doses of superagonist APL. The MBP(Ac1–9) model allows us to address this due to the range of APL available with well-defined antigenic properties in vitro. We can assume that T cells primed against wild-type Ac1–9 must express high-affinity TCRs to compensate the peptide's vanishing affinity for A^u and allow formation of a productive MHC:peptide:TCR complex.

Immunization with Ac1–9 expands these highly sensitive T cells required to induce EAE. Paradoxically however, superagonists failed to prime effectively for EAE induction.^13,20 We analysed the in vivo fate of the highly sensitive T cells by transferring fluorescently labelled naïve Tg4 T cells into non-transgenic recipients prior to immunization with either wild-type Ac1–9 or the 4Tyr superagonist.²⁰ After 4Tyr immunization more than 70% of Tg4 cells showed signs of apoptosis compared with less than 10% in response to wild-type Ac1–9. Highly sensitive Ac1–9-reactive cells therefore undergo negative selection (through antigen-induced cell death) on in vivo exposure to superagonist ligands.

This provides an explanation for the early results that co-immunization of Ac1–9 with the 4Ala superagonist failed to induce EAE.¹³ Highly sensitive T cells would be deleted on 4Ala before they had a chance to initiate an inflammatory lesion in the CNS. Peripheral deletion also explains the failure to transfer protection using T cells from 4Ala-primed mice.²¹ Apoptosis in response to superagonist stimulation has also been shown in vitro using an independently derived Ac1–9-specific TCR transgenic mouse.⁶⁷ Similar in vivo findings have been reported using a superagonist APL in the MBP(87–99) EAE model.⁶⁸ Antigen-induced cell death has also been shown using relatively low doses of an oligomerized form of influenza virus haemagglutinin peptide.⁶⁹ The development of ‘killer APL’ may therefore provide a therapeutic avenue for human diseases. One drawback would be that superagonists generally induce a cytokine burst prior to apoptosis. Such a burst, although hopefully only transient, might lead to a significant clinical exacerbation. It is encouraging to note therefore that APL have been developed that are capable of inducing apoptosis in activated T cells specific for pigeon cytochrome c, but without the production of cytokines seen in association with apoptosis induced by native peptide.⁷⁰

An avidity model for clonal selection

Our studies of Ac1–9 superagonists revealed a further striking feature.²⁰ T-cell populations generated by immunization with Ac1–9 or any of the APL all showed identical sensitivity to the antigen they were originally primed against (with responses first apparent at ∼10 nanomolar). Staining antigen-reactive T cells with fluorescently labelled peptide:MHC class II tetrameric complexes provides a measure of TCR affinity for that complex.^71,72 Using tetrameric A^u:Ac1–11 complexes we determined that the T cells that avoid deletion after immunization with superagonist expressed distinct (not Vβ8), low-affinity TCRs compared to the pathogenic T cells expanded by wild-type Ac1–9. Thus T-cell lines generated against Ac1–9 were heterogeneous populations containing cells expressing low-, moderate- and (around 50%) high-affinity TCRs. T cells generated against the 4Tyr superagonist, in contrast, only used low-affinity TCRs. This low affinity explains why superagonist-primed T cells are relatively insensitive to superagonist, responding with the same kinetics to superagonist as Ac1–9-primed cells respond to Ac1–9.

These results led us to propose an avidity model for clonal selection during peripheral immune responses to antigen (summarized in Fig. 3).²⁰ This model argues that due to the high level of TCR cross-reactivity for antigen,⁷³ the immune repertoire will contain a heterogeneous population of T cells capable of antigen recognition with sensitivities ranging from low to high. Relatively insensitive T cells will not receive a strong enough stimulus and will remain ignorant of the presence of antigen. At the other extreme, T cells that are too highly sensitive (due to having high-affinity receptors) will undergo antigen-induced cell death. Only T cells displaying sensitivity for the priming antigen within a defined range will therefore be allowed to undergo successful clonal expansion and constitute the productive repertoire. This process would explain why we and others consistently generate antigen-specific T cells with this constricted sensitivity for the priming antigen. Such effects of avidity are of course not new. They form the basis for our perception of positive and negative selection during thymic development.⁷⁴ What our results tell us is that mature T cells are also under avidity-based control once in the periphery. We have described the constriction of T cells into the productive sensitivity window as ‘tuning’. We use this word to describe the effects on the entire antigen-reactive repertoire via TCR-based selection. This is distinct from tuning at the single cell level which presumably involves rewiring of components of the cell signalling machinery.^74–76

Figure 3.

An avidity model for clonal selection. The näve T-cell repertoire will contain a heterogeneous population of potentially antigen-reactive T-cell precursors (Tp) expressing diverse TCRs. These TCRs will have a range of affinity for the antigen. T cells with low-affinity TCRs (and therefore low sensitivity) for antigen will remain unaffected by exposure to antigen (via infection, immunization, or exposure of self antigen during inflammation/tissue damage). T cells bearing TCRs with high affinity will become activated but are speedily deleted from the repertoire by apoptosis. Only T cells expressing TCRs with ‘moderate’ affinity will be activated without signals for apotosis and so can expand and form the productive immune response. This serves to maintain T-cell sensitivity for the inducing antigen within a predetermined (and as yet undefined) ‘window’.

Are changes necessary? From APL to unaltered peptides and back again

Perhaps the best therapeutic approach is to induce autoantigen-specific T-cell tolerance using native antigen. It has been evident for some 40 years that the form in which antigen is administered has a dominant influence on the decision whether to mount a response or not.⁷⁷ Antigen in aggregated form, or mixed with adjuvant, will provoke an overt response, whereas soluble monomeric antigen induces a state of antigen-specific T-cell tolerance. We and others have given soluble peptides via the oral, intravenous, intraperitoneal, or intranasal routes.^78–89 This approach can effectively inhibit T-cell activation in response to antigen in adjuvant. Moreover, it has proved effective in inhibiting the onset of disease in EAE^{78–80,83,84} as well as experimentally induced models of arthritis,^85,86 uveoretinitis⁸⁷ and MG.³⁰ In addition, administration of antigenic peptides from insulin or glutamate decarboxylase (GAD) 65 can inhibit the spontaneous development of diabetes in the non-obese diabetic (NOD) mouse.^88,89 The exact mechanism underlying tolerance induction with soluble peptides remains an issue of debate and has been proposed to involve the induction of apoptosis, anergy, or a regulatory phenotype in antigen-specific T cells.⁹⁰ Tolerance may in fact involve all of these processes with different cells being killed, anergized or diverted to regulatory function during the tolerance process.

The paucity of knowledge of the T-cell epitopes that are critical to pathogenesis raises an obstacle to the use of peptide-induced tolerance in human autoimmune disorders. For certain diseases the autoantigens are relatively well defined, such as the AChR in MG⁹¹ and thyroglobulin and thyroperoxidase in autoimmune thyroid disease.⁹² For other diseases, candidate autoantigens have been proposed but we cannot definitively assess their role in disease or rule out the contribution of other as yet undefined autoantigens. Recent evidence using ‘humanized’ mice have revealed that the immunodominant MBP83–99 epitope, which is implicated in MS pathogenesis, can indeed elicit EAE.⁹³ Other epitopes have yet to be assessed in this way, however. A further complicating factor is epitope spreading,⁹⁴ a mechanism by which autoreactivity spreads to new epitopes due to tissue damage as disease progresses. This not only increases the number of epitopes recognized, but precludes identification of the epitope(s) involved in initiation of autoimmune attack once the clinical signs have developed, making prophylaxis difficult.

The question facing peptide-based tolerance, therefore, is how can we tolerize T cells against all known and unknown epitopes involved in disease? The answer may lie in the induction of bystander suppression. By this we mean the suppression of responses to multiple epitopes after administration of a single epitope. This phenomenon has been described using the Der p I house dust mite allergen as the model antigen. Intranasal administration of the immunodominant epitope of Der p I resulted in tolerance to this epitope and subdominant epitopes upon subsequent immunization with the intact protein.^81,82

We tested the ability of three encephalitogenic peptides to induce suppressive effects in EAE: Ac1–9 and 89–101 of MBP and PLP(139–151).⁸⁰ Intranasal administration of Ac1–9 suppressed responses to both Ac1–9 and 89–101 on subsequent immunization with intact myelin, but did not suppress responses to PLP(139–151). The PLP epitope proved an even more effective tolerogen, suppressing responses to itself and both MBP epitopes. So bystander suppression could be induced at the level of T-cell priming in the draining lymph node. Intranasal administration of a single peptide did not block T-cell priming but could suppress EAE after immunization with a different peptide. Thus PLP(139–151) suppressed EAE induced with itself or either of the two MBP epitopes, whereas Ac1–9 only suppressed disease induced with each of the MBP peptides but not the PLP peptide. Figure 4 supplies a model for bystander suppression that can function in either the lymph node draining the site of immunization with whole myelin or in the lymph nodes draining the CNS after immunization with peptide. The increased effectiveness of the PLP epitope over the MBP epitope may well reflect a comparatively higher frequency of PLP(139–151)-specific precursors in naïve mice due to a lack of thymic expression of this region of PLP.⁹⁵ Thus, a high frequency of tolerant PLP-reactive cells could suppress a low frequency of MBP-reactive cells. A low frequency of tolerant MBP-reactive cell may not however, suppress a high frequency of PLP-reactive cells.

Figure 4.

Mechanism for bystander suppression active either in the periphery or site of autoimmune attack. Induction of tolerance with soluble peptide leads to the expansion of an epitope-specific T regulatory (T reg) cell population. Subsequent immunization with intact antigen will lead to presentation of this epitope (red star) in addition to other pathogenic epitopes to which tolerance has not been induced (green triangle) in the draining lymph nodes. This leads to recruitment of pathogenic T cells capable of inducing disease. However, T reg cells are also recruited and suppress activation of the pathogenic T cells either directly, or indirectly via regulation of APC function. Immunization with only the epitope against which the immune sytem has not been tolerized will fail to recruit T reg cells, therefore allowing the pathogenic T cells to be fully activated. These cells then migrate to the target organ (in this case the CNS) and initiate an inflammatory lesion. Subsequent tissue damage releases the epitope recognized by the T reg cells which then are recruited to the target organ and/or lymph nodes draining the target organ where they can now exert their suppressive effects.

Bystander suppression has also been reported in other studies of administration of soluble myelin-derived peptides in EAE.^83,96 These studies, coupled with those in the Der p I model^81,82 and the L144/R147 APL of PLP described earlier,⁴² present a persuasive argument that the induction of active suppression using soluble peptides will be the most effective therapeutic approach.

The precise nature of the suppressive effects is an area of controversy. In the NOD mouse soluble peptides have been reported to drive the immune response towards a Th2 phenotype.^88,89 T cells expanded as a result of oral tolerance have been reported preferentially to secrete TGF-β (the so-called ‘Th3’ phenotype).³⁹ We were unable to detect a shift towards Th2 or Th3 immunity in normal mice.⁸⁰ Studies in Tg4 transgenic mice, however, demonstrated down-regulation of IL-2, IL-4 and IFN-γ production, but up-regulation of IL-10 production following peptide treatment.⁹⁷ Furthermore, the administration of a neutralizing anti-IL-10 antibody restored susceptibility to EAE in peptide-treated Tg4 mice. The immunosuppressive properties of IL-10 have been implicated in control of pathology in EAE.^36,98 T cells that preferentially express IL-10 (T reg or Tr1 cells) have also been shown to have potent suppressive effects in models of inflammatory bowel disease.⁹⁹

A note of caution to the systemic application of peptides comes from a recent description of anaphylaxis in EAE models.¹⁰⁰ EAE was induced with either PLP(139–151) or MOG(35–55). Anaphylaxis and death resulted when the same peptides were given in solution intraperitoneally after disease resolution (but not when given earlier during active disease). This may reflect a Th2 response associated with the later remission stages of EAE. Indeed, in our studies repeated administration of the PLP peptide after onset of EAE induced with whole myelin led to a significant reduction in incidence and severity of relapses without signs of anaphylaxis.⁸⁰ This relapsing form of disease induced with myelin may not involve a switch to Th2. Immunization with peptide may well induce a significantly larger antigen-reactive T-cell population, many of which may play no role in pathogenesis due to recognition of the peptide in a form not generated by processing of native antigen. We have evidence that immunization with MOG(35–55) expands a heterogeneous population of cells of which only a fraction respond to intact MOG and it is only these MOG-reactive cells that transfer EAE (S.M.A. unpublished observations). Anaphylaxis in the peptide-induced models may thus be the result of reactivation of cells not directly involved in disease. This may not therefore be directly relevant to the human situation in which autoreactive T cells will be activated by presentation of processed self antigen rather than by immunization with a high dose of synthetic peptide.

It is worth noting that there are situations in which we might be able to improve the efficacy of soluble peptide therapy by making changes to the antigen sequence. Hence APL may after all be an attractive approach. Firstly, improving the affinity of a peptide for MHC creates a stronger tolerogen.^78,79,84 Wild-type MBP(Ac1–9) is relatively poor at inducing tolerance compared to the position 4 superagonists with increased affinity for class II. Our experiments showing bystander suppression therefore used the 4Tyr APL to induce tolerance.⁸⁰

In our studies of the suppressive effects of MBP and PLP peptides we observed that the MBP(89–101) peptide failed to prevent EAE induced even with itself.⁸⁰ Thorough analysis reveals complex T-cell responses to this peptide (S.M.A. manuscript in preparation). There appears to be three overlapping epitopes within this short 13mer peptide recognized by distinct T-cell populations. Two of these epitopes are generated from antigen processing of native MBP whereas the third epitope is ‘cryptic’, i.e. it is not generated form intact MBP.¹⁰¹ Peptides containing the two naturally processed epitopes are capable of inducing EAE, whereas the cryptic epitope is not. This presumably reflects a requirement for autoreactive T cells to recognize epitopes processed from native MBP in the CNS. T-cell lines derived from MBP(89–101)-immunized mice show preferential recognition of the cryptic epitope, suggesting that this is the dominant configuration assumed by the class II:89–101 complex. Thus the encephalitogenic T-cell population constitutes only a minor fraction of the total 89–101-reactive T-cell pool. Furthermore soluble application of 89–101 appears to tolerize T cells recognizing the dominant cryptic epitope, but not those recognizing the minor disease-relevant epitopes. This would account for the failure to prevent EAE when tolerizing with this peptide. One potential way to make this peptide tolerize to EAE would be to generate APL with appropriate substitutions that ablate the formation of the cryptic peptide:class II conformation. The disease-relevant epitopes should then become dominant and capable of delivering the tolerance signal to the pathogenic T cells. Such an approach using APL would minimize the risk of anaphylaxis due to expansion of high numbers of T cells responsive to cryptic epitopes as described above. For instance, in the MOG(35–55) system we are currently exploring APL that preferentially stimulate those T cells that are responsive to intact MOG and induce EAE. These APL would be ignored by 35–55-reactive cells that are not relevant to disease. Indeed it may well be that tolerizing these cells would be undesirable as they might perform useful functions due to cross-reactive recognition of foreign antigens not related to autoimmunity.

Another potential area for the use of APL involves T-cell recognition of antigens that have been post-translationally modified. Such antigens may be present in the target organ but not in the thymus, allowing the escape from central tolerance of T cells that specifically recognize the post-translationally modified form of the antigen. Thus collagen-induced arthritis has been demonstrated to involve T-cell recognition in glycosylated epitopes.¹⁰² More recently, dominant T-cell responses to α-gliadin in coeliac disease have been shown to require the conversion of glutamine residues to glutamate via the action of tissue transglutaminase.¹⁰³ Such enzyme-mediated alterations in antigen sequence may have important roles in the activation of autoaggressive T cells. The use of APL may therefore improve not only the identification of autoantigenic epitopes, but also our attempts to control the activities of self-reactive cells.

Concluding remarks

The use of APL over the past decade has formed a central plank in our efforts to probe the molecular requirements for T-cell activation. Identification of APL that act either as TCR antagonists or immune deviators prompted their successful use in animal models of autoimmunity. These experiments, however, have also highlighted the considerable complexity and diversity of the self-reactive T-cell repertoires in what first appeared to be simple systems. These complexities will be greater in the outbred human population and require extreme caution when designing APL-based therapies as highlighted by recent clinical trials.

A more favourable approach therefore is the use of natural peptide sequences to tolerize the immune system, preferably by expansion of self-reactive T regulatory cells capable of bystander suppression. Even this approach may need to be improved, however, providing new avenues for the application of APL to allow: (a) the focusing tolerance on those cells relevant to disease; (b) the avoidance of hypersensitivity reactions to cryptic epitopes within the drug; and (c) investigation of the role of modified self antigens in autoimmune pathogenesis.

Abbreviations

APL

altered peptide ligand

CFA

complete Freund's adjuvant

Cop1

co-polymer 1

EAE

experimental autoimmune encephalomyelitis

IFA

incomplete Freund's adjuvant

MBP

myelin basic protein

MG

myasthenia gravis

MOG

myelin oligodendrocyte glycoprotein

MS

multiple sclerosis

PLP

proteolipid protein

TCR

T-cell receptor