Antigen-specific immunotherapy for autoimmune disease: fighting fire with fire?

Introduction

The defining characteristic of an autoimmune disease is the existence of T- and B-cell autoreactivity directed against self proteins (autoantigens). The early events that precipitate autoreactivity in human autoimmune disease remain obscure. Once established, the autoreactive process itself may or may not be directly relevant to the pathological events involved in the disease. In patients with established rheumatoid arthritis, for example, on-going joint damage is most likely the result of a self-perpetuating inflammatory process that has few features of tissue-specific autoreactivity. In type 1 diabetes mellitus (DM), on the other hand, autoreactivity directed against the target cell in the pancreas reaches its peak of intensity around the time of clinical diagnosis and typically wanes only after the target cell has been destroyed. These are examples at opposite ends of a continuum but most theoretical models used to explain autoimmune disease invoke a period of autoreactivity against a specified protein autoantigen(s) as a critical phase in disease development. Work in animal models indicates that this phase may be susceptible to regulation by administration of the inciting autoantigen.

In this review several key questions will be addressed in relation to antigen-specific immunotherapy (ASI) for autoimmune disease. First, does it work in animal models, and by what mechanism(s) of action? How should we best harness these effects for human disease? How will we know when it has worked? And finally, will it work in established disease or will prophylaxis be required? In terms of autoantigen administration, we will focus mainly upon systemic rather than mucosal routes, which have been discussed extensively in other recent reviews.^1,2 A particular emphasis will be placed upon the issues to be faced in transposing ASI as a therapeutic option from animal into clinical studies.

Developing the principle of antigen-specific immunotherapy for autoimmune disease

One of the major goals for therapies designed for autoimmune disease is to restrict the immunosuppressive elements of the treatment so that they act upon the relevant autoimmune response without engendering a state of generalized immunosuppression. It was noted several decades ago that prior systemic administration of a protein antigen could inhibit the subsequent generation of an immune response to the same antigen. As antigen-induced models of autoimmune disease were devised over the subsequent years, the effect of pre-treatment with the same antigen was explored. Seemingly without exception, this strategy has resulted in protection from disease in a large range of animal models tested.^3–7

However, there is an important point in relation to these studies that should not be overlooked. These were not spontaneous models of human autoimmune disease, but models induced by administration of antigen under immunizing conditions. In this sense, the outcome was perhaps entirely to be expected on the basis of the original studies on antigen-specific responses and the effect of pre-treatment with the same antigen. It has been an important step forward, therefore, that similar findings have been made in spontaneous animal models of human autoimmune disease. These are much less common, the best characterized spontaneous animal model of autoimmune disease being the non-obese diabetic (NOD) mouse, which typically develops islet infiltration and diabetes due to complete destruction of the insulin-producing β cells between the ages of 12 and 25 weeks.⁸ Several putative β-cell autoantigens have been identified and evidence exists that CD4 T-cell tolerance to some of these is lost as an early feature of the disease.^9,10 Administration of the same antigens as whole proteins^9,11 and peptides^12–14 by a variety of routes^11–13,15 markedly prevents the development of diabetes.

In summary, then, an experimental principle was established during the infancy of modern immunological studies, namely that pre-administration of antigen in a tolerogenic form prevented subsequent attempts to induce active immune responses to the same protein. The principle adapted perfectly to the advent of animal models of human autoimmune disease that were induced by active induction of an autoimmune response, and it has now been shown to be sufficiently robust as to cure spontaneous rodent autoimmune disease.

Mechanisms of therapeutic effect

Several mechanisms have been invoked to explain the beneficial effects of administration of autoantigens on autoimmune diseases.

Induction of clonal anergy and deletion

Anergy is defined as antigen-specific non-responsiveness. At the CD4 T-cell level, it describes the inability of T cells to respond [e.g. by proliferation or interleukin-2 (IL-2) production] to stimulation with an appropriate whole antigen or peptide. T-cell anergy was first described as an in vitro phenomenon in which high doses of specific peptide epitopes rendered T-cell clones unresponsive to the same peptide in subsequent challenges.^16,17 Anergy induced in vivo after antigen administration is more difficult to demonstrate than in vitro effects on clones, since the detection of ex vivo unresponsiveness by proliferation or cytokine production in a polyclonal T-cell population could be attributable to other effects, such as regulation (see below). Convincing evidence of anergy induced by antigen (peptide or whole) has been provided in mice transgenic for a T-cell receptor (TCR) specific for a peptide of cytochrome C.¹⁸ After intravenous administration of peptide there was no reduction in the number of clonotypic T cells but there was marked T-cell hyporesponsiveness (both proliferation and cytokine production). Others have made similar findings.¹⁹ What other considerations are there in relation to anergy induced by antigen administration? First, it is clear that in terms of peptides, the same peptide that induces an active immune response can induce anergy. As a second, related, point it appears that the route of administration is important, with intravenous favoured by several groups.^18,19 Third, the administration of major histocompatibility complex (MHC) class I peptides can achieve a similar effect, i.e. anergy induction in CD8 T cells. Finally, for anergic hyporesponsiveness to be maintained, the continued presence of the antigen is required.^20,21 In the absence of antigen, hyporesponsiveness may be relatively short-lived (a matter of several weeks), but can be maintained by chronic administration of antigen (e.g. by weekly injections), which acts to sustain peripheral tolerance very effectively.²²

It is probably artificial to attempt to delineate cleanly between particular mechanisms of antigen-specific tolerance induction. For example, in several of the studies cited above,^19,22–24 the development of hyporesponsiveness to stimulation could be explained by both the induction of anergy and the disappearance of responsive cells. Just as for anergy induction, deletion of cells is most readily demonstrable in transgenic TCR models, and is evident in both thymic and peripheral T-cell populations. Disappearance of CD4⁺ T cells from the thymus and peripheral lymph node organs following peptide administration is associated with evidence of apoptosis,²⁴ although under other conditions peripheral deletion may depend upon an intact death pathway mediated by the Fas/Fas ligand (FasL) and tumour necrosis factor (TNF) receptor pathways.^25,26

Induction of regulatory cells and immune deviation

There is considerable evidence of natural regulation of autoimmune processes. This includes studies demonstrating that lymphopaenic animals develop autoimmune disease;^27,28 that autoreactive T-cell responses often wax and wane, and induced disease often resolves spontaneously; and that adoptive transfer of pathogenic T cells into naive recipients often requires irradiation to remove endogenous disease-preventing cellular mechanisms (i.e. regulatory cells). In man there is a body of literature indicating that the development of cellular and humoral organ-specific autoimmunity need not inevitably lead to frank disease.²⁹ Thus it is entirely consistent that in addition to the direct effects on pathogenic T cells described above, administration of autoantigens could also generate antigen-specific T cells that down-regulate the autoaggressive response.

There are at least two models by which such an effect can be explained. First, the induction of ‘regulatory T cells’, which exert their effects through the secretion of immunomodulatory cytokines, such as IL-10 and transforming growth factor-β (TGF-β).^30–33 The exact nature of these regulatory cells, sometimes referred to as T helper type 3 (Th3) cells³² remains to be defined but they have the potential to home to the target tissue of autoimmune disease driven by their specificity for autoantigen available at this site, and once stimulated, to inhibit the entire autoimmune response by the local release of cytokines (‘bystander suppression’).^30,34 The relationship between these induced regulatory cells, the regulatory populations currently being defined by expression of CD4, CD25 and CTLA-4³² and occasionally referred to as T regulatory (Tr) cells, and the T cells rendered anergic by interaction with peptide as described above, which share features with CD4⁺ CD25⁺ T cells, will probably take some years to unravel. As discussed, the best evidence that regulatory cells of potential benefit in the context of autoimmune disease can be induced is in relation to Th3-like cells, and most of that evidence relates to antigen administration by the oral or intranasal routes.² It is not yet clear whether, in the context of autoimmunity, Tr cells are inducible by exogenous autoantigen administration, although relevant studies are now beginning to appear. Seddon and Mason have shown in a model of autoimmune thyroiditis generated by induction of lymphopenia that the presence of thyroid autoantigens in the periphery is a prerequisite for induction of Tr-like regulatory cells,³⁵ which perhaps suggests that they may be inducible by ASI. In addition, a recent study by Thorstenson et al.³⁶ indicates that oral antigen ovalbumin (OVA), and to a lesser extent intravenous OVA peptide administration are able to induce CD4⁺ CD25⁺ cells with immunoregulatory properties. Woods et al. have demonstrated that a peptide from the gastric H⁺/K⁺ ATPase β-subunit delivered via a cutaneous route suppresses the development of autoimmune gastritis induced by neonatal thymectomy,³⁷ although the nature of the induced regulating cells is not known. Little information is available as to constraints upon the nature of the peptide epitope itself in terms of inducing regulatory cells. Jordan et al.³⁸ suggest that endogenous intrathymic peptides with a high affinity for TCR are required for induction of regulatory CD4⁺ CD25⁺ cells. However, the overall avidity of the TCR–ligand interaction is likely to be the determining factor, and this could be enhanced by provision of exogenous peptide.

Factors related to peptide administration that promote the generation of Th3-like regulatory T cells include the absence of adjuvant, the use of highly soluble peptides and administration by a tolerogenic route (e.g. oral or intranasal).³⁹ This combination of factors suggests that the nature of the antigen-presenting cells (APCs) involved in T-cell activation may be critical: the avoidance of inflammatory stimuli that might activate APCs or the targeting of tolerance-promoting APCs at mucosal sites both favour the generation of regulatory cells.⁴⁰

The second model invoked to account for the regulatory effects of antigen administration on autoimmune responses is that of ‘immune deviation’, discussed by Larché⁶⁰ in this issue. It has become apparent that administration of autoantigens, either as soluble proteins or peptides, can achieve immune deviation in association with arrested development or protection from autoimmune disease. As a generalization, the majority of animal disease models investigated in this way are thought to be Th1-mediated, and deviation towards Th2 is associated with disease abrogation. Examples of the successful application of this approach include the demonstration that administration of the diabetes-related islet autoantigen, glutamic acid decarboxylase (GAD-65) in incomplete Freund's adjuvant induces IL-4-secreting autoreactive T cells in association with disease protection¹⁵ and that GAD-65 peptide-induced prevention of diabetes in the NOD mouse is IL-4-dependent.¹⁴

Important questions in relation to transferring ASI into man remain. For example, from the putative mechanisms discussed above, is it possible to ‘cherry-pick’ an ideal mechanism of controlling disease? If this is the case, can a therapeutic strategy be targeted to maximize exploitation of that particular mechanism? As Anderton⁵⁹ argues persuasively in this issue, targeting cells for deletion or anergy may prove difficult, as may the use of APLs to attempt immune deviation of polyclonal populations of autoreactive T cells. Furthermore, it will be difficult in humans with multiple class II human leucocyte antigen (HLA) alleles to be certain as to the nature and efficacy of all autoantigens and their epitopes. The most attractive strategy therefore may be the targeted induction of regulatory cell populations, which will be favoured by using relatively low doses of highly soluble peptides/antigens. The opportunity these cells offer to foster bystander effects makes them an attractive goal for ASI.

Whole antigen, peptide epitope, or altered peptide ligand?

In animal models, similar results have been achieved whether administering a short peptide corresponding to a T-cell epitope on a given autoantigen, or the whole antigen itself, and in these studies it is perhaps difficult to define one approach as better than the other. When considering human disease, there is at least one good argument for taking a ‘whole autoantigen’ approach. For example, in most human autoimmune diseases, whereas the autoantigens are identified, the disease-initiating and disease-perpetuating epitopes on these are frequently not. Administering the whole antigen will allow all potential epitopes to be realized through natural processing and presentation. On the other hand, there may be several theoretical advantages to using peptide epitopes, if these are known, rather than whole antigen. Peptides can be produced in large quantities and in highly purified forms much more readily than whole recombinant antigens. Peptide sequences can be selected to ensure solubility and to target specific disease-associated HLA molecules. Using peptides avoids the risk of dose-limiting biological side-effects since peptides do not retain the bioactivity of the intact molecule. In the case of an autoantigen such as insulin, for example, a 0·5-mg dose is 10–15 IU, sufficient to induce hypoglycaemia. Equally, peptide epitopes offer greater flexibility than whole antigen: where immunogenic epitopes and tolerogenic sequences are known,⁴¹ these could be selected and avoided as required. Weight for weight, short synthetic peptides will deliver 20–50 times more epitope than whole autoantigens. An additional consideration relates to bioavailability, i.e. how effectively the antigen or peptide can be delivered to the sites of greatest potential effectiveness. Soluble peptides with sequences based on known epitopes have the advantage over whole autoantigen that they are able to bind directly to HLA class II molecules. This may theoretically enhance presentation at sites of high HLA class II expression, such as the local lymph nodes and target tissue. Whole antigens, on the other hand, will have a generalized tissue distribution and require processing by APC before becoming bioavailable.

Using peptide sequences in immunotherapy will, of course, depend upon the reliable identification of disease-relevant epitopes. Most studies employ sets of overlapping peptides based on the primary sequence of an autoantigen, and either a proliferation or ELISPOT format for response detection in bulk peripheral blood T cells. The technology for cytokine ELISPOT assays is constantly being refined, and it offers functional information on cytokine secretion that may be relevant to epitope selection. A novel approach to epitope identification has recently been described using HLA class II tetramers.⁴² It will be important that epitopes identified in this way are subsequently shown to be naturally processed and presented, since there is some evidence that cryptic epitopes induce or exacerbate disease.⁴³ In addition, peptides which require further processing for the generation of epitopes will not have the advantages alluded to above. Biochemical analysis of peptides eluted from HLA proteins on APCs pulsed with whole antigen represents a novel and direct method for identifying HLA allele-specific naturally processed and presented T-cell epitopes.⁴⁴

There are many persuasive arguments for the use of altered peptide ligands as therapy for autoimmune disease (see review by Anderton,⁵⁹ this issue). However, APL technology for human autoimmune disease is currently limited by several constraints. First, it requires the accurate definition of critical disease-related epitopes and their T-cell contact residues. Second, it requires experimental evidence of efficacy in antagonizing appropriate T cells. Both of those features are difficult to circumvent in the absence of extensive panels of disease-relevant T-cell clones, which are not always easy to generate, particularly from the peripheral blood in disorders in which the target organ is inaccessible to study.

An alternative approach to peptide therapy has been explored in multiple sclerosis. Glatiramer acetate (GA) is a random co-polymer of the amino acids alanine, lysine, glutamic acid and tyrosine which has been licensed for the treatment of MS in the USA since 1996.⁴⁵ GA does not constitute an autoantigen or an APL, but was originally conceived as a mimic of the biophysical properties of the autoantigen myelin basic protein (MBP). Daily subcutanous injections of 20 mg of co-polymer 1 (COP-1) are remarkably well tolerated, and clinical benefit, in terms of reduced relapse rate, is sustained for up to 6 years. At least four major mechanisms of action of COP-1 have been identified and include competition with MBP peptides for binding to MHC class II and/or T-cell receptor, APL-like activity and induction of Th2-like regulatory cells.⁴⁶

The timing of therapy; will it work for established disease and how will we know?

A further consideration in the development of antigen-specific or peptide immunotherapy in man relates to the timing of treatment in relation to the natural history of the disease. Peptide immunotherapy in rodents has generally been administered at an early stage of disease (e.g. 4–6 weeks of age in NOD mice, when islet inflammation is barely detectable) or before priming in models in which disease is induced by immunization. At these stages it has been relatively easy to achieve major effects on disease with a single peptide. In human autoimmune diseases with an insidious, asymptomatic onset, a form of therapy is required which is effective when organ infiltration and damage are already established, and the autoimmune response has spread to include multiple epitopes on multiple antigens.⁴⁷ To achieve a detectable immunological benefit under these circumstances, peptide immunotherapy will probably need to be administered by the optimal protocol and target multiple epitopes of more than one antigen. Experimental evidence supports this, with recent studies showing that two peptides are better than one and can be effective in halting advanced disease.^14,30

One of the important issues to be faced in clinical trials in man is the development of surrogate markers of therapeutic efficacy. It is entirely conceivable that ASI will require repeated administrations to achieve the best effects, and to use the onset or worsening of clinical disease as an outcome measure will slow progress.

Taking antigen-specific immunotherapy into the clinic: safety first

If antigen or peptide immunotherapy is to be used in man, it must be safe. In addition to the importance of recent reports of allergic hypersensitivity after peptide (APL) therapy in man and in mice (discussed by Anderton⁵⁹ in this issue), there are other issues in relation to safety, especially the potential for precipitating, exacerbating or inducing de novo autoimmune disease.

Exacerbation of disease: generating tolerance rather than immunity

ASI may be applied to patients with established autoimmune disease, or to those at high risk or with subclinical disease where predictive genetic or immunological tests exist. Therapy in these latter groups carries the risk that administration of a disease-relevant autoantigen will exacerbate established disease, or precipitate nascent disease. This theoretical possibility is perhaps unlikely if soluble antigens are given without adjuvant. For example, in an extensive analysis of the literature in relation to the NOD mouse model of autoimmune diabetes (in which, for example, over 45 published studies since 1996 have involved injection or ingestion of whole or peptide autoantigen, either as simple solutions or in conjunction with powerful adjuvants) has revealed that it is extremely unusual to accelerate disease; in most cases the manoeuvres are protective or have no effect. Only in one reported circumstance did such a manoeuvre accelerate disease. In this case, two peptides of GAD-65 (denoted p34 and p35), administered intrathymically, caused a mild acceleration of diabetes onset in NOD mice.⁴⁸ The authors point out that immunization with whole GAD-65 does not induce CD4 T cells reactive with either p34 or p35, implying that they are cryptic epitopes not naturally processed and presented by NOD APCs. The authors speculate that p34 and p35 administration may have activated low abundance mature T cells emigrating from the thymus, enabling additional anti-islet T cells to be recruited and thus accelerating the disease. The route of administration may also be important in this outcome – it should be noted that we have injected a single dose of p34 intraperitoneally and found a strong protective effect.¹³ The use of naturally processed and presented epitopes may be a safeguard against the danger of priming additional T cells. Other examples of disease exacerbation by autoantigen administration in models of autoimmune demyelination⁴⁹ and autoimmune diabetes⁵⁰ have also been reported, and it remains to be seen whether these exceptions to the general rule are just that.

An additional consideration relates to CD8 T-cell epitopes contained within CD4 T-cell epitopes that are being used as tolerogens. There is the theoretical possibility that the simultaneous provision of a helper and cytotoxic T lymphocyte (CTL) epitope could induce a damaging cytotoxic response. The proinsulin molecule contains such a peptide region, and only when the peptide is modified to disable the CTL epitope (by removal of the terminal amino acid) does it tolerise and protect NOD mice (Len Harrison, personal communication). As described above, CD8 T cells are susceptible to modulation by peptide epitope administration. Future studies will be required to address whether targeting autoreactive CD8 T cells is as effective as for CD4, and whether the simultaneous provision of CD4 and CD8 epitopes on the same or separate peptides is a good or bad thing.

Successful peptide immunotherapy may require activation of the pathogenic T cells before they are rendered tolerant.³⁰ This appears to be the case for the therapeutic effects of allergen peptides (see review by Larché,⁶⁰ this issue). Transient activation of pathogenic T cells in autoimmune lesions may be an acceptable side-effect if the therapy is successful. In addition, since autoimmune diseases are often characterised by flitting, focal pathology,⁵¹ such activation is unlikely to have global effects on the target organ, or damage areas that are not already infiltrated.

Precipitation of new autoimmune diseases

A theoretic complication of antigen-specific immunotherapy hinges upon the fact that some autoimmune diseases occur together and share autoantigens. An example is the relationship between type 1 DM and the neurological disease Stiff–Man syndrome (SMS), which are both associated with an antibody and T-cell response to the autoantigen GAD-65.^52–54 Approximately two thirds of SMS cases also have type 1 DM. A major safety consideration in using GAD-65 in ASI for diabetes therefore is the theoretical possibility of priming against GAD-65 epitopes capable of being presented within the CNS. GAD-65 autoantibodies have also been described recently in nondiabetic individuals with idiopathic temporal lobe epilepsy.⁵⁵

Hypersensitivity

A second safety consideration is the induction of hypersensitivity. This may manifest itself as irritation at the site of injection, rash, hives, wheeziness, chest tightness, or syncope. It is likely to result from production of antipeptide antibodies. As discussed in the review by Larché⁶⁰ in this issue, hypersensitivity has been reported in some patients receiving repeated doses of altered peptide ligand (APL) therapy for multiple sclerosis (MS).⁵⁶ The side-effects were seen with repeated injections at the higher doses (> 250 µg) and were accompanied by development of immunoglobulin G antibodies against the APL. It should be borne in mind that in this study, the peptides were extensively modified from the native sequence, and may therefore have a greater potential for induction of an immune response. For example, it is worth noting that the natural peptide sequence (MBP₈₅₋₉₆) administered intravenously or subcutaneously in doses as high as 500 mg was not associated with hypersensitivity symptoms.⁵⁷

In the light of these results, it may be advisable to adopt a more restricted approach to peptide therapy, particularly when APLs are being used. Doses of 250 µg or less may be best, with peptide sequences as short as possible to reduce the chance of antibody binding. Peptides should be soluble to avoid Th2 priming, and the intradermal rather than the subcutaneous route should be used which showed less hypersensitivity in the study of Haselden et al.⁵⁸ All peptide injections should be performed in the presence of practitioners trained in and equipped for resuscitation from anaphylaxis. Anaphylaxis should not necessarily be a contra-indication to the therapy continuing, as it should be borne in mind that it is a recognized and well-managed complication of allergen immunotherapy. Moreover, it is conceivable that this evident induction of peptide-reactive Th2 cells is a key part of the therapeutic effect.

In summary, there are theoretical risks associated with ASI. The pursuit of tolerogenic protocols should avoid the problems of exacerbation of existing disease and induction of new disease. Hypersensitivity may be avoidable with short, soluble peptides administered at low doses.

Conclusions

It is likely that ASI trials in autoimmune disease will be a major feature of clinical immunology research in the next decade. These trials, in conjunction with further basic immunology research in the fields of epitope discovery, mechanisms of immune tolerance and regulatory cell populations, will create an environment in which the safe and effective employment of antigens to treat autoimmune disease is developed.