Five key questions arise in the debate: what is the evidence that vaccination can restore immune tolerance in human disease; is there any reason it will not work in Type 1 diabetes; what antigens could be of use in a diabetes-centric approach; what are the current experiences; and how can they be built upon and optimized?
Currently, the lead setting for using antigens to impact upon an inflammatory disease in man is clinical allergy. This was the setting for the very first recorded description of the approach, in 1911 [26
], and it has now entered mainstream clinical practice (as “allergic desensitization” or “allergen-specific immunotherapy”). The antigen(s) (allergen extract) are injected as simple solutions into the subcutaneous tissue, typically in the clinical setting of prevention of intractable seasonal pollen allergy or anaphylactic responses to wasp and bee stings. Treatment is typically undertaken for a 1-2 year period, during which the majority of patients show sustained clinical benefit. Numerous mechanisms of action have been described, including the generation of Tregs and the production of IL-10 [27
]. New developments include the administration of allergens, via the sub-lingual route, and the use of short peptide sequences that represent the key regions (termed “epitopes”) within an allergen molecule that are presented to, and recognized by, T-lymphocytes [30
]. The latter (“peptide immunotherapy”) has the advantage that only the relevant parts of the allergen need be used in the vaccine, avoiding the risk of antibody-allergen complex formation that can precipitate anaphylactoid responses. Peptide selection can also be tailored to a particular immune response, thus introducing a step towards personalized medicine. This could be in the form of selection of patients by human leukocyte antigen (HLA
) genotype (the HLA
genes encode HLA class II molecules that present peptide epitopes to T-lymphocytes), or in the avoidance of CTL epitopes to limit any risk of disease exacerbation [32
]. Thus, the clinical allergy experience is that this simple vaccination-like manipulation can bring significant symptom relief, reduced inflammation and reduced dependence on drug therapy to many thousands of patients worldwide.
As has been noted already, there is a considerable body of evidence that Treg function is defective in patients with Type 1 diabetes, and several of the underlying diabetes-associated single nucleotide gene polymorphisms identified in genome-wide studies are considered likely to impact upon immune regulatory pathways [33
]. This highlights the question as to whether, if the immunological brakes have a design fault, there is any prospect that just pumping the brake pedal harder will achieve any notable effect. In other words, can functional antigen-specific Tregs be induced in patients with Type 1 diabetes? To counter this argument, there are several indications that other aspects of immune regulation may be amenable to a vaccine approach. We have reported for example, that there are non-classical islet autoantigen-specific CD4+ Tregs, characterized by IL-10 production, that are present and functional in many adult patients with Type 1 diabetes, their presence being associated with a later age of onset of disease [34
]. These cells potently regulate pro-inflammatory T cell responses [35
]. Unpublished work in our laboratory shows that cells of this phenotype are also a frequent finding in autoantibody-negative (i.e. low risk) unaffected siblings of Type 1 diabetes patients, hinting at a role in active regulation of islet autoimmunity. Our interpretation of these data is that this pathway of immune regulation can be functional on a genetic background predisposing to Type 1 diabetes, but that its development may require environmental cues that may be stochastically missing or insufficient in those developing the disease. Intriguingly, it is widely thought that one of the pathways of therapeutic effect of the vaccine approach is induction of IL-10 secreting antigen-specific CD4+ T cells. Future studies will need to address the question of whether these induced regulators, the naturally-arising IL-10+ regulatory cells that we have described, and so-called Tr1 cells [36
] are all one and the same.
To address the third key question, it is clear that selection of the appropriate antigen(s), around which to build the vaccine, is a key step. Tregs utilize a conventional T-cell receptor for antigen, through which signaling is required to initiate suppressor functions. Thus, a successful vaccination will be one which yields Tregs present at the site of inflammation that can be activated by the antigens being presented. In essence, this implies that the vaccine should be based on the same autoantigens that are the target of the inflammatory autoimmune response. Much progress has been made in the last two decades in defining these β-cell molecular targets, and, in several cases, the immunodominant peptides recognized by T-lymphocytes are also known [37
]. This provides options to use whole antigens or peptides as the vaccine with the caveat that autoantigens that are also bioactive molecules, such as insulin, may be limited to certain routes of administration.
Currently we are at an early stage in developing vaccines/antigen-specific immunotherapies in Type 1 diabetes, but several developments are noteworthy. First, where patients have been given autoantigens as part of a vaccine strategy, there have been few, if any, safety concerns [38
]. There has, for example, been no evidence of hypersensitivity induction. Acceleration of disease is also a theoretical concern, but it has been observed only rarely in animal models (in comparison to the number of studies conducted) [39
], although it may have been a factor in the relapse of a minority of patients treated with altered peptides in patients with multiple sclerosis [42
]. This experience suggests that care must be exercised in relation to dose (lower doses being better) and in relation to any attempt to enhance immunogenicity by deviation from the wild-type autoantigen sequence. Nonetheless, there has been no specific observation that antigen-induced disease acceleration is a risk in the Type 1 diabetes setting, even when there is very clear evidence that immunization has been achieved, as occurred for example in the studies of glutamic acid decarboxylase-alum immunization [44
]. A second noteworthy point is that a sub-study of oral insulin administration to at-risk first-degree relatives (the Diabetes Prevention Trial-1) demonstrated a significant delay of progression to diabetes in those subjects with high levels of insulin autoantibodies [45
], suggesting that a therapeutic effect is achievable, and again that appropriate immunological staging of participants is a key component of trial design. Further encouragement is derived from unpublished reports that the heat shock protein-derived peptide 277 (Diapep), a putative diabetes autoantigen [46
], has beneficial metabolic effects when administered parenterally in the context of a recent phase III study [47
]. A third point is that proinsulin and insulin peptides have also proved safe at early stages of clinical development, opening up the potential for epitope-based vaccines [48
]. These positive interpretations must be balanced, however, by other notable failures of antigen-specific immunotherapy. Nasal insulin has not proved effective in a large study of at-risk children in the prevention setting [50
] and alum-conjugated glutamic acid decarboxylase-65 (GAD65, a known islet autoantigen in Type 1 diabetes [51
]) therapy failed in separate studies at phase II conducted by Diabetes TrialNet [52
] and phase III conducted by Diamyd [53
]. There may, however, be important clues and lessons in these failures. Nasal insulin was given daily, a strategy which is known to be sub-optimal for Treg induction and might even be counter-productive by favoring deletion of Tregs, a well-described effect of abundant or frequent antigen administration [54
]. In addition, our own unpublished mechanistic studies on samples from the TrialNet study of GAD-alum show that it is a potent inducer of GAD-specific pro-inflammatory CD4 T cells, which may not be the required phenotype for a vaccine for autoimmune disease.
The final question is the key one: what are the translational steps that are currently required to realize the potential of vaccine-based therapeutics for autoimmune diseases, such as Type 1 diabetes, and avoid the pitfalls? A series of research imperatives can be outlined to address this.
First, the clinical activity needs to be maintained and extended. It is important that studies in man continue to be conducted, but if these are in the setting of recent-onset diabetes then the expectations of funders, patients, investigators and journal reviewers and editors will need to be carefully managed. The success of these studies should be judged by the development of robust data regarding safety and the identification of biomarkers of vaccine administration, with preservation of C-peptide (a measure of residual β-cell function) a secondary end-point. Thus, there should be a clear delineation of what “success” looks like for an early stage vaccine study in new-onset disease. Thereafter, it should be a relatively straightforward progression of a vaccine approach to move to the next stage. This will either be as a mono-therapy in a secondary prevention setting (i.e. subjects are identified as being at risk of progression to diabetes using autoantibodies; the best current example of this approach being oral insulin conducted by Type 1 diabetes TrialNet) or as a combination approach (see below), either in secondary prevention or as an early intervention after diagnosis.
Second, it will benefit the field if experimental models can be refined to examine vaccine approaches in a humanized setting, using the same agents as those that will go into man. To date, for example, there is no report of a humanized mouse that regulates glucose via synthesis of endogenous human preproinsulin, whilst preproinsulin is the basis for several vaccine approaches. These models could be particularly useful for understanding dose/dose frequency, route of administration and biomarkers, which will otherwise limit development of vaccine-based strategies. Third, there are considerable basic research questions about the fate of exogenous antigen administered to man, and its interaction with the immune system, that will need to be addressed. And finally, there is the question of whether antigen-alone vaccines should be used in combinations and with what additional immune modulators. Finding the best combination or adjunct therapy will be challenging. Preclinical models suggest that combinations of antigen and anti-CD3 monoclonal antibody therapy can synergise to induce sustained disease protection via induction of immune regulatory pathways [55
]. Theoretically, any manipulation that reduces inflammation has the potential to aid the development of a regulatory response to antigen, and thus combinations of antigens plus biologics that mediate cytokine blockade or cell depletion might be effective. This may present a conundrum to the field, however, should these single agents prove ineffective on their own; at this stage, preclinical models of antigen combinations will be an important part of the puzzle. In addition, there is the possibility that highly safe and moderately effective single agent biologics (in particular co-stimulatory blockers, such as CTLA-4Ig [57
]) will be counter-productive in combinations because they block pathways required by both effector and regulatory cells alike. Finally, scant attention has been paid in the past to “regulatory adjuvants”, but this is an area that may expand as commercial interests focus on therapies designed for autoimmune and inflammatory diseases, such as Type 1 diabetes.