SEARCH

SEARCH BY CITATION

Keywords:

  • autoimmunity;
  • diabetes;
  • viruses/viral immunity

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References

The hypothesis that a virus might in some way be involved in the causation of type 1 diabetes has a long history, but decades of research have failed to resolve the issue beyond reasonable doubt. Viruses could potentially play a primary role in the pathogenesis of type 1 diabetes by initiating autoimmunity, a secondary role by promoting established immune responses, or a tertiary role by precipitating the onset of hyperglycaemia. There is currently little evidence to suggest that viruses play a primary role in the causation of type 1 diabetes, let alone a necessary or sufficient role. Secondary or tertiary roles remain possible, but have yet to be confirmed in prospective studies.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References

The notion that a virus or viruses might be involved in the pathogenesis of type 1 diabetes has a long and chequered history. Mumps was suspected in the early part of the last century, based upon epidemiological associations between outbreaks and the incidence of diabetes, together with the known propensity of the virus for the pancreas [1]. Fifty years later, diabetes was noted to be a feature of the congenital rubella syndrome, proof of principle that a virus could cause diabetes, although the diabetes may have had more to do with overwhelming viraemia than with the induction of autoimmunity [2].

Viral infection emerged as the catch-all explanation for unexplained illness in the 1960s. Dennis Burkitt's work led to the discovery that his eponymous lymphoma was related to infection with the Epstein–Barr virus, and Peyton Rous was hauled out of retirement shortly afterwards to receive a Nobel Prize for work performed on his sarcoma virus 50 years previously. In the United States, the National Cancer Institute launched a Special Cancer Virus Program that was to absorb 10% of its budget over the coming years, for little perceived benefit [3].

The autoimmune paradigm emerged at around the same time. This linked organ-specific autoimmunity with human leucocyte antigen (HLA) susceptibility, and bridged the gap between the two with known or hypothetical environmental triggers. Viral infection was an obvious candidate. The first clear expression of the emerging paradigm for type 1 diabetes was formulated in 1974 as follows:

one or more immune-response genes associated with HLA-A8 and/or W15 might be responsible for an altered T-lymphocyte response. The genetically determined host response could fail to eliminate an infecting virus (Coxsackie B4 and others) which in turn might destroy the pancreatic beta-cells or trigger an autoimmune reaction against the infected organ [4].

Forty years on, as Yogi Berra might have remarked, it is ‘déjà vu all over again’. Revolutionary new methods of genetic analysis have confirmed that genetic susceptibility to type 1 diabetes is overwhelmingly determined by the HLA system, the prehistory of type 1 diabetes has been traced forward from the womb in prospective studies, and increasingly sophisticated animal models and bench techniques have added plausibility to the concept that a virus might mediate diabetes. The only thing we lack is certain knowledge as to whether a virus, or more specifically an enterovirus, is involved in the causation of type 1 diabetes. Nor are we alone in this, for viruses have been sought actively in multiple sclerosis and many other autoimmune conditions, yet not one single candidate has yet been found to satisfy Koch's postulates for a causal relation between microbial infection and autoimmune disease [5].

Three potential explanations should be considered. The first is that there is little or no causal association between virus infection and human type 1 diabetes, and that the literature has accumulated by a combination of wishful thinking, random noise and publication bias. The second is that such an association exists, but we lack the technology or the experiment needed to prove it. The third is that we are asking the wrong question.

More to the point, what exactly is the question? The viral hypothesis has rarely, if ever, been formulated in terms which are capable of disproof. There is indeed no single hypothesis, merely a cluster of epidemiological, experimental and clinical observations clustered around the proposition that viral infection is in some way implicated in the pathogenesis of human type 1 diabetes. This degree of imprecision is easily explained by the staggering array of possible relationships between virus and disease, and the variety of possible mechanisms underlying any such relationship.

Here are some possibilities. Viral infection might play a direct role in the causation of type 1 diabetes. It might do this directly, for instance by triggering a mechanism (e.g. lysis of beta cells) which initiates some sort of doomsday scenario within the immune system. It might do this indirectly, by modifying the activity of the cell in such a way as to make it show up as a target on the immune radar screen. A virus might be a necessary co-factor in the development of diabetes, or just one among a number of possible co-factors, contributing to some cases but not others. It might play different roles at different stages of the process, whether by initiating, promoting or precipitating the development of disease. Alternatively, the genetic background that favours the development of diabetes might simply happen to provide a convenient loophole for invasion by certain types of virus, or such a loophole might be created by the immune and metabolic defects associated with the onset of diabetes; in either case we will have been focusing our attention upon an innocent fellow passenger. The list could easily be extended.

A second list begins when we come to possible mechanisms. Are we dealing with a specific viral strain, or with a range of viruses producing comparable effects? Is the infection acute, chronic or latent? Is there direct cellular damage, with exposure of intracellular antigens, activation of death mechanisms within the cell, altered signalling at the cell surface, molecular mimicry, generation of a proinflammatory environment mediated by cytokines and chemokines, or several of the above? Alternatively, do viruses cause harm simply by not being there at the right moment in ontogenesis of the immune system? Almost any known mechanism involving viral invasion or the ensuing immune response can plausibly be invoked, and the aspiring investigator need only identify the experimental system in which his or her preferred model actually works, as against those in which it does not.

The striking feature of the debate, if debate it has been, is its one-sided nature. Glory beckons for those who can find a virus, or who believe that they have done so, whereas anonymity is reserved for those who doubt the hypothesis, or generate negative results. The argument has therefore been driven almost exclusively by proponents of the viral hypothesis. This has given it certain characteristic features: for example, the tendency for those who are engaged in detailed mechanistic studies to overstate the epidemiological evidence, or vice versa; for mutually inconsistent reports to be recruited to the cause, for negative studies to remain uncited, and for obvious counterarguments to be overlooked.

How, then, do we escape from this game of smoke and mirrors? To their great credit, the contributors to a recent Mini-Review series in Clinical and Experimental Immunology have avoided most of the pitfalls listed above, and the collection offers an excellent and balanced overview of current thinking in the area. Their contributions, and indeed all research in this area, revolve around two central questions: what is the evidence that a virus (or viruses) plays a causal role in human type 1 diabetes; and if so, by what mechanism? This Commentary will focus on the first of these questions, subdivided into two parts: is there a smoking gun; and is there a bullet?

Is there a smoking gun?

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References

Epidemiology originated with the investigation of infectious disease, and epidemiologists are therefore well equipped to recognize the footprints of infection in the natural history of a disease. These footprints have, for the most part, been lacking in type 1 diabetes. Enteroviruses, the preferred candidate group, are transmitted by the faecal–oral route among others, and their transmission is reduced by improved hygiene. Furthermore, as exemplified by polio, they show an epidemic pattern of spread. On this basis, one could predict confidently that a disease caused directly by enterovirus would show a low and declining incidence in affluent countries and an epidemic pattern of spread. Instead, and as we all know, type 1 diabetes has risen more or less continuously in many populations since World War 2, and the increase has been most striking in children of European descent living in relatively affluent environments [6]. There have been no reliable reports of epidemics. Instead, we face a pandemic.

This may not close the book on enterovirus and type 1 diabetes, but it certainly narrows the field of possible explanation. Investigators have responded by looking for evidence of exposure before birth or in early infancy, and this has given rise to a series of major birth cohort or longitudinal studies, reviewed critically by Stene and Rewers in this issue [7]. They point out that viral infection might potentially influence the onset of diabetes at two key transitions: from genetic susceptibility to the onset of autoimmunity, and from established autoimmunity to the onset of hyperglycaemia. Here, again, the summary can be brief, for they found no consistent evidence that enterovirus infection can predict the development of islet autoimmunity, and (with one important exception [8]) no evidence that viral infection can accelerate the progression of autoimmunity.

The further possibility they discuss, as do Coppieters and colleagues [9], is that viral infection might be protective if encountered in early infancy, whereas delayed exposure is more likely to be diabetogenic. This, the polio hypothesis, actually does not work that well for polio [10], and has the further limitation that prospective studies from birth show that the first islet autoantibodies have typically appeared by the age of 2 years, and that multiple autoantibody positivity, the necessary prelude to most cases of childhood diabetes, is typically established by the age of 3 years [11], an observation which does not allow much opportunity for delayed exposure to the virus. A broader possibility is that early development of a robust pattern of immune response relies upon ‘education’ of the immune system by a wide range of antigens, viral infection included. Absence of microbial or parasitic infection is indeed necessary for the spontaneous development of diabetes in the non-obese diabetic (NOD) mouse or BioBreeding (BB) rat, and fits reasonably well to the epidemiology of type 1 diabetes and other autoimmune diseases [12,13].

Is there a bullet?

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References

Although little smoke appears to have issued from the barrel of the gun, this step in the argument could be skipped if there is evidence that a viral bullet is indeed lodged in the child with diabetes. Serological studies have provided no clear evidence of recent viral infection in the newly diagnosed, but meta-analysis of studies of circulating viral RNA has reached a more positive conclusion. A signal was (on aggregate) present in 42% of cases, with a 10-fold odds ratio compared with controls. An equally high ratio was observed in long-standing cases, but – an important but – no excess of viral infection was seen in those studied prior to the onset of diabetes [14]. As Stene and Rewers point out, this particular analysis has its limitations [7]. It calculated pooled estimates across studies despite considerable heterogeneity, for example, and lumped together studies which had used different methods to demonstrate a viral signature. Many cases came from a study that has, as yet, been published only in abstract. Few analyses had been matched for HLA, and so on. Compelling, therefore, but not yet conclusive.

The unequivocal place for a viral bullet to be lodged is within the beta cells of recently diagnosed individuals who come to autopsy. This mercifully rare situation is well described in the contribution by Grieco et al. [15], who provide strong evidence that viral material can sometimes be found within the beta cell, and review a range of mechanisms by which viral infection might modulate beta cell function.

A variant form of type 1 diabetes has been described in Japan and East Asia. The characteristics include very abrupt and aggressive onset of symptoms, high blood glucose with acidosis but a relatively low HbA1c at diagnosis (implying acute onset of hyperglycaemia), absence of circulating islet autoantibodies and increased levels of pancreatic enzymes in the circulation, consistent with inflammation of the exocrine pancreas. These observations are strongly suggestive of acute viral infection, although conclusive proof is still lacking [16]. At present this appears to be a distinct disease variant, of limited relevance to the chronic immune-mediated process seen in typical type 1A diabetes.

Taken together, these studies suggest that viral RNA can be detected in a proportion of those with recent or longer-term ‘classic’ type 1 diabetes, but its role in the pathogenesis of the disease remains unclear. Methodological difficulties abound. This phenomenon clearly requires further investigation, and may in time provide some fascinating answers.

An inconclusive conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References

Francis Crick remarked that ‘impossible’ areas of science were populated almost exclusively by optimists, everyone else having departed. The quest for viral involvement in type 1 diabetes is well suited to optimists, but should be considered difficult rather than impossible. Our ignorance has become incomparably better educated, and methodological advance can be expected to bring much greater clarity to an area still plagued by inconsistent results. Furthermore, the launch of a series of major prospective studies from pregnancy or birth has allowed the pathway from genetic susceptibility to immune activation, and from immune activation to onset of disease, to be mapped in increasing detail.

Viruses could potentially be involved at three stages in the pathogenesis of classic type 1 diabetes: they could play a primary role in initiating autoimmunity, a secondary role in promoting it or a tertiary role in precipitating the onset of disease. On present evidence, and despite one or two hints in the opposite direction, there is no strong reason to believe that viruses play any role in primary causation of type 1 diabetes, and certainly not a necessary or sufficient role. Secondary or tertiary roles remain possible, but have yet to be confirmed in prospective studies; much of the argument rests upon evidence acquired in those with recently acquired or longer-standing diabetes. If confirmed by future studies, such evidence remains open to the possibility of reverse causation. There is a long and difficult road ahead, but the landmarks are slowly emerging from the mist.

Disclosure

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References

No potential conflicts of interest have been disclosed in relation to this manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Is there a smoking gun?
  5. Is there a bullet?
  6. An inconclusive conclusion
  7. Disclosure
  8. References