The idea of acute infectious illness being caused by rampant overproduction of inflammatory cytokines that, in lower concentrations, mediate innate immunity, was first argued a quarter of a century ago 1, and has generated a large literature. Once recombinant tumour necrosis factor (TNF) and interleukin-1 (IL-1) became available in the late 1980s, and assays based on them replaced earlier methods, the concept spread to other pro-inflammatory cytokines, and from malaria and sepsis to viral and certain autoimmune diseases. It is now also well-entrenched in the literature of the post-trauma syndrome. Clearly, different triggers for cytokine generation and release can be expected to generate different foci, profiles, concentrations and kinetics of these mediators – now numerous enough to form superfamilies – and thus clinical variation within the same general principle is to be expected. But an accepted fundamental pattern has emerged.

At a time of shifting perceptions on the interaction between sickness and host activity (reviewed by Dantzer 2), Hart 3 argued that the distinctive behaviour of sick humans and animals (lethargy, anorexia, depressed motor activity etc.) was not simply another untoward aspect of being ill. Instead, it was reasoned to be an adaptive syndrome that had evolved as a protective mechanism to maximize chances of survival through encouraging the sick animal to seek out and remain in a safe resting place, and not search for food, until a survivable infectious episode had passed. Hart also proposed that sickness behaviour was caused by the inflammatory cytokines, TNF and IL-1. This was later confirmed by others 4, and investigated further through showing that IL-1 contributes significantly to the anorexia caused by both endotoxin and influenza infection 56. The literature on this field is now considerable. Indeed, Cavadini and co-workers 7 note that this link between TNF and IL-1 and sickness behaviour induced them to investigate if these inflammatory cytokines suppress expression of the clock genes that regulate circadian rhythm. It has recently 8 been realized that the circadian cycle controls many more biological functions than previously supposed.

Circadian rhythm and hibernation are variations of a theme of genetic control of activity and metabolism, and the degree to which these two clocks are functionally linked is often studied 9. The first example of a human disease being thought of in these terms was ischaemic heart disease, in which the concepts were adopted to explain the self-protective downregulation of metabolic function seen in cardiomyocytes after repeated bouts of ischaemia. Like the seasonal hibernation seen in Arctic mammals, it involves biochemical changes directed at preserving energy, so in 1997 was termed myocardial hibernation 10. Functional hibernation as a survival strategy in severe sepsis was proposed in 2003 11, and subsequently documented in myocytes in the mouse caecal ligation and puncture model of systemic inflammatory disease 12. Levy's recent reasoning 13 that broadens the possible relevance of cytokine-induced cardiomyocyte hibernation to other organs is, as Singer suggested 11, essentially sickness behaviour 3, with altered circadian rhythm being considered in terms of cellular bioenergetics. The consequence is to reduce demand for energy, in the form of ATP, to a level commensurate with its supply.

As outlined above, it seems reasonable that sickness behaviour, including lowering the metabolic rate to conserve energy and turning off appetite to avoid the urge to forage, will help tide the ill animal or human over an illness crisis. Recently, Cavadini and co-workers 7 reported that TNF and IL-1β suppress the expression of genes (Dbp, Tef, Hlf, Per1, Per2 and Per3) in vivo, and a nuclear receptor (Rev-erbα) involved in controlling circadian rhythm. The authors proposed that these cytokines thereby provide a link between the fatigue associated with autoimmune diseases, such as rheumatoid arthritis and Crohn's disease.

Nevertheless these cytokines, and others related to them, can cause a much wider spectrum of harmful changes than this 14. Why these changes, so familiar from the literature and everyday observation, should occur as a group, forming a distinct syndrome, currently lacks a rationale. It was therefore considered how widely the implications of severe sickness behaviour, through suppressed circadian rhythm 7, might extend, and whether this approach might be able to explain the composition of the severe systemic inflammatory syndrome common to malaria, sepsis, viral infections, and also seen post-trauma. The literature to date has focussed on mechanisms for individual changes, but not questioned why the mosaic contains the components one is accustomed to observing.

In other words, are inflammatory cytokines simply innately harmful in excess, pathophysiology arising from excessive manifestation of the physiological roles they possess when in trace amounts? Or, as suggested here (Figure 1), do the wider functions of the nuclear receptor (Rev-erbα) and genes (Dbp, Tef, Hlf, Per1, Per2 and Per3) 7 lead to an understanding of the syndrome to be expected when a severe inflammatory illness occurs? In documenting this suppression Cavadini and co-authors administered, to mice, only a tenth of the dose of bacterial lipopolysaccharide (LPS) required for a severe response in this species. Accordingly, the literature was examined to see whether these cytokine-suppressed nuclear receptor and genes involved in circadian rhythm have other functions that help us understand the pattern of changes routinely observed in the systemic inflammatory syndrome.

An active brain consumes much more energy than any other organ of similar size, largely through the demands of the Na⁺/K⁺ ATPase that runs the cell membrane Na⁺/K⁺ pump 15. Hence, it is likely to have been a prime conservation site for any set of adaptive changes that have evolved to conserve energy in order for the organism to survive, whether at the level of sickness behaviour or hibernation, both of which can be conveniently approached in terms of altered circadian rhythm. The fatigue of chronic inflammatory diseases such as rheumatoid arthritis is consistent with the capacity of TNF, to which this fatigue is linked in the literature, to suppress circadian rhythm genes 7. Sleep saves energy and pathways that lead to sleep are induced by TNF 16 or IL-1β 17. Also, electroencephalogram (EEG) delta wave power is decreased for several days when short interfering RNA (siRNA) targeting TNF is microinjected into the primary somatosensory cortex 18. This is consistent with TNF increasing cortical EEG delta wave power and, therefore, 19 being involved in sleep regulation.

The TNF and IL-1β-susceptible genes through which these changes are controlled have not been identified, but the above observations imply the same principles apply as the need to reduce ATP expenditure increases in urgency with more severe illness, progressing through the increased fatigue and sleep periods to the clinical signs of encephalopathy, including coma, in severe malaria, sepsis, or influenza. Clearly, this encephalopathy can be associated with a harmful outcome in these systemic inflammatory states, but its evolutionary function may simply be to save energy. Indeed, this strategy is likely to be successful in, for example, most adult cerebral malaria patients, who recover with negligible neurological deficit if other organ failure is not involved 20. TNF 21 and IL-1 22 are increased in proportion to the degree of coma in falciparum malaria. The attractiveness of impaired consciousness in this disease being based on awareness of the basic physiology and broad effects of these and functionally similar cytokines across acute disease in general – rather than on primary vascular blockage by sequestered parasitized red cells 23 – is enhanced by recent evidence, from Papua New Guinea 24 and West Papua 25, of the essentially non-sequestering Plasmodium vivax causing this encephalopathy to the same degree as does Plasmodium falciparum.

How might this coma come about? Biology's known gaseous signalling molecules offer possibilities, since they are increased in inflammatory states, and induce a hibernation-like reversible state of suspended animation that protects against low ATP levels in Drosophila melanogaster larvae (NO 26), Caenorhabditis elegans (CO 27) and mice (H₂S 2829). All three gases have the potential to operate through inhibiting mitochondrial cytochrome oxidase 303132. A striking contrast exists, however, in the roles of these three gases in mammalian inflammation – both NO 33 and CO 34 inhibit NF-κB, whereas H₂S upregulates it 35. This dichotomy may prove to explain why both NO and CO are anti-inflammatory in mice, including protecting against the mouse model of malaria encephalopathy 3637, whereas H₂S is pro-inflammatory, increasing plasma TNF levels in mice 38, and its inhibition protects in sepsis models 3839. These three gases warrant a detailed comparative investigation in this context.

A major unaddressed question is the origin of the widespread gene silencing that develops after the initial illness crisis of severe systemic inflammation 40, including H5N1 influenza in mice 41. Immunosuppression is common to haemoprotozoan diseases such as malaria 42 and trypanosomiasis 43, as well as in sepsis 44, influenza 45, and trauma 46. Patients who survive the initial acute effects of excessive levels of inflammatory cytokines often succumb during a subsequent period of immunological and metabolic shutdown, in a state variously termed anergy, immune paralysis, or monocyte deactivation, with monocytes unable to generate HLA antigens or enough cytokines for normal immunological responses 47, let alone the excess that might be harmful to the host 48. In short, a systemic hyperinflammatory response turns into a harmful hypoinflammatory state of unknown origin 49. Explanations for these events in sepsis have included anti-inflammatory cytokines out-producing inflammatory cytokines 50 and severe lymphoid cell apoptosis 51. However, both pro- and anti-inflammatory cytokines proved to switch off as the disease progresses 52, and the point of maximum T cell anergy correlates to a diminished apoptotic response 53.

LPS tolerance is a common model for gene silencing, and repressed gene expression, arising from disrupted transcription, is as widespread in LPS tolerance as in severe inflammatory disease. Indeed the two coincide, with patients recovering from malaria being tolerant to LPS 54 as well as to high loads of malaria parasites 55. In landmark studies in 1975 Lewis Thomas widened our awareness of the broad relevance of the changes seen in LPS tolerance by studying it in parallel to tolerance to haemorrhagic shock and trauma shock 56. Arguably, therefore, LPS tolerance is simply a particular way of detecting a process common to many disease states. For instance, the patterns of NF-κB expression in sepsis and LPS tolerance closely resemble each other 57, and monocyte inactivation is as generalized, with the same characteristics, in LPS tolerance as in chronic sepsis 58. LPS tolerance, and therefore sepsis gene silencing, has a functional link to the theme of this review, in that it has a distinct circadian rhythm 5960, and can be created by depleting mice of Per2, one of the circadian genes that TNF suppresses 7.

More recently, LPS tolerance has been explained in terms of RelB induction generating transcriptionally inactive NF-κB p65/RelB heterodimers 61 and epigenetic silencing of TNF 62. The latter seems a particularly cogent argument, with LPS tolerant, or silenced, cells exhibiting repressed production of TNF mRNA, retained binding of heterochromatin binding protein 1α, sustained methylation of histone H3 lysine 9, reduced phosphorylation of histone H3 serine 10, and diminished binding of NF-κB RelA/p65 to the TNF promoter. This group has combined these elements, which are consistent with basic studies on gene silencing 63, into a compelling case to explain this phenomenon in severe systemic inflammation 40. It is, moreover, consistent with the hypoinflammatory state being a logical consequence of the earlier cytokine storm that caused the excessive sickness behaviour.

These arguments strengthen the concept that the pro-inflammatory cytokines are the primary driver of the pathophysiology of severe infectious, including malaria, and post-trauma illness. An alternative view, that each pathogen causes disease through a different primary mechanism (eg death of viral-infected host cells or vasculature blockage by parasitized red cells in malaria 23), with cytokines providing only non-specific changes such as fever, still exists. This view is further weakened by the above literature and reasoning, since it is difficult to credit vascular obstruction (falciparum malaria) or cell death (viral infection) with the ability to generate the above changes, such as coagulopathy, hypoalbuminaemia or immunosuppression, in a single infectious disease, let alone a range of them, or post-trauma.

While sickness behaviour, comprising a series of changes caused through pro-inflammatory cytokines shutting down circadian rhythm, has survival advantages, its more intense expression in severe disease generates a range of more harmful alterations, recognizable as the severe systemic inflammatory response. These are controlled by more severe effects of the same cytokines (TNF and IL-1β) affecting the same genes and nuclear receptor(s) as suppress circadian rhythm. Accordingly, the coexistence of reversible encephalopathy, gene silencing, dyserythropoiesis, seizures, coagulopathy, hypoalbuminaemia and hypertriglyceridaemia in severe infectious disease, including malaria, can best be understood as an extreme form of sickness behaviour. It is further proposed that this pattern, familiar in such conditions, has persisted in evolution because it has a survival advantage when, most commonly, the syndrome is less severe. In terms of a basis for developing treatments for human disease, this implies that the observed hypoinflammatory changes seen in these conditions require no primary explanation other than they are logical consequences of excess earlier production of pro-inflammatory cytokines, although local variations in secondary, cytokine-induced mechanisms occur.

As knowledge of the range of genes controlled by the superfamilies of pro-inflammatory cytokines and the genetic control of pathological changes both expand, it seems likely that other components of the systemic inflammatory syndrome will be added to the list begun in this paper.

The authors declare that they have no competing interests.

All three authors contributed to wide-ranging discussions on the ideas contained in this manuscript. IC wrote the manuscript, to which LA and AB made invaluable suggestions.