Human malarial disease: a consequence of inflammatory cytokine release

Nearly thirty years ago, ignorance of accepted malaria wisdom, in a London tumour/virology lab where this attribute was universal, allowed a novel tumour-oriented interpretation to be put on the observation that the non-lethal mouse parasite, Babesia microti, was killed by the immune system in circulating red cells [9]. This also happened in some species of malaria parasites with which it cross-protected. Encouragement that this phenomenon was worth pursuing came from the Guy's Hospital group, then at the top of the malarial immunity tree, who observed the same unexpected and puzzling phenomenon when they challenged malaria-immunized rhesus monkeys [10].

Of particular concern for the then official Guy's dogma (malarial immunity operates through a specific antibody focussed on the merozoite surface, adopted unquestioned by other major vaccine groups), even unrelated parasites died en masse inside red cells when previously immunized monkeys were challenged [11]. In their view this could offer an explanation of why some monkeys had high levels of antibody expected to be protective, yet failed to resist a challenge infection [11], while others with little or no anti-merozoite antibody were immune [12]. A report from the US told a similar tale [13]. Clearly, some powerful influence beyond specific antibody, and inconvenient for mainstream thinking on immunity and vaccine development, was reproducibly occurring.

The observation that pre-treatment with the Bacillus Calmette-Guérin (BCG) strain of Mycobacterium tuberculosis controlled a subsequent infection with any of several strains of babesia or malaria in mice (no antibody, parasites dying in red cells, not phagocytes) was fortuitously timed with the publication of the first paper on TNF [14]. This allowed us, in collaboration with these New York tumour researchers, to propose novel roles for TNF in immunity and disease pathogenesis in malaria and sepsis [15, 16]. In summary, through linking the protective capacity of agents such as BCG with the degree to which they sensitised to bacterial lipopolysaccharide (LPS) it came to be realised that the pathology of LPS toxicity (and subsequently that of rTNF) and Plasmodium vinckei infections in mice were largely identical, cytokine-mediated, events. As noted in a contribution to Brian Maegraith's 1988 Festschrift Proceedings [17], the cytokine approach gave teeth to his inflammation-based arguments on malaria disease of forty years earlier [18].

As reviewed in 1987 [19], this experience with mouse babesiosis and malaria provided the insight that this anti-tumour mediator arguably had roles in both cell-mediated immunity (CMI) and the pathogenesis of infectious disease in general. As well as malaria, this concept was reasoned in 1981 to explain the mechanism of typhoid [15], sepsis in general [16], and viral diseases in 1989 [20], and it eventually spread across all acute infectious diseases (see [21] for a recent review). For example, within a few years it began to dominate the sepsis literature [22, 23], and the virulence of different strains of influenza, a disease that is a standard clinical misdiagnosis for imported malaria, has now been expressed in terms of their capacity to induce TNF [24]. While still engendering strong opposition from some malaria researchers [4, 25], these ideas have been readily accepted by scientists working on bacterial, rickettsial or viral diseases. A broad literature across infectious disease now describes inflammatory cytokines as having a beneficial role in host defence, but being harmful to the host if produced excessively. Indeed, the acceptance and applicability of this concept is now general enough for its biological evolution to be an independent subject for research [26].

Once neutralizing anti-TNF antibodies became available for human use, they were tested by others for efficacy against malarial parasites and disease. Unfortunately a central tenet of the concept (that the pro-inflammatory cytokines that cause disease are the same mediators that, in lower concentrations, are responsible for the innate immunity that controls parasite growth – see also tuberculosis etc. in next paragraph) was not adequately considered. TNF has been shown to inhibit a mouse malaria parasite in vivo [27], and Plasmodium falciparum in vitro, provided white cells to generate the next down-stream mediator, possibly nitric oxide [28], were present [29]. This is consistent with findings in human subjects [30]. Thus it is not surprising that anti-TNF antibody, by removing inhibitory pressure from the pathogen, can enhance the disease in falciparum malaria [31], as shown five years earlier in human sepsis [32].

The broad relevance of these malaria-origin concepts in immunity and disease is best illustrated by noting the consequence of passive vaccination against TNF for Crohn's disease and rheumatoid arthritis, now a large-scale routine treatment [33]. Its practical success puts the relevance of these pro-inflammatory cytokines in human inflammatory disease beyond doubt, and the major side effect (pre-existing or acquired tuberculosis, salmonellosis, or listeriosis becoming fulminant) nicely demonstrates its relevance to CMI against many pathogens. It is unlikely to be coincidental that all three that flare are on our list of organisms that protected against haemoprotozoan parasites, causing intra-erythrocytic death, and priming mice for TNF production [34, 35]. Evidently the host was setting up a cell-mediated response that would protect against these organisms. Being non-specific in nature, it also protected against haemoprotozoa as well. From this reasoning Coxiella burnetii, a crude extract of which was an extremely good protectant [36], and primer for TNF (E. Carswell, pers. comm.) will also predictably flare if anti-TNF is given, long term, to an arthritis patient harbouring this human pathogen.

In this text, TNF is used as a term of convenience to designate the pro-inflammatory cytokines as a whole. Other cytokines, such as lymphotoxin (LT), interleukin-1 (IL-1), interleukin-6 (IL-6) and soluble Fas ligand (FasL) serve similar functions. In passing, it warrants noting that the term TNF-alpha, while still common, has been obsolete ever since LT ceased being referred to as TNF-beta and reverted to its original name, allowing TNF to do the same. Although the literature connecting the pro-inflammatory cytokines other than TNF to malaria [37‐40] is as yet much smaller than that for TNF, this does not imply that their potential for understanding this disease is correspondingly minute. A TNF superfamily of 19 members signalling through 29 receptors has more recently been described [41]. Many of these mediators induce other cytokines and enzymes that add to the inflammatory cascade. For example, TNF induces migration inhibitory factor (MIF) [42, 43], and TNF, IL-1 beta and LT generate the inducible form of nitric oxide synthase iNOS [44]. Anti-inflammatory cytokines such as IL-10, IL-4, and transforming growth factor-beta (TGF-beta), also play active roles, and an imbalance between these and their pro-inflammatory counterparts often determines outcome in disease. Some tens of thousands of publications on inflammatory cytokines and systemic inflammatory and other disease now exist.

The pro-inflammatory cytokines most closely investigated in malaria, such as TNF, usually act as homeostatic agents, but can cause pathology if produced excessively. A recently defined example is MIF (see above paragraph), belonging to an ancient gene family, with structural homologues in bacterial organisms. As an indication of the broad relevance and complexity of these cytokines and their interactions, MIF is also induced in mammals by E. coli lipopolysaccharide [42], staphylococcus toxic shock toxin, and streptococcal pyrogenic toxin A [45]. Whether this is direct, or via the TNF they induce, appears not to have been ascertained. Likewise, MIF can act directly, through TNF, or in synergy with it, in generating anaemia, as discussed below. It is, as noted above, TNF-induced, and remarkable for several reasons, one being that its description 40 years ago [46, 47], began the concept of what are now called cytokines. Ten years later, and long before MIF was realised to have functions other than migration inhibition and pathogenicity in sepsis (which its inhibition suppresses, in a realistic model, most impressively [48]), it was the first cytokine described in a malaria infection [49]. A few years after its rediscovery as a homeostatic glucocorticoid antagonist [50, 51], it has become central to understanding malarial anaemia, as discussed below. It is also increased in malarial placentas [52].

Cytokines such as TNF and IL-1, both increased in a wide range of systemic inflammatory diseases, including falciparum malaria, can induce a late-onset, but long-acting wave of a cytokine termed the high mobility group box 1 (HMGB1) protein, which prolongs and amplifies inflammation [53, 54]. This molecule, previously known for several physiological functions, now shows great promise as a therapeutic target in sepsis, in that countering it after the onset of illness has been reported to protect well in experimental sepsis [55, 56]. HMGB1 has been shown to be increased, in proportion to degree of illness, in serum from African children infected with falciparum malaria [57]. Like TNF, HMGB1 has roles in other inflammatory diseases [58], reaffirming malaria's position within their ranks. The malarial context of HMGB1 is reviewed more fully elsewhere [21].

The idea of malarial disease being caused by parasites releasing a toxin is even more venerable than that of vaso-occlusion, since it is based on a report by Golgi in 1886 [59], in which he noted onset of fever and rigors at a predictable short interval after the regular shower of new parasites escape from bursting red cells. These principles were much discussed in the first decade of the 20th century [60]. Clearly, something like this was needed to explain how tissue not invaded by the parasite was nevertheless damaged during falciparum malaria. Examples are sites such as the adult kidney and lung, where dysfunction can be catastrophic, yet sequestration never obvious, and often absent. The toxin idea lay fallow for many decades, not helped, in hindsight, by the underlying assumption that toxicity arose directly from a parasite product, in the manner of tetanus toxin.

The proposal that malarial products were not harmful in themselves, but only through causing the infected host to harm itself through generating toxic amounts of molecules that, in lower concentrations, inhibit growth of malarial parasites, gave the toxin concept new impetus [15, 16, 19]. These papers predicted that the nature of the malarial product that triggers illness could be defined through its ability to induce release of TNF from mammalian cells. A group in London did much work along these lines in the late 1980s and early 1990s [61, 62], and concluded it was closely related to phosphotidylinositol (PI) [63]. Others extended this argument to the glycosylated form of this molecule (GPI) [64]. The original proposal that malarial toxin operates through inducing generation of TNF and related cytokines was greatly strengthened when immunizing mice against GPI and then infecting them with one of the mouse malaria parasites protected against certain pathology that TNF causes on injection [65]. Indeed, this study reports having established that GPI appears sufficient and necessary for the induction by malarial parasites of host pro-inflammatory responses in vitro. The field has been well reviewed recently [66], with these authors and others expressing doubts about the wisdom of vaccinating against GPI to prevent malarial disease [21, 66, 67]. As noted some years ago, the need for sufficient TNF to allow immune activation to proceed normally during infections is plausibly why this potentially lethal mediator has survived 300–500 million years of evolution [68]. However, despite recent reaffirmation of the GPI/cytokine/disease concept [69], the group that first suggested that GPI was the main TNF inducer in malaria appear to have recently [25] changed their disease model to one that eliminates a requirement for inflammatory cytokines. In view of GPI having been identified through its capacity to induce pro-inflammatory cytokines, it would have been remarkable to chance upon a molecule that induces these mediators, yet mimicks their actions in their absence. Clarification awaits a more detailed report.

The extensive parallels that exist between the sepsis and malaria literature can be viewed from the perspective of the wide range of functionally-important inflammatory cytokines present in the circulation in both conditions (Table 1). This strengthens the view that the two diseases operate through very similar mechanisms. Nevertheless, a group working with African children has recently advocated that in order to understand falciparum malaria disease one must return to the pre-cytokine era. They evidently still espouse the idea that local vaso-occlusion uniquely sets the organ pathology of this disease apart from others with which it is clinically confusable, in particular sepsis [3‐5, 70]. Since the failure of treatment with corticosteroids to ameliorate severe cerebral malaria has been used as evidence against cytokine involvement [4], it warrants recalling that MIF, known to be high in this circumstance [71], antagonizes glucocorticoids [72], and nitric oxide (noting iNOS is also high [71]) inhibits glucocorticoid binding to its receptor [73]. Moreover, this data rationalizes the failure of corticosteroid as a treatment.

Analogy with other diseases is still an under-exploited tool in malaria. The original interest in TNF as a possible mediator of both innate immunity and disease pathogenesis in infectious disease came from analogies between the ability of BCG to protect against both tumours and intra-erythrocytic protozoa [9]. When introducing the excess inflammatory cytokine concept in 1981 [15], a common case for malaria, gram-negative bacteria and the Jarisch-Herxheimer reaction, all of which withstood the test of time, was argued by analogy. As reviewed recently [21], the range of infectious diseases that come under the systemic inflammation umbrella now extends beyond bacterial diseases to those caused by rickettsias, protozoa other than malaria, and viruses. Moreover, increased circulating levels of TNF and functionally similar cytokines have been measured in the serum very soon after onset of illness in virtually all those infectious diseases in which it has been sought. In addition, essentially all of the signs and symptoms involved in the clinical confusability of malaria and other causes of fever were inadvertently reproduced during the era when rTNF was being injected in volunteers as an antitumour agent [74, 75]. This includes headache, fever and rigors, nausea and vomiting, diarrhoea, anorexia, myalgia, thrombocytopaenia, immunosuppression, coagulopathy and central nervous system manifestations, all of which have a literature on a mechanism through inflammatory cytokines. The rate, timing and intensity of cytokine (pro- as well as anti-inflammatory) release will vary in different disease states, and also between individuals, and provide them with somewhat distinctive clinical pictures, but the fundamentals remain. The clinical patterns generated are remarkably close, in that, at least in some populations, clinical features cannot predict a diagnosis of malaria from other causes of fever [76].

The principle extends beyond infectious diseases. A number of non-infectious states fit this pattern, with excessive release of pro-inflammatory cytokines producing a systemic inflammatory response. As in malaria and sepsis, metabolic acidosis [77, 78], hyperlactataemia [79, 80] and encephalopathy are seen in tissue injury syndromes such as heatstroke, trauma, and burns. As reviewed [21], all of these conditions are ripe for an explanation in terms of HMGB1, liberated from the nuclei of damaged tissue [81], setting the scene for a broad range of inflammatory cytokine release. Iatrogenic cytokine release syndromes, such as the side effects of OKT3 therapy [82], and acute graft versus host disease reaction [83, 84] can also exhibit these changes, including a reversible encephalopathy. In both of these conditions the relevant pro-inflammatory cytokines are produced excessively, and where tested (side effects of OKT3 therapy [85], and acute graft-versus-host disease reaction [86]), prior exposure to neutralizing antibody directed against TNF prevents illness.

As with much research on neutralizing antibody to TNF, this outcome does not imply that TNF is more important than, for example, IL-1 in this context, since anti-IL-1 antibodies have rarely been tried. Blocking IL-1, and indeed IL-1 and TNF simultaneously, are in their infancy, but show promise [87, 88]. Likewise, research into the disease aspects of LT, present in falciparum malaria, and relevant to the mouse model of cerebral malaria, is relatively ignored, largely through difficulty of obtaining reagents. Additional strong evidence for inflammatory cytokines and falciparum malaria being functionally intertwined comes from studies on variation in the human genome in Africa [89‐91]. It is now accepted that falciparum malaria, historically the major fatal endemic disease in much of this continent, is associated with polymorphisms of these pro-inflammatory cytokines and iNOS, which are induced in this disease. Not surprisingly, sepsis and meningococcal disease have a similar literature [92, 93]. Like any other DNA trying to survive, that of humans uses trial and error to adapt itself to its surroundings, leaving a trail of evidence as it does so.

In summary, illnesses arising from excessive systemic production of inflammatory cytokines include not just malaria and sepsis, but many more infectious, and non-infectious, diseases. Insights gained by recognizing the value of argument by analogy across this wide spectrum have been immense, and the general concept is now so firmly entrenched that, as noted earlier, its influence on the evolutionary effects of infectious disease is a research topic in its own right [26]. It would, therefore, be most unexpected were the illness of falciparum malaria, so clinically confusable with other infectious diseases, and known to generate the same inflammatory cytokines as they do, to arise from an unrelated mechanism [5, 25].

One of the many actions of the cytokines responsible for systemic inflammation is to disable oxidative phosphorylation within mitochondria. This is reflected in the hyperlactataemia commonly seen in severe infectious disease, and correlating with outcome. This is not to suggest that oxygen supply and its utilization are not often limited simultaneously, and interact. Indeed, if toxin is replaced by its downstream consequence (the effects of pro-inflammatory cytokines) such interaction was proposed for malaria by Meleney over 60 years ago [94].

Both sepsis and malaria researchers have shown that injecting TNF, the prototype inflammatory cytokine, increased in both diseases, causes hyperlactataemia [95, 96], and blood lactate levels in severe malaria have proved to correlate closely with levels of both TNF and interleukin-1 [97]. Nevertheless, it is fair to say that sepsis researchers, without the tradition of a primary role for sequestration to defend, have been more receptive than malariologists to the systemic inflammatory explanation for altered carbohydrate metabolism, and more readily pursued it. Thus, they were more open than most of their malaria counterparts for the insights of the early 1990s, when newer techniques demonstrated that oxygen tension in tissues was actually increased (not decreased, as the literature predicted) in septic rats [98], patients [99], and pigs [100]. This implied an inability of mitochondria to utilize oxygen, forcing glycolysis to compensate, as best it can, for the energy deficit. The next insights came from groups who developed the cytokine-induced mitochondrial dysfunction model of disease [101‐103] and, thus, provided an inflammation-based explanation for a shutdown of aerobic glycolysis, a consequent increased rate of glycolysis, and thus lactate production, metabolic acidosis and cellular energy depletion. In any disease with high levels of inflammatory cytokines this mimics poor oxygen supply. An important difference, however, is that the effects of the nitric oxide through which mitochondrial shutdown largely operates are reversible [104, 105], whereas frank hypoxia, through vaso-occlusion, as evidenced by the stroke literature, is less so. Thus, the former, not the latter, is consistent with the marked reversibility of metabolic comas [106], a term advocated [21, 107] to include human malaria.

This approach to understanding energy balance in sepsis has been followed successfully on a number of cell types, including hindlimb skeletal myocytes, gut wall cells, and hepatocytes. The wide range of tissues in which these concepts have been demonstrated adds to the arguments on systemic origins of lactate in sepsis and malaria. For example, inflammatory cytokines have been shown to cause contractile dysfunction [108, 109] and also energy depletion [110, 111] through effects, often mediated through induced nitric oxide [109, 112], on cardiac muscle. Likewise, in diaphragmatic skeletal muscle there is evidence of cytokine-induced nitric oxide [113, 114] and oxygen-derived free radicals [115] combining to form peroxynitrite [116, 117] and this causing dysfunction of mitochondria in myocytes, leading to energy depletion and thus muscular contractile failure. The outcome here is to reduce the patient's ability to counter acidosis by blowing off CO₂.

An additional pathway through which the inflammatory cytokines may reduce oxygen consumption is through peroxynitrate (OONO^-), a product of NO (from iNOS induced by these cytokines) and superoxide, overactivating poly(ADP ribose) polymerase-1 (PARP-1) [118]. This can deplete cellular stores of NAD⁺, and efforts to resynthesise it can deplete ATP as well (reviewed in reference [119]). Moreover, NAD⁺ is essential for glycolysis, so its depletion can be expected to impair glycolytic input into mitochondria. These concepts were reviewed in depth in a malaria context a few years ago [107].

It is well accepted that upregulation by inflammatory cytokines of adhesion sites on endothelial cells invites susceptible circulating blood elements to attach to the inner wall of blood vessels. In many diseases, including malaria, this includes activated leukocytes and platelets, both of which play important roles in promoting procoagulant activity. For example, in malaria this activity is seen on circulating monocytes [120] and placental macrophages [121], and the thrombin so formed enhances adhesion by increasing expression of CD36 on platelet surfaces [122].

As recently reviewed [176], critical illness associated with an inflammatory response invariably causes multifactorial anaemia. It has often been noted that anaemia could contribute to poor oxygenation of tissues in malaria [177] and there is general acceptance that it can be severe enough to reduce supply of oxygen to mitochondria to dangerously low levels. Thus it can be a major component of malarial pathology. Obviously a high parasite load indicates imminent widespread lysis, but anaemia does not correlate with parasitaemia, and sometimes is extreme when very few parasites are present.

Cytokine-induced myocardial depression frequently accompanies severe sepsis (see [220]). Whereas it was previously considered a pre-terminal event, it is now clear that cardiac dysfunction, as evidenced by biventricular dilatation and reduced ejection fraction, is present in most patients with severe sepsis. It has been known for some time to be caused by soluble factor(s) released by macrophages exposed to endotoxin [221]. Once cloning of cytokines occurred its activity was attributed to IL-1, then also to TNF [222], then the two synergistically [223], and finally to IL-6 [224], a macrophage product induced by both of these mediators. A literature exists on these effects being minimised by blocking MIF, which reduces the feedback inhibition of TNF production by glucocorticoids [225, 226]. As discussed in above, excess inflammatory cytokines have been shown to cause cardiomyocyte mitochondrial dysfunction. Since TNF [38], IL-1 [38], IL-6 [39] and MIF [71] are all highly expressed in falciparum malaria, it can be expected that these cardiac-depressing activities would be acting in this disease as well as in sepsis. Evidence of this, in terms of circulating cardiac proteins, has accumulated in the past few years [227, 228], although its clinical impact is yet to be evaluated. It may be present, but its potential clinical impact is over-ridden by the effects of hypovolaemic shock, as summarized in the next section.

Insufficient intravascular volume is ultimately of concern in disease because, through poor perfusion, it leads to poor oxygen supply where it matters, the microcirculation that feeds the mitochondria within the cells that form the tissues the capillaries pass through. The major therapeutic option is volume resuscitation. As recently reviewed [229], it is under the control of a number of autoregulatory mechanisms, and these are known to be disrupted in sepsis. As well as the effects of changed red cell deformability, and adherence of platelets and leukocytes, as discussed above, variation in iNOS induction, leading to more or less nitric oxide, essential for local degrees of the vasodilation that perfusion depends on, are major controlling factors.

Using a range of indicators, workers in Kenya have confirmed the older observation [230] that shock is not rare in severe falciparum malaria [231], and that the haemodynamic changes in children with severe malarial anaemia complicated by the respiratory distress were more characteristic of hypovolaemia than of biventricular failure [232]. They have also demonstrated that while administering albumin did not improve acidosis, it did reduce mortality [233]. This conceptual approach has been strenuously questioned by others who detected only a mild fall in total body water volume and extracellular water volume [234], as well as the relative rarity of severe hypotension in falciparum malaria compared to the shock that can accompany trauma or sepsis [4]. However, local effects may be much more important in falciparum malaria than in sepsis – for example these could in part arise from vasodilation being much more uneven in malaria than sepsis because of patchy local foci of post-schizogony malaria toxin release from sequestered parasites, and local generation thus of the inflammatory cytokines that induce endothelial iNOS [71]. It is recognized that during treatment one would need to be cautious of fluid overload if the patient displays evidence of cerebral oedema [235] or cardiac insufficiency [236].

The detailed arguments on both sides of this debate are beyond the scope of this review, except to note that they are currently a major fault line between those who propose that malaria has a fundamentally similar pathophysiology to other acute systemic infections [21, 233, 237] and those who see it as unique [238]. This issue cannot be resolved until recognition is given to the need to research the pathophysiology of malaria and other systemic infectious states in parallel rather than, as at present, in isolation.

From this and the previous section, it is not hard to visualise the combined harmful effect on the patient when systemic inflammation reduces oxygen supply to their cells also makes these cells worse at using it. As shown in Figure 3, the initiating pathophysiological lesion is the onset of the systemic inflammatory response, and it is difficult, from the evidence, to envisage sequestering parasitized red cells, per se, initiating malarial disease before it is focussed to sensitive organs by systemic release of inflammatory cytokines. Sequestering parasitized red cells may then, in part through locally released cytokines, exacerbate the illness if the patient survives long enough.

Hyperlactataemia, a recognized marker of falciparum malaria severity, is at the centre of controversies relevant to the theme of this review. Its discussion requires some basic biochemical background. The lactate anion has complex roles in biology. Hyperlactataemia may be associated with acidosis, a normal pH, or alkalosis [239], and can occur in viral and rickettsial diseases [240], as well as (see below) sepsis and malaria. In synopsis, most lactate is generated during glycolysis, which essentially consists of oxidising glucose, a six-carbon structure, into two three-carbon molecules of pyruvate. This is reduced to lactate through the action of pyruvate dehydrogenase, a reaction that avoids pyruvate accumulating, and supplies NAD⁺ to keep glycolysis going. Thus lactate can be formed as a byproduct of glycolysis, which can occur in all metabolically active tissues and supplies ATP, albeit in small amounts, independently of the presence of oxygen. Every mole of glucose metabolised by anaerobic glycolysis to carbon dioxide and water yields 4 moles of ATP, whereas oxidative phosphorylation within mitochondria yields 32 moles of ATP. When oxygen usage falls (whether through poor supply or poor utilisation) ATP generation falls, and glycolysis is accelerated to compensate, as much as possible, for this energy loss. A consequence is oversupply of the byproduct, lactate, but from a disease perspective this is a side issue compared to insufficient ATP generation, even though the two may correlate well. Enhanced glycolysis under aerobic conditions can also increase lactate production. The metabolic acidosis secondary to this failure of mitochondrial energy production, which high lactate often accompanies, is a consequence of this energy failure, and inevitably accompanies it in severe inflammatory illnesses, including malaria and sepsis.

The body's supplies of glucose, including stores of its polymer, glycogen, are not unlimited, so when glycolosis is enhanced for any period it sooner or later runs out of fuel. Gluconeogenesis fills the breech as much as possible, but it soon fails because substrate supplies are limiting [241]. These events are reflected in the hypoglycaemia that has often been reported in severe malaria [242] and sepsis [243, 244]. When seen in this context hypoglycaemia in these diseases is no longer a primary cause of harm, such as when driven by hyperinsulinaemia, but an inevitable consequence of over-exuberant, typically anaerobic, glycolysis.

There are real differences in the nature of the disease in children and adults. For example, severe anaemia is much more common for a given parasite density in young children than in older children and adults [300], and seizures follow a similar pattern [357]. In contrast, cerebral malaria is less common in the very young, and the pulmonary and renal complications seen in adults are very uncommon in children [300]. Moreover, children can experience a type of clear lung respiratory distress not reported in adults [358], implying a more severe metabolic acidosis driven by the need to blow off excess CO₂ and restore normal pH. This is consistent with hypoglycaemia (from high glucose consumption feeding the anaerobic state that is causing proton accumulation) being a characteristic of childhood malaria more than of adult cases.

The cytokine theory of disease provides directions that could lead to an understanding, currently lacking, for these age differences. For instance, generation of nitric oxide, which has been instrumental in the Na⁺/K⁺-ATPase inhibition thought to cause poor red cell deformability, and thus shorten erythrocyte life span, is higher in younger children [359, 360]. Additional nitric oxide generated by a systemic inflammatory response would add to this higher base level, reducing red cell deformability correspondingly more in the very young. This could be tested. Likewise, pulmonary and renal organ failure being common in severely affected adults, but virtually absent in children [300], has an exact parallel in the age differences for the inflammatory response that follows trauma [361]. This has been explained in terms of the different balance of inflammatory cytokines generated from normal peritoneal cells collected from children vs adults [362, 363]. There seems no reason why these data, derived from normal human cells, should not be equally applicable to any human systemic inflammatory state, including malaria. This enhanced anti-inflammatory response of young children can also be expected to minimize the proposed focussing of sequestration during severe illness in this age group, contributing to their low incidence of cerebral malaria [300].