Complexity of immunological processes in the pathogenesis of malaria

In their letter, Geoff Butcher and Janice Taverne speculate on the potential dangers of glycosylphosphatidylinositol (GPI)-based vaccines. The topic of GPI-based vaccines was only a minor component of our Review¹, which focused on the immune mechanisms elicited during infection with the malaria parasite Plasmodium falciparum that might cause disease. We did include a brief discussion about the possibility of developing GPI-based vaccines, and followed this by a brief summary of the opposing views, which are reiterated in the letter from Butcher and Taverne. We noted that safety concerns apply to all malaria vaccines under development, including GPI-based vaccines. Therefore, we do not believe that we ignored safety concerns about GPI-based vaccines. In addition, although Butcher and Taverne assert that the host, rather than the parasite, produces the toxins that initiate disease, we believe that the parasite produces the toxins that initiate disease. Because malaria GPI is highly conserved, ubiquitous and produced in excess of the amount required for anchoring, we questioned whether it is the target of immune responses that efficiently control parasitaemia, and discussed this in our Review¹.

Rather than focus on GPI-based vaccines, our Review¹ of malarial pathogenesis builds on and extends previous discussions of the topic by suggesting a role for complex cellular infiltrates, rather than simple systemic cytokine excess, as the fundamental immunological event in disease pathogenesis. Sequestered parasites secrete products (including toxins) that initiate cellular recruitment and activation, contributing to cerebral and placental malaria, and to anaemia. These processes are subject to complex homeostatic regulation. We discussed a range of parasite molecules with potential toxic, inflammatory and counter-regulatory activity (see TABLE 2 of our Review¹). Engagement of host pattern-recognition receptors by these molecules induces expression of numerous host genes, of which tumour-necrosis factor (TNF) is but one. TNF might therefore simply be a marker of pattern-recognition receptor activation, which is historically useful as a read-out for in vitro assays, but might not have a key causal role in cerebral malaria and anaemia. Indeed, TNF might be beneficial during malaria, and toxins clearly induce pathogenic gene expression independently of this molecule². Our view therefore differs substantially from the systemic inflammation model that was originally proposed by Maegraith³, rather than by Clark⁴ as suggested by Butcher and Taverne.

Although Butcher and Taverne claim that GPI was described as a malaria toxin by Bate et al.⁵, this paper and subsequent studies from the same group⁶ specifically state that GPI is not the malaria toxin (pages 133 and 1889, respectively). In contrast to others^7,8,9,10, these researchers did not detect TNF induced by malaria GPI, did not detect competitive inhibition by monosaccharide fragments, and found that their treatments aimed at achieving nitrous acid deamination and deglycosylation had no effect on toxic activity. They also deduced that malaria GPIs were not toxic because they found that another molecule containing a GPI anchor from a different species of parasite, Leishmania spp., was inactive, despite structure–activity analyses showing that the activities of various parasite GPIs are highly specific^8,10. Furthermore, this group proposed that phosphatidylinositol was the malaria toxin and not GPI⁶. Phosphatidylinositol, which is not a malaria toxin⁶, and GPI, which seems to be a malaria toxin, have completely different biology, therefore regarding these two as interchangeable is not correct.

Oversimplified models of malarial pathogenesis can indeed be dangerous. Georges Grau and colleagues showed that intervening against TNF could exacerbate infectious disease¹¹, and this was later confirmed in the study cited by Butcher and Taverne¹². Despite this, it was considered that there was sufficient evidence in the literature to support Clark's hypothesis⁴ that the pathology of malaria was a result of high levels of TNF, and so patients with cerebral malaria were treated with TNF?specific antibodies. However, these provided no protection and, in fact, markedly worsened neurological sequelae¹³. Therefore, rather than being a cause of disease⁴, high levels of TNF might be a physiologically appropriate, beneficial response during malaria, and blocking TNF might be an inappropriate intervention. However, intervening with antibodies against host molecules such as TNF is fundamentally different from intervening against microbial products such as GPI and lipopolysaccharide (LPS), as host molecules are not regulated by antibodies arising following natural exposure to the pathogen, whereas pathogen products are. Although Butcher and Taverne suggest that antibodies specific for GPI might promote disease and increase pathogen load, naturally acquired antibodies specific for bacterial toxins, such as LPS, reduce disease severity and risk without increasing bacterial load¹⁴. Antibodies arising following natural exposure to such molecules do not cause Toll-like receptor non-responsiveness or host incapacity. Rather, they seem to be beneficial, mopping up excess toxins and feeding into a regulatory network that dampens the variance and amplitude of the pathogenic pro-inflammatory response. Therefore, acquired immunity to pathogen products has a role in reducing the risk of excessive inflammation, whereas intervening against the host molecule TNF adversely perturbs the homeostatic regulation of immune and physiological responses.

The evidence so far indicates that toxins promote the expression of genes implicated in the pathogenesis of malaria and that toxin-based vaccination (such as with GPI-based vaccines) provides clinical protection in the best available preclinical models of disease¹⁵. The criticism of GPI-based vaccines suggested by Butcher and Taverne is that they might actually promote disease pathogenesis. However, this was not observed in the preclinical testing¹⁵. Their view would also predict that antibodies specific for GPI that arise following natural exposure to the pathogen will be associated with higher parasite densities and increased risk of disease, whereas the view that GPI is a target of the immune response to malaria predicts that these antibodies will be associated with reduced risk of disease¹⁶. So far, these issues have only been addressed in cross-sectional or case-control studies, using serological assays that require considerable further validation¹⁶. On balance, most^17,18,19, but not all²⁰, data indicate an association of GPI-specific antibodies with clinical protection. Nonetheless, longitudinal prospective cohort studies are by far the best epidemiological tool for assessing the relationship of immunological variables with disease risk. The way forward therefore is not to block the development of these and other candidates based on oversimplified models of malaria pathogenesis. Instead, the effects of vaccination should be addressed in further appropriate pre-clinical studies, and disease risk in relation to GPI-specific immunity arising following natural exposure to the pathogen should be assessed using well-validated assays in longitudinal prospective cohorts. Such studies are underway.