Anti-Inflammatory Cytokines Predominate in Acute Human Plasmodium knowlesi Infections

Plasmodium knowlesi has entered the human population of Southeast Asia. Naturally acquired knowlesi malaria is newly described with relatively little available data, including data on the host response to infection. Therefore pre-treatment cytokine and chemokine profiles were determined for 94 P. knowlesi, and for comparison, 20, P. vivax and 22 P. falciparum, patients recruited in Malaysian Borneo. Nine, five and one patient with P. knowlesi, P. falciparum and P. vivax respectively had complicated malaria as defined by World Health Organisation. Patients with uncomplicated P. knowlesi had lower levels of the pro-inflammatory cytokines IL-8 and TNFα than those with complicated disease (both p<0.05, Dunn's post test, DPT). The anti-inflammatory cytokines IL-1ra and IL-10 were detected in all patients in the study. IL-1ra, the most abundant cytokine measured, correlated with parasitaemia in P. knowlesi (rs = 0.47, p =  <0.0001), P. vivax (rs = 0.61, p = 0.0042) and P. falciparum (rs  = 0.57,p = 0.0054) malaria. IL-10 correlated with parasitaemia in both P. knowlesi (rs = 0.54, p =  <0.0001) and P. vivax (rs = 0.78, p =  <0.0001) infections. There were between group differences in soluble markers of macrophage activation (MIP-1β and MCP-1). P. knowlesi patients had significantly lower levels of MIP-1β than P. falciparum (DPT, p =  <0.01). Uncomplicated P. knowlesi patients had significantly lower levels of MCP-1 than uncomplicated P. falciparum patients (DPT, p =  <0.001). There was no significant difference between complicated and uncomplicated P. knowlesi infections. MCP-1, MIP-1β, IL-8 and TNFα increased in complicated P. knowlesi but decreased in complicated P. falciparum infections. Descriptions of human knowlesi malaria provide a comparative means to discover mediators of pathophysiology in severe P. knowlesi as well as P. falciparum malaria. Crucially, P. knowlesi may be the disease and experimental primate model for severe malaria.


Introduction
Plasmodium knowlesi infection is a major cause of malaria in humans in Malaysian Borneo [1,2]. Approximately 10% of patients develop severe symptoms when classified by WHO criteria for disease severity and approximately one percent of cases have a fatal outcome [3]. Recent post mortem findings from a fatal P. knowlesi infection show remarkable histological similarities with P. falciparum, suggesting commonality of some pathophysiological processes between severe infections caused by different species of Plasmodium [4]. P. knowlesi and P. falciparum infections can both lead to hepatorenal dysfunction, acute respiratory distress syndrome (ARDS), metabolic disturbance and the accumulation of large numbers of heavily pigmented parasitized red blood cells in brain and other microvasculatures [4,5,6,7]. ARDS, jaundice renal dysfunction and anaemia can also occur in complicated P. vivax malaria [8,9,10,11]. However, despite accumulation of P. knowlesi infected erythrocytes in brain vessels, coma is not a feature of fatal knowlesi infection. The severity of anaemia associated with P. knowlesi may also be less than that seen with P. falciparum and P. vivax infection although this complication has not been studied systematically.
In fatal knowlesi malaria, progression from the onset of symptoms to the development of high parasitiaemia and death can be swift, and may be as short as 3 days [1,4]. After reviewing data from cases of severe and fatal knowlesi malaria, we observed that parasitaemia might be a better marker of disease severity than for other species of human malaria [3]. Sequestration of mature stage parasites from peripheral blood circulation in P. falciparum infections precludes the possibility of determining accurate estimates of asexual stage parasitaemia by microscopy and therefore of using parasitaemia as a reliable marker of disease severity in falciparum malaria [12]. Difficulty in making direct associations between parasitaemia and pathophysiology in falciparum malaria have hampered research on identifying parasite virulence, particularly in relation to estimating thresholds for the development of severe disease [13]. Within the context of host response to acute malaria, quantifiable measures of disease severity, parasitaemia and soluble mediators of immune response were measured in clinically and parasitologically well-characterised P. knowlesi, P. falciparum and P. vivax infections. Although falciparum and vivax patients had mostly imported malaria they were retained in the study to provide contemporaneous and other information for comparison. We test the hypothesis that soluble immune mediators correlate with measures of disease severity at presentation, including circulating parasitaemia, in P. knowlesi, P. vivax and P. falciparum infections. We also comment on the influence of innate immunity on disease outcome. Our results provide new insights into the pathophysiology of this emerging human pathogen, which can be viewed both as a human model of an animal infection as well as the reverse.

Patients
All patients recruited into the study had PCR confirmed single species infections and had not received anti-malarial treatment within the preceding 14 days of admission. Pre-treatment serum and plasma samples were available for 94 P. knowlesi, 20 P. vivax and 22 with P. falciparum patients and were included in the study presented here. Admission clinical and laboratory variables for these patients are summarised in Table S1 as some of these data have been published previously [3].
The number of P. falciparum patients recruited in Malaysian Borneo was small (n = 22), so we assessed how representative the information from this group was for P. falciparum infections by comparing the data with that available from 304 P. falciparum patients admitted to St George's Healthcare NHS Trust. There were no significant differences between patients from St. George's (all available data) and Malaysia for parasitaemia, haemoglobin, total white cell and platelet counts (Table S2).
Eight (8.5%) patients with P. knowlesi had complicated malaria, with one (1%) fatal case that was included in the complicated group for analysis. One patient (5%) with P. vivax malaria had anaemia and 5 (22.7%) with P. falciparum had complications defined as severe by WHO criteria (summarised in Table S3).

Characteristics of host response to infection
The cytokines IL-1a, IL-4, IL-17 and GM-CSF in all types of malaria in this study were not detected or were detected in too few patients to be informative. Table 2 summarises the cytokine and chemokine profiles by species and disease severity. Significant  correlates of immune mediators and parasitaemia for each infecting parasite species are summarised ( Figure 2, and Table  S4). Parasitaemia was associated with a different cytokine profile in each infecting species of Plasmodium. P. knowlesi and vivax malaria patients produced similar associations except that MCP-1 weakly correlated with parasitaemia in P. knowlesi (rs 0.22, p = 0.0318) but strongly in P. vivax infections (rs 0.72, p = 0.0003). By contrast, parasitaemia in falciparum malaria was a poor correlate of immune mediators and, apart from IL-1ra, was different from that observed for knowlesi and vivax malaria patients, including an association with IL-1b that was not seen in infections with the other two species (Figure 2).  Anti-inflammatory cytokines IL-1ra was the most abundant cytokine measured and was associated with complications in P. knowlesi patients (r s 0.34, p = 0.0007) but not in the P. falciparum group (Table 3). IL-1ra in all three species was a correlate of cytokines, including IFNc, and chemokines in P. knowlesi (Table 3). There were significant between group differences for IL-1ra (K-W, p = 0.001), but not IL-10. IL-10 was not associated with complications but was associated with neutrophil and lymphocyte counts in P. knowlesi and P. vivax patients and with other circulating immune mediators in all three species (Table 3). IL-1ra and IL-10 were detected in all patients recruited into the study ( Figure 3).

Macrophage activation
There were significant between group differences for MIP-1b and MCP-1 levels (K-W, p = ,0.0001 for both). P. knowlesi patients had significantly lower levels of MIP-1b (median 110.0; IQR 73.48-155.9 pg/mL) than P. falciparum (median 382.3; IQR 129.5-708.0 pg/mL: DPT, p = ,0.01) but not P. vivax patients (median 144.6 ; IQRI 80.30-838.4 pg/mL). There was no significant difference between the P. vivax and P. falciparum groups. MIP-1b was a strong correlate of TNFa in P. vivax (rs, 0.66) and P. falciparum  (Table 4). Four vivax and two falciparum patients had MCP-1 levels greater than the detection level of the assay.  Immune mediators and complicated disease Changes in immune mediators between patients with uncomplicated and complicated disease in P. knowlesi and P. falciparum infections are summarised in Figure 4. The pro-inflammatory mediators TNFa and IL-8 increased in P. knowlesi patients but decreased in P. falciparum complicated infections (Figure 4, columns 'a' in left and right panels). The anti-inflammatory cytokines were comparable between the two infections ( Figure 4 columns 'b' in both panels). MIP-1b and MCP-1 decreased in patients with complicated P. falciparum but increased in those with complicated P. knowlesi (Figure 4, columns 'c' in both panels). VEGF decreased in P. knowlesi but increased in P. falciparum patients (Figure 4, columns 'd' in both panels).

Discussion
We have noted previously that thrombocytopenia, mild to moderate hyponatremia, deranged liver function jaundice, renal impairment and hypotension are complications of human knowlesi malaria [1,3,4]. We now demonstrate that measures of disease severity at presentation are more often associated with parasitaemia in P. knowlesi than in P. vivax and P. falciparum infections.
Soluble immune mediators associate with parasitaemia in P. knowlesi and P. vivax but less so in P. falciparum malaria. The falciparum and vivax patient groups were equally small and the lack of association, particularly in soluble immune mediators, between parasitaemia in patients with P. falciparum is possibly a reflection of sequestration of all but immature trophozoite infected cells and in consequence the unreliability of estimating parasitaemia from peripheral blood samples [12].
Soluble immune mediators measured pre-treatment were used as surrogate markers of immune activity in our patients. In this study host response profiles of symptomatic patients, in all three species of Plasmodium, were those of anti-inflammatory regulation by IL-1ra and IL-10. IL-1ra regulates the proinflammatory biological properties of IL-1 and was the most abundant circulating cytokine. IL-1ra concentrations also correlate with parasitaemia in all three species. Interestingly IL-1ra and Il-10 serum concentrations are positively associated with percent immature trophozoites in P. knowlesi and P. vivax infections indicating host response to schizont rupture events. It is possible that anti-inflammatory cytokines are released to modulate the expected immune surge during schizont rupture and reinvasion. IL-10 limits the inflammatory responses provoked by many stimuli including normal homeostasis and IL-10 has been variously implicated in malaria pathophysiology [14,15,16,17]. IL-10 was detected in all patient groups in our study. IL-10 was positively associated with parasitaemia in P. knowlesi and P. vivax infections. In both falciparum and knowlesi patients IL-10 levels, while elevated, were not associated with markers of disease severity. Our results contrast with findings from Brazil where parasitaemia was significantly higher in severe compared with mild vivax infections and IL-10 levels significantly lower [18].
Different macrophage subsets in patients, including those secreting IL-10, have been shown to alter malaria pathology [19]. In our study there were differences in both the manifestations of disease severity in P. knowlesi and P. falciparum patients and the markers of macrophage activation MIP-1b (CCL-4) and MCP-1 (CCL2). Both chemokines were lower in P. knowlesi compared with P. falciparum patients. In falciparum patients MIP1b but not MCP-1 was strongly associated with IL-1ra while the opposite was true for vivax and knowlesi patients. Subtle differences in macrophage activity may account for some of the differences in manifestations of disease severity between patients infected with different Plasmodium species, for example hypotension and renal impairment in P. knowlesi and severe anaemia in P. falciparum infections. MCP-1 attracts monocytes, irrespective of subset, therefore MCP-1 (CCL2) activity is not necessarily characterised by a heightened pro-inflammatory IFNc response [17,19]. In support of an anti-inflammatory macrophage sub-set activity in our patients, circulating IFNc was not predominant in P. knowlesi, P. falciparum or P. vivax malaria in this study. MIP1b (CCL-4) is produced by macrophages in response to foreign particle stimulus and can induce IL-1, IL-8 and TNFa. MIP1b was strongly correlated with these cytokines in P. falciparum patients in our study but not with IL-1 in P. vivax and P. knowlesi and weaker association with TNFa in P. knowlesi patients. Interestingly, neither chemokine was associated with markers of disease severity in P. knowlesi but MCP-1 (CCL2) was strongly associated with haemoglobin in P. falciparum. Severe anaemia was a feature of P. falciparum but not P. knowlesi patients. The association between MCP-1 (CCL2) and haemoglobin, one of the few clinical associations found in falciparum patients in this study, adds support to the idea that the relative weight of macrophage subset activation may influence malaria pathophysiology between patients and between infecting Plasmodium species.
Neutrophil counts strongly correlate with P. knowlesi parasitaemia and neutrophils were also increased in our P. vivax and P. falciparum patient groups. Also P. knowlesi patients were more thrombocytopenic than those with P. vivax and P. falciparum malaria. Here we make a case for the interplay of innate immune mediators, particularly platelets and neutrophils, in controlling pre-treatment parasitaemia.
In the non-immune the host-parasite interface early in infection would be expected to involve pattern recognition motifs on merozoite surfaces, infected erythrocyte surfaces and other substances released during asexual life-cycling [20]. In acute malaria, these interactions are initiated and contained within the red blood cell circulatory compartment with additional parasite clearance activity in the spleen and other well protected endothelial surfaces including those serving the liver, lung, brain and bone marrow. Monocytes, macrophages, endothelial cells, neutrophils and platelets are all equipped with pattern recognition receptors, the Toll-like receptors (TLR) [21]. Of these, platelets and neutrophils are the most numerous, mobile and likely to encounter merozoites and parasite-altered erythrocytes early in infection.
Thrombocytopenia is often studied in bacterial sepsis and is associated with features of disseminated intravascular coagulation and a poor prognosis [22]. These features are not consistent with thrombocytopenia in malaria, especially in acute uncomplicated knowlesi malaria where thrombocytopenia is a characteristic. An anti-parasitic associated, rather than coagulopathy -associated, thrombocytopenia would better explain thrombocytopenia in uncomplicated malaria together with the rapid recovery of circulating platelets following treatment [3]. In support of this, platelets with TLR 2,4 and 9 [22] can actively kill P. falciparum (in vitro) and P. chabaudi (in vivo) [23]. During the killing process platelets bind to infected red blood cells and we propose that thrombocytopenia in uncomplicated P. knowlesi infection reflects a killing process beneficial to the host rather than deleterious platelet depletion [24]. Further investigation of the contribution of platelets in controlling parasitaemia and the events leading to a breakdown of this type of control may help to inform the rational development of intervention methods in malaria.
Platelet activation provides the stimulus for leucocytes, including neutrophils, to migrate out of the capillary beds to sites of infection again not particularly applicable in intravascular Plasmodium infections. Similarly a phagocytic function for activated neutrophils, and associated host tissue damage through the respiratory burst, may be expected to produce more harm than good within the intravascular compartment [25]. In our study, neutrophils were strongly associated with P. knowlesi parasitaemia yet there is a lack of evidence for host endothelial cell damage in human cases of knowlesi malaria, even at post mortem [4].
The strong association between knowlesi parasitaemia and neutrophils may be explained by considering the pattern of circulating cytokines. We observed a strong correlation between neutrophils and the anti-inflammatory cytokines IL-1ra and IL-10 with less evidence for the expected IFN c and TNF a proinflammatory response. Rather, an IL-10 driven response, possibly with a monocyte origin. Even so it remained difficult to reconcile an anti-inflammatory, rather than the characteristic pro-inflammatory, role for neutrophils in malaria, or to understand the associated strong IL-1ra signal in our patients. Recently Bazzoni et al, [26] presented an eloquent anti-inflammatory pathway for neutrophil activation based on the inter-relationship between neutrophils, IL-10 and IL-1ra. The IL-10 receptor on neutrophils, IL-10R2, is constitutively expressed but is not responsive to IL-10 without the co-expression of IL-10R1. IL-10R1 co-expression requires up-regulation. Bazzoni et al., use the example of lipopolysaccharide (LPS) binding to TLR-4 for stimulating IL-10R1co-expression. In the presence of both receptors IL-10 induces neutrophils to synthesise the anti-inflammatory IL-1ra cytokine [26]. IL-1ra blocks the potent inflammatory IL-1b response which, if unregulated in infection, would result in host tissue damage [27,28]. An analogous TLR mediated antiinflammatory neutrophil response during the asexual Plasmodium life cycle would prevent rather than augment an overwhelming pro-inflammatory response in the vulnerable circulatory compartment. Anti-inflammatory neutrophil function, regulated by IL-10, would help explain the up-regulation of IL-1ra in our patients and limitation of disease severity, not only in uncomplicated P. knowlesi, but also in uncomplicated P. falciparum and P. vivax malaria.
In addition to contributing to understanding human knowlesi malaria, the human disease, we have demonstrated important similarities and differences between the three virulent agents of malaria in humans. Our study was limited by small numbers of comparator falciparum and vivax patients which may under power some of the between group statistical analyses and by single time point sampling. Furthermore, not all possible immune mediators were measured. However, we provide unequivocal evidence for an association between parasitaemia, markers of disease severity and immune response in P. knowlesi. We hypothesise that platelets and neutrophils are both anti-parasitic and anti-inflammatory in acute uncomplicated P. knowlesi, P. vivax and P. falciparum infections. We also add some support to the idea that activated monocyte and macrophage subsets direct organspecific immuno-pathology and may influence manifestations of severe malaria between infected individuals and infecting Plasmodium species. We propose that P. knowlesi be both the disease and the experimental primate model to discover how best to properly examine and understand these processes with a view to improving the lives of those most vulnerable to malaria.

Patient consent and ethics
Informed written consent was obtained from all patients entered into these studies that were approved by the Malaysian Ministry of Health's Medical Research and Ethics Committee.

Patient recruitment
Adult patients presenting in Kapit district, Malaysian Borneo with fever and suspected of having malaria were recruited into this study between July 2006 and February 2009 and subsequently confirmed by PCR to carry single species Plasmodium infections [3]. Patients with multiple species infections, other co-morbidities or those who were PCR negative, were excluded retrospectively from further study. Patients in this study were reported previously in studies describing clinical features of P. knowlesi infection and response to therapy [3,29]. Only patients with stored pretreatment serum or plasma samples were included in the study presented here. As described, patients were classified as being severe based on WHO criteria for severe falciparum malaria in non-immune individuals [3,30], with the following exception: For the study presented here we did not consider hyperparasitaemia (.100,000 parasites/mL) [3] in the absence of other clinical or laboratory evidence of severe disease as sufficient to classify patients as having complicated disease.
Falciparum and vivax patients recruited in Malaysian Borneo were relatively small, demographically different and with a history of prior exposure to malaria and may not have been comparable to the mostly untraveled P. knowlesi group [3]. For this reason, archival haematology and parasitology results from 304 P. falciparum cases admitted to St George's Healthcare Trust between 1993 and 1999 were compiled for validation purposes. Some data conversion was required. Percentage parasitaemias were converted to parasites /mL by assuming there were 4.5610 6 RBC/mL blood. Parasitaemias recorded as ,0.01% and ,0.0001% were adjusted 450 and 45 parasites /mL respectively.

Cytokine analysis
Pre-treatment serum was collected and transported in liquid nitrogen or on dry ice and stored at 240uC or colder. In four patients, cytokine analysis was carried out on plasma as serum was unavailable. Comparisons of cytokines in other patients' plasma and serum confirmed that only VEGF levels were ,18% lower in plasma than in serum. This was therefore the only correction factor applied to the four sets of results from plasma.

Measurement of human cytokine and chemokine levels
FluorokineH MAP MultiAnalyte profiling ELISA for LUMI-NEX Technology (R&D Systems, Inc.) with Bio-Rad Bio-Plex System (Bio-Rad Laboratories Inc.) output were used according to manufacturer's instructions. The following cytokines and chemokines were measured: IL-1 a, IL-1 b, IL-1ra, IL -2, IL-4, IL-6, IL-8, IL-10, IL-17, GM-CSF, TNF a, Interferon c, CCL4 (MIP-b), CCL2 (MCP-1) and VEGF. Human Fluorokine MAP Multiplex Control Panel A (R&D systems, Inc) containing low, medium and high mixtures of the cytokines and chemokines measured in this study was used to validate the results.

Statistical analysis
Data were entered into Excel and were analysed using STATA software version 8 (Stata, Corporation, USA.) and Prism (GraphPad Inc.). Non-normal data were log transformed and most remained non-normal (Shapiro-Wilks test). Non-parametric tests (Spearman's rank correlation, Kruskal-Wallis with Dunn's post test) were used throughout. For Spearman's rank correlation the correlation coefficient 'r s ' is given. Significant correlations are represented by *(p = 0.01-0.05), **(p = 0.001-0.01) and *** (p = ,0.001). For between group differences only groups with a significant Kruskal-Wallis (K-W) p value (p = ,0.05) were reported. Significant Dunn's post-test (DPT) p values (p = ,0.05) for multiple comparisons are given.

Supporting Information
Table S1 Summary of demographic data, clinical characteristics and laboratory results for patients in the study. * One P. vivax patient had complicated disease. Pre-treatment values and measurements are summarised as median (inter quartile range) except where stated differently. Numbers in square brackets, [ ], = n when different from total number of patients in each groups. (DOC)

Table S4
Correlation between parasitaemia and immune mediators. Spearman's rank correlation 'r s ' value is given; *, ** and *** denote the level of significance of the correlation, '-' denotes inverse associations and ns = no significant correlation. (DOC)