Whole blood angiopoietin-1 and -2 levels discriminate cerebral and severe (non-cerebral) malaria from uncomplicated malaria

Individuals (≥ 13 years of age) presenting with falciparum malaria to Hospital for Tropical Disease clinics (Mahidol University, Thailand) were eligible for enrolment into this study. Patients were classified into uncomplicated (UM), cerebral (CM) or severe (non-cerebral) malaria (SM) according to WHO criteria [2]. Whole blood samples (pre-treatment) were collected from all patients for thick and thin blood film preparation, PCR, and complete blood counts. A heparinized aliquot was frozen at -80°C for determination of subsequent angiopoietin levels. An expert microscopist, who was blinded to the results of additional diagnostic testing, examined the blood films. Smears were considered negative if no parasites were seen in two consecutive thick blood films. Parasite density was calculated by thick or thin film determining the number of parasites per 200 white blood cells for thick blood films or per 1,000 red blood cells (RBC) for thin blood films. Baseline white blood cell counts or RBC counts were used to calculate parasitaemia (parasites per μL). Plasmodium falciparum diagnosis was confirmed by PCR as previously described [16]. The institutional review board of the Faculty of Tropical Medicine, Mahidol University approved this study and informed consent was obtained from all patients or their legal guardians prior to specimen collection.

Whole blood concentrations of ANG-1 and ANG-2 were measured by ELISA (R&D Systems, Minneapolis MN) according to the manufacturers' instructions. Concentrations were interpolated from 4-parameter-fit standard curves generated using a standard curve of recombinant human angiopoietin proteins. The lower and upper limits of detection for each assay were as follows: ANG-1 (156.25 - 10,000 pg/mL) and ANG-2 (54.69 - 3,500 pg/mL).

Statistical analysis was performed using SPSS v17.0. Comparative statistics are reported from a non-parametric univariate analysis (Kruskal-Wallis test, followed by Mann-Whitney U test with adjustment for multiple comparisons or Chi-squared test, as appropriate). Multivariate binary logistic regression was used to examine differences in whole blood protein levels between groups (UM, SM, CM), adjusting for differences due to ethnicity, age, parasitaemia and sex. Spearman's correlation was used to examine correlations between angiopoietin levels and parasitaemia or cumulative organ injury scores. Receiver operating characteristic (ROC) curves and area under the ROC curves were generated using SPSS v.17.0. Optimal test thresholds were derived mathematically from the ROC curves. To adjust for multiple comparisons at a family-wise simultaneous error rate of α = 0.05, Bonferonni corrections were applied as appropriate.

193 patients with P. falciparum malaria were enrolled during the study. Patients were classified as having UM, (n = 70), SM (n = 36) or CM (n = 87) [2]. In patients with complicated disease (SM and CM), 22 participants (17.9%) had evidence of acute renal failure, 71 had evidence of liver dysfunction (57.7%), three had algid malaria with a systolic blood pressure less than 80 mmHg (2.4%), 34 were anaemic and required a blood transfusion (27.6%) and 87 were in a coma (70.7%). Of study participants with severe malaria, 46 had a single organ system involved, 41 had two organ systems affected, 22 had three and five patients had involvement of four organ systems. There was a single fatality in the study with the study participant having cerebral malaria and algid malaria. Demographic and baseline clinical characteristics of the study participants are shown in Table 1.

Whole blood ANG-2 levels were significantly higher in individuals with SM or CM compared to individuals with UM (Figure 1A; p < 0.001). Furthermore, ANG-1 levels were significantly lower in those with SM compared to individuals with either UM or CM, and in individuals with CM compared to UM (Figure 1B; p < 0.001). Since ANG-2 and ANG-1 have divergent roles in vascular activation, the ratio of ANG-2:ANG-1 was used as an additional measure for each patient. This ratio was significantly higher in patients with SM or CM compared to those with UM (Figure 1C; p < 0.001).

ROC curves were plotted for ANG-2, ANG-1, and the ratio of ANG-2:ANG-1 to assess the ability of each biomarker to discriminate between UM, SM and CM. Comparing individuals with UM vs. individuals with SM, all markers have an area under the curve (AUC) that differs significantly from that of a chance result (AUC: 0.5) (Figure 1D: AUC, p-value: ANG-2, 0.763, p < 0.001; ANG-1, 0.884, p < 0.001; Ratio, 0.857, p < 0.001). Similarly, all markers discriminated between UM and CM (Figure 1E: AUC, p-value: ANG-2, 0.772, p < 0.001; ANG-1, 0.778, p < 0.001; Ratio, 0.820, p < 0.001). Finally, ANG-1 but not ANG-2 or the ANG-2:ANG-1 ratio, was able to discriminate between SM and CM (Figure 1F: AUC, p-value: ANG-1; 0.735, p < 0.001: ANG-2; 0.527, p = 0.663: Ratio; 0.599, p = 0.084) with ANG-1 levels being lower in SM than CM. In this population, ANG-1 discriminated between UM and severe malarial syndromes better than either ANG-2 or the ratio of ANG-2:ANG-1, and was able to distinguish SM from CM.

The sensitivity, specificity, and positive and negative likelihood ratios (LR(±)) were determined for ANG-1, ANG-2, and the ratio of ANG-2:ANG-1. Based on sensitivity and specificity, ANG-1 was best able to discriminate between UM and SM (cut-off (12.38 ng/mL): sensitivity: 0.861, specificity: 0.857, LR(+): 6.028, LR(-): 0.162)); whereas ANG-2 was better at discriminating between UM and CM (cut-off (1.33 ng/mL): sensitivity: 0.759, specificity: 0.771, LR(+): 3.319, LR(-): 0.313)). The ratio of ANG-2:ANG-1 was best at discriminating between complicated disease of mixed phenotype (SM+CM) (cut-off 0.08): sensitivity: 0.789, specificity: 0.829, LR(+): 4.60, LR(-): 0.264)) (Table 2). Collectively, our data suggest that the ratio of ANG-2:ANG-1 may be the best predictor of patients with uncomplicated disease versus patients with complicated (severe or cerebral) disease, whereas ANG-1 may have utility in differentiating between cerebral malaria vs. severe (non-cerebral) disease.

In order to examine whether endothelial activation was independent of parasite burden, levels of ANG-1 and ANG-2 were correlated with parasitaemia. Overall, there was a positive correlation between ANG-2 and parasitaemia (Spearman's rho 0.339, p < 0.0001) and a negative correlation between ANG-1 and parasitaemia (Spearman's rho -0.446, p < 0.0001). Furthermore, ANG-2 and ANG-1 were negatively correlated with one another (Spearman's rho -.320, p < 0.0001). Multivariate logistic regression analysis revealed that the observed differences in ANG levels between groups remained even after adjusting for differences due to ethnicity, age, gender, and parasitaemia. There remained a significant increase in ANG-2 (and the ratio of ANG-2:ANG-1) in SM vs. UM, and CM vs. UM, with no significant difference between SM and CM. For ANG-1, there was a significant decrease in SM vs. UM, CM vs. UM, and SM vs. CM (p < 0.0001).

To address whether ANG levels are a measure of disease severity, an organ injury score ranging from 0-5 was assigned to each participant based on the number of organ systems involved. A value of 1 was added to the organ injury score for each participant based on the presence of each of the following: acute renal failure requiring haemodialysis, clinically apparent jaundice, severe anaemia requiring a blood transfusion, circulatory collapse with a systolic pressure <80 mmHg, and coma. ANG-2 and the ratio of ANG-2: ANG-1 were positively correlated with cumulative organ injury (Spearman's rho, p-value: ANG-2; 0.378, p < 0.0001: ANG-2:ANG-1 ratio; 0.288, p = 0.001). Interestingly, ANG-1 also trended towards a positive correlation (Spearman's rho 0.166, p = 0.071) despite its ability to discriminate between individuals with cerebral vs. severe non-cerebral malaria. To investigate this further, a post-hoc analysis was performed to examine how ANG-1 correlated with the organ systems included in the analysis. There were no associations observed between ANG-1 levels in participants with hepatic dysfunction, severe anaemia or circulatory collapse, although the latter was underpowered. However, ANG-1 was positively associated with renal dysfunction (Spearman's rho, p-value: 0.243, p = 0.007) and coma (Spearman's rho, p-value: 0.370, p < 0.0001).