Severe childhood malaria syndromes defined by plasma proteome profiles

Abstract

Background: Cerebral malaria (CM) and severe malarial anemia (SMA) are the most serious life-threatening clinical syndromes of Plasmodium falciparum infection in childhood. Therefore it is important to understand the pathology underlying the development of CM and SMA, as opposed to uncomplicated malaria (UM). Different host responses to infection are likely to be reflected in plasma proteome-patterns that associate with clinical status and therefore provide indicators of the pathogenesis of these syndromes.

Methods and findings: Plasma and comprehensive clinical data for discovery and validation cohorts were obtained as part of a prospective case-control study of severe childhood malaria at the main tertiary hospital of the city of Ibadan, an urban and densely populated holoendemic malaria area in Nigeria. A total of 946 children participated in this study. Plasma was subjected to high-throughput proteomic profiling. Statistical pattern-recognition methods were used to find proteome-patterns that defined disease groups. Plasma proteome-patterns accurately distinguished children with CM and with SMA from those with UM, and from healthy or severely ill malaria-negative children.

Conclusions: We report that an accurate definition of the major childhood malaria syndromes can be achieved using plasma proteome-patterns. Our proteomic data can be exploited to understand the pathogenesis of the different childhood severe malaria syndromes.

Figure 1. Discriminatory accuracy of proteome profiles across study groups.

CM = Cerebral Malaria; SMA = Severe Malarial Anemia; UM = Uncomplicated Malaria; DC = Disease Controls; CC = Community Controls. Blue bars compare (CM, SMA and UM) with non-parasitemic CC. MP(−ve) = malaria parasite negative; MP(+ve) = malaria parasite positive. Green bar compares DC and CC (both non-parasitemic). Orange bars compare (CM, SMA and UM) with DC (non-parasitemic). Yellow bars compare CM, SMA and UM (all parasitemic) among themselves. In brackets are shown the number of relevant m/z clusters that make up the discriminatory proteome profile. (a.) Accuracy of predictive models built with relevant m/z clusters using the discovery cohort data. Error bars indicate +/− standard deviations obtained by 100 train/test randomizations of the data. (b.) Accuracy of the best predictive model from (a.) when applied to the validation cohort data.

Figure 2. Discriminatory performance of proteome profiles across study groups in ROC space.

CM = Cerebral Malaria; SMA = Severe Malarial Anemia; UM = Uncomplicated Malaria; DC = Disease Controls; CC = Community Controls. (a.) Discriminatory performance in ROC space of predictive models built with relevant m/z clusters using the discovery cohort data. Error bars indicate +/− standard deviations obtained by 100 train/test randomizations of the data. (b.) Discriminatory performance in ROC space of the best predictive model from (a.) when applied to the validation cohort data.

Figure 3. Visualization of community control (CC, non-parasitaemic) children versus other study groups.

Each sphere represents an individual child proteome profile plotted in 3D space defined by the first three principal components. CM = Cerebral Malaria (red); SMA = Severe Malarial Anemia (purple); UM = Uncomplicated Malaria (yellow); DC = Disease Controls (blue); CC = Community Controls (green). (a.) CC vs. CM; (b.) CC vs. SMA; (c.) CC vs UM and (d.) CC vs. DC.

Figure 4. Visualization of (a-c) disease control (DC, non-parasitemic) children versus parasitemic children groups; (d-f) among CM, SMA and UM parasitaemic children groups.

Each sphere represents an individual child proteome profile plotted in 3D space defined by the first three principal components. CM = Cerebral Malaria (red); SMA = Severe Malarial Anemia (purple); UM = Uncomplicated Malaria (yellow); DC = Disease Controls (blue). (a.) DC vs. CM; (b.) DC vs. SMA; (c.) DC vs. UM; (d.) CM vs. SMA; (e.) CM vs. UM and (f.) SMA vs. UM.

Figure 5. Discriminatory accuracy of proteome profiles across six anionic plasma fractions (f1 to f6) among CM, SMA and UM (all parasitemic) groups of children.

CM = Cerebral Malaria; SMA = Severe Malarial Anemia; UM = Uncomplicated Malaria. f1 to f6 represent anionic plasma fractions at pH 9.0 (f1), pH 7.0 (f2), pH 5.0 (f3), pH 4.0 (f4), pH 3.0 (f5) and organic phase (f6). In brackets are shown the number of relevant m/z clusters that make up the discriminatory proteome profile. (a.) CM vs. SMA; (b.) CM vs. UM; (c.) SMA vs. UM.

Figure 6. Proteome-pattern overlap.

The Venn diagram represents the level of overlap between the patterns that individually discriminate CM, SMA, UM and DC from CC. CM = Cerebral Malaria; SMA = Severe Malarial Anemia; UM = Uncomplicated Malaria. f1 to f6 represent anionic plasma fractions at pH 9.0 (f1), pH 7.0 (f2), pH 5.0 (f3), pH 4.0 (f4), pH 3.0 (f5) and organic phase (f6).