Gene expression analyses reveal differences in children's response to malaria according to their age

Abstract

In Bandiagara, Mali, children experience on average two clinical malaria episodes per year. However, even in the same transmission area, the number of uncomplicated symptomatic infections, and their parasitemia, can vary dramatically among children. We simultaneously characterize host and parasite gene expression profiles from 136 Malian children with symptomatic falciparum malaria and examine differences in the relative proportion of immune cells and parasite stages, as well as in gene expression, associated with infection and or patient characteristics. Parasitemia explains much of the variation in host and parasite gene expression, and infections with higher parasitemia display proportionally more neutrophils and fewer T cells, suggesting parasitemia-dependent neutrophil recruitment and/or T cell extravasation to secondary lymphoid organs. The child's age also strongly correlates with variations in gene expression: Plasmodium falciparum genes associated with age suggest that older children carry more male gametocytes, while variations in host gene expression indicate a stronger innate response in younger children and stronger adaptive response in older children. These analyses highlight the variability in host responses and parasite regulation during P. falciparum symptomatic infections and emphasize the importance of considering the children's age when studying and treating malaria infections.

Fig. 1. Percentage of the variance in host and parasite gene expression explained by each variable considered.

Each violin plot shows the percentage of variance (y-axis) explained by each variable (x-axis) for each (A) human gene (n = 9205) and (B) P. falciparum gene (n = 2484 genes) based on the analyses of 136 infections. Each black dot represents one gene. The internal boxplot shows the median as a horizontal bar and the interquartile range of variance for each variable. “Residuals” indicates the percentage of the variance in gene expression not explained by any of the variables considered (i.e., driven by remaining inter-individual differences).

Fig. 2. Host gene expression and parasitemia.

The volcano plots show the association of host gene expression with the log of the parasitemia before (A) and after (B) adjusting for immune cell composition using a quasi-likelihood negative binomial generalized model. Each point represents one gene, displayed according to its p-value (y-axis) and log fold-change (x-axis). Blue and red points represent genes that were significantly more expressed in low and high parasitemia infections, respectively, corrected for multiple testing using false discovery rate (FDR) (FDR = 0.1). Correlation of the proportion of neutrophils (C) or T cells (D) (y-axis), estimated by gene expression deconvolution, with the log of the parasitemia (x-axis) using linear regression (respectively, Pearson’s R² = 0.05, p = 0.006 and Pearson’s R² = 0.09, p = 0.0004). *DEG = differentially expressed gene. Note that since we measured gene expression correlated with parasitemia as a continuous variable, the log fold-change reflects the change in expression of each gene with each unit of parasitemia, which can be smaller than typical log fold-change values that measure differences in expression between two groups. (N = 136 individuals).

Fig. 3. Parasite gene expression and parasitemia.

The volcano plots show parasite gene expression associated with the parasitemia, unadjusted (A) and adjusted (B) for developmental stage composition using a quasi-likelihood negative binomial generalized model. Each point represents one gene, displayed according to its p-value (y-axis) and log fold-change (x-axis). Blue and red points represent genes that were significantly more expressed in low and high parasitemia infections, respectively, correcting for multiple testing using FDR (FDR = 0.1). Correlation of the proportion of rings (C) or trophozoites (D) (y-axis), estimated from gene expression deconvolution, with the log of the parasitemia (x-axis) using linear regression (respectively Pearson’s R² = 0.13, p = 1.74 × 10⁻⁵ and Pearson’s R² = 0.08, p = 0.0009). *DEG = differentially expressed gene. (N = 136 individuals).

Fig. 4. Host gene expression and child’s age.

The volcano plots show the association between each human gene’s expression and the child’s age at the time of the infection, before (A) and after (B) adjusting for differences in immune cell composition using a quasi-likelihood negative binomial generalized model. Each dot represents one gene and is displayed according to its log10 p-value (y-axis) and fold-change (x-axis). Blue and red points represent differentially expressed genes that were more expressed in younger and older children, respectively, corrected for multiple testing using FDR (FDR = 0.1). Correlation of the proportion of neutrophils (C), B cells (D), NK cells (E), and Plasma cells (F) (y-axis), estimated from gene expression deconvolution, with the age of the child in years (x-axis) using linear regression (respectively, Pearson’s R² = 0.06, p = 0.002; Pearson’s R² = 0.13, p = 2.10 × 10⁻⁵; Pearson’s R² = 0.13, p = 1.15 × 10⁻⁵; and Pearson’s R² = 0.06, p = 0.004). Note that different ranges for the y-axis in (C, D, E, F) due to differences in cell proportions. *DEG = differentially expressed gene. (N = 136 individuals).

Fig. 5. Parasite gene expression and child’s age.

The volcano plots show parasite gene expression associated with the child’s age at the time of infection, before (A) and after (B) adjusting for developmental stage composition using a quasi-likelihood negative binomial generalized model. Each point represents one gene, displayed according to its p-value (y-axis) and fold-change (x-axis). Blue and red points represent differentially expressed genes that were more expressed in younger and older children, respectively, corrected for multiple testing using FDR (FDR = 0.1). Correlation of the proportion of male gametocytes (y-axis) inferred from gene expression deconvolution with the participant’s age in years (x-axis) using linear regression (Pearson’s R² = 0.05, p = 0.0093) (C). Note that different ranges for the y-axis in C due to differences in cell proportions. *DEG = differentially expressed gene. (N = 136 individuals).