Figures and data
Divergent malaria exposure histories in coastal Kenya
Monthly malaria case counts from active surveillance in two adjacent regions of Kilifi County, Kenya, between 1998 and 2017. Junju (red) maintained sustained malaria transmission throughout the study period, while Ngerenya (blue) experienced a rapid collapse in transmission beginning in 2004. Points represent total cases per month; lines show smoothed trends generated using locally weighted regression (loess) in R (span = 0.8).
Baseline characteristics of the longitudinal cohorts from Junju and Ngerenya
AMA1-specific IgG trajectories mirror individual and regional malaria exposure
(A–B) Longitudinal AMA1 IgG levels in individual children measured by protein microarray. Vertical red lines denote confirmed febrile malaria episodes. Panel A shows a Junju child with multiple documented infections; Panel B shows a Ngerenya child who remained malaria-free throughout follow-up. Each blue spot is a single antibody measurement. Each time point was measured in quadruplicate (C) Mean AMA1-specific IgG levels with 95% confidence intervals for all children in the microarray subset plotted by year of sampling. Junju children showed persistently elevated antibodies, while AMA1 antibody levels in Ngerenya declined sharply after 2004.
Malaria exposure is associated with reduced measles-specific antibody levels.
Example plots of temporal changes in measles-specific antibody in vaccinated and unvaccinated children are shown (A) Longitudinal IgG levels in an individual child measured by microarray. The dashed vertical red line indicates the timing of the last dose of the routine measles vaccine. (B) Shows a similar temporal trend for a child with no history of measles vaccination. (C) Longitudinal IgG responses to measles virus by cohort, measured by protein microarray in vaccinated children. Junju children exhibited consistently lower levels of measles-specific antibody than Ngerenya counterparts. The circles indicate means, and the whiskers denote 95% confidence intervals. (D) Measles-specific IgG levels in 3-year-old children from Junju and Ngerenya, measured by ELISA in children with a documented history of measles vaccination, including a negative control (PBS) and pooled adult serum as a positive reference Each dot represents an individual participant.
Early-life malaria exposure is associated with reduced antibody responses to multiple antigens.
(A) Forest plot showing the average age-adjusted difference in log₂ antibody responses between children from Ngerenya and Junju, estimated using mixed-effects models incorporating all available longitudinal measurements. Points represent model estimates and horizontal bars indicate 95% confidence intervals. (B) Summary table of model-derived estimates, including log₂ differences with 95% confidence intervals and false discovery rate (FDR)-adjusted significance across antigens. (C) Cross-sectional comparison of antibody responses at 10 years of age, shown for reference. Boxes indicate interquartile ranges, centre lines denote medians, and whiskers represent 1.5× the interquartile range. Asterisks indicate significance from Wilcoxon rank-sum tests (* P < 0.05, ** P < 0.01, *** P < 0.001, ****. P < 0.0001).
Early-life malaria exposure predicts long-term suppression of antibody responses within the same geographic region
(A) Active malaria surveillance records for children in the Ngerenya cohort. Each row represents an individual child, and each column represents a surveillance timepoint. Dark red boxes indicate one or more confirmed febrile malaria episodes; light grey boxes indicate surveillance visits without malaria detection. Children are grouped by early-life exposure status (top: malaria-naïve; bottom: previously exposed). (B) IgG levels at 10 years of age among Ngerenya children, stratified by early-life malaria exposure. Children with ≥1 confirmed febrile malaria episode during early childhood (n = 20) show significantly lower titres to multiple unrelated pathogens compared to malaria-naïve peers (n = 42). All children lived in the same geographic area and received identical vaccines and follow-up. The black dots are means and error bars are 95% confidence intervals.
Population-level comparison of anthropometric profiles between Junju and Ngerenya.
(A) Age-specific distributions of height-for-age (HAZ), mid-upper arm circumference (MUAC), and weight-for-age (WAZ) among children aged 0–59 months, derived from contemporaneous hospital-based surveillance data. Boxplots show median and interquartile range, with whiskers extending to 1.5× the interquartile range; points represent individual observations. Distributions are shown separately for children from Junju and Ngerenya. (B) Summary of anthropometric indices by age band and location. Values are presented as mean (standard deviation) for MUAC and mean (95% confidence interval) for WAZ and HAZ. (C) Adjusted differences in anthropometric indices between children from Junju and Ngerenya. Points represent model-derived estimates for the difference (Junju − Ngerenya), and horizontal lines indicate 95% confidence intervals. Estimates were obtained from regression models adjusting for age (modelled using splines), calendar year, and concurrent infections (RSV, parainfluenza, influenza A, human metapneumovirus, OC43, and malaria)
Febrile malaria episode burden is inversely associated with heterologous antibody responses.
Scatterplots show the relationship between cumulative febrile malaria episode count and standardised antibody responses (z scores) to eight heterologous antigens. Each point represents an individual observation, and dashed lines indicate fitted linear trends. For most antigens, higher malaria episode burden was associated with lower antibody responses, with no evidence of opposing trends.
A summary of the sampling frame for the study cohorts (Junju and Ngerenya).
Each vertical line represents the longitudinal time series for a single individual, and each dot represents a timepoint where a serum sample was collected.
