Results
Association of Antibody Titres and Breadth of the Humoral Response With Age and Parasitaemia
A longitudinal study was performed in 99 children during one malaria transmission season in Samako, a village in a hyperendemic area of Mali. The frequency of P. falciparum infections was highest during the July cross-sectional visit at the beginning of transmission season, with a frequency of slide positive children of 27.6% (27/98), and a proportion of parasite-positive children by PCR of 78.6% (77/98) (Table 1). Both the demographic and parasitological parameters of this longitudinal cohort were highly similar to the original total cohort of 171 children, from which this longitudinal cohort was selected based on attendance of all four cross-sectional visits (Additional file 3 http://www.malariajournal.com/content/14/1/56/additional). The number of clinical malaria episodes recorded during longitudinal follow-up of this cohort from July to Dec 2012 peaked in October (July n = 1, Aug n = 8, Sept n = 9, Oct n = 19, Nov n = 17, Dec n = 4).
Antibody responses against five P. falciparum malaria antigens were assessed in the cohort at enrolment, the beginning of the transmission season (baseline), during and after the transmission season. At all time points there was a positive correlation between age and antibody titres for AMA-1, MSP-3, CSP and GLURP-R0, while responses to MSP-119 were not associated with age (Additional file 4 http://www.malariajournal.com/content/14/1/56/additional). Stratification of children into three age categories of two to five years, six to nine years and ten to 15 years revealed that the main increase in AMA-1 titres occurred before the age of six, while antibody levels for MSP-3, CSP and GLURP-R0 rose in a more continuous manner, which was evident both at baseline (Figure 2) and at the peak and end of the transmission season (Additional file 5 and Additional file 6 http://www.malariajournal.com/content/14/1/56/additional).
(Enlarge Image)
Figure 2.
Humoral responses by age at the beginning of the transmission season. Antibody reactivity against P. falciparum antigens was assessed in samples from (n = 99) children collected at baseline. A pool of sera from 100 hyperimmune Tanzanians (HIT) was used as a standard positive control. Reactivity for each antigen in undiluted HIT serum was set at 100 arbitrary units (AU). Humoral reactivity was assessed against (A) AMA-1, (B) MSP-119, (C) MSP-3, (D) CSP and (E) GLURP-R0. Children were divided into three different age groups: two to five years (n = 36), six to nine years (n = 27) and ten to 14 years (n = 36). Responses between the three age groups were compared using Kruskal-Wallis with Dunn's multiple comparison post-test. *p < 0.05; **p < 0.01; ***p < 0.001. Scatter plots show individual data points, horizontal lines indicate the median of the group and error bars the interquartile range (IQR).
The next question was whether the breadth of the response also increased with age. Indeed, at all time points during follow-up there was a significant correlation between age and the number of antigens to which children showed an antibody level that reached at least 10% (>10 AU) of the hyperimmune reference serum (Additional file 4 http://www.malariajournal.com/content/14/1/56/additional). Stratification into age groups showed that this increase occurred again before the age of six, with no difference between the older age groups (Figure 3A). Early in the transmission season (July 2012) 47.2% (17/36) of the children in the youngest age group had no high reactivity to any of the antigens, which decreased to 29.6% (8/27) in six to nine years old and 16.7% (6/36) in ten to 14 years old (p for trend = 0.007). While only 13.9% (5/36) of the two to five years old had high levels of antibodies against three or more antigens, this proportion was more than twice as high in six to nine year old children (29.6%, 8/27) and ten to 14 years old children (30.6%, 11/36); Figure 3B) (p for trend = 0.10). The breadth of the humoral response was significantly associated with parasitaemia at time of sampling (Figure 3C): 57.1% (12/21) of children with no detectable parasites recognized not a single antigen strongly, while this was only true for 28.0 (14/50) and 14.8% (4/27) of children with sub-microscopic or microscopic parasitaemia, respectively (p for trend = 0.007), after adjustment for age. While there was no significant difference in microscopically detectable parasitaemia between the different age categories (p = 0.79), PCR detectable parasitaemia, regardless of thick smear positivity, however, increased with age (p = 0.04; Figure 3D).
(Enlarge Image)
Figure 3.
Relationship of multiple antigen recognition with age and parasitaemia early in the transmission season. The humoral response was analysed according to the number of antigens strongly recognized by each individual (n = 99) at the beginning of transmission season (July 2012). An arbitrary cut-off of 10 AU was used for each antigen and subjects were stratified by age. The three age groups: two to five years (n = 36), six to nine years (n = 27) and ten to 14 years (n = 36) were compared using Kruskal-Wallis with Dunn's multiple comparison post-test. *p < 0.05; **p < 0.01; ***p < 0.001. (A) Scatter plots show individual data points, horizontal lines indicate the median of the group. (B) For each age group, the percentage of children strongly recognizing 0 or more antigens at baseline is shown and was analysed as count data by Poisson regression (overall association between age group and number of antigens recognized: p = 0.001). (C) Parasitaemia was assessed at the beginning of the transmission season (July 2012) by thick smear (TS) and PCR in 98 children (one TS negative child was excluded from analysis since no filter paper was available for PCR analysis). The proportion of children recognizing different numbers of antigens strongly (>10 AU) was stratified based on parasitaemia at time of sampling - not detectable (TS-PCR-; n = 21), sub-microscopic (TS-PCR+; n = 50) or microscopic (TS + PCR+; n = 27) and analysed as count data by Poisson regression (overall association between parasite status and the number of antigens recognized, p = 0.002). (D) The proportion of children with either no detectable (TS-PCR-; white), sub-microscopic (TS-PCR+; grey) or microscopic parasitaemia (TS + PCR+; black) in the three different age categories was analysed by logistical regression for the different diagnostics separately (p = 0.47 for TS+, p = 0.042 for total PCR+).
Dependency of Boosting of Humoral Responses on Early Season Antibody Levels
The breadth of the antibody response, i.e., the number of antigens strongly recognized (>10 AU) per individual, was stable over the time of follow-up, as evidenced by the strong correlation between this readout at enrolment (December 2011) and the beginning (July 2012; Spearman r = 0.59, p < 0.0001), peak (September 2012; r = 0.70, p < 0.0001) and end of the transmission season (February 2013; r = 0.67, p < 0.0001). Next it was investigated whether exposure during the transmission season altered antibody levels for the individual antigens. During the seven months of follow-up, 91/99 children became parasite-positive by either PCR or thick smear during one or more visits. Of these, 47 children remain asymptomatic while 44 experienced one or more episodes of clinical malaria (Figure 1).
Comparing antibody responses in all n = 91 malaria exposed children, there were no statistically significant difference for any of the antigens were observed between the beginning and end of the transmission season, (median AU [July 2012, February 2013]: AMA-1 [7.8, 7.2] p = 0.058; MSP-119 [2.37, 2.51] p = 0.46; MSP-3 [11.6, 12.0] p = 0.27; CSP [6.14, 4.66] p = 0.18; GLURP-R0 [3.40, 3.09] p = 0.80). Since boosting of antibody responses may depend on the strength of the pre-existing response, all exposed children (n = 75) were stratified into low (<1 AU), intermediate (1–10 AU) and high responders (>10 AU) for each antigen. Low early season-responders for AMA-1, MSP-119 and GLURP-R0 showed higher antibody levels after the transmission season, while these titers remained largely unchanged in intermediate early season-responders (Table 2). For MSP-3, there was only a single low early season-responder. For this antigen, intermediate early season-responders showed boosted antibody levels after the transmission season. The opposite was observed for children showing high early season antibody responses: for these children, post-transmission season antibody levels for MSP-119, MSP-3 and CSP were significantly lower than their titers at the beginning of the season (Table 2). In contrast, boosting or waning of antibody titers was not depend on age, since without stratification by early season antibody titers, there was no significant difference between early and post-season antibody titers in any of the different age categories (Additional file 7 http://www.malariajournal.com/content/14/1/56/additional).
Preferential boosting in the group of children with low antibodies at the beginning of the season could be due to the particularly high frequency of clinical malaria episodes during the transmission season in this group, compared to the intermediate or high early season-responders. This was indeed observed regardless of which antigen was used for stratification (Additional file 8 http://www.malariajournal.com/content/14/1/56/additional). One possible explanation for lower antibody levels after the transmission season compared to baseline, as found in the high early season-responder group, might be temporarily elevated antibody levels at baseline due to an ongoing malaria infection. Indeed, there was a trend that the frequency of children with microscopic parasitaemia was highest in the early season high responder group and lowest in the early season low responder group for GLURP (p = 0.02), AMA-1 (p = 0.06) and MSP-1 (p = 0.11) (Additional file 9 http://www.malariajournal.com/content/14/1/56/additional). For AMA-1 (p = 0.04) and MSP-3 (p = 0.03), this distinction was even found for the total proportion of children with any (either sub-microscopic or microscopic) parasitaemia, despite the fact that the majority of children (77/90) included in this longitudinal analysis was PCR positive at baseline.
Association of Antibody Responses With Clinical Protection
Because the youngest children showed overall weaker antibody responses and recognized a smaller number of antigens strongly than older children, it was next verified whether age might be a determinant for protection from clinical malaria during the transmission season. Overall, children remaining asymptomatic were older than those experiencing clinical malaria episodes, although this difference was not statistically significant (Figure 4A; p = 0.09). When age was dichotomized, children two to five years of age had a higher proportion of clinical episodes than the two older age groups (p = 0.049; Figure 4B). Therefore, all the following statistical analyses were adjusted for age.
(Enlarge Image)
Figure 4.
Relationship between age and clinical protection. (A) The age of all children who developed parasitaemia by PCR or thick smear during follow-up (total n = 91) in the asymptomatic (n = 47) and symptomatic groups (n = 44) is plotted. Scatter plots show individual data points, horizontal lines indicate the median of the group and error bars the interquartile range (IQR). The two groups were compared using Mann–Whitney U test. (B) The proportion of children experiencing 0, 1, 2 or 3 clinical episodes was calculated for the three different age groups (two to five years, six to nine years and ten to 14 years) and analysed as count data by Poisson analysis.
Compared to children developing clinical malaria episodes, asymptomatic children showed significantly higher levels of antibodies against AMA-1, and non-significantly higher antibody responses for MSP-3 and GLURP-R0, at the start of the season (Figures 5A, C and E). These differences in antibody levels found early in the transmission season were also observed at enrolment and during longitudinal follow-up for AMA-1, MSP-3 and GLURP-R0 (Additional file 10A, C and E; Additional file 11A and C; Additional file 12A and E http://www.malariajournal.com/content/14/1/56/additional). However, none of these differences were statistically significant when adjusted for age. The levels of anti-MSP-119 and anti-CSP antibodies were comparable between children remaining asymptomatic and symptomatic at enrollment (Additional file 10 http://www.malariajournal.com/content/14/1/56/additional) and during the entire follow-up (Additional files 11B and 12B http://www.malariajournal.com/content/14/1/56/additional). When antibody responses were dichotomized as high responders (>10 AU) and low responders, only high antibody responses to AMA-1 (AU > 10) in July 2012 were associated with a lower risk of clinical malaria episodes in the subsequent season (OR 0.37, 95% CI 0.15-0.91, p = 0.03). No such association was observed MSP-1 (OR 0.92, 95% CI 0.35-2.37, p = 0.86), MSP-3 (OR 0.60, 95% CI 0.25-1.40, p = 0.24), GLURP (OR 0.62, 95% CI 0.21-1.80, p = 0.38) or CSP (OR 1.00, 95% CI 0.39-2.57, p = 0.99).
(Enlarge Image)
Figure 5.
Relation between antibodies at the beginning and clinical protection during the transmission season. Humoral responses at the beginning of the transmission season were assessed by ELISA against (A) AMA-1, (B) MSP-119, (C) MSP-3, (D) CSP and (E) GLURP-R0. The n = 91 children who developed parasitaemia by PCR or thick smear during follow-up were divided into two groups: a group consisting of asymptomatic children who experienced no clinical episode of malaria during the transmission season (n = 47), and a group of children who had a symptomatic malaria episode at least once during the transmission season (n = 44). Differences between the two groups were analysed by linear regression of log-transformed (log10) data, adjusting values for age. Age adjusted P values are shown for each plot, with p-values without age adjustment (Mann–Whitney U test) in brackets. Scatter plots show individual data points, horizontal lines indicate the median of the group and error bars the interquartile range (IQR).
While recognition of individual antigen was therefore no good predictor of protection from clinical disease, children remaining asymptomatic recognized a broader repertoire of antigens compared to those that developed clinical disease (Figure 6A (December 2011, p = 0.03); B (July 2012, p = 0.12); C (February 2013, p = 0.008)). Although not always statistically significant, across all time points analysed, a greater proportion of children that became symptomatic during the season lacked high reactivity to any of the antigens compared to asymptomatic children, while strong responses to three or more antigens were found in a greater proportion of asymptomatic children compared to those that were clinically unprotected (Figure 6). Amongst those children with high reactivity for three to five antigens, 95.4% had strong responses for MSP-3 (median across all time points), 81.0% for AMA-1, 72.3% for CSP, 66.3% for GLURP-R0 and 53.8% for MSP-119. Children strongly recognizing three or more antigens has a reduced risk of developing clinical malaria, both when antibody titers were assessed prior to (prospective analysis, December 2011 and July 2012) or after (retrospective analysis, February 2013) the 2012 transmission season, in which clinical malaria was detected (Figure 6D). Again, these data did not all reach significance, but showed a clear trend across time points.
(Enlarge Image)
Figure 6.
Relationship between strong humoral responses to multiple antigens and risk of clinical malaria. Children who developed parasitaemia by PCR or thick smear during follow-up (n = 91) were divided into asymptomatic children who experienced no clinical episode of malaria during the transmission season (n = 47), and children who had a symptomatic malaria episode at least once during the transmission season (n = 44). For both groups, the percentage of children strongly recognizing 0 or more antigens is shown at (A) enrollment, (B) in the beginning and (C) after the transmission season. An arbitrary cut-off of 10 AU was used for each antigen. Horizontal bars indicate the percentage of children in each group recognizing three or more antigens strongly. The number of antigens to which high reactivity (>10 AU) was observed was analysed as count data by Poisson regression. (D) The risk of developing clinical malaria during the transmission season was calculated by logistic regression analysis for children that were high responders (>10A AU) for 1, 2, 3, 4 or 5 antigens simultaneously in reference to those that recognized not a single antigen strongly, with adjustment for age. Analysis was performed either prospectively (using December 2011 or July 2012, i.e. before clinical episodes were recorded, excluding one volunteer that experienced a clinical episode in July from the analysis), or retrospectively (using post-season February 2013 antibody titers). Symbols depict odds ratios, error bars indicate the upper and lower 95% confidence interval.