This study was carried out at Kilifi District Hospital (KDH) on the Kenyan coast. Ethical clearance for the study was given by the Kenya Medical Research Institute ethics review board. Forty eight children admitted to the pediatric ward of KDH with a primary diagnosis of malaria, but who did not fulfill the World Health Organization criteria for severe malaria 7, were recruited, if their guardian gave written consent. A venous blood sample was taken from each child at recruitment and, subsequently, at as many of the time points as possible 1, 2, 3, 6, 9, and 12 wks after treatment with sulphadoxine/pyrimethamine (SP). The samples were centrifuged at 700 × g for 5 min to obtain plasma, which was stored at 20°C. The children were examined by a clinician and a thick malaria film prepared during the follow-up visits or any other time during the study when they were unwell. Malaria treatment (SP) was given for parasitaemia in the presence of fever (axillary temperature ≥ 37.5°C). Seven children from whom weeks 1 and 2 samples could not be obtained were considered lost to follow up, so the cohort for analysis comprised 41 children.

IgG1 and IgG3 antibody reactivity to recombinant ectodomain of P. falciparum apical merozoite antigen 1(AMA-1), the 11 kDa carboxyl portion of merozoite surface antigen 1 (MSP-1₁₉), region II of the 175 kDa erythrocytes binding antigen (EBA-175 RII), and two recombinant proteins representing the two major allelic types of MSP-2 was assessed in plasma samples from 41 children (age range = 7 – 107 months, median = 34 months). Levels of IgM reactivity against the two allelic types of MSP-2 were also assessed. The AMA-1 antigen and EBA-175 region II were kind gifts from Sheetij Dutta and Arnoldo Barbosa (WRAIR, Maryland, USA) and have been previously described 89, while MSP-1₁₉ and MSP-2 proteins were provided by Jana McBride and David Cavanagh (University of Edinburgh, UK) and have also been previously described 1011. Plasma was assayed at 1:4000 dilution for antibodies to AMA-1 and at 1:500 for antibodies to the other antigens using ELISA protocols described previously 31112. Briefly, the wells of 96-well plates (Immulon4; Dynatech, Chantilly, VA) were coated overnight at 4°C with either 50 ng (MSP-1₁₉ and MSP-2) or 100 ng (EBA-175 and AMA-1) of antigen in sodium carbonate buffer (pH 9.3), washed, and then blocked with 1% skimmed milk in coating buffer for 5 h and washed again. The wells were subsequently incubated overnight at 4°C with 100 μl of plasma diluted in 1% skimmed milk, followed by a three-hour incubation with 100 μl of horseradish peroxidase-conjugated rabbit antibodies against human antibodies (Dako Ltd, Cambridge, UK) at 1/1000 for anti-human IgG1 and IgG3 or 1/5000 for anti-human IgM, with washing between the incubations. Colour was developed with o-phenylenediamine, the reaction stopped after 15 minutes with 20 μl of 2 M H₂SO₄, and absorbance was measured as optical density (OD) at 492 nm. All samples were simultaneously assayed in duplicate for IgG1 and IgG3 antibodies against a given antigen. Absorbance readings for responses against MSP-1₁₉ and MSP-2 antigens were corrected for responses to the GST fusion carrier by deducting the absorbance of the same plasma assayed against GST alone. The cut-off for absorbance readings for each antigen was set as mean plus three standard deviations of absorbance readings of sera from eight Europeans without any history of malaria infection. In order to convert absorbance to concentration (μg/ml), plates were coated with purified human myeloma proteins IgG1, and IgG3 (at threefold serial dilutions from 10 to 5.1 × 10^-4 μg/ml).

The resulting standard absorbance versus concentration curves were used for interpolation of the absorbances of the plasma samples to estimate concentrations.

To estimate the half-life of antibody responses to the test antigens, response profiles that showed a decline of antibody levels over at least four consecutive sampling points were selected and the decay phase analysed up to the sampling point where the ODs were too low for concentration to be estimated from standard curves, or the decline was obviously interrupted by boosting. After conversion of absorbance to concentration as described above, one-phase exponential decay curves defined by the equation below were fitted to the decay phase of the selected response profiles using Graphpad Prism software ver. 3.2 (San Diego, USA).

[C]_t = [C]₀e^-kt + Plateau

Where [C]₀ and [C]_t are concentrations at time 0 and t respectively, k is a rate constant and plateau is the concentration at the end of the decay and half-life of the decay (t_1/2) is calculated as follows:

ln ⁡ 1 2 [ C ] 0 [ C ] 0 = − kt 1 2   therefore t 1 2 = ln ⁡ 2 k MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=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@5A67@

Association between kinetics half-life of the responses and age of responder or peak level of antibodies at the beginning of the decay was assessed by linear regression using Stata ver. 8 (Stata Corporation, TX, USA), while differences between half-lives of response to the various test antigens was assessed using one-way ANOVA. In addition, ANOVA was used to test if the mean level of responses to the test antigens differed between individuals. The profiles were then grouped together depending on whether or not they had peaked before the child presented to hospital, and the mean half-lives of the groups compared using a student t-test.

There was considerable variation in the children's overall pattern of responses to the five test antigens but for each antigen, the children generally responded in one of three ways. Either, (1) they already had moderate to high levels of antibodies to the antigen by the time they presented to hospital (e.g. IgG1 responses to MSP-1₁₉ in subjects 23, 36, 38, 40, 41,47, 48 and 56, Figure 1a; IgG3 response to MSP-2 in subjects 10, 44, and 56, Figure 2) or, (2) they presented with low levels of antibodies but made sharp responses that peaked one week after the episode and then declined rapidly (IgG1 responses to MSP-2 in the plots in Figure 1b) or, (3) they made only very low responses (many of the responses plotted in Figures 1a and 1b and Figure 2). IgM antibodies to the two MSP-2 antigens was also assayed to see if the children were making primary responses. Only in two subjects did the IgM absorbance rise above the cut off value during the study period: in subject 28 anti-MSP-2 type B IgM peaked at week one with an absorbance of 0.74, which was much lower than that of the strongest IgG responses that the child made (absorbance of 2.42 for IgG1 to MSP-2 type A); child number 44 had anti-MSP-2-type B IgM absorbance of 0.74 at presentation, which was also much lower than the strongest IgG response (absorbance of 2.55 for IgG1 to MSP-2 type B).

After peaking, the responses generally showed a steady decline but eventually some boosting was observed. Boosting of responses to individual antigens was often associated with the appearance of malaria parasites in the IgG3), but not always (in 15 instances for IgG1, and six for IgG3). Only in 10 of the 43 individual antigen-specific boosting instances were IgG1 responses boosted to similar or higher levels than the initial peak (for example, anti-MSP-2 type A response in subject 38, anti-EBA-175 response in subject 47, Figure 1a), and only five of 15 instances did such boosting occur in the case of IgG3 responses (e.g. anti-MSP-2 type B response in subject 47, Figure 2). 28/41 (68%) children became positive for malaria parasites at one or more sampling time points, mainly 5 weeks or longer after the clinical episode.

Sixty nine IgG1 response profiles from 28 children showed monophasic decay over four consecutive time points. Of these, 40 profiles from 23 children had four consecutive sampling points with absorbance within the interpolation range of the standard curves and were used in the half-life estimation. Of the 26 IgG3 response profiles from 15 children that exhibited monophasic decay over four consecutive time points, 16 profiles from nine children had absorbance within interpolation range of the standard curve, and were used for analysis of half-life.

The exponential decay model used had a fit of R² > 0.90 for the profiles analysed. The mean half-lives of the 40 IgG1 responses analysed individually was 9.8 days (95% confidence interval [CI]: 7.6 – 12.0). There was no association between half-life and age (r² = 0.055, P = 0.130), peak level of antibody (r² = 0.003, P = 0.748), or antigen specificity (F = 0.547, P = 0.653). The half-lives of the responses did not differ significantly between different individuals (F = 2.198, P = 0.140; ANOVA on data from five individuals for each of whom the half-lives of responses to three antigens were estimated). The mean half-life of the IgG3 responses analysed was lower (6.1 days, 95% CI: 3.7 – 8.4) than that of IgG1 but not significantly (Mann-Whitney U test, P = 0.424). As with IgG1, there was no association between half-life and age (r² = 0.13, P = 0.210), peak levels of antibody (r² = 0.74, P = 0.273), or antigen specificities (F = 0.448, P = 0.647) (Figure 3). For both isotypes, the small number of data points analysed for each profile did not allow reliable estimation of the confidence intervals of the half-lives within each individual child.

The profiles were grouped depending on whether or not they had peaked by the time the child was presented to hospital and compared their mean half-lives. IgG1 profiles which peaked before presentation to hospital (n = 8) had a longer mean half-life (13.1 days [95% CI: 8.2 – 21.4]) than those that peaked after presentation (n = 32) (8.9 days [95% CI: 5.5–10.8]) but the difference was not significant (P = 0.17). The mean half-life of IgG3 responses that peaked before presentation to hospital (n = 5) was 7.3 days (95% CI: 5.0 9.5), while that of the responses that peak post presentation (n = 10) was 8.1 days (95% CI: 4.8 – 11.5), a small difference that was not significant (P = 0.685)

Data from 11 children where responses to two or more antigens were analysed for half-life was used to determine the correlation between the half-life of responses to different antigens within the same child. A high level of correlation between responses to different antigens by the same child was observed (R² = 0.63, P = 0.004).