Serological signatures of declining exposure following intensification of integrated malaria control in two rural Senegalese communities

Recent control scale-up has reduced malaria in many areas but new tools are needed to monitor further progress, including indicators of decreasing exposure to parasite infection. Although serology is considered a promising approach in this regard, the serological impact of control interventions has been so far studied using indirect quantification of exposure. Cohort surveys concomitantly recording entomological and malariometric indices have been conducted in two Senegalese settings where supervised control intensification implemented in 2006 shifted malaria from historically holoendemic in Dielmo and mesoendemic in Ndiop to hypoendemic in both settings by 2013. We analyse here serological signatures of declining transmission using archived blood samples. Responses against ten pre-erythrocytic and erythrocytic antigens from Plasmodium falciparum and P. malariae alongside an Anopheles gambiae salivary gland antigen were analysed. Cross-sectional surveys conducted before (2002) and after (2013) control intensification showed a major impact of control intensification in both settings. The age-associated prevalence, magnitude and breadth of the IgG responses to all antigens were village-specific in 2002. In 2013, remarkably similar patterns were observed in both villages, with marginal responses against all parasite antigens in the 0-5y children and reduced responses in all previously seropositive age groups. Waning of humoral responses of individuals who were immune at the time of control intensification was studied from 2006 to 2013 using yearly samplings. Longitudinal data were analysed using the Cochran-Armittage trend test and an age-related reversible catalytic conversion model. This showed that the antigen-specific antibody declines were more rapid in older children than adults. There was a strong association of antibody decline with the declining entomological inoculation rate. We thus identified serological markers of declining exposure to malaria parasites that should help future monitoring of progress towards malaria elimination.

To gain insight into the consequences of control interventions and transmission decline on the antibody responses, we analyse here two communities living in a rural area of Senegal under the same climatic conditions albeit distinct historical malaria endemicity [42][43][44][45][46]. In both settings, P. falciparum dominates, but P. malariae and P. ovale are endemic as well [42,47,48]. High coverage, supervised control interventions have been introduced simultaneously in both communities. In 2006, an artemisinin-based combination therapy (ACT) was introduced as first-line treatment. Long-lasting insecticide-treated nets (LLINs) were provided to each household in July 2008 and renewed in 2011 [46]. The incidence of malaria attacks decreased after 2006 and dropped to low levels after 2008 [46,47,49]. Some rebound was observed in 2011, the highest incidence being in the older children and young adults [42,46]. Importantly, parasite rates dramatically declined a few months after LLIN deployment, and transmission of parasites dropped markedly after LLIN renewal in 2011 [46,48,49].
We profile here the IgG responses of archived blood samples using a multistage, multispecies approach. We first analyse cross-sectional blood samplings made at the end of the dry season in the years 2002 and 2013and monitor responses to a panel of antigens used in previous studies as inducing high seroprevalence in malaria-exposed communities [10,14,34]. We then analyse on a yearly basis the waning of responses to three antigens during the years 2006-2013 in previously immune age groups. This data shows that control intensification induced major changes in the age-stratified antibody profiles, reflecting age-dependent seroreversion of previously immune age groups and minimal responses in young children. The similar response profiles of both communities in 2013 provide serological signatures of efficacious transmission recombinant MSP1p19. Procedures were performed using sera diluted 1/200 as described [5,15,16]. A pool of sera from immune adults from Dielmo and a pool of European and African non-immune sera were included in each assay as positive and negative controls, respectively. IgG Levels were expressed as OD ratio = OD sample / mean OD naive pool . Sera showing an OD ratio >2, corresponding to the mean OD naive pool + 3 SD were considered positive.

Bead-based multiplex assay
A multiplex assay was used to monitor the response the panel of individual antigens listed in Table 2. Covalent coupling of antigens to carboxylated magnetic Luminex beads and the custom magnetic bead-based Luminex multiplex assay including the antigens listed in Table 2 were done as described [10,14,16,55]. IgG levels were expressed as Mean Fluorescence Intensity (MFI). The background signal against BSA was negligible (around 50-80 for all antigens) and was subtracted from all MFI values (net value). The positivity cut-off was set as above the net MFI + 3 SD of naïve controls.

Statistical analysis
Categorical variables were compared using the Fisher exact test, continuous variables of antibody responses were analysed using the Kruskal Wallis and the Spearman rank correlation test for non-normally distributed data. Cochrane-t linear and logistic regressions adjusted for age groups or age were used for comparison of categorical and continuous variables. P values <0.05 Asymptomatic carriage (presence of circulating blood stages detected by microscopy in healthy subjects) was detected by examining Giemsa-stained blood smears by microscopy, the microscopic detection level being in our hands 2 parasites/uLblood [48] and in 2013 by nested PCR [50].
were considered significant. Longitudinal data were analysed using the Cochran-Armittage trend test. Statistical analyses were performed with R and Statview 5.0 1 software. Seroconversion (SCR) and seroreversion (SRR) rates were calculated using an age-specific reversible catalytic conversion model [12]. The model was constructed based on the longitudinal data stratified into pools of 2-year for a given period and for a given antigen.

Results
In 2002, malaria was typically holoendemic in Dielmo [46] and mesoendemic in Ndiop [42,44,45]. Between 2002 and 2013, malaria case incidence decreased by 14.7-fold in Dielmo and 6.3-fold in Ndiop (Table 1) (Table 1). For expression profiles and protein characteristics, see www.plasmodb.org c The LSA1 41 peptide is derived from the well-conserved central repeat region of LSA1 (Liver Stage antigen 1) encoded by gene PF3D7_1036400. The SALSA peptide is derived from the well-conserved region encompassing residues 76 to 101 of the merozoite surface protein 4 encoded by gene

Decreased seroprevalence in all age groups in 2013
We first analysed the cross-sectional blood samplings made in the years 2002 and 2013.A multiplex bead-based approach was used to monitor the responses against individual antigens derived from erythrocytic and exoerythrocytic P. falciparum stages, P. malariae exoerythrocytic stages as well as an A. gambiae salivary gland antigen ( Table 2). This panel included peptides and recombinant antigens known to elicit high seroprevalence in these settings [10,15,16,53]. In parallel, we monitored the responses against the complex pool of erythrocyte stage antigens included in SE using ELISA also previously used in these communities [5,52,[56][57][58] and providing interesting information in a previous study of immunity decay in Dielmo and Ndiop [5].
From 2002 to 2013, the community seroprevalence decreased in both villages by 26% to 66% depending on the antigen, and this was significant for all antigens (Table 4). Between-village comparison showed higher community seroprevalence for SE, PfCSP, LSA1 41 Table). Seroprevalence against all antigens except gSG6 was very low (0 to <20%) in the 0-6y children in 2013. Seroprevalence against PfCSP, SALSA, GLURP, AMA1, PF13 and PmCSP was much lower in 2013 than in 2002 in the 7-14y old children. The age-associated profiles, which differed between the villages in 2002 for SE, LSA1 41 , PF13 and PmCSP (Chi-square test P<0.01 for each comparison), were similar in 2013, except for PmCSP.  Table). In both villages, age was weakly correlated (correlation coefficient R = 0.30, P<10 −2 ) in 2002 with antibody levels against SE, MSP1p19, GLURP, AMA1, PfCSP as well as against LSA1 41 and PmCSP in Ndiop. In2013, a stronger correlation with age was observed for all responses except against gSG6 (R = 0.55 to 0.62, depending on the antigen P<10 −2 ). The 2013 profiles were characterised by marginal IgG levels to all parasite antigens in the <6y children, low in the 7-14y children, higher in the 15-29y individuals and apart from PF13, further increased in the oldest group. Specific profiles were observed for gSG6. In Dielmo, a wide range of IgG levels was observed in the 0-6y olds and a low response in the older age groups. In Ndiop, the IgG levels against gSG6 tended to decrease with age and to be lower in 2013 compared to 2002.

Longitudinal analysis of antibody decay in previously immune age groups
To investigate how the changes of transmission after control intensification affected the acquired antibody responses, we focused on the period from 2006 to 2013 in Dielmo. The EIR declined substantially after LLINs deployment in 2008, rebounded in 2010 and declined to low levels after new LLINs were distributed in 2011 (Fig 3). Samples from age groups that were immune in 2006, namely 17 children aged 7-14y and 47 adults aged 15-63.5 y in 2006 were analysed on a yearly basis. IgG responses against LSA1 41 , MSP1p19 as well as SE were studied in parallel assays by ELISA. ELISA was used here to homogenise the experimental procedures in this study and to provide a basis of comparison with published studies [4, 5, 9, 12, 15, 16, 19, 26-30, 32, 36, 38, 39, 52, 56, 58].
Antibody levels against the three antigens fluctuated with transmission and overall declined with time (Kruskal Wallis test within each age group, P<10 −3 ) (Fig 3A). The magnitude of the responses to the three antigens decreased in both age groups after each LLIN implementation ( Fig 3B). There was a significant and strong (R>0.9, P<10 −3 ) correlation between the EIR values and the magnitude of the response (i.e. the geometric mean of the OD ratios) against the three antigens in both age groups evidenced by polynomial regression analysis (Fig 4A). The decreases were antigen-and age-specific, more rapid and larger in children than in adults for the three antigens. The decrease of antibody levels over the 8-year period was confirmed using the Cochran-Armittage trend test, which showed a general significant decreasing trend (P<10 −4 ). However, when restricted to adults, the P values were lower for MSP1p19 (P = 0.015) and met borderline significance for SE (P = 0.055). These data showed different kinetics of antibody decline in previously immune children and adults depending on the antigen. The best association with EIR was observed for the anti-MSP1p19 response in children (R = 0.99, polynomial regression, P<10 −3 ).
Mathematical modelling of the seroprevalence data using an age-specific reversible catalytic conversion model [12] using a 2-year stratification combining both age groups indicated antigen-specific and period-specific seroconversion (SCR) and seroreversion (SRR) rates (Fig 4B). The individual SCRs and SRRs were unrelated to the mean EIR of the corresponding 2y period but interestingly, the SRR/SCR ratio for LSA1 41 and for MSP1p19 increased as the mean EIR decreased.  Antibody levels against the ten antigens observed in four age groups <7y (white), 7-14 y (light grey), 15-29 y (dark grey) and !30 y (black) in Dielmo and Ndiop are shown as box-and-whisker plots. The box shows the first to the third quartile range, the band inside represents the median, the ends of the whiskers represent the 95% percentiles. Asterisks above the

Discussion
The Dielmo and Ndiop cohort surveys, which uninterruptedly recorded entomological and malariometric indices using a common methodology for more than twenty years, allow informative comparisons of the effect of control intensification on the immune responses at the community level. Seroprevalence, magnitude and breadth of the responses had decreased in 2013 in all age groups and the age-associated profiles were similar in both settings for many antigens. Thus, the humoral profiling of the communities in 2013 reflected the significant progress in malaria control in the recent years resulting in low incidence of cases, uncommon asymptomatic carriage and low transmission level [5,44,46,49].
The low seroprevalence and marginal levels of antibodies to parasite antigens, including SE, in the youngest age group in 2013, is in line with observations several years after control intensification in The Gambia [36,38] and after eradication of malaria in Mauritius [35]. These children were born after introduction of ACT as first line treatment and for many of them after LLINs implementation. Their response profile reflects the cumulative effects of reduced exposure in the first years of life as they slept under nets early at night, rare occurrence of clinical malaria, which was rapidly and efficiently drug cured [46] and short half-life of the antibody response in this age group [9,26].
The marked drop in seroprevalence, magnitude and breadth of antibody responses in the 7-15y old children between 2002 and 2013 confirms observations about the anti-MSP1p19 response in The Gambia [36,38]. By 2013, the children 7-15y old had lost a good fraction of the responses against SE, LSA1 41 , MSP1p19, SALSA, PF13, AMA1, PfCSP and gSG6observed in the youngest age group 11 years earlier. This occurred in both settings where acquired immunity was different before control intensification, likely reflecting the short half-life of the responses and the similar exposure/ transmission conditions in the few years preceding the 2013 blood sampling. At the community level, seroprevalence, magnitude and breadth of the responses declined. This included a decreased response to MSP1p19, consistent with observations in The Gambia [38], but conflicting with the very long half-life (about 50 years) of anti-MSP1p19 antibodies estimated from comparisons of cross-sectional studies in settings with differing transmission conditions [29].
The longitudinal data provide interesting insights into the temporal decay of responses. Cases and parasite prevalence dropped in 2009 but a dramatic reduction of transmission occurred after 2011 only [46]. The responses against LSA1 41 and MSP1p19 and to a lesser extent against SE correlated with the EIR in the preceding 12 months (Fig 4A) suggesting that they are affected by recent exposure to parasite infection and hence more labile than appreciated. Mathematical modelling showed that the SCRs and SRRs were antigen-specific and poorly related to the mean EIR in the previous 24 months, but interestingly the SRR/SCR ratio increased as the EIR collapsed. The sample size of the longitudinal study did not allow modelling the antibody decay on a yearly basis in the two age groups considered. Additional studies are needed to clarify whether the SRR/SCR ratio is age-specific as suggested by both the crosssectional data and the relationship of antibody levels with EIR (Figs 3 and 4A).
Limitations of this study include the timing of the cross-sectional surveys. We studied the antibodies present after 5-7 months of dry season, and consequently short-lived responses and responses elicited by recent infections could remain undetected. This limitation should  [46,48]. It would thus be interesting to investigate in Dielmo the responses to antigens identified recently as signing exposure in the past 3-6 months[24], and to carry out additional surveys at the end of the rainy season to track possible recent infections. A second limitation is the uncertainty about residual asymptomatic carriage, known to affect the immune response [3,34]. Parasitological surveillance showed uncommon microscopy-positive smears after 2008 individuals (left part) and older individuals (right part). The box shows the first to the third quartile range, the band inside represents the median value; the limits of the whiskers represent the 95% percentiles. The temporal evolution of the median is shown by the dotted line. The temporal evolution of the EIR (right y Axis) is plotted as the background grey area. In part (B) antibody responses are summarized as box-and-whisker plots (as above) by age group, and before and after LLIN distribution in 2008 and 2011. Asterisks indicate significant different levels for a given age group between two successive time periods (P<10 −2 , Mann Whitney Test).
https://doi.org/10.1371/journal.pone.0179146.g003 Polynomial regression curves for the relationship between the median antibody responses to SE (black circles), LSA1 41 (grey squares), MSP1p19 (grey triangles) and EIR (log scale) for the two age groups (y = y 0 + ax + bx 2 + cx 3 ). B. Seroconversion (SCR) and seroreversion (SRR) rates calculated using a 2-year stratification illustrated as bar plot for each antigen (black = SCR, grey = SRR). The temporal evolution of the SRR/SCR ratio is plotted as the plain line connecting the dot point values. The mean EIR (right y axis) for each period is plotted as the step vertical dashed red line. The black crosses indicate the mean prevalence of antibody responses for a given period (the scale used for 100% and 50% is 3.0 and 1.5, respectively on the left y axis). https://doi.org/10.1371/journal.pone.0179146.g004 Profiling decay of anti-malarial immunity [46,49], none were detected in 2013 and very few sub-microscopic infections were detected using PCR (which in our hands detects less than 1 parasite per microliter of blood). This does not exclude possible infections with lower densities or that were not detected by our snapshot samplings. However, the low responses observed in 2013 in the 7-15y old children who historically had a high rate of asymptomatic infections [59], is consistent with uncommon asymptomatic carriage in the recent years.
A limited panel of individual antigens was studied here compared to published highthroughput studies [6,11,13,24]. We studied the responses to SE and to an array of individual antigens known to elicit responses reaching high seroprevalences in these settings [5,10,15,16,53] such that any decay would embrace a large dynamic range. Apart from the variant PF13 [53], the antigens used here are well conserved. Temporal variation in seroprevalence/ levels against these conserved antigens is thus unlikely due to temporal genetic variation of the parasites circulating in the villages. Data regarding anti-SE responses are consistent with the observed decrease from 2000 to 2010 in the 5-19 year old villagers from Dielmo and Ndiop [5]. We show here that by 2013 the decay had occurred in all previously immune groups, including adults and importantly, that the children under 6 years were essentially seronegative against SE indicating quite limited exposure to parasites in this age group. The information gathered using SE is reminiscent of observations using parasite-based assays such as immunofluorescence [35] or indirect hemagglutination [7] assays. SE has the advantage of containing a broad panel of parasite antigens, which moreover are presented in their native conformation and native post-translational modifications. The high historical seroprevalence and levels observed in both villages is consistent with previous studies [5,56] and with the high responses observed against the various erythrocytic antigens studied here but contrasts with the low figures reported in a Kenyan setting [22], possibly reflecting differences in extract preparation. The sensitive SE-assay provides a rapid mean to monitor the anti-parasite-response, but the exact specificity of the observed signals is unknown, precluding identification of specific conversion and reversion events.
The 2002 survey confirms the high historical seroprevalence against the recombinant antigens and peptides used here [10,15,16,53,55]. All responses apart from gSG6 had declined by 2013. The response against PmCSP followed a pattern similar to PfCSP, consistent with the marked drop in P. malariae cases and infections after control scale-up [47].
The response against gSG6 presented specific age-stratified patterns and temporal trends. The peptide allows to capture of antibodies elicited by An. gambiae [17] as well as An. funestus [60], which are the main vectors in Dielmo and Ndiop [43,46]. The lower community seroprevalence in 2013 compared to 2002 is consistent with the published association with transmission intensity [27, 61,62]. The lower responses in the older age groups observed in both villages in 2013 are in line with the reported tolerance response in the older subjects [33,63,64]. The large heterogeneity of antibody levels observed in the young children in Dielmo in 2013 is surprising, as antibodies to gSG6 have been reported to be negatively associated with the use of LLIns in young children [65]. It contrasts with the more homogeneous response observed in the same age group 11 years earlier and suggests substantial heterogeneity of exposure of young children to Anopheles bites in 2013, as reported in low transmission conditions [62]. Presence of antibodies against gSG6 indicates continued exposure to mosquito bites in 2013 is consistent with the recorded vector density in the villages, and probably indicates exposure to Anopheline bites at a time when children are not using LLINs. Indeed, the low/nil responses of these young children against parasite antigens reflect the combined benefit of limited/nil exposure to infectious bites thanks to LLINs and limited exposure to parasites thanks to the use of efficacious antimalarial treatments.
The panel of antigens analysed here provided a consistent picture of declining immunity and allowed to identify clear signatures of progress towards malaria elimination. The list of potential biomarkers is not closed and could obviously be completed with additional antigens identified recently in other settings as associated with recent infections [24]. Our data illustrate the value of community surveys for documenting malaria decline. A few consecutive years of reduced exposure decreased specific immunity against pre-erythrocytic and blood stage antigens whatever the pre-existing level of immunity in the community. The young children born after control intensification had quite limited if any acquired immunity against parasite antigens. This age group is particularly vulnerable to possible failures in control and highly informative in case of such circumstances, but is poorly informative for monitoring further progress towards elimination. The older children have lost a good fraction of their pre-existing immunity as transmission declined. There was clear evidence of waning immunity in adults, with declining antibody levels but still substantial seroprevalence levels, consistent with the residual local transmission and the high antibody levels before control intensification. The previously immune age groups thus appear sensitive indicators of declining transmission and it will be important to continue monitoring in the future years to study how their immune responses evolve as transmission further declines.
Supporting information S1