Increased circulation time of Plasmodium falciparum underlies persistent asymptomatic infection in the dry season

The dry season is a major challenge for Plasmodium falciparum parasites in many malaria endemic regions, where water availability limits mosquito vectors to only part of the year. How P. falciparum bridges two transmission seasons months apart, without being cleared by the human host or compromising host survival, is poorly understood. Here we show that low levels of P. falciparum parasites persist in the blood of asymptomatic Malian individuals during the 5- to 6-month dry season, rarely causing symptoms and minimally affecting the host immune response. Parasites isolated during the dry season are transcriptionally distinct from those of individuals with febrile malaria in the transmission season, coinciding with longer circulation within each replicative cycle of parasitized erythrocytes without adhering to the vascular endothelium. Low parasite levels during the dry season are not due to impaired replication but rather to increased splenic clearance of longer-circulating infected erythrocytes, which likely maintain parasitemias below clinical and immunological radar. We propose that P. falciparum virulence in areas of seasonal malaria transmission is regulated so that the parasite decreases its endothelial binding capacity, allowing increased splenic clearance and enabling several months of subclinical parasite persistence. Malaria cases are predominant during the rainy seasons in many endemic regions owing to the life cycle of the mosquito vector. How Plasmodium falciparum adapts in humans during the intervening dry season, without causing malaria symptoms or killing the host, offers new insights into its persistence in humans.

T he mosquito-borne P. falciparum parasite is responsible for over 200 million malaria cases annually, and in 2018 it killed nearly 400,000 individuals, most of whom were African children under 5 years of age 1 . P. falciparum causes disease while multiplying asexually within red blood cells (RBCs) and exporting its variant surface antigens (VSAs) to the RBC surface. VSAs mediate adhesion to vascular endothelium, thereby helping the parasite avoid splenic clearance 2,3 . During each ~48-h replicative cycle in RBCs, P. falciparum follows a regulated transcriptional pattern, starting from the invading merozoite, through the ring and trophozoite stages and to the multinucleated schizont 4,5 , which yields 16-32 new merozoites. In parallel with a predictable transcriptional pattern, the parasite develops a network of membrane structures 6 in the infected RBC (iRBC), and, at the trophozoite stage, the host cell membrane presents knobs 7 exposing parasite-derived P. falciparum erythrocyte membrane protein 1 (PfEMP1), encoded by the multigene family var 8 . var genes are expressed in a monoallelic fashion, coding for PfEMP1s that bind host endothelial cell receptors, with different binding phenotypes associating with varying virulence and pathological outcomes 9 . In Mali and many African regions, malaria cases are restricted to the rainy season when the mosquitoes transmitting P. falciparum are present 10 , and P. falciparum induces a minimal immune response during the dry season. To test the hypothesis that host immunity might contribute to the suppression of parasitemia during the dry season, we compared the immune responses of subclinical carriers of P. falciparum (May + ) versus non-infected children (May − ). We profiled individuals between 7 and 17 years of age, as this was the group in which most subclinical infections were detected, for serological markers of inflammation and cytokines, circulating immune cells and humoral responses to P. falciparum VSAs of age-and gender-matched children who did or did not carry P. falciparum during the dry season, detected retrospectively by PCR. Inflammation markers previously reported to be elevated in clinical cases of malaria, such as C-reactive protein (CRP) 15 , von Willebrand factor (vWF) 16 and hepcidin 17 , were quantified in plasma samples obtained at the beginning (January) and end (May) of the dry season. None of the three markers was significantly different in subclinical carriers compared to uninfected children at either time point (Fig. 2a). We complemented these serological analyses with a multiplex bead array to detect 32 cytokines and chemokines and observed no differences between children with or without P. falciparum at the end of the dry season (May) in all but one of the quantified analytes (Supplementary Table 2). Only CXCL1, a pro-inflammatory chemokine known to recruit neutrophils, which has thus far not been associated with malaria in the clinical setting, was significantly increased in children with P. falciparum-persistent parasitemias at the end of the dry season (Fig. 2b). In contrast, CCL3, IL-10, IL-6 and IL-1β, previously associated with clinical malaria 18-20 , were similar in the plasma of infected versus uninfected children at the end of the dry season ( Fig. 2b and Supplementary Table 2). We next quantified the proportions of major leucocyte populations from thawed peripheral blood mononuclear cells (PBMCs) collected at the end of the dry season from children with or without subclinical P. falciparum (gating strategy in Extended Data Fig. 3). We observed that monocytes, T cells, B cells and natural killer (NK) cell sub-populations were not significantly different between children who carried (May + ) or did not carry (May − ) P. falciparum ( Fig. 2c and Extended Data Fig. 4). To interrogate differences in cell function, we quantified intracellular cytokines, activation or cytotoxicity markers, transcription factors or exhaustion markers of freshly collected PBMCs from P. falciparum subclinically infected and non-infected children at the end of the dry season. The levels of the activation marker CD25, transcription factor T-bet or cytokine IL-2 of CD4 T cells, granzyme B of CD8 T and NK cells and exhaustion marker FCRL5 of atypical memory B cells were similar between children who carried or did not carry P. falciparum (Fig. 2d and Supplementary Table 3). We further questioned whether memory B cells (MBCs, defined as CD19 + , CD10 − and CD21 − and CD27 + or − or CD21 + and CD27 + ; gating strategy in Extended Data Fig. 3) specific for P. falciparum were affected in subclinical carriers compared to non-infected individuals at the end of the dry season. Using biotinylated P. falciparum blood stage antigens apical membrane antigen 1 (AMA-1) and merozoite surface protein 1 (MSP1) 21 , we quantified AMA-1-or MSP1-specific MBCs in children who carried P. falciparum parasites and non-infected children at the end of the dry season. We found that the proportion of class-switched P. falciparum-specific MBCs (AMA1 + or MSP1 + IgG + IGM − MBCs) was significantly increased in subclinical carriers at the end of the dry season, whereas no such difference was found in the non-class-switched MBC population (AMA1 + or MSP1 + IgG − IGM + MBCs) (Fig. 2e). Within the IgG + MBC sub-populations, we did not observe differences between P. falciparum carriers and uninfected individuals at the end of the dry season in P. falciparum-specific classical and atypical MBCs, but we detected increased P. falciparum-specific activated MBCs in subclinical carriers (Extended Data Fig. 4). Using another multiplex bead array, we quantified humoral responses of P. falciparum subclinical carriers and uninfected individuals, at the beginning

NaTurE MEDiciNE
and end of the dry season, to 35 domain types of the VSA multigene family var, which were grouped according to their endothelial receptor affinity (CD36, EPCR or unknown receptor) and PfEMP1 upstream promotor sequence (UPS) type (A, B or B/A types) 22 (Supplementary Table 4). We observed that more subclinical carriers (May + ) than non-infected individuals (May − ) were reactive against PfEMP1 domains binding to CD36, EPCR or to unknown receptors at both time points, and also that the proportion of individuals reactive to the different PfEMP1 domains decreased over the dry season independently of individual infection status ( Fig. 2f and Extended Data Fig. 4). These differences parallel our previously published data on P. falciparum-specific humoral responses to non-VSAs 13 , suggesting similar humoral dynamics for PfEMP1s and non-VSAs. Additionally, we observed that the magnitude of IgG reactivity to A, B or B/A types of PfEMP1 declined similarly from the beginning to the end of the dry season in children who carried subclinical infection (May + ) or were uninfected (May − ) during the dry season ( Fig. 2g and Extended Data Fig. 4). Antibodies against PfEMP1 domains (Fig. 2f,g), against a large set of P. falciparum non-VSAs 13 and also particularly against RBC invasion-related proteins 23 (Extended Data Fig. 4) were consistently higher in subclinical carriers compared to non-infected children at the end of the dry season, so we questioned whether the difference in humoral response at the end of the dry season could impose variance in inhibition of merozoite invasion in vitro. We tested merozoite invasion of a laboratory-adapted P. falciparum strain in the presence of plasma from Malian children who carried parasites or not during the dry season and evaluated the antibodies' ability to block RBC invasion. Testing complete and antibody-depleted plasma, we observed that complete Malian plasma inhibited invasion of merozoites ~fivefold more than antibody-depleted Malian plasma, whereas antibody depletion had no differential effect on the control German plasma used ( Fig. 2h and Extended Data Fig. 4). Notably, however, plasma from Malian children carrying subclinical infections (May + ) or not carrying parasites (May − ) had similar ability to inhibit merozoite invasion, suggesting that the antibodies remaining elevated at the end of the dry season have no significant effect on inhibiting merozoite invasion of RBCs and are unlikely to contribute to the maintenance of low parasitemias through this mechanism.
P. falciparum genetic diversity is maintained throughout the year. Next, we asked whether P. falciparum parasites persisting through the dry season are genetically distinct from those causing acute malaria during the transmission season. To that end, we measured the size of the merozoite surface protein 2 (msp2) gene, which is highly polymorphic and discriminates different P. falciparum genotypes 24,25 . Through nested PCR followed by fragment analysis using capillary electrophoresis, we compared paired samples from 93 subclinical carriers at the beginning (January) and end (May) of the dry season, with 136 samples from clinical cases of malaria in the ensuing transmission season (MAL). The number of clones detected per individual did not significantly differ between parasites isolated during the dry season or transmission season nor did the percentage of individuals with different numbers of clones (Fig. 3a, b). Furthermore, the size and distribution of msp2 clones identified during the dry season were similar to those isolated from clinical malaria cases during the transmission season ( Fig. 3c,d), with the most frequent clone sizes being the same at any of the time points analyzed.
Transcriptome of circulating subclinical P. falciparum at the end of the dry season differs from that of P. falciparum during clinical malaria. We could not investigate the few malaria cases diagnosed in the dry season because the study protocol did not include venipuncture blood draws from clinical cases during this time. However, we performed RNA sequencing (RNA-seq) of leucocyte-depleted blood from 12 children with persistent subclinical P. falciparum at the end of the dry season (May) and from 12 age-and gender-matched children presenting with their first clinical malaria case in the ensuing transmission season (MAL) (Supplementary Table 5 Table 6). Validation of the RNA-seq data was performed by RT-qPCR of eight high-expressing and variable-in-function DEGs, and the correlation between the two methods resulted in highly significant r 2 = 0.929 ( Fig. 4d and Supplementary Table 7). Furthermore, samples from additional children (six from the end of the dry season and 12 malaria cases during the transmission season) were used to quantify expression of three of the eight DEGs above, in parallel with the initial 24 samples, and revealed similar fold changes by RT-qPCR (Fig. 4e). We investigated similarities in the DEGs obtained in this study with those of previous reports comparing parasite physiological states and transcriptomes from a range of clinical malaria severities 26 , or parasites causing malaria in high-versus low-transmission areas 27 , but no enrichment was found (Extended Data Fig. 5), suggesting that singular mechanisms might be at play during the dry season. Functional and Gene Ontology analysis of the dry season DEGs revealed a significant enrichment of transcripts involved in cellular processes related with several metabolic pathways and also with phagosome, DNA replication or homologous recombination (Fig. 4f). Indeed, DEGs involved in metabolic pathways suggest that glycolysis, glycerophospholipid, purine and pyrimidine pathways were increased in parasites from the end of the dry season (May), whereas fatty acid biosynthesis appeared downregulated compared to parasites from clinical malaria (MAL) in the wet season (Fig. 4g)

NaTurE MEDiciNE
transcripts that appear to be exceptions to the trend of upregulation or downregulation (Fig. 4g); however, we observe that these particular transcripts also present exceptional patterns of expression within the 48-h intraerythrocytic developmental cycle compared to other pathway transcripts 4 (Extended Data Fig. 6). To follow-up on possible metabolic differences between parasites persisting through the dry season and parasites causing malaria in the transmission season, we used liquid chromatography-mass spectrometry (LC-MS) to profile both hydrophilic and hydrophobic metabolites from the plasma of 12 subclinical children with P. falciparum infections at the end of the dry season (May) and of 12 children presenting with their first clinical malaria case (MAL) in the rainy season. We

NaTurE MEDiciNE
in the wet season (Fig. 5d). Finding later developmental parasite stages at earlier times in this short-term in vitro experiment during the dry season could indicate a faster than 48-h intraerythrocytic replicative cycle, or, alternatively, that dry season parasites circulate longer without adhering to the host vascular endothelium and were more developed than circulating parasites in clinical malaria cases at the time of the blood draw. To test the latter, we used the RNA-seq data described in Fig. 4 to estimate, with a likelihood-based statistical method previously described 28 , the age in hours post-invasion (hpi) of circulating parasites from subclinical children at the end of the dry season (May) and from clinical cases during the wet season (MAL). We determined that parasites circulating in the dry season had a transcriptional signature of ~12.5 hpi, 95% CI (11.2, 13.8), whereas parasites circulating in malaria cases during the wet season had a transcription profile similar to parasites with ~6.4 hpi, 95% CI (6.16, 6.8) (Fig. 5e). Accordingly, imaging the thick blood films made in the field at the time of the blood draw, we confirmed that the more developed trophozoite stages of P. falciparum were present on subclinical samples collected at the end dry season, whereas clinical malaria samples in the transmission season presented much smaller ring stages of P. falciparum ( Fig. 5f,g). All together, these data show that, at the end of the dry season, P. falciparum can be found circulating at later stages of the ~48-h asexual cycle than what is seen during clinical malaria cases in the wet season.
Infected erythrocytes in circulation at the end of the dry season are at higher risk of splenic clearance. To investigate if longer circulation of iRBCs in the dry season would affect host RBC deformability and potentiate splenic clearance, we used a microsphiltration system mimicking the narrow and short inter-endothelial slits of the human spleen with different-sized microspheres 29 . Using freshly collected blood samples from asymptomatically infected children at the end of the dry season and from children presenting with febrile malaria during the transmission season, we assessed retention in the microspheres and flow-through of circulating iRBCs at time 0 and after 6, 18 and 30 h in vitro. We observed that iRBCs collected from malaria (MAL) cases were not significantly retained in the spleen-like system at 0, 6 or 18 h after culture and that only after 30 h was the percentage of iRBCs in the flow-through reduced, indicating splenic retention of iRBCs (Fig. 6a). Conversely, iRBCs in RDT + blood collected at the end of the dry season (May) had significantly reduced flow-through immediately after the blood draw (~25% retention of 0-h iRBCs in the microsphere system) and over 50% retention of iRBCs after 6 or 18 h in culture (Fig. 6a). Accordingly, we observed that trophozoites or schizonts that fail to flow through the microsphere system were circulating (at 0 h) only in the dry season samples (Fig. 6b). We then investigated whether differences in cyto-adhesion, affecting the length of time that parasitized cells remain in circulation, could explain the observed parasite age distributions and microsphiltration results. For this, we used a mathematical model to describe the within-host growth and removal of iRBCs from circulation through cyto-adhesion in the vasculature and through splenic retention (Methods). Both processes were assumed to be dependent on the parasite's developmental stage, increasing as the parasite starts to express adherence-mediating surface antigens, and RBC modification leads to cell rigidity. Whereas cyto-adhering parasites still replicate, those filtered out by the spleen were assumed to be removed. As shown in Fig. 6c, effective growth rates and population sizes of low-cyto-adhering parasites are significantly lower than those of high-cyto-adhering parasites, which can avoid splenic clearance before parasitized cells become too rigid to pass through the spleen. We then obtained estimated parasite age distributions for both scenarios (Fig. 6d) by sampling from the modeled parasite population at random points over the simulated infection time course, akin to blood sampling from a population. As low-adhesion parasites are predominantly removed by the spleen toward the end of their life cycle, they show a much broader age range than high-cyto-adhesion parasites, which are removed from circulation earlier through cyto-adhesion and, therefore, show a narrower and younger age range, in agreement with the observed age distribution from thick blood smears (Fig. 5e,f). Next, we simulated a microsphiltration experiment by 'growing' our sampled model parasites older and evaluating their projected average flow-through based on our assumed, age-dependent splenic retention function (Methods and Extended Data Fig. 7). The throughput of high-cyto-adhering parasites is high for the first 6-10 h before dropping off gradually as parasites grow older (Fig. 6e). In contrast, samples from low-cyto-adhering parasites, with their more uniformly distributed age range, already have a much reduced flow-through at 0 h, which, however, remained more stable as parasites mature over the next 30 h-again, in line with the empirical observations (Fig. 6a). These mathematical results suggest that cyto-adhesion alone can explain the differences between parasites sampled during the dry season (low-adhesion) and parasites sampled from malaria cases (high-adhesion).
To investigate whether the expression of cyto-adhesion-mediating PfEMP1 proteins differed in abundance or quality in subclinical parasites from the dry season compared to parasites found in malaria cases during the wet season, we assembled the var genes from the RNA-seq reads of the 24 samples from the end of the dry season and the malaria cases (Methods). Expression of several var genes has been shown to remain fairly stable between ~10 and 20 h after invasion 30 , which should be close to the average ages estimated for parasites in malaria cases and dry season samples, respectively. Using a recently developed analytical pipeline 31 , we could detect LARSFADIG motifs identifying PfEMP1 coding genes 32 in eight of 12 samples from the dry season and in ten of 12 malaria cases (Extended Data Fig. 8). We were able to annotate full-length var genes, including both the start N-terminal sequence (NTS) and the acidic terminal sequence (ATS) domains, and also many isolated fragments, and we observed more contigs with LARSFADIG motifs and var gene fragments in the wet season samples (Supplementary Table 10). We used different methods to access enrichment of higher expressed var genes in the wet versus the dry season samples. Although we did not see statistically significant enrichment,

NaTurE MEDiciNE
we observed a trend that the top expressed var genes in each individual in the wet season are more highly expressed (Fig. 6f), and we plan further studies with larger samples size to verify this trend. Furthermore, we identified known var gene domains such as Duffy binding-like (DBL) and cysteine-rich interdomain region (CIDR), as well as the NTS and ATS, and searched for typical 5′

NaTurE MEDiciNE
UPS sequences associating with different pathological outcomes 22 in the dry season and malaria case assembled vars. Although we were unable to determine the UPS type of expressed vars owing to the short assembly of the 5′ UTR region, we did find more var genes with a DBLz domain in the malaria samples (13 of 61 in var fragments longer than 3.5 kb) compared to the dry season (one of 11 in var fragments longer than 3.5 kb) (Supplementary Table 10).

Discussion
Asymptomatic individuals carrying P. falciparum at the end of the dry season in areas of seasonal malaria have been broadly described 13,33-37 , but how the parasite bridges two rainy seasons without promoting malaria symptoms or being cleared remained elusive. In this study, with samples from Malians exposed to alternating 6-month dry and transmission seasons, we show that, within each 48-h replicative cycle, P. falciparum iRBCs circulate longer in the bloodstream during the dry season, allowing increased clearance in the spleen and thus preventing high parasitemias, which could lead to immune activation or malaria symptoms 38,39 .
Although asymptomatic parasitemia at the end of the dry season associates with a lower risk of clinical malaria in the ensuing wet season 13,33-35 , clearance of parasitemia with anti-malarials before the transmission season does not increase subsequent malaria risk, and the persistence of infection during the dry season does not prevent

NaTurE MEDiciNE
or slow the decline of P. falciparum-specific antibodies 13 (Fig. 2f,g). Consistent with these observations, low parasitemia during the dry season did not elicit detectable inflammation or affect immune cell function (Fig. 2), indicating that chronic low parasitemia in seasonal endemic settings might differ from controlled human malaria infections (CHMIs) in naive individuals, where low parasitemias Mean ± s.d.; two-tailed Mann-Whitney test. c, Within-host dynamics simulation of growth rates and population sizes over five replication cycles of low-cyto-adhering (left) and high-cyto-adhering (right) parasites, stratified as circulating (red lines), cyto-adhering (orange dashed lines) and total biomass (black lines). d, Simulation of circulating parasite age distribution over two replication cycles after repeated sampling of low-cyto-adhering parasites (Low, n = 100) and high-cyto-adhering parasites (High, n = 100). e, Simulation of circulation and passage through the spleen of independently sampled parasites aging over time, with low-cyto-adhering (Low, n = 50) and high-cyto-adhering (High, n = 50) parasites. f, Expression level of the highest expressed var gene at the end of the dry season and during a clinical malaria case (n = 8 May, 10 MAL). Box plots (d, e) indicate the median ± IQR and the minimum (Q1 − 1.5× IQR) and maximum (Q3 + 1.5× IQR) of the data range (whiskers). IQR, interquartile range.

NaTurE MEDiciNE
appear to induce immunity 40 , and also suggesting that slow and continuous stimulation of the immune system is less effective than sudden changes in antigenic stimulation 41 . Nevertheless, cumulative immunity might be required to sustain the dry season reservoir of P. falciparum. Dry season subclinical carriers have higher anti-P. falciparum humoral immunity (ref. 13 and Fig. 2f,g) and higher P. falciparum-specific MBCs (Fig. 2e) than non-infected individuals, suggesting that a certain level of cumulative exposure is necessary to carry dry season subclinical infections. Additionally, we and others have shown that end-of-dry-season parasitemias are more frequent in older than younger children 13,42 , which is consistent with an age-dependent decrease in parasitemia and increase in anti-parasitic immunity 43,44 . It is possible that, within each P. falciparum infection, sequential presentation of different VSAs on the surface of iRBCs and its corresponding ordered acquisition of antibodies 45,46 favor progressively less virulent parasites. Accordingly, a recent study of CHMI, including naive and semi-immune individuals, observed clinical cases in naive individuals, whereas chronic infections appeared in semi-immune individuals with intermediate antibody levels 47 .
Reports from the transmission season show that increasing malaria severity associates with different parasite transcriptional profiles 26,48-50 , but the persisting dry season reservoir had not been investigated. Our data show that, whereas P. falciparum-causing malaria in the transmission season has a ring stage transcriptional signature, parasites persisting at the end of the dry season resemble more developed intraerythrocytic stages, which we confirmed both visually and through differential growth kinetics in vitro (Fig. 5). Future single-cell transcript analysis of iRBCs 5 will allow comparing stage-matched pools of parasites to better understand how P. falciparum achieves low cyto-adhesion in the dry season. Also of interest will be to revisit earlier reports of transcriptional differences between parasites, inducing varying degrees of malaria severity 26,48,49 , and to question whether these could be partially imposed by the hpi of circulating parasites. In fact, Tonkin-Hill and colleagues found a bias toward early trophozoite transcription in non-severe malaria samples compared to the ring stage transcriptional profile of severe malaria cases 48 , which could be due to differing adhesion efficiencies in vivo. Interestingly, in vitro replication rates of severe and uncomplicated malaria-causing parasites 51,52 have not consistently explained the higher parasitemias observed in severe malaria cases, suggesting that adhesion efficiency differences might also contribute. Continued asexual replication (independently or coupled with immunity) might lead to progressively less adhesive iRBCs, as observed in parasites collected during the dry season. In a rodent-malaria model, uninterrupted asexual-stage growth led to bias in gene expression of VSA and parasite virulence 53 , and the transition between acute and chronic phases is suggested to be independent of adaptive immunity 54 . Also, in humans, it has been suggested that continued asexual replication can skew the PfEMP1 expression profile 47,55 , which is consistent with our var genes RNA-seq data ( Fig. 6f and Extended Data Fig. 8). The mechanisms by which the parasite adapts to the dry season, and how transmission is assured as the rainy season ensues, remain to be investigated. In a varying or unpredictable environment, organisms can overcome unfavorable conditions by sensing environmental changes and adapting their individual developmental program to increase survival. Alternatively, stochastic population heterogeneity can increase the probability of survival under changing conditions 56 . P. falciparum might sense and respond to environmental cues of transmissibility opportunity, as has been described for detection of nutrient availability 57 , sexual commitment 58 or appropriate environment for gametogenesis 59 . Such a mechanism could act through epigenetic modulation of VSAs, or be seasonally imposed by different metabolic states of the host, driving a shift of the parasite from a fast-to a slow-growing program as the transmission season ends and persistence is required, and returning to fast growth as transmission resumes. In an avian-malaria model, chronic Plasmodium relictum was shown to respond to bites from uninfected mosquitoes and increase its replication, promoting transmission 60 . Parasite survival during the dry season is imperative but will be efficient in resuming transmission only if these retain the ability to produce gametocytes that mosquitoes can uptake. Thus, investigating potential adaptive changes in the sexual stages of P. falciparum during the dry season will likely also reveal seasonal adjustments. Investigating the transcriptional profile of the few malaria cases diagnosed in the dry season ( Fig. 1a and Extended Data Fig. 2) should be the focus of future studies.
In conclusion, the survival of P. falciparum-infected individuals during the dry season is advantageous for the parasite, and low adhesion of infected erythrocytes is likely a central feature to the subclinical carriage of P. falciparum, demonstrating the adaptability of P. falciparum parasites to the vector-free period.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41591-020-1084-0.   13 , and from individuals at their first malaria episode of the ensuing transmission season, was collected into EDTA tubes (Vacutainer K3EDTA Tubes, BD) and processed directly at the field site. Plasma (used for metabolomic analysis) was separated by centrifugation and immediately frozen in liquid N 2 . The buffy coat was discarded, and leucocytes were removed from the RBC pellet in a two-step procedure: first by density gradient on Lymphoprep solution (Fresenius Kabi), followed by Plasmodipur (EuroProxima) filtration, all according to the manufacturer instructions and as previously described 61 . RBC pellets were then frozen in liquid N 2 and were later used for the RNA-seq and RT-qPCR validation.

Detection of clinical malaria and subclinical P. falciparum infection.
Plasmodium parasites per microliter of whole blood were counted in Giemsa-stained thick blood smears against 300 leukocytes of all symptomatic participants presenting to the study clinic, based on a mean leukocyte count of 7,500 cells per μl. Each smear was evaluated separately by two expert microscopists, and a third resolved discrepancies. Subclinical infections during the dry season were detected at cross-sectional time points by RDT (SD BIOLINE Malaria Ag P.f test of histidine-rich protein II) with a sensitivity of ~100 parasites per μl 62 ) once the blood was in the laboratory and by Giemsa-stained smear and nested PCR amplification of P. falciparum DNA retrospectively as previously described 43 from filter papers (2017)   IgG-AlexaFluor700 (clone G18-145). Aqua Dead Cell Stain was added for live/ dead discrimination (Thermo Fisher Scientific). Stained samples were run on a LSR Fortessa X20 (BD), and data were analyzed using FlowJo 10.2 or higher versions. PfMSP1-or PfAMA1-specific MBCs were identified after exclusion of CD3 + CD4 + CD8 + CD14 + CD16 + CD56 + non-B cells and CD10 + immature and IgD − B cells. To increase the frequency of specific B cells detected in any given sample, PBMCs were stained simultaneously with PfAMA1 and PfMSP1 probes. Therefore, PfAMA1 and PfMSP1 probe-binding cells are indistinguishable by flow cytometry analysis and are referred to together as 'Pf-specific' cells. Influenza hemagglutinin (HA) antigen was used as a non-Pf-specific cells control.
P. falciparum culture. 3D7 P. falciparum parasites were maintained in fresh human O Rh+ erythrocytes at 5% hematocrit in RPMI 1640 complete medium (with L-glutamine and HEPES 7.4% sodium bicarbonate, 100 μM hypoxanthine (C.C.Pro) and 25 mg ml −1 gentamycin (all Gibco)) added with 0.25% Albumax II (Gibco), at 37 °C either in the presence of a gas mixture containing 5% O 2 , 5% CO 2 and 90% N 2 or using the candle jar system method described by Trager and Jensen 63 .
P. falciparum invasion assay. 3D7 P. falciparum parasites were cultured and regularly synchronized by the use of 5% sorbitol 64 and heparin to prevent re-invasion. At the early schizont stage, E64 compound (Sigma-Aldrich) was used to prevent merozoite egress, and, later, merozoites were purified through filtration as previously described 65 . Merozoites were cultured with non-infected RBCs and RPMI medium supplemented with 25% human plasma from different donors for 30 min to allow invasion, and then RPMI medium supplemented with Albumax II was provided to all parasites for 30 h. iRBCs were fixed as previously described 66 and stained with 5% SYBR Green, and successful invasion by merozoites was measured 30 h after invasion using FACS Canto II (BD) and analyzed using FlowJo software 10.2 or higher versions. Antibody depletion from human plasma was achieved using Protein G and Protein L Plus Agarose beads (Pierce, Thermo Fisher Scientific). Successful depletion was obtained after three incubations of plasma with the beads and was verified with a Ready-SET-Go! ELISA Kit (eBioscience) measuring total IgG and IgM quantities in the plasma before and after depletion read. Results were read on a Cytation3 plate reader.
Transcriptome analysis. RNA was extracted from frozen RBC pellets using TRIzol as previously described 4 and was performed from RBC pellets that were frozen in liquid N 2 immediately after blood draw. RNA quality was tested using a Bioanalyzer (Agilent). The average RNA integrity number value was 5.4 for the dry season samples and 5.3 for the clinical malaria cases. The average Bioanalyzer yield was 27.4 ng on the dry season samples, whereas the average Bioanalyzer yield on the clinical malaria cases was 25.0 ng. Twenty-four samples were selected based on parasitemia, the highest 12 titers for the wet and dry seasons. The RNA input going into next-generation sequencing (NGS) sample preparation ranged from 100 pg up to 100 ng. The samples were prepared for transcriptome analysis following a previously described protocol 67 with minor modifications. Initially, samples were treated with TURBO DNase as described. After DNase treatment, Agencourt RNAClean SPRI beads (Beckman Coulter) were resuspended in 19 μl to modify the protocol for a low-input ribosomal RNA depletion step. Ribosomal RNA was depleted following the Clontech Modified Protocol for Removal of rRNA from Small Amounts of Total RNA (100 ng) using Human/Mouse/Rat Ribo-Zero Magnetic Kit (Epicentre). The purification beads were resuspended in 21 μl to proceed with the fragmentation step as described 67 . The first-and second-strand cDNA synthesis steps were followed without modification, except that the final bead purification was eluted in 55 μl. Aliquots from the DNase, Ribo-Zero treatment and fragmentation were analyzed on Agilent Bioanalyzer Pico RNA chips, along with the final cDNA constructs. The purified cDNA was below detectable levels of the Bioanalyzer Pico chips. Next, 50 μl of purified cDNA was prepared for NGS using the KAPA Hyper Prep Kit (KAPA Biosystems). The adaptor stock concentration was 300 nM for the ligation with 4-h incubation at 20 °C. The USER digestion step was omitted as the Hi-Fi polymerase does not amplify uracil-containing products. The number of amplification cycles for PCR was determined to be 14 based on quantification of the amount of post-ligation product with a KAPA Quant Kit for Illumina Sequencing (KAPA Biosystems). The purified amplified libraries were visualized on Agilent DNA 1000 chips. Libraries were quantified using the KAPA Quant Kit for Illumina Sequencing and normalized to 2 nM. The samples were pooled based on parasitemias, denatured and diluted to 11 pM for cluster generation and paired-end 100-cycle sequencing on a HiSeq 2500 Rapid flow cell (Illumina) producing ~13 million reads per sample. Raw reads were trimmed of adapter sequence and low-quality bases and filtered for low-quality reads using the FASTX-Toolkit v0.0.14. Remaining reads were mapped to the P. falciparum genome, build ASM276v1, using HISAT2 v2.0.4 (ref. 68 ). Reads mapping to genes were counted using htseq-count: v0.6.1 (ref. 69 ). For each sample, parasite age (hpi) was estimated based on gene expression using a previously described maximum likelihood method 28 , applying a script that was kindly provided by Chris Newbold (lemieux_et_al_pipeline_functions.r and lemieux_et_al_pipeline.r). Differential expression analysis was performed using the Bioconductor package DESeq2 v1.26.0 (ref. 70 ), with adjusted P values (P adj ) < 0.05 considered significant. DEGs were analyzed for enrichment in Gene Ontology categories using DAVID 71 . RNA-seq data (normalized counts data and raw sequencing reads) are available at the National Center for Biotechnology Information's Gene Expression Omnibus (project ID no. GSE148125). Validating RT-qPCR was performed on the 24 above-mentioned samples and on an extra set of 18 samples: six dry season and 12 clinical cases during the wet season. VILO cDNAs were synthesized using the SuperScript VILO cDNA Synthesis Kit (Invitrogen) and purified according to QIAquick 96-well protocol (Qiagen) with a modified centrifugation protocol 72 . Expression levels of eight transcripts (sir2, rex3, Pfsec23, PFB0100c/KARHP, PF07_0006/STARP antigen, PFB0900c/ PHISTc GEX20, PF08_0020/UBE4B and MIF) were determined by RT-qPCR using Invitrogen Express qPCR SuperMix with premixed ROX reference dye (Invitrogen) in 20-μl reactions. All gene-specific oligo sequences were designed using Beacon Designer software (Premier Biosoft) and purchased from LGC, Biosearch Technologies, with double quencher BHQnova fluorescent probes owing to AT-rich P. falciparum sequences (Supplementary Table 7). In a multiplex format, we used reference gene glycine-tRNA ligase%2C putative (PF14_0198) and a standard made from pooled SPIA cDNA. RT-qPCR reactions were carried out at 95 °C for 2 min, 55 cycles of 95 °C for 15 s and 60 °C for 1 min. Data were analyzed using the 7900HT version 2.4 sequence detection system software per the manufacturer's recommendations.
Metabolite profiling. Plasma metabolomics was performed using both targeted and untargeted approaches, across several LC-MS platforms, for small molecules and lipids, to obtain full metabolite coverage. Each plasma sample was split into two independent samples for metabolite extraction. For hydrophilic metabolites, 50 µl of plasma was extracted by the addition of 9× volumes of ice-cold methanol. Samples were briefly vortexed before centrifuging for 10 min to remove precipitated protein. The clarified supernatants were dried under N 2 gas and resuspended in 100 µl (1:2 dilution). Sample groups were pooled to create a group quality assurance (QA), and all samples were pooled to create a batch quality control (QC), which were injected periodically throughout each run. The hydrophilic extracts were randomized and analyzed using reverse-phase high-performance LC-MS by injecting 10 µl onto an AB SCIEX 5600 (QTOF) TripleTOF in positive ESI mode before injection on a Thermo Exactive Plus Orbitrap in negative ESI mode. Samples were separated on the AB SCIEX 5600 by reverse-phase HPLC using a Prominence 20 UFLCXR System (Shimadzu) with a Waters (Milford) BEH C18 column (100 mm × 2.1 mm, 1.7-µm particle size) maintained at 55 °C and a 20-min aqueous acetonitrile gradient at a flow rate of 250 µl min −1 . Solvent A was HPLC-grade water with 0.1% formic acid; solvent B was HPLC-grade acetonitrile with 0.1% formic acid. The initial conditions were 97% A and 3% B, increasing to 45% B at 10 min and 75% B at 12 min where it was held at 75% B until 17.5 min before returning to the initial conditions. The eluate was delivered into a 5600 TripleTOF using a Duospray ion source (all AB SCIEX). The capillary voltage was set at 4.5 kV in negative ion mode, with a declustering potential of 80 V. The mass spectrometer was operated in information-dependent acquisition mode with a 100-ms survey scan from 100 to 1,200 m/z and up to 20 MS/MS product ion scans (100 ms) per duty cycle using a collision energy of 50 V with a 20-V spread. Metabolite separation was performed on the Thermo Exactive Plus Orbitrap as previously described 73 using a Waters XSelect HSS T3 column (2.1 × 100 mm, 2.5 µm). For hydrophobic metabolites, 25 µl of plasma was extracted by the addition of 3× volumes of isopropanol. Samples were briefly vortexed and allowed to sit at room temperature for 10 min. Samples were then placed at −20 °C to precipitate overnight. Precipitated samples were centrifuged for 20 min, and the clarified supernatant was diluted to 50% water in a glass LC-MS sample vial (1:6 dilution). Sample groups were pooled to create a group QA, and all samples were pooled to create a batch QC, which were injected periodically throughout each run. The hydrophobic extracts were randomized and analyzed using reverse-phase high-performance LC-MS by injecting 10 µl of sample onto an AB SCIEX 5600 TripleTOF in positive and negative ESI modes. Metabolite separation was performed as previously described 74 using a Waters Acquity UPLC CSH C18 column (100 × 2.1 mm, 1.7 µm). Ammonium formate and formic acid were added to the positive ESI solvents, and ammonium acetate was used for negative ESI. Targeted analysis of the Orbitrap data was performed as previously described 75 . Untargeted analysis from the 5600 TripleTOF was performed using the default settings in MS-DIAL 76 . The built-in databases were used for putative identification of metabolites at the MS/MS level. QA/QC samples were evaluated to minimize systematic and technical issues. All data were normalized to the total ion chromatogram and blank subtracted, to remove background noise, before statistical analysis.
P. falciparum field isolates short-term culture. RBC pellets isolated from CPT tubes of RDT + samples in the dry season and of malaria case samples in the transmission season were cultured in fresh human O Rh+ erythrocytes at 7% hematocrit in complete RPMI supplemented with 0.25% Albumax II at 37 °C in a candle jar for 36 or 48 h. Malaria case samples were cultured undiluted and diluted 1:10, 1:25 and 1:50 with non-infected blood to assure that initial parasitemia was low (0.5-1%) and that all cultured parasites could grow to their maximum potential. Parasitemia and parasite development were assessed at 0, 16, 24, 30, 36 and 48 h in culture by thin blood smears and flow cytometry. Parasitemia fold change was determined for each sample (ratio of %iRBCs at each time point over its preceding one). The time of highest increase of parasitemia was the time point at which the ratio of %iRBCs at a given time point over its preceding one was the highest for each sample. Progeny number was determined by dividing SYBR Green fluorescence of multi-nucleated schizonts before or at the time of the highest increase in parasitemia in vitro by the fluorescence of the smallest ring stage population that sample presented. Fiji software was used to measure P. falciparum area. Articles NaTurE MEDiciNE diagonally on the tip and wet by pushing 200 μl of complete medium through the filter. The bead suspension was vortexed, and 400 μl was loaded onto the filter tips, yielding a 1.5-mm layer of microsphere beads. The tips were then filled up with medium and connected to a three-way stopcock. Microsphiltration tips were used within 12 h of preparation. At each time point, 600 μl of the 2% hematocrit culture was loaded onto the microbead layer and perfused with 5 ml of complete medium at 1 ml min −1 using a syringe pump (AL-4000, World Precision Instruments). The upstream and downstream samples were collected at the different time points and stained for P. falciparum quantification by flow cytometry. Samples in which parasitemia increased (fold change > 1) between 0 and 30 h (May) or 0 and 48 h (MAL) were included in the analysis.

Microsphiltration of
Simulation of P. falciparum cyto-adhesion and splenic clearance. We developed a discrete-time within-host infection model to monitor parasite replication, cyto-adhesion and splenic clearance in the absence of host immune responses. The parasite's 48-h life cycle was divided into eight 6-h time steps, and we assumed that the parasite population was fully synchronized. The dynamics of circulating, B i , and cyto-adhering, V i , parasites are described through the following iterative scheme are the numbers of circulating and cyto-adhering parasites of age i and at time t, respectively. γ is the intrinsic parasite growth rate-that is, the average number of newly infected RBCs arising from a single infected cell. As seen from the above equations, only cyto-adhering and freely circulating iRBCs replicate and contribute to parasite population growth, whereas the spleen is assumed to remove retained parasites. Removal of infected RBCs from circulation by means of cytoadhesion, η i , and splenic clearance, σ i , is assumed to be dependent on the age of the parasite, i, where the rate of removal increases as the parasite starts to express cyto-adhering VSAs, PfEMP1, shortly after invasion, and where RBC modification by the parasite gradually increases the cell's rigidity and, hence, splenic retention. Both removal functions are given by the following sigmoidal forms and are visualized in Extended Data Fig. 7. σ max and η max are the maximum removal rates by the spleen and cyto-adhesion, respectively. p σ,η and T σ,η are the shape and location parameters of the sigmoidal functions, where T determines the age at which 50% of the parasites are removed. The factor κ∈[0,1] is included to investigate the effect of cyto-adhesion on the within-host growth dynamics. All parameters and values used are listed in Supplementary Table 12.

PfEMP1 genes expression.
To exclude human reads, all reads were mapped against the human genome (hg19) using bwa mem 77 (-k 32). Reads and their pair that did not map the human genome were used further. Reads were correct with quorum 78 , parameter -k 35. Adapters were trimmed with cutadapt 79 , using the TruSeq LT adapter sequences. To assemble the var genes, the RNA-seq reads (fastq files) were first assembled with the pipeline recently published in Otto et al. 31 . Owing to the variable coverage of the sequence reads, the results were not satisfactory. Therefore, an older approach was used to assemble all the non-human sequencing reads with velvet 80 , following parameters Kmer 41, exp_cov 999999999, ins_length 420, cov_cutoff 5, ins_length_sd 30 and min_pair_count 5. Different k-mer were tested, and k-mer of 41 returned the best results. The obtained contigs were annotated, and the domains were called as previously described 31 . For each var gene, domains and subdomains were annotated as defined by Rask et al. (vardom 81 ). The expression per var gene was calculated with two methods. In both methods, the reads back with bwa mem were mapped against the assemblies. In the first method, the number of mapped reads was counted on var genes larger than 3,500 and normalized by the length of the var genes and the number of reads mapped against all contigs, similar to RPKM. In the second method, the coverage over the middle of the LARSFADIG motifs (samtools depth) was addressed. These values were divided by the number of reads mapped on each assembly and multiplied by 10 million. As a complementary validation, a mapping approach was used: the sequencing reads were mapped using bwa mem, parameter -k 31 -a (to allow multiple hits), against a combined database of var genes from Otto et al. 31 (varDB. fulldataset.1 kb.nt.fasta.gz) and the var genes from Otto et al. 82 , limited to var genes of the length of at least 3 kb. Next, the number of reads mapped with an AS score of at least 95 against the var genes from this combined database were counted.

Statistical analysis.
Mann-Whitney or Kruskal-Wallis was used to test for differences between two or more groups, respectively. Differences between the number of clones and clone sizes were tested by Mood's median test. PfEMP1 slopes between the beginning and the end of the dry season of subclinical carriers and non-infected individuals were compared using a linear non-interaction model. Spearman's rank correlation between linear RT-qPCR normalized data and linear RNA-seq normalized data of each gene was obtained along with P values using GraphPad Prism 8.0 software. Metabolite significant differences were determined by a two-way analysis of variance (ANOVA) corrected for multiple comparisons controlling for a false discovery rate of 0.05%. Statistical significance was defined as a two-tailed P value of ≤ 0.05. All analyses were performed with GraphPad Prism versions 6.0 or 8.0, JMP 14.0.1 or R.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
RNA-seq data (normalized counts data and raw sequencing reads) have been deposited in the National Center of Biotechnology  October 2018 This work is supported by NIH grant U2C-DK119886. The data file of assembled var gene fragments of all isolates are available at: https://github.com/ThomasDOtto/varDB/tree/master/Otherdatasets/ Andrade_DryWet2020 As detailed in the Data availability section of the manuscript.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
No statistical method was used to predetermine sample size. The study included ~600 individuals every year and all clinical data presented in Fig. 1a-c and Supplementary Table 1 include data from all individuals participating in those years with data for each analysis. The decision of size of the study was made based on outcomes of the clinical studies not related with this manuscript. Within the experiments presented in this article no sample size calculations were performed. For data in Fig. 2c-f, Fig 3, and Fig. 5a and c, the number of donor samples used in each experiment was from a minimum of 10 children per group to all the subjects possible to include in each time-point out of the 600 total, during the different cross-sectional time points or during the visit to Bamako in the transmission season, allowing to make the comparisons age-and gender-matched, which was estimated based on our preliminary experiments and previous experience. For all other experiments, a minimum of 10 patient samples were chosen as a sample size to ensure adequate power, unless stated otherwise. Data obtained from each patient sample represents an independent experiment.
Data exclusions There is no exclusion based on race, ethnicity, or gender. Pre-established exclusion criteria at enrolment in May consisted of haemoglobin concentration <7 g/ dL, axillary temperature >37.5°c, acute systemic illness, or use of antimalarial or immunosuppressive medications in the preceding 30 days. Pregnant women were excluded because of the increased risk of anemia during pregnancy and the possible immunomodulatory effects of pregnancy on immunity.

Replication
Data presented includes biological replicates in all experiments (always each point on a graph represents a different biological replicate), with number of samples detailed in the Fig. legends and along the text. All in vitro assays were performed at least twice with different subjects. The data presented is representative of one assay or combined data from multiple, as detailed in the figure legends. Data from serological analyses, RT-qPCR and microsphiltration also included technical replicates of each of the biological samples included.
Randomization Individuals were age-stratified, and randomly selected (computer-generated randomization) from the previously collected census data who meet the eligibility criteria and have provided written informed consent or assent (including parental consent) were enrolled in the study.

Blinding
Data shown in Fig. 2 was performed and analysed without the experimenter knowing to which group each samples belonged, and only at the time of plotting the data was each individual assigned to its group. Metabolomic data (Fig.4 h and supTable 8 and 9) was obtained and analysed with the experimenters blinded to which group each samples belonged to. Parasite areas measured in Fig. 5g were determined by microscopists blinded to the groups the images belonged to. Var gene bio-informatic data was analysed blinded to the groups the RNAseq reads belonged to.
Reporting for specific materials, systems and methods We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Recruitment
A random sample of the village population was drawn in an age-stratified manner to enrol in the study, including individuals aged 3 months to 45 years. The names of potential subjects were selected at random (computer-generated randomization to avoid any self-selection or other biases) from an age-and gender-stratified census (previously collected) of the entire village. The study team contacted the selected individuals in person by visiting their respective homes and inviting them to participate in the study. Written informed consent was obtained from all participants and the parents/guardians of included children.

Ethics oversight
The Ethics Committee of Heidelberg University Hospital, the Faculty of Medicine, Pharmacy and Odontostomatology (FMPOS) at the University of Bamako, and the National Institute of Allergy and Infectious Diseases of the National Institutes of Health Institutional Review Board approved this study.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Clinical data
Policy information about clinical studies All manuscripts should comply with the ICMJE guidelines for publication of clinical research and a completed CONSORT checklist must be included with all submissions.

Clinical trial registration
The study is registered at ClinicalTrials.gov (identifier NCT01322581).

Study protocol
The protocol of the observational study "A Longitudinal Systems Biological Analysis of Naturally Acquired Malaria Immunity in Mali" can be found here: https://clinicaltrials.gov/ct2/show/NCT01322581

Data collection
Samples and clinical data were obtained in a cohort study conducted between 2011 and 2019 in Kalifabougou, Mali, a rural village where malaria transmission occurs from June through December. A single clinic and pharmacy provided the only access to antimalarial drugs. Clinical malaria episodes were detected prospectively by passive surveillance and were defined by axillary temperature ≥37.5°C, ≥2500 asexual parasites/μL of blood, and no other cause of fever upon physical examination. Malaria episodes were treated with a standard 3-day course of artemether/lumefantrine according to national guidelines. Crosssectional clinical visits and blood draws were performed at the beginning (January), mid (March) and end (May) of each dry season.

Outcomes
The pre-defined primary outcome of the clinical study (identifier NCT01322581 onClinicalTrials.gov) was to Identify genomewide expression profiles induced by Pf infection that are associated with immunity to malaria, and it is not related nor was it addressed in this particular manuscript. Our analyses and manuscript relate to the Exploratory Objectives of NCT01322581, to compare immune parameters in Pfinfected and -uninfected individuals at the end of the dry season to investigate host and parasite factors associated with chronic asymptomatic Pf infection. And these measures were accessed monitoring all malaria cases through passive surveillance and have all asymptomatic infection at three different times of the dry season actively monitored.

nature research | reporting summary
October 2018 Flow Cytometry Plots Confirm that: The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided. Cell population abundance population abundance was determined by % of parent population, and a minimum of cells was set for some more rare populations, which did not happen often. Samples with less than 50 iRBCs when quantifying P. falciparum were not included in the analysis. PBMC samples not counting more than 500 monocytes were not included in the monocyte analises. Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information.