Distribution of Plasmodium falciparum apical membrane antigen-1 cluster one loop haplotypes and their association with malaria symptoms in Bandiagara, Mali

Abstract

Although overall malaria incidence has decreased sharply over the past two decades, progress has stalled since 2015. The human host maintains a complex set of interactions with Plasmodium falciparum, the parasite species causing most malaria deaths. Understanding the genetic factors of parasites that play a role in malaria pathogenesis and host immune responses could help identify proteins that are essential for parasite survival and aid in the design of more effective tools against malaria, including more effective malaria vaccines. Clinical malaria phenotypes are a function of parasite density, genetic makeup of both parasites and hosts, malnutrition, coinfection, environmental factors, and potentially parasite and host epigenetic factors. We aimed to investigate the relationship between parasite genetics, focused on apical membrane antigen 1 (AMA1), a blood-stage vaccine candidate antigen involved in erythrocyte invasion, and clinical phenotypes. AMA1 haplotype dynamics have shown that the encoding gene cluster one loop (c1L) is the target of invasion-inhibiting antibodies. Herein, we hypothesized that variants of c1L may be associated with malaria symptoms. Using ama1 gene sequences generated from samples from 100 children during a three-year malaria incidence study conducted in Bandiagara, Mali, we assessed whether specific haplotypes were associated with disease phenotypes of children presenting with clinical malaria. After adjusting for study participant location, specific c1L haplotypes were not associated with malaria symptoms or disease phenotypes. None of the most frequent haplotypes were linked to clinical malaria, defined as any temperature greater than 37.5 °C and parasitemia greater than or equal to 2,500 parasites per microliter. Our findings suggest that malaria symptoms may be driven by infections with new allele(s) rather than specific allele(s) already seen by an individual.

Background

Although the overall malaria incidence has decreased sharply from 2000 to 2015, it has stalled since 2017 [1]. To reach the World Health Organization (WHO) Global Malaria Program’s goal of eradication in at least 35 countries by 2030, new tools are needed in addition to current control approaches. Vaccination is an effective strategy that eliminates or mitigates multiple infectious diseases, including smallpox and poliomyelitis [2]. Only two malaria vaccines have been tested in phase 3 trials [3], and one has undergone a phase 4 trial [4]. The WHO has recommended the use of two partially effective pre-erythrocytic malaria vaccines in endemic countries: RTS, S/AS01 [5] and R21/Matrix-M [6]. The pursuit of a vaccine with an overall efficacy greater than 75% in areas with high rates of seasonal malaria is now accomplished with R21/Matrix-M, but efforts toward a more efficacious malaria vaccine with a long duration of protection are still vital [7]. A better understanding of parasite genetic factors that interact with the host immune system and play a role in malaria pathogenesis could help identify proteins that are essential for parasite survival (virulence) and design more effective malaria vaccines.

Clinical malaria occurs when parasites interact with the host immune system, resulting in parasites escaping the host immune response [8–10]. Factors involved in pathogenesis include parasite density and the pyrogenic threshold [11]. Recent studies have shown that var genes and their protein products, PfEMP1s, a group of encoded parasite proteins expressed on the surface of infected red blood cells, play important roles in malaria pathogenesis [12]. The approximately 60 var genes per genome demonstrate high sequence variation [13]. In addition, evidence from genomic and microarray studies has revealed the associations of specific PfEMP1 domain cassettes with malaria severity [14], including cerebral malaria and severe malarial anemia. Furthermore, studies have shown that parasite antigenic variability is a result of immune pressure exerted by the human host [15]. The clinical phenotypes of malaria in humans, including clinical disease, asymptomatic carriage, or submicroscopic parasite carriage, could be impacted by parasite density along with genetic determinants, including proteins that are the target of neutralizing malaria antibodies [16–18]. We hypothesize that the genetic makeup of parasite populations could matter in shaping such phenotypes and vaccine efficacy. Hence, identifying malaria parasite epitope(s) that are the target of the immune response may help select the next generation of malaria vaccine candidate(s).

Prior studies have demonstrated that malaria vaccine efficacy can be strain/variant specific [19–21]. Thus, it is reasonable to evaluate genetic variation when screening malaria vaccine candidates. Although recent progress has highlighted the role of epigenetic factors in susceptibility to malaria [22] and the mediation of antimalarial immune responses [23], it is unclear how these epigenetic regulatory mechanisms could be incorporated into malaria control measures. Therefore, in this study, we assessed whether specific AMA1 protein variants are associated with malaria symptoms by measuring the associations between specific AMA1 c1L variants and the clinical phenotypes of the resulting malaria.

AMA1 is a blood stage malaria vaccine candidate that has been tested in field trials, including two in Mali [24, 25]. The protein is expressed at the surface of the merozoite and is critical for parasite survival, as it is involved in red blood cell invasion [26, 27]. Although the associations of AMA1 variants with clinical malaria have been assessed in vaccine trials [25], little is known about the associations of AMA1 c1L variants with malaria symptoms. In addition, antigenic escape modeling, growth inhibition-invasion assays [28], and field studies [29] have shown the impact of AMA1-c1L polymorphic sites on risk of clinical malaria [29] and strain-specific efficacy [25]. Moreover, there is a lack of knowledge on the impact of AMA1 genetic variants and clinical malaria, which are defined both objectively and subjectively. Finally, the kinetics of AMA1 alleles over time and space have not been assessed in this study site. Further investigation of the role of AMA1 c1L haplotype in relation to malaria symptoms is needed for a rational development of an AMA1-based vaccine.

We hypothesized that intrinsic factors of parasites, particularly their genetic makeup, impact the human host response, and subsequent clinical phenotype. Hence, parasite populations can be differentiated not only by their own phenotype derived from their genetic and epigenetic variables but also by the symptomatic phenotype of the host during clinical presentation. To identify these potential genetic patterns, herein, we explore the relationship between AMA1 genetic variants and malaria clinical symptoms to understand the underlying parasite pathogenesis and to potentially identify variants that may be incorporated into a multivalent malaria vaccine.

Methods

Study site

The study took place at the Bandiagara Malaria Project, which is located within the Bandiagara Health District Hospital. Inpatient, outpatient, and basic laboratory facilities have been established by the Malaria Research and Training Center (MRTC) at this site since 1997. For study purposes, Bandiagara was divided into 8 sectors, and the eight sectors were clustered into two areas depending on their proximity to the river. In Bandiagara, malaria transmission is seasonal, with peak transmission occurring between September and December each year [30].

Subject recruitment and enrollment

The current study is a retrospective analysis of a malaria cohort study performed in Bandiagara, Mali, from 1999 to 2001. The parent study sample size was 450 volunteers. Using 1999 census data collected by an MRTC team, approximately 225 subjects of both sexes aged ≥ 3 months to < 10 years and 225 subjects aged 10–20 years were recruited for the parent study. Study subjects were recruited from all districts of Bandiagara in proportion to the populations of those quartiers to achieve a representation of malaria incidence throughout the community. After providing written informed consent, venous blood, malaria smears, and filter paper blood spots were obtained, and hemoglobin levels were measured for each volunteer. For this study, 100 children (thirty children aged ≤ 5 years, 32 aged 6 to 10 years, and 38 aged ≥ 11) from the parent study were randomly selected for genetic sequencing of each malaria infection during the three-year incidence study [29]. Assuming a rate of 2.5 Plasmodium falciparum-positive time points per participant per year (active sampling), a 50% population-level prevalence of parasitemia, we anticipated a total of 650 symptomatic infections from the cohort. With a sample size of 500 and alpha = 0.05, we will have 80% power to detect a 17.2% difference in incidence/prevalence based on Chi-squared tests.

Inclusion and exclusion criteria

Subjects in the target age group presenting with the following conditions/criteria were not eligible for inclusion in the study: known pregnancy; hemoglobin < 6 g/dl; parasitemia > 100,000/mm³; prostration; respiratory distress; bleeding; recent seizures, coma, or obtundation; inability to drink; or persistent vomiting. Subjects without these symptoms were eligible for enrollment in the study. Those with concurrent illnesses other than malaria but with no exclusion symptoms were included unless severe or chronic illness was present.

Participant follow-up

All the subjects were actively followed by weekly clinical evaluations and monthly blood smears during the rainy season. Filter paper blood spots were collected monthly via digital puncture. All the subjects were followed passively by the continuous availability of study clinicians to evaluate any medical complaints. A malaria smear was performed on any subject presenting at the health clinic with fever defined as an axillary temperature greater than or equal to 37.5 °C using a digital thermometer. Subjects who were diagnosed with uncomplicated or severe malaria were treated in accordance with the National Malaria Control Program treatment policy, using standard doses of oral pyrimethamine and sulfadoxine for uncomplicated malaria or intravenous quinine dihydrochloride for severe malaria. During the dry season, all study participants were evaluated monthly. Any subject with malaria symptoms, including an axillary temperature greater than 37.5 °C or severe anemia or jaundice, was given a full medical and laboratory assessment. If another condition, including respiratory infection or a diarrheal illness, was suggested on the basis of the patient’s clinical history, appropriate referral for treatment was performed.

Sample collection

Malaria smears (thin and thick), venous blood, and filter paper samples were collected at the time of enrollment and monthly every year from July to December (rainy season). The location of samples collected (study participants house location) was assessed by Global Positioning System (GPS) using a portable GPS (E1 RTK Base station^®). Thick smears were stained via Giemsa stain and examined immediately by microscopy for initial parasitemia screening for study participants with fever. Thin smears were preserved for field staining and parasite species identification. Parasitemia was determined onsite via standard methods, either by counting parasites against leukocytes for parasite count/mm³ or by counting the percentage of parasitized erythrocytes on the thin smear when the parasitemia was too high to count on a thick smear. The filter papers used for ama1 gene sequencing were air dried and stored for nucleic acid extraction.

Delineation of objective and subjective malaria clinical phenotypes

For treatment decisions, a broad clinical phenotype for uncomplicated malaria was defined as any level of parasitemia and malaria symptoms in subjects seeking care during unscheduled illness visits or during weekly follow-up visits. These symptoms included fever at the time of evaluation, severe anemia (conjunctival or palmar pallor); jaundice; fever within the last two days; and lassitude, headache, body aches, cough, diarrhea, or abdominal pain. However, in this study, the clinical phenotype of each episode was further defined as objective and subjective clinical malaria. Malaria with symptoms such as increased body temperature, vomiting, and diarrhea, which are observable by others and quickly lead to care-seeking behavior, is defined as “objective” (signs observed/measured objectively), whereas subjective malaria symptoms are those that are challenging to capture, including headaches, abdominal pain, and myalgia (symptoms declared by the patient, not measured by clinician).

Parasitemia and fever data collected from cases of malaria as defined above, cases of fever or symptoms consistent with malaria but with negative malaria smears, and cases of asymptomatic parasitemia detected during monthly parasitemia surveys were used to define thresholds of parasitemia associated with symptomatic malaria infection following published methods [3]. Parasitemia groups were defined as those with fewer than 2,500 parasites/µl; and 2,500 parasites/µl or more.

Parasite DNA extraction and ama1 gene sequencing

Malaria parasite DNA extraction, ama1 gene PCR and sequencing were performed as described previously [20]. Briefly, malaria parasite DNA was extracted from dried blood spots via a QIAamp DNA Mini Kit (Qiagen, Valencia, CA). To sequence ama1, PCR was used to amplify the full-length gene from all samples representing clinical episodes and from asymptomatic infections detected during monthly cross-sectional surveys. The primers and cycling conditions have been described in a previous study [31]. The 1,860 bp PCR products were purified via filter plates (Edge Biosystems, Gaithersburg, MD) attached to a vacuum manifold, eluted in water, processed in a 10 µl reaction mixture via BigDye v3.1 (Applied Biosystems, Carlsbad, CA), precipitated in ethanol, and run in 7 µl HiDi formamide on an ABI 3730XL 96-capillary sequencer. Following sequencing, the contigs were assembled, edited, and aligned to the 3D7 reference sequence via Sequencher v.4.7 (Gene Codes Corporation, Ann Arbor, MI). Infections were classified as single/predominant allele infections or as multiple allele infections if the secondary peak height was > 50% of the primary peak height at any polymorphic position [29]. Thick smear-positive and sequence-negative samples were reamplified and sequenced from the original sample to rule out PCR error. Sequence alignment was performed via BioEdit version 7.2. Haplotypes are defined on the basis of amino acids in the cluster one loop (c1L) of the AMA1-encoding gene, which has been shown to be a target of anti-AMA1 allele-specific humoral immune responses [20, 32, 33]. All infections, including both symptomatic and asymptomatic cases, were sequenced. Minor haplotypes (less than 5% incidence/prevalence) were excluded to avoid statistical bias and improve the reliability of the analyses. This cutoff was chosen as rare variants are generally present in few individuals and may not allow an assessment of the association of these variants with malaria symptoms.

Haplotype dynamics

To assess c1L haplotype dynamics, we measured the percentage of haplotypes at each timepoint and compared their relative frequencies across time. The major haplotypes included 3D7, Fab9, Dd2, 102-1, M24, 7G8, M5, FVO and L32. Haplotype dynamics were initially assessed in both infections and symptomatic cases to outline general trends, followed by analyses limited to infections and clinical malaria cases. Nonreference haplotypes with percentages less than 5% in the whole dataset were excluded from the association analyses.

Data analysis

The haplotype dynamics were assessed by comparing AMA1 c1L variant incidence and prevalence during the three years of follow-up. To evaluate whether the same variants were observed in malaria patients defined objectively and subjectively, we assessed the association of variants of c1L with these two definitions via a chi-square test (low incidence haplotypes). Generalized estimating equations were used to perform logistic regression to estimate whether AMA1 c1L alleles were associated with malaria clinical symptoms (high incidence haplotypes: 3D7, Fab9, and Dd2). The primary outcome was defined as presentation with any malaria symptoms, whereas the primary predictor for the analysis was the presence of a specific AMA1 c1L allele. The models included potential confounding variables such as age group (five years and less and more than five years old) and study year. A compound symmetry covariance structure was used to model the correlation between measurements from the same individual. Analysis was performed via SAS v9.4, SAS Institute Inc. 2023 (Cary, NC).

Results

Out of the 100 randomly selected children, 792 malaria clinical episodes and infection sequences were observed during the three-year follow-up. The study participants had an average of three clinical episodes and/or infections during the study. Overall, 67.5% (345) of the 511 samples with single or predominant alleles had a clinical malaria episode, whereas the remaining (32.5%) sequences were from asymptomatic infections or symptomatic infections with multiple alleles that could not be resolved as a predominant haplotype. We observed (35.55%, 281/792) genetically mixed infections. Overall, mixed infection incidence was higher during the transmission season compared to the dry season. During the three-year follow-up, the numbers of sequences generated from infections or malaria episodes in 1999, 2000, and 2001 were 146, 171, and 194, respectively (Table 1). As previously reported [29], c1L haplotypes were defined as a concatenation of the eight polymorphic residues located between amino acid positions 196 and 207. In the whole dataset, the major c1L allele was the 3D7 reference strain, present in 74 (14.5%) of the 511 samples, whereas the 7G8 haplotype was observed in 5.28% of the samples. Samples without whole gene sequences (missing nucleotides) and minor haplotypes represented 29% (149) of the whole dataset and were excluded from the analyses. The haplotypes of the Dd2, Fab9, M5, M24, 7G8, and 102-1 strains were all present at a frequency greater than 5% (Table 2).

Table 1.

Summary table

Objective	Sample size	Statistics	Key findings
Prevalence of AMA1 c1L haplotypes	511	Chi square	Nine major AMA1 c1L haplotypes are observed at each time point.
Dynamics of AMA1 c1L haplotypes	511	Chi square GLM	AMA1 c1L haplotype is stable across years.
Role of AMA1 c1L haplotype in relation to malaria defined objectively	345	Chi square GEE	AMA1 c1L haplotypes not associated with malaria symptoms or disease phenotypes.
Role of AMA1 c1L haplotype in relation to malaria defined subjectively	345	Chi square GEE	AMA1 c1L haplotypes were not associated with malaria symptoms or disease phenotypes.

GLM: Generalized Linear Model; GEE: Generalized Estimating Equation

Table 2.

Distribution of apical membrane antigen-1 (AMA1) cluster one loop (c1L) haplotype collected from study participants during the three years of follow up

Haplotype	Amino acid sequence	Frequency (%)
3D7	DERHFDKY	74 (14.5%)
Fab9	DQRHFDKY	67 (13.1%)
Dd2	DDRLLDED	65 (12.7%)
102-1	DDRRFDEY	34 (6.6%)
M24	DQRDFNEY	31 (6.1%)
7G8	DHRDFNEY	27 (5.3%)
M5 FVO L32 Others	NGRDLNEY NGRDFNEY DRRDFNED NA*	27 (5.3%) 25 (4.9% 11 (2.2%) 150 (29.4%)

*NA: Not applied

AMA1 c1L haplotype dynamics during the three years of follow-up

The number of AMA1 c1L haplotypes identified in malaria clinical episodes (defined as an axillary temperature greater than or equal to 37.5 °C with any parasitemia) was assessed in 1999 (Fig. 1). While the three major haplotypes (3D7, Dd2, and Fab9) were consistently present at high frequencies, the remaining haplotype relative abundances were variable across time (Fig. 1). Overall, the dynamic pattern based on c1L haplotype frequencies identified in clinical malaria cases was comparable to analyses that included all sequence data (monthly p values < 0.05) and those limited to asymptomatic infection cases. Thus, all the clinical definitions of malaria were merged for analytical purposes.

Fig. 1.

Distribution of c1L haplotypes over the three-year follow-up period. Each bar represents the monthly distribution of identified haplotypes, with colors indicating distinct c1L variants. Colors are consistent across panels to represent the same haplotype throughout. Overall, haplotype distribution patterns remained stable across months and years of follow-up. A: all malaria cases, B: clinical malaria defined subjectively, and C: clinical malaria defined objectively

Dynamics of all AMA1 c1L variants (clinical episodes and infections combined)

The most abundant variants, 3D7, Dd2 and Fab9, accounted for 40.3% of all c1L haplotypes in the dataset with the remaining six variants combined, contributing to 30.4% of the dataset, and the nonreference strain (NRS) haplotypes (defined as any haplotype different from 3D7, Dd2, Fab9, and the six variants listed above) represented 29.4% of the c1L variants. Figure 1 (A-C) shows, on a monthly (beginning of the malaria transmission season to the end of the season) and yearly basis, the dynamics of nine major c1L haplotypes between 1999 and 2001. In samples collected in 1999, the beginning and end of malaria transmission were characterized by fewer haplotypes at higher frequencies, whereas at the height of malaria transmission, all haplotypes were present at high frequencies. The 3D7 reference strain frequency was 7.7% at the beginning of the transmission season, significantly increased to 20% during peak transmission (p = 0.01) and decreased to 11.1% at the end of malaria transmission. The same trend was observed for all haplotypes, except Fab9, which started at its highest level at 31% and significantly decreased to zero at the end of malaria transmission (p = 0.01). Likewise, in 2000 and 2001, a significant increase in the number of haplotypes was observed during the peak of malaria transmission, followed by a reduction toward the end of the rainy season. In general, the differences in amino acid composition among c1L frequencies within the same season were significantly greater than those among seasons (Fig. 1A-C).

Dynamics of AMA1 c1L variants generated from P. falciparum infections

For the three-year follow-up period, the frequencies of c1L haplotypes collected from malaria infection cases were stable, even during high malaria transmission and across years. The peak malaria transmission (September to October) was characterized by many haplotypes at low frequencies, ranging from 0.1% to 14.5% (Fig. 2A-D). The frequencies for each haplotype except for nonreference variants (which were not assessed due to low frequencies) were not significantly different (each p > 0.05).

Fig. 2.

Dynamics of Plasmodium falciparum c1L haplotype collected from study participants with fever and malaria parasite during follow up. Each color represents a c1L variant or group of variants. The relative frequency of each variant did not change during the three years follow-up. A: 1999. B: 2000, C: 2001, D: Total. Blue: 3D7, Gray: Fab9, light blue: L32, dark blue: M5, dark gray: 7G8, orange: Dd2, yellow: FVO, green: M24, brown:102-1, light brown: Others

Dynamics of AMA1 c1L variants collected during malaria clinical episodes

Among all the clinical cases observed in 1999, the dynamics of haplotypes followed the same distribution as the whole dataset. The frequency of the 3D7 haplotype, although low, was dependent on the malaria transmission season. In addition, as described in the whole dataset, the Fab9 haplotype frequency decreased while malaria transmission increased (Fig. 2A-D). Overall, haplotype dynamics at the whole dataset level followed the same trend as an analysis limited to clinical malaria cases. Moreover, the same dynamics were observed for the c1L variants collected in 2000 and 2001 (Fig. 2A-D).

AMA1 c1L haplotypes and study participant characteristics

An unadjusted association of c1L haplotypes with participant characteristics was assessed (Table 3). The predictors included sex, age group (less than five years; and five years and older), parasitemia as a continuous variable, and parasitemia group (fewer than 2,500 parasites/µl; and 2,500 parasites/µl or more) in addition to living place (divided into 8 sectors). While parasitemia, parasitemia groups, and age groups were not associated with any of the haplotypes, study participants living closely to the river (p = 0.02) and of the male sex (p = 0.01) had higher frequencies of the 3D7 and M24 haplotypes, respectively. However, after adjusting for study participant sample collection location (quartier closer to the river vs. distant from the river), the prevalence of c1L haplotype variants was evenly distributed between males and females (p = 0.41) and between locations (p = 0.37).

Table 3.

p-values of the association of apical membrane antigen-1 (AMA1) cluster one loop (c1L) haplotypes with malaria, gender and parasitemia groups

Haplotype(N)	Malaria (Objective)	Malaria (Subjective)	Parasitemia group	Gender	Location
3D7*(74)	0.1	0.4	0.6	0.51	0.02
Fab9* (67)	0.9	0.8	0.9	NA+	NA+
Dd2* (65)	0.6	0.7	0.3	NA+	NA
102-1 (34)	0.6	0.3	0.1	0.09	0.17
M24 (31)	0.1	0.1	0.7	0.01	0.16
7G8 (27)	0.1	0.07	0.1	NA+	NA+
M5 (27) FVO (25) L32 (11)	1 0.5 NA+	0.9 0.6 NA+	0.4 0.3 NA+	0.15 0.43 NA+	0.96 0.34 NA+

N: Total number of sequences included in the analysis

* For these haplotypes with larger sample sizes, a generalized estimating equations was used by controlling for age and time of sample collection. A crude association was conducted for the remaining haplotypes

+ Very low sample size. Association cannot be assessed

Association of AMA1 c1L haplotype with malaria symptoms

The percentages of study participants with specific haplotypes who had malaria morbidity parameters, including fever, parasitemia, vomiting, diarrhea, headache, myalgia, and abdominal pain, were determined. Using generalized estimated equations adjusted for age, time of sample collection, and parasitemia groups, the distributions of 3D7 c1L haplotypes indicate that none of the major haplotypes were associated with malaria, which was defined as having a temperature greater than 37.5 °C and a parasitemia greater than or equal to 2,500 parasites/µl (p = 0.61) (Table 3). An identical analysis based on Dd2, Fab9, L32, M5, M24, 7G8, and 102-1 c1L revealed a lack of association between the c1L haplotype of these strains and clinical malaria (Table 3). Overall, the findings suggest a lack of association between AMA1 haplotypes and any malaria symptoms (Table 3). Similarly, the same lack of association was observed when analyses were conducted based on subjective malaria clinical phenotypes (data not shown).

To assess whether specific haplotypes were associated with subjective malaria, which was defined as malaria parasite carriage with headaches, abdominal pain, or myalgia, the same analyses conducted above were repeated. None of the participants’ characteristics or symptoms were associated with the nine major haplotypes studied (Table 3).

Discussions

To reduce the malaria burden and incidence, an understanding of the genetic background and dynamics of functionally relevant proteins is important. Understanding the associations of Plasmodium falciparum gene variants with malaria symptoms could help develop new control measures. Moreover, assessing how proteins exposed to the immune system evolve in natural settings could help identify new malaria vaccine candidates that may not be subjected to vaccine escape [34]. Additionally, such analyses could help identify variants/alleles/epitopes that could be part of a multivalent vaccine. In this study (Table 1), the AMA1 c1L haplotype distribution was determined among children of different ages over time, along with the proportion of children infected with particular haplotypes and their associations with fever, parasitemia, vomiting, diarrhea, headache, myalgia, and abdominal pain morbidity parameters. This study aimed to assess whether a given haplotype was associated with a morbidity parameter with the goal of identifying AMA1 c1L variants that may be involved in malaria pathogenesis and, furthermore, designing a multivalent AMA1 c1L anti-symptomatic malaria vaccine. The AMA1 c1L haplotype was chosen because amino acid residues located in this cluster have been shown to be the main targets of naturally acquired anti-AMA1 protective antibodies [35], although this type of analysis could be performed on other antigens with similar qualities to further explore the associations with parasite genetic variability and clinical malaria. PfAMA1 c1L analyses in Bandiagara, Mali, have shown extensive genetic diversity among infections. The levels of genetic variability were comparable with values published in studies conducted at the same site, using a different protein [36] or the same protein but with a different study objective [37]. In addition, the results of this study show that, based on polymorphisms in the c1L, the distribution and prevalence of the nine most frequent haplotypes in Table 2 was very similar to results found at another vaccine testing site in Mali [20]. These parallels could suggest selection similarities among seasonal malaria settings located more than 700 km apart. These findings confirm that AMA1 c1L genetic variability is high in this area of high malaria transmission [37, 38]. Studies from Bancoumana (malaria meso-endemic transmission level) [31] in Mali and Tanzania (holoendemic for malaria) [39] have also shown that patterns of allele distribution are correlated with transmission level. This correlation is supported by the hypothesis that diversity depends on the proportion of genetically mixed infections (35.55%, 281/792) and the number of infectious clones per human host [40].

As expected, these findings suggest that the frequency distribution of PfAMA1 c1L haplotypes was variable within malaria transmission seasons but was stable across study years. Merozoite surface protein-1 (MSP1), a polymorphic antigen on the surface of the merozoite, was shown to have the same dynamics in a study conducted in Bandiagara, Mali [41]. While the distribution of alleles within the same transmission season was not stable, there was no shift in the relative allele frequencies of AMA1 c1L haplotypes in Bandiagara during the three years of follow-up. Overall, the minor strain concept of malaria transmission, which holds that selection of less common parasites by immune selection promotes the emergence of novel polymorphisms [42], is consistent with these findings. Hence, genetic diversity may peak during one transmission season but remains constant across years. It is likely that within a given season different new mutations are created, and selection works to purify them away over time.

After the sample collection location and study participant characteristics were adjusted to address confounding variables, none of the malaria-related symptoms were associated with a specific AMA1 c1L allele. Although studies by Engelbrecht [43] and Ofosu-Okyere [44] have shown that certain MSP2 genotypes are associated with higher parasite density and that the FCF27 allele is twice as likely to be found in symptomatic malaria cases than the 3D7 allele is, these findings could be explained by the background exposure of the participants (their acquire immune repertoire), environmental factors and the genetic background of the study participants. In fact, multiple studies [26, 38] have shown no evidence of associations between genotypes and parasitemia, and it has been difficult to identify a virulent haplotype [45], although the lack of capacity to demonstrate haplotype-based virulence could be explained by the technology used. As polymorphic residues in AMA c1L have been shown to be the primary focus of naturally acquired protective antibodies for the AMA1 protein [46], we hypothesized that amino acids in this cluster could be associated with morbidity parameters. After we corrected for characteristics that could bias the associations, no associations were observed. Malaria symptoms are multifactorial. In addition to the natural immunity acquired as a result of multiple consecutive infections, human and parasite factors including var genes make it difficult to identify a single cause. A concurrent analysis of the parasite and human genomics and proteomics may help identify factors that are involved in malaria symptomatology. Likewise, owing to the diversity of interactions, much larger sample sizes might be required to assess the importance of each characteristic or symptom. Finally, AMA1 haplotype assessment alone cannot rule out the possibility that parasite genetic factors are involved in clinical symptoms. The role of AMA1 in parasite biology could be limited to red blood cell invasion and not triggering pathogenetic mechanisms of the parasite. Identifying more genetic markers, such as binding proteins, that are associated with cerebral malaria pathogenesis could help unveil parts of these complex interactions.

The AMA1 sequences used in this study originated from a country where multiple infections are the norm. Thus, the major sequence variants were selected as representative of all the sequences collected from the study participants. While the same criteria have been applied to those with and without malaria symptoms, the exclusion of minor variants may affect this analysis. In addition, the limited number of reference variants did not allow a multivariate analysis. However, when possible, controlling for variables that may influence an outcome did not change the association. While age and location were adjusted for, other confounders including host genetics and co-infections were not addressed in these analyses. These exclusions may lead to misleading results and missed relevant biological mechanisms. Thus, the lack of association could reflect insufficient power or true biological irrelevance of AMA1 c1L to symptoms. Alternatively, factors including epigenetics [47] and var genes [8, 48] may play a central role in malaria symptoms by enabling immune evasion and contributing to severe disease through cytoadherence and antigenic variation.

Conclusion

This study characterizes the dynamics of Plasmodium falciparum AMA1 c1L haplotypes and their association with clinical malaria phenotypes in children. The findings indicate that AMA1 c1L haplotype frequencies remained stable over the three-year follow-up, despite ongoing clinical monitoring and antimalarial treatment pressure. This suggests that an AMA1-based vaccine which include representative variants of the circulating haplotypes could potentially remain effective throughout different transmission seasons. Furthermore, no individual AMA1 c1L variants were significantly associated with clinical symptoms. Although no direct association was observed between AMA1 c1L variants and malaria symptoms, additional integrative studies incorporating host genomic and proteomic data in addition to more parasite genetic markers may be necessary to further investigate the potential role of parasite virulence factors in malaria pathogenesis.

Abbreviations

AMA1

Apical membrane antigen 1

c1L

Cluster 1 loop

P. falciparum

Plasmodium falciparum

Pf

Plasmodium falciparum

WHO

World Health Organization

R21/MM

R21/Matrix-M

PfEMP1s

Plasmodium falciparum membrane protein 1

MRTC

Malaria Research and Training Center

DNA

Deoxyribonucleic acid

PCR

Polymerase chain reaction

NRS

Nonreference strain

MSP1

Merozoite surface protein-1

MSP2

Merozoite surface protein-2

Author contributions

AO designed the study, analyzed the data, wrote the paper, and revised all drafts of the manuscript; MA analyzed the data, wrote the paper, and reviewed the manuscript; AN, DC, AKK, KT, YC, and BK analyzed the data and reviewed the manuscript; MBL, DAD, OKD, AD designed and supervised the field study; S T-H designed the molecular epidemiology study and generated the study sequences; CVP and MAT designed and supervised the field study; supervised the data analysis; and conducted in-depth editing of the paper. All authors attest that they meet the ICMJE criteria for authorship.

Funding

This research was funded by the National Heart, Lung and Blood Institute K01 award 1K01HL140285-01A1 for support for training in vaccinology for AO and additional support provided by the NIH Fund under Award Number U54CA272205 award to AO. This work was also supported by the following grants from the National Institutes of Health: NIH HHSN272201300022I contract, PI: Kotloff, Protocol PI: MBL, NIH U01AI112367 grant, PI: Laurens, NIH N01-AI-85346, contract PI Plowe; Fogarty International Center D43-TW-001589, grant PI Plowe; NIH 1R01AI145852 granted to S.T-H, and NIH R01AI099628 grant PI, Thera.

Data availability

The genetics data used during the current study is available in GenBank (FJ898536 - FJ899041) and from the corresponding authors upon reasonable request.

Declarations

Ethics approval and consent to participate

The original study protocol was reviewed and approved by the Institutional Ethics Committee of the University of Mali (Reference No. FWA#00001769) and the Institutional Review Board of the University of Maryland (Assurance of Compliance No. M1174-01NR). The study was conducted in accordance with the ethical principles outlined in the 2018 Declaration of Helsinki [49] and study participants provided informed consent for the use of genetic data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.