Natural History of Malaria Infections During Early Childhood in Twins

Abstract

Background

The frequency and clinical presentation of malaria infections show marked heterogeneity in epidemiological studies. However, deeper understanding of this variability is hampered by the difficulty in quantifying all relevant factors. Here, we report the history of malaria infections in twins, who are exposed to the same in utero milieu, share genetic factors, and are similarly exposed to vectors.

Methods

Data were obtained from a Malian longitudinal birth cohort. Samples from 25 twin pairs were examined for malaria infection and antibody responses. Bayesian models were developed for the number of infections during follow-up.

Results

In 16 of 25 pairs, both children were infected and often developed symptoms. In 8 of 25 pairs, only 1 twin was infected, but usually only once or twice. Statistical models suggest that this pattern is not inconsistent with twin siblings having the same underlying infection rate. In a pair with discordant hemoglobin genotype, parasite densities were consistently lower in the child with hemoglobin AS, but antibody levels were similar.

Conclusions

By using a novel design, we describe residual variation in malaria phenotypes in naturally matched children and confirm the important role of environmental factors, as suggested by the between-twin pair heterogeneity in malaria history.

Factors including host genetics and environment are thought to influence susceptibility to malaria burden. We compared malaria infection rates in twins during early childhood and show similarities in infection history within twin pairs, but not between twin pairs.

Children living in endemic areas differ in at least 3 important aspects with regard to malaria infection history [1, 2]: the number of falciparum infections experienced, the levels of parasitemia when blood-stage infections are established, and the clinical outcomes of these infections. Several factors have been suggested to contribute to these differences. For example, host genetic factors [3–5] are thought to influence clinical expression and differences in parasitemia. Factors that modify in utero environment have also been linked to between-child differences in malaria outcomes [6, 7]. Although many epidemiological studies aim to quantify these different types of variations and identify their causes, the potentially high number of important host factors complicates the understanding of the different and common processes that lead to variability in these phenotypes.

Here, we propose that comparing the history of malaria infections in twins throughout early childhood would generate new insights into malaria pathogenesis. Twins are exposed to the same in utero milieu and share genetic factors, either partially or entirely, when monozygotic. Furthermore, in some areas, twins often sleep together throughout childhood, implying similar exposure to malaria vectors. By studying the frequency of infections, parasite levels, and clinical malaria presentations in twins, we learn about residual variability of malaria phenotypes in children naturally matched for disease-modifying factors and exposure.

METHODS

Study Population and Clinical Procedures

The longitudinal cohort study of mother-infant pairs was conducted from February 2011 to July 2016 in the health district of Ouelessebougou, located 80 km south of Bamako, Mali, an area of intense seasonal malaria transmission. Study protocol was approved by the institutional review boards at the National Institute of Allergy and Infectious Diseases, National Institutes of Health, and by the Ethics Committee of the Faculty of Medicine, Pharmacy and Dentistry, University of Bamako. Written informed consent was obtained from the study participants or the parents/guardians of pregnant adolescent after receiving a study explanation form and oral explanation from a study clinician in their native language. The protocol is registered at Clinicaltrials.gov, under identifier NCT01168271.

Children were followed up from birth up to 5 years of age. Scheduled and unscheduled visits included clinical examination and blood smear microscopy for the detection of malaria parasites. Children were seen monthly during the malaria transmission season, and every 2 months during the dry season; note (1) that during the study period before August 2013 follow-up was more frequent for children aged <24 months, who were seen every 2 weeks during the transmission season, and (2) that the same visit schedule applied to children within pairs. Throughout the study, any time a child in the study was sick, she or he would also be seen (unscheduled visit). Severe malaria was defined as parasitemia detected by means of blood smear microscopy and at least one of the following World Health Organization definitions of severe malaria: coma (Blantyre score <3), ≥2 convulsions in the past 24 hours, prostration (inability to sit unaided or in younger infants inability to move or feed), hemoglobin level <6 g/dL, or respiratory distress (hyperventilation with deep breathing, intercostal recessions, and/or irregular breathing).

Antibody Detection With Enzyme-Linked Immunosorbent Assay

Blood samples were taken, and plasma was used to detect total immunoglobulin (Ig) G against apical membrane antigen 1 and merozoite surface protein 1 [8, 9]. Enzyme-linked immunosorbent assays were performed using Immulon 4HBX (high-binding), flat-bottom, 96-well microtiter plates (Thermo Scientific). Briefly, plates were coated with 200 ng per well (50 μL per well) of each of the antigens, diluted in 0.05 mol/L carbonate-bicarbonate buffer (pH 9.6) and incubated at 4°C overnight. Plates were blocked with 120 μL per well of blocking buffer containing 5% skim milk diluted in Tris-buffered saline (Fisher Scientific) and incubated at room temperature (RT) for 1 hour. Wells were washed with Tris-buffered saline with 0.05% Tween 20 and 50 μL per well of plasma diluted in antibody diluting buffer was added in duplicate and incubated at RT for 1 hour. After washing, 50 μL per well of goat anti-human antibody (KPL, SeraCare) was added at 1:300 and incubated at RT for 1 hour. After the plates were washed, 50 μL per well of phosphatase substrate (Sigma Aldrich) diluted in coating buffer was added and incubated at RT for 20 minutes. Optical density was measured at 405 nm, using a microplate photometer (Infinite M200 PRO; Tecan).

Assays for Antibodies to P.falciparum Glutamic-Acid-Rich Protein and P. falciparum Schizont Egress Antigen 1

To measure levels of IgG antibodies to recombinant P. falciparum glutamic-acid-rich protein (rPfGARP) and recombinant P. falciparum schizont egress antigen 1 (rPfSEA-1), we developed a bead-based assay, using methods published elsewhere [10, 11]. In brief, the beads in 96-well plates were washed 3 times with Assay Buffer E (ABE) to remove particulates. Plasma samples were diluted 1:80 in ABE buffer, and 50 μL of diluted sample was added to the beads and incubated in the dark on a shaking platform at 1000 rpm for 30 seconds, followed by 300 rpm for 30 minutes at RT. These conditions were used in successive incubations. Excess antibody was removed using a vacuum manifold followed by 3 washes in ABE buffer. After that, 25 μL of biotinylated human IgG detection antibody (Pharmingen) diluted 1:500 in ABE buffer was added to the beads and incubated as indicated. Next, 50 μL of phycoerythrin-conjugated streptavidin (BD Pharmingen) diluted 1:500 in ABE buffer was added to the beads and incubated as indicated. Excess phycoerythrin conjugated streptavidin was removed, followed by 3 washes in ABE buffer. The beads were then resuspended in 125 μL of ABE buffer and analyzed using the BioPlex100 system.

The reader was set to read a minimum of 50 beads with identical unique detection signal. Fluorescence values for bovine serum albumin beads were subtracted from rPfGARP and rPfSEA-1 beads, and the results were expressed as median fluorescence intensity. The cutoff for detectable anti–PfGARP/PfSEA-1 antibody levels was defined as a fluorescence value greater than the mean fluorescence level of 10 healthy children plus 2 standard deviations. Anti-rPfGARP and anti-rPfSEA-1 antibody levels are presented in the Supplementary Materials.

Quantitative Polymerase Chain Reaction

After DNA extraction, real-time quantitative polymerase chain reaction (qPCR) to amplify the 18S small subunit ribosomal RNA gene of Plasmodium falciparum was carried out on a 7500 Real-Time PCR System (Applied Biosystems). Briefly, a PCR mixture was prepared with 2.5 µL of DNA template, a 5-µmol/L concentration of each primer, and a 1.5-µmol/L concentration of TaqMan TAMRA Probe labeled with 6-carboxyfuorescein (FAM) as a reporter. The results were analyzed using default settings on the Applied Biosystems 7500 Fast Real-Time PCR System Sequence Detection software (version 1.4.1).

Statistical Analysis

Descriptive analyses were performed using Pandas and matplotlib libraries in Python version 3.7. Positive blood smears within 28 days of a previous positive smear, and with no negative smears during the same period, were considered to represent a single infection. Bayesian statistical models were fit, using PyStan library, to explore whether observed patterns were consistent with Poisson, negative binomial, or hierarchical Poisson models, to account for the grouping of children in twin pairs. In all models, the prior used for the log-mean parameter (for the hierarchical Poisson model, presented in Equations 1 and 2, this corresponds to μ_g) was Normal (0, 5²), which covers all plausible values and beyond; for the scale parameter (σ_g) used in the hierarchical Poisson model, we used a weakly informative prior of Normal (0, 1). Sensitivity analyses were performed with different prior assumptions, with similar results (see Supplementary Table 1 and Supplementary Figure 8); models were also fit that used the log number of routine visits as the offset term, rather than the log follow-up duration, and were consistent with findings in the Results section.

RESULTS

Study Participants

We report data from a birth cohort study with intensive follow-up performed in Southwest Mali. A total of 25 twin pairs, 2 of which monozygotic, were identified and included in this analysis. These participants were recruited at birth between April 2011 and February 2015 and were followed up for a median of 30.7 months (interquartile range [IQR], 18.5–39.9 months). The median number of study visits, scheduled or unscheduled, was 33 (IQR, 16–44), and the range was 7–91. Half of these children (25 of 50) were female. Of the 50 children, 36 had AA genotype, 9 were heterozygous for hemoglobin C mutation (AC genotype), and 5 had AS genotype. In 9 twin pairs, at least 1 child had hemoglobin mutations; in 4 of these 9 pairs, only 1 of the twins had hemoglobin mutations. Additional information about the pairs is presented in Table 1.

Table 1.

Characteristics and Follow-up of Twins in Study

Characteristic	Dizygotic Twins (n = 46)	Monozygotic Twins (n = 4)
Female sex, no.	25	0
Hemoglobin genotype
ȃAA	32	4
ȃAC	9	0
ȃAS	5	0
Mother's gravidity, no.
ȃPrimigravid	6	2
ȃSecundigravid	6	2
ȃMultigravid	34	0
Placental malaria, no.	6	0
Maternal age, y	Median (IQR), 27 (21–34.5)	20 and 22
Age at first infection, mo	Median (IQR), 10.4 (5.9–13.5)	14.1, 19.8, and 26.2
Age at last study visit, mo	Median (IQR), 30.7 (18.0–40.3)	28.6 and 33.3 in 1 twin pair each
No. of study visits	Median (IQR), 34 (15.2–46.2)	27, 28, 35, and 36
Total no. of positive blood smearsa	183	7
Total no. of clinical malaria episodes	135	6
Total no. of severe malaria episodes	2	0

Abbreviation: IQR, interquartile range; mo, months.

Positive blood smears refer only to falciparum infections

Malaria Infections During Childhood

Overall, 1717 blood smears were performed, 126 in monozygotic twins. Asexual falciparum parasites were identified in 190 smears—137 during scheduled and 53 during unscheduled, walk-in visits. Forty of 50 children had ≥1 falciparum-positive visit; the median age at the first falciparum-positive smear was 11.4 months (IQR, 5.9–14.4 months). In 24 of 25 pairs, at least 1 twin had a positive smear. Both children in the pair with no parasitemia were followed up for >15 months. Clinical malaria developed in 36 of 40 children with ≥1 positive smear (1–13 clinical episodes per child).

Figure 1 (the subset of twin pairs with longest follow-up) and Supplementary Figure 1 (all twin pairs) show patterns of infection and parasitemia. Study pairs were broadly categorized: pairs in which both children were infected at least once (concordant group; group 1 in Figure 1 and Supplementary Figure 1); pairs where only 1 twin had infection (discordant group; group 2 in Figure 1 and Supplementary Figure 1); twins with different hemoglobin genotypes (AA vs AS [ie, heterozygous for sickle cell mutation] hemoglobin type; group 3 in the same figures).

Figure 1.

Falciparum infections during early childhood for 8 twin pairs. In each panel, smear results (parasite counts per 300 white blood cells [WBCs]) for the 2 twins in the pair are represented by blue and red lines. Hemoglobin type and ID numbers are indicated in the upper left corner. Circles represent time points when smears were performed; black squares, clinical episodes. Orange shaded areas represent the transmission season (July–December). To facilitate visualization of low-density infections, infections with <100 parasites per 300 WBCs are represented by stars. Groups defined in the text are shown on the top right corner in each panel. Pairs in group 1 were subcategorized in those with similar parasite levels during infection (group 1a) and those with discordant levels (group 1b), in which the twin with highest parasitemia had ≥1 smear with parasite counts ≥5-fold the maximum count of her or his sibling. Group 2 corresponds to pairs in which only 1 twin had infection. Group 3 includes pairs with discordant hemoglobin S mutation status. Note that here only the 2 pairs in each group with the longest follow-up are presented (the same information is shown for all twin pairs in Supplementary Figure 1). In group 3, only individuals with AA genotype had hyperparasitemia (here defined as > 3750 parasites per 300 WBC; see also Supplementary Figure 3). Clinical episodes in which only nonfalciparum parasites were detected are represented by black squares that are not linked to the plot lines and whose y-axis coordinate is defined by the nonfalciparum parasitemia.

Although numbers of infections in children in the same pair were often similar (Supplementary Figures 1 and 2), in 8 of 25 pairs, all with follow-up longer than 1 year, parasitemia was detected in only 1 twin. The number of positive smears in this group was low: in 5 of 8 pairs the infected sibling had 1 or 2 positive smears, which might suggest lower exposure to sporozoite-infected mosquitoes in the household or area where these pairs lived. Note that 16 of 28 births in group 1 and 4 of 16 births in group 2, as defined above, happened during the transmission season months (July–December); furthermore, the median (IQR) numbers of scheduled visits per month of follow-up in the 2 groups were similar: 0.84 (0.71–0.97) and 0.74 (0.66–0.88) scheduled visits per month, respectively. To assess whether this pattern, in which only 1 twin had infections detected by microscopy, represents reduced exposure or better control of parasitemia at submicroscopic levels, we tested available samples from children with no positive blood smears, using a sensitive molecular assay (qPCR targeting 18S; 2 samples for 3 children, 1 sample for 4 children). Parasites were detected by molecular method in only 1 child without patent infection.

To further understand the distribution of infections within twin pairs, we fit a series of models to the number of independent malaria infections, accounting for differences in follow-up duration: the Poisson model (in which the same incidence parameter is assumed for all children in the study), the negative binomial model (which also assumes the same incidence parameter but allows for more between-child variability, without regard to twin pairs), and the hierarchical Poisson model (in which a separate incidence parameter is allowed for each twin pair). The hierarchical Poisson model is shown below in Equations 1 and 2:

$${\mu }_j\sim Normal({\mu }_g,{\sigma }_g^2)$$ (1)

and

$$y_{ij}\sim Poisson(e^{({\mu }_j+offset)})$$ (2)

where log rate μ_j for pair j is assumed to follow a normal distribution, with location parameter μ_g and scale σ_g. The number of infections detected for child i in pair j is represented by y_ij; the offset term corresponds to the log of the follow-up duration.

The Poisson model, assuming the same incidence for all twin pairs, does not fit the data well. As can be seen in Figure 2A, the data are consistent with the hierarchical Poisson model, suggesting that there is between-pair variability in incidence parameters, but not necessarily within-twin pair variability. In particular, consistency with the latter suggests that the observed degree of discordance in malaria infection history among twins in some pairs could have arisen even if there were no differences in the underlying incidence within pairs of siblings, that is, if both twins in each pair were equally exposed to malaria parasites and had on average the same number of events per time unit. In other words, the relatively low expected numbers of infections per time unit estimated under the hierarchical Poisson model for pairs in group 2, defined in Figure 1, are consistent with the pattern in which one twin in a pair had no infections detected during follow-up and the other twin in the same pair had only a few infections. On the other hand, the underlying infection rates estimated in the hierarchical Poisson model suggest significant variation between pairs (Figure 2B).

Figure 2.

Results of statistical models on the number of malaria infections. A, We fit 3 count data models: a Poisson model, which assumes the same incidence parameter for all individuals; a negative binomial model, which allows for additional variability in the incidence but does not explicitly allow for the clustering of children within pairs; and a hierarchical Poisson model, which allows the incidence parameter to vary between pairs of twins but not within pairs. The y-axis represents the number of infections (black circles) during the follow-up of children, grouped in pairs (x-axis); red stars indicate children with hemoglobin AS genotype. For each child, posterior predictive distributions are presented for the 3 models discussed above; the different degrees of transparency of the colors represent different posterior intervals: 2.5–97.5, 25–75, 40–60. While the Poisson model (orange) does not fit the data well, data are consistent with both the negative binomial (green) and the hierarchical Poisson (purple) models. Of note, we also fit a Poisson model that included as covariate birth before versus after August 2013, when follow-up frequency changed (see Supplementary Figure 4). B, Estimated rate of infection, per month (x-axis; posterior median and 95% interval), based on the hierarchical Poisson model, presented for each of the 25 twin pairs (y-axis). Note that the rate in pair 5 is estimated to be relatively high, owing to infection detection during the relatively short follow-up of this pair. The ordering of this panel, from higher to lower y-axis coordinates, corresponds to the same ordering, from left to right, in A.

The total number of clinical malaria episodes was 141. Of the 4 children who became infected during follow-up and did not develop clinical malaria, 2 were followed up for only a few months, another child became infected only at the age of approximately 6 months, and the fourth child had only 1 infection at the age of approximately 13 months, even though her twin had 2 clinical malaria episodes during the same period. As can be seen in Figure 1 and Supplementary Figure 1, children continued to experience clinical malaria throughout follow-up, sometimes with >1 episode during the same transmission season. The numbers of clinical episodes were generally similar for twins, except for those pairs in which 1 twin had AS and the other had AA genotype (Figure 3).

Figure 3.

Numbers of clinical malaria episodes for twins in the same pair (x- and y-axes). Red circles represent pairs with discordant hemoglobin S status; in these pairs, twin 2 (y-axis) is the AS child. Noise was added to x- and y-coordinates so that circles for pairs with similar coordinates do not fully overlap.

Nonfalciparum malaria infections were also observed: 5 positive smears for Plasmodium malariae, all in different pairs, and 7 positive smears for Plasmodium ovale (3 children in 2 pairs). In 3 clinical malaria episodes, only nonfalciparum malaria species were observed. In addition, malaria transmission stages, gametocytes, were detected in 18 visits (11 children in 9 pairs).

Twins With Different Hemoglobin Types

There were 3 twin pairs with discordant hemoglobin S mutation status. In 1 pair, while 8 of 10 positive smears from the AA child showed >1000 parasites per 300 white blood cells, none of the 17 positive smears from the AS child showed >1000 parasites per 300 white blood cells. The second twin pair with discordant hemoglobin S status had differences in follow-up duration: while the AA child was followed up for nearly 4 years, the AS child died at age 2.5 years. If we consider only the first approximately 2.5 years of follow-up in the AA child, there were 10 positive blood smears, while the AS child only had 1 falciparum-positive visit. We tested 2 samples from this child with AS genotype, using qPCR, and did not observe submicroscopic malaria infection. In the third pair, no infections were detected in either child.

Longitudinal Seroreactivity

In addition to quantifying parasite carriage throughout childhood, we also tested samples, including cord-blood samples, for antibodies against the malaria blood-stage antigens. Cord-blood levels of antibodies against 4 blood-stage antigens were similar between twins (Supplementary Figure 5). In Figure 4 and Supplementary Figures 6 and 7, we show longitudinal changes in antibodies as children age. Samples collected during infancy had lower antibody levels compared with birth, as maternal antibodies are cleared from infants’ blood. For a subset of children, in samples collected after infancy, antibody levels that were as high as those observed at birth were detected, at least temporarily. In 5 of 8 pairs in the discordant group (ie, pairs in which only 1 twin had evidence of infection), the twin with ≥1 positive smear had higher antibody levels after an infection than the twin with no infections. In 1 pair of twins with dissimilar hemoglobin genotypes, although the AS child had consistently lower parasitemia, that child’s antibody levels were similar to those of the AA sibling.

Figure 4.

Levels of antibodies (y-axes) against the malaria antigens apical membrane antigen 1 (AMA-1; gray lines) and merozoite surface protein 1 (MSP-1; green lines). As in Figure 1, each panel presents data for a twin pair. The ordering of the panels in the same in both figures; continuous and dashed lines represent data from the twins represented with red and blue lines, respectively, in Figure 1. Blue and red circles, in the lower half of each panel, indicate the timing (x-axes) of positive smears. Black crosses on the serological measurements represent serological samples coinciding with smear-positive results. Of note, there were twins in whom a high humoral response developed despite low cumulative parasite burden. Note that, as for Figure 1, only data for the twin pairs in each group with the longest follow-up are shown; similar data for all twin pairs are shown in Supplementary Figure 6. Hemoglobin type and ID numbers are indicated in the upper left corner. Abbreviation: OD, optical density at 450 nm.

DISCUSSION

Falciparum malaria burden is not equally distributed in endemic areas [12]. Here, we provided a detailed description of the history of malaria infections in children who have both similar long-term exposure to vectors and share individual-level factors thought to influence infection outcome.

Although there is substantial evidence for the role of host genetics in modifying malaria disease, with genome-wide association studies in recent years identifying several new protective variants [13], only a handful of malaria epidemiological studies have focused on twins, none of which, to our knowledge, included intensive longitudinal follow-up from birth. These studies primarily aimed to quantify the host genetic contribution to specific malaria phenotypes by comparing monozygotic twin pairs and dizygotic twins. For example, in The Gambia, twin pairs, recruited on average at age 5 years, were followed up during rainy seasons, and a stronger correlation in the development of fever was observed for monozygotic than for dizygotic twins [14].

In another malaria twin study, it was shown that correlations of the IgG isotype levels were generally higher in monozygotic than in dizygotic twins, which was interpreted as reflecting a degree of genetic regulation [15]. In addition to studies on twins, other studies have assessed familial aggregation of malaria phenotypes to understand the joint contribution of shared genetic and environmental factors [16]. While our sample size of 25 twin pairs is relatively small, the data set included >1500 parasitological observations during the age period when most clinical burden occurs. More importantly, by having information from birth, we were able to describe patterns of infections in these pairs of siblings, as these children acquire partial immunity from repeated exposure.

One advantage of the design adopted is that study participants in each twin pair were expected to have similar exposure to Anopheles vectors. Indeed, although previous studies suggest that even within households there is variation in the numbers of mosquito bites different individuals receive [17], twins do not differ in 2 important determinants of frequency of malaria vector exposure: age, which is associated with attractiveness to mosquitoes, possibly through body size, and bed net use, which is likely to be shared by twins in the same room.

Several previous epidemiological studies have described considerable variability in the numbers of malaria episodes among children living in the same endemic area. In our study area, we observed that while in some twin pairs both children had multiple clinical episodes, in nearly a third of the pairs, only 1 twin had parasites during follow-up. While previous studies [18, 19] suggest that the average duration of malaria infection is longer than the interval between scheduled study visits, we cannot rule out the possibility that children with no parasites detected during the study had short-duration or subpatent infections. However, statistical analyses suggest that these observations are consistent with models that assume a common incidence for twins in the same pair.

The presence in our study population of siblings with discordant hemoglobin types also allowed us to compare parasitemia during children’s first infections by hemoglobin S mutation status. The observation of consistently low parasitemia in a child with AS hemoglobin genotype, while the twin with AA genotype had multiple infections with high parasitemia, provide additional evidence for the effect of hemoglobin S heterozygosity on parasite growth as a mechanism of protection against malaria [20]. This observation also parallels results of a Kenyan study showing that children with more clinical episodes than average often lived in the same homestead as children with average numbers of clinical episodes and that sickle cell mutation was an important factor that differentiated these groups [21]. Furthermore, after the consistent decrease in antibody levels during the first months of life as maternal antibodies wane, our antibody data show variable patterns of longitudinal changes during early childhood, including for children in the same pair (although changes in antibodies were often similar for different target antigens studied), reflecting between-pair differences in parasite exposure.

In conclusion, by using a novel design, including longitudinal follow-up of twins from birth, with >1500 parasitological observations, and a series of statistical models involving different assumptions on the distribution of the underlying incidence risk in the study population, we described heterogeneity in malaria burden between children from different families. This could be related to differences in vector exposure, host genetics, and preventive measures, such as bed net use. Our detailed longitudinal data from naturally matched children also suggested similarities in malaria infection history within twin pairs, although important differences were observed for twins with different hemoglobin genotypes. More detailed phenotyping (ie, host and parasite transcriptomics) of infections among intensively followed up matched children, for whom history of infection exposure would thus be documented, is the next methodological step that could provide additional insights into susceptibility to repeated malaria episodes during early childhood and generate hypotheses to be tested in larger population-based studies.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. We thank the women and children in Ouelessebogou, Mali, for participation in the study. Jillian Neal managed quantitative polymerase chain reaction analysis, Rathy Mohan managed the clinical data, and staff at the community health center supported follow-up of study participants. We thank J. Patrick Gorres for editing and assistance in preparing the manuscript.

Financial support. This work was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health.