Study Design

The samples used in this study were collected during a cohort study designed to measure the age-specific incidence of malaria infection and disease in children and young adults in Bandiagara. The study was conducted prospectively during the years 1999, 2000, and 2001. Study participants were aged 3 mo to 20 y and were recruited from all eight sectors of Bandiagara town in proportion to the populations of those sectors. From July to January of each year, individuals were followed actively by weekly visits. Each visit included a brief clinical examination, and those with symptoms consistent with malaria were given a full medical and laboratory assessment. Each participant contributed approximately 0.1 ml of blood, collected on 3MM Whatman filter paper (Whatman, http://www.whatman.com), at least monthly and at every episode of clinical malaria. Samples were collected under protocols reviewed and approved by Institutional Review Boards of the University of Maryland School of Medicine and the University of Bamako Faculty of Medicine. Informed consent was obtained from all study participants or their guardians.

For this study, 100 individuals with at least two years of follow-up during the malaria incidence study were randomly selected within three age strata. Thirty children aged 5 y or younger, 32 aged 6–10 y, and 38 aged 11 y or older were selected. Individuals selected for analysis did not differ from those not selected with respect to age, gender, or incidence of clinical malaria (unpublished data). Blood samples (n = 2,309) corresponding to all monthly surveys (n = 1,801) and clinical episodes (n = 508) occurring during the transmission season of the three years of the incidence study underwent DNA extraction (QIAamp DNA Mini Kit, Qiagen, http://www.qiagen.com) and PCR, and those yielding PCR products underwent pyrosequencing and haplotype estimation.

PCR

As described previously [38], PCR to amplify the nucleotide sequence encoding MSP-1₁₉ was performed on samples collected at all monthly surveys (regardless of microscopy results) and clinical episodes. A single PCR was used to amplify the region of interest from samples with parasitemia over 1,000 parasites/μl using biotinylated PCR primers (forward, 5′-CAATGCGTAAAAAAACAATGTCC-3′; reverse, 5′BIOTIN-TTAGAGGAACTGCAGAAAATACCA-3′). For samples with parasitemia below 1,000 parasites/μl and microscopy-negative samples, a nested PCR was used to amplify this region, using the above primers for the internal reaction and the following PCR primers for the external reaction: forward, 5′-TTAAACATTTCACAACACCAATG-3′; reverse, 5′-GAGTATTAATAAGAATGATATTCCTAAG-3′. Experiments comparing allele frequencies determined via single versus nested PCR showed no significant difference in allele frequencies (n = 64, mean difference in allele frequencies −0.5%, −0.2%, 1.6%, −0.7%, 0.4%, and −0.3% for the six SNPs of interest). The same reagent concentrations were used for both the single and nested PCR reactions and are described in a previous study [38], as are the cycling conditions for the single PCR. Cycling conditions for the external PCR included ten touchdown cycles beginning at a 63 °C annealing temperature with incremental changes of −0.5 °C/cycle, followed by 35 cycles with a 58 °C annealing temperature. The internal PCR used 2 μl of external PCR as template and ran for 25 cycles at a 60 °C annealing temperature. The 272 bp PCR products were visualized on 1.5% agarose gels. Samples that did not yield a product using the nested PCR were considered parasite-negative.

Pyrosequencing

Using previously published methods, all PCR-positive samples underwent pyrosequencing to determine allele frequencies at each of the six SNPs in the nucleotide sequence encoding MSP-1₁₉ [38]. Pyrosequencing (Biotage, http://www.pyrosequencing.com) is a high-throughput, real-time sequencing method that allows sequencing of short stretches of nucleotides (10–20 bp) surrounding known polymorphisms and quantification of the proportions of alternative nucleotides at each SNP. Four pyrosequencing reactions were used to genotype the six SNPs of interest.

Haplotype Estimation

A mathematical model was used to estimate the frequency of 14 haplotypes in each genotyped sample given the allele frequencies generated by pyrosequencing [38]. The haplotype-estimating algorithm uses maximum likelihood methods to determine the most probable combination of haplotypes given the allele frequencies for an infection, the haplotypes known to be circulating in the population, and a probability distribution of the measurement errors.

Multiplicity of Infection

Figure 3 shows the average MOI in three age groups during the three years of the incidence study. Average MOI fluctuated over time, but overall tended to decrease over the course of the transmission season (Mantel-Haenszel Chi-square test for trend p = 0.074, p < 0.001, p = 0.027, for 1999, 2000, and 2001, respectively, for all age groups combined). In addition, MOI was higher in the oldest age group (≥11 y) than in the two younger age groups. In a multivariable regression model including month, study year, age, and parasite density and taking into account repeated measurements from the same individual, persons 11 years of age or older had a 2.7 times greater odds of having a mixed infection (defined as having more than one MSP-1₁₉ haplotype) compared to children 5 years of age or younger (odds ratio [OR] 2.71, 95% CI 1.76–4.15, p < 0.001). The odds of having a mixed infection was not significantly different between the two younger age groups (OR 1.22, 95% CI 0.84–1.78, p = 0.24). The odds of a mixed infection was also greater in infections with moderate parasitemia (500–5,000 parasites/μl) compared to low-parasitemia infections (<500 parasites/μl) (OR 1.66, 95% CI 1.20–2.31, p = 0.0022). The odds of a mixed infection was not significantly different between higher-parasitemia infections and low-parasitemia infections (OR 1.11, 95% CI 0.80–1.55, p = 0.53 for 5,000–25,000 parasites/μl, and OR 1.07, 95% CI 0.75–1.53, p = 0.70 for >25,000 parasites/μl).

Open in a separate window

Figure 3

MOI by Age

Average MOI over time in three age groups.

MSP-1₁₉ Haplotype Frequencies in Symptomatic and Asymptomatic Infections

Figure 4 shows the frequency of symptomatic and asymptomatic infections in the cohort for all three study years combined. As seen in Figure 4A, the incidence of symptomatic infections peaked in October and decreased to almost zero by January, and was lower in individuals aged 11 years or older. Asymptomatic infection was less common in individuals aged 5 years or younger, and even in the absence of clinical malaria, asymptomatic parasitemia persisted, with a prevalence as high as 60% in the oldest individuals (Figure 4B).

Open in a separate window

Figure 4

Frequency of Symptomatic and Asymptomatic Infections during the Malaria Transmission Season in Three Age Groups

(A) Incidence of symptomatic infection over the course of the transmission season.

(B) Prevalence of parasitemia among asymptomatic monthly surveys. Values in the figure represent the average for all three years of the study, since patterns were similar in all years. Bars indicate 95% CIs.

When the frequency of MSP-1₁₉ haplotypes was compared between symptomatic and asymptomatic infections, three haplotypes (ETSSRL [3D7], QKNNGL, and EKSNGF) had a significantly lower prevalence among symptomatic infections compared to asymptomatic infections (Figure 5). The same results were observed in a multivariable model adjusting for time and age, while taking into account longitudinal measurements. With QKNSGL (FVO) as the reference group, the ETSSRL (3D7), QKNNGL, and EKSNGF haplotypes had decreased odds of being present in a symptomatic infection (ETSSRL [3D7]: OR 0.67, 95% CI 0.49–0.92, p = 0.012; QKNNGL: OR 0.48, 95% CI 0.35–0.68, p < 0.001; and EKSNGF: OR 0.29, 95% CI 0.09–1.02, p = 0.053). When parasite density was added to the model, these associations were no longer significant, suggesting that the associations between these three haplotypes and asymptomatic infection may be accounted for by differences in parasitemia. To test this hypothesis, GEE was used to model the association between each of the three haplotypes and parasite density, adjusting for time, age, and MOI. Indeed, both ETSSRL (3D7) and QKNNGL were more prevalent in the lowest-parasitemia infections (<500 parasites/μl) compared to the highest-parasitemia infections (>25,000 parasites/μl) (ETSSRL [3D7]: OR 2.12, 95% CI 1.41–3.20, p < 0.001; QKNNGL: OR 3.65, 95% CI 2.32–5.72, p < 0.001). QKNNGL was also more prevalent in moderate-parasitemia infections (500–5,000 parasites/μl) than in the highest-parasitemia infections (OR 2.44, 95% CI 1.38–4.31, p = 0.0021). There were no statistically significant associations between the EKSNGF haplotype and parasite density, although as one of the more rare haplotypes, the data may have been too few for meaningful multivariable analysis. The raw data indicate a frequency of 2.8% in the lowest parasitemia infections compared to a frequency of 0.9% in the highest frequency infections.

Open in a separate window

Figure 5

Prevalence of MSP-1₁₉ Haplotypes among Symptomatic and Asymptomatic Infections

Bars indicate 95% CIs. *Chi-square p-value = 0.08 (ETSSRL [3D7]), p < 0.001 (QKNNGL), and p = 0.006 (EKSNGF).

Implications for Malaria Vaccine Design and Testing

By providing information about the prevalence and dynamics of vaccine antigen polymorphisms in a population being targeted for malaria vaccines, this study will inform choices about which MSP-1₁₉ haplotypes to include in future vaccine formulations, and will allow more accurate interpretation of the efficacy of current formulations of MSP-1–based vaccines being tested in clinical trials. For example, the FMP1 vaccine antigen is based on the C-terminal 42 kDa of MSP-1 from the 3D7 strain of P. falciparum (MSP-1₁₉ haplotype ETSSRL) [31]. In the Malian cohort examined in this study and in other malaria endemic areas, this haplotype is not the most common haplotype circulating in the parasite population [20,22,23]. The FVO (QKSNGL) and FUP (EKSNGL) haplotypes were the most prevalent in this study and the FUP (EKSNGL) haplotype had the highest frequency in Vietnam [20], Kenya [22], and Thailand [23]. If immunity conferred by such monovalent vaccines is allele specific, a vaccine with high allele-specific efficacy would have low overall efficacy in populations where the target allele is in the minority. Without an understanding of the distribution of vaccine target haplotypes, this scenario could result in the premature abandonment of a promising vaccine that could be modified (perhaps by including additional target alleles) to be more universally protective. This possibility highlights the need to include molecular endpoints in addition to conventional efficacy endpoints in clinical trials of malaria vaccines.

Analyses of the within-host dynamics of MSP-1₁₉ polymorphisms indicate that changes in amino acids at positions 1691, 1700, and 1701 are associated with a shorter time to next clinical infection and with the development of clinical symptoms, suggesting that these residues may be particularly important in determining allele specificity of anti-MSP-1₁₉ immune responses. These results are consistent with previous studies of anti-MSP-119 immunity, which suggest that there is some differential recognition of MSP-119 haplotypes [24–27], particularly based on polymorphic amino acids in the second EGF-like domain [24,25]. However, sera from malaria-experienced adults immunized with the 3D7-based FMP1/AS02A malaria vaccine reacted to FVO and FUP as well as to 3D7 parasites [30]. Coordinated molecular epidemiological and immunoepidemiological studies are needed to confirm the importance of these residues in determining the dynamics of MSP-1₁₉ diversity, particularly at sites like ours where malaria transmission is seasonal, since MSP-1₁₉ antibody responses wane quickly in the absence of exposure [39].

If the amino acids present at positions 1691, 1700, and 1701 do define an MSP-1₁₉ “serotype,” then instead of 14 haplotypes, there would be six pertinent serotypes, with the KNG serotype (FVO and FUP) having a prevalence of 88% and the TSR serotype (3D7) having a prevalence of 20% in this Malian cohort. This finding greatly reduces the amount of genetic diversity vaccine developers would need to take into account in developing a polyvalent MSP-1–based vaccine, allowing them to focus on the residues most relevant to cross-protection.

Natural and Vaccine-Induced Selection

While temporal variation in the prevalence of individual haplotypes was observed in the cohort from Mali, the relative frequency of the haplotypes remained stable—that is, the most prevalent haplotypes, FVO and FUP, remained most prevalent, and rare haplotypes remained rare. Our results contrast with the temporal changes observed in 138 isolates from Brazil, where in 1985–1986 the FUP haplotype was absent and the 3D7 haplotype was present in 40% of samples collected, but in 1997–1998 the FUP haplotype was more common than 3D7 (37% versus 10%, respectively) [19]. These different patterns may reflect true differences in haplotype dynamics in these locations, but may also have resulted from differences in sample size, sampling interval, and study duration.

It is possible that the stability of MSP-1₁₉ polymorphism prevalences in Mali might be due to balancing selection acting to maintain genetic diversity at this locus, as has been demonstrated for other regions of MSP-1 [40–42]. If vaccine-induced immunity is allele specific, then vaccination could disturb this equilibrium, imposing directional selection that favors alleles not targeted by the vaccine. Such directional selection could induce changes in haplotype frequency and cause a reduction in vaccine efficacy over time, although this effect would likely be diluted by the substantial parasite population maintained in unvaccinated older children and adults. By decreasing the effective population size, vaccination could also increase the effect of genetic drift in the parasite population, resulting in decreased stability of haplotype frequencies. Use of insecticide-treated nets or prolonged effective drug treatment might be expected to have the same effect.

Genetic Diversity and Malaria Epidemiology

Although MOI based on MSP-1₁₉ is likely an underestimate of true MOI, the observed pattern of greater MOI in older individuals is in agreement with studies that have documented similar associations between MOI and age [43,44]. It has been hypothesized that fevers experienced during clinical malaria episodes may play a role in clearing parasitemia in younger children, while older children and adults have acquired immunity that limits disease but does not clear genotypes below the threshold of detection [43,44]. This hypothesis is consistent with the epidemiology of malaria in this area of Mali, where there is a lower incidence of symptomatic infection in individuals aged 11 years or more. However, the increased MOI in this group could also be due to less frequent treatment. Whatever the cause, these data emphasize the possibility that older children and adults serve as an important reservoir of genetic diversity.

MOI was highest at the beginning of the transmission season and decreased over the course of the transmission season, particularly in the two younger age groups. This initial decrease in MOI may be due to an increase in clinical malaria episodes that lead to drug treatment and clearance of parasites, followed by a continued decrease in MOI as EIR drops after the rains cease in October. Since vaccination could reduce MOI, as it has in previous malaria vaccine trials [35,45,46], it is important to document the reduced frequency of mixed infections over the course of the transmission season at this site to allow these natural fluctuations to be distinguished from the effects of a vaccine.

The associations between MOI, age, and time do not appear to be confounded by parasite density, though the data suggest that MOI is higher in infections with moderate parasitemias. The apparent decrease in MOI in higher-parasitemia infections may reflect the rapid growth of a disease-causing parasite clone overtaking the infection and masking the presence of minor alleles.

This study also demonstrated a reduced frequency of three specific MSP-1₁₉ haplotypes in symptomatic infections compared to asymptomatic infections, including one haplotype (ETSSRL [3D7]) currently being targeted for vaccines [29]. This association appears to be due to these haplotypes having a higher prevalence in low parasitemia infections. Vaccination against haplotypes associated with favorable clinical outcomes instead of those associated with poor clinical outcomes could result in increased morbidity and mortality in the human population. Although analyses of the within-host dynamics of MSP-1₁₉ diversity suggest that a change in amino acid at specific residues may be more important than having any particular haplotype, it will be important to examine the association between MSP-1₁₉ haplotypes and clinical malaria in other endemic regions and in the context of genome-wide analyses to determine whether these three haplotypes are inherently less virulent (or are linked to important virulence factors) or whether these associations are an artifact of the haplotype dynamics at this particular site.

We thank the population of Bandiagara, Mali for their continued participation in our studies, as well as the regional and district health authorities of Bandiagara, Mali for their continued support. We thank Drs. Ananias Escalante, Laura Hungerford, and J. Glenn Morris, Jr. for helpful comments on the manuscript. We also thank Dr. Alan Shuldiner and the Division of Endocrinology, Diabetes, and Nutrition, University of Maryland School of Medicine, for use of the pyrosequencer.