Introduction

Exposure to malaria – i.e. the number of blood-stage infections acquired over time – determines to a large extent the frequency of disease and can also explain seasonal patterns and age trends. The number of infections acquired over time depends on transmission intensity, with frequent bites by infected mosquitoes resulting in high numbers of infections in a short period of time. In P. vivax, additional blood-stage infections are caused by relapses from dormant hypnozoites.

As people who live in malaria endemic areas achieve immunity to disease after several years of exposure, newly acquired infections do not always result in clinical episodes. The speed of acquisition of immunity depends on transmission intensity [1]–[3] and also differs between parasite species, with immunity to P. vivax appearing to be acquired faster than immunity to P. falciparum [4], [5]. In numerous field studies conducted in areas co-endemic for both species, the burden of P. vivax infections and disease was found to peak at a younger age than that of P. falciparum [6]–[10].

The mechanisms underlying the immunity to malaria are not entirely understood [2], [11]. Whereas semi-immune people still develop blood stage parasitemia, parasite densities are considerably lower and rarely cause fever. Differences in the rate of natural acquisition of immunity to P. falciparum and P. vivax may be the result of differences in the immune responses induced by either species, and/or a consequence of different numbers of infections acquired over time.

Malaria infections consist of different parasite clones that can infect individuals successively or simultaneously as multiple clone infections. Clones differ in their polymorphic surface antigens as well as in neutral genetic markers. Genotyping of such markers allows discrimination of individual parasite clones during multiple clone infections. The diversity of antigens is well documented in P. falciparum [12]–[14] and P. vivax [15], [16]. The large number of antigen variants is thought to allow parasites to escape the immune system, as new infections likely express alleles different from previous clones, and thus are not recognized by the present humoral response [17], [18]. Immunity is thus assumed to be largely clone-specific, providing little cross-protection against heterologous clones [4], [19]–[22].

Plasmodium falciparum clones transmitted by a mosquito bite appear in the blood stream within 7–10 days. Albeit some of the transmitted sporozoites might not establish blood-stage infections, the number of genetically distinct clones detected in the blood-stream over time (_molFOB, as determined by genotyping) is a direct measure of the molecular force of infection (_molFOI, defined as number of distinct clones entering the body over time). It is thus closely linked to intensity of transmission [23].

In tropical and sub-tropical areas, P. vivax infectious mosquito bites also lead to primary blood stage infections within 10–14 days. In addition, a proportion of parasites remain dormant in the liver. Relapses from such hypnozoites can then occur months later and result in delayed blood stage infections. In high transmission regions, such relapsing parasites are mostly genetically different from parasites detected during the last acute blood stage infection [24], [25]. The relapsing clones might have been present earlier as blood stage infection [26], or the primary and relapsing infections are genetically different clones that were jointly transmitted in a single mosquito bite with only one of these clones having emerged from the liver. Thus, relapses may not only boost existing immune responses by repeated exposure to the same parasite clone, but also lead to a broader immune repertoire. No genotyping approach allows differentiating a relapse from a primary blood stage infection. The number of P. vivax parasites detected in the peripheral blood in a given interval (_molFOB) is therefore a combination of primary blood-stage infections and relapses from the same and/or earlier mosquito bites.

In order to assess the association between (individual) exposure and risk of malaria, we followed 264 children aged 1 to 3 years at enrolment in an area of high endemicity of P. falciparum and P. vivax in Papua New Guinea (PNG) over 16 months [23], [27], [28]. As in earlier studies in PNG [6], [10], [29], P. vivax incidence in this cohort peaked in younger children compared to P. falciparum incidence [27]. While P. vivax incidence decreased throughout the age group, P. falciparum incidence increased between the ages of 1 to 3.5 years with little change thereafter [27]. Previously we had estimated the P. falciparum _molFOI by msp2 genotyping, and shown that children acquired 5.9 new P. falciparum infections per year-at-risk [23]. P. falciparum _molFOI increased significantly with age and was highly predictive of incidence patterns [23].

The diversity and multiplicity of P. vivax in this cohort was published previously [28]. In this paper, we present estimates of the P. vivax molecular force of blood-stage infection (_molFOB).

Methods

Data analysis

In a previous analysis of the samples collected 24 hours apart, we found that not all clones present in a host were detected within a single sample. Twenty-one percent of all msp1F3 alleles and 28% of all MS16 clones were missed on a single day [30]. Thus, we used the combined genotyping data from both day 1 and day 2, except for samples from enrolment and final visits, where only 1 venous blood sample was taken.

The force of new P. vivax blood-stage infections (_molFOB) was generated by counting the number of genotypes in each interval that had not been present in the preceding interval. An 8 to 9 weeks interval started on the first day after a regular cross-sectional visit and included all samples collected during passive case detection over two months plus the samples collected at the end of the interval. _molFOB was also determined for both markers combined, msp1F3 and MS16. In case of discrepancy between the markers for an 8-weeks interval the higher estimate from either marker was used. This approach corrected for imperfect resolution and detectability of a single marker.

Genotyping cannot directly identify relapses; _molFOB measures the combination of primary blood-stage infections and those caused by relapsing hypnozoites. Thus, homologous relapses occurring in two subsequent 2-month intervals would be misclassified as persisting clones. New Guinean P. vivax strains are known to relapse rapidly [31], however in regions of high transmission, relapsing clones are usually of a different genotype than the initial blood stage infection [24]. As a consequence the number of homologous relapses that were not detected is expected to be relatively small.

In line with the pharmacokinetic properties of the drugs [32]–[34], children were not considered at risk for two weeks after treatment with artemether-lumefantrine and four weeks after treatment with amodiaquine (AQ) plus sulphadoxine-pyramethamine (SP). The force of blood-stage infection for each child and interval was subsequently converted into the number of new clones acquired per year-at-risk.

Similar to previous analyses of P. falciparum _molFOI [23], generalized linear mixed models (GLMMs) were used for analyses of force of blood-stage infection as well as for incidence of P. vivax episodes. These models were chosen because they allowed the fixed effects to be specified separately from the random effects (i.e. repeated measurements from the same child over time and unmeasured village factors). Furthermore, the random-effects model allowed for decomposition of the error into between-village and within-village variation.

We fit a Poisson GLMM model with a log link to relate the fixed and (Gaussian) random effects to the number of clinical episodes experienced during a two month interval (defined as febrile illness plus P. vivax >500 parasites/µl). Covariates were selected based on earlier analyses of the same data [27]. Seasonality was characterized by two readily interpretable parameters: the amplitude, which was half the range between the peak and trough, and the phase, which was the location of the first zero crossing in a cycle relative to the origin in time (in this case, the first week of the year). For computational convenience, they were replaced by sine and cosine terms with fixed phases. For all outcomes except prevalence, an offset was fit to adjust for years at risk. Estimation of these models was done using the LME4 package in R version 2.12 [35]. All point estimates provided throughout the text (except those for seasonal effects) were obtained from cubic splines fit using generalized additive models (Figure 1) using the MGCV package in R version 2.12 [36]. For a more detailed description of the statistical approaches see [23]. Point estimates for seasonal peaks and troughs were obtained from the GLMMs by setting all other values of the covariates at their means. For the analyses of the effect of exposure on the relationship between age and incidence of P. vivax malaria, children were stratified into terciles according to the average _molFOB during the entire follow-up.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Age and seasonal patterns of molecular force of blood-stage infection (_molFOB, Panel A) and incidence of clinical P. vivax malaria (Panel B), and effects of ITN use on these parameters.

Smooth splines from generalized additive models with single predictors and 95% confidence intervals.

https://doi.org/10.1371/journal.pntd.0002403.g001

Results

A total of 264 children aged 0.9 to 3.2 years at baseline were enrolled and followed up for 69 weeks. Out of 264 children, 248 (93.9%) were retained until the end of the study with 96.0%–100.0% of children seen at each scheduled two-monthly survey. Over the entire follow-up period, the age ranged from 0.9 to 4.5 years. A detailed description of this cohort was published previously [27]. All but five children had at least one P. vivax msp1F3 or MS16 PCR positive sample during the 16 months of follow-up. Of all samples collected, 51.6% and 52.7% were positive for msp1F3 and MS16, respectively, and 54.8% were positive for either marker. In a total of 1448 P. vivax positive samples, 2,305 and 3,372 distinct clones were detected by msp1F3 (65 different alleles) and MS16 (113 alleles) genotyping, respectively.

Force of blood-stage infection

Excluding any period with residual drug levels from the time at risk, each child was at risk of acquiring new infections for an average of 0.93 years during the cohort (95% confidence interval (CI₉₅) [0.91, 0.96] range: 0.11–1.32). On average, 8.7 P. vivax msp1F3 (CI₉₅ [8.1, 9.4], range: 0–30) and 12.8 MS16 clones (CI₉₅ [11.8, 13.7], range: 0–35) were found per child over the entire follow-up period. When both markers were combined, an average of 14.0 P. vivax clones were observed per child (CI₉₅ [13.0, 15.0], range 0–38).

The average _molFOB was 9.4 new P. vivax infections per child per year-at-risk by msp1F3 (CI₉₅ [8.7, 10.0]), 13.8 by MS16 (CI₉₅ [12.8, 14.8]) and 15.1 by both markers combined (CI₉₅ [14.1, 16.2]). All further analyses were done for both markers combined. In addition, children acquired an average of 5.1 different P. falciparum msp2 clones during the cohort (CI₉₅ [4.6, 5.6], range: 0–19), resulting in a corresponding _molFOB of 5.5 new P. falciparum infections per child per year-at-risk (CI₉₅ [5.0, 6.1]).

P. vivax _molFOB showed a very pronounced seasonality (Figure 1, Table 1, P<0.0001), peaking in early January (week 1, 17.1 clones/year-at-risk) and was lowest in early July (week 27, 11.3 clones/year-at-risk, Figure 2). _molFOB was also significantly lower in 2007 compared to 2006 (incidence rate ratio (IRR) 0.84, CI₉₅ [0.77, 0.92], P = 0.0002, Table 1, Figure 1). Regular ITN use was associated with a significant reduction in acquisition of new clones (IRR 0.66, CI₉₅ [0.56, 0.77], P<0.0001, Table 1). Children with antimalarial treatment in the preceding four weeks had a higher _molFOB than those that were not treated (IRR 1.24, CI₉₅ [1.15, 1.33], P<0.0001, Table 1). _molFOB did not vary significantly with age (Figure 1, P (GLMM) = 0.6). There was significant variation in _molFOB between children in a village (P<0.001) but not between villages (P = 0.3, Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Seasonality of P. vivax clinical episodes (as compared to the first week of January, where incidence peaked) excluding _molFOB (blue) and adjusted for _molFOB (red).

_molFOB accounts for approximately 50% of the seasonal variation in incidence. IRR = incidence rate ratio.

https://doi.org/10.1371/journal.pntd.0002403.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Parameter estimates from GLMMs predicting the molecular force of P. vivax blood-stage infections (_molFOB) and the number of incident clinical episodes of P. vivax malaria with density >500 parasites/µl with and without adjustment for _molFOB.

https://doi.org/10.1371/journal.pntd.0002403.t001

Predictors of clinical P. vivax illness

Over the 69 weeks of follow-up, a total of 1134 febrile episodes with parasitemia of any parasite species and any parasite density by light microscopy were observed, resulting in an incidence rate (IR) of 4.60 episodes/year-at-risk [27]. P. vivax was the second most common cause of malarial illness, causing 605 episodes (IR = 2.46) with any parasite density (supplementary Table 1). Of these, 391 episodes (IR = 1.59) fulfilled the more specific definition of P. vivax malaria (i.e. febrile illness plus P. vivax parasitemia >500 parasites/µl) [37]. P. falciparum caused slightly more clinical episodes (any density: 630 episodes (IR = 2.56); >2,500 parasites/µl: 472 episodes (IR = 1.92)).

As in earlier analyses [27], age and season were significant predictors of clinical episodes of P. vivax malaria (Table 1). The incidence of P. vivax malaria decreased log-linearly with age (Figure 1, Table 1, P<0.0001) from 2.9 episodes/year-at-risk at 1 year of age to a minimum of 0.6 episodes at 3.5 years of age. It peaked at the beginning of the rainy season (early December, week 49, 1.7 clinical episodes, Figure 1) and was lowest in the early dry season (early June, week 23, 0.8 clinical episodes, Figure 1). Insecticide treated net (ITN) use was not significantly associated with incidence of P. vivax malaria. The incidence of P. vivax malaria varied significantly between villages (P<0.0001) and between children living in the same village (P<0.0001).

When _molFOB was added to the model (defined as the rate of new clones acquired per year-at-risk), it was highly significantly associated with the incidence of P. vivax malaria (Table 1, P<0.0001). Adjusting for _molFOB resulted in a 45% decrease in seasonal differences in P. vivax incidence (Table 1, Figure 2). Contrary to what was observed for P. falciparum [23], adding P. vivax _molFOB did not significantly alter the association of age or ITN use with incidence of clinical P. vivax episodes. Comparable results were seen when P. vivax episodes with any parasite density were considered (Table S1).

In order to determine the association of exposure with the rate of immune acquisition we stratified the cohort in children with high (_molFOB: 18.1–39.0/year-at-risk), medium (_molFOB: 10.7–18.0/year-at-risk) and low exposure (_molFOB: 0–10.7/year-at-risk) based on their average P. vivax _molFOB. The reduction of incidence of P. vivax episodes >500/ul with age was less pronounced in the one third of children with the lowest average _molFOB (IRR = 0.63/year increase in age, p = 0.02) than those with the highest exposure (IRR = 0.49/year increase in age, p<0.001) (Figure 3). Even stronger differences were observed in associations of age with all P. vivax episodes, where no significant reduction in incidence with increasing age was found for children with low exposure (IRR = 0.93, p = 0.6), but strong reductions in children with medium and high exposure (IRR = 0.50–0.52, p<0.001, Figure 3). These differences in IRR for P. vivax episodes of all densities with age were statistically significant for children with low versus medium (p = 0.03) and high exposure (p = 0.02), respectively. Due to overall lower number of episodes with a parasitemia >500 parasites/ul (and thus reduced power), the differences between children with low and medium exposure (p = 0.40) and low and high exposure (p = 0.15) did not reach statistical significance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Association of incidence of P. vivax episodes with age in children with low (_molFOB: 0–10.7), medium (_molFOB: 10.7–18.0) and high average exposure (_molFOB: 18.1–39.0) to P. vivax infections.

Incidence rate ratio and 95% confidence intervals for changes associated with 1 year increase in age.

https://doi.org/10.1371/journal.pntd.0002403.g003

Discussion

By genotyping all blood-stage parasites detected over 16 months of follow-up, this study provides the first direct estimate of the molecular force of P. vivax blood-stage infections. Children aged 0.9 to 4.5 years acquired 14.0 P. vivax clones per year-at-risk, thus approximately twice as many as P. falciparum blood-stage clones [23]. Differences between P. vivax and P. falciparum were not only evident in the absolute number of infections, but also in the associations between _molFOB, incidence of clinical episodes and age. While P. falciparum _molFOB increased with age and thus paralleled the trend in incidence [23], P. vivax _molFOB did not change with age, but incidence of disease decreased dramatically over the age range of the cohort.

Minor differences in the typing techniques applied for these two Plasmodium species could account for some differences in the estimates. Typing was based on length polymorphic marker genes. Their diversity was high, yet slightly differed: P. falciparum was typed using msp2 (expected heterozygosity H_E = 0.933) [38], while msp1F3 (H_E = 0.881) and MS16 (H_E = 0.978) were used for P. vivax typing [28]. For all three markers, we previously determined the clone detectability, i.e. the proportion of clones detected in both of two bleeds collected 24 hours apart. Detectability differed between markers: 79% for Pfmsp2, 61% for PvMS16 and 73% for Pvmsp1F3 [30]. The overall diversity of the Pfmsp2 marker was therefore intermediate to that of the two P. vivax markers, while Pfmsp2 detectability was highest.

The most obvious difference in typing strategies was that our analysis of P. vivax was based on two loci with the combined _molFOB determined from maximal number of alleles per sample observed by any marker. Thus, the ability to detect clones of both P. vivax markers combined was higher than by the single P. falciparum marker.

To assess the effect of the use of 2 markers, P. vivax _molFOB was also calculated using a single marker only. The average _molFOB was 8.7 and 12.8 clones/year-at-risk for msp1F3 and MS16, respectively. Both of these values are substantially higher than the 5.5 clones/year-at-risk detected by Pfmsp2 genotyping. It is worth noting that a 1.6 times higher _molFOB was obtained with P. vivax marker msp1F3 compared to P. falciparum despite its lower diversity and detectability. Therefore, the differences in allelic diversity and detectability of genotyping markers can not account for the large differences between P. vivax and P. falciparum _molFOB.

Given imperfect detectability and resolution of markers plus infrequent sampling, our estimates of _molFOB of both species were likely underestimates of the true burden and complexity of Plasmodium spp. infection. However, for 97.5% (1638/1680) of all regular follow-up bleeds (except baseline and the final round), two paired samples taken 24 hours apart were available, thus the probability of missing a clone in both of two paired samples dropped to 0.07 for P. vivax msp1F3, 0.15 for P. vivax MS16 and 0.044 for P. falciparum msp2 [30] (assuming independence in the chance to detect a clone from any one sample). In addition, clones detected during passive case detection between regular visits were also included, further reducing the chance of missing a clone during an 8-week interval. This suggests that the estimates of _molFOB were biased downwards as a result of imperfect detectability only by between 6 and 13% for P. vivax and 4% for P. falciparum.

In PNG, P. vivax and P. falciparum are transmitted by the same mosquito vectors. While no entomological studies were conducted in parallel with this cohort, earlier studies in different PNG lowland populations reported comparable sporozoite rates for P. falciparum and P. vivax [6], [39], [40]. However, the likelihood that individuals become infected with multiple P. vivax clones by a single mosquito bite is high, because 75% of all P. vivax positive individuals from our cohort carried multiple clone infections [28]. It is therefore likely that a mosquito takes up different P. vivax gametocyte clones during a blood-meal, resulting in sexual recombination and the transmission of a genetically diverse inoculum to a new host. This contrasts to only 33% of multi-clone infections in the P. falciparum positive children (Schoepflin, Mueller & Felger unpublished results).

P. vivax parasites detected in the blood stream do not always derive from a recent mosquito bite; they can also result from relapsing hypnozoites. New Guinean P. vivax strains were reported to relapse rapidly, and 63% of 1 to 5 year old children had a recurrent parasitemia within six weeks after treatment of blood-stage parasites [41]. Two thirds of those recurrent infections (all post day 28) were of a different genotype [42]. Although genotyping cannot differentiate between true new infections and relapses of an unrelated genotype, P. vivax relapses likely contribute significantly to the higher _molFOB. Evidence for this comes from a recent cohort of Papua New Guinean children aged 1 to 5 years, in which the contribution of relapses to the burden of infection was assessed directly by randomising one third of children to receive anti-hypnozoite drug therapy with a 14 day course of high-dose primaquine. While the primaquine treatment was not 100% efficacious, it did nevertheless result in a 34–57% reduction in the incidence of new P. vivax infections [43].

As a consequence of relapses, _molFOB measures different processes in P. vivax and P. falciparum. P. falciparum _molFOB is directly linked to transmission intensity in a given interval, whereas _molFOB of P. vivax represents a composite measure of both transmission intensity and frequency and genetic complexity of relapsing parasites (some of which may have been acquired months or years earlier). P. vivax _molFOB is therefore not a direct measure of _molFOI (i.e. the number of new parasites acquired by the human host), but the difference between the two measures is small if the total follow-up period is substantially longer than the average relapse frequency. In the Southwest Pacific, P. vivax infections are thought to relapse very rapidly (i.e. within a few weeks [31]), thus P. vivax _molFOB is a good surrogate marker for _molFOI if calculated over the entire 16 months of follow-up.

Interestingly, P. vivax _molFOB did not change with age, while P. falciparum _molFOI was strongly age dependent and increased from 3 to 8 clones per year-at-risk over the age range of 1–4.5 years of this cohort [28]. As a consequence, _molFOI largely explained the age trend in P. falciparum incidence (increasing from 1 episode per year in children one year of age to 2.5 in children three years of age) [23]. In contrast, P. vivax _molFOB did not explain the decrease in P. vivax incidence from over 3 episodes per year in children one year of age to less than 1 episode in children older than three years.

The age shift in incidence in clinical disease caused by P. vivax and P. falciparum observed in this study – i.e. P. vivax incidence peaking in younger children - parallels earlier findings in Papua New Guinea [6], [10], [29] and other regions where both species are co-endemic [7]–[9]. This indicates a rapid acquisition of immunity to P. vivax in individuals with life-long exposure to both P. falciparum and P. vivax. As immunity to malaria builds up gradually and is thought to be strain-specific [4], [5], it is likely that the number of distinct infections acquired over an individual's lifetime is a major driving force for acquisition of immunity [20].

In their first years of life, the children in our cohort were estimated to have acquired three times more genetically distinct P. vivax than P. falciparum infections. In both species a higher _molFOB was itself associated with a significant increase in incidence of clinical disease. However, a more rapid decrease in incidence of P. vivax malaria was observed in children exposed to high compared to low levels of P. vivax infections (Figure 3). Although notable against P. vivax episodes with >500 parasites/µl, this effect of exposure on age-specific risk of P. vivax malaria was more pronounced when all P. vivax episodes (irrespective of parasitaemia) were considered. While the incidence of P. vivax malaria >500 parasites/µl decreased, even in children with low exposure, the incidence of P. vivax of any density did not change significantly with age in the children with the low exposure. This indicates that, in children with low exposure, those >3 years acquired some immunity against high-density clinical P. vivax episodes, but not against those associated with low levels of parasitaemia. Together these observations suggest that high _molFOB, and consequently the overall genetic diversity to which children were exposed in early childhood, contribute substantially to the rapid acquisition of clinical immunity to P. vivax across the entire age range [27].

The high exposure to P. vivax may even be sufficient for children to acquire a certain degree of immunity against P. vivax infection. With at least some malaria vectors, biting rates increase with host body size [44]. Therefore, as children grow older, their exposure to malarial infection also increases, as supported by the strong age-dependence of P. falciparum _molFOI [28]. The lack of an association of P. vivax _molFOB with age suggests that, in older children, some of the P. vivax sporozoites transmitted by mosquitoes do not succeed in establishing detectable blood stage infections.

Supporting Information

Acknowledgments

We would like to thank the study participants and their parents or legal guardians, and the field teams of the PNG Institute of Medical Research. We thank Thomas A. Smith for critical discussion of the manuscript.