Plasmodium falciparum and Plasmodium vivax Infections in the Peruvian Amazon: Propagation of Complex, Multiple Allele-Type Infections without Super-Infection

Abstract

Outcrossing potential between Plasmodium parasites is defined by the population-level diversity (PLD) and complexity of infection (COI). There have been few studies of PLD and COI in low transmission regions. Since the 1995–1998 Peruvian Amazon epidemic, there has been sustained transmission with < 0.5 P. falciparum and < 1.6 P. vivax infections/person/year. Using weekly active case detection, we described PLD by heterozygosity (H_e) and COI using P. falciparum Pfmsp1-B2 and P. vivax Pvmsp3α. Not being homologous genes, we limited comparisons to within species. P. falciparum (N = 293) had low (H_e = 0.581) and P. vivax (N = 186) had high (H_e= 0.845) PLD. A total of 9.5% P. falciparum infections and 26.3% P. vivax infections had COI > 1. Certain allele types were in more mixed infections than expected by chance. The few appearances of new alleles could be explained by stochastic polymerase chain reaction detection or synchronization/sequestration. The results suggest propagation of mixed infections by multiple inocula, not super-infection, implying decade-long opportunity for outcrossing in these mixed infections.

Introduction

Despite near eradication of human malaria pathogens from the Peruvian Amazon between 1960 and 1990, Plasmodium falciparum and Plasmodium vivax re-emerged in 1991 and 1994, respectively. The brunt of the epidemic occurred between 1995 and 1998, when P. falciparum malaria dominated P. vivax malaria transmission 2:1. However, as the epidemic was curbed, likely because of effective intervention efforts of fumigation combined with free and highly controlled drug treatment, P. falciparum accounted for less than one third of the malaria infections in this region (1:3 ratio of P. falciparum infections compared with P. vivax).^1,2 Since 2000, there has been sustained low transmission. Active and passive case detection from 2003 to 2004, in one of the highest malaria-endemic regions during the malaria transmission season, showed a prevalence rate of P. falciparum = 0.13 and P. vivax = 0.39 infections/person/malaria season.²

Several features of malaria transmission at this study site make it particularly interesting. First, the low transmission rates enable us to determine the malaria parasite genetic diversity in discrete (non-overlapping) infections within a relatively newly exposed population.² Second, the independent characterization of P. falciparum and P. vivax genetic diversity in a sympatric population where there is only one mosquito vector, Anopheles darlingi, allows for more direct interspecies comparisons.^1,2 Finally, malaria infections in this geographic region are surprisingly mild, with > 50% of both P. falciparum and P. vivax infections detected by active case detection being asymptomatic.² Together, the unique features of malaria in this study region may provide clues to how malaria genetic diversity is maintained in low transmission and the relationship between genetic diversity and the development of immunity and possible drug resistance.

Population-level diversity (PLD) is typically defined by examining genetic variation occurring in variable antigenic loci and/or neutral (purportedly non-antigenic) loci. Because of the low transmission in the Peruvian Amazon, it is also possible to examine PLD occurring in a single mixed-clone infection. Genotyping methods that identify one or more polymorphic, single-copy loci can be used to assess the minimum number of parasite allele-types present within a given infection; this is termed the complexity of infection (COI).^3–8 In this study, we studied the PLD and COI for P. falciparum and P. vivax independently, using the P. falciparum merozoite surface protein-1 block 2 region (Pfmsp1-B2) and P. vivax merozoite surface protein-3α (Pvmsp3α), respectively.

PfMSP1-B2 is a highly variable repeat-length region near the N terminus of PfMSP1 (PlasmoDB gene ID: PF1475w). Three allelic families (K1, Mad20, and RO33), identifiable by polymerase chain reaction (PCR) of the encoding single copy gene using respective conserved flanking sequences, have been detected globally in all transmission regions.^3–13 K1 and Mad20 allelic families can be further distilled by base pair length/size differences observed after electrophoresis because of the expansion and contraction of internal repeating motifs, whereas RO33 is monomorphic in size.^8,9 Both PLD and COI can be defined by the identification of size-specific alleles within each allelic family.

PvMSP3α also contains an internal region comprised of repeating motifs of varying length (PlasmoDB gene ID: PVX_097720).^14,15 Three allelic families have been observed after PCR using P. vivax conserved primers that broadly characterize Pvmsp3α: A = 1.8–1.9 kb, B = 1.5 kb, and C = 1.1–1.2 kb.^16–22 Diversity within these allelic families can be further exposed by digesting the PCR product with the restriction enzymes Hha I and AluI.¹⁸Pvmsp3α genotyping patterns can be compiled by using both PCR and restriction fragment length polymorphism (RFLP) to identify infecting alleles and define the PLD and COI.

High PLD (high levels of polymorphism) in both Pfmsp1-B2 and Pvmsp3α has been suspected to arise because of variant-specific immunity, resulting in diversifying selection in the population of parasites.^{14,19,21,23–26} Alternatively or additionally, the high PLD could be attributable to asexual rapid diversification of repeat motifs and modest functional constraints on the encoding merozoite surface protein repeat regions.^10,27–31 The high levels of polymorphism in these antigenic loci results in a fine characterization of infecting parasites, which explains their frequent use in malaria genetic diversity studies. Further characterization of the PLD and COI can be achieved by sequencing these regions. Determining PLD and COI of P. falciparum and P. vivax provides insight into the transmission dynamics, population structure, and potential for sexual recombination within the mosquito vector.

In this study, we examined PLD and COI using Pfmsp1-B2 and Pvmsp3α. Uniquely, we have weekly sampling to consider the changing parasite genotypes overtime. The low transmission in this Peruvian population suggests that a COI > 1 is caused by one inocula; that is, two genotypes inoculated by one mosquito versus overlapping infections (super-infection). The purpose of this study is to establish the baseline genetic diversity levels in this region of the Peruvian Amazon. We hypothesized that there would be low P. falciparum and low P. vivax PLD and COI. We propose a molecular test and discuss the evolutionary consequence of mixed infections without super-infection.

Materials and Methods

Study design

Our study was conducted in Iquitos, Peru (Zungarococha community, N = 1907).² Blood samples were collected during the malaria transmission season (January–July) in 2003 and 2004, for a total of 10 months of active case detection.² All the samples were obtained after informed consent. Details on the study site and design are described in Branch and others.² In brief, active case detection and passive case detection obtained 3,574 samples from 1,735 individuals participating in this study. Active case detection included a beginning and ending malaria season community-wide cross-sectional survey and also a selection of ∼200 individuals each month during the malaria season for weekly visits for 1 month. Additionally, there was passive case detection executed in the health center; therefore, it is possible that a detected malaria infection did not have weekly sampling available.

In the active case detection, blood slides from individuals who had a body temperature of ≥ 38.3°C, reported having a high fever within 2 days, or had hematocrit < 30% pcv, had their blood slides read by expert microscopists within 1 day and were treated within 1 day if positive. If an individual was asymptomatic, there could be 6 days before reading the microscopy slide. At the next scheduled visit (1 week later), another blood sample was collected, and any individual who was found with malaria parasites on the week before visit had their blood slide read immediately. Therefore, asymptomatic individuals had more than one blood slide collected during a given infection before treatment. Treatment regimens are reported in Branch and others.²

DNA extraction and testing for Plasmodium species

All blood samples, positive or negative by microscopy, underwent DNA extraction by Qiagen DNeasy Blood and Tissue Kits and were tested for presence of Plasmodium species using a nested-multiplex PCR method targeting DNA encoding the small subunit ribosomal protein (ssrDNA) (Qiagen Inc., Valencia, CA).³² For this species-specific ssrDNA PCR diagnosis and PCR for genotyping (described below), PCR mastermixes were comprised of primers at the indicated concentration (Integrated DNA Technologies, Coralville, IA): 2.0 mmol/L concentration of deoxynucleoside triphosphate (dNTP) mixture (Invitrogen, Carlsbad, CA); MgCl₂ at a concentration of 1 mmol/L (Promega, Madison, WI); 5× PCR buffer, Taq polymerase (GoTaq Flexi; Promega), molecular grade water, and blood extracted DNA (genomic DNA adjusted to 10–20 ng/μL) or externally amplified (PCR amplicon) DNA. All PCRs were conducted in an Eppendorf Mastercycler ep (Westbury, NY).

PCR genotyping Pfmsp1-B2

Pfmsp1-B2 genotyping was attempted on all available samples that were microscopy and/or PCR positive for P. falciparum (293 P. falciparum infection-months). Previously reported genotyping methods were used for main allelic family analysis, using all permutations of primers to detect any allele families that were one allele type in the 5′ direction and a different allele type in the 3′ direction.^8,10

PCR products for the K1, Mad20, and R033 allelic families were observed; we did not observe the Mad20-R033 recombinant (MRrec) allele.¹⁰ The amplicons from each PCR reaction were individually visualized by electrophoresis using a 1.5% concentration of UltraPure Agarose 1000 (Invitrogen). The alleles were characterized by allelic family and size of the amplified product using the Gel Doc gel analysis system (Bio-Rad Laboratories, Hercules, CA). The lengths were assigned by rounding up or down to the nearest 10 bp characterization based on the 25 bp ladder (Invitrogen, Carlsbad, CA). Final allele assignment was determined by performing electrophoresis a second time, arranging the samples in order of base pair length determined from the first electrophoresis and confirmed by sequencing. Samples with band sizes between 250 and 300 bp for both K1 and Mad20 allelic families (not shown) were found to be artifacts of the first external PCR and therefore were removed from the analysis.

PCR genotyping Pvmsp3α

The nested-PCR protocol of Bruce and others¹⁶ was used on samples P. vivax positive by microscopy and/or PCR. Primers were used at a concentration of 5 μmol/L. For the further-discerning RFLP assay, 8 μL of the PCR products was digested with HhaI and AluI enzymes, individually, overnight at 37°C (4 units of enzyme/reaction; Promega). The PCR-amplified products underwent electrophoresis on a 1.8% agarose gel (Low Range Ultra Agarose; Bio-Rad). Fragment sizes were determined using the Gel Doc gel analysis system (Bio-Rad). Additionally, gel images were evaluated by hand and compared with the software-based analysis to ensure fragment calling consistency. The resulting fragment sizes were assembled into an Excel file where there was a column for all possible sizes (one bin for every 10 bp). Each RFLP pattern was represented by having fragment sizes present/absent in the array of bins. Patterns with fragment sizes that overlapped or were within 20 bp across fragments were considered to be the same pattern.

Sequencing of Pfmsp1-B2 and Pvmsp3α

We sequenced a selection of infections (selecting without any apparent bias) representing the alleles detected for both P. falciparum and P. vivax. Using the product of the second PCR reaction (second in the two-step nested method), the amplified PCR products were isolated (Gel Extraction Kits; Qiagen). In all cases, the 5′ and 3′ nested-PCR primers were used in the sequencing reaction at a concentration of 2 μmol/L. Sequencing was performed on an ABI 3730 Genetic Analyzer (Applied Biosystems, Foster, CA).

Analysis of PLD and COI

We defined infections as one or more malaria-positive sample point(s) during the follow-up, which in many cases included time periods of the weekly active case detection. We considered all sampled time points within 1 month before and after the infection. The sampling could begin before the initial malaria positive sample or after, with an overall sampling time frame of ≤ 30 days (4 weeks of sampling follow-up planned). For this reason, infections are called “infection-months,” and this is the denominator used when considering number of infections followed and frequency of the different parasite alleles detected.

The expected heterozygosity, a measurement PLD, was calculated by the following formula: H_e = 1/(1 – N)(1– σp²_i), with p equal to allele frequency and N equal to the number of alleles detected. The H_e was used to quantify the overall genetic diversity in each species using the respective genotyping targets and also to compare the genetic diversity within each allelic family within each species. A COI > 1 could be determined by observing multiple alleles within a single sample or by concatenating alleles observed in an infection-month.

We considered the transmission level by using both the active and passive case detection prevalence determined for the whole Zungarococha community in an earlier study² and by calculating the maximum transmission rate from the active and passive case detection prevalence in these 1,735 individuals (dividing infections detected in these individuals by 1,735, with the actual rate being calculated by determining the rate of becoming infected during the active sampling). The range of these measures was 0.022–0.069 infections/person/month for P. falciparum and 0.065–0.107 infections/person/month for P. vivax. The highest transmission level was when considering the active case detection transmission rate calculation, which was 0.069 and 0.107 infections/person/month for P. falciparum and P. vivax, respectively. Therefore, to test the probability of mixed infections, we used this high-end transmission rate calculation.

Results

Plasmodium falciparum and P. vivax infections detected

Between 2003 and 2004, there were 1,735 individuals contributing to 3,574 visits, including both active case detection and passive case detection. Species-specific PCR analysis targeting the genes encoding 18 ssrDNA was conducted on all sample visits (microscopy positive and negative). Given the design of the study, with frequent follow-up sampling after infections are detected as well as active case detection visits in the absence of symptoms, it was possible that a single infection could be sampled multiple times. To allow for this possibility in our analysis, we used the term infection-months instead of infections. Infection-months were defined from the sampling protocol: individuals were followed for up to 30 days with weekly sampling over 1 month. The dynamics within this unit-month is used to describe the PLD and COI and determine whether any detection of new alleles might suggest that there was a super-infection during this infection-month.

There were 308 P. falciparum infection-months identified within 2003–2004, of which 293 were selected for genotyping in this study. Of these 293 infection-months selected, 180 were identified by microscopy through active case detection, 63 were sub-microscopic and identified by 18 ssrDNA methods through active case detection, and an additional 50 infection-months were identified by microscopy and/or 18 ssrDNA methods through passive case detection (passive case detected infection-months do not have follow-up data; Table 1).

Table 1. Plasmodium falciparum and P. vivax infections detected and genotyped between 2003 and 2004.

Species	Sample collection specification	N	No. included in genotyping study	No. with genotype present in > 1 sample of infection-month	No. with new alleles detected in follow-up
P. falciparum	Mx+ with weekly follow-up	182	180	51	2
	PCR+ (Mx−) with weekly follow-up	63	63	7	–
	Missing weekly follow-up	63	50	–	–
	Total	308	293	58	2
P. vivax	Mx+ with weekly follow-up	265	71	35	7
	PCR+ (Mx−) with weekly follow-up	97	65	5	10
	Missing weekly follow-up	390	50	–	–
	Total	752	186	40	17

Mx+ = microscopy positive; Mx− = microscopy negative.

For P. vivax, there were 752 infection-months identified within 2003–2004, of which 186 (∼25%) were selected for genotyping in this study. Of these 186 infection-months, 136 were identified through active case detection and 50 were identified through passive case detection. Seventy-one of the 136 active case detected infection-months were identified by microscopy and 65 were sub-microscopic and identified by 18 ssrDNA methods. The 50 passive case detected infection-months were detected by microscopy and/or 18 ssrDNA methods (Table 1).

Of those infection-months that had multiple sample points, new alleles were detected in the successive sample points in only 2 of the 243 P. falciparum (0.8%) and 17 of the 136 P. vivax (12.5%) infection-months where there were multiple sample points. Most P. falciparum and P. vivax infection-months were single allele infections and/or did not have new alleles detected over time; as will be described in detail below, this could be explained by stochastic PCR detection, allele synchronization/sequestration, or, in the case of P. vivax, relapse.

Pfmsp1-B2, population diversity

The K1, Mad20, and RO33 allele families that have been detected globally in prior studies were detected in Zungarococha. We observed two K1 allele types (170 and 195 bp), two Mad20 allele types (200 and 210 bp), and one R033 allele type (140 bp). There were a total of 322 parasite allele detections: 202 (62.7%) K1, 112 (34.8%) Mad20, and 8 (2.5%) RO33. The most frequent K1 allele type detected in this population was K1-195 (N = 176; 54.7%); K1-170 was observed less frequently (N = 26; 8.1%). Of the Mad20 allelic family, the Mad20-210 allele type was most frequent (N = 87, 27.0%), whereas Mad20-200 was observed less frequently (N = 25, 7.8%; Figure 1).

Figure 1.

Frequencies of Pfmsp1-B2 allele types. Shading is used to discriminate between individual alleles within each bar on the graph. Individual allele frequencies are shown, with allelic families on the x-axis and respective frequencies on the y-axis.

To quantify the PLD, H_e was calculated for the entire population (including all three allelic families) and within each allelic family. When considering the entire population, H_e was 0.581, whereas individually, the H_e within K1, Mad20, and RO33 allelic families was 0.221, 0.283, and 0.000, respectively.

To consider accuracy/resolution of the genotyping results, we sequenced 11 K1-170 (GenBank accession numbers: FJ612009–FJ612019), 18 K1-195 (FJ612020–FJ612037), 2 Mad20-200 (FJ612038 and FJ612039), 24 Mad20-210 (FJ612040–FJ612063), and 2 RO33-140 (FJ612064 and FJ612065). The sequences coincided with the genotyping results; there was only one polymorphism, a synonymous nucleotide substitution within a Mad20-210 variant at position 101 (FJ612041).

Pfmsp1-B2, complexity of infection

Of 293 genotyped P. falciparum infection-months, there were 265 (90.5%) single genotype infections and 28 (9.5%) mixed genotype infections (Table 2). Of the mixed infections, 4 (1.4%) were mixed alleles of the same allelic family and 24 (8.2%) were mixed alleles of at least two allelic families. The COI was equal to 1 in 90.5% of infection-months, COI = 2 in 9.2% of infection-months, and COI = 3 in 0.3% of infection months.

Table 2. Frequencies of single and mixed P. falciparum alleles.

	Pfmsp1-B2 alleles	No. of infection-months (N = 293)	Allele frequency	COI
Single Genotype infections	K1-170	22	0.0751	90.5% single infections
	K1-195	153	0.5222
	Mad20-200	15	0.0512
	Mad20-210	74	0.2526
	RO33-140	1	0.0034
Mixed genotype infections	K1-170, K1-195	2	0.0068	9.5% mixed infections
	K1-170, Mad20-200	1	0.0034
	K1-170, Mad20-210	1	0.0034
	K1-195, Mad20-200	7	0.0239
	K1-195, Mad20-210	8	0.0273
	Mad20-200, Mad20-210	2	0.0068
	K1-195, RO33-140	5	0.0171
	Mad20-210, RO33-140	1	0.0034
	K1-195, Mad20-210, RO33-140	1	0.0034

Certain mixed infections were detected at a rate more than expected by super-infection based on the allelic frequencies of the single counterparts as a factor of the transmission rate (Table 2). The Mad20-200 allele type was observed 24 times in this population; of those, it was observed mixed with the K1-195 allele-type in seven instances. The expected rate of a K1-195 + Mad20-200 mixed infection occurring by super-infection would be 0.0018 (or [(0.5222 × 0.0512) × 0.069]), respectively. The observed rate of these two alleles occurring concurrently in the same infection-month was 0.0239, which was significantly greater than the expected rate of two separate inoculating events (P < 0.01, Fisher exact test). Similarly, of the eight times R033 was observed in this population, six times was in a mixed infection with K1-195. The expected rate for a K1-195 + RO33 mixed infection occurring by super-infection would be 0.0001 (or [(0.5222 × 0.0034) × 0.069]); yet, this combination was observed in 0.0204 of the infections, which was significantly greater than expected (P < 0.01, Fisher exact test). Overall, inter-family mixed infections (mixed alleles of different allelic families) were observed 8.64 times more frequently than intrafamily mixed infection (mixed alleles of the same allelic family).

Pvmsp3α population diversity

Initially, we identified allelic families by size differences in the undigested PCR product. Further characterization was made by RFLP digestion of these amplified products with HhaI and AluI digestive enzymes. Genotyping results (allele types) were described by a combination of both PCR and RFLP pattern codes: “ Allelic Family_Hha I pattern_Alu I pattern ” (Figure 2A and 2B). By this method we observed all three Pvmsp3α allelic families found globally (A = 1.9 kb, B = 1.5 kb, C = 1.2 kb). There were 198 of 222 allele detections (from 168 of 186 infection-months) where the PCR-RFLP combination pattern (Allelic Family_Hha I pattern_Alu I pattern) suggested a single, discernible allele present. These 198 alleles were used when considering the allele frequency (Table 3). We did not include patterns suggesting multiallele detections (the 24 allele detections shown in bold in Table 3) when we could not discern the components of mixed infections.

Figure 2.

RFLP patterns by (A) HhaI and (B) AluI. Represented are the fragment patterns as found by observing RFLP fragments on an electrophoresis gel and placing the fragment patterns in 20 bp bins for comparison. Patterns were described as waves, where the bands were detected. If a pattern was of bands that overlapped within 1 SD of another pattern in more than three waves, it was considered the same. *It was explained by a mix of HhaI Patterns 5 and 6. AluI was less allele-type discriminating because of lower size range and variation; the lines separating Waves 5–7 were removed because of overlapping patterns. Hatch marked blocks are where there were two fragments in the same 20 bp bin.

Table 3. Frequencies of single and mixed P. vivax alleles.

Pvmsp3α alleles						No. of infections N = 186)	Allele frequency	COI
Single genotype infections					A_1_12	1	0.0054	73.7% single infections
					A_3_11	2	0.0108
					A_3_12	1	0.0054
					A_3_6	9	0.0484
					A_3_7	1	0.0054
					A_4_1	1	0.0054
					A_4_13	1	0.0054
					A_4_4	21	0.1129
					A_4_5	1	0.0054
					A_4_6	1	0.0054
					A_5_1	31	0.1667
					A_5_10	1	0.0054
					A_5_15	1	0.0054
					A_5_5	1	0.0054
					A_5_8	8	0.0430
					A_6_14	1	0.0054
					A_6_2	1	0.0054
					A_6_4	32	0.1720
					A_6_7	1	0.0054
					B_1_12	18	0.0968
					C_1_14	3	0.0161
Mixed genotype infections appearing at one sample point			A_5_1	+	A,B_5_4	1	0.0054	26.3% mixed infections
	A_6_4	+	A_4_4	+	A,B_2_12	1	0.0054
			A_5_8	+	B_6_12	1	0.0054
			A_5_1	+	A_10_13	1	0.0054
	A_5_6		A_5_1	+	A_6_4	1	0.0054
			A_5_1	+	A_6_4	8 *	0.0431
	A_5_14	+	A_5_1	+	A_6_4	1 †	0.0054
	A_6_11	+	A_6_4	+	A_5_1	1 †	0.0054
					A,B_5_3	1	0.0054
					A,B_5_4	1	0.0054
					A,B_6_4	1	0.0054
					A_9_1	2	0.0108
					A_9_15	4	0.0215
					A_9_4	1	0.0054
					A_10_5	3	0.0161
					A_10_9	1	0.0054
					A_11_1	1	0.0054
					A_12_1	1	0.0054
					A_13_1	1	0.0054
					B_6_12	1	0.0054
Mixed genotype infections observed caused by multiple sampling over time			A_5_1	+	A_3_11	1	0.0054
			A_5_1	+	A_3_6	1	0.0054
			A_3_13	+	A_3_6	1	0.0054
			A_4_4	+	A_3_6	1	0.0054
			A_5_8	+	A_3_6	1	0.0054
			A_8_4	+	A_4_4	1	0.0054
			A_5_8	+	A_4_4	2	0.0108
			A_6_4	+	A_5_1	1	0.0054
			A_6_14	+	A_6_4	1	0.0054
			A_6_7	+	A_6_4	1	0.0054
			B_1_12	+	A_6_4	2	0.0108
			C_1_14	+	C_1_4	1	0.0054
			C_7_4	+	C_3_4	1	0.0054
	A_6_4	+	A_5_8	+	A_4_4	1	0.0054

Bold is used to show known mixed infections or infections mixed by the third determinant.

^*Our analysis assumes that the observed A_10_1 and A_10_4 genotypes (infections classified mixed by the third determinant in the study) are a mixed infection of A_5_1 + A_6_4.

^†A_6_4 or A_5_1 was added in place of A_10_1.

In the 168 infection-months considered in the Pvmsp3α allele frequency calculation (excluding 18 individuals who had a RFLP pattern that could not be discerned to a single allele type), there were 172 (86.9%) A-type, 20 (10.1%) B-type, and 6 (3.0%) C-type alleles found (number exceeds total because there were identifiable mixed infections; Figure 3). In this analysis, significant intrafamily diversity was found within the A-type allelic family (23 alleles), whereas little diversity was observed in the B- and C-types (1 and 3, respectively).

Figure 3.

Frequencies of Pvmsp3α allele types. Shading is used to discriminate between individual alleles within each bar on the graph. Individual allele frequencies are shown, with allelic families on the x-axis and respective frequencies on the y-axis. Frequencies exclude alleles that are part of indiscriminable mixed infections.

As with Pfmsp1-B2, the H_e was calculated to determine the overall PLD and the PLD within each allelic family. Because the HhaI fragments were more discernible, we calculated H_e based solely on the Allelic Family_Hha I combination. The overall H_e observed within the population was 0.845, whereas individually, the H_e values were 0.808 and 0.600 for the A-type and C-type allelic families, respectively.

To determine how the 44 PCR-RFLP genotype combinations reflected sequence differences, we sequenced 68 samples representing 14 of the 44 different allele types. Of the 68 samples, there were 40 different sequences observed (FJ612066–FJ612105), where the presence of at least one non-synonymous substitution defined a sequence as “unique.” Sequences were correlated to the original PCR-RFLP allele types observed by the selection of eight regions of variability (Figure 4 Supplementary Data Figure 1 available online at www.ajtmh.org). Of the 40 sequences, there were eight distinct groups. These eight sequence groups were well described by the PCR-RFLP study (six sequences could not be included because of incomplete sequencing data). If a sequence within a group defined a unique allele type or if a genotype defined a sequence, this was considered concordant. The PCR-RFLP genotype combination and sequencing data was 84.9% concordant (Figure 4; Supplementary Data Figure 1), indicating that > 80% of the variation within the sequence is reflected in our PCR-RFLP pattern. Although sequencing methods provide a more accurate assessment of the PLD present in Pvmsp3α, this PCR-RFLP genotyping method allows for the differentiation between single genotype and mixed genotype infections, which cannot be resolved by sequencing methods alone.

Figure 4.

Regions of Pvmsp3α sequence diversity correlated to PCR-RFLP genotyping data. Groupings were made based on a similarity of 90% between each region. Varying shades of gray within these PCR-RFLP patterns indicate relatedness within the group. Group numbers are given such that the smallest numbers represent the most frequent sequence code in the region. Differences of a single amino acid within a group are denoted by the grouping number followed by a lowercase letter. Numbers in bold indicate a difference greater than a single amino acid change. GenBank accession numbers: FJ612066–FJ612105.

Pvmsp3α complexity of infection

In total, there were 137 (73.7%) single genotype infections and 49 (26.3%) mixed genotype infections (Table 3). Of the mixed infections detected, 11 (5.9%) were due to the detection of more than one allelic family (determinant 1), 23 (12.4%) were due to alternating alleles concatenated into the same infection-month (determinant 2), and 15 (8.07%) were mixed as a result of the summation of RFLP fragments exceeding the allelic family size by 2 SD (determinant 3). A COI = 1 was observed in 73.7%, a COI = 2 in 15.6%, and a COI = 3 in 2.7% of the infections (frequency excludes infections where allele types could not be surmised).

There were some alleles detected together in mixed infections that were observed in this population more than expected by super-infection. Of the 31 times that A_5_1 and the 32 times that A_6_4 were detected in this population, there were 12 infections where they were observed mixed. The expected rate of A_5_1 + A_6_4 occurring by super-infection (where allele frequency is a factor of the transmission rate ([0.1667 × 0.1720] × 0.107) was 0.0030. The observed rate of A_5_1 + A_6_4 mixed infections occurring in this population was 0.0645, which is greater than expected by super-infection (P < 0.0002, Fisher exact test). Intrafamily mixed infection-months were observed 5.10 times more frequently than inter-family mixed infection-months.

New alleles detected over time

There were 58 P. falciparum infection-months that could be evaluated based on more than one genotype result taken over multiple sampling points (Table 1). Forty-eight (82.8%) of these had a COI = 1, 8 (13.8%) has a COI > 1 with no new allele types detected over time, and 2 (3.5%) had a COI > 1 with new allele types detected over time (Table 4, left). Of the 40 P. vivax infections we could evaluate over time, there were 19 (47.5%) that had a COI = 1 (Allelic family_Hha I PCR-RFLP patterns), 4 (21.1%) with a COI > 1 with no new allele types detected over time, and 17 (42.5%) with new allele types detected over time (Table 4, right).

Table 4. Analysis of successive genotypes over infection-months for P. falciparum (left) and P. vivax (right).

New Allele? When Tx?	P. faIciparum successive pfmsp-1B2 gerotyped FU points													P. vivax succcssive Pv msp-3α gcnotypcd FU points
	Part Anon ID#	Genotype FU1			Genotype FU2			Genotype FU3			Genotype FU4			Part Anon ID#	Genotype FUl		Genotype FU2		Genotype FU3		Genotype FU4
		Kl	M20	R0	Kl	M20	R0	Kl	M20	R0	K1	M20	R0		Allele 1	Allele 2	Allele 1	Allele 2	Allele 1	Allele 2	Allele 1	Allele 2
NO, Tx at 1st FU	0001	195			195			195						0001	A_5_8		A_5_8
	0002	195			195									0002	B_1_12		B_1_12
	0003	195			195
	0004	195			195
	0005	170			170
	0006		210			210
	0007		200			200			200
NO, Tx later b/c ASYM at 1st FU	0008	195			195									0003	A,B_5_4		A_5_l
	0009	195			195									0004	A_4_4		A_4_4
	0010	195			195									0005	A_5_l		A_5_l
	0011	195			195									0006	A_6_4		A_6_4
	0012	195			195									0007	A_6_4		A_6_4
	0013	195			195									0008	A_6_7		A_6_4
	0014	195			195									0009	A_6_4		A_6_4		A_6_4		A_6_4
	0015	195			195									0010	A_3_13		A_3_6
	0016		210			210								0011	C_1_14		C_1_4
	0017		210			210
	0018		210			210
	0019	195	200			200
	0020	195	200		195	200
	0021	195	200		195				200
	0022	170	200		170	200			200
	0023	195	210	140		210	140
	0024	195	200		195			195				200
NO, Tx later b/c PCR at 1st FU	0025		210			210			210					0012	A_5_8		A_5_8
	0026		210			210			210					0013	A_6_4		A_6_4
	0027		210			210								0014	B_1_12		B_1_12
	0028		210			210			210					0015	B_1_12		B_1_12
	0029		210			210			210			210		0016	A_6_4		A_6_4		A_6_14
	0030	195	210		195				210
	0031	195			195			195
	0032	195			195			195
	0033	195		140	195		140	195		140	195
	0034	195			195			195
	0035	195			195			195			195
	0036	195			195			195
	0037	195			195
	0038	195			195
	0039	195			195
	0040	195			195
	0041	195			195
	0042	195			195
	0043	195			195
	0044	195			195
NO, selflimited	0045		210			210			210					0017	A_5_l	A_6_4	A_6_4		A_6_4
	0046	195			195									0018	A_6_4		A_6_4
	0047	170			170									0019	A_10_13		A_5_l
	0048	195			195									0020	A_5_1	A_6_4	A_5_1
	0049	195			195									0021	A_5_1		A_5_l		A_5_1
	0050	195			195			195			195			0022	A_5_l		A_5_1
	0051	195			195			195						0023	A_6_4		A_6_4
	0052	170			170
	0053		210			210
	00J4		210			210
	0055		210			210
	0056		210			210
YES, Tx at 1 st FU	0057		210		195									0024	A_4_4		A_8_4		A_4_4
YES, Tx later b/c ASYM at 1st FU														0025	A_6_4		A,B_2_12		A_4_4
														0026	A_5_6		A_5_1	A_6_4	A_5_1	A_6_4
														0027	A_6_4		A_5_l
														0028	B_1_12		B_1_12		A_6_4
														0029	A_5_8		A_4_4
														0030	A_4_4		A_6_4		A_6_4		A_5_8
YES, Tx laterb/c PCR at 1st FU	0058		210		195	210		195	210					0031	A_3_l1		A_5_1
														0032	A_5_8		A_4_4
														0033	C_7_4		C_3_4
														0034	A_3_6		A_5_l
														0035	A_5_8		A_3_6
														0036	A_5_8		B_6_12
														0037	B_l_12		A_6_4
														0038	A_4_4		A_4_4		A_3_6
														0039	A_5_4		A_5_14		A_5_l	A_6_4
														0040	A_6_11		A_6_4		A_5_1	A_6_4

Tx = treatment; Part. anon ID no. = participant anonymous ID number; ASYM = asymptomatic (no reported or measured fever > 37.5°C, hematocrit > 30%, parasitemia < 5,000/mL); M20, Mad20; R0, RO33; FU = follow-up.

We hypothesized that the detection of new alleles over time was attributable to a multiple inoculation event from one mosquito, resulting in the stochastic PCR detection of alleles/alternating alleles based on parasite density rather than superinfection. To test this, we calculated the expected mixed genotype infection frequency occurring by super-infection, using the formula ∑_i[p (1 – p) × x], with p equal to allele frequency of the ith allele and x equal to the maximum transmission rate observed in this study (P. falciparum = 0.069; P. vivax = 0.107). The expected frequency of super-infection for P. falciparum and P. vivax was 4.24% and 9.01%, respectively. Testing the hypothesis of no super-infection (0), neither the P. falciparum nor the P. vivax expected frequency of super-infection was significantly different from zero (P = 0.24 and P = 0.12, respectively, Fisher exact test). The possibility of relapsing infections in P. vivax does not limit our conclusion that, over the infection-month, the appearance of new alleles by super-infection is expected to be rare.

Discussion

The low transmission rates of both P. falciparum and P. vivax (0.13 and 0.39 infections/person/month, respectively) in the Peruvian Amazon prompted us to test the hypothesis that there would be low PLD and COI in both species. Here we independently compare the PLD and COI within P. falciparum and P. vivax using the polymorphic antigens Pfmsp1-B2 and Pvmsp3α, respectively. This genetic diversity study was not meant to compare PLD between infecting species as in Ord and others,³³ which used homologous apical membrane antigen-1 (AMA1) encoding genes to elegantly characterize and compare the PLD between species in the mesoendemic Venezuelan Amazon. Although Pfmsp1-B2 and Pvmsp3α are not homologous encoding genes, we used these markers because of their frequent use in earlier studies and their shown value as highly polymorphic markers that make fine distinctions between different parasites to define allele types. By defining the PLD and COI of P. falciparum and P. vivax species, correlative studies can be performed by drawing on this reported diversity as a baseline; for example, it provided the foundation for the future characterization of PLD and COI using neutral (purportedly non-antigenic) loci markers.

Using Pfmsp1-B2 as a marker for genetic diversity in P. falciparum, we quantified PLD and found low levels of genetic diversity within the entire population (H_e = 0.581) and within each allelic family (K1 = 0.221, Mad20 = 0.283, and RO33 = 0.000). The Pfmsp1-B2 diversity was significantly less in this Peruvian cohort than has been observed in hyper/holoendemic geographic regions, such as Senegal, ⁷ Papua New Guinea, ³ and western Kenya.⁸ For example, in western Kenya, during a time with > 300 infected P. falciparum bites/person/night, there were as many as 21 P. falciparum allele types defined by Pfmsp1-B2 at a population-level and up to 6 Pfmsp1-B2 allele types detected in one infection.⁸ Other studies in South America have reported similar or even lower PLD than observed in this study.^9,12–14,34 Recently, increased Pfmsp1-B2 diversity was reported in a collaborative study using some samples from our cohort.¹² However, the results from this collaboration are not directly comparable because the previous study 1) included samples from different communities and years and 2) only reported genetic diversity based on sequencing methods. Defining PLD solely by sequencing eliminates the possibility of detecting mixed genotype infections and may result in the exclusion of rare alleles that are found paired in the population, such as RO33-140 paired with K1-195. In our study, sequencing was used to further characterize the genetic diversity on a codon level and not as the primary method of allele detection.

We found high overall P. vivax PLD, quantified by expected heterozygosity (H_e = 0.845) and similarly high levels of intra-family PLD (A-type = 0.808, C-type = 0.600) with the exception of the B-type allelic family (H_e = 0.000). Our results were consistent with other Pvmsp3α PLD studies.^16–22 Relatively high PLD is expected within this polymorphic marker even in hypoendemic regions; from non-synonymous amino acid substitutions to the expansion and contraction of the sequence by inclusion or exclusion of minor indels, increased diversity in Pvmsp3α is a consequence of balancing selection or slipped-strand mismatch repair during replication.²¹ In hyperendemic regions like Papua New Guinea (PNG), Bruce and others¹⁶ reported that there were as many as 11 different alleles types detected in only 12 infections (133 bites/person/year) by a PCR-RFLP pattern defining method. More recently, Ord and others reported nine different Pvmsp3α alleles from 131 infections in the mesoendemic Venezuelan Amazon using sequencing methods. In our study, we found 44 different allele types from 186 infection-months defined by the PCR-RFLP method and were even able to identify many complex infections. Sequencing confirmed the high P. vivax PLD and also showed additional genetic variation, permitting the identification of sub-alleles. Based on sequence results, we were able to cluster the 34 strains into eight different groups based on relatedness. Ord and others²¹ suggested that A-type alleles are the putative progenitors of the smaller B- and C-type alleles. Based on the correlative PCR-RFLP/sequencing methods in this study, we clearly observed relatedness between these allelic families (Figure 3) and submit that perhaps increased selection on A-type alleles decreases selection on the B- and C-type alleles, explaining their lack of diversity.

The few detections of new genotypes over the month sampling in this study might be attributable to 1) a new infection (super-infection), 2) relapsing infection, or 3) PCR not detecting all alleles that were inoculated in one mosquito bite (stochastic PCR results or alternating alleles caused by parasite densities in the host at the time of sample collection).^18,35,36 The increased frequency of new alleles detected over time in P. vivax might be attributable to infection by relapse, despite treatment with Primaquine. However, because of 1) the low transmission rate (P. falciparum = 0.13 and P. vivax = 0.39 infections/person/malaria season), 2) high frequency of single genotype infections (P. falciparum = 90.5%, P. vivax = 73.7%), 3) the observation of allelic pairs appearing together more frequently that expected by chance, and 4) the few observations of new alleles introduced over time (P. falciparum = 2/58, P. vivax = 17/40), we predict that mixed genotype infections are likely the result of a multiple inoculation event. We calculated the expected rate of super-infection causing mixed genotype infections in this population based on the observed allele type frequencies and transmission rate; this calculation was even biased toward the null hypothesis by using the maximum observed transmission rate in this study. We found that the mixed genotype infections in this study are likely propagated by a single multiple inoculation event rather than by super-infection. Further study of the propagation of mixed genotype infections by a single inoculation event is necessary to confirm the intermittent role of super-infection in this region. Our future studies will investigate the homogenization of genomes as a consequence of alleles existing together over time and the role of natural selection on the alleles within these mixed infections.