A general SNP-based molecular barcode for Plasmodium falciparum identification and tracking

Nearly all malaria-associated mortality in children is due to infection with Plasmodium falciparum, which causes over 300 million clinical infections and a million deaths per year in African children under five years of age [1]. Genome sequencing of multiple parasite isolates [2‐5] indicates that the parasite population is highly diverse. Genetic diversity in P. falciparum is manifested in the form of single nucleotide polymorphim (SNPs), microsatellite repeats, insertions, deletions and a range of gene duplication events. Much of this diversity segregates independently. Analysis of the progeny from a genetic cross suggests that the parasite genome is approximately 50-fold more 'recombinogenic' than the human genome [2]. This genetic diversity underlies the ability of the parasite to escape both immune clearance and drug treatment.

Parasite genetic variation can be exploited epidemiologically to provide a means of uniquely identifying parasites that infect individuals and to follow these parasites through the course of infection as drug or other interventions are applied. For example, these methods are useful for identifying parasites in an individual prior to treatment to later determine if post-treatment parasitaemia are the result of failure to clear the original parasites by the drug(s) (recrudescence) or of re-infection by another parasite form. These methods are critical for determining the efficacy of drugs in the field.

The first powerful and easily deployable tool for assessing the identity (i.e., genotype) of a given parasite isolated from a patient was published in 1994 by Snounou using three polymorphic loci (merozoite surface protein (msp)-1, msp-2, and glutamine rich protein, glurp) in a nested polymerase chain reaction (PCR) to identify length polymorphisms within six potential alleles (K1, MAD20, and RO33 for msp-1; IC and FC27 for msp-2; and glurp) [3]. The technique requires gel electrophoresis, fluorescent primers in conjunction with capillary electrophoresis to resolve the length of the individual allele polymorphisms, or a Luminex fluorescent microsphere assay [4]. The practical use of this genotyping system – whole or in part – was reviewed by Collins et al for 91 of 384 studies between 1995 and 2005, focused on antimalarial clinical trials [5]. They conclude from this meta-analysis that the great deal of variation found in the use and interpretation of this system was statistically related in multivariate analysis to the polyclonality of infections, treatment employed, geographic location, and duration of follow-up [5]. Pyrosequencing has also been used to evaluate short stretches of DNA sequence from patient samples surrounding a single locus to quantify alleles and haplotypes of MSP-1 in a population [6]. A modification of the standard MSP/GLURP genotyping technique, the heteroduplex tracking assay, demonstrated an ability to track complex infections in an endemic area with high polyclonality (3.82 per patient). However, this assay currently depends on the use of radioactive tracers, making it difficult or impossible to employ in many field settings [7]. Although others have offered refinements on these basic approaches or focused on microsatellite polymorphisms for determining polyclonality, or some combination thereof, the current approaches are either labour-intensive, difficult to carry out in a field setting, or require subjective interpretation [8, 9].

As part of an ongoing effort to map genomic diversity of P. falciparum, genome sequencing has identified more than 112,000 SNPs from about 18 parasite genomes [10‐12]. Current technology has made genotyping of SNPs by real-time PCR using dual probes in an end-point detection assay a standard practice. We filtered the discovered SNPs to create a panel of genotyping assays capable of defining a "molecular bar code" or signature for a given malaria parasite. Ideal SNPs for such an assay panel segregate independently, are common (i.e. exhibit a high minor allele frequency (MAF)) and are broadly distributed across the genome. The assay method has to be easy to use, inexpensive, and applicable to a wide variety of both field and laboratory derived material. The TaqMan system was choosen as a genotyping methodology, with the eventual goal of developing these assays into a simple, end-point PCR process that could be performed in the field where a PCR machine and a plate reader were available.

This work describes the first P. falciparum molecular bar code composed of 24 SNPs that in combination create a unique fingerprint or signature for a parasite genome. This methodology can be applied to a variety of laboratory and field samples including direct culture-adapted material, genomic DNA, frozen blood from patients or filter-paper collected samples with over 99% success. This methodology is extremely sensitive, requiring only a small amount of input material. Human DNA within the sample does not interfere with the results. This molecular barcode is also capable of identifying mixtures of parasite genomes within samples that would otherwise be identified as single parasite infections by conventional msp-1 and msp-2 genotyping, and thus provides a robust, inexpensive, facile method for evaluating parasite genomes within patient samples.

Leveraging the large amount of sequence data derived from numerous parasite genomes, we identified a set of SNPs which were common among the sequenced parasites. Briefly, a set of unlinked SNPs from broadly distributed genomic locations was selected with an overall average minor allele frequency that exceeded 35% (Figure 1A and Additional Files 2 to 4). The "molecular bar code" is a collection of 24 of these markers that can be assayed simply and inexpensively using TaqMan technology (Figure 1B and Additional Files 2 to 4). For the development and evaluation of these assays, a method to quantify both parasite and human DNA within mixtures was required. Thus, a quantification probe for P. falciparum based upon the PF07_0076 gene was developed and a commercially available human quantification probe based on RNase P was used.

The parasite quantification probe corresponds to a region of the PF07_0076 gene (see Materials and Methods), which exhibits low genetic variation and is a single copy gene within a variety of parasites derived both from laboratory culture and patient samples [15]. For the purposes of standardizing the quantification, a parasite DNA standard using HB3 DNA that was used as a control for all assays in the analysis was created. Quantitation of this reference DNA sample was linear over a large quantitation range and was generally comparable to both standard pico green and spectrophotometric methods. This method has the advantage of being able to quantify P. falciparum DNA within a patient sample that contains human DNA. Thus, the qPCR method showed more consistency for assessing P. falciparum DNA contributions from sample to sample and across dilutions of samples than the fluorescent or spectrophotometric methods. To confirm that the parasite and human quantification probes worked independently, we created standard mixtures of genomic DNA from P. falciparum and human and demonstrated that quantification within the mixtures was consistent with quantification of both pure parasite and pure human DNA samples. Thus, the TaqMan quantification probe for P. falciparum was reliable and consistent both in samples containing parasite DNA alone and in samples containing a mixture of parasite and human DNA (Additional File 5).

DNA samples from 12 sequenced parasites were used to validate the original 88 candidate bar code assays (Figure 1B). Assays were discarded for poor or late amplification or lack of ability to detect either allele. From this reduced set, individual assay performance to discriminate the major and minor allele was assessed (Figure 2 and Additional File 6,7), and each of the 24 assays showed clear differentiation of the two alleles for each position. The bar code provided robust and reliable discrimination of alleles in mixed genome samples (Figure 3). To evaluate specific performance in mixtures of known ratios, DNAs from two sequenced parasites – HB3 and Dd2 – were mixed in various proportions. Assays where the major and minor alleles were distinct were assessed for their ability to detect the minor allele in the mixture (Figure 3A). In mixtures of these DNAs ranging from a 3:1, 1:1 or 1:3 ratios, the assays were clearly able to quantify the proportion of the major and minor allele within the mixture. The ability of the assays to perform in mixed genome samples was evaluated by mixing DNA from sequenced parasites at known concentrations (1:20, 1:10, 1:5, 1:1, 1:5, 1:10, 1:20) to test the ability of the assay to detect both alleles (Figure 3B and 3C). Assays unable to detect both the major and minor allele, or an individual allele representing at least 20% of the total alleles in the population were also removed from consideration (Additional File 8). The final criterion for inclusion was that the SNPs assayed in the final bar code were broadly distributed across the genome and not in linkage disequilibrium with each other. To ensure that these assays were unlinked and independent, we performed linkage disequilibrium analysis (Additional File 9), and only unlinked loci were included. The final set of 24 SNPs selected for the molecular bar code meet all of the above criteria.

The molecular bar code assays were applied to 114 independent parasite lines (see Additional File 10) to identify samples with mixed genomes and derive unique signatures for the cultured parasite lines comprised of a single genomic background. The methodology worked on > 99% of samples requiring as little as 5 ng of genomic DNA. The utility of the bar code assay for patient-derived samples was determined by assessing its performance across a variety of sample types. DNA derived from culture-derived samples, from patient-derived samples, and from experimental samples were tested in the assay. For the culture-derived samples either genomic DNA derived from the parasite or direct culture material was used. For the patient-derived samples either blood spots dried onto filter paper or frozen blood samples were used. The experimental samples were mixtures of genomic DNA (parasite and human or multiple parasites), mixtures of culture with whole human blood, or WGA material from a variety of samples. The molecular bar code was robust and reliable for all of these sample types. In general a reliable bar code genotype was detected with less than 5 ng of genomic DNA from culture-adapted or patient-derived material. WGA products from a variety of samples (extracted DNA from culture or patient samples) also reliably and reproducibly provided a molecular bar code with the minimum input of ~1 ng DNA.

The molecular bar code's utility for identifying mixed parasite samples was evaluated relative to standard MSP-1 and MSP-2 genotyping (Figure 4). MSP-1 and MSP-2 represent highly polymorphic surface molecules within the parasite that vary in length, and thus can provide a measure of whether the sample has a mixture of parasite genomes. In most cases the two methods were in agreement, but a few cases were found where MSP-1 and MSP-2 genotyping indicated that the sample contained a single genome while the bar code demonstrated a mixed genotype or conversely, a mixed genome sample by MSP-1 and MSP-2 had a single genotype by the molecular bar code method. A collection of 61 patient samples were tested and compared for their findings of "single" or "mixed" genome using MSP typing or molecular bar code methods. Of these 61 samples, 35 (57.3%) were single/single (MSP/molecular bar code); 5 were single/mixed (8.2%); 5 were mixed/single (8.2%) and 16 were mixed/mixed (26.5%) [Pearson's chi-squared value of 24.74, with a p-value of < 0.001]. In using the bar code assay to track parasites in patient samples as they are being adapted to culture, it was observed that individual parasite genomes as indicated by single alleles at each of the 24 assays in the molecular bar code could be isolated from these mixtures during subcloning. It was also observed that MSP-1 and MSP-2 genotyping can occasionally detect a second genotype with greater quantitative sensitivity than the bar code assay, presumably due to fact that MSP-1 and MSP-2 genotyping involves a nested PCR methodology. The molecular bar code provides greater power to discriminate among strains than MSP-1 and MSP-2 genotyping because many more possible alleles exist among the 24 markers than length polymorphisms within these regions, and provides a clear, objective and reproducible result.

This work describes the development of a molecular barcode assay that leverages genome-wide diversity data to create a concise and unique signature for every parasite genome. This field-deployable method employs 24 TaqMan genotyping assays of independently segregating SNPs sampled from across the genome, and requires only a small amount of easily collected material such as filter-paper-collected blood spots. The same SNPs can, of course, be assayed by any other available method. The TaqMan system was chosen for its robustness, potential to deploy with common laboratory equipment, and straightforward data interpretation.

Although this method is based upon currently available sequence data, it is inherently adaptable in that new genomic diversity information from yet-unsampled P. falciparum populations can be used to generate assays suited for strain identification in those populations in the event the current method is insufficient. Although the assays were designed based on data from Africa and Thailand, the chosen set of SNPs also performed well to distinguish parasites from Brazil. This method can also flexibly incorporate assays for phenotypes of interest, such as SNPs in pfcrt and dhfr known to directly confer drug resistance, or assays at non-polymorphic residues that quantitatively assess copy number of genes such as pfmdr1 that have been associated with some forms of drug resistance. As new markers of different phenotypes are identified through QTL or association studies, this simple method can be extended to include these new markers. Given that this method is simple, inexpensive and field-deployable, it can greatly enhance our ability to follow epidemiological or therapeutic efficacy trials. Furthermore, this simple and inexpensive assay system can be used in the laboratory to follow parasites through the culture-adaptation and subsequent subcloning processes to ask questions about selection and competition as parasites grow in mixtures.

As the molecular bar code is applied and refined it will be critical to continue to make comparisons with standard genotyping currently in the field including the MSP-1, MSP-2, GLURP combination or microsatellite genotyping. Though these initial assays are able to distinguish the relative proportions of major and minor alleles in simple mixtures, we need to improve upon the ability to detect multiple genomes within patient samples. Among culture-adapted parasite samples the molecular bar code was useful for tracking parasites during the culture-adaptation and subcloning process. Finally, this method was easily able to detect instances of cross contamination in the laboratory, thus providing a powerful quality control step within the laboratory setting.