Evolving as a Haploid Genome

The present findings underscore the extraordinary capability of the parasite to evolve during a human infection as a haploid asexual population. In nature, during a single human infection, a few hundred parasites entering the liver expand successfully to become many billions in the face of both drug and immune pressure. Once established in the blood, the parasites can increase and decrease in waves even without a reinfection. In order to evolve during these expansions, haploid parasites must do so with minimal damage to their genome. Similarly to what was first proposed for bacteria [36], the initial random sampling of duplications in the malaria genome under selective pressure serves as an effective first step to locate and identify genetic targets for resistance and generates enough of a foothold for the haploid parasites to proliferate under lethal pressure. The randomness of the initial duplication step in this organism is evident in our detailed molecular characterization of independently-selected resistant parasites from round 1 selections. In addition, these early events also capture the large size of amplicons that are initially sampled (Fig. 2). Assuming one duplicated region of approximately 50–100 kb per parasite, in principle, it is possible to cover the entire 23 Mb P. falciparum genome with a few hundred parasites. However, this is clearly not the whole story: the large parasite populations of roughly a million cells required for a successful DSM1 resistance event (Table S1) points to possibly extensive number of “trial duplications” that are non-productive or even lethal to the parasite. The success rates of about 1∶10,000,000 from round 1 selections against this completely novel evolutionary challenge (DSM1) are similar to a previous semi-quantitative estimation of the initial amplification rate in this organism that were inferred from challenges with a traditional aminoquinoline class of antimalarials in clinical use [37].

The few parasites that can identify a productive locus by chance in the first step then rely on a second more efficient step to achieve evolutionarily more robust levels of resistance (Fig. 5, Step 2). Based on survival numbers from round 2 selections (Table S8), this second step appears at least 100-fold more efficient once pseudo-polyploids have been established around a high priority locus. This second process also allows continual fine-tuning of amplicon unit numbers based on the level of antimalarial pressure (Fig. 4). For a haploid blood-stage parasite, when necessary, pseudo-polyploidy could even allow for the safe introduction of point mutations within the amplified region before amplicon units decrease to single copies (Fig. 5, Long Term). Indeed, during laboratory selections, amplifications of the target gene often are observed alongside point mutations in the same gene ([22], [28], [35] and our unpublished observations).

Both during in vitro selection and in natural human infection, these productive genomic alterations must take place independent of meiosis. Meiosis is the stage of the life cycle where “textbook” chromosomal crossover mixes different genomes from coinfections to bring together beneficial new traits and to remove damaged DNA in the progeny. However, the sexual stages at which meiosis occurs are not available to the parasite until the transmission of gamete stages to the mosquito. Prior to this stage of the life cycle, how does the evolving haploid parasite avoid large collateral damage as it is under pressure to change in the human? Our detailed characterization of clones from carefully-controlled independent experiments reveals a powerful evolutionary strategy to make precise changes in its genome while expanding in the human, away from the mosquito. At its core, the strategy involves the creation of a single significant new genetic amplification in an individual parasite, even as the entire genome is being sampled by a large starting population. Through controlled laboratory experiments, we directly observed that the amplicon responsible for resistance was the only new amplicon in every individual successful DSM1 resistant parasite. By avoiding adventitious new amplicons elsewhere in the genome, collateral damage is minimized during a time when meiotic cleansings are not available to the parasite. This precise genetic modification is not without cost: every event is accompanied by millions of parasites that do not amplify a useful portion of the genome and do not survive. Whether the initial rearrangements are occurring continuously during the life of the parasites or only in response to stress is a question that remains to be answered.

Benefits of an AT-rich Genome

The first step in the generation of the DHODH amplicon was clearly mediated by stretches of polyA sequences or polyT sequences (Figs. 5 (Step 1) and S5). Previously, similar homopolymeric A/T tracks have also been identified at the borders of naturally-occurring P. falciparum amplicons on chromosome 5 [32], [33], [38], solidifying the relevance of the current laboratory based observations to naturally occurring genomic amplifications in this organism. The A/T-based strategy revealed by these data is uniquely matched with the high AT content of the P. falciparum genome, which averages 81% AT but can reach upwards of 90% in introns and intergenic regions [39]. Exactly such approaches are probably not utilized by other Plasmodium species that cause human malaria or by other protozoan parasites. Of note, the genomes of the haploid blood stage of P. vivax, the second most prevalent human malaria species worldwide [40], averages ∼60% A/T content [41] and Leishmania species that are prone to drug resistance and gene amplifications average ∼40% A/T content [42], [43].

This A/T-dependent approach likely applies to many successful evolutionary selections of different P. falciparum parasites; genomic amplifications have been observed during the characterization of both lab-adapted and field-derived parasites from various regions of the world [15], [33], [34], [35], [38]. In the previous studies, however, the exact mechanistic origin of the genomic rearrangements was often ambiguous. First, amplicons were generated in response to antimalarials in clinical use, and independent founder events could not be distinguished from later rearrangements. Second, in some cases, parasites were isolated from clinical infections and thus information on both the clinical drug pressures and the life history of the parasite leading to observed mutational patterns (including passage through a mosquito and recombination with other genotypes) were lost. In the present study, since the DHODH amplicons were selected entirely in the asexual blood stage of P. falciparum, we can definitively conclude that the A/T track-mediated step is important in the initial acquisition of a new amplification and not in changing or rearranging amplicons later in evolution.

Potential Recombination Mechanisms

In addition to showing a general strategy of how a population of parasites narrows in on a resistance-conferring DNA locus, data from the present study points to the importance of two distinct biochemical processes that must operate in each parasite for overall evolutionary success. During replication, A/T tracks are known to cause polymerase pausing due to the rigid bend of the DNA structure [44], [45]. Events that follow could include the creation of a double strand break and recognition by a DNA repair pathway. Alternatively, the rigidness of A/T tracks may prevent adequate histone interactions, leaving DNA open to proteins that may trigger recombination pathways [46]. Recombination pathways generally require large regions of homology to mediate strand invasion but shorter stretches of repetitive bases have also been implicated in the initiation of various mitotic DNA rearrangements [47], [48], [49]. Recent studies of E. coli under stress also implicate very short G-rich sequences in template switching between stalled replication forks that leads to the duplication of large genomic regions [50]. In addition to a microhomology-mediated recombination pathway that repairs DNA breaks, a similar replication-based mechanism has been implicated in the generation of complex genomic rearrangements in yeast and humans [49], [51], [52]. Both of these processes appear to get by with extremely small stretches of homology (<10 bp), which are significantly shorter than the A/T tracks observed at the borders of P. falciparum amplicons (∼30 bp, Fig. S5) [32], [33], [38]. A/T tracks as large as 60 bp are estimated to make up ∼5% of the parasite genome [31], [39], [53] and although these sequences may take part in template switching or microhomology-mediated recombination as in other organisms, their significant length could also be enough to trigger more canonical recombination pathways requiring as little as 50 bp of homology [54].

Amplification Verses Point Mutation

Our selection studies with DSM1 show a clear preference for pathways that generate DHODH amplifications even though point mutations have been shown to prevent DSM1 binding to a recombinant catalytically active version of DHODH [55]. Given that round 2 clones display a broad range of DSM1 resistance (Table S9), we had to be sure that hidden point mutations (either in the DHODH gene itself or at other locations in the genome) were not contributing to survival in the presence of high levels of DSM1. Based on the very deep coverage of our whole genome sequencing studies (Table S3, Fig. S4), we are confident that even low frequency mutations within large amplicons would be detected. A few additional observations suggest that resistance is truly due to DHODH amplification and there are no other undetected causal mutations in the DSM1 resistant genome: 1) parasites maintain sensitivity to a number of additional antimalarials (Table S11) indicating that they are not employing a pleotropic resistance mechanism such as drug efflux, and 2) EC₅₀ against DSM1 and DHODH copy number decrease in a parallel fashion (Fig. 4), which emphasizes the contribution of the chromosome 6 amplicons to the resistance phenotype (as opposed to changes in other regions of the genome). Despite our confidence in the sequencing data, we cannot rule out that variations in amplicon sizes, and related physiological effects, contribute to the relationship between amplicon copy number and drug resistance.

Why are genome amplifications favored over the acquisition of point mutations in the DSM1 model? The ease with which one can find the correct locus that confers drug resistance and the lack of severe penalties for expanding copy numbers in the neighborhood of the DHODH gene may allow the gene amplification path to dominate. Additionally, the pharmacodynamics of drug exposure during selections could also play a role in favoring amplifications over mutations. Continual, unrelenting drug pressure demands an immediate sustained solution from the parasite population with little tolerance for wrong guesses. Although there is a measurable fitness cost of maintaining many amplicons in the absence of drug pressure (Fig. S7), parasites thrive following the increase in copy numbers of dozens of genes by an order of magnitude. Intuitively, intermittent cycling of increasing antimalarial levels, as is applied in many in vitro selection systems, may provide parasites with a chance to acquire mutations that confer high level resistance, and possibly even lose a relevant amplicon that had served its purpose in the early stages of resistance evolution. Beyond this, the nature of the drug, the nature of the target, and its location in the genome could all contribute to the optimum pathway to resistance since continual, uninterrupted application of some antimalarials during laboratory selections (as utilized in our selection scheme) has successfully generated point mutations in various target genes ([19], [23] and additional unpublished work).

Memory and Generality

The present laboratory-controlled studies show that, in the absence of drug pressure, malaria parasites lose extra copies of amplicons. However, as often seen with field-derived amplicons (Table S13 and [56]), malaria parasites do not always revert back to single copies of the target gene but instead retain a low number of amplicons in the absence of drug pressure (Fig. 4). This act has important implications for future survival: when a parasite population encounters drug pressure that it has successfully overcome before, the population is poised to rapidly re-amplify relevant amplicons quickly and efficiently without heavy collateral damage associated with A/T-based reshuffling between genes near the target.

While the evolution of malaria parasites is studied here in the context of drug resistance as a selection force, the versatile parasite-specific mechanisms that are used to achieve evolutionary success must help the parasites deal with a diverse set of challenges. In the forward direction, acquisition of appropriate beneficial amplifications could help parasites survive antimalarial drugs but also other potential challenges such as host immunity [57]. It may not be a coincidence that the liver stage expansion of an incoming parasite first allows a few hundred parasites to expand to about 100,000 to a million parasites before the population faces unexpected immune-reactions or unusual erythrocyte genotypes of the human patient. In addition to a gain of genetic material through asymmetric recombination, the reverse direction could also have public health relevance. For example, deletions of specific genes in changing parasite populations could render rapid diagnosis tests ineffective [58] thereby misguiding diagnosis-based chemotherapy campaigns.

Conclusion

The initial two-step evolutionary strategy of P. falciparum identified here, likely driven by two different molecular pathways with different biochemical preferences, assists the parasite in finding productive solutions to new and unexpected evolutionary challenges. The strategy is well suited for a parasite population to evolve with minimum collateral damage in surviving cells, it can act to anticipate and mount a rapid response to repeat threats, and it may offer universal advantages to parasite populations that need to withstand multiple threats beyond drug pressure.

Parasite Culture

For each experiment, erythrocytic stages of P. falciparum (previously cloned HB3 or Dd2) were freshly thawed from frozen stocks and maintained as previously described [59], [60], [61]. Briefly, parasites were grown in vitro at 37°C in solutions of 2 to 2.5% hematocrit (serotype A positive human erythrocytes) in RPMI 1640 (Invitrogen) medium containing 28 mM NaHCO₃ and 25 mM HEPES, and supplemented with 20% human type A positive plasma in sterile, sealed flasks, flushed with 5% O₂, 5% CO₂, and 90% N₂. Cultures were maintained with media changes 3 times each week and sub-cultured as necessary to maintain parasitemia below 5%.

Initial DSM1 Challenge

The highest concentration of DSM1 to which clonal Dd2 and HB3 parasites could develop resistance was determined empirically as previously described [17]. To ensure genetically pure populations, aliquots of 10 infected erythrocytes of each clonal parasite line were allowed to proliferate to about 10⁸ infected erythrocytes. From these populations, 10²–10⁷ infected erythrocytes were challenged in flasks with 0.1–10 µM DSM1 (results from 10⁷ are displayed in Table S1). Additionally, 10 infected erythrocytes were challenged with these same concentrations, to ensure that DSM1 was effective and lethal. To confirm that the parasites could proliferate normally under these experimental conditions, one flask of 10 infected erythrocytes did not receive DSM1. This experiment was performed in triplicate, using three independent biological samples of both Dd2 and Hb3 clones. Media was changed 3 times each week (receiving fresh DSM1 each time) and cultures were split 1∶2 once a week to guarantee a continuous supply for fresh erythrocytes during the experiment. Parasite proliferation was monitored by Giemsa-stained thin smear blood samples taken at each media change. Selection flasks were cultured until parasites were observed proliferating or until 90 days, whichever occurred first.

Selection of DSM1 Resistant Parasites (Rounds 1 and 2)

Using limiting dilution, 10² to 10⁷ (Dd2) or 10⁷ (HB3) populations of genetically pure parasites (see above) were plated across 24 wells of a 96-well plate (each clone was selected in quadruplicate on a single plate). Additionally, a control plate, containing 10 infected erythrocytes per 24 wells, was set up for each clone. To ensure that DSM1 was effective and lethal, the upper half of the control plate was treated with 0.3 µM DSM1. To show that the parasites could proliferate normally under the test conditions, the lower half of the control plate received no DSM1. Plates were cultured (as described above) until parasites were observed proliferating or up to 120 days, whichever occurred first. As soon as parasites were observed (Round 1 results are displayed in Table S2), the well contents were transferred to a new 10 ml culture flask for expansion, sample storage and sub-cloning. During this expansion, DSM1 resistant parasites were kept under continuous 0.3 µM DSM1 pressure and freeze-thawing and culturing for >1month at a time was avoided as much as possible. Four DSM1 resistant clones isolated in round 1 were submitted to another round of selections (round 2). Parasite populations of 10 and 10⁷ were selected with 1, 3.3 and 10 µM DSM1 as described for round 1. In addition, 10⁵ parasites were also challenged with 3.3 µM (Round 2 results are displayed in Table S8). Resistant parasites were isolated as described above before sub-cloning for further analysis.

Parasite Sub-cloning

To isolate genetically pure populations of DSM1 resistant parasites for further analysis, aliquots of 10–20 infected erythrocytes were plated across an entire 96-well plate. These plates were maintained (as described above) and as soon as parasites were observed proliferating, the well contents were extracted from the plate and transferred to a new 10 ml culture flask for further expansion, sample storage and analysis.

EC₅₀ Determination by Hypoxanthine Uptake Assay

A parasite solution at 0.5–1% parasitemia (0.5% hematocrit) from the clone of interest was plated into a 96-well culture plate. An appropriate range of concentrations of DSM1 (from 0.02–200 µM), depending on the level of resistance of the parasites being tested, were then added to the parasites (because of solubility issues, 100× DSM1 concentrations (in 100% DMSO) were first diluted 1∶10 into RPMI (final 10% DMSO) before being diluted again into the parasite-containing wells (final 1% DMSO)). Each concentration of interest was performed in triplicate and included solvent-only controls. After incubating for ∼48 hours, wells were pulsed with 0.35 µCi each of ³H-hypoxanthine. Following an additional 24–40 hours, well contents were extracted and radioactivity was measured. Parasite proliferation in each test well was expressed as a percentage of the solvent control well. EC₅₀ values were fit using the GraphPad PRISM software, according to the equation: Y = Bottom+(Top-Bottom)/(1+10^{((LogEC50−X) * HillSlope)}).

Genomic DNA Isolation for Downstream Genomics Methods

For microarrays and quantitative PCR (qPCR) protocols, clonal asynchronous P. falciparum-infected erythrocytes were lysed with 0.15% saponin (Akros) for 5 min and genomic DNA (gDNA) was extracted using the DNeasy kit (Qiagen) according to the manufacturer's instructions. For whole genome sequencing, clonal P. falciparum cultures (30 mls in T75 flasks, 3% hematocrit) were synchronized with 5% sorbitol for two consecutive cycles (∼45 hrs apart) and then once more (3–4 hr later) before harvesting for gDNA purification. These highly synchronous cultures (∼3% parasitemia at >90% rings) were washed with PBS and frozen at −80°C prior to red blood cell lysis with saponin as above. Isolated parasites were washed 3× with PBS before resuspension in 150 mM NaCl, 10 mM EDTA, and 50 mM Tris-HCl pH 7.5. Parasites were lysed with 0.1% L-loril sarkosil (Teknova) in the presence of 200 µg/ml proteinase K (Fermentas) overnight at 37°C. Nucleic acids were then extracted with phenol/chloroform/isoamyl alcohol (25∶24∶1) pH 7.8–8.1 (Acros) using phase lock tubes (5 Prime). Following RNA digestion (with 100 µg/ml RNAse A (Fermentas) for 1 hr at 37°C), gDNA was extracted twice more as above, once with chloroform, and then ethanol precipitated by standard methods.

DNA Microarrays and Comparative Genomic Hybridization (CGH)

Spotted DNA microarrays (used for both CGH and expression analysis) consisted of 10,416 −70mer oligonucleotides designed from the P. falciparum 3D7 sequence with increased coverage for long ORFs [62]. Additional custom oligonucleotides were included in the microarray to increase coverage of genes involved in folate and nucleic acid metabolism. DNA was spotted on poly-lysine coated slides and post-processed using methods described previously [25], [63]. For hybridizations on spotted DNA microarrays, 5 µg of gDNA from each clone was sheared, labeled with 5-(3-aminoallyl)-2′-deoxyuridine-5′-triphosphate, and coupled to Cy-dyes as was done previously [64]. Uncoupled Cy-dyes were removed using the DNA Clean and Concentrate-5 kit (Zymo Research) before hybridization to the microarray at 62°C for 16–18 h. After washing, slides were dried, scanned at 10 µM resolution using the GenePix 4000B scanner and fluorescent images were quantified with GenePix Pro 3.0 (Axon Instruments). Further analysis, including normalization and statistical methods were performed as described previously [25]. Spotted microarray data are presented in MIAME-compliant format on the NCBI-based Gene Expression Omnibus (GEO) database (Accession # GSE35732).

Commercially manufactured mid-density CGH microarrays containing 385,585 oligonucleotide probes ranging in size from 15- to 45-mer were purchased from NimbleGen Systems, Inc. These microarrays are sufficient to detect copy number variations but not single nucleotide polymorphisms (SNPs) [65], [66]. For hybridizations to mid-density microarrays, gDNA was labeled with Cy3 and Cy5-labeled random nanomers (Trilink Biotechnologies) and hybridized to the current CGH design Plasmodium_3D7_WG_CGH as described previously [65] except hybridization was performed overnight (∼16–18 h) in a 42°C water bath and microarrays were dried and scanned as above (at 5 µM resolution). Normalization and analysis was performed using NimbleScan version 2.6 (SegMNT CGH) and plotted using GraphPad PRISM. Mid-density microarray data are presented in MIAME-compliant format on the GEO database (Accession # GSE37306).

Quantitative PCR

For DHODH qPCR, two separate sets of primers were used to amplify a 206 bp amplicon beginning at nucleotide +656 of the DHODH coding sequence (DHODH front), and the second set amplified a 158 bp amplicon beginning at nucleotide +1423 of the DHODH coding sequence (DHODH rear) (see Table S11 for all primer sequences). The qPCR protocol was 95°C for 10 min, followed by 39 rounds of 95°C for 15 sec and 60°C for 1 min. For all experiments, we performed melt curves (55°C to 85°C in 0.5°C steps with 1 s hold at each step) to ensure a single amplicon was produced, and standard curves (10× dilution ladders of Dd2 gDNA) to determine the amplification efficiency. Relative copy number was determined for 1 ng of gDNA, using the Pfaffl method [67] according to the equation (E_target)^{ΔCt, target (control−test)}/(E_ref)^{ΔCt, reference (control−test)}, where Seryl t-RNA Synthetase (PF07_0073) and 18 s Ribosomal RNA (MAL13P1.435) served as reference genes. DSM1 resistant clones served as the test, and the Dd2 parent served as the control. Significance was determined from multiple experiments with one-way ANOVA analysis and values from individual clones were compared using the Tukey's Multiple Comparison Test in GraphPad PRISM.

Whole Genome Sequencing (WGS)

Accession Numbers

Plasmodium falciparum Dihydroorotate dehydrogenase (DHODH), Genbank Accession Number: AB070244.