Development and Optimization of a Selective Whole-Genome Amplification To Study Plasmodium ovale Spp.

ABSTRACT

Since 2010, the human-infecting malaria parasite Plasmodium ovale spp. has been divided into two genetically distinct species, P. ovale wallikeri and P. ovale curtisi. In recent years, application of whole-genome sequencing (WGS) to P. ovale spp. allowed to get a better understanding of its evolutionary history and discover some specific genetic patterns. Nevertheless, WGS data from P. ovale spp. are still scarce due to several drawbacks, including a high level of human DNA contamination in blood samples, infections with commonly low parasite density, and the lack of robust in vitro culture. Here, we developed two selective whole-genome amplification (sWGA) protocols that were tested on six P. ovale wallikeri and five P. ovale curtisi mono-infection clinical samples. Blood leukodepletion by a cellulose-based filtration was used as the gold standard for intraspecies comparative genomics with sWGA. We also demonstrated the importance of genomic DNA preincubation with the endonuclease McrBC to optimize P. ovale spp. sWGA. We obtained high-quality WGS data with more than 80% of the genome covered by ≥5 reads for each sample and identified more than 5,000 unique single-nucleotide polymorphisms (SNPs) per species. We also identified some amino acid changes in pocdhfr and powdhfr for which similar mutations in P. falciparum and P. vivax are associated with pyrimethamine or cycloguanil resistance. In conclusion, we developed two sWGA protocols for P. ovale spp. WGS that will help to design much-needed large-scale P. ovale spp. population studies.

IMPORTANCE Plasmodium ovale spp. has the ability to cause relapse, defined as recurring asexual parasitemia originating from liver-dormant forms. Whole-genome sequencing (WGS) data are of importance to identify putative molecular markers associated with relapse or other virulence mechanisms. Due to low parasitemia encountered in P. ovale spp. infections and no in vitro culture available, WGS of P. ovale spp. is challenging. Blood leukodepletion by filtration has been used, but no technique exists yet to increase the quantity of parasite DNA over human DNA when starting from genomic DNA extracted from whole blood. Here, we demonstrated that selective whole-genome amplification (sWGA) is an easy-to-use protocol to obtain high-quality WGS data for both P. ovale spp. species from unprocessed blood samples. The new method will facilitate P. ovale spp. population genomic studies.

INTRODUCTION

Malaria is a vector-borne infectious disease transmitted by Anopheles mosquito bites. The main agent of human malaria in terms of number of clinical cases and related deaths is Plasmodium falciparum (1). P. ovale spp. represents 0.77% of Plasmodium cases worldwide and up to 1.69% in Africa (2). According to the French National Malaria Reference Center (FNMRC) data, P. ovale spp. is the second most frequently detected malaria species among imported malaria cases during the last decade and represented from 6% to 10% of total malaria infections in France (3). Since 2010, P. ovale spp. has been divided into two genetically distinct species, P. ovale wallikeri and P. ovale curtisi (4). These two species are sympatric in Africa and differ by their clinical and biological characteristics in infected travelers (5 to 8).

Like P. vivax, P. ovale spp. parasites have the ability to cause relapse, defined as recurring asexual parasitemia originating from liver-dormant forms (9). The first reported cases of P. ovale spp. relapses date back to 1955 (10). A retrospective study on patients treated for neurosyphilis with sporozoite and trophozoite-induced P. ovale spp. malaria infections (also called malaria therapy) revealed that half of them had a relapse event after a complete chloroquine course (11). The existence of hypnozoites in P. ovale spp. is still debated (9), but Soulard et al. recently noticed late-developing schizonts in humanized mice infected with P. ovale spp. sporozoites (12).

The study of P. ovale wallikeri and P. ovale curtisi genomes could provide a better understanding of the genomic diversity and population structures of the two species, and help to date the species separation event and identify putative molecular markers associated with the relapse mechanism. In 2017, the first assembled genomes of P. ovale wallikeri and P. ovale curtisi were published, allowing to branch with confidence these species within the Plasmodium phylogeny (13). Of note, the study reported P. ovale curtisi genes orthologous to P. falciparum and P. vivax, but P. ovale wallikeri orthologs have not yet been published (13), limiting orthologous-based comparisons at both the Plasmodium and P. ovale spp. scales.

Whole-genome sequencing (WGS) of P. ovale spp. remains challenging due to the low parasite density commonly found in P. ovale spp. infections, compared to P. falciparum and P. vivax, and the absence of ex vivo culture protocols to amplify the parasites in vitro (14). In addition, parasite DNA extracted from unprocessed whole blood is highly contaminated with human DNA, requiring parasite DNA isolation or selective parasite genome amplification before sequencing. Some techniques have been developed to enrich the Plasmodium genome from clinical blood samples by filtering out leukocytes carrying human DNA before DNA extraction (15), selectively amplifying the parasite genome using specific primer sets, referred to as the selective whole-genome amplification (sWGA) (16) or hybrid selection (17). The sWGA approach has already been successfully developed for P. falciparum (18), P. vivax (19), P. malariae (20), and P. knowlesi (21) human-infecting malaria parasites.

In this work, we developed specific sWGA protocols for P. ovale wallikeri and P. ovale curtisi. We showed that this method is efficient and cost-effective to obtain high-quality P. ovale spp. genomic data, with blood leukodepletion by filtration used as the comparative gold standard (22). Digestion of genomic DNA with the restriction enzyme McrBC, an endonuclease that cleaves DNA containing methylcytosine (23), improved sWGA quality, as previously observed for low-density P. falciparum samples (24). Through the 11 new P. ovale spp. genomes sequenced using this method, we identified more than 5,000 single-nucleotide polymorphisms (SNPs) in both species, and we assigned P. ovale wallikeri genes orthologous to P. ovale curtisi, based on the previously published P. ovale spp. assemblies (13).

RESULTS

Sample collection.

A total of five P. ovale curtisi (here named Poc1 to Poc5) and six P. ovale wallikeri (Pow1 to Pow6) isolates were included in the study, covering a wide range of parasite density (P. ovale curtisi: 1,790 to 26,700 parasites/μL [p/μL] and cycle threshold (Ct) 23.2 to 28.1; P. ovale wallikeri:198 to 90,000 p/μL and Ct 23 to 31.2) (Table 1). Isolates were originated from patients who travelled to West or Central Africa. All samples were used to develop the sWGA strategy, with or without McrBC digestion. Poc1 and Pow1 were selected to perform leukodepletion as positive controls for high-quality WGS.

TABLE 1.

Isolates included in the study^a

Sample	Date of inclusion	Country of contamination	Parasite density (p/μL)	Ct P. ovale spp.	Ct human	Delta	sWGA	sWGA + McrBc	Filtration	Chromosomes SNPs filtered (sWGA/filtration)
Poc1	April 2021	Republic of the Congo	8,000	25	21	4	X	X	X	3,732/6,980
Poc2	January 2019	Gabon	8,000	24, 1	20, 9	3, 2	X	X		3,893/NA
Poc3	March 2015	Central African Republic	20,000	28, 1	21, 9	6, 2	X	X		4,823/NA
Poc4	November 2016	Cameroon	26,700	23, 2	20, 3	2, 9	X	X		2,558/NA
Poc5	September 2018	Ivory Coast	1,790	27	21, 3	5,7	X	X		1,580/NA
Pow1	April 2021	Cameroon	22,000	24	20	4	X	X	X	6,045/6,145
Pow2	October 2018	Mali	90,000	23	22, 4	0,6	X	X		5,292/NA
Pow3	December 2015	Benin	198	31, 2	23, 6	7, 6	X	X		5,430/NA
Pow4	April 2018	Cameroon	2,961	28	21, 5	6, 5	X	X		5,055/NA
Pow5	December 2019	Central African Republic	8,000	25	21	4	X	X		6,402/NA
Pow6	August 2018	Republic of the Congo	450	29, 8	22	7, 8	X	X		4,864/NA

^aPoc1 to Poc5 and Pow1 to Pow6 were tested on both sWGA and sWGA + McrBc conditions. Poc1 and Pow1 were also tested for the filtration condition. qPCR cycle thresholds (Ct) have been obtained with the Plasmodium typage kit (see Materials and Methods). Delta is the difference between the Ct of P. ovale spp. and the Ct of the human. Chromosome SNPs filtered column indicates the number of SNPs obtained on the reconstructed chromosome for the sWGA + McrBc condition. A “X” symbol indicates that the experiment has been performed. NA, not applicable.

Custom-made cellulose-based filtration is a suitable leukodepletion method for P. ovale spp. clinical samples.

In order to evaluate sWGA in producing accurate WGS data, the leukodepletion-based method was used as the gold standard (15), as it was already successfully used for P. ovale spp. genome sequencing (13). When applied to Poc1 and Pow1 red blood cells, filtration successfully removed human leukocytes, with a 1,000- to 10,000-fold loss of human template DNA as quantified by qPCR targeting the human beta actin gene, whereas parasite DNA loss was very much less (Table S1 in the supplemental material). WGS of those samples revealed that most of the reads mapped to the parasite genome (74% for Poc1, 95% for Pow1). More than 95% of the parasite genome was covered with a depth of coverage ≥10×.

Newly developed sWGA primers and McrBC preincubation enrich P. ovale spp. DNA over human DNA.

Despite blood leukodepletion being a suitable method for getting high-quality P. ovale spp. genomes, it is time-demanding, especially for large-scale biobanking, and cannot be used when the available blood volume is low. To fill this gap, we aimed to develop an sWGA approach for both P. ovale curtisi and P. ovale wallikeri that can be applied to archived DNA biobanks from unfiltered clinical blood samples.

Primers’ sets with phosphorothioate bonds (*) that reached the best score consisted of five primers for both P. ovale curtisi (Poc set: ATATTTT*C*G, CGTAT*C*G, TAATTCG*T*A, TATTTCG*T*A, and TCGTATA*T*A) and P. ovale wallikeri (Pow set: ATATACG*A*A, CGATAAA*A*A, CGATA*C*G, TACGAAA*T*A, and TATAACG*A*A). For the Poc set, the mean distances between two primers on human and P. ovale curtisi genomes were respectively 128,170 and 6,964 nucleotides. For the Pow set, the mean distances between two primers on human and P. ovale wallikeri genomes were respectively 115,670 and 6,082 nucleotides. Number of primers’ hits per genome was significantly higher in P. ovale wallikeri compared to P. ovale curtisi (median [10th percentile to 90th percentile]: 4,551 [4,479 to 4,628] versus 4986 [4934 to 5003]; P = 0.004, Mann-Whitney U test) (Table S2).

After WGS, mean depth of coverage with sWGA for the five P. ovale curtisi and the six P. ovale wallikeri was respectively 32× and 24×. For P. ovale curtisi, approximately half of the reads mapped to the parasite genome, whereas one third only mapped to the parasite genome for P. ovale wallikeri. For both P. ovale curtisi and P. ovale wallikeri, half of their genome was covered at ≥10×, suggesting that some improvements in the method could be done (Table S3a).

In a previous study, it was demonstrated that digesting the genomic DNA with McrBC endonuclease enzyme prior to sWGA greatly improved WGS data for low P. falciparum parasite densities (24). Considering that P. ovale spp. samples were commonly associated with low parasite density, we decided to explore the effect of McrBC digestion prior to sWGA on WGS results. WGS of McrBC-treated samples revealed a significant improvement in data quality, with a rise in the percentage of reads mapping to the parasite genome (P < 0.001 for both species, χ² test), in the normalized mean coverage (P = 0.002; Mann-Whitney U test) and in the percentage of the genome with a depth of coverage ≥10× (P < 0.001, χ² test) (Fig. 1A and B and Table S3A). For each sample, each chromosome had a mean coverage ≥10× for both species (Fig. 2A and B) except for the chromosome 4 of P. ovale curtisi. With the McrBC enzyme, P. ovale wallikeri and P. ovale curtisi had equivalent normalized mean coverage (P = 0.66, Mann-Whitney U test), but P. ovale wallikeri displayed a higher percentage of genome covered with at least 10 reads (P < 0.001, χ² test). The only sample for which the McrBC enzyme did not improve the overall WGS data quality was the sample Pow2 (Table S3A and S3B and Fig. S1). This sample was associated with the highest parasite density, the lowest qPCR Ct for P. ovale wallikeri, and the lowest ΔCt (Table 1). Hereafter, McrBC condition was chosen for downstream analyses. The specificity of the short reads generated by sWGA was confirmed by aligning the raw reads of Poc1 and Pow1 with the P. falciparum and P. malariae reference genomes (Fig. S2).

FIG 1.

(A) Percentage of mapped reads (left) and normalized mean coverage (right) for sWGA in red and sWGA with preincubation with the McrBC enzyme in blue. Normalized mean coverage is the mean coverage divided by the number of total reads produced. (B) Proportion of the genome covered with or without the McrBC endonuclease enzyme for both P. ovale curtisi (left) and P. ovale wallikeri (right). The blue dashed line represents a depth of coverage of 10. *, P < 0.05; **, P  < 0.01; ***, P < 0.001.

FIG 2.

Mean depth of coverage on each chromosome P. ovale curtisi (A) and P. ovale wallikeri (B) with sWGA (in red) or sWGA with preincubation with the McrBC enzyme (in blue). The red dashed line represents a depth of coverage of 10.

We finally compared WGS data between leukodepletion and McrBC-treated sWGA approaches. Normalized mean coverages were similar between the filtration and the McrBC conditions and equivalent percentage of reads mapped to the P. ovale spp. genomes. In contrast, depth of coverage at ≥10× was significantly better by using filtration (Fig. S3) in concordance with a better homogeneity in reads mapping (Fig. 3), as already reported for other Plasmodium species (18, 25).

FIG 3.

Comparison of the reads mapping homogeneity for the sWGA + McrBC condition (in red) and the filtration condition (in green) among the 14 chromosomes of P. ovale curtisi (on the left) and P. ovale wallikeri (on the right).

Orthologous genes and interspecies phylogenetic tree.

To perform comparative genomics at both the Plasmodium and P. ovale spp. scales, we first determined P. ovale wallikeri orthologous protein-coding genes that were still to be characterized (13). We first excluded the 1,742 pir proteins from the PocGH01 reference proteome (13), which are encoded by a multigenic family with numerous amino acid sequence variations, similarly to var genes in P. falciparum (26). By aligning P. ovale curtisi on P. ovale wallikeri protein sequences with BLAST+ (27), we identified 4,420 proteins with a high sequence similarity. We validated the orthology with OrthoMCL (28) for 4,175 proteins (3,626 orthologs between P. ovale wallikeri and P. falciparum, 3,776 orthologs between P. ovale wallikeri and P. vivax). We found 121 P. ovale curtisi and P. ovale wallikeri proteins in distinct OrthoMCL ortholog groups, and 124 P. ovale wallikeri proteins did not belong to any ortholog group (Table S4).

Using a subset of 216 orthologous protein sequences across 13 Plasmodium species and including our 11 P. ovale spp. samples, a phylogenetic tree was built (Table S5). The phylogenetic relationships obtained with this set of orthologous sequences were largely consistent with the acknowledged phylogeny of Plasmodium species (Fig. 4).

FIG 4.

Maximum likelihood phylogenetic tree using protein sequences of the Plasmodium genus rooted on P. gallinaceum showing the P. ovale spp. clade in green. Bootstrap values are indicated at nodes. Pictures indicate the host specificities.

Identification of SNPs in P. ovale curtisi and P. ovale wallikeri clinical samples.

To identify SNPs from sWGA reads, we performed a variant calling based on the reconstructed chromosomes (excluding the unassigned contigs). From sWGA-treated samples, a mean of 3,317 per sample and 5,515 SNPs per sample were identified in the chromosomes of the five P. ovale curtisi and the six P. ovale wallikeri isolates, respectively. No evidence of multiclonal isolates was detected according to the nonreference allele frequency (NRAF) plots (Fig. S4A and S4B).

To confirm the accuracy of sWGA, we compared the NRAF metrics for the positions with SNPs (NRAF > 0) for samples subjected to filtration and sWGA approaches (i.e., Poc1 and Pow1). We called 3,732 and 6,980 SNPs in sWGA and leukodepletion for P. ovale curtisi and 6,045 and 6,145 for P. ovale wallikeri. We obtained highly correlated NRAF for both species (ρ = 0.75 for P. ovale curtisi, ρ = 0.87 for P. ovale wallikeri; P < 0.001, Spearman’s rank correlation test, Fig. S5), indicating that filtration and sWGA in the context of P. ovale spp. led to very similar variant calling. Most of the differences between the two methods were due to coverage or quality issues (Fig. S6A and S6B). Positions that had different genotypes (i.e., a wild-type allele in one method and heterozygous or homozygous mutant allele at the same position with the other method) were mainly in noncoding regions (62% of discordant positions for P. ovale curtisi and 68% for P. ovale wallikeri) and with low NRAF (Fig. S6C).

We finally looked specifically for SNPs in P. ovale spp. genes orthologous to major P. falciparum drug resistance genes. SNPs were detected in the intron and/or coding sequence of the chloroquine resistance transporter (powcrt and poccrt), dihydrofolate reductase (pocdhfr and powdhfr), dihydropteroate synthetase (pocdhps and powdhps), and multidrug resistance protein 1 (pocmdr1 or powmdr1) genes (Table S6). No mutation was detected in the kelch13 genes (pock13 and powk13). Amino acid changes were found in pocdhfr (A15S and S58R), powdhfr (F57L, S58R, S113N), pocdhps (K189E, D275G), powcrt (C19G, L216F), and powmdr1 (F34N) (Table 2). Remarkably, the mutations pocdhfr A15S, pocdhfr S58R, powdhfr S58R, and powdhfr S113N corresponded to positions known to confer pyrimethamine and cycloguanil resistance in P. falciparum (pfdhfr A16V, C59R, and S108N) (29, 30) (Fig. S7). The mutation powdhfr F57L corresponded to the pvdhfr F57L/I known to confer pyrimethamine resistance in P. vivax (31). Reads from previous P. falciparum and P. malariae WGS (ERR636035 for P. falciparum and ERR4019168 for P. malariae) did not map to those P. ovale spp. orthologous genes, confirming the specificity of the SNPs generated (Fig. S8).

TABLE 2.

Amino acid changes in P. ovale spp. orthologous proteins to major P. falciparum drug-resistance genes

Species	Isolate	Crt	Dhfr	Dhps	Mdr	k13
P. ovale curtisi	Poc1		H98P	K189E and D275G	I1303V
	Poc2		H98P		I1303V
	Poc3		A15S, S58R and H98P
	Poc4		A15S
	Poc5
P. ovale wallikeri	Pow1	C19G			F34N
	Pow2	C19G
	Pow3	C19G
	Pow4	C19G and L216F
	Pow5		F57L and S58R
	Pow6	C19G and L216F	S113N

DISCUSSION

P. ovale spp. was divided in 2010 into P. ovale curtisi and P. ovale wallikeri based on distinct genetic patterns, absence of hybrid forms despite sympatry, and potential incompatibility of sexual-encoding proteins (4, 8, 31). The two P. ovale spp. have been precisely studied in terms of epidemiology, diagnostics, biology, and epidemiology, based on imported malaria cases (5 to 8). However, very few genomic data are available and molecular evolutionary analysis remains incomplete for these species.

We aimed to develop sWGA for both P. ovale curtisi and P. ovale wallikeri to sequence the parasite genome from archived DNA that had not been filtered. Specific P. ovale spp. DNA amplification through sWGA with McrBC preincubation is a suitable method to obtain high-quality sequences and SNPs for variant analysis, within the parasite density tested (up to 1,790 p/μL for P. ovale curtisi and 198 p/μL for P. ovale wallikeri). In this study, we identified an average of 3,317 and 5,515 SNPs per sample in the chromosomes of P. ovale curtisi and P. ovale wallikeri, respectively, less than what was previously described in P. vivax (14,000 per sample in a study on 18 samples [19]) and P. malariae (5,800 per sample in a study on 18 samples [20]). Due to imperfect assembly, only 60% of the P. ovale spp. genomes are assembled into chromosomes, and the rest of the genomes were grouped in nearly 700 contigs (13). As contigs contain hypervariable regions like pir genes, SNPs were called only on chromosomes, probably explaining the lower number of SNPs compared to P. malariae or P. vivax, since the core genome (i.e., the genome region used for variant calling analysis) of these species represents 80% and 92% of their entire genome, respectively (20, 32). Although the NRAF plot did not show evidence of polyclonal isolates, some positions were heterozygous, especially for P. ovale wallikeri (Fig. S4). Those features might relate to sequencing errors or alignment artifacts such as incorrect mapping of reads due to paralogous regions (32), described by Rutledge et al. (13). Better reconstructions of the two P. ovale spp. genomes and more genomic data will probably help in the future to improve variant filtering, such as previously described by Pearson et al. for P. vivax (32), and to reduce the gap in the SNP number with the other human-infecting Plasmodium species. Finally, to ensure that sWGA on P. ovale spp. provides correct sequencing data, we compared the SNPs obtained from this strategy with those called from the same samples but processed in parallel with the blood-filtration strategy. NRAF estimated from these paired samples was strongly correlated for both species (Fig. S5). We missed some SNPs in sWGA mainly due to insufficient coverage related to nonhomogeneity of the reads mapping (Fig. 3).

We further demonstrated the efficiency of sWGA to amplify orthologous genes to those associated with drug resistance in P. falciparum or P. vivax and identify SNPs in powcrt, pocdhfr, powdhfr, pocdhps, pocmdr1, and powmdr1. Some P. ovale spp. dhfr mutations led to amino acid changes at positions known to reduce susceptibility to pyrimethamine and cycloguanil in P. falciparum or P. vivax (A15S, F57L, S58R, and S113N; see Fig. S7) (31, 33). The biological effect, the prevalence, and the geographical origin of these mutations are unknown in P. ovale spp., and both molecular epidemiology and in vitro testing studies should be performed. Of note, a previous study reported amino acid change S113C in two P. ovale curtisi isolates from the China-Myanmar border area (34), and in another study, three P. ovale curtisi with A15S + S58R and one P. ovale wallikeri with F57L + S58R in Africa (35).

Although sWGA could be of great help to perform P. ovale spp. genomic study, it presents two major limitations. First, amplification of Plasmodium coinfections is really challenging. Even though theoretically possible, sequencing of P. ovale spp. in coinfection with P. falciparum (more frequent in areas of endemicity [36] than in imported malaria [37]) might be difficult. While P. ovale spp. sWGA primers bind more frequently to P. ovale spp. than to P. falciparum genome (Table S2), differences in parasite densities (10 times higher for P. falciparum than P. ovale spp. [38]) remain a major obstacle for both sWGA and leukodepletion. Capture methods to selectively collect one Plasmodium species DNA in a mixed infection should be further explored (17, 39). Second, due to preamplification with Phi29 polymerase resulting in nonhomogeneity of reads mapping, gene copy number variation (40) could not be studied with sWGA, despite its importance in Plasmodium resistance to treatments (41).

Genomic analysis of P. ovale curtisi and P. ovale wallikeri is still in its infancy. There is a need to better understand the differences between the two species, including a dating of the separation event, and, more importantly, to identify molecular determinant(s) of medically relevant traits such as dormancy. Further studies, with a larger number of isolates from different geographic areas, are needed to characterize the genetic diversity of P. ovale spp., with the potential to discover features that will help to control the disease. The new tools and genome data we produced here will help to perform such studies. The sWGA method provides a simple and efficient way to study the genomes of P. ovale curtisi and P. ovale wallikeri mono-infections. We recommend using it in cases where leukodepletion is not applicable, such as low volume of blood sample available, low parasite density, or studies on archived DNA.

MATERIALS AND METHODS

Sample collection.

P. ovale spp. isolates were selected from the FNMRC database based on parasite densities and the countries where the patients got contaminated.

Genomic DNA was extracted from 200 μL of whole blood using MagNA Pure automaton (Roche diagnostics, USA) and eluted in 100 μL. P. ovale spp. mono-infection was confirmed with the species-specific quantitative PCR (qPCR) Plasmodium typage kit (Bio-Evolution, France) targeting the 18s rRNA for P. ovale spp. and the human beta actin gene to evaluate human DNA contamination. The reaction was carried out on a ViiA 7 thermocycler (Applied Biosystems). One positive and one negative control were included in each run. ΔCt (cycle threshold) was defined as the difference between the Ct of P. ovale spp. and the human Ct. We performed in-house qPCR high resolution melting (HRM) to differentiate P. ovale wallikeri from P. ovale curtisi (42).

Ethical statement.

No specific consent from patients was required since clinical and biological data were collected from the FNMRC database in accordance with the common public health mission of all National Reference Centers in France, in coordination with the Santé Publique France organization for malaria surveillance and care. The study was considered noninterventional research according to article L1221–1.1 of the public health code in France and only requires the nonopposition of the patient (per article L1211–2 of the public health code). All data were anonymized before use. Human DNA was not analyzed.

Sample filtration.

As a positive control of P. ovale spp. high-quality WGS, we adapted to P. ovale spp. the parasitize whole blood filtration protocol previously developed by the MalariaGEN consortium for P. falciparum infections (https://www.malariagen.net/resources/partner-study-resources/archive-partner-study-resources). This filtration procedure removes the leukocytes carrying human DNA from the infected blood sample. Briefly, 1 g of cellulose powder (MN2100ff cellulose powder, Macherey-Nagel) was transferred into a 10-mL column (BD Emerald syringe 10 mL) and washed with 4 mL of 1× PBS. We slightly modified the filtration protocol and used 200 to 400 μL red blood cells instead of 2 mL of whole blood to fit our requirements. Blood samples were diluted in 1× PBS to reach a final volume of 2 mL and transferred into the column. A plunger was used to help the blood pass through the column. The column was rinsed three to four times with 4 mL of 1× PBS, and the eluate was centrifuged at 4,000 rpm for 10 min. DNA was then extracted following previously described protocol (see Sample Collection).

Primer design.

We used the sWGA program available on GitHub (https://github.com/eclarke/swga) to define two sets of primers that bind preferentially to foreground P. ovale curtisi (PocGH01, GenBank assembly accession: GCA_900090035.2) or P. ovale wallikeri (PowCR01, GenBank assembly accession: GCA_900090025.2) genomes over background human genome (GRCh38, UCSC genome browser) (43). The following sWGA parameters were selected to generate the sets of primers: primers’ melt temperature between 18 and 35°C, minimum foreground primers’ binding at 414 (equivalent to a binding every 50,000 nucleotides), maximum background primers’ binding at 12,837 (equivalent to a binding every 250,000 nucleotides), and a minimum primer size of six nucleotides. We then computationally tested different sets composed of 5 to 10 primers with the following parameters: minimum binding distance between two primers on the background genome of 30,000 nucleotides and maximum binding distance between two primers on the foreground genome of 90,000 nucleotides. Each set is scored using the average distances between primer binding sites on the foreground and background genomes and the Gini index of foreground binding sites (43). The closer the score is to 0, the more specific the primer set is to the parasite genome.

Selective whole-genome amplification.

The sWGA reactions were performed using one set of primers for P. ovale curtisi and one set of primers for P. ovale wallikeri. Each reaction was performed in 0.2 mL PCR tubes containing at least 20 ng of template genomic DNA, 0.1 mg/mL bovine serum albumin (BSA) (New England Biolabs), 1 mM dNTPs (New England Biolabs), 2.5 μM each primer, 1× Phi29 reaction buffer (New England Biolabs), 30 units of Phi29 polymerase (New England Biolabs), and molecular biology-grade water to reach a final reaction volume of 50 μL. The reaction was carried out on a thermocycler (Mastercycler Gradient, Eppendorf) with the following step-down program: 5 min at 35°C, 10 min at 34°C, 15 min at 33°C, 20 min at 32°C, 25 min at 31°C, 16 h at 30°C, then heating for 15 min at 65°C to inactivate the Phi29 polymerase before cooling to 4°C. Amplified products were quantified using the Qubit dsDNA high-sensitivity kit (Thermo Fisher Scientific). Amplified products were cleaned using Agencourt Ampure XP beads (Beckman Coulter) as follows: 1.8 volumes of beads were added to 1 volume of amplified products, briefly mixed, and then incubated for 5 min at room temperature. A magnetic rack was used to capture the DNA binding beads. The beads were then washed twice using 200 μL of 80% ethanol, and DNA was eluted with 60 μL of EB buffer.

Methylation digest.

Twenty microliters of genomic DNA extracted from P. ovale spp.-infected blood samples were digested before sWGA with 10 units of McrBC (New England Biolabs, United Kingdoms) in a 30-μL reaction mix containing 1× NEBuffer 2, 0.5 μL of 100× BSA, and 0,5 μL of 100× GTP (New England Biolabs, United Kingdoms) as previously performed on P. falciparum isolates (24). Samples were incubated for 2 h at 37°C then inactivated at 65°C for 20 min according to the manufacturer’s recommendations.

Whole-genome sequencing.

Sequencing libraries were prepared with 250 ng or 50 μL of DNA for sWGA samples and filtered controls, respectively, using the KAPA HyperPrep Library Preparation kit (Kapa Biosystems, Woburn, MA) following manufacturer’s instructions. Mechanical DNA shearing was performed with the Covaris S220 through microTube-50 AFA Fiber Screw-Cap (Covaris) using the following settings: 30% duty factor, 100W peak incidence power, and 1,000 cycles per burst for 150 s. DNA libraries were then checked for quality and quantity using Qubit for concentration and BioAnalyzer 2100 Agilent for fragment size. Libraries were sequenced on an Illumina NextSeq 500 System using 150-bp paired-end sequencing chemistry at the GENOM’IC platform from Institut Cochin (Paris, France).

Raw fastq files were aligned to the PowCR01 or PocGH01 reference genomes using the BWA-mem (Burrows-Wheeler Aligner) algorithm (default parameters) (44). Aligned reads were processed using SAMtools v.1.4 (45). Coverage statistics and depth estimates were obtained using BEDtools v.2.26.0 (46). To confirm the specificity of the short reads generated, raw reads were aligned to a concatenated genome of P. falciparum (Pf3D7, PlasmoDB release 57), P. malariae (PmUG01, PlasmoDB release 57), and P. ovale curtisi/P. ovale wallikeri (PocGH01 or PowCR01). We finally took fastq from previously published P. malariae (ERR4019168) or P. falciparum (ERR636035) data and aligned them against the concatenates resistances-associated gene sequences (PF3D7_0417200, PF3D7_0810800, PF3D7_0709000, PF3D7_0523000, PF3D7_1343700, PmUG01_05034700, PmUG01_14045500, PmUG01_01020700, PocGH01_05028400, PocGH01_14036800, PmUG01_10021600, PmUG01_12021200, PocGH01_01016900, PocGH01_10018700, PocGH01_12019400, PowCR01_050023500, PowCR01_140031200, PowCR01_010011800, PowCR0100013900, and PowCR01_120015100).

Determination of P. ovale wallikeri genes orthologous to P. ovale curtisi.

To identify P. ovale wallikeri protein-coding genes orthologous to P. ovale curtisi, we aligned PocGH01 P. ovale curtisi proteins (extracted from PlasmoDB release 49) to the P. ovale wallikeri proteome (GenBank assembly accession: GCA_900090025.2) with BLAST+ (version 2.11.0) (27). Homologous proteins were identified using two criteria: a significant BLAST E-value of <10⁻³ and a similar protein sequence size (difference length of maximum 20%). We used the OrthoMCL algorithm to confirm that previously identified proteins shared common functions and could therefore be defined as orthologous (28).

Variant calling and analysis.

Duplicate reads were tagged and removed using Picard MarkDuplicates (v. 2.26.10). Single-nucleotide polymorphisms (SNPs) were identified using BCFtools mpileup version 1.13 (minimum mapping quality for an alignment to be used [q]: 20; minimum base quality for a base to be considered [Q]: 20; coefficient for downgrading mapping quality for reads containing excessive mismatches [C]: 50) (47, 48), followed by BCFtools call with the option –V indels to discard indels. The following filtration quality criteria were used with GATK (v.4.1.8) (49): mapping quality of 40, a Phred-scaled quality score of 30, and a Phred-scaled P value using Fisher’s exact test to detect strand bias of >60. Variant calling analysis was restricted to chromosomes. Positions covered with less than five reads were filtered out. For each isolate, we considered a position as biallelic if a minimum of five reads mapped to both reference and alternative alleles. Following the Pearson’s method (32), we calculated and plotted the nonreference allele frequency (NRAF) and the percentage of heterozygote calls across the chromosomes for each isolate for the SNPs obtained with the sWGA + McrBC approach.

Phylogenetic analysis.

The interspecies phylogenetic tree was reconstructed using the top covered gene of P. ovale spp. Gene sequences of our P. ovale spp. samples were determined by producing pileup files (containing information on matches, mismatches, indels, strand, and mapping quality) using SAMtools mpileup from bam files, then consensus sequences were determined using a Perl homemade script. Nucleotide sequences were then translated into protein sequences, and for each was identified the orthologous to P. berghei, P. chabaudi, P. cynomolgi strain M, P. falciparum 3D7 (50), P. gallinaceum, P. knowlesi strain M, P. malariae, P. reichenowi, P. vivax P01 (51), P. vivax-like, and P. yoelii parasites. Each set of protein sequences for a given Plasmodium was independently aligned using MAFFT v.7.307 (52) with default options, then cleaned using GBlocks v.0.91b (53) to automatically remove noninformative and gapped sites. The cleaned, nonzero length alignments were then concatenated. The maximum likelihood phylogenetic tree was inferred using IQTREE (v. 1.6.12) after determining the best-fitting amino acid exchange rate matrix (54). Branch supports were assessed with 1,000 bootstrap replicates using the ultrafast bootstrap approximation implemented in IQTREE. The phylogenetic tree was manipulated and visualized with treeio and ggtree R packages respectively (55, 56).

Data analysis.

For all statistical tests, the number of reads was expressed in millions of reads. We normalized the depth of coverage with the overall number of reads in millions (normalized coverage) to compare WGS data sets having different numbers of total reads. Quantitative variables were expressed as median (10th to 90th percentile). Mann-Whitney U test was used to compare medians. Proportions were compared using the χ² or Fisher's exact tests. The Kolmogorov-Smirnov test was used to assess the normality of variable distributions, and the Levene’s test to verify the homogeneity of the variances. When both criteria were validated, we used the Pearson correlation test; otherwise, a Spearman’s rank test was performed. All statistical analyses and graphs were performed on R software v. 4.0.3 (57). To visualize read sequence distribution across parasite chromosomes, the average read depth in 2-kb windows across the 14 chromosomes was calculated then displayed with Circos software (58).

Data availability.

The genome Illumina sequencing reads from the P. ovale curtisi and P. ovale wallikeri samples produced in the sWGA + McrBC condition were deposited in the European Nucleotide Archive under the accession number PRJEB51041. All scripts used in this study were deposited on GitHub (https://github.com/Rcoppee/P_ovale_sWGA_project).