Samples were collected from patients presenting with uncomplicated P. vivax malaria as determined by microscopy at the Royal Darwin Hospital, Australia (returning travellers), the Rumah Sakit Mitra Masayarat Hospital in Papua Indonesia, and the Shoklo Malaria Research Unit in Thailand. All samples were collected from patients presenting to outpatient clinics with P. vivax parasitaemia as determined by microscopy. Exclusion criteria included severe malaria, administration of anti-malarial or antibiotic treatment in the previous 2 weeks, and age less than 3 years. The volume of venous blood collected was tiered by patient age and weight. In children (age <5 years), blood volumes were specified as 1 ml blood per kg body weight, with a maximum of 6 mls. In children >5 years of age and adults, 12 and 18 mls were drawn, respectively. All blood samples were drawn into lithium heparin Vacutainer™ tubes (Becton-Dickinson).

All biological samples and patient data were collected with written, informed consent from the patient or a parent or guardian. Ethical approval for this study was obtained from the Human Research Ethics Committee of the NT Department of Health & Families and Menzies School of Health Research, Darwin, Australia (HREC 09-83), the Eijkman Institute Research Ethics Commission (EIREC 2010-47), the Ethics Committee of the Faculty of Tropical Medicine, Mahidol University, Thailand (MUTM 2006-029) and the Oxford Tropical Research Ethics Committee, Oxford University, UK (OXTREC 27-05).

All samples were processed within 6 hours of venepuncture. The plasma and buffy coat layer were removed after separation by centrifugation at 1,500–2,000 rpm for 10–15 minutes at ∼22°C. The remaining pellet was resuspended in an equal volume of phosphate-buffered saline (PBS) and subject to CF11 (Whatman) filtration, as previously described [20]. Briefly, the re-suspended blood was added to a CF11 column which had been pre-washed with PBS, and the RBCs were gently washed through with PBS until the liquid dripping into the collection tube was clear. The filtered blood sample was then centrifuged at 1000 rpm for 5 minutes, the supernatant was removed, and the pellet was re-suspended in an equal volume of PBS. The filtration process was then repeated with a new CF11 column. After the study was completed, Whatman discontinued the production of CF11. However, as discussed elsewhere [23], unpublished trials in P. vivax field isolates indicate that medium-sized cellulose fibre powders available from other companies (e.g. Sigma product number C6288), provide equivalent leukocyte retention to Whatman CF11. In a subset of the Indonesian and returning traveller samples, aliquots were collected at different stages of blood processing. A summary of the sample numbers available for each processing step and study site is presented in Table 1.

After double CF11 filtration, the parasites in the remaining RBC preparation were cultured ex vivo to enable early blood stages to mature to schizonts, as described elsewhere [24]. Incomplete culture medium was prepared with McCoy’s 5A medium supplemented with HEPES (25 mM), L-glutamine (2 mM), gentamicin (40 mg/L) and D-glucose (0.24%). Twenty percent AB+ human serum (inactivated at 56°C for 30 mins) was added to the incomplete medium to prepare the complete medium. The RBC pellet was added to complete medium at a 2% haematocrit. The resultant blood medium mixture was split accordingly into culture flasks (75 cm³) and incubated at 37.5°C for ∼40–48 hours, using a candle jar to control O₂ and CO₂ levels. After culture, the RBC pellet was obtained by centrifugation at 2,000 rpm for 10 minutes. Thick and thin blood films were made on pre-, peri- and post-culture samples in order to monitor the extent of parasite maturation in culture.

DNA was extracted from the RBC pellets using QIAamp Blood Midi or Maxi kits (Qiagen) as per the manufacturer’s protocol. Plasmodium species was confirmed by PCR and real-time quantitative PCR (RT-qPCR) [25], [26]. Total sample DNA concentration was measured using Invitrogen’s Quant-iT™ dsDNA HS assay as per the manufacturer’s protocol.

The percentage of human (and P. vivax) DNA in the Indonesian and traveller samples was estimated by RT-qPCR with human (Toll-like receptor 9 [TLR9]) and P. vivax (P. vivax fructose 1, 6-bisphosphate aldolase [pvaldo]: PVX_118251) specific primers. Primer sequences (5'-3') were as follows: TLR9-F: ACGTTGGATGCAAAGGGCTGGCTGTTGTAG, TLR9-R: ACGTTGGATGTCTACCACGAGCACTCATTC, pvaldo-F: GACAGTGCCACCATCCTTACC, pvaldo-R: CCTTCTCAACATTCTCCTTCTTTCC. The Indonesian samples were processed at the Eijkman Institute, and the traveller samples were processed at the Menzies School of Health Research. Cloned plasmid TLR9 and pvaldo fragments surrounding the targeted RT-PCR amplicons were generated as standards for the assays using the TOPO TA Cloning® Kit (Invitrogen). The plasmids were diluted as appropriate to encompass the expected concentration range of the test samples. Separate SYBR® Green-based reactions comprising 1 µl DNA template, 2x SensiMix No Ref (Quantace), 200 nM each primer, and Nuclease-free water to a 10 µl final volume, were prepared for each of the TLR9 and pvaldo assays and run on a Corbett Rotor-Gene 6000 machine with the following cycle conditions: 1×95°C/15 mins; 40 × (95°C/30 sec; 60°C/30 sec; 72°C/45 sec); 1×68°C - 90°C in 0.5°C increments. All samples and standards were run in duplicate. The concentration of the test sample’s TLR9 and pvaldo amplicons were approximated by standard curve analysis using the Corbett Rotor-Gene 6000 series software 1.7. In order to calculate the relative proportions of human and P. vivax DNA, the estimated concentration of P. vivax DNA was multiplied by a factor of ∼0.01 to account for the relative differences in the size of the TLR9 (4550 bp) and pvaldo (3115 bp) plasmids, and the human (∼3.2 Gb) and P. vivax (∼28 Mb) genomes. The percentage of human DNA was calculated using the adjusted P. vivax DNA quantity.

In the Thai samples, the percentage of human DNA was estimated at the Wellcome Trust Sanger Institute using a RT-qPCR assay with human standards and primers [19]. The P. vivax DNA quantity was then extrapolated as the total DNA quantity minus the estimated human DNA quantity.

Excluding the mixed-species infections, the significance of the difference between Indonesia (n = 15) and Thailand (n = 19) in each of ex vivo culture duration, percentage of schizonts post-CF11+ culture, percentage of human DNA post-CF11+ culture, and P. vivax DNA yield post-CF11+ culture was assessed using the Mann-Whitney U test.

Using the Indonesian isolates for which both pre- and post-culture aliquots were available (n = 8), the Wilcoxon signed-rank test was used to assess the difference between paired pre- and post-culture isolates in each of human DNA percentage and P. vivax yield.

Comparisons between the RT-qPCR and Illumina estimates of human DNA percentage were undertaken on data derived from the post-CF11+ culture aliquots of the pure P. vivax isolates (n = 22 [3 Traveller, 19 Thailand]) for which Illumina sequence was available (see Table 1). The correlation between the RT-qPCR and Illumina human DNA percentage estimates was assessed using Spearman’s rank correlation coefficient. The significance of the difference in magnitude between the percentage of human DNA determined by RT-qPCR and Illumina was assessed using the Wilcoxon signed-rank test. Significance for all tests was determined at the 5% threshold. All analyses were undertaken using R software [27].

Twenty-four pure P. vivax and 2 P. vivax+P. falciparum infections (see Table 1) were sequenced on the Illumina GA II or HiSeq 2000 platform (see Table S1) (http://www.illumina.com/systems/genome_Analyzer.ilmn) at the Wellcome Trust Sanger Institute. Pre- and post-culture aliquots were sequenced for two of the pure P. vivax isolates, leaving 22 independent samples (see Table 1). Paired-end multiplex or non-multiplex (see Table S1) libraries were prepared from 500 ng - 1 µg total DNA, as per the manufacturer’s protocol, with the exception that genomic DNA was fragmented using Covaris Adaptive Focused Acoustics rather than nebulisation. Multiplexes comprised 12 tagged samples. Cluster generation and sequencing were undertaken as per the manufacturer’s protocol for paired-end 75 bp, 76 bp or 100 bp sequence reads (see Table S1).

The total yield of raw, unaligned, sequence data was recorded for each sample. The data was then aligned against each of the P. vivax Sal-1 (PlasmoDB release 9.1 [http://plasmodb.org/common/downloads/release-9.1/PvivaxSaI1/fasta/data/PlasmoDB-9.1_PvivaxSaI1_Genome.fasta] [10], and NCBI accession AB649419.1 [apicoplast genome: http://www.ncbi.nlm.nih.gov/nuccore/AB649419.1] [28]), P. falciparum (3D7 PlasmoDB release 9.1 (apicoplast genome included) [http://plasmodb.org/common/downloads/release-9.1/Pfalciparum3D7/fasta/data/PlasmoDB-9.1_Pfalciparum3D7_Genome.fasta] [29]), and Homo sapiens (NCBI build V37 [30]) reference genomes using the bwa read mapping algorithm (version 0.6.2), with the default execution parameters (available from http://bio-bwa.sourceforge.net) [31]. The percentage of reads mapping against each of the P. vivax, P. falciparum, and human genomes was derived from the resultant bam alignment files and recorded for each sample [32]. The P. vivax bam files (and P. falciparum bam files for the 2 mixed-species samples) were analysed further with the ‘depth’ command provided in the SAMtools application (version 0.1.18) to compute the per-base depth. The default SAMtools depth parameters were used with the exception of the mapping quality threshold (Q) of 20, corresponding to 1% mapping error). The resultant depth files were processed using R software [27] to derive summary statistics on sequence depth (fold-coverage of a given region) and breadth (percentage of a given region covered by a given number of reads). Breadth statistics were derived for a minimum of 1 and of 20 reads. Breadth and depth statistics were determined for the full nuclear reference genome (i.e. excluding the mitochondrial and apicoplast sequences) and coding and non-coding regions. Coding and non-coding positions were defined in each of P. vivax and P. falciparum with the Sal-1 PlasmoDB release 9.1 (http://plasmodb.org/common/downloads/release-9.1/PvivaxSaI1/gff/data/PlasmoDB-9.1_PvivaxSaI1.gff) and 3D7 PlasmoDB release 9.1 [http://plasmodb.org/common/downloads/release-9.1/Pfalciparum3D7/fasta/data/PlasmoDB-9.1_Pfalciparum3D7_Genome.fasta] annotation files, respectively. Summary plots were generated using R software [27].

In total, 39 independent P. vivax isolates were processed: 4 from returning travellers, 15 from Indonesian patients and 20 from Thai patients (Table 1). One of the returning traveller isolates (Papua New Guinea) and one of the Thai isolates were found to be mixed-species infections of P. vivax and P. falciparum by PCR and thus were not included in the sample preparation analyses. In the remaining 37 pure P. vivax samples, parasite densities ranged from 2,600 to 99,480 parasites µl⁻¹, with a median of 18,640 parasites µl⁻¹ (Figure 1A). Sampling was undertaken largely within the framework of a complementary study investigating P. vivax ex vivo drug susceptibility testing, which requires processing predominantly ring stage parasites (see [24]). Therefore the majority of samples in the current molecular study exhibited a high proportion of ring stage parasites (median 94% [range: 7–100%]) (Figure 1B). Short-term ex vivo culture was successful in all except one sample (DRW-002), the median duration of incubation being 43.5 hours [range: 39–53] with a final median percentage of schizonts of 52% [range: 12–88%] (Figure 1C and D). The median incubation period in the Thai samples (42.5 hours [range: 40–46]) was significantly shorter than in the Indonesian samples (median = 48 hours [range: 43–53) (Mann-Whitney U test: p = 5.6×10⁻⁷). In addition, the Thai samples produced a greater percentage of schizont stages (median = 63% schizonts [range: 27–88]) than the Indonesian samples (median = 42% [range: 20–80]) (Mann-Whitney U test: p = 6×10⁻⁷). Details on the clinical and laboratory properties of each sample are presented in (Table S2).

In each of the Indonesian, Thai and traveller sample sets, the distribution of human DNA percentages (as defined by RT-qPCR) at each processing step is illustrated in Figure 2. The human DNA contamination assessed in three samples from travellers prior to processing showed all samples to have >90% human DNA. After CF11 filtration, human DNA contamination fell from a median of 95.2% [range: 91.9–99.4] to 54.6% [range: 24.3–60.6]. This dropped further to a median of 38.1% [range: 6.29–56.5] following short-term culture.

The impact of the short term culture step relative to CF11 filtration alone was further assessed in 8 Indonesian isolates with paired pre- and post-culture aliquots, and found to reduce human DNA by 4.91-fold [range: 1.87–13.48]; from a median of 33.9% [range: 1.58–68.6] in post-CF11 samples to 6.22% [range: 0.20–23.0] following culture (p = 0.008). Of the 3 traveller and 8 Indonesian samples, 81.8% (9/11) of the post-culture aliquots exhibited less than 30% human DNA, in contrast to 45.5% (5/11) of the post-CF11 aliquots.

In the full set of 37 post-CF11+ culture samples, the median human DNA content was 3.85% [range: 0.20–56.5]. The median human DNA percentage in the post-CF11+ culture Thai samples was 2.65% [range: 0.40–53.4]. This was lower than in the full set of 15 post-CF11+ culture Indonesian samples (4.31% [range: 0.20–44.6]) but the difference was not significant (p = 0.086).

In each of the Indonesian, Thai and traveller sample sets, the distribution of the P. vivax DNA yield at each processing step is illustrated in Figure 3. In the 3 unprocessed samples from returning travellers, the median P. vivax DNA yield was 1.29 ng µl⁻¹ packed red blood cells (pRBCs)[range: 0.36–3.93], rising to 1.71 ng µl⁻¹ pRBCs [range: 0.24–1.91] following CF11 filtration, and 2.00 ng µl⁻¹ pRBCs [range: 0.57–3.82] following the addition of short term culture.

In the 8 Indonesian isolates with paired pre- and post-culture aliquots, the P. vivax DNA yield increased a median 34-fold from a median of 0.15 ng µl⁻¹ pRBCs [range: 0.01–1.39] in the post-CF11 aliquots to a median of 6.2 ng µl⁻¹ pRBCs [range: 2.34–14.4] after additional culture (p = 0.001). Extrapolating from the yield estimates per microliter pRBCs, in a “standard” 5 ml (50% haematocrit) whole blood sample, amongst the 3 traveller and 8 Indonesian isolates, 100% of the post-culture aliquots would give a sufficient P. vivax DNA yield (≥500 ng) for Illumina genome sequencing compared to 54.5% (6/11) of the post-CF11 aliquots. Including the human DNA yield, 72.7% (8/11) of the post-CF11 aliquots would yield sufficient total DNA for whole genome sequencing. Considering both the yield of total DNA (500 ng threshold) and percentage of human DNA (30% threshold), 81.8% (9/11) of the post-culture aliquots would meet the requirements for whole genome sequencing in contrast to 36.4% (4/11) of the post-CF11 aliquots.

In the full set of 37 post-CF11+ culture samples, the median P. vivax DNA yield was 3.12 ng µl⁻¹ pRBCs [range: 0.56–14.4]. The median yield in the Indonesian samples (5.33 ng µl⁻¹ pRBCs [range: 0.56–14.4]) was significantly higher than in the Thai samples (2.34 ng µl⁻¹ pRBCs [range: 0.64–6.61]) (p = 0.010).

Sequence data was successfully generated on 22 post-CF11+ culture P. vivax isolates, 2 post-CF11 (pre-culture) P. vivax sample aliquots, and 2 mixed-species P. vivax+P. falciparum isolates. The data are available in the European Nucleotide Archive (ENA) (http://www.ebi.ac.uk/ena/data/search?query=plasmodium) [33]. For ethical purposes, the data submitted to the ENA are filtered to remove the human alignments. Sample accession numbers for the raw data are presented in (Table S3). Details of the sequence yield and mapping statistics for each sample are presented in (Table S1). The median yield of total sequence data generated was 2321 million bases (Mb) [range: 902–6911 Mb]. The observed variation in total sequence yield may reflect variation in a range of factors including read length, platform (Illumina GAII versus HiSeq) and multiplex versus non-multiplex sequencing (see Table S1). Figure 4 presents bar-plots illustrating the distribution of reads which aligned against each of the P. vivax, human, and P. falciparum reference genomes, as well as “other”, non-aligned reads, in each sample. In the 22 independent, pure P. vivax samples, the median percentage of reads which mapped to the P. vivax reference was 79.7% [range: 11.0–89.1]. Across these samples, the median sequence depth was 54X [range: 18–116X], as illustrated in Figure 5. In terms of read breadth, the median percentage of the P. vivax reference covered by a minimum of 1 base was 90.9% [range: 89.1–92.6%], dropping to a median of 86.9% [range: 45.6–89.1%] covered by a minimum of 20 bases (Figure 6).

As expected, the percentage of reads which mapped against the human reference was positively correlated with the percentage of human DNA estimated by RT-qPCR (rho = 0.94, p = 4.7×10⁻¹¹) in the 22 independent pure P. vivax isolates. The percentage of human reads was only a median 1.34-fold [range: 0.35–4.02-fold] higher than the RT-qPCR estimates of human DNA. However, the difference between the median percentage of human reads (4.40% [range 0.18–83.2%]) and RT-qPCR human DNA estimates (2.7% [range 0.40–56.5%]) was significant (p = 0.0053). Also as expected, the post-CF11 aliquots of DRW-001 and DRW-003 demonstrated a substantially higher median percentage of human reads (2.55 and 5.36-fold difference, respectively) and accordingly lower median P. vivax read depth (2.14 and 3.76-fold difference, respectively) than their respective post-culture aliquots (Table 1).

In the 22 independent, pure P. vivax samples, a median of 0.30% reads [range: 0.17–0.61%] mapped to both the P. vivax Sal-1 and P. falciparum 3D7 reference genomes. The median percentage of reads which did not map to the human, Sal-1 or 3D7 reference genomes (“other”) in these samples was 14.4% [range: 5.79–20.0%].

The coding and non-coding sequence depth distributions for P. falciparum and P. vivax in the 2 mixed-species sample are presented in Figure 7. In both samples, the P. vivax non-coding and coding sequences peaked at similar depths of coverage (56 and 59 respectively in THA-009; both 5 in DRW-004), close to the expected (nominal) sequence depths (53.0 and 5.5 in THA-009 and DRW-004, respectively). In contrast, the P. falciparum non-coding sequences peaked at moderately lower depth than the corresponding coding sequences (14 and 29 respectively in THA-009; 22 and 36 respectively in DRW-004) and nominal sequence depths (24.3 and 29.8 in THA-009 and DRW-004, respectively). The coding and non-coding coverage distributions in the 22 pure P. vivax samples (Figure S1) confirmed this reduced coverage bias across the P. vivax genome.