Leukocyte depletion in dried blood spot cards enables enrichment of parasite DNA for improved sequencing

Abstract

Expanding access to simple blood collection tools is essential to monitor, control, and eliminate malaria in low resource settings where the disease is endemic. The most common method to preserve blood is depositing fingerstick samples onto filter paper—the dried blood spot (DBS) card. While DBS cards offer more optimal storage solutions than venous blood in vacutainers, they do not provide sample cleanup or enrichment of Plasmodium DNA. These samples retain high host-to-parasite DNA ratios, which negatively affect the quality of downstream sequencing. We developed a Leukocyte Depletion Card (LDC) that substantially depletes host white blood cells from whole blood to enrich Plasmodium-infected red blood cells in a hematocrit-independent volume (9.0 ± 0.5 μL). Using quantitative PCR, we evaluate the performance of the LDC using blood collected from 16 P. falciparum-infected patients at a clinic in Cape Coast, Ghana. The LDC achieved an average 32.5-fold parasite enrichment over venous blood. Promisingly, the LDC also provides a 36.6-fold parasite enrichment over a DBS card. Initial testing of targeted sequencing demonstrates significant (p < 0.01) improvement in P. falciparum read counts and coverage for the LDC. The LDC represents a unique microsampling device with potential applications in epidemiological studies of malaria.

Graphical Abstract

Introduction

Malaria remains a major public health challenge, particularly in sub-Saharan Africa, where Plasmodium falciparum is the most prevalent and deadly species of parasite^[1]. Malaria parasites infect red blood cells and can typically be detected via peripheral blood smear once parasitemia exceeds approximately 50 parasites μL⁻¹.^[2] Levels of parasite burden in patients can be categorized as low (≤ 1,000 parasites μL⁻¹), moderate (1,000–4,999 parasites μL⁻¹), high (5,000–99,999 parasites μL⁻¹), or hyperparasitemia (≥ 10,000 μL⁻¹).^[3] Beyond its role in facilitating a primary diagnosis via rapid tests or blood smears, sampling blood from infected patients is crucial to molecular and epidemiological studies that aid in understanding, managing, and controlling malaria. Currently, the primary method for blood collection and storage for molecular analysis is the dried blood spot (DBS) card (e.g., Whatman 903 Protein Saver cards). Despite the convenience and ease of sampling and storage, 903 cards are limited by their simple design, leading to a range of possible user errors (e.g., unreliable zone filling) and inconsistent sample extraction.^[4–7] Moreover, DNA extracted from 903 cards is often significantly contaminated with host DNA, which dilutes the targeted parasite DNA.^[8] Even in samples with moderate levels of parasitemia (e.g., 1%), where the parasites are more abundant than host white blood cell (WBC) DNA, the larger size of the diploid human genome (6.4 GB) overshadows the smaller, haploid genome of the parasite (22.8 MB)—a 280-fold difference in size.^[9] While challenges related to analyzing these relatively small amounts of parasite genetic material have been extensively addressed by modifying current sequencing methods (e.g., amplification additives^[10]) or improving sample enrichment methods (e.g., hybrid capture using whole genome baits^[8], or synthetic oligonucleotide probes^[11]), contamination from the host remains an outstanding obstacle. This issue is particularly pronounced in technologies available to those in limited resource settings. Additionally, the presence of host genes and proteins severely affects the efficiency of downstream analyses, leading to off-target sequencing and reduced sequence coverage of the target Plasmodium genome.^[12]

Methods to remove host cell contamination in field-collected samples of blood include a range of chemical and physical processes. Enzymatic treatments (e.g., methylation dependent enzyme digestion)^[8] can enrich Plasmodium DNA up to 9-fold by selectively degrading host DNA. Common density gradient media (e.g., Ficoll, Lymphoprep) have been shown to offer between 40–80% WBC depletion,^[13,14] while Pall Acrodisc WBC syringe filters and Plasmodipur filters have been shown to provide 40–100%^[15] and 12–54%^[12] depletion, respectively. In addition to single isolation methods, combination methods of density gradient media and filter units (e.g., Lymphoprep + Plasmodipur + antiHLA1)^[12] have been shown to deplete 75–98% of WBCs. Unfortunately, approaches relying on filtration require relatively large volumes of venous blood (1–10 mL, at minimum) and active processing at the point of sampling by a healthcare worker using intensive protocols. Leukosorb, a fibrous membrane used in commercially-available filter units to recover WBCs selectively,^[16,17] is translatable to the development of paper-based devices and can accommodate fingerstick volumes of blood (40–70 μL cm⁻¹).^[18] We have previously used Leukosorb as a pre-filter for plasma separation devices that can recover liquid^[19] or dried^[20] plasma from whole blood. We hypothesized that using Leukosorb as the basis of a simple, field-deployable card (a Leukocyte Depletion Card; LDC) would cause the spatial separation of host WBCs from parasite-infected red blood cells (RBCs) and result in improved recovery and analysis of Plasmodium DNA.

We first tested the LDC using in vitro, contrived samples of parasitized blood to demonstrate how card design elements (i) enriched the recovery of parasite DNA over host DNA compared to the 903 card, while (ii) maintaining quantitative recovery of parasite DNA from infected RBCs. Following this characterization, we assessed the clinical performance of the LDC with whole blood samples from 16 patients, which varied in WBC and parasite counts. In comparison to the 903 card, the LDC showed superior WBC depletion and fold parasite enrichment, resulting in significant improvements to Plasmodium sequencing results (both read count and read coverage) for patients with parasitemias above 4046 parasites μL⁻¹. By intrinsically depleting WBCs and retaining parasitized RBCs—without requiring any additional steps by a patient or healthcare worker collecting blood in a field setting—the LDC has the potential to advance genomic, epidemiological studies of malaria, such as improving the accuracy and scope of malaria drug- and drug-resistant^[21,22] tracking and public health monitoring.

Experimental Section

Live subject statement

We obtained samples of whole blood from Research Blood Components (Watertown, MA). The vendor follows the American Association of Blood Banks guidelines for all donors, which includes IRB approved consent to the use of collected blood for research purposes. All research was approved by the Institutional Biosafety Committees of Tufts University and Uniformed Services University of the Health Sciences.

Patient participation and consent

The clinical study in Cape Coast, Ghana was approved by the Institutional Review Board of the Noguchi Memorial Institute for Medical Research (CPN# 050/12–13). Participants of all age groups were eligible to be enrolled into the study. Prior to enrollment, we obtained written informed consent from all adults with parental consent provided by the parent/guardian of all children below the age of 18 years. All samples were obtained from Ewim Polyclinic in Cape Coast, Central Region of Ghana.

Fabricating leukocyte depletion cards

We designed four dried blood spot cards in Adobe Illustrator (prototypes 1–4). We utilized a double-sided wax transfer method to pattern the TFN with unique designs on each side.^[23] Briefly, we printed the top and bottom designs onto Avery laminate sheets using a Xerox ColorQube 8580 wax printer. Next, we aligned a sheet of TFN with the top and bottom designs using a custom acrylic alignment jig. Finally, we used a VEVOR P8200 T-shirt press (50 s at 142 °C) to transfer the wax from the laminate sheets to the paper to form hydrophobic barriers through the full thickness of the paper. We cut Leukosorb circles or channels using an OMTech laser engraving machine (Anaheim, CA). We fabricated each card by attaching layers with laser-cut adhesive sheets and sealed each card using Fellowes laminates.

Measuring and adjusting the hematocrit from whole blood samples

We measured the initial hematocrit of each donor whole blood sample upon arrival using centrifugation (3 μL into a capillary and spun for 3 min at 12,000 rpm in an LW Scientific ZipCombo centrifuge). We created samples of whole blood at different hematocrit values (25–55%) by adjusting the volume of native plasma in the sample using previously described methods.^[28] We then confirmed the hematocrit value by centrifugation (N = 2 capillary tubes per contrived hematocrit).

In vitro P. falciparum culture conditions

We obtained P. falciparum strain NF54 through BEI Resources (MRA-1000, Patient Line E, contributed by Megan G. Dowler). In brief, we maintained parasites in an atmosphere of N₂/CO₂/O₂: 90/5/5 and complete RPMI medium (RPMI 1640, 25 mM HEPES, 100 μg mL⁻¹ hypoxanthine, 0.3 mg mL⁻¹ glutamine) supplemented with 25 mM NaHCO₃, 5 μg mL⁻¹ of gentamicin, and 10% human serum or 0.5% Albumax II. We evaluated parasitemia (Infected RBCs/Total RBCs*100) through microscopy after Giemsa staining of smears (10%, 15 minutes). We used sorbitol (5 wt%, 20 min at 37 °C) to synchronize parasites cultures in the ring-stage.

Making blood samples with contrived parasitemias to add to 903 cards and LDCs

We harvested sorbitol-synchronized, ring stage parasite cultures, spiked these cultures into donor whole blood, and adjusted the hematocrit to 40%. We then used these samples to prepare samples of blood comprising parasitemias of 0%, 0.001%, 0.01%, 0.1%, 1% and 5% (Table S1). We applied 50 μL of each parasitemia to each replicate zone of a 903 card and a LDC (N = 5 replicates per card type) and stored an additional 100 μL of whole blood to serve as a control specimen.

Analyzing 903 cards and LDC with contrived parasitemias

We stored 903 cards and LDCs in foil bags at 4 °C until analysis. For LDCs, we used a standard 6 mm hole punch to remove the extraction zones from a LDC (N = 5 replicates, one punch per replicate zone). For 903 cards, we used a ⁵/₈” manual punch to remove the entirety of each blood spot (N = 5 replicate zones per card). We extracted genomic DNA (gDNA) from each punch (1 punch per extraction) and from liquid controls (9 or 50 μL) using a Qiagen QIAamp DNA Mini kit and according to Qiagen’s dried blood spot and liquid whole blood extraction protocols. We extracted each punch separately (one punch per reaction). We used 100 μL of Qiagen QIAamp DNA Mini Kit Buffer AE (water-based elution buffer) for each final elution step and stored the purified gDNA from each sample at −20 °C until use. We amplified the β-actin and sbp1 genes from each purified DNA sample using a QuantStudio3 Real Time PCR system. Briefly, each 20 μL of qPCR reaction mix contained 10 μL Fast SYBR Green Master Mix, 1.6 μL of mixed forward and reverse β-actin or sbp1 primers (5 μM total per reaction, Table S2), 6.4 μL Type I water, and 2 μL of purified DNA. We used RNase P as the non-template control. Instrument cycling conditions are shown in Table S3. Example amplification and melt curves are shown in Figure S1.

Clinical sample collection

We first identified malaria-infected patients using a blood smear. Then, we collected approximately 1 mL of blood from each consenting participant. For each sample, we spotted five 50-μL drops of blood onto unique collection zones of individually labeled Whatman 903 Protein Saver Cards and LDCs. We dried both card types at room temperature for 4 hours. Upon drying, we put both card types in individual foil bags containing a desiccant and stored each at room temperature pending analysis. The remaining blood samples were stored at −20 °C.

Determining parasite density of clinical samples using microscopy

We processed and stained blood films according to WHO guidelines.^[24] We had two independent malaria microscopists read each smear, with any discordant calling of positive or negative smears broken by a third microscopist. We estimated parasite density as the number of parasites counted per 200 WBCs, multiplied by 40 based on the assumption that 1 μL of blood contains 8000 WBCs.

Analyzing 903 cards and LDCs from clinical patient samples

We handled clinical samples in an identical manner to in vitro samples (see Section 3.8), with a slight modification for the analysis procedure due to the number of samples we needed to process, which were analyzed across 8 PCR plates to include all 16 patients and replicates. On each plate, we included a sample (N = 3 replicates) with a single, known WBC and parasite concentration and then corrected all raw cycle threshold (Ct) values using Equation (1).^[35] Using these corrected raw Ct values, we then determined fold parasite enrichment.

$${\text{Ct}}_{\text{i}}^{\text{corrected}}={\text{Ct}}_{\text{i}}^{\text{uncorrected}}-{\text{Ct}}_{\text{i}}^{\text{IPC}}+\frac{1}{\#\text{of plates}}{\sum }_{\text{i}=1}^{\#\text{of plate}}{\text{Ct}}_{\text{i}}^{\text{IPC}}$$ (1)

Analysis of card drying time

We obtained the initial mass of 903 cards and LDCs (N = 3 replicates per card type) using a Cole Palmer LA-164 balance. Then we applied 50 μL of blood to each card via pipette (at a contrived 40% hematocrit) and tracked their wet volume loss due to evaporation using RADWAG R-Lab balance tracking software until the measurements reached a plateau indicative of the residue dry volume mass. We plotted the blood spot mass as a function of time to obtain the average dry time for each card type. We approximated the average, initial rate of loss due to drying by fitting the initial linear portions of each curve.

Selective whole genome amplification

We performed selective whole genome amplification (sWGA) on extracted gDNA from three clinical patient samples, across all specimen types (903 card, LDC, and matched liquid reference samples; N = 2 replicates per patient), as described previously^[40] with minor modifications. In brief, sWGA is an enrichment method to selectively amplify a target genome (here, P. falciparum DNA) over background DNA (here, the human genome in WBC) using a pool of primers designed to amplify frequently occurring motifs of short nucleotides in the P. falciparum reference genome (Table S2). We performed the sWGA experiment in two steps: first, we combined 8 μL of each purified gDNA sample, 0.25 μL of primers (final solution concentration of 20 μM of each primer in the pool), 0.5 μL of 10X ThermoFisher EquiPhi29 reaction buffer, and 1.25 μL of nuclease-free water. We then denatured the 10 μL reaction for 3 minutes at 95 °C. Next, we mixed the denatured product with 1 μL (10 units) of EquiPhi29 DNA polymerase, 2 μL reaction buffer, 0.2 μL of 100 μM MDT, 2 μL of 10 mM dNTPs, and nuclease-free water to make a total pool volume of 20 μL. We incubated the reaction at 45 °C for 3 hours and then at 65 °C for 10 minutes to suspend further enzyme activity. We validated amplification success by quantifying DNA before and after enrichment using a Qubit fluorometer and Qubit 1X dsDNA High Sensitivity Kit.

MIP sequencing and data analysis

We targeted, captured, and sequenced sWGA-amplified DNA using molecular inversion probes (MIP) targeting key P. falciparum drug resistance genes associated with artemisinin and partner drug resistance, including pfkelch13, pfmdr1, pfcrt, pfdhfr and and pfdhps genes. We performed MIP capture and library preparation as previously described.^[41,42] In brief, we conducted sequencing using an Illumina NextSeq 550 instrument (150 bp paired-end reads). We demultiplexed the raw data generated using MIPs using MIPtools software,^[25] which is a computationally suitable tool for MIP data processing and analysis. We further processed the data using MIP Wrangler software,^[26] in which sequence reads sharing the same Unique Molecular Identifiers (UMIs) were collapsed to generate a single consensus. We analyzed each dataset by mapping sequence reads to the P. falciparum 3D7 reference genome using Burrows-Wheeler Aligner (BWA) to generate a total number of sequenced reads per sample and sequencing coverage per samples per MIP probe used.

Results and Discussion

Designing and testing consistent filling of card prototypes across a range of hematocrits

Cards must optimally (i) fill across a wide hematocrit range to accommodate the variable composition of blood anticipated in a broad patient population, (ii) dry in less than 3 hours to promote biomolecule stability (i.e., DNA, RNA, proteins), (iii) provide a hematocrit-independent, single extraction punch volume to allow for quantitative leukocyte and parasite counts, and (iv) be operationally simple in the environments where they are intended to be handled (i.e., field settings and clinical laboratories).^[27] Based on these design criteria, we used our previously reported observations of blood cell transport in paper-based devices to design candidate prototypes.^[28–30] Each candidate comprised an (i) inlet zone, (ii) separation channel, (iii) extraction zone, and (iv) overflow channel as shown in Figure 1A. The overflow channel, which extends beyond the extraction zone, ensures complete filling of extraction zones regardless of the hematocrit of the applied sample of blood.^[31,32] Further, designing devices that provide a patient-independent and reproducible volume stored in the extraction zone enables more quantitative analyses and inference of absolute input amounts (e.g., parasitemia per microliter of peripheral blood).^[32] As a result of these design criteria, we evaluated four prototypes of a DBS card that could deplete WBCs: prototype 1, TFN channel only (Figure S2A,B); prototype 2, TFN channel and 1 Leukosorb disc (Figure S3A,B); prototype 3, TFN channel and 2 Leukosorb discs (Figure S4A,B); and prototype 4, Leukosorb channel only (Figure 1A,B). These prototypes differed in material choice (e.g., TFN cardstock or Leukosorb) or assembly (e.g., channels with or without separate sample addition layers) to promote RBC transport to the extraction punch zone while restricting WBC movement.

Figure 1.

Design of the Leukocyte Depletion Card (LDC, “prototype 4”) and quantification of the volume contained in the sample extraction zone. (A) The LDC design comprises a single-layer of Leukosorb laser-cut into a channel that is reinforced within an identically-cut layer of TFN (yellow dashed lines represent the Leukosorb inlaid within the laser-cut TFN support). The design is tiled five times across the card. (B) Image of the upper portion of the LDC depicting an area to fill out patient information (white rectangle) as well as an (i) inlet zone (indicated to user with green arrow), (ii) separation channel, (iii) extraction zone, and (iv) overflow channel. (C) Representative images showing LDC channels filling across a range of hematocrits from 25–55%. (D) Punch zone volumes for all hematocrits were determined using Drabkin’s assay (N = 3 replicate punches per hematocrit). Markers represent individual measurements, the solid green line represents the average punch volume, and the dotted green lines represent the 95% CI.

We challenged each prototype with 50 μL of whole blood via pipette with contrived hematocrits ranging from 25–55%. We ensured that the entirety of the extraction zone filled for all four prototypes (Figures 1C, S2C, S3C, S4C). After drying the cards overnight, we determined the volume of blood contained in each extraction zone by quantifying the amount of hemoglobin contained in the punch. Based on the results of our earlier studies with designing DBS cards that meter whole blood, we expected that if the hematocrit did not have a significant effect on the volume of blood recovered in a punch, we could use a single volume as a liquid reference for any patient to compare to the extracted punch.^[32] Prototypes 1, 2, and 3 provided punch volumes of 13.0, 18.4, and 9.7 μL, respectively (Figures S2D, S3D, S4D), when averaged across the entire range of hematocrits (Table S4); however, these numbers varied significantly between hematocrits (ANOVA, p < 0.001 for all three prototypes). Alternatively, prototype 4 provided an average punch volume of 9.0 μL that did not vary significantly between hematocrits (p = 0.11, Figure 1D). Although prototype 3 did not provide hematocrit independence across the entire range tested, volumes were statistically indifferentiable across a narrower range of hematocrits (25–50%; p = 0.06). As it is anticipated that patients infected with malaria are unlikely to have high hematocrits,^[33,34] we chose to move forward with both prototype 3 and prototype 4 to test the ability of each card to deplete WBCs.

WBC depletion using healthy donor blood

To determine WBC depletion provided by prototypes 3 and 4, we quantified the amount of β-actin, a common human reference gene, present in the sample via quantitative PCR using a single donor (WBC count of 6300 μL⁻¹). To ensure analyses were comparable across different card types, we compared raw Ct values measured from a punch from each card to a matched liquid reference volume: 50 μL (903 card as a microsampling comparator), 9.7 μL (prototype 3), and 9.0 μL (prototype 4) (Table S5). Prototype 4 depleted significantly more WBCs (higher δCt value) than the 903 card (ANOVA, p < 0.001) or prototype 3 (ANOVA, p = 0.009). We hypothesize that this increase in depletion is due to the greater amount of Leukosorb that cells interact with in prototype 4 (i.e., 20-mm pathlength in a single channel) compared to prototype 3 (i.e., ~1 mm pathlength in two vertical zones), allowing for increased capture of WBCs. These data suggest that by depleting more WBCs, prototype 4 may provide better overall parasite enrichment when challenged with blood containing parasitized RBCs. Based on these results, we chose to move forward with prototype 4 as our final Leukocyte Depletion Card (LDC) to be tested against both contrived samples of whole blood containing RBCs infected with in vitro cultures of P. falciparum and also clinical samples of whole blood collected at a field site.

Analyzing parasite recovery using Plasmodium falciparum-infected samples from in vitro cultures

We generated contrived samples of Plasmodium falciparum-infected blood by supplementing whole blood from a single, healthy donor with cultures of synchronized, ring-stage parasites (NF54). Parasite counts are reported as percentages rather than parasite per microliter counts in this experiment because the RBC count of the healthy donor was unknown. We prepared contrived samples with parasitemias of 0.001%, 0.01%, 0.1%, 1%, and 5% infected RBCs (approximately 45–225,000 parasites μL⁻¹ assuming an RBC count of 4,500,000 RBC μL⁻¹), encompassing low- to hyperparasitemias. Because we used a single donor to prepare samples, the total WBC count and hematocrit (40%) did not vary across parasitemias. We applied 50 μL of each sample of blood to replicate zones of a 903 card and an LDC (N = 5 zones per parasitemia). We quantified WBC and parasite DNA (β-actin and spb1 genes, respectively) from punches and matched volumes of a liquid reference sample (50 μL and 9 μL for 903 card and LDC, respectively) via qPCR (Table S6) to determine the fold parasite enrichment, which we define as the relative amount of parasite DNA compared to host DNA, between the card and its reference, as described in Equations (2)–(6).

$$\Delta \text{Ct}={\text{Ct}}_{\text{sbp}1}-{\text{Ct}}_{\beta -\text{actin}}$$ (2)

$$\Delta \Delta {\text{Ct}}_1=\Delta \Delta {\text{Ct}}_{\text{LDC or}903\text{card}}-{\text{Ct}}_{\text{Liq. Ref.}}$$ (3)

$${\text{Fold parasite enrichment}}_1=2^{-\Delta \Delta {\text{Ct}}_1}$$ (4)

$$\Delta \Delta {\text{Ct}}_2=\Delta \Delta {\text{Ct}}_{\text{LDC}}-{\text{Ct}}_{903\text{card}}$$ (5)

$${\text{Fold parasite enrichment}}_2=2^{-\Delta \Delta {\text{Ct}}_2}$$ (6)

We normalized the measured Ct of the parasite gene (sbp1) to the host gene (β-actin) using Equation (2). To compare the 903 card and LDC performance versus liquid blood, we calculated the difference in ΔCt between each card and its volume-matched, liquid reference using Equation (3). Then, we calculated the fold parasite enrichment for each card versus its matched liquid reference using Equation (4). As the 903 card is the standard for collection and dried storage of malaria-infected blood samples, we also calculated the difference in ΔCt between the LDC and the 903 card—with the 903 card as the reference sample—using Equation (5). We obtained a second fold parasite enrichment for the LDC over the 903 card using Equation (6). A fold parasite enrichment of greater than 1 represents an increase in the amount of parasite DNA present in each sample type compared to its control, while a fold parasite enrichment of less than 1 represents a decrease in the amount of the parasite DNA. Results indicate a negligible, average fold parasite enrichment of 1.2 for the 903 card versus its liquid reference (Figure 2A). Promisingly, we demonstrate a substantially improved fold parasite enrichment of 243.2 and 139.6 for LDC compared to its liquid reference and the 903 card, respectively.

Figure 2.

Fold parasite enrichment from in vitro and clinical samples. The Y-axis has been log-transformed for clarity. Boxes represent the 25^th–75^th percentiles, the middle line represents the median value, and whiskers represent the 5–95% percentiles. We calculated relative parasite to host enrichment by first comparing the parasite gene target (sbp1) with the human gene target (β-actin) for each sample (ΔCt), then comparing the difference between these values for each card type (ΔΔCt), and finally calculating fold gene expression (2^−ΔΔCt). (A) Fold parasite enrichment from in vitro samples using blood from a single donor supplemented with cultures of parasitized RBCs. Each data set represents five parasitemias (0.001%, 0.01%, 0.1%, 1%, and 5%) with five replicates per parasitemia for a total of 25 measurements per comparison. Visualized data points represent values (2/25 points per comparison) that fall outside of the 5–95% percentiles. (B) Fold parasite enrichment from clinical samples collected from malaria-infected patients. Each data set represents 16 patients (N = 3 replicates per patient for a total of 48 measurements). Visualized data points represent values (4/48 points per comparison) falling outside of the 5–95% percentiles.

Analyzing parasite recovery using a clinical population

We obtained venous blood from 16 malaria-infected patients attending the Ewim polyclinic in Cape Coast, Ghana with WBC and parasite counts ranging from 3,800–20,100 WBC μL⁻¹ and 33–251,100 parasites μL⁻¹ (Table S7). At the clinic site, we applied 50 μL of venous blood from each patient via pipette to each replicate zone of a 903 card and an LDC (N = 5 zones per card type per patient, Figure S5 and Figure S6). After drying in ambient conditions for 4 hours, cards were bagged in barrier pouches with silica desiccant and stored for up to 7 days at ambient conditions in the laboratory in Accra. In addition to the cards, we saved and froze a liquid blood sample from each patient. We shipped both card types together by FedEx at ambient conditions between the Accra laboratory and our laboratory at Tufts University (Medford, MA), which took 7 days. We shipped liquid blood samples on dry ice. Upon arrival, we immediately stored cards and liquid samples at −20 °C until they could be processed. We extracted DNA from both card types and matched, whole blood reference samples (50 μL and 9 μL for 903 card and LDC, respectively), and we performed qPCR to obtain β-actin and sbp1 Ct values (Table S8). Unlike in vitro samples, DNA extracted from clinical samples had to be amplified across multiple PCR plates due to the evaluation of 16 patients, each with 3 sample types (903 card, LDC, and liquid blood), and with 3 technical replicates each. We used an inter-plate calibrator sample (IPC) with a known WBC and parasite count to normalize for plate-to-plate variability. Using the IPC, we normalized the raw Ct values from each plate (N = 9 plates) using Equation (1)^[35] to obtain “corrected” Ct (Table S9) that we used for subsequent calculations (Table S10). Because volumes of all samples are known precisely, we can use measured Cts to determine parasite enrichment (Table S11). Results again indicate negligible fold parasite enrichment of 1.2 for the 903 card versus the liquid reference (Figure 2B). Encouragingly, the LDC demonstrated improved enrichment of parasite DNA with average fold parasite enrichment of 32.5 and 36.6 compared to its matched liquid reference and the 903 card, respectively.

We directly compared the Ct values obtained from the sbp1 gene for in vitro and clinical samples to better understand the cause of differences in fold parasite enrichment observed between for the LDC. While we observed a statistically significant loss of parasites in the 903 card (Figure 3A), as determined by linear regression comparison analysis (ANCOVA, ɑ = 0.05, p = 0.002), we observed no significant loss in parasites for LDC (Figure 3B, p = 0.497) (Table S12). To provide a better context for the clinical samples, we converted raw Ct values into WBC and parasite counts using calibration curves (Figure S7). We report WBC and parasite counts as total counts per extracted sample (Table S13). We then calculated the recovery of WBC and parasites in punches from each card using the matched liquid reference sample (Table S14 and Table S15, respectively). Table 1 illustrates the average recovery of WBCs and parasites from 903 cards and the LDC. Interestingly, we observed that both cards substantially deplete WBCs (< 18% recovered from both cards); however, the data also suggest that 903 cards deplete parasites with an average of only 18% of the parasites recovered. Since the entire 50 μL zone is extracted from the 903 card, we would not anticipate any parasite or WBC loss. This result highlights a potential problem with DNA stability in the 903 card. In contrast, LDC showed 126.2% parasite recovery, suggesting that LDC stabilizes DNA and may enrich parasites at the extraction zone in comparison to whole blood. For the LDC, only 1 of the 16 patients (Patient 16) demonstrated poor WBC depletion, with 88.5 ± 42% WBC recovery at the extraction zone. The 15 other patients retained an average of only 12.4% of WBCs at the extraction zone. Based on visual inspection of the LDC for Patient 16 (Figure S6), we believe this discrepancy may be due to overloading this LDC with > 50 μL of blood as their hematocrit (31.5%) should not result in the complete saturation of the device overflow channel.

Figure 3.

Comparison of P. falciparum gene amplification trends using in vitro and clinical samples of whole blood. We amplified the malaria sbp1 gene by qPCR and plotted the measured Ct as a function of parasite count for samples collected with (A) 903 cards and (B) LDCs using in vitro (n = 5 contrived parasitemias, N = 5 replicates per parasitemia) and clinical samples (n = 15 unique patients, N = 3 replicates per patient). We removed the results from one patient, Patient 14, from this analysis due to an undefined sbp1 Ct value.

Table 1.

WBC and Parasite Recovery from Clinical Blood Samples by 903 Cards and LDCs

	WBC Recovery (%)		Parasite Recovery (%)
Card	Average	95% CI	Average	95% CI
903	11.4	[8.8–14.0]	18.3	[12.2–25.2]
LDC	17.2	[9.1–25.3]	126.2	[109.8–142.5]

Analysis of blood drying times for each card type

While DBS cards like the 903 card can provide advantages over vacutainers for storing blood, they are not intended for long term-storage at ambient conditions, and it is recommended that 903 cards are stored at −20 °C with desiccant to minimize the possibility of sample degradation over time.^[36] As both cards types were stored for 14–16 days at ambient temperatures prior to storage at −20 °C, it is possible that a combination of drying speed^[37] and storage conditions^[36] (e.g., temperature) led to poor DNA stability in the 903 card, while the construction of the LDC could have overcome these same conditions. A polyester-based membrane (Leukosorb) rather than a cellulosic material (Whatman 903) was used in the LDC. In addition, the LDC design provided a larger sample collection area than the 903 card (180 mm² versus 125 mm²), while approximately maintaining the same material thickness: Leukosorb is between 356–559 μm thick, while the 903 card is 440–520 μm thick. These factors could enable faster drying of an equal volume of collected blood for the LDC over the 903 card (ca. 2–4 hours).^[38,39] Importantly, while maximizing the LDC surface area, we kept in mind a geometry that would: (i) allow for hematocrit-independent filling with a 50 μL sample and (ii) present clearly defined sample application and extraction zones. To directly test the drying time, we applied 50 μL of blood to the 903 card and LDC (N = 3 replicates per card type) and quantified loss of mass due to evaporation over time with an analytical balance. Results show LDCs dry over two times faster than 903 cards (Figure 4) with initial drying rates of 0.65 and 0.31 mg min⁻¹, respectively. These differences in rates led to average drying times of 72 ± 2 minutes and 193 ± 13 minutes (Student’s t-test, p < 0.001), respectively. In addition to drying time, the storage of these cards at ambient and likely fluctuating conditions over a period of many days during transport between laboratories may have also played a role in degrading DNA stored in the 903 card, which would present as a depletion of WBCs or parasites by qPCR. Even if shipping conditions were responsible for the reduced recovery of DNA by 903 cards, these identical conditions did not have a negative impact on the performance of the LDC.

Figure 4.

Dry time as a function of card type. We applied 50 μL of healthy, whole blood from a healthy donor (40% hematocrit) to 903 cards and LDCs. Each blank card mass was subtracted to obtain the mass of blood in each case. Curves represent the average dry time from n = 3 cards per card type and error bars reflect the standard error of the mean (SEM). We collected measurements every minute until sample masses plateaued. 903 cards dried in an average of 193 ± 13 minutes while LDCs dried in an average of 72 ± 2 minutes. 903 cards and LDCs have average, initial drying rates of 0.65 and 0.31 mg min⁻¹, respectively, determined by linear regression (black lines, 903 card: y = −0.31x + 50.27, LDC: y = −0.65x + 51.59).

Targeted Illumina sequencing of P. falciparum

To examine the benefit of leukodepletion in the LDC, we selected 3 of the 16 patient samples to analyze by targeted sequencing of P. falciparum DNA. These patients presented with a range of WBC counts (5,340–7,600 μL⁻¹) and parasite counts (56–46,909 μL⁻¹). We performed selective whole genome amplification (sWGA)^[40] to amplify P. falciparum DNA and then subjected all amplified samples to molecular inversion probe (MIP) assay^[41,42] prior to targeted next generation sequencing. Post-sequencing, we obtained read counts, which describes the number of times our sequence of interest (P. falciparum DNA) is identified, and read coverage, which describes the extent to which a genome or region is represented by reads, indicating how many times each base is sequenced for each sample (N = 2 replicates per patient). It is best to see higher read count and coverage when sequencing a gene of interest, as this ensures greater accuracy in detecting variants and quantifying gene expression. Lower read count and coverage are usually observed when host or human DNA overwhelms parasite DNA in a sample, making it harder to capture sufficient sequencing data from the target organism. We compared the read count and read coverage across different parasite densities for each sample set (903 card versus LDC) using R software and a p-value of ≤ 0.05 as statistically significant. Results demonstrate a statistically significant improvement in read count (Figure 5A) and coverage (Figure 5B) for moderate and high parasite counts: Patient 02 (4,046 parasites μL⁻¹) and Patient 11(46,909 parasites μL⁻¹) (Figure 4, Table S16). For Patient 09 (56 parasites μL⁻¹), there was no significant difference in sequencing performance between the two card types. Additionally, we analyzed several known and validated resistance markers across five key antimalarial drug resistance genes, including dhfr-ts and dhps (for SP resistance), crt (CQ marker), k13 (for partial artemisinin resistance), and mdr1 (a marker for various drugs, including CQ and AL). We determined smoothness of coverage by accessing how uniformly the coverage spans the sequenced loci per sample for the 903 card and LDC. Promisingly, the sWGA results from the LDC samples with significant improvements in read count (Patient 02 and 11) show both improved coverage and smoother coverage (i.e., uniform coverage spanning the sequenced loci per sample, Figure 5C) across loci than the sWGA results from the 903 card (Figure 5D). Overall, these results demonstrate the potential of LDC to outperform traditional DBS cards in improving the quality of targeted next generation sequencing of parasite genes through leukodepletion.

Figure 5.

Using clinical samples collected by 903 DBS cards and LDCs to support next generation sequencing of P. falciparum genomes. (A) We analyzed Patients 09, 02, and 11 to cover a wide range of parasite counts (56, 4046, and 46909 parasites μL⁻¹, respectively). We plotted sample read count from Illumina sequencing as a function of parasites μL⁻¹, and performed statistical comparisons of read counts from both cards for each patient (N = 2 replicates per run). (B) We calculated the average target coverage by dividing the paired read count by the number of probes (185) in each sample. (ns = p > 0.05, ** p < 0.01, *** p < 0.001) (C) Heat map coverage of sequenced loci from five known, antimalarial drug resistance genes for the LDC and 903 card. We analyzed several known and validated nonsynonymous resistance markers including dhfr-ts and dhps (SP resistance), crt (CQ marker), k13 (partial artemisinin resistance), and mdr1 (a marker for various drugs, including CQ and AL). Data are plotted on a log10 scale for clarity. Grey boxes reflect loci that were not present in a sample.

Conclusion

We developed a blood microsampling device that is capable of selectively depleting host leukocytes to enrich the ratio of parasite to human DNA in malaria-infected patients, with the goal of supporting efforts in malaria epidemiology by enabling the generation of high-quality sequences derived from samples collected at field sites. We designed our Leukocyte Depletion Card (LDC) to fill with blood across a wide range of hematocrits and provide a clear extraction zone for reproducible DNA extraction from the dried sample in a laboratory. We ensured the volume of blood within the extraction zone was reproducible, regardless of patient hematocrit, which enabled us to directly compare the extracted sample to a matched liquid reference volume and calculate counts of WBCs and parasites stored in each zone. In both in vitro models of parasite-infected blood and in a panel of blood samples collected from Plasmodium falciparum-infected patients at a clinical field site, the LDC outperformed the 903 card in both leukocyte depletion and the stability of parasite DNA during storage and transport of materials between laboratories. Ultimately, we demonstrated that this significant fold parasite enrichment improved read counts and read coverage for the LDC over the 903 card for parasites sequenced using selective whole genome amplification and molecular inversion probes targeting P. falciparum drug resistance genes. This improvement in sequence quality is enabled even by the smaller volume of blood contained in the extraction zone in the LDC (9 μL) in comparison to the 903 card (50 μL), further highlighting the advantages of simultaneous leukodepletion and DNA stabilization. While we designed the LDC to operate with fingerstick volumes of blood, this first study was enabled by blood obtained from venipuncture due to the need for approximately 1 mL of blood to permit all testing. Future research efforts that leverage the performance of the LDC will require developing protocols for the direct application of fingersticks (e.g., treatment with anticoagulant) in addition to optimizing the analysis of different parasite species, co-infections, or lower parasitemias. While we demonstrate here that the LDC can enumerate parasite burden and improve parasite sequence quality in patients infected with malaria, we expect these results will lead to new applications in clinical malaria research and global health. In light of the well-documented performance gaps of 903 cards, the LDC represents a promising advancement in DBS technology for decentralizing sample collection and testing.

Supplementary Material

The following file is available free of charge:

Supporting Information: Supplemental figures and tables including initial prototype designs, punch volume analysis, WBC depletion analysis by qPCR, raw Ct values and calculated and ΔCt values for in vitro and clinical samples, patient demographics, scans of patient card samples, fold parasite enrichment values, WBC and parasite calibration curves, WBC and parasite recovery from cards, card drying time, read counts and coverage for clinical samples including heat map coverage of sequencing coverage from five known antimalarial drug resistance genes, and examples of qPCR amplification.

SYNOPSIS.

Drug resistance hinders malaria control efforts and makes population surveillance crucial. Dried blood spot (DBS) cards support these efforts but host DNA makes collected samples inadequate for molecular analysis. A Leukocyte Depletion Card (LDC), which separates parasitized red blood cells from white blood cells, provides superior sequencing results over the traditional DBS card.