EDTA blood samples from 345 patients between November 2019 and February 2020 were obtained from a malaria monitoring study that aimed to compare the performance of two malaria RDTs used for diagnosis in Pingilikani dispensary (25). Pingilikani is located within the Kilifi Health and Demographic Surveillance System (KHDSS), approximately 30 km south of Kilifi town (26). The malaria monitoring study (approved by the ethics review committee of the Kenya Medical Research Institute under protocol number SERU 2617) routinely consents patients with malaria for a dried blood spot (DBS) and EDTA blood sample alongside the diagnostic Carestart™ RDT and a microscopy test. The microscopy was conducted in the KEMRI-Wellcome Trust Research Programme GCLP accredited labs by competent expert microscopists, as per the WHO criteria, who routinely support clinical trials. The lab is also registered for quantitative Polymerase Chain Reaction (qPCR), UKNEQAS. The qPCR assay targeted the P. falciparum 18S ribosomal DNA. The 345 samples were categorized as either RDT and microscopy positive, RDT positive and microscopy negative, RDT and microscopy negative, or RDT negative and microscopy positive. The latter category (RDT negative and microscopy positive) was of interest as it highlighted a discordant sample set that were suspected false negatives, hence explored further for Pfhrp2 and hrp3 gene deletions, the primary targets in the Carestart™ RDT. Within the broader sample set, we explored the utility of using clinical surveillance samples for routine molecular monitoring of Pfhrp2 and Pfhrp3 gene deletions. Since the samples had been analyzed by qPCR alongside RDT for clinical diagnosis, all the samples positive by qPCR were genotyped for the Pfhrp2/3 gene deletions along with 30% of the qPCR negative samples. The latter samples were first genotyped for Pfdhfr to confirm parasite DNA quality and those that tested positive were subjected to hrp2/3 PCRs. We also analyzed a subset of 25 samples that were positive by qPCR but negative by RDT and 23 samples that were positive by RDT but negative by both qPCR and microscopy (to determine the rate of detecting false negative results at very low parasitemia). The sample selection criteria based on RDT and qPCR are shown in Figure 1.

Following the manufacturer's instructions, parasite genomic DNA was extracted from the EDTA whole blood samples using the QIAamp® DNA Blood Mini kit (QIAGEN). The QIAcube HT (QIAGEN) platform was used for automated extraction, and the Nanodrop, ND 1000 Spectrophotometer was used to assess the concentration and purity (A260/A280 ratios) of the isolated genomic DNA. To create working concentrations of 5 ng/µl that could be used in the following procedures, the extracted DNA samples were diluted in sterile, double-distilled water and aliquots stored at −20°C. The Pfhrp2/3 PCRs were conducted in a semi-nested approach, using previously published primers (targeting Exon 2 of both genes) (17) listed in Table 1. The Plasmodium falciparum dhfr gene was also included in the PCRs, to confirm the presence and quality of parasite DNA in samples with possible Pfhrp2/3 deletions. The Expand High Fidelity PCR system (Roche™) was used to conduct conventional PCR. The annealing temperature for each primer combination (Supplementary Table S1) was determined through gradient PCR, conducted using Pf3D7 DNA as the template. Given that the Pfhrp2 and Pfhrp3 PCRs were semi-nested, the first round of amplification was conducted using the F1&R1 primer combination for each gene. Amplicons from the first amplification round were then used as templates for the second PCR which was performed using the F2&R1 primer combinations. A single reaction volume of 10.5 µl consisted of 1μl of template DNA (5 ng/μl), 6.56 µl of ultrapure DNase/RNase free distilled water, 0.2 µl of 10 mM deoxynucleotides triphosphate (dNTPs), 1 µl of 25 mM MgCl₂, 0.3 µl of 10pmol/µl forward primer (F1or F2), 0.3 µl of 10pmol/µl reverse primer, 1 µl of 10X PCR buffer with 15 mM MgCl₂ and 0.14 µl of 3.5 U/µl Expand High fidelity PCR Taq polymerase (Roche). The thermal cycling conditions for each round of PCR were 94°C for 2 min, 10 cycles of 94°C for 15 s, 54°C - 30 s and 72°C for 2 min, 25 cycles of 94°C for 15 s, 54- 30 s and 72°C for 2 min then a final extension at 72°C for 2 min. The positive controls were DNA from the P. falciparum 3D7 (with both Pfhrp2 and Pfhrp3), Dd2 (Pfhrp2 deleted) and HB3 (Pfhrp3 deleted) laboratory isolates. The Pfdhfr PCRs were performed using published primers (27) in a 10.5 µl reaction (dhfr had one forward primer, unlike hrp2/3 which had two, each). The thermal cycling conditions included 94°C for 2 min, 10 cycles of 94°C for 15 s, 58°C for 30 s and 68°C for 2 min, 25 cycles of 94°C for 15 s, 58- 30 s and 68°C for 2 min then a final extension at 72°C for 2 min. P. falciparum 3D7 DNA was similarly used as the positive control. Nuclease-free water was used as the template for the negative controls, for all three genes. 2% (w/v) agarose gels stained with 5 µl of 20,000X RedSafe DNA staining solution, were used to visualize amplification products. 5 µl of each PCR product was used for gel electrophoresis. All gels were run at 100 V for 40 min then imaged using the Bio-Rad ChemiDoc XRS + TM imaging system.

Real-time amplification for detection and absolute detection of P. falciparum were performed on the ABI 7500 Real-Time PCR system (Applied Biosystems) as previously described (28) using TaqMan based P. falciparum specific probes. The reaction was performed in a final volume of 25 μl containing: 2.5 µl each of P. falciparum 18S rDNA forward and reverse primers (10 pmol/µl), 0.625 µl of 18S probe (10 pmol/µl), 12.5 µl TaqMan universal PCR master mix (2X), 6.75 µl of sample or 3D7 control samples, with the remaining volume PCR clean water. The following RT-PCR conditions were used: 50°C for 2 min, 95°C for 10 min, and then 45 cycles of 95°C for 15 s and 60°C for 1 min. All qPCR assays were run with appropriate controls including the Non-Template Control (NTC). For parasite quantification, eight defined standards with varying parasitemias were included, Standard 1: 460,000 parasites/μl, Standard 2: 27,000 parasites/μl, Standard 3: 15,000 parasites/μl, Standard 4: 10,000, Standard 5: 100 parasites/μl, Standard 6: 10 parasites/μl, Standard 7: 1 parasites/μl and Standard 8: 0.1 parasites/μl. Following an extrapolation from the standard curve, samples with a Ct value of above 34 were considered P. falciparum negative (<5 parasites/ul).

We sequenced samples that were Pfhrp2 and Pfhrp3 positive, from the 12 that were microscopy positive but RDT negative. Following PCR, amplicons were purified using the ExoSap-IT™ (Affymetrix™) reagent, then sequenced using the BigDye® Terminator v3.1 cycle sequencing kit from Applied Biosystems, UK. Sequencing was conducted using the same primers utilized for PCR. Briefly, a single reaction volume of 10 µl consisted of 2 µl, purified template DNA, 0.5 µl BigDye® Terminator enzyme mix, 1.75 µl 5X BigDye® sequencing buffer, 4.75 µl ultrapure DNase/RNase-free distilled water and 1 µl of 5pmol/µl sequencing primer. The cycling conditions for sequencing were as follows; an initial denaturation at 95°C for 30 s, 25 cycles of 96°C for 10 s, 50°C for 5 s and 60°C for 4 min, then a final hold at 15°C for 10 min. The products of sequencing were thereafter subjected to ethanol precipitation then capillary electrophoresis on the 3730xl DNA analyser (ILRI, Kenya), to generate sequence chromatograms. Finally, sequence analysis was conducted using CLC Workbench Version 7.7.1 (QIAGEN).

Statistical analyses were conducted using Version 4.2.1 of the R software (29), where frequencies were calculated. A log₁₀ transformation of parasitaemia was applied and used to compare the different Pfhrp2/3 genotypes using a Student's t-test with the statistical significance set at a p value <0.05. Box plots were used to show the 25th percentiles, geometric mean and 75th percentiles of the log transformed parasitemia.

In this study, DNA from 345 samples tested for malaria by RDT, microscopy and qPCR were used and 249 (72.2%), 170 (49.3%) and 242 (70.1%) samples were positive for each test, respectively. There were 12 (3.47%) discordant samples that were P. falciparum negative by RDT but positive by microscopy (Figure 1), considered as false negative RDTs, were genotyped for Pfhrp2 and Pfhrp3. Pfdhfr was used as a secondary gene to confirm the presence of parasite DNA in samples that tested negative for both Pfhrp2 and Pfhrp3. One of the 12 discordant samples was missing and was therefore excluded from subsequent analysis. Among the remaining 11 samples, 3 (0.87%) tested positive for both Pfhrp2 and Pfhrp3, 3 (0.87%) were positive for Pfhrp2 but negative for Pfhrp3, while 4 (1.16%) were negative for Pfhrp2 but positive for Pfhrp3 (Table 1). Six of these samples had low parasitemias of <1,000 parasites/µl and the remaining sample was negative for all the genes tested (Pfhrp2, Pfhrp3 and Pfdhfr). Notably, this sample was negative by qPCR but contained 3,520 parasites/µl.

There were 25 (7.25%) samples that were positive by qPCR but negative by RDT, considered as discordant based on the qPCR test, and these were assessed for Pfhrp2 and Pfhrp3 deletions. Within this sample set, 6 samples were also microscopy positive. The results obtained, identified 11 (3.19%) samples that were positive for both Pfhrp2 and Pfhrp3, 9 (2.61%) were positive for pfhrp2 but negative for Pfhrp3, while 4 (1.16%) were positive for pfhrp3 but negative for Pfhrp2. None of the samples were negative for both Pfhrp2 and Pfhrp3 but positive for dhfr (Table 1). We extended the analysis further and grouped the samples based on a Cycle threshold (Ct) cut-off of 34, which represented at least 5 parasites/µl (an extrapolation from our standard curve). This stringent cut-off yielded 17 discordant samples, from which the Pfhrp3 negative only samples reduced to 4 (1.16%) with no change to the number of to the Pfhrp2 negative only samples. There were no samples that tested negative for both hrp2 and hrp3 but positive for dhfr. In addition, the lone sample that was Pfhrp2−ve/Pfhrp3−ve (Pfdhfr−ve) but qPCR positive when Ct values were not considered as a criteria was further confirmed as lacking parasite DNA and dropped off with the stringent Ct cut-off.

In addition to the discordant samples above, the analysis included a larger sample set that were P. falciparum positive by qPCR. A total of 242 samples were qPCR positive, the large majority, 44.4% were positive for both hrp2 and hrp3, while only 6 (1.74%) were negative for both genes but positive for Pfdhfr. The remaining samples were either Pfhrp2 negative only, Pfhrp3 negative only, negative for all three genes or missing with no genotyping data. The stringent cut-off of Ct ≤ 34 resulted in 218 samples that met the criteria of a parasite positive sample. Similar to the qPCR discordant samples, there was reduction by about half [13, (4.1%)] in the numbers of Pfhrp2 negative only samples. Only 1 sample was not accurately determined as Pfhrp2 and Pfhrp3 negative plus Pfdhfr positive, this sample was also microscopy negative. The four samples that were Pfhrp2−ve/Pfhrp3−ve (Pfdhfr−ve) were further confirmed as parasite negative samples and dropped off with the stringent Ct cut-off.

As a final confirmation of the variation in qPCR and PCR specificity, 48 (∼47%) samples that were negative for P. falciparum by qPCR were selected and tested for the Pfhrp2 and Pfhrp3 deletions. All the 48 samples were genotyped for Pfdhfr to confirm that parasite DNA was present, yielding 15 positive samples, 69% of which were microscopy negative. Only 1 (0.29%) sample was negative for both genes but positive for Pfdhfr and further examination of the sample revealed that it was negative by microscopy and thus a false positive. Most, 8 (2.3%), of the remaining samples were negative for Pfhrp3 only and only 2 of these samples were microscopy positive; and within the 5 Pfhrp2 and Pfhrp3 positive samples, 3 were microscopy positive.

We had an additional category of samples (23) that were positive for P. falciparum by RDT, but negative by both microscopy and qPCR. Like the qPCR negative group, the samples were genotyped at the dhfr locus to select positive DNA samples for Pfhrp2 and Pfhrp3 testing. Of the 23, only 5 (1.45%) were positive for dhfr and of the 5, Pfhrp2/3 data was obtained from 4 samples. Two (0.6%) of the four samples was positive for both hrp2 and hrp3, 2 (0.6%) were Pfhrp3 negative only, while 1 (0.3%) was Pfhrp2 negative only. None of the samples was negative for both Pfhrp2 and Pfhrp3.

We further assessed if the level of parasitemia influenced whether a sample tested positive or negative for Pfhrp2, Pfhrp3 or both. There were 231 samples with Pfhrp2/3 genotypes out of the total 345 samples. The geometric mean parasitemia of the respective Pfhrp2/3 genotypes was 61,965 parasites/μl (Pfhrp2/3 + ve), 6,088 parasites/μl (Pfhrp2 + ve, Pfhrp3−ve), 44,247 parasites/μl (Pfhrp2−ve, Pfhrp3 + ve), and 32,212 parasites/μl (Pfhrp2/3−ve). Overall, the analysis revealed that the Pfhrp2 and Pfhrp3 negative and Pfhrp2−ve only samples had similar levels of parasitemia as those that tested positive for both genes (Figure 2). However, there was a significant difference in parasitemia between the Pfhrp3−ve only and Pfhrp2 + ve/hrp3 + ve samples (p = 0.033, Figure 2). Parasitemia levels were generally lower between samples that were discordant by either microscopy or qPCR and those that were non-discordant. Samples that were microscopy positive, but RDT negative had significantly (p = 0.0087) lower levels of parasitemia compared to those that were both microscopy and RDT-positive (Figure 3A), though cognizance is taken of the expected low numbers of discordant samples. There were 25 qPCR + ve/RDT−ve samples in total, however only 6 of these had microscopy data, and only 150 of the 216 qPCR + ve/RDT + ve samples had microscopy data. The low numbers of qPCR + ve/RDT−ve samples are likely to have attributed to the lack of significance observed in the comparison with qPCR + ve/RDT + ve samples (p = 0.065, Figure 3B).

We sequenced 12 and 13 samples for Pfhrp3 and Pfhrp2, respectively. The 7 and 9 good quality sequences obtained for Pfhrp2 and Pfhrp3 respectively, were subsequently analyzed for the presence of 3-, 6- and 9-amino acid repeats. A total of 20 repeat types were identified. Type 4 (AHH) and type 7 (AHHAAD) repeats were common in Pfhrp2 and Pfhrp3 and an additional 8 repeat types were described for Pfhrp2, while 5 more repeats were defined for Pfhrp3 (Supplementary Table S2). Thus, demonstrating the similarity between both proteins and with the global population (30, 31).