DNA of P. falciparum and P.vivax was obtained from blood samples of patients previously diagnosed as having a single infection by well-trained microscopists, and PCR using ribosomal and non-ribosomal targets (5, 16). One hundred and eleven blood-positive samples were used from patients from different regions of the Brazilian Amazon infected with P. vivax and P. falciparum or stored at the biorepository of Laboratory of Malaria at Instituto René Rachou ( Supplementary Table 1 ). Parasite density was determined as the number of asexual parasites observed per 200 white blood cells on a thick smear and was estimated by assuming a leukocyte count of 8,000 per µL. Parasitemia ranged from 90 to 16,950 parasites/µL for P. falciparum-infected individuals, and from 30 to 17,100 parasites/µL for P. vivax-infected individuals. The inclusion criteria were mild malaria or asymptomatic, more than 5 years old, absence of pregnancy, and a signed informed consent form.

The collection of human samples for DNA extraction was performed in accordance with relevant guidelines and regulations. All participants and/or their legal guardians provided written informed consent. Ethical and methodological aspects of this study were approved by the Ethical Committee of Research on Human Beings from the IRR (N° 2.243.058), in accordance with the Brazilian National Council of Health (Resolutions 196/96 and 466/12). The INI-Fiocruz Ethical Board approved the study concerning the patients from the Atlantic Forest (number 0062.0.009.000-11).

DNA extraction from blood samples was performed using a QIAamp® DNA Blood Mini Kit (Qiagen, Hilden, Germany) or Gentra® Puregene® Blood Kit (QIAGEN, Chatsworth, CA, USA), using 300 µL of blood and a final volume of 50 µL, in accordance with the manufacturer’s instructions. Both blood and extracted DNA material were stored at –20°C.

Seventeen blood samples were included from patients from the Atlantic Forest area diagnosed as being positive for Plasmodium simium by PCR analysis (17, 18). Samples from 26 non-infected individuals (i.e., human negative controls tested with distinct malaria PCR protocols) from a non-transmission malaria area (i.e., Belo Horizonte, Minas Gerais, Brazil) were included as negative controls in the assays.

Thirteen P. simium and 15 P. malariae/P. brasilianum DNA samples obtained from free-living and captive non-human primates (NHPs) from the Brazilian Atlantic Forest were included; these samples were stored at the biorepository of the Laboratory of Malaria (18–20). Samples from 12 NHPs from areas without malaria transmission, negative in distinct PCR protocols (5, 18), were included as negative controls in the assays. All samples were processed to obtain DNA, as mentioned above, and stored at –20°C. NHPs were diagnosed by only PCR (17, 18), as their low parasitemia hampers the diagnosis by light microscopy. The Brazilian government (Ministry of Environment) authorized the capture, handling, and collection and transport of biological samples from NHPs (SISBIO numbers: 43375–4/2015, 54707–137362–2 and 52472–1; and INEA license 012/2016012/2016). This study was approved by the Institutional Ethics Committee of Animal Use (CEUA license L037/2016).

qPCR assays for the diagnosis of P. vivax and P. falciparum, targeting coding and non-coding mt genome regions were designed. Whole-mt sequences of P. falciparum (GenBank accession number: AY282930.1), P. malariae (GenBank accession number: AB354570.1), and P. vivax (GenBank accession number: PvP01, GCA_900,093,555.1) were aligned using ClustalW software in the BioEdit package. For primer design, regions of mt genome that contained polymorphisms between Plasmodium species, preferentially in the 3ʹ ends, were used to ensure assay specificity. The primer design followed the optimal primer recommendations, such as: (i) a primer length around 18–30 bases; (ii) a GC content > 40 (whenever possible, because of the high A/T content of Plasmodium genomes); (iii) a melting temperature between 50°C and 62°C; and (iv) an absence or reduced regions of secondary structure, intra-primer homology (self-dimer), or inter-primer homology (primer dimers). Primers/probes with the highest stringency in the parameters of the Primer Express™ software (Applied Biosystems™) and Oligo (Molecular Biology Insights, Inc) were chosen ( Supplementary Figure 1 ). Probes for the cox1 gene assays contained minor groove binders (MGBs) and for the cytb gene and non-coding regions assays, two different quenchers were used: an internal quencher, ZEN, and a Black Hole Quencher™ at the 5′ end.

For the qPCR assays, the optimum concentrations of the primers and probes were defined using a concentration gradient (0.3–0.9 µM of each primer and 0.15–0.25 µM of each probe) ( Supplementary Table 2 ). A gradient of annealing temperatures (52°C–63°C) was also tested. To improve the specificity of qPCR amplification, PCR additives/enhancers [e.g., dimethyl sulfoxide (DMSO), glycerol, bovine serum albumin (BSA)] were used in different concentrations (21) ( Supplementary Table 2 ). All assays used TaqMan^TM Universal PCR Master Mix (Applied Biosystems), and the reactions were standardized on 384-well plates using QuantStudio® 12K Flex Real-Time PCR System (Applied Biosystems) for the cox1 gene assays or the ViiA™ 7 Real-Time PCR System (Applied Biosystems) for the cytb and non-coding regions assays. All results were analyzed using QuantStudio Real-Time PCR Software (Applied Biosystems) or QuantStudio 12K Flex Real-Time PCR Software (Applied Biosystems).

Positive and negative (no-DNA) controls were used in each round of amplification. The positive controls in the qPCR assays used DNA from Plasmodium species previously diagnosed by well-trained microscopy and/or by other molecular tests (nested PCR and qPCR) (5, 16): (i) P. falciparum DNA strain 3D7, which was maintained in the Laboratory of Malaria at IRR, (ii) DNA extracted from the blood of patients with high parasitemia for P. vivax, (iii) DNA of P. simium from a NHP (Alouatta guariba clamitans—MB), and (iv) DNA of P. brasilianum from the MR4 Malaria Research and Reference Resource Center [American Type Culture Collection (ATCC), USA].

To prevent cross-contamination, the DNA extraction and reaction mix preparation were performed in “parasite DNA-free rooms” distinct from each other. Each of these separate areas had different sets of pipettes and all procedures were performed using plugged pipette tips. Furthermore, DNA extraction was performed twice on different days, and for each sample with non-specific amplification, two or three independent PCR reactions were performed using the same conditions (i.e., with reagents and thermal cycler).

A cutoff value for each assay was established by the receiver operating characteristic (ROC) curve given by MedCalc for Windows, version 20.123 (MedCalc Software, Ostend, Belgium), which considered the positive Cq values of specific and non-specific amplifications. The Cq values from positive samples were considered “true positives”, whereas for the negative samples (i.e., negative for the species to be tested) were considered “true negatives”. These samples were previously diagnosed by molecular tests with ribosomal (5) and non-ribosomal targets (16). Thus, the combined results of these two molecular tests were used as the reference to estimate the sensitivity and specificity of each mt assay. From each cutoff value, the true- and false-negative/-positive samples were determined using the R package, as well as the sensitivity and specificity values. The optimal value for the cutoff was determined as a specificity of ≥ 90%.

The overlap in the Cq values between specific and non-specific amplifications for each assay was established as the percentage of samples showing similar Cq values of non-specific and specific amplifications, that is, overlap occurred when Cq values between specific and non-specific amplifications superimposed each other and included the same Cq data (Cq min. non-specific < x < Cq max. specific).

Seven TaqMan-based qPCR assays were designed for the detection of P. falciparum and P. vivax. Coding genes and non-coding regions of the Plasmodium mt genome were targeted in the assays: cox1 gene (PV_COX1_1; PV_COX1_2), cytb gene (PV_CYTB, PF_CYTB), and non-coding regions (PF_NC_1; PF_NC_2; PV_NC) ( Figure 1 and Supplementary Figure 1 ).

Although the mt genome is highly conserved among Plasmodium species, the sequence alignments of the amplicon, primers, and probes of each assay showed the presence of some single-nucleotide polymorphisms (SNPs) used to distinguish the different Plasmodium species (particularly at the 3′ ends of the primers) ( Supplementary Figure 1 ).

A wide screening to verify the cross-reaction between the Plasmodium species or amplification using human malaria-free samples was performed in all Plasmodium mt assays. A panel of 153 Plasmodium samples, which had been diagnosed by well-trained microscopists and/or had been previously assayed with ribosomal and non-ribosomal targets and samples from 26 healthy volunteers, were used. Non-specific amplification was observed in all assays for other Plasmodium species and/or human negative control samples, as described below. The geometric mean of parasitemia was 2,528 parasites/μL (CI₉₅ 1,914–3,338) for P. vivax samples and 1,697 parasites/μL (CI₉₅ 1,009–2,854) for P. falciparum samples used in the assays.

For the cox1 gene assays (PV_COX1_1; PV_COX1_2), optimization included different primer and probe concentrations and DNA quantities. To improve the specificity of qPCR amplification, PCR enhancers (DMSO, glycerol, BSA) were also evaluated. For both cytochrome oxidase 1 assays, no non-specific amplifications with human malaria-free samples (0/15) were observed. The Plasmodium vivax PV_COX1_1 assay was tested with 40 P. vivax samples, all of them amplified (Cq mean = 29.0; range = 20.6–38.2) and did not cross-react with any of 30 P. simium samples. However, non-specific amplification was identified in five out of nine (55.5%) reactions using well-characterized P. falciparum samples (with known parasitemia and confirmed by other PCR protocols to exclude co-infection) amplified with a Cq mean value of 39.2 (range = 36.6–40.8) ( Supplementary Table 2 ). To reduce non-specific amplification, the probe PV_COX1_2 was redesigned by including nucleotides spanning a more polymorphic region. Using the new probe, 49/50 (98%) P. vivax samples were amplified correctly and did not cross-react with any of the 20 P. simium samples or with the 15 P. malariae/P. brasilianum samples. Nevertheless, 8 out of 38 (21.0%) reactions showed non-specific amplification using P. falciparum samples, even using the enhancer DMSO ( Supplementary Table 2 ). The Cq mean values were 42.3 (range = 35.3–44.6) and 32.4 (range = 24.2–42.0), for non-specific and specific amplifications, respectively ( Supplementary Table 2 ).

The assay targeting the non-coding region, PV_NC, was tested with 19 P. vivax samples, all of them amplified (Cq mean = 23.0; range = 18.0–32.4). Non-specific amplification was identified in 8 out of 29 (27.6%) reactions using well-characterized P. falciparum samples (Cq mean = 37.4; range = 35.2–39.9) and in 1 out of 56 (1.8%) reactions using human uninfected controls (Cq = 39.3) ( Supplementary Table 1 ).

For the PV_CYTB assay, which uses the cytb gene as target, 22 out of 23 (95.6%) P. vivax samples were amplified correctly (Cq mean = 30.6; range = 27.0–37.4). However, non-specific amplification was identified in three out of six (50%) reactions using P. falciparum, P. simium, or P. brasilianum (Cq mean = 34.8; range = 32.7–38.8) and in 8 out of 68 (11.8%) reactions using human uninfected controls (Cq mean = 34.5; range = 30.7–38.6) ( Supplementary Table 2 ).

In the assay targeting the P. falciparum cytb gene (PF_CYTB), 33 out of 48 (68.7%) reactions using P. falciparum samples (Cq mean = 23.8; range = 14.7–33.3) were amplified correctly. However, non-specific amplification was identified in five out of six (83.3%) reactions using P. vivax, P. simium, or P. malariae/ P. brasilianum (Cq mean = 38.3; range = 38.0–38.4) and in 29 out of 104 (27.9%) reactions using human uninfected control (Cq mean = 37.2; range = 36.1–39.8) ( Supplementary Table 2 ). To reduce non-specific amplification, different DMSO concentrations were used: 1%–5% without significant improvement for specificity (data not shown).

For the PF_NC_1 assay, based on different regions of the non-coding mt genome, 12 out of 13 (92.3%) reactions using P. falciparum samples were amplified correctly (Cq mean = 24.9; range = 19.3–29.7), while 7 out of 18 (38.9%) and 18 out of 42 (42.9%) reactions showed non-specific amplification for P. vivax (Cq mean = 35.0; range = 29.2–37.4) and human uninfected control samples (Cq mean = 35.3; range = 34.6–38.7), respectively ( Supplementary Table 2 ). For the PF_NC_2 assay, also based on the non-coding region, 10 out of 14 (71.4%) reactions using P. falciparum samples were amplified correctly (Cq mean = 20.9; range = 12.7–34.6), whereas two out of nine (22.2%) P. vivax reactions (Cq mean = 34.7; range = 33.4–36.0) and 9 out of 43 (20.9%) reactions using human uninfected control (Cq mean = 36.3; range = 33.0–39.7) showed non-specific amplification ( Supplementary Table 2 ).

A cutoff value for each assay was stablished by the ROC curve considering the positive Cq values of specific and non-specific amplifications. From each cutoff value, the true and false negative/positive samples were determined, as well as the sensitivity and specificity values ( Figure 2 ). The Figure 2A shows the accuracy for each assay. Comparing all assays, PV_CYTB presents the best accuracy for P. vivax detection, with 95% of the sensitivity and 96% of the specificity. The PF_CYTB assay was the only P. falciparum assay with a specificity of > 90%. For these two assays with best results for each Plasmodium species, the cutoff was not able to totally discriminate the Cq values for specific and non-specific amplifications ( Figure 2B ).

Initially, to minimize the non-specific amplification, a set of parameters was modified, such as the concentration of primers/probes in each assay, the denaturing and annealing conditions, and the number of cycles. The primer concentrations varied from 0.3–0.9 µM and from 0.15–0.25 µM for probes. The annealing temperature tested varied from 52°C–63°C. An incremental annealing temperature in progressive cycles was also tested ( Supplementary Table 2 ). These modifications decreased the number of non-specific amplifications in some cases, but did not eliminate all of them. However, the Cq values for the specific amplifications increased, showing that changes in annealing temperature and primer/probe concentrations interfered with assay sensitivity ( Supplementary Table 2 ). A variety of PCR additives/enhancers (DMSO, glycerol and BSA) has also been used to increase the specificity, yield, or consistency of reactions ( Supplementary Table 2 ). The reduction of non-specific amplification was notable in most of the samples; however, the use of these additives hampered the efficacy of specific amplifications. For the PV_COX1_2 assay, 5% of DMSO, glycerol, or BSA was individually tested, and only DMSO was able to reduce non-specific amplifications. Nevertheless, the use of additives also had an impact on the specific amplifications, increasing the Cq values ( Supplementary Table 3 ). Different concentrations of DMSO and DNA were also tested. The best combination to avoid non-specific amplification without interference with specific amplification was using 3.5% DMSO and 2 μL of DNA ( Supplementary Table 3 ). Then, a large panel (n = 30) of well-characterized P. falciparum samples were screened using this combination of DMSO, and eight samples were amplified, with a mean Cq of 42.3 (range = 35.3–44.6) ( Supplementary Table 2 ). None of the adopted strategies eliminated the non-specific amplification. For the PF_CYTB, all concentrations of DMSO tested (1–5%) were not able to eliminate the non-specific amplification (using 52 samples, including P. vivax and human uninfected control, data not shown). In a single experiment with 1% DMSO, 3 out of 29 human uninfected control samples were non-specifically amplified for P. falciparum, with a Cq mean of 37.4 (range = 37.2–37.8), whereas 13 out of 19 P. falciparum samples were amplified specifically (Cq mean = 24.5, range = 14.7–32.3). Once again, the use of DMSO hampered the efficiency of specific amplifications, and it explains the decreased sensibility of PF_CYTB assay, which amplified only 68% of specific samples ( Supplementary Table 2 ).