Background
The transmission of malaria from humans to mosquitoes requires the presence of mature sexual stage parasites (gametocytes) in the peripheral blood. Upon ingestion by
Anopheles mosquitoes and following sporogonic development, these gametocytes can render mosquitoes infectious to humans. Quantifying the density of these gametocytes in the peripheral blood is a fundamental part of estimating the infectiousness of individuals with malaria infections. For over a century, light microscopy was the only method available to quantify malaria parasites, including gametocytes. The development of sensitive molecular methods has uncovered some of the limitations of microscopy. In endemic areas, a substantial proportion of infected individuals carry parasites at levels below the microscopic threshold of detection [
1‐
4], and many of these individuals also have low, sub-microscopic, densities of gametocytes [
5,
6]. Together with early observations that individuals with no gametocytes detected by microscopy could be infectious to
Anopheles mosquitoes [
7], these studies suggest that the use of more sensitive methods is necessary to characterize the infectious reservoir of malaria [
8,
9].
The most widely used target for molecular gametocyte detection and quantification is
Pfs25 mRNA [
9], which is gametocyte-specific and highly conserved among different parasite isolates [
6]. Although expression of
Pfs25 mRNA is upregulated in female gametocytes [
10], the female-biased gametocyte sex ratio and the high abundance of
Pfs25 mRNA relative to the male specific marker
Pfs230p make it an attractive target for sensitive gametocyte detection and quantification [
11].
Various molecular techniques have been used for gametocyte detection and quantification. For
Pfs25 mRNA there are two commonly used protocols, one based on quantitative reverse transcriptase PCR (qRT-PCR) and another based on quantitative nucleic acid sequence based amplification (QT-NASBA) [
6,
12]. Although previous studies on other malaria-specific targets suggest that the accuracy, precision and operational attractiveness may differ between assays [
13], QT-NASBA and qRT-PCR, as currently routinely used, have never been directly compared for
Plasmodium falciparum gametocyte quantification. Additionally, differences in
Pfs25 transcript levels between malaria parasite strains have been hypothesized but never directly examined [
14].
Here,
Pfs25 mRNA QT-NASBA and qRT-PCR were compared for intra- and inter-assay variation using gametocytes from three different isolates [
6,
12]. The association between gametocyte densities estimated by these two molecular methods in natural malaria infections and infectiousness quantified by membrane feeding assays was also determined.
Discussion
In this study, Pfs25 mRNA QT-NASBA and qRT-PCR were evaluated using culture-derived gametocyte samples and samples from naturally infected individuals. In in vitro experiments, both intra- and inter-assay variations were lower for qRT-PCR compared to QT-NASBA. When analysing samples from naturally infected individuals, gametocyte densities estimated by qRT-PCR were higher than densities estimated by QT-NASBA and microscopy in individuals with high gametocyte levels. Gametocyte densities estimated by the two molecular assays, as well as by microscopy, correlated well with mosquito infection rates. Overall, the findings indicate that both assays are useful for quantifying sexual stage parasite densities in samples from naturally infected individuals, but qRT-PCR showed better reproducibility.
Determining the variability of quantitative assays is important for the interpretation of results. The experiments here suggest that qRT-PCR has lower intra- and inter-assay variation compared to QT-NASBA. The higher precision of qRT-PCR is evident for the entire range of gametocyte densities used in the dilution series, despite involving extra reaction steps, such as DNase treatment and cDNA production, which are known to be a source of variation [
27]. These two steps were performed once for each dilution series density, for the three different NF54 cultures and the NF135 and the NF166 cultures. There was considerably less between-strain and -culture variation for similar densities in qRT-PCR
versus QT-NASBA. In QT-NASBA reactions, from the time of primer depletion, amplification will mostly depend on the initial number of transcripts, resulting in a linear increase of amplicons. This exact moment when amplification starts may differ between samples depending on when and how the enzyme is added into the reaction wells, which adds a level of uncertainty to this methodology [
28]. The range of TTP values in dilution series is also sensitive to the KCl concentration in the mix, which requires regular optimization with new targets, batches of primers or molecular beacons [
29].
Despite qRT-PCR being a more precise method for gametocyte quantification, QT-NASBA might detect a larger number of very low gametocyte densities (<0.1 gametocytes per μL). QT-NASBA was originally developed for highly sensitive qualitative detection. The apparent higher sensitivity of QT-NASBA compared to qRT-PCR may be explained by methodological differences: for QT-NASBA it is not necessary to dilute samples, in order to complete DNase treatment and cDNA production (Additional file
1: Figure S1) and the equivalent of a larger volume of the original sample is therefore added into the final reaction in QT-NASBA compared to qRT-PCR. This additional dilution in the qRT-PCR protocol could be avoided by using column-based RNA extraction with on-column DNase digestion [
6]. Also, while qRT-PCR, in the protocol used here, carries on for a set amount of 40 cycles, the QT-NASBA reaction is continuous at a stable temperature for the duration of 90 min. This might allow for more sensitive amplification of low numbers of transcripts, but could also lead to the detection of single stranded DNA, in cases in which genomic DNA has been degraded and exists in the sample in a single stranded form. A stringent cut-off density below which samples are considered negative must therefore be applied in QT-NASBA, as with all other molecular assays.
Whether qRT-PCR overestimates true gametocyte levels or QT-NASBA and microscopy underestimate densities is not clear. When qRT-PCR is compared with the other methods, high mean densities correspond to the qRT-PCR measurement being larger than the other measurements. There are two possible reasons for observing this pattern: (1) qRT-PCR overestimates high gametocyte concentrations; or, (2) qRT-PCR measurements have greater variability [
30]. Since the in vitro experiments suggest that qRT-PCR has lower variability than QT-NASBA, it is likely that qRT-PCR has an upward bias at high gametocyte concentrations (>100 gametocytes per μL). The overestimation is unlikely to be caused by residual DNA in the RNA samples used in the qRT-PCR analysis, as the success of the DNase treatment was tested for two to five samples per plate, with no amplification detected.
The value of molecular methods in detecting epidemiologically relevant malaria infections is becoming increasingly well established. A recent meta-analysis comparing PCR, microscopy and rapid diagnostic test (RDT) sensitivities showed that on average RDTs detect less than half of PCR-positive
P. falciparum infections, and microscopy detects slightly less infections than RDTs [
8]. The density of these sub-microscopic infections and their concurrent gametocyte levels are of relevance in estimating the contribution to malaria transmission. For
P. falciparum parasite density (including asexual parasites and gametocytes), RNA-based QT-NASBA was previously compared to DNA-based qPCR for the quantification of
P. falciparum 18S RNA and DNA, respectively [
13]. Another recent study, this time on
Leishmania parasites, compared
18S rRNA-based quantification by qRT-PCR and QT-NASBA to
18S rRNA gene DNA-based detection with qPCR [
31]. In the first study, QT-NASBA was found to be more convenient and equally applicable for quantification purposes as real-time PCR, with strong correlation with quantification by microscopy and similar inter-assay variation to both methods [
13]. In the second study, intra- and inter-assay CV were deemed equal for all three molecular methods and qRT-PCR was preferred over the other two methods for convenience reasons [
31]. One explanation for these conflicting conclusions is the method used for detection of the amplified target in QT-NASBA: in the study comparing
Leishmania quantification, the electrochemiluminescence (ECL) read-out added to the general workload of the assay. Additionally, co-amplification with a known amount of quantitative (Q)-RNA, as was originally done for the quantification of the different gametocyte stages as well as the
Leishmania parasites, added an extra factor to the analysis of results [
12,
31]. The direct comparison to a standard curve with microscopically determined density of parasites, in addition to the use of molecular beacons, simplified the quantification process of QT-NASBA results [
13]. As mentioned earlier in the context of assay sensitivity, also convenience-wise, an important advantage of QT-NASBA is that it can be performed directly on extracted NA.
Several factors, such as the type of anticoagulant used in sample collection and the storage conditions, might influence the outcome of molecular assays on samples [
9,
32]. The analysis of samples from Burkina Faso provides indirect evidence for this: the extremely low assigned gametocyte levels, including in samples collected from children with microscopically detectable gametocytes, suggest that these samples might have gone through more than one freeze-thaw cycle. In these situations, sample quality may be compromised: mRNA may be degraded and, although abundant transcripts, such as
Pfs25 mRNA, might still be detectable, quantification becomes less reliable [
33]. Quantification of constitutively expressed human RNA targets and comparison of transcript levels among samples is one method that could have confirmed whether RNA degradation occurred and is recommended for future studies where there is uncertainty about sample integrity.
Ultimately, the goal in quantifying gametocytes is to indirectly estimate human infectivity, as mosquito feeding assays are logistically complex and only a handful of research institutes currently have the infrastructure to perform these experiments in sub-Saharan Africa. In this study, it was observed that QT-NASBA- and microscopy-defined densities explain a slightly higher proportion of the variation in infectivity compared to gametocyte levels assigned by qRT-PCR but this difference should be interpreted with caution. The statistical method used here to evaluate this relationship was recently employed to assess the association between viraemia and mosquito infection risk in feeding assays involving dengue-infected individuals [
34]. While the sigmoidal curves estimated by these models seem to fit mosquito infection data well for QT-NASBA and microscopy, it is possible that qRT-PCR-defined high densities were overestimated (Fig.
2b) and that could possibly explain the poorer fit of the model for that assay. These comparisons are based on a statistical model. More complex models [
35], allowing for more flexible sigmoidal curves, would be required to formally assess the shape of the association between gametocyte density and mosquito infection rates. The quantification of other parasite (e.g., sex ratio) or host factors (e.g., transmission-blocking immune responses and haematological factors) that might influence infectivity is likely to improve the predictive value of these models [
36]. Of note, the enrolment criterion of multiple gametocytes being observed by microscopy (two or more gametocytes per 500 WBC) makes it difficult to directly extrapolate these results to sub-microscopic infections. By definition, microscopy has no quantitative value in sub-patent infections and molecular methods will need to be used to characterize the gametocytaemia-infectivity curve at low gametocyte densities.
Despite being the most commonly used amplification target for estimating mature
P. falciparum gametocytes counts, the use of
Pfs25 transcripts for quantifying gametocyte densities has drawbacks.
Pfs25 mRNA levels are much lower in male compared to female gametocytes [
11]. Although it is generally assumed, due to the female bias in sex ratios (4–5:1) in natural infections, that female gametocyte levels are a good surrogate for all-sex gametocytes counts, quantification of the less abundant male gametocyte-specific
Pfs230p mRNA or other male targets would provide more accurate estimates of the total number of mature gametocytes as well as sex ratios [
11], a factor known to influence infectiousness [
37]. In addition to these targets, transcribed only in mature gametocytes, transcripts of other genes have been used for the study of sexual stage malaria: an alternative marker of female gametocytes is
pfg377, the expression of which begins in the sequestered stage III [
38]; a traditional marker for commitment to gametocytogenesis is
Pfs16 [
12,
39], which is present in all gametocyte stages. Another early marker of commitment is
PfGEXP5, the expression of which starts in ring-stage parasites committed to gametocytogenesis, is a potentially useful tool to be used in studies aiming to identify factors influencing commitment to sexual stage [
40]. For all of these targets, transcript detection is possible both by the more convenient QT-NASBA and by the more laborious but more reproducible qRT-PCR. Design of intron-spanning primers may close the gap between the attractiveness of both methods and retain the advantage of the more reproducible qRT-PCR [
41].
Authors’ contributions
HEP and BPG contributed equally to the completion of this work, taking part in interpretation and analysis of data as well as writing. AD, IC, JoelleB, TB, and RG contributed to the study design for the trial in Mali. AD and SFT oversaw the data collection in Mali. HD, AM, HMS, and IB contributed to the conduct of the study and the data collection in Mali. BPG, IN, ABT, SBS, TB, and CD took part in coordinating the trial in Burkina Faso. KL completed the laboratory work, taking part in the development of methodologies. JohnB contributed to the analysis and interpretation of data. RB provided the facilities for the laboratory work and took part in useful discussions on data interpretation. IF provided critical assessment and input on the content of this work during completion process. TB contributed to and supervised interpretation and analysis of data, as well as writing. All authors read and approved the final manuscript.