Reverse transcription PCR to detect low density malaria infections [version 3; peer review: 2 approved with reservations, 1 not approved] transcription PCR to detect low density

Background: Targeted malaria elimination strategies require highly sensitive tests to detect low density malaria infections (LDMI). Commonly used methods for malaria diagnosis such as light microscopy and antigen-based rapid diagnostic tests (RDTs) are not sensitive enough for reliable identification of infections with parasitaemia below 200 parasites per milliliter of blood. While targeted malaria elimination efforts on the Thailand-Myanmar border have successfully used high sample volume ultrasensitive quantitative PCR (uPCR) to determine malaria prevalence, the necessity for venous collection and processing of large quantities of 95.1%. Conclusion: Malaria detection in areas of low transmission and LDMI can benefit from the increased sensitivity of ribosomal RNA detection by RT-PCR, especially where sample volume is limited. Isolation of high quality RNA also allows for downstream analysis of malaria transcripts. Any reports and responses or comments on the article can be found at the end of the article. Abstract: What was the target used? Was all PCR targeted to a specific gender? What was the gold standard used? The method


Background
As malaria burden reduces globally, the international community is working toward its elimination. Successful targeted malaria elimination strategies will require increased surveillance and highly sensitive tests capable of detecting asymptomatic and low density malaria infection (LDMI). These infections are often well below 200 parasites per milliliter and are an important disease reservoir capable of transmitting malaria that must be detected and eliminated 1,2 . Light microscopy and antigen based rapid diagnostic tests (RDTs) are the most common tests used in malaria prevalence surveys, and usually assess 5 μL of whole blood per test, a volume which precludes reliable detection of LDMI. Ultrasensitive RDTs improve detection sensitivity of patients with a parasitaemia between 200 parasites/mL and 10,000 parasites/mL 3 , but are still limited by their low input volume. Molecular methods using the polymerase chain reaction (PCR) remain the only common and reliable method to detect LDMI. The sensitivity of PCR is due to its ability to detect single, specific nucleic acid molecules and use of concentrated DNA from a large sample. Widespread use of PCR based assays, namely real-time quantitative PCR (qPCR) and reverse-transcription PCR (RT-PCR), have revealed a new landscape of malaria prevalence particularly in low transmission areas 4,5 .
The targeted malaria elimination project (TME) on the Eastern Myanmar border used a high blood volume ultrasensitive qPCR (uPCR) to consistently detect parasitaemia down to 22 parasites per mL 6 , and revealed a high proportion of LDMI 7 . Major features of uPCR are its 7 copies of gene target, the high volume of blood tested and the ability to accurately quantify low density parasitaemia. Although increasing the blood volume of a PCR leads to higher sensitivity 6 , the collection of large numbers of high volume samples have their own specific limitations. These can include, the ethics approval required for venous blood draw, sample logistics, increased nucleic acid extraction cost and increased sample processing time. Another way to increase the sensitivity of PCR is to increase the number of specific nucleic acid targets per parasite by targeting specific RNA and DNA using RT-PCR. As previously reported by Kamau et al. 8 , the primer set used in uPCR can be made more sensitive by incorporating reverse transcription prior to qPCR, enabling detection of the 7 genes encoding Plasmodium 18S ribosomal nucleic acid (rRNA) and its rRNA transcripts.
In this study, we compare the sensitivity and specificity of high sample volume, low target copy number ultrasensitive qPCRs, with reduced sample volume, high target copy number RT-qPCR for the detection and quantification of Plasmodium spp. and P. falciparum.

Methods
We selected 304 samples with previously reported uPCR results: 21 P. falciparum positive and 283 Plasmodium spp. negative for comparison to RT-qPCR with increased target numbers per parasite but 30% of the sample volume.
Study area, sample collection and malaria screen Active case detection samples were collected from rural Eastern Kayin (Karen) state of Myanmar between 2014 and 2015 as part of an international malaria elimination project. Full methods have been published 9 , briefly, 3 ml of blood was drawn into an EDTA container from each adult, and transported on ice to the Shoklo Malaria Research Unit in Mae Sot, Tak, Thailand. Within 48 hours the samples were processed, and two aliquots of up to 500 μL of packed red blood cells (PRBC) were stored at -80°C. The rapid diagnostic, SD Bioline Malaria Ag P.f/Pan POCT and light microscopy of giemsas stained thick and thin smears was also done. Only 500 μL sample aliquots were accepted for uPCR detection and quantification while RT-qPCR included samples with a second cryopreserved 150 μL PRBC aliquot.
Ultrasensitive qPCR (uPCR) DNA was extracted from 500 μl of cryopreserved PRBC using QIAamp DNA Blood Midi Kit (Cat. No. 51185, Qiagen™) according to manufacturer's instructions. The DNA template was then dried in a vacuum concentrator, re-suspended in 10 μL of PCR grade water and stored at -20°C prior to qPCR. Separate uPCRs specific for Plasmodium spp., P. falciparum and P. vivax were performed over 3 years from 2013 as previously described 6 . Briefly, uPCR was done in 10 μL reactions that contained 2 μL of DNA template with 1x QuantiTect Multiplex PCR No ROX mastermix (Cat. No. 204743, Qiagen™), 0.4 μM each primer, and 0.2 μM Taqman probe. Thermal cycling and signal acquisition was done on an ABI 7500 Fast real-time PCR machine with initial denaturation and enzyme activation at 95°C for 15 min, then 50 cycles of denaturation at 94°C for 15 sec followed by annealing and extension at 60°C for 60 sec. A reaction with exponential signal increase before cycle 40 was considered positive.

Sample selection
Within 3 years of sample storage at -80°C, nucleic acid extraction and RT-qPCR assays were performed on the second aliquot of PRBC for selected samples. Selected samples included 283 uPCR negative samples and 21 uPCR P. falciparum positives, this includes 18 mixed infections with P. vivax and 13 LDMI defined as >50 parasites/μL by uPCR, parasitaemia of samples ranged from 17.5 -9,907,000 parasites/μL.

REVISED
DNase enzyme wasn't used, and RNA was eluted in 20 μL of molecular grade water. Two RT-qPCRs were performed on each extract in duplicate: the Plasmodium spp. specific assay using the same primer and probe set as uPCR, and a P. falciparum specific set targeting the DNA of the A-18S rRNA genes and its rRNA transcripts 5 . Both reactions had a final volume of 15 μL and contained 4 μL of RNA template, 1x Superscript III One-Step RT-PCR System master mix (Cat. No. 12574018, ThermoFisher Scientific™), 0.4 μM forward and reverse oligo primer and 0.2 μM of MGB Taqman probe. Amplification and signal acquisition were done on an ABI 7500 Fast real-time PCR machine with cycling conditions as follows: reverse transcription at 45°C for 30 min, enzyme activation at 95°C for 2 min, followed by 50 cycles of denaturation at 95°C for 15 sec and combined annealing and extension steps at 60°C for 60 sec.
A reaction with exponential signal increase before cycle 40 was considered positive.

Standard curve
Standard reference curves for the RT-PCR and uPCR were made using aliquots of 10,000 flow cytometry sorted small ring stage P. falciparum (3D7) parasites 11 . The method used to extract the nucleic acids from these parasites depended on the assay used (RT-PCR or uPCR). For the uPCR standard curve, Qiagen DNA Blood Mini Kit (Cat. No. 51106, Qiagen) was used to extract DNA, this was eluted in 200 μL of sterile water, dried in a partial vacuum at 30°C for consistency with sample extraction, and re-suspended in 200 μL of Qiagen AE buffer 6 . Nucleic acid for the RT-PCR standard curve was extracted using the Zymo Quick-RNA Miniprep (Plus) (Cat. No. R1058, Zymo Research) kit as above but eluted in 40 μL of water. Serial one in five dilutions were made with these extracts to make 7 standards. The uPCR standard curve ranged from 100 to 0.032 parasites per qPCR reaction and the RT-PCR standard curve ranged from 1000 parasites to 0.064 parasites per reaction.

Analysis
Plasmodium spp. and P. falciparum diagnostic accuracy was calculated by comparing RT-PCR results to the uPCR result in 2 × 2 cross tabulation with uPCR as gold standard in STATA version 15.1 (StataCorp LLC, USA). Results of both RT-PCRs were also combined and compared to uPCR, where a positive test by at least one RT-PCR was considered positive. Continuous non-parametric data were compared with Kruskal-Wallis equality-of-populations rank test. P values <0.05 were considered statistically significant. Analyses were performed using STATA version 15.1. Agreement between the two Plasmodium specific PCR quantifications (spp RT-PCR and uPCR) was determined using Bland-Altman analysis of log transformed data (difference vs average) using GraphPad Prism version 6.07 for Windows, GraphPad Software (La Jolla California USA).

Discussion
LDMI detection is essential for effective targeted elimination programs, necessitating the careful selection of a detection assay that is appropriate to the setting and study requirements. Sample volume, storage conditions and transit time are important factors, as well as the required assay sensitivity, specificity and its cost. RNA is generally less stable than DNA, and accurate RNA quantification often requires normalization due to variable transcription rates. It is for these reasons that DNA based qPCR was chosen for the original study as this approach enabled accurate quantification of LDMI in a setting where samples from remote locations would likely experience delays.
The targeted malaria elimination (TME) project on the Eastern Myanmar border used conserved regions of the 18S rRNA genes as the target for qPCR. This high sample volume uPCR assay was modified from the RT-qPCR published by Kamau et al. 8 for their detection of low density malaria infections (LDMI). After uPCR detected a high prevalence of LDMI in the region 9 , and with continued surveillance in mind, we wanted a lower sample volume assay with similar LDMI detection sensitivity. As Kamau et al. has shown, using this primer set as a RT-PCR significantly increases the sensitivity of the assay. The increased sensitivity of RT-PCR is due to the increased number of targets per parasite (compared to uPCR). These amplification targets include the 18S rRNA genes, and the structural RNA of each ribosome. Plasmodium genus specific uPCR amplifies type A and S 18S ribosomal RNA genes distributed on chromosomes 1, 5, 7, 11 and 13. A positive qPCR reaction requires at least one of these genes to be included in the assay. Alternatively, RT-PCR can detect these gene loci, and their gene transcripts (rRNA). The increased target copy number per parasite means a smaller fraction of parasite is needed to provide a positive reaction, leading to less false negatives and an opportunity for further downstream applications. The downside to increased sensitivity and target copy is the variable nature of gene expression and relative fragility of RNA molecules. In general this makes accurate quantification of parasitaemia by RT-qPCR more challenging, and is one of the reasons for outliers in our inter-assay quantification comparison. In future studies the use of a comprehensive methodology tailored specifically to RNA preservation, extraction and amplification would improve sensitivity and allow for downstream applications involving whole-transcriptome profiling important for studies on transmission, pathogenesis and virulence 13 .
While there were obvious differences in quantification for some samples by each method, these did not reach statistical significance. One reason for these outliers was the use of different stored aliquot volumes between assay types. The uPCR aliquot was fixed at 500 μL while the second aliquot, which was used for RT-qPCR ranged from 150 μL to 1 mL prior to freezing. Because RT-qPCR targeting rRNA is capable of detecting tiny fractions of a single parasite, the original lysed sample volume becomes an important detail. Assuming a single freeze thaw lyses blood stage Plasmodium and a Plasmodium assay has a hundred thousand targets per parasite, then a single lysed parasite divided into one hundred aliquots can produce one hundred positive reactions, if the parasite wasn't lysed beforehand then only 1 of 100 would be positive. This can lead to people describing their assay sensitivity well below the sample volume used, a theoretical impossibility unless detecting free circulating nucleic acids outside of parasite cells.
Alternatively, LDMI detection relying on DNA, will have a significant reduction in sensitivity if only a fraction of the DNA template is tested. This is exemplified by the lowest concentration standard used for quantification in uPCR. This standard theoretically contains 0.032 parasites per PCR reaction, at this concentration there is a 1 in 5 chance for the reaction to be positive (7 target genes per parasite × 0.032 = 0.2 copies per reaction). These factors need to be considered when setting up a qPCR standard curve for LDMI quantification. A reliable standard curve for qPCR requires at least one copy of its target at the lowest concentration.

Figure 2. Bland-Altman analysis (difference vs average) of uPCR and RT-qPCR quantification data from the TME malaria survey of Kayin state, Myanmar.
While many consider light microscopy the gold standard for malaria diagnosis, its lack of sensitivity at parasitaemia below 50 parasites/μL makes its use impractical for analysis of LDMI (13 of 21 uPCR positive samples were LDMI). Ultrasensitive PCR has been assessed for sensitivity in previous studies with a limit of detection at 22 parasites per mL 6 .

Conclusion
The success of any LDMI detection protocol relies on the careful consideration of the following factors: sample volume, elution volume, template volume per assay reaction, primer set target and assay type (uPCR or RT-qPCR). Our experience of the different types of assay suggest that for a LDMI program requiring highly sensitive, accurate quantification and where venous blood collection is possible, uPCR is recommended. In an environment where blood volume is limited (i.e. finger prick sampling) and quantification accuracy of parasitaemia is less important, RT-qPCR is a suitable alternative.

Declaration
Ethics approval and consent to participate First, community engagement teams sought community approval ahead of the survey date. Then, survey participants received individual information in their language, and informed consent was obtained from each individual before they provided a venous blood sample. Appropriate treatment for Plasmodium falciparum or Plasmodium vivax was available for all RDT-positive individuals. This project contains the following underlying data: -Plasmodium quantifications_Parasites per mL.xls (Contains Plasmodium spp., P. falciparum quantifications (parasite per mL whole blood) and the threshold cycle for all RT-qPCRs) Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).
could use microscopy but that has much lower sensitivity. I would need to repeat the Kamau et al. paper with changes to the sample preparation, sample volume, elution volume and kit brand if I want to determine and compare the low limits of detection. Although PCR is a less sensitive technology, the uPCR contains DNA from a larger sample volume and is more likely to include a parasite from patient samples with very low parasitaemias. To confuse matters further, both assays can detect a single parasite in a given sample volume, if the whole sample is tested. Problems arise when the PCR template is diluted and split into multiple assays.
If contamination is the cause of ultra-sensitive detection, is there a way to prevent it? If not, your RT-qPCR in endemic areas may not be useful. Please specify where in your manuscript the reviewer's comments were or were not reflected.

1.
Reply: The uPCR is high throughput and used to survey many 1000s of samples regularly. The flow through contains multiple negative and positive controls (DNA extract controls and mastermix controls) per test plate. They also have sample left over to analyse and check.
Using this data they found standard contamination to be a factor in some runs. There's no way to determine if the unique RT-qPCR positives are true or not in this study. The RT-qPCR assay included negative controls in all runs (and they were negative).
Removing the standard and making the assay qualitative is one way to remove possible contamination issues, but a positive control will always be needed and represent a source of contamination. I'm also not convinced that these are all contamination. RT-qPCR is a more sensitive technique, and could be detecting true positives (especially when taking an aliquot from a frozen/lysed sample -unfortunately I didn't record which samples had more than 150ul PRBC remaining and is a limitation). There's just no way to know when using clinical samples from asymptomatic people for evaluation. In this paper I wanted to answer a simple question "can I use a lower volume sample and a more sensitive technique to achieve similar sensitivity to uPCR?" I can add a point in the discussion on the importance of including multiple true negative controls, especially in field settings when using highly sensitive techniques. In Abstract, you have defined the real-time reverse transcription PCR you performed as RT-qPCR. qPCR is an abbreviation for quantitative PCR. If this PCR method is qualitative, it should be clearly stated and changed to an appropriate abbreviation.

1.
Reply: The 2 assays were run in duplicate with a high quality standard curve. There has been some discussion amongst authors as to what we are allowed to call the assay. In my opinion it is a quantitative assay (and bland-altman analysis showed reasonable agreement between the 2 assays). Unfortunately the assay doesn't follow MIQE guidelines for RT-qPCR, so one of the authors wanted it changed to RT-PCR. I recently changed all back to RT-qPCR when referring to the assay, after comments from another reviewer. Dear Reviewer, thanks for you time reviewing our paper. I have responded to your questions below. Figure 1 are these data in triplicate, can error bars be added? 1.
Unfortunately not. This is one reason we called it RT-PCR not RT-qPCR (the other is a lack of internal control). We were trying to add as much concentrated sample into the reaction as possible to have detection sensitivity similar to uPCR. This was the catch, we are forced to use as much sample as possible per reaction to achieve high sensitivity, but we don't want the logistics issues of large sample volumes. So we had to decide; either we decrease the assay sensitivity to increase its quantification accuracy, or put the whole sample into the PCR reaction. As this PCR will be used for 'Hot Spot' detection at the village level, we prioritized high sensitivity over high precision.

2.
Figure 2 besides Bland-Altman can further analysis be added to add some more depth to the paper?

3.
I'm hesitant to analyse the quantifications further as they're not high quality (not triplicate, not normalised) and accurate quantification of LDMI is not particularly important (low detection limit is more important).

4.
I'll be adding microscopy and RDT results to the next draft and more information on the species present. Some of these samples were mixed Pf+Pv infections.
there any risk of detecting rRNA of other pathogens by performing RT-qPCR? It is important to evaluate whether this method is suitable for actual blood samples The 18SrRNA primers are unique to Plasmodium, so false positives are either real positives or, more likely, contamination from the standard during plate preparation (false positive from uPCR (Imwong 2014) were found to be caused by standard contamination). Unfortunately, we were trying to include as much sample as possible to to increase sensitivity of the PCR, so we had no sample left over to investigate these false positives further.

4.
In the paper, what should be written as RT-qPCR was written as RT-PCR. The authors need to revise this.

5.
Our PCR doesn't meet the criteria for RT-qPCR so one of the authors wasn't happy naming it 'qPCR' (The PCR is not in triplicate, and doesn't have an internal control for normalising quantification). I understand the confusion, because we go on to compare quantification (which I assumed would be different but to my surprise wasn't). In my opinion the quantification of LDMI is less useful than the low limit of detection and could be removed entirely from the paper (especially without knowing the life stage -ring vs schizont will have vastly different ribosome quantities).
malaria? Which parasitemia of the people studied to study low parasitemia? If used, was gold pattern used as a thick blood smear or RDT? The target for RT-PCR was for Plasmodium? Two RT-PCRs were performed on each extract in duplicate: Plasmodium spp. and are they sensitive enough? Using uPCR as a gold standard would not be the most appropriate use of the thick blood smear as a control. It is not clear that he will use a method for gender and species for methodology.

Results:
Using uPCR as a gold standard is not the most suitable. How to be sure of positive positives and false negatives? Why was parasitemia not compared in the studied methods to verify the comparison between the methods to verify the smallest parasitemia that it would detect? In quantification, it is more appropriate to perform correlation analysis to compare a parasitemia between the methods.

Discussion:
The targets should have been described in the methodology as well. What are the limitations of the study?

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly Competing Interests: No competing interests were disclosed. I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Author Response 22 Apr 2021
Peter Christensen, Mahidol University, Maesot, Thailand Thank you for your time reviewing this paper, I appreciate the effort.