The extended recovery ring-stage survival assay provides a superior association with patient clearance half-life and increases throughput

Background Tracking and understanding artemisinin resistance is key for preventing global setbacks in malaria eradication efforts. The ring-stage survival assay (RSA) is the current gold standard for in vitro artemisinin resistance phenotyping. However, the RSA has several drawbacks: it is relatively low throughput, has high variance due to microscopy readout, and correlates poorly with the current benchmark for in vivo resistance, patient clearance half-life post-artemisinin treatment. Here a modified RSA is presented, the extended Recovery Ring-stage Survival Assay (eRRSA), using 15 cloned patient isolates from Southeast Asia with a range of patient clearance half-lives, including parasite isolates with and without kelch13 mutations. Methods Plasmodium falciparum cultures were synchronized with single layer Percoll during the schizont stage of the intraerythrocytic development cycle. Cultures were left to reinvade to early ring-stage and parasitaemia was quantified using flow cytometry. Cultures were diluted to 2% haematocrit and 0.5% parasitaemia in a 96-well plate to start the assay, allowing for increased throughput and decreased variability between biological replicates. Parasites were treated with 700 nM of dihydroartemisinin or 0.02% dimethyl sulfoxide (DMSO) for 6 h, washed three times in drug-free media, and incubated for 66 or 114 h, when samples were collected and frozen for PCR amplification. A SYBR Green-based quantitative PCR method was used to quantify the fold-change between treated and untreated samples. Results 15 cloned patient isolates from Southeast Asia with a range of patient clearance half-lives were assayed using the eRRSA. Due to the large number of pyknotic and dying parasites at 66 h post-exposure (72 h sample), parasites were grown for an additional cell cycle (114 h post-exposure, 120 h sample), which drastically improved correlation with patient clearance half-life compared to the 66 h post-exposure sample. A Spearman correlation of − 0.8393 between fold change and patient clearance half-life was identified in these 15 isolates from Southeast Asia, which is the strongest correlation reported to date. Conclusions eRRSA drastically increases the efficiency and accuracy of in vitro artemisinin resistance phenotyping compared to the traditional RSA, which paves the way for extensive in vitro phenotyping of hundreds of artemisinin resistant parasites.

filters to select for merozoites [13], and using a dual layer Percoll gradient, as has been done previously in other malaria assays [14,15]. Throughput of these various methods is dictated by the number of samples that can be simultaneously synchronized and prepared for processing. Another major bottleneck and source of variability in the final readout of the RSA is counting viable malaria parasites by microscopy [8,11,14]. To increase throughput, flow cytometry has become heavily utilized as an alternative to counting viable parasites by microscopy, removing hours of counting slides and human error [14,16]. However, staining of cells for flow cytometry to detect viable parasites is time sensitive and requires samples to be prepared immediately after the 66 h incubation, which can be time consuming and inconvenient [14].
Despite these advances in the protocol, the RSA is still far from being both high-throughput and highly reflective of PC 1/2 . Recently, Mukherjee et al. used the RSA to measure the percent survival of 36 culture-adapted parasites, but only showed a correlation with PC 1/2 data of 0.377, suggesting there is still significant room for improvement (Spearman's Rho, internal calculations based off of supplemental data) [17].
Here a modified RSA is presented: the extended ring-stage recovery assay (eRRSA). This modified RSA protocol utilizes a simple single layer Percoll synchronization, flow cytometry to determine the stage and parasitaemia for assay setup, a 96-well plate format for the assay, and a SYBR Green-based quantitative PCR (qPCR) method as the final readout. These modifications allow for an increased throughput in vitro experiment that better correlates with PC 1/2 , allowing for improved segregation of resistant and sensitive parasites, as well as improved sorting of moderately resistant parasites.
Further, efficiency improvements in the eRRSA allow for a higher throughput in vitro testing of ART resistance, accelerating our understanding of artemisinin resistance in the laboratory and providing a more accurate method to track the spread of resistance.

Parasite isolates
To evaluate the eRRSA methods, P. falciparum isolates with varying kelch13 mutations and PC 1/2 were chosen. These isolates were derived from cloning by limiting dilution from patient samples. A total of the most common kelch13 mutation found in Southeast Asia currently), and a PC 1/2 distribution between 1.67 and 9.24. All 15 parasite isolates were isolated from patients on the Thailand-Myanmar border between 2008 and 2012. 3D7 was used as a control for comparison to the 15 Southeast Asian isolates (Table 1) [18,19].

Parasite culture
Plasmodium falciparum isolates were cultured using standard methods in human red blood cells (RBC) Technologies.), 10 mg/L gentamicin (Gibco, Life Technologies) and 0.225% NaHCO 3 (Corning, VWR) at 5% haematocrit. Cultures were grown separately in sealed flasks at 37˚C under an atmosphere of 5%

Percoll synchronization
Parasites were synchronized using a density gradient method as previously described with slight modifications [14,15,20]. Briefly, 350 μl of packed, infected erythrocytes at high schizogony (>50% schizonts) was suspended in 2 ml of RPMI. Cultures were layered over a single 70% Percoll (Sigma-Aldrich) layer in 1x RPMI and 13.3% sorbitol in phosphate buffer saline (PBS) and centrifuged (1561xg for 10 min, no brake). The top layer of infected late stage schizonts was then removed and washed with 10 ml of RPMI twice. Cultures were then suspended in 2 ml of CM at 2% haematocrit and placed in culture flasks on a shaker in a 37 °C incubator for 4 h to allow for re-invasion.

Flow cytometry
Four hours after Percoll synchronization (unless noted otherwise), samples were measured by flow cytometry as previously described with slight modifications to determine parasitaemia [14,16].
Briefly, 80 μl of culture and an RBC control incubated for at least 8h at 5% haematocrit in CM were stained with SYBR Green I (SYBR) and SYTO 61 (SYTO) and measured on a guava easyCyte HT (Luminex Co.). Analysis was performed with guavaSoft version 3.3 (Luminex Co.). 50,000 events were recorded for both the RBC control and samples to determine relative parasitaemias.

eRRSA setup
Two hours post-cytometric quantitation (or 6 h after Percoll synchronization) samples whose stage composition was >70% rings as determined by flow cytometry were diluted to 2% haematocrit and 0.5% parasitaemia (unless otherwise noted), and 200 μl of culture was aliquoted into 6 wells of a flat bottom 96-well plate. Each treated and untreated sample had three technical replicates: RBC controls were aliquoted into 2 wells at 2% haematocrit and 200 μl. Three wells of parasites and 1 well of the RBC control were treated with 700 nM dihydroartemisinin (DHA) (Sigma-Aldrich); an additional 3 wells of parasites and 1 well of RBC control was treated with 0.02% dimethyl sulfoxide (DMSO) (ThermoFisher) as untreated controls. Parasites were incubated for 6 h, and then washed three times with 150 μl of RPMI to remove drug. Samples were then suspended in CM and placed back in the incubator. Sixty-six hours after drug removal, 20 μl of sample from each well was collected and frozen for qPCR amplification (72 h sample) without any media changes. Plates were then placed back in the incubator for another 48 h, after which 20 μl of sample was again collected and frozen for qPCR amplification (120 h sample).

qPCR Amplification
Ring-stage samples were quantified at 72 and 120 h post-drug treatment. qPCR was performed using the Phusion Blood Direct PCR kit (ThermoFisher, cat # F547L), supplemented with 1x SYBR. Three microlitres of a 1:3 culture dilution was used in a 10 μl reaction and amplified using forward and reverse primers of the pfcrt gene. PCR amplification was measured using the fast mode of the ABI 7900HT, with a 20 s denaturation at 95 °C, followed by 30 cycles of 95 °C for 1 s, 62.3 °C for 30 s, and 65 °C for 15 s (Additional file 1). Cycle threshold (Ct) values were calculated using the ABI SDS 2.4.1. Fold change (2 ΔCt ) was calculated by determining the average ΔCt for the three technical replicates for the untreated and treated samples by applying the following equation: (see Equation 1 in the Supplementary Files) All statistics were performed and figures generated using GraphPad Prism version 8.2.1.

Results
Using a SYBR Green-based quantitative PCR method to quantify the fold-change between treated and untreated samples Percent proliferation is the standard RSA measurement to determine whether a parasite is artemisinin resistant or sensitive. This is calculated by dividing the percent parasitaemia in the treated (DHA) sample over the percent parasitaemia in the untreated (DMSO) sample. Parasitaemia is determined either by counting the number of viable and nonviable parasites using blood smears and microscopy or by flow cytometry. Determining parasitaemia with microscopy is cost effective and convenient but is also highly variable and time-consuming. Flow cytometry is typically very accurate; however, the process must begin upon reaching a timepoint, adding a substantial time investment at the point of sampling. To find another measurement of parasitaemia that could be automated and have less variability, qPCR on parasite genomic DNA was tested. A standard curve of percent parasitaemia (as measured by flow cytometry) shows excellent inverse correlation with Ct values as measured by qPCR ( Fig. 1). To quantify the difference between treated and untreated final RSA samples, fold change (2 ΔCt ) was calculated according to equation 1.

RSA readout at 120 h provides superior differentiation between sensitive and resistant isolates
The standard RSA determines parasite viability at 72 h after drug treatment. However, a common problem in the final readout (using either microscopy, flow cytometry, or qPCR) is the difficulty in differentiating between pyknotic (nonviable) parasites and viable parasites 72 h after drug treatment.
It is also difficult to measure viable parasitaemia when it can be as low as 0.01% (or even 0% in some cases), especially when measuring ART sensitive parasites [21]. To address these issues, the time to readout was extended by an additional intraerythrocytic development cycle (48 h); previous methods have extended the time to readout by 24 h in specific cases [13,22]. This extension was added to both allow parasites an additional expansion cycle, creating larger differences to distinguish resistant and sensitive isolates, and to allow erythrocytes to clear pyknotic parasites. This additional cycle provides a much greater separation between resistant and sensitive parasite isolates (Fig. 2).

Assay setup conditions have a substantial impact on RSA outcomes
The RSA is a growth assay targeted at a very narrow window of the parasite intraerythrocytic development cycle. In order to maximize the precision of the assay, the growth and the timing of the target window were carefully optimized. First, as it has been established that growth rates can vary based on parasitaemia, the effect of varying starting parasitaemias on RSA outcome was observed [23]. RSA was performed on three parasite isolates (3D7, 1337, and 4673) at varying starting parasitaemias as determined by a microscopist, and samples were collected at 72 h and 120 h (Additional file 2A and 2B, respectively). For the three parasite isolates tested, 3D7 was the ART sensitive control and 1337 and 4673 were two ART resistant parasite isolates as determined by their PC 1/2 values (1337 PC 1/2 = 7.84 and 4673 PC 1/2 = 5.34). The 0.25% starting parasitaemia showed the most distinguishable phenotype between the ART sensitive 3D7 control and the two ART resistant isolates at 120 h (Additional file 2B). It was noted internally that determinations of parasitaemia by microscopy varied widely and underestimated parasitaemia compared to flow cytometry, likely due to the difficulty of correctly identifying new invasions. A comparison of RSA results from isolates set up at 0.25% parasitaemia determined by microscopy and 0.5% parasitaemia determined by flow cytometry showed no difference between the two (Additional file 3). As a result, in subsequent setups starting parasitaemia was determined by flow cytometry and was normalized to 0.5% parasitaemia.
A key factor in the RSA is applying the drug treatment in the tight 0-3 h window of the parasite life cycle that can differentiate ART resistant parasites from ART sensitive parasites. Therefore, the time from Percoll synchronization to drug treatment was also varied to find the optimal time for drug

eRRSA correlates better with PC 1/2 than RSA
The RSA was introduced as an in vitro method that better captures the gold standard for in vivo ART resistance, PC 1/2 , than the traditional IC 50 . For a new assay to be relevant, it should perform at least comparably to the existing standard. Therefore, the eRRSA was used to assay 15 isolates from Southeast Asia with varying known PC 1/2 and kelch13 mutations collected between 2008 and 2012 (Table 1). NHP1337 was used as a resistant parasite isolate control and 3D7 was used as a sensitive parasite control. Three biological replicates were collected (each with three technical replicates) for each isolate and collected samples at 72 h and 120 h post-drug treatment and compared the viability of treated and untreated parasites at each stage.
The fold change data was then compared to PC 1/2 : at the 72 h timepoint across 15 isolates, the eRRSA has a Spearman correlation coefficient of -0.6071 (Fig. 3A). This is comparable to other RSA correlations in the field, showing that the improvements made to the RSA protocol do not significantly affect the outcome while increasing efficiency and ease of the assay [8,17]. The 120 h correlations, however, improved Spearman correlation between fold change and PC 1/2 to -0.8393 (Fig. 3B).
Traditional RSA uses a value of ≥ 1% parasite survival to indicate ART resistance [8]. To determine a value for ART resistance as measured by the eRRSA, the best-fit line for the 120 h post-drug treatment (Fig. 3B) was used to calculate the fold change when PC 1/2 is 5 (given that PC 1/2 ≥ 5 h defines clinical ART resistance [7]). Therefore, a fold change of 30 or less defines an ART resistant parasite by eRRSA.

Discussion
With the gradual spreading of artemisinin resistance throughout Southeast Asia, it is imperative that resistance can be accurately measured both in the field and in the lab. To date, the RSA has been the golden standard for in vitro measurement of artemisinin resistance. In the clinic, PC 1/2 is the standard for in vivo measurement of artemisinin resistance. The PC 1/2 comprises contributions by both the human host and the parasite. Because the RSA is an in vitro measurement of only the parasite component, the RSA cannot perfectly correlate with PC 1/2 [24-27]. Despite this, there is a need for a more accurate, higher throughput in vitro measurement alternative to the current RSA to accelerate understanding of artemisinin resistance. The eRRSA was developed for this purpose and the study shows that it can outperform the RSA in both accuracy and efficiency.
A major bottleneck and source of variability in the RSA is the final readout to determine the ratio of viable parasites in the treated and untreated cultures. RSA uses microscopy or flow cytometry as the final readout, while the eRRSA uses qPCR. When using microscopy to compare viable and nonviable parasites, the presence of the ring-like structure of healthy, viable parasites are compared to the collapsed, nonviable parasites which can be highly subjective and time consuming, as shown by the original RSA paper which required two to three microscopists [14]. Flow cytometry utilizes the DNA stain SYBR and the mitochondrial stain MitoTracker red to differentiate between viable and nonviable parasites. This eliminates the requirement of microscopists and drastically decreases the labor required to determine percent proliferation of parasites [14,16]. However, staining of cells for flow cytometry is time sensitive and must be done immediately following the end of the RSA (at 72 h), which can limit flexibility and lengthens an assay that already demands long hours. qPCR measure viable parasites solely by concentration of genomic material, comparing the efficiency of parasites to proliferate post-artemisinin perturbation. Here, the efficacy of qPCR is demonstrated as a readout for proliferation in a survival assay context. The use of qPCR allows for both smaller sample sizes and a delayed readout, rendering the protocol easier and more precise. The demonstration of the effect of various setup conditions on the final outcome required an optimization of these parameters in the eRRSA. With no sorbitol synchronization required and only one single layer Percoll synchronization to select schizonts, parasites are synchronized easier, quicker, and closer to the ring-stage so that they can be set up in the assay sooner to avoid losing their synchronization. The variability in assay outcome caused by varying the delay between synchronization and treatment is likely due to the short window (0-3 h) that differentiates ART resistant parasites from ART sensitive parasites. The eRRSA assay is set up 6 h post-Percoll, which allows for a high percentage of early ring-stage, tightly synchronized parasites. Using flow cytometry for parasitaemia measurements post-Percoll synchronization allows for rapid and accurate parasitaemia determination and staging for many parasites at once at the setup of the assay, which is an essential factor for the results of both RSA and eRRSA. Starting parasitaemia has a substantial effect on growth throughout the assay; by using a lower starting parasitaemia (0.5%) compared to the in vitro RSA, the eRRSA has a lower volume and parasitaemia requirement while also allowing for more precise measurement of growth with and without drug treatment. Lower culture volume requirements permit the use of 96-well plates, which uses less reagents, time, and space.
The RSA is a multi-step and time-consuming assay. Even with the optimizations of the eRRSA, the four hours required to wait until schizonts have reinvaded after single layer Percoll synchronization to obtain 0-3 h ring-stage parasites and the six hours of drug incubation time cannot be avoided.
However, by automating aspects of the assay that have a certain run time regardless of how many samples are being assayed (e.g., flow cytometry in a 96-well plate to set-up parasites and qPCR as the final readout), running larger numbers of samples in parallel does not add a significant amount of time to the assay. In addition to automation, decreasing the culture volumes (by using a 96-well plate) and using a quick and tight synchronization method (single layer Percoll) makes it possible for one researcher to set-up 12-15 parasites samples with technical replicates in a ~12 h day (Additional file 4).
Finally, it was demonstrated that the eRRSA shows superior correlation with the clinical phenotype PC 1/2 . As ART resistance is currently presenting as a continuous phenotype, the ability to accurately determine intermediate phenotypes is critical in understanding ART resistance and identifying contributing genotypes beyond kelch13 propeller mutations. The efficiency of the eRRSA makes it an excellent replacement for the traditional RSA in any study of ART resistance requiring accuracy and higher throughput. The 15 cloned parasites examined include three which lack kelch13 mutations but show PC 1/2 >5. These include one clone (NHP4373) with PC 1/2 = 7.1. Interestingly, these clones also show high eRRSA values, confirming their ART resistant status. These results provide further support that ART resistance may result from mutations elsewhere in the parasite genome, or perhaps from non-coding regulatory changes controlling kelch13 activity [17].

Conclusions
The eRRSA method described here provides a more robust in vitro representation of PC 1/2 while also providing vastly improved throughput. Widespread adaptation of the eRRSA should significantly accelerate our understanding of artemisinin resistance, allowing for both high throughput surveillance of the spread of resistance and for the precise phenotyping necessary to uncover complex genetic contributors to resistance.    Southeast Asia with varying PC1/2 were assayed using the eRRSA. 72 h post-drug treatment samples were measured to give a fold change for each isolate and those fold changes were correlated with each isolate's PC1/2 (Spearman r =-0.6071). Isolates with red boxes are kelch13 mutants and the 95% confidence interval around the best-fit line (y=-6.025x+61.14) is denoted with dotted lines. (B) The same 15 isolates were assayed using the eRRSA and 120 h post-drug treatment samples were measured to calculate a fold change for each isolate. The fold changes were correlated with PC1/2 (isolates marked with red boxes are kelch13 mutants and the 95% confidence interval of the best-fit line is denoted with dotted lines) (Spearman r = -0.8393); the 120 h eRRSA samples showed an increased correlation with PC1/2 compared to the 72 h samples.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.