Artemisinin-resistant malaria parasites show enhanced transmission to mosquitoes under drug pressure

Resistance to artemisinin combination therapy (ACT) in the Plasmodium falciparum parasite is threatening to reverse recent gains in reducing global deaths from malaria. Whilst resistance manifests as delayed asexual parasite clearance in patients following ACT treatment, the phenotype can only spread geographically via the sexual cycle and subsequent transmission through the mosquito. Artemisinin and its derivatives (such as dihydroartemisinin, DHA) as well as killing the asexual parasite form are known to sterilize male, sexual-stage gametes from activation. Whether resistant parasites overcome this artemisinin-dependent sterilizing effect has not, however, been fully tested. Here, we analysed five P. falciparum clinical isolates from the Greater Mekong Subregion, each of which demonstrated delayed clinical clearance and carried known resistance-associated polymorphisms in the Kelch13 gene (PfK13var). As well as demonstrating reduced sensitivity to artemisinin-derivates in in vitro asexual growth assays, certain PfK13var isolates also demonstrated a marked reduction in sensitivity to these drugs in an in vitro male gamete activation assay compared to a sensitive control. Importantly, the same reduction in sensitivity to DHA was observed when the most resistant isolate was assayed by standard membrane feeding assays using Anopheles stephensi mosquitoes. These results indicate that ACT use can favour resistant over sensitive parasite transmission. A selective advantage for resistant parasite transmission could also favour acquisition of further polymorphisms, such as mosquito host-specificity or antimalarial partner–drug resistance in mixed infections. Favoured transmission of resistance under ACT coverage could have profound implications for the spread of multidrug resistant malaria beyond Southeast Asia. ONE SENTENCE SUMMARY Artemisinin-resistant clinical isolates can also demonstrate resistance to the transmission-blocking effects of artemisinin-based drugs, favouring resistance transmission to the mosquito.


INTRODUCTION 52
Malaria kills more than 400,000 people each year (1). Whilst there has been a marked reduction in 53 global rates of malaria disease since the new millennium, progress has stalled recently even 54 reversing in some regions (1). A critical factor threatening future gains is the emergence and 55 spread of drug resistance in the most virulent parasite Plasmodium falciparum (2). Of most 56 concern is the reported spread of resistance to frontline artemisinin-based drugs in the Greater 57 Mekong Subregion (GMS) of Southeast Asia (3, 4). Artemisinin has revolutionised treatment for 58 severe malaria. The drug acts rapidly to clear the clinical symptoms of malaria by killing the 59 asexual parasite in host red blood cells. Although a precise mechanism of action is contested, it is 60 thought that iron-mediated activation of artemisinin arising from parasite metabolism of 61 haemoglobin causes the drug to be both highly reactive and consumed rapidly in the process of its 62 action(5). Consequently, use of artemisinin or its derivatives requires coformulation with longer-63 lasting partner drugs as artemisinin-based combination therapies (ACTs). In recent years, 64 however, resistance to both artemisinin and partner drugs, including piperaquine and mefloquine, 65 has increased in prevalence throughout Southeast Asia (4, 6-8). The spread of such multidrug 66 resistant parasites beyond the GMS region could prove catastrophic for global malaria control. 67 68 Resistance to artemisinin is strongly associated with non-synonymous single nucleotide 69 polymorphisms (SNPs) in the propeller domain of P. falciparum Kelch 13 (PfK13) (9) a protein with 70 multiple likely functions in the parasite cell (5). Based on the SNP analysis, several PfK13 variants 71 (PfK13 var ) have been defined displaying different degrees of delayed parasite clearance in patients 72 under ACT treatment. PfK13 variants include mutually exclusive SNPs giving rise to amino acid 73 changes C580Y, R539T, I543T and Y493H (4, 7, 10, 11). Whilst the precise mechanism by which 74 PfK13 var determines resistance remains ill-defined (5), PfK13 var parasites show an upregulation in 75 the unfolded protein cell stress response (12). Given the importance of this pathway to general cell 76 viability, PfK13 var parasites may be better able to deal with stresses arising from drug damage on 77 cell function (12). Persistence of parasites in the blood of infected individuals will lead to their 78 delayed clearance and ultimately treatment failure. Among PfK13 polymorphisms, the PfK13 C580Y 79 genotype is the most widely spread variant currently circulating in eastern Southeast Asia (7). 80 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint APS3G and APL5G both had copy number variants of mdr1 ( Table 2). This suggests that APS3G 137 and APL5G will likely show resistance to mefloquine, whilst APL4G will likely show resistance to 138 piperaquine (21,24). 139 140 Variation in PfK13 results in a growth defect in asexual blood stages but not in mosquito 141 stages 142 Polymorphisms in the PfKelch13 gene (PfK13 var ) that are associated with artemisinin resistance 143 are known to also show reduced asexual blood stage growth (25,26). To validate this in selected 144 isolates, parasites were set up in synchronised ring stage cultures, at a starting parasitaemia of 145 2%, and followed over the course of eight days. Parasitaemia was analysed every second day by 146 flow cytometry and cultures re-diluted to 2%. NF54 parasites showed a cumulative parasitaemia as 147 expected under standard laboratory conditions ( Figure 1A). Parasites with a PfK13 var showed a 148 significantly reduced replication rate in in vitro compared to NF54 ( Figure 1A) agreeing with 149 previous studies (25,26). To explore the underlying mechanism of slowed growth, we measured 150 the number of merozoites per schizont, which will directly determine potential growth rates (27). 151 Late synchronised schizonts were blocked from merozoite egress using the protein-kinase G 152 (PKG) inhibitor, compound 2(28). Thin smears, 12 hours later, were then made of each culture and 153 stained with the nuclear stain 4′,6-diamidino-2-phenylindole (DAPI) to count nuclei per schizont. 154 PfK13 var isolates displayed fewer nuclei per schizont than the PfK13 WT isolate and NF54 control, 155 suggesting that the observed reduced growth rate may at least be partly explained by a reduction 156 in the number of progeny ( Figure 1B). 157

158
To investigate the transmission capability of each P. falciparum field isolate, we induced 159 gametocytes at a starting parasitaemia of 2% (29) and, 14 days post induction, fed cultures to 160 Anopheles stephensi mosquitoes by standard membrane feeding assay (SMFA)(30). No significant 161 differences were noted in the stage V (mature) gametocytaemia for isolates (in terms of relative 162 numbers of gametocytes to asexual parasites). However, upon activation, male exflagellation rates 163 were reduced in PfK13 var isolates compared with PfK13 WT (Figure S2). Ten days post-feeding, 164 mosquito midguts were dissected, and oocysts numbers recorded. All field isolates were found to 165 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint be capable of infecting mosquito midguts at varied intensity levels, i.e. oocyst counts per midgut, 166 as reported previously for Cambodian field isolates (31). To test if PfK13 var led to a reduced 167 replication rate in mosquito stage growth (following the reduced merozoite count), we measured 168 the diameter of each oocyst in these infections as a proxy for replication. Oocyst size showed no 169 consistent pattern of variation when compared to controls other than a PfK13 R539T variant which 170 displayed significantly larger oocysts than the PfK13 WT control ( Figure S2). This shows that whilst 171 variations in PfK13 may reduce parasite multiplication rate in the asexual blood stage it does not 172 appear to directly influence transmission and growth in the mosquito-stage. It has previously been shown that artemisinin and its derivatives have an inhibitory effect on male 177 gamete exflagellation, irreversibly sterilizing male gametocytes from activation (15, 16). To explore 178 whether PfK13 var isolates were resistant to this sterilizing effect we tested mature gametocyte 179 culture capacity to activate in the dual gamete formation assay (PfDGFA) (15, 16). 24-hour 180 incubation of cultures with the artemisinin derivative dihydroartemisinin (DHA) was found to be 181 insufficient to elicit a complete inhibition of exflagellation for PfK13 WT NF54 parasites (Figure S3), 182 likely as a result of the rapid instability of the drug (32). To improve activity and allow for a 183 comparative analysis between isolates, gametocytes were exposed to a second compound dose 184 24 hours after the first, resulting in a double-dose regimen with a readout after 48 hours. Double 185 exposure consistently gave complete inhibition of male activation with DHA at the highest 186 concentration tested ( Figure S3). Female gamete activation was unaffected as found previously 187 (15). In parallel, two other artemisinin derivatives and four other antimalarial drugs were tested by 188 PfDGFA (Figure S4 and Figure S5). Exflagellation rates for PfK13 var isolates varied in the 189 presence of drug, with two PfK13 R539T and PfK13 C580Y isolates consistently showing tolerance to 190 artemisinin-derivatives (Figure 2A) Figure S2) and also 203 showed a high level of male gamete activation resistance to DHA (Figure 2). Gametocyte cultures 204 of both parasites were exposed to different concentrations of DHA for 48 hours using our double-205 dosing regimen, before feeding to An. stephensi mosquitoes by SMFA. At day 10 post-feed, 206 mosquitos were dissected, and midguts examined for oocyst load ( Figure S6). Generalised linear 207 mixed effects models were used to analyse infection intensity (number of oocysts per midgut) and 208 infection prevalence (proportion of midguts with oocysts) in response to treatment with DHA, in 209 order to incorporate data from 18 individual SMFA experiments within the same modelling 210 framework. A decrease in both intensity and prevalence of infected mosquitoes was observed for 211 both parasite isolates with increasing DHA concentration ( Figure 3A and Figure 3B). A significant 212 decrease in both the oocyst intensity (ratio of oocyst intensity = 0.73, 95% CI: 0.66-0.80) and 213 prevalence (odds ratio = 0.46, 95% CI: 0.41-0.52) of mosquito infection was observed for NF54 214 with increasing drug concentration. In contrast, APL5G parasites (C580Y) showed no evidence for 215 a significant decrease in oocyst intensity with increasing DHA (ratio of oocyst intensity = 0.84, 95% 216 CI: 0.65-1.07). Increasing concentrations of DHA did still reduce the oocyst prevalence for APL5G 217 (odds ratio = 0.72, 95% CI: 0.56-0.96), however, this effect was significantly less than the effect 218 seen for WT parasites ( Table 3). The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint dose for the ACT, DHA-piperaquine (33-35). In the absence of DHA, NF54 parasites were 224 consistently observed to have an increased infection prevalence (55.1%, 95% CI: 50.5% -57.8%) 225 compared to APL5G parasites (19.2%, 95% CI: 13.3% -26.0%) i.e. all things considered, NF54 226 transmits better in the absence of drug ( Figure 3C). This suggests that a fitness cost is associated 227 with the APL5G genotype, which causes a sizeable reduction in the onward probability of infection 228 relative to WT parasites in the absence of DHA. However, this changes sigifnificantly in the 229 presence of drug. With 2µM DHA, no significant difference was observed between NF54 parasites 230 (20.6%, 95% CI: 27.7% -14.3%) and APL5G parasites (12.1%, 95% CI: 4.1% -25.4%) (  However, the transmission-resistance phenotype is nonetheless robust for certain isolates. 251

252
As with previous studies, we first started with investigation of the asexual growth rate, confirming a 253 consistent reduced rate in resistant isolates. The asexual growth rate reduction seen in PfK13 var 254 isolates likely acts as both a selective cost for parasite growth (in being out competed in normal 255 infections) but also likely explains how these parasites persist during drug treatment, i.e. explaining 256 delayed clearance(5). Switching our focus to sexual commitment and development, we next 257 explored gametocyte production. With the caveat that different parasite isolates always show 258 marked differences in gametocyte formation capacity, we did not observe any obvious reduced 259 capacity among PfK13 var isolates in gametocyte production. All five field isolates produced 260 equivalent numbers of mature gametocytes (stage V gametocytes) after day 14 upon induction. 261 Indeed, the reverse correlation between drug resistance and sexual commitment has been 262 consistently reported. Clinical isolates with demonstrated drug resistance and delayed clearance 263 have been consistently reported to produce a higher gametocytaemia, suggesting a potentially 264 elevated potential for transmission(10). 265 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint Having confirmed capacity to form gametocytes, we assessed the capacity to generate 266 exflagellation centres, a marker of male gamete activation capacity in the mosquito. We found no 267 direct correlation between gametocytaemia and exflagellation count. This lack of correlation may 268 be due to differences in the sex ratio between parasite isolates (reduced males mean less 269 exflagellation centres though gametocytaemia may be the same). Whilst commitment of 270 gametocytes to either male or female is poorly understood, it is entirely conceivable that 271 gametocytes mature or differentiate into either male or female at differing rates in each isolate. 272 Unfortunately, sex ratios were untested here due to a paucity of markers for male and female 273 gametocytes and challenges with definitive differentiation of the sexes using Giemsa stain. 274 Irrespective, gametocyte conversion rates have been shown to be sensitive to asexual stage 275 replication, which itself is affected by drugs. This suggests that there is the potential for a trade-off 276 between asexual stages and sexual stages in ensuring the spread of the artemisinin resistant 277 parasites(36). 278

279
With sexual commitment and exflagellation in vitro seemingly uncompromised in resistant isolates 280 we next sought to explore transmissibility directly. We saw no obvious defect in either the 281 transmission capacity (number of oocysts) or the transmission replication rate (as measured by 282 oocyst size) among the five parasite field isolates and NF54. This latter point is noteworthy since it 283 is clear that artemisinin-resistant parasite isolates show a lower asexual growth rate and merozoite 284 (progeny) rate (Figure 1), however, number of oocysts and sporogony in the mosquito doesn't 285 appear to be affected. Thus, PfK13 var parasites appear able to commit to sexual reproduction, 286 activate and transmit to mosquitoes at levels that don't differ dramatically to those commonly seen 287 in sensitive parasites (i.e. beyond variability usually seen between isolates). 288 289 Shifting our attention to transmission under drug coverage, tests of the viability of gametocytes for 290 gamete activation using the dual gamete formation assay (PfDGFA) with artemisinin derivatives; 291 DHA, Artemether and Artemisone clearly found that certain PfK13 C580Y and PfK13 R539T parasites 292 demonstrated significantly higher resistance compared to sensitive controls. Of note, whilst 293 undertaking this work, a parallel study made similar observations. Testing male exflagellation 294 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint sensitivity to DHA in unrelated culture-adapted PfK13 var Cambodian field isolates, Lozano et al 295 found that PfK13 var isolates showed a reduced sensitivity of exflagellation rates to DHA treatment, 296 though onward mosquito infectivity was not tested(37). Extending this observation to transmission 297 directly we took the most competent transmissible field isolate representing the resistant 298 phenotypes (APL5G, C580Y) compared to the laboratory reference strain NF54, and tested 299 whether transmission resistance plays out in terms of capacity to infect mosquitoes over a range of 300 drug concentrations. Controlling for gametocytaemia, number of cells and haematocrit level for 301 each, infection prevalence in mosquitoes could then be tested and compared between lines. Of 302 note, the infection intensity (number of oocysts found in each mosquito) was consistently different 303 between NF54 and APL5G, as it is for each different culture-adapted parasite strain (see (31)). 304 These differences make direct measures of mixed infections challenging. Nonetheless, we found 305 that artemisinin resistant, PfK13 C580Y (APL5G) was consistently more likely to transmit malaria 306 under drug pressure compared to its DMSO treated controls than the control NF54 parasite 307 ( Figure 3C). This was due to the greater impact of DHA on oocyst infection exhibited by the wild-308 type isolate, which served to offset the decreased transmission potential for APL5G in the absence 309 of artemisinin. This demonstrates that the artemisinin resistant phenotype of APL5G impacts both        For the growth assay, asexual parasites were sorbitol-synchronised at least twice 16 hours apart to 572 create an 8-hour growth window. Starting parasitaemia was seeded at 1-2% early ring stages in 573 triplicates that were treated separately. The assay was performed twice using 3 replicates each. 574 Every other day, parasites were fixed in 4% formaldehyde and 0.2% Glutaraldehyde for at least 10 575 minutes. After washing with PBS, and DNA was stained with SybrGreen1 (diluted 1:10'000) in the 576 dark for 20 minutes at room temperature. After incubation, cells were washed three times with PBS 577 and resuspended in 80ul PBS. Flow cytometry was performed counting a total of 100'000 cells per 578 condition. 579 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint

Nuclei count 581
Parasites were synchronised twice using 5% Sorbitol to obtain a 10-hour life cycle window. 10 µM 582 compound 2 was added to late trophozoite stages for a maximum of 12 hours to block egress of 583 the red blood cells (RBC) (28). Resulting segmented schizonts were thinly smeared, fixed with 4% 584 Formaldehyde and 0.2% Glutaraldehyde for 20 minutes. Smears were then stained with 1 µg/ml 585 DAPI for 5 minutes and mounted in Vectashield (Vector Laboratories). Z-stacks were taken using a 586 Leica microscope at 100x magnification. Nuclei of arrested segmented schizonts were counted 587 using the plugin tool "Manual counting" on ICY (40). Only singly-invaded RBC were counted. 588 589

Trophozoite maturation assay (TMI) 590
The trophozoite maturation assay was performed according to (17). Briefly, P. falciparum infected 591 blood was collected into heparin tubes and centrifuged at 800 g at 4°C for 5 minutes to allow the 592 removal of the plasma and buffy coat. This was followed by three washes in RPMI 1640 (without 593 serum supplement) and adjusted to 3% cell suspension in 10% A+ human serum supplemented 594

Statistical Modelling of Oocyst infection Intensity and Prevalence 664
To assess the impact of artemisinin on the ability of each parasite line to form oocysts, we used 665 generalised linear mixed effects models in order to incorporate data from different experimental 666 replicates within the same modelling framework. These models have previously been used to 667 model transmission blocking interventions (43). We modelled either oocyst intensity or prevalence 668 as the response with treatment (DHA concentration) included as a fixed effect and 0 µM DHA 669 represented by control groups treated with DMSO. The parasite line treated (PfK13 WT or 670 PfK13 C580Y ) was included as a fixed effect to assess the differential impact of artemisinin on 671 transmission success. The impact of treatment between experimental replicates was allowed to 672 vary at random between replicates. A logistic regression (binomial error structure) was used to 673 model the prevalence of mosquito infection, i.e. the presence or absence of oocysts, and a zero-674 inflated negative binomial distribution was used to model the intensity of infections, i.e. the 675 numbers of mosquito oocysts (44). 95% confidence interval estimates were generated for the 676 impact of drug concentration by bootstrapping methodology (with 100,000 replicates). 677 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint stephensi mosquitoes were infected with field isolates, and oocyst numbers and prevalence are 690 shown. C. The size of day 10 oocysts was measured in (B). Graph shows diameter of each oocyst 691 found. PfK13 genotype is not related to oocyst size. Isolate APS2G shows significantly bigger 692 oocysts than NF54 (unpaired t-test, ** p<0.01). 693 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.04.933572 doi: bioRxiv preprint