Plasmodium falciparum sexual differentiation in malaria patients is associated with host factors and GDV1-dependent genes

Plasmodium sexual differentiation is required for malaria transmission, yet much remains unknown about its regulation. Here, we quantify early gametocyte-committed ring (gc-ring) stage, P. falciparum parasites in 260 uncomplicated malaria patient blood samples 10 days before maturation to transmissible stage V gametocytes using a gametocyte conversion assay (GCA). Seventy six percent of the samples have gc-rings, but the ratio of gametocyte to asexual-committed rings (GCR) varies widely (0–78%). GCR correlates positively with parasitemia and is negatively influenced by fever, not hematocrit, age or leukocyte counts. Higher expression levels of GDV1-dependent genes, ap2-g, msrp1 and gexp5, as well as a gdv1 allele encoding H217 are associated with high GCR, while high plasma lysophosphatidylcholine levels are associated with low GCR in the second study year. The results provide a view of sexual differentiation in the field and suggest key regulatory roles for clinical factors and gdv1 in gametocytogenesis in vivo.

C ontrolling malaria transmission is key to global malaria eradication efforts and continues to be highlighted as a critical research area by the recent Malaria Eradication (MalERA) refresh panel 1 . Human to human malaria transmission via a mosquito requires the production of sexual stage parasites, called gametocytes, which are produced during the erythrocytic cycle. After erythrocyte invasion by a merozoite 2 , the parasite replicates producing new merozoites that will either continue asexual multiplication or develop into gametocytes during the next erythrocytic cycle [3][4][5][6] . The regulation of this balance between asexual propagation and sexual differentiation is key to understanding the dynamics of malaria transmission and pathology in human hosts 7,8 . After a P. falciparum gametocyte-committed merozoite invades an erythrocyte, it differentiates over 10-12 days from a ring stage parasite through 5 distinct morphologic stages (I-V) into a single mature male or female gametocyte 9 . For malaria transmission a male and a female gametocyte must be taken up in a blood meal by a mosquito, fertilize and begin differentiation into sporozoites that can be spread to another human host 10 .
In the human host, immature stage I-IV P. falciparum gametocytes sequester and cannot be detected in peripheral blood samples, making it difficult to monitor sexual differentiation in the field until mature, transmissible stage V gametocytes are released into the circulation 8-10 days later [11][12][13][14] . This delay in the ability to monitor early gametocyte formation makes it challenging to define the factors that influence gametocytogenesis in patients before they become infectious 15,16 . Studies of stage V gametocyte prevalence in the field have clearly demonstrated wide variability between individuals 12 , but it is hard to know if this is due to gametocyte production or survival. An association between stage V gametocyte levels and low hematocrit 17,18 or a history of illness >2 days 15,19 has been reported, but whether these factors are causal is difficult to determine 20,21 .
In culture, both asexual and sexual differentiation are intrinsic parts of the life cycle with each successive erythrocytic cycle producing a mixture of schizonts with merozoites destined for either gametocyte production or asexual replication [22][23][24] suggesting an underlying stochastic or constitutive control mechanism. However, differences in sexual differentiation rates between P. falciparum strains can range from 0 to 20% 25,26 , suggesting parasite genetics play a role and variation within strains also indicates modulation by host factors such as lysophosphatidylcholine (LysoPC) levels 5,21,27 . Again, the underlying regulatory mechanisms remain unknown. Two genes have been reported to be required for P. falciparum gametocyte production in vitro, gametocyte development protein 1 (gdv1, PF3D7_0935400) 24 and the apetela-2 transcription factor (ap2-g, PF3D7_1222600) 25,28 . The activity of gdv1 was originally demonstrated by genetic complementation of a gametocytedeficient line (G def ) that had a spontaneous deletion in the gdv1 locus on chromosome 9 24 . Gdv1 deletions are frequently associated with the loss of gametocyte production in lab strains, while overexpression increases gametocyte production 24 . Field populations also exhibit high gdv1 allelic divergence that has been suggested to vary with transmission patterns 29 . Transcription factor ap2-g was also identified as the target of a mutation in a gametocyte-deficient parasite line in both P. falciparum 28 and P. berghei 30 . Notably, AP2-G protein has been shown to enhance its own transcription 25,28,30 as well as to be epigenetically repressed by heterochromatin protein 1 (HP1, PF3D7_1220900) in schizonts producing asexually committed merozoites 31,32 . Recently, episomal expression of GDV1-tagged with green fluorescent protein (GFP) and a ligand stabilized-degradation domain (DD) in a parasite line with an intact gdv1 locus has been shown to increase gametocyte production 33 . The GFP.DD-tagged GDV1 also associated with HP1 and localized to heterochromatin sites throughout the genome, including the ap2-g locus suggesting that GDV1 and HP1 interact to regulate gametocyte production.
In the work reported here, we developed a gametocyte conversion assay (GCA) to quantify gametocyte-committed ring stage parasites in human blood samples, which allows direct comparison of sexual differentiation rates with clinical characteristics and gene expression levels. Gametocytes were detected in 76% of the samples from uncomplicated malaria patients and there was a wide range of asexual-to gametocyte-committed ring ratios (GCR). GCR was positively associated with parasitemia and negatively influenced by temperature. RNA levels of ap2-g, AP2-G-dependent gene, msrp1 (PF3D7_1335000) 24,25 and AP2-G-independent gametocyte-specific gene, gexp5 (PF3D7_0936600) 34,35 were also higher in high GCR samples. To more carefully analyze the regulation of gametocytogenesis, we generated a parasite line that allowed tight control of gametocyte production by tagging the endogenous gdv1 gene with GFP and a ligand-dependent degradation domain (Pfgdv1.gfp.dd). In the absence of Shield 1 ligand (Shld1), Pfgdv1.gfp.dd gametocyte production was markedly decreased. Using this clonal Pfgdv1.gfp. dd line to perform GDV1 protein dosing experiments, we demonstrate that GDV1 acts as a 'molecular rheostat' during schizogony to control gametocyte production and gene expression. Together the data suggest that in humans there is a low basal level of GCR that is modulated by an interplay of parasitemia, body temperature, and P. falciparum gdv1.

Results
Gametocyte-committed ring quantification in patient samples. To directly assess gc-rings in patient blood samples, we developed an ex vivo GCA to screen blood from uncomplicated malaria patients (0.5-13 years old) presenting at Ewim Health Center in Cape Coast, Ghana during two malaria seasons, July-August 2016 and 2017 ( Fig. 1 and Supplementary Table 1). Briefly, one aliquot of a subject's blood sample is used to quantify white blood cells (WBC) and hematocrit, another is used for RNA and DNA isolation, while the remaining material is cultured ex vivo in the presence of n-acetyl glucosamine (NAG) to block further asexual replication (Fig. 1a). The cultures are followed by Giemsa-stained smears over the next 8 days for gametocyte development. Only 1 sample in 2016 and none in 2017 had microscopically detectable stage V gametocytes on a thick blood smear made on Day (D) 0 and by D4 stage V gametocytes were not detected in the ex vivo smear. Schizonts are also not observed in the NAGtreated ex vivo cultures indicating NAG had successfully blocked asexual development. Morphologically distinct stage II gametocytes are occasionally observed on D3 of ex vivo culture and became the predominant stage by D4. The parasites continue to progress to a mixture of III-V on D8, which is similar to the in vitro time course of standard lab-adapted cultures (Fig. 1b). Most of the gametocytes mature from stage II to III-V (Fig. 1c) with a median difference between the D4 stage II-III and the D8 stage III-V gametocytemia of 0 (interquartile range −0.02 to 0.008, Fig. 1d), which is comparable to that seen in standard in vitro culture. GCR is calculated by dividing the D4 stage II and III gametocytemia or the D8 stage III-V gametocytemia by the total ring stage parasitemia on Day 0 (D0P).
As an internal check of the GCR calculations we compared the D4 and D8 gametocyte conversion rates for each sample and found they are significantly positively correlated both by Pearson and regression analysis (R 2 = 0.73, p < 0.00001) providing confidence in the counting method used to identify gametocytes (Fig. 1e). The GCR distribution for both collection years is also similar with a total of 74% and 79% of the ex vivo samples in 2016 and 2017, respectively, containing detectable gametocytes. While 20% of the samples both years have GCR above 4%, the median GCR is only 0.6% in 2016 and 0.86% in 2017 consistent with high gametocyte formation in only a small number of individuals (Fig. 1c).
The D0 blood samples are also assessed for the presence of mature gametocytes using RT-qPCR to determine Pfs25 (PF3D7_1031000 [https://plasmodb.org/plasmo/app/record/ gene/PF3D7_1031000]) transcript levels. As with the GCR a majority of the population have low levels of circulating stage V gametocytes and there is a large range of Pfs25 transcript levels from undetectable to 1,205,521/μl blood which corresponds to 1102 stage V/μl ( Fig. 1f and g and Supplementary Fig. 1). In total 26% of the samples have greater than 2 gametocytes per µl, while only 4 samples have > 100 gametocytes per μl and the median is 0.638 stage V gametocytes per μl with an interquartile range of 0. 18-2.20. The sample with the highest Pfs25 RNA level is also the one with microscopically detectable gametocytes. However, high stage V gametocyte levels on enrollment in the study (D0) do not predict high GCR demonstrating the need to further evaluate this association longitudinally in asymptomatic patients that do not have to be treated.
Host/parasite factors modulating in vivo gametocytogenesis. The relationship between patient age, temperature, parasitemia, WBCs levels, and hematocrit on D0 and gametocyte production and conversion on D4 or D8 is evaluated first by Pearson correlation (Fig. 2a) to identify significant bivariate associations. Before analysis gametocytemia, GCR and D0P are log transformed to normalize the distribution as assessed by histogram and normal quantile plots 36 . As previously reported 37 , D0 parasitemia is positively correlated with temperature and WBCs levels and negatively correlated with age. Age is also negatively correlated with temperature and WBCs and positively correlated with hematocrit. This pattern is consistent with the age dependent development of partial immunity that reduces parasite growth resulting in lower D0P and a decrease in associated symptoms,   The results indicate a significant negative interaction between D0P and temperature on D4 as well as D8 GCR (D4GCR p = 0.016 and D8GCR p = 0.002) and also D4 and D8 gametocytemia (D4GC p = 0.004 and D8GC p = 0.009). The statistical analysis is summarized in Fig. 2b and the complete analysis is shown in Supplementary data (Supplementary Tables 2 & 3). The final model indicates that there is a positive correlation between D0P and GCR for patients with lower temperatures, but this association decreases as temperature increases. The negative influence of temperature on the association between D0P and GCR is demonstrated graphically by comparing the slopes of the trend lines for the lnD0P verse lnGCR plots for patients with temperatures < or ≥ the median temperature of 38.5°C ( Fig. 2c and d).
Metabolomic analysis of host plasma. Both asexual parasitemia and temperature have been reported to be associated with serum lysophosphatidylcholine (LysoPC) levels in malaria patients 41 and LysoPC levels have recently been implicated in gametocytogenesis in vitro 27 . To focus the initial analysis, we first compared individuals with high and low to undetectable GCR. Forty plasma samples (Supplementary Table 4) with a range of D0P (>0.35%), 20 (10 from each year) that have consistently high GCR (D4 GCR > 4.9%) and 20 (10 from each year) with low to undetectable GCR were selected for metabolomic analysis using the Biocrates AbsoluteIDQ® p180 kit and liquid chromatography tandem mass spectroscopy (LC-MS/MS). On first analysis there is no significant difference between the mean of the 13 LysoPC isoforms in the high and low GCR samples. However, there is a significant difference between the mean total LysoPC iso- Early gametocyte gene expression is associated with high GCR.
We then sought to evaluate potential in vivo molecular markers for early P. falciparum gametocytes in field samples using the same 40 samples used in the metabolomics study that included 20 with high and 20 with low GCR (Supplementary Table 4). In addition to gdv1 and ap2-g, we selected one ap2-g-dependent gene, msrp1 24 , and one ap2-g-independent early gametocyte gene, gexp5 35 to evaluate in the D0 blood samples using 18s rRNA as well as a ring stage-specific gene, skeleton-binding protein (sbp1, PF3D7_0501300 [https://plasmodb.org/plasmo/app/record/gene/ PF3D7_0501300]) as constitutive controls for parasitemia ( Fig. 4a & b). Ring-stage Pfgdv1.gfp.dd parasites grown in the absence of Shld1 to inhibit gametocyte production are used as the reference group to calculate relative abundance for the high and low GCR samples. Ap2-g, msrp1, gexp5 transcript levels are all significantly higher in the high GCR samples than the low GCR samples using either 18s rRNA or Sbp1 as the control, while there is no difference in gdv1 RNA levels in the two groups. Pfs25 transcript levels are also tested to evaluate stage V gametocyte levels in the high and low GCR samples and found not to differ (Fig. 4c). This work is the first demonstration that the in vivo expression levels of these gene correlate with ex vivo sexual differentiation, providing support for their use as biomarkers in the field and indicating that the changes in gene expression underlying gametocytogenesis are similar in vivo and in vitro.
Inducible regulation of gametocyte production. We next wanted to extend our analysis to the regulation of these genes during the transition from schizogony to gc-rings. However, due to sequestration of mature intraerythrocytic asexual parasites, schizonts are not accessible in human peripheral blood and therefore cannot be studied directly in patient samples. Given the similar  Fig. 3C) as well as gametocyte production in the presence and absence of Shld1 (Fig. 5c). As anticipated, gdv1.gfp.dd transcript levels are insensitive to Shld1, while Shld1 is required for GFP expression confirming successful integration into the gdv1 locus and Shld1-dependent GDV1.GFP. DD expression. The perinuclear pattern of GFP expression in schizonts is the same as that observed previously in parasites transformed with a plasmid directing episomal expression of GFPor HA epitope-tagged GDV1 24 demonstrating that protein localization is not affected by the addition of the DD at the Cterminus. GFP expression is first observed as DNA replication began,~36 h post invasion, which is consistent with peak gdv1 transcript levels (www.PlasmoDB.org 42 ) 43 ( Supplementary Fig. 4).
The GFP signal is observed by microscopy in all schizonts and remained high for 8 h before dissipating during merozoite segmentation and GFP expression is not observed in the newly formed ring stage parasites. Schizogony is monitored by flow cytometry using DNA stain SYTO 59, but the GDV1.GFP signal is not strong enough to be detected above the background green autofluorescence that also increased during schizogony. As expected Shld1 stabilization of GDV1.GFP.DD is also required for wild type levels of gametocyte production in both Pfgdv1.gfp. dd clonal lines (Fig. 5b & c). In the absence of Shld1 only an occasional gametocyte is observed, while exposure to Shld1 for only the last 12 h of the asexual cycle (36-48 h post RBC invasion) is sufficient to generate merozoites that are fully competent to invade red blood cells (RBCs) and differentiate over the next 10-12 days through the 5 morphologically distinct stages of gametocyte development (I-V) (Fig. 6).
GDV1 regulates ap2-g as well as ap2-g-independent genes. The ability to control gametocyte production in a clonal parasite line using Shld1 allowed direct analysis of the initial stages of sexual differentiation without contaminating subpopulations of older gametocytes present in wild type P. falciparum cultures. The expression of a range of stage-specific and constitutive transcripts were compared in MACS/sorbitol synchronized cultures that were grown in the presence or absence of Shld1 ( Fig. 7 and Supplementary Fig. 5). RNA obtained during schizogony [36 ± 2 h postinvasion (hpi)] and 24 h later, during the subsequent ring stage, was tested for gametocyte-associated transcripts, ap2-g, msrp1, gexp5, Pfs16 (PF3D7_0406200) 24,44 Pfge3 (PF3D7_1477700) 45 and ap2-g3 (PF3D7_1317200) 46,47 , which was reported in a piggyBac screen for gametocytogenesis related genes in P. falciparum and has recently been shown to be upstream of ap2-g expression in the rodent malaria P. yoelii. In addition, 18s rRNA (PF3D7_0725600) and two asexual ring specific genes, kahrp (PF3D7_0202000) 14 and sbp1 48 were tested as well as hp1-specific primers and conserved primers that amplify multiple var 31 genes which are regulated by hp1 (Supplementary Table 5).
In schizonts from Shld1-treated vs untreated Pfgdv1.gfp.dd.T1 cultures we found a significant increase only in mRNA corresponding to ap2-g, and msrp1 ( Fig. 7b & Supplementary  Fig. 5A), which is not observed in parental NF54 parasites ( Fig. 7d and Supplementary Fig. 5B). This Shld1-dependent increase is maintained in RNA harvested 24 h later from the ring stage Pfgdv1.gfp.dd.T1 parasites ( Fig. 7c and Supplementary  Fig. 5A), while there continued to be no Shld1-dependent difference in NF54 parasites ( Fig. 7e and Supplementary Fig. 5B). Transcripts for gexp5 as well as an additional early gametocyte gene, Pfs16, are also found to be significantly up-regulated in ring  Fig. 4 In vivo, high gametocyte conversion is associated with high ap2-g, msrp1 and gexp5 transcript levels. a, b The relative abundance (2 −ΔΔC T ) of the transcript levels of the indicated genes were determined for malaria patient samples in the high GCR (H, triangles, n = 20) and low GCR (L, circles, n = 20) cohorts using 18s rRNA (a) or sbp1 (b) RNA levels as the endogenous control. The reference sample was RNA from Pfgdv1.gfp.dd.T1 parasites grown in the absence of Shld1 to inhibit gametocyte production. c The relative abundance (2 −ΔΔC T ) of Pfs25 transcripts in the high GCR (H, triangles, n = 20) and low GCR (L, circles, n = 20) cohorts was determined using 18s rRNA or sbp1 RNA as the endogenous control and, again, the reference was RNA from Pfgdv1.gfp.dd.T1 parasites grown in the absence of Shld1 to inhibit gametocyte production. The experiment was performed in duplicate and the means and s.e.m. are shown. A Kruskal-Wallis test followed a Dunn's multiple comparison test was used to compare between high and low converters and probability is indicated, p > 0.05 (ns), p < 0.05 (*) and p ≤ 0.0001 (****) stage parasites from the Shld1-treated Pfgdv1.gfp.dd.T1 group. At this early time point Pfge3 RNA levels are not significantly elevated. As expected there is no Shld1-dependent change in transcript levels for 18s rRNA, kahrp, or sbp1, which have not been associated with early gametocytes, but are included as controls for parasite stage. Additionally, RNA levels for gdv1, and var genes are not affected by Shld1 treatment, indicating that ligand-mediated stabilization of GDV1 is not required for expression of these genes. Ap2-g3 and hp1 transcript levels are also not affected by Shld1 treatment during schizogony, but in the subsequent ring stage RNA levels are slightly lower for hp1 and slightly higher for ap2-g3.
The expression of ap2-g, msrp1, hp1 and ap-g3 in Pfgdv1.gfp.dd. T1 parasites was further investigated by isolating RNA from schizonts when gdv1 transcript levels are increasing (~1-3 nuclei, 36 ± 2 hpi) and then again 10 h later during late schizogony when ap2-g levels increase and during the subsequent ring stage another 14 h later. Consistent with the previous experiment, in the presence of Shld1 transcript levels for ap2-g and msrp1 levels increase while in rings hp1 transcripts decrease and ap2-g3 transcripts increase slightly (Fig. 8a-d). Notably, even in the absence of Shld1 and in a gametocyte-deficient line (3D7G def ) that lacks the gdv1 locus 24 ap2-g and msrp1 transcript levels increase significantly during schizogony suggesting that GDV1 is not required to initiate transcription ( Fig. 8a-b) Next we performed a GDV1 protein-dosing experiment by varying Shld1 concentrations (0-1500 nM) to evaluate the role of GDV1 protein levels on ap2-g, msrp1 and gexp5 expression levels. An anti-GFP monoclonal antibody (mAb) is used to assess GDV1 levels in Pfgdv1.gfp.dd schizonts following 14 h treatment with different Shld1 concentrations and demonstrates a dose dependent increase (Fig. 9a & b). GDV1.GFP.DD migrates as two bands and can only be extracted by adding 8 M Urea and 5% SDS, indicating it tightly associates with an insoluble complex, possibly the nuclear envelop. The estimated molecular weight of the anti-GFP positive upper bands from 5 independent experiments is 113 ± 2.8 kDa (mean ± s.d.), which is consistent with the predicted molecular weight of the full length chimeric protein (111.8 kDa) suggesting the lower band (100 ± 1.7 kDa) is processed, but this needs to be further tested.
To test for a dose-dependent effect of Shld1 on transcript levels, RNA was collected from late schizonts (46 ± 2 hpi) and 14 h later from the resulting ring stage parasites. Transcripts for ap2g, msrp1 and gexp5 genes, as well as gametocyte production (Fig. 9c-e & Supplementary Fig. 6) increase in a dose dependent manner with a plateau in RNA levels at the concentration of Shld1 normally used to stabilize GDV1 in culture (500 nM). The similar Shld1 dose response curve for gametocyte production and RNA levels suggests that these transcripts might be potential markers for early gametocytes. We test this correlation directly using the data obtained over the course of our various different Shld1 experiments (Fig. 9f-h). The R 2 values for correlation between gametocyte conversion and the RNA levels of each of the 3 genes and are > 0.44 in ring stage parasites. In particular, ap2-g levels in ring stage parasites and gametocyte conversion have an R 2 of 0.92. In total, these data suggest the capacity of GDV1 protein levels to modulate gametocyte formation and transcription levels of downstream genes.
Gdv1 allele frequency in high and low GCR samples. The ability of GDV1 to act as a rheostat for gametocyte production coupled with prior field work that identified a SNP in gdv1 (Chr9:1378602) 29 whose population frequency varied with transmission intensity prompted us to evaluate this SNP in patient samples in the high and low GCR cohorts that were used for metabolic analysis (Supplementary Table 4). Dried blood spots for DNA analysis were available for all twenty 2017 samples and all but one of the twenty 2016 samples. The high and low GCR samples have significantly different gdv1 allele distributions (Fisher exact test, p = 0.022) (Fig. 10), while there is no difference in msp2 allele frequencies in the two cohorts (Fig. 10). The frequency of the H 217 GDV1 allele previously associated with a limited seasonal malaria in The Gambia is higher in the high GCR samples (n = 20, 73%), while the frequency of the P 217 allele previously associated with continual malaria transmission in the Republic of Guinea is higher in the low GCR cohort (n = 19, 76%). This finding suggests that in areas with limited seasonal transmission H 217 parasites may be selected for the ability to produce a higher ratio of gametocyte-to asexually committed rings to maximize transmission.

Discussion
Here we report the first direct comparison of the presence of gametocyte-committed ring stage parasites in human blood samples with clinical parameters and parasite transcript levels.
The GCA allowed detection and quantification of gametocyte generating capacity 10 days before the circulation of transmissible stage V gametocytes. The results indicate that gc-rings circulate in most patient's peripheral blood, however the levels vary dramatically, from 78% to no detectable sexually committed rings. This pattern is consistent with a low basal level of sexual differentiation that is modulated by D0 parasitemia, patient temperature, gdv1 allele, GDV1-dependent gene expression and, possibly, LysoPC levels. The GDV1-dependent genes are also strong candidates for sensitive molecular markers that could be used to identify symptomatic and asymptomatic gc-ring carriers to ensure treatment before the production of circulating transmissible stage V gametocytes. The GCA assay developed in this work is a valuable first step toward evaluating gametocytogenesis in the field. The strong concordance of the GCR values calculated in the 2016 and 2017 malaria seasons as well as on days 4 and 8 demonstrate the reproducibility of the gametocyte counts and the ability of the stage II/III gametocytes detected on D4 to continue to mature to D8. The results are consistent with the circulation of gc-rings prior to sequestration in the bone marrow 49 . Although we cannot rule out a subpopulation of sequestered gc-rings that never circulate, the circulating gc-rings provide a window to assess gametocytogenesis in blood samples within 24 h of the production of gametocyte-committed merozoites. This early detection of gc-rings allows much closer association with host and clinical parameters that may contribute to the balance  between asexual and sexual differentiation. The further development of sensitive molecular gc-ring markers coupled with the existing stage V gametocyte markers (Pfs25 and mget, Pf3D7_1469900) 50 , should make it possible to track gametocyte maturation over the next 10-12 day in asymptomatic parasite carriers. Such longitudinal analysis was not possible in this study of symptomatic malaria patients because they had to be treated. At enrollment most of the study population had RT-qPCR detectable circulating stage V gametocytes as well as gcrings and there was a wide range of levels for both stages. However, there was no significant correlation between the number of stage V gametocytes in a D0 sample and its GCR suggesting that 10-12 days earlier their GCR was different. As mentioned above the only way to directly study the relationship between GCR and stage V gametocyte production is a longitudinal study monitoring asymptomatic individuals for the progression of gc-rings to stage V gametocytes two weeks later. We anticipate that young children at the beginning of the malaria season before the development of immunity would have a direct correlation between gc-ring markers/GCR, and stage V gametocytes 2 weeks later, which would provide in vivo validation for the GCR assay and gc-ring markers.
The positive association we found between D0P and GCR is consistent with previous in vitro data demonstrating that sexual differentiation increases with high parasitemia 51,52 , but attempts to identity a consistent density dependent trigger have been challenging 21 . Recently, LysoPC-depleted culture media has been shown to increase conversion rate in vitro 27 , but was not examined in humans. In the 40 samples we tested, the total plasma LysoPC levels ranged between (40-275 μM) and varied significantly between collection years. In 2017, high LysoPC levels are associated with low GCR, which is consistent with high plasma LysoPC having a negative influence on gametocyte production. The reason for the difference in LysoPC levels in the 2016 and 2017 plasma samples is unclear. As there was no significant difference in the other clinical parameters between the two years, it is possible that different plasma storage conditions in 2016 and 2017 could have been responsible. However, LysoPC levels have also been shown to be influenced by body weight 53,54 and infection; decreasing during severe malaria 41 , African trypanosomiasis 55 , and sepsis 56 so alternative explanations cannot be ruled out completely.
In contrast to low D0 parasitemia 51 and high LysoPC levels 27 , high temperatures have not previously been specifically ARTICLE associated with a decrease in sexual differentiation 57 , but have been shown to decrease parasite growth and viability 58,59 . The negative influence of high D0 temperature on the positive relationship between D0 parasitemia and GCR was consistently observed on D4 and D8, but the effect was more pronounced on D4. This difference in the influence of temperature could be due to a temperature-induced delay in growth decreasing the number of gametocytes that reached stage II by D4. Stage II is the first morphologically distinct gametocyte stage so a delay would reduce the D4 gametocyte count, but could have less effect on the D8 counts which included stages III-V. An inverse relationship of D0 parasitemia and fever on GCR suggests that for a given parasitemia gametocyte production would be higher in asymptomatic individuals, which is consistent with recent work showing a positive correlation between parasite density and stage V gametocytes in asymptomatic infections 50 and a negative correlation in a meta-analysis of symptomatic cases 40 . This inverse relationship between D0P and temperature also suggests that sexual differentiation would increase as an individual's pyrogenic threshold to Plasmodium infection increases 39 and this may contribute to the relative increase in mature gametocytes to asexual parasites that has been observed with age 16 .
In addition to the negative effect of temperature on sexual differentiation, another unexpected finding was the lack of correlation between high GCR and low hematocrit. Low hematocrit has previously been reported to be associated with an increase in circulating stage V gametocytes in patient blood smears 15,17,18 .
The lack of correlation between GCR and hematocrit suggests that low hematocrit is not the major independent driver of sexual differentiation in this study. It is possible that the low hematocrit previously associated with circulating stage V gametocytes could have been due to the prolonged infection required for enough gametocytes to mature 10 days and be released into the circulation as stage Vs for detection by Giemsa-stained blood smear. Such differences between the factors associated with gc-rings and circulating stage V gametocytes highlight the need to evaluate sexual differentiation longitudinally at 2 week intervals in asymptomatic individuals to assess both gc-rings and the resulting stage V gametocytes that circulate 2 weeks later.
Sexual differentiation in vivo was also associated with an increase in RNA levels of three genes shown to be dependent on GDV1 protein levels in vitro, as well as the H 217 GDV1 allele previously reported to be over represented in a region with limited seasonal malaria transmission 29 . Together, this data is consistent with GDV1 playing a key role in gametocytogenesis in vivo and the H 217 GDV1 allele inducing sexual differentiation more effectively than the P 217 allele in the field. In contrast to histidine, proline restricts the flexibility of the peptide chain and could alter protein structure potentially affecting function 60 , but this needs to be further evaluated by more detailed genetic studies. A recent field study extending the analysis of the gdv1 locus to additional West Africa locations found a structural dimorphism downstream from the gdv1 coding region that also demonstrated strong geographical divergence, but was not clearly linked to transmission intensity 61 . In contrast, an association between   Fig. 8 Gene expression patterns in the absence of GDV1. The fold change (2 −ΔΔC T ) of ap2-g a, msrp1 b hp1 c, and ap2-g3 d RNA levels was tested in early schizonts (36 ± 2 hpi) (light grey), late schizonts (46 ± 2 hpi) (medium grey) and 12 ± 2 h later in ring stage parasites (dark grey) in Pfgdv1.gfp.dd.T1 parasites in the absence (Off) and presence (On) of Shld1 (a-d) as well as the clonal 3D7.G def parasite line that lacks the gdv1 locus due to a spontaneous chromosome 9 deletion 24 . Expression levels in early schizonts in the absence of Shld1 are used as the reference for ap2-g a and msrp1 b, while the Shld1 minus ring stage expression levels were used for hp1 c and ap2-g3 d. The data are from 2 independent experiments performed in duplicate. To compare expression between stages and conditions in the Pfgdv1.gf.dd parasite line a two-way ANOVA with Sidak's multiple comparisons test was used. For the 3D7.G def line, a one-way ANOVA with Dunnett's multiple comparisons test was performed to compare expression between stages. For both analyses a Bonferroni post-hoc test to correct for the analysis of 4 genes was used and probability is indicated, p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***) low malaria transmission and high expression levels of gametocyte-associated genes, including ap2-g, has recently been reported in a publication comparing the transcriptomes of parasites isolated in areas with different patterns of malaria transmission 62 . However, they did not directly assess gametocyte production or gdv1 allele. Clearly further work is needed to integrate these intriguing findings.
To extend our molecular analysis of the regulation of gametocytogenesis, we used Pfgdv1.gfp.dd parasites that allow inducible gametocyte production by Shld1-dependent protection of GDV1 expression. GDV1 was required only during schizogony for the induction of gametocytes. Since schizonts are sequestered in vivo and not present in human blood samples we assessed regulation in vitro and found ap2-g and msrp1, not gexp5, transcript level were significantly increased during schizogony in the presence of GDV1. However, close analysis of the in vitro data during schizogony in the Pfgdv1.gdf.dd line and a gdv1 deficient line (G def ) indicates that, although ap2-g RNA levels are lower in the absence of GDV1, ap2-g transcripts still increase significantly during the transition from early to late schizonts. This low but significant GDV1-independent increase in ap2-g transcript levels suggests GDV1 is required to augment ap2-g mRNA levels during schizogony, rather than being directly involved in initiating transcription or required to release repression by displacing or preventing HP1 binding to H3K9Me3. The ability of Shld1mediated stabilization of GDV1 to enhance ap2-g RNA levels not var gene expression, both of which are repressed by HP1 31 , also argues against a direct effect of GDV1 on all HP1 repression sites. Based on recent data from P. yoelii it is possible that another ap2 transcription factor, ap2-g3 is required for this initial ap2-g expression and here we show that ap2-g3 levels are not affected during schizogony by GDV1 levels. It is possible that ap2-g3 initiates ap2-g transcription, while GDV1 is required to augment transcription or stabilize RNA levels, which is an established regulatory mechanism in P. falciparum [63][64][65] . Future functional studies will include clinically relevant parasitemias, temperatures, LysoPC levels and gdv1 alleles in vitro to better define the signals influencing gametocyte production.
One of the major limitations of this study is that all the volunteers were recruited from a single clinic in an area of high seasonal transmission with a low levels of microscopically detectable stage V gametocytes 66 . Reproducibility in different regions with different malaria transmission patterns and a wider range of ages is needed to validate the clinical parameters  Fig. 9 Shld1 dose-dependent increase in GDV1 protein levels. a Immunoblot of saponin-treated Pfgdv1.gfp.dd or wild type NF54 schizonts extracted in 8 M urea/5% SDS after 24 h incubation in the indicated amount of Shld1. The immunoblot was first probed with anti-GFP mAb and imaged, then reprobed with rabbit anti-Histone 3 (anti-H3) antibody as a loading control as described in Methods. The anti-GFP and anti-H3 positive bands are shown and the complete gel is shown in Supplementary Fig. 6A. b The density of the upper (dot) and lower (x) bands was determined after background subtraction using Image J software, normalized to the anti-histone 3 antibody signal and plotted. c-e Shld1 dose-dependent increase in gametocyte production and ap2-g, msrp1 & gexp5 transcription. c After growth in the indicated Shld1 concentration for~50 h, NAG was added and gametocyte production (average ± s.d.) was determined 9 days later by counting Giemsa-stained smears. d, e The average relative abundance (2 −ΔΔC T ) of ap2-g (blue), msrp1 (red), gexp5 (purple) and gdv1 (black) transcripts in RNA obtained from schizonts (46 ± 2 hpi) d and 14 h later from ring stage parasites (e) grown in the indicated concentration of Shld1 are plotted using early schizonts (36 ± 2 hpi) as the reference. The experiment was run twice with duplicate flasks and a representative data set is shown with the trend lines indicated. The second experiment is shown in Supplementary Fig. 6B-D. f-h Gametocyte conversion rate is positively correlated with ap2-g, msrp1, gexp5 mRNA expression. The gametocyte conversion rate and the relative abundance (2 −ΔΔC T ) of f ap2-g, g msrp1 or h gexp5 RNA in ring stage (14 ± 2 hpi) Pfgdv1.gfp.dd clone T1 parasites cultured in the absence (open circle) or presence (black dot) of 500 nM Shld1 are plotted. The data are from 3 independent experiments performed in triplicate or quadruplicate. The trend line and correlation coefficient (R 2 ) is shown and gene expression profiles found to be associated with gc-ring stage parasites. To extend the study to include lower parasitemias and remote areas without access to tissue culture facilities or trained staff to maintain and quantify the ex vivo cultures, robust molecular markers are needed to identify gcrings. Such markers will also allow the direct comparison of gcring production and maturation to stage V gametocytes two weeks later.
In summary, circulating gc-rings were detected in 76% of the 260 malaria patients tested using the GCA developed here. The ratio of gametocyte-to asexual-committed rings varied widely and was associated positively with D0 parasitemia and negatively influenced by patient temperature. GDV1 allele H 217 and transcript levels of three GDV1-regulated genes, ap2-g, msrp1 and gexp5 were significantly higher in individuals with high ratios of gametocyte-to asexual-committed rings, while in the 2017 plasma samples high levels of LysoPC were associated with low levels of gametocyte production. In vitro, GDV1 was found to play a critical role in the regulation of the balance between asexual and sexual development during schizogony, prior to the production of gametocyte-committed rings. The combined findings support consistent low-level gametocytogenesis in the human host that is modulated by parasitemia and fever as well as GDV1 levels and allele.

Methods
Study site and Population. Blood samples were collected from children attending the Ewim Health Center in Cape Coast, Ghana 66 . Briefly, children ( ≤ 13 yrs) with P. falciparum parasitemia ranging between 1,000 and 250,000 per µl of blood based on the WHO standardized protocol 67 were recruited into the study and their axillary temperature recorded.
Ethics Statement. The study was approved by the Institutional Review Board of the Noguchi Memorial Institute for Medical Research and the Ghana Health Services and reviewed by DMID, NIAID, NIH. Before recruitment each parent/ guardian was informed of the objectives, methods, anticipated benefits and potential hazards of the study. The parents/guardians were encouraged to ask questions about any aspect of the study that was unclear to them and informed about their liberty to withdraw their children at any time without penalty. Children were enrolled only after written parental consent had been obtained. All patient information is treated as confidential.
Sample collection. Blood samples were obtained from consented children prior to drug treatment (D0). For each subject, blood was collected into acid citratedextrose solution (ACD) (2.0 ml) for parasite cultures, a tube containing ethylenediaminetetraacetic acid (EDTA) (0.5 ml) (BD Biosciences, San Jose, CA) for hematological analysis and in 2016 a PAXgene RNA blood tube (2.5 ml) (BD Biosciences) for RNA isolation. A Urit 3000 Plus (Urit Medical Electronics, Guilin, Guangxi 541004.P.R.China) was used to assess the hematological parameters, including the WBC count and Urit-12 hemoglobin meter (Urit Medical Electronics) was used to determine hemoglobin levels according to manufacturers' instructions. Immediately after removing the samples for the hematological indices and making a thick blood smear, the EDTA sample was centrifuged and the plasma stored at −20°C for immunological studies. The thick blood smear was stained with Giemsa and used to quantify asexual and sexual stage parasites against 200 WBCs. In 2017 the cell pellet was resuspended in NucleoZOL and stored frozen until RNA analysis. All study participants were given a standard curative dose of artemether-lumefantrine (20/120 mg/kg) or artesunate-amodiaquine (4/10 mg/kg) and scheduled for a follow up visit 7 days later.
Ex vivo analysis of gametocyte commitment. The samples collected in ACD tubes were centrifuged for five minutes at 2000 rpm to harvest plasma. The cells were washed twice with 5 ml of sterile RPMI 1640 media without serum or AlbuMAX II (Thermo Fisher Scientific, Waltham, MA) and centrifuged at 2000 rpm for five minutes to remove white blood cells and the buffy coat. The red blood cell pellet was resuspended to a 3% hematocrit in RPMI 1640 media supplemented with 2% inactivated human serum and 0.5% AlbuMAX II. Aliquots of the resuspended RBCs (3 ml) were added into 2 separate wells of a 12 well plate. NAG was added and then the plate was incubated at 37°C in an atmosphere of 5% CO2/5% O 2 /90% N 2 for 8 days with daily media changes and thin smear preparation but without RBC supplementation. On the 8 th day, the cultures were harvested and the parasites preserved in NucleoZOL (Macherey-Nagel). D0, D4 and D8 parasite thin smears were Giemsa-stained and the distinct parasite stages observed by microscope (×1000) were counted. A total of 2000 RBCs were counted to determine the D0 ring stage parasitemia, while 20,000-30,000 RBCs were counted per smear from each of the duplicate D4 and D8 cultures. A sample was classified as gametocytedeficient if no gametocyte was observed in 30,000 RBCs. Gametocyte conversion rates were calculated for each sample by dividing the average D4 stage II-III gametocytemia or the average D8 stage III-V gametocytemia from the two NAGtreated ex vivo cultures by the D0 parasitemia. For quantitative reverse transcriptase-polymerase chain reaction (RT-qPCR) analysis of the D0 blood sample, RNA was purified using the PAXgene RNA kit (Qiagen, Hilden, Germany) according to the manufacture's protocol and converted to cDNA for RT-qPCR as described below in the Quantitative reverse transcriptase-polymerase chain reaction section. The gene specific primers were designed using Primer3 version 4.1.0 and are listed in Supplementary Table 5 and the 2 −ΔΔC T for each gene was calculated using 18s rRNA or sbp1 as the endogenous control and RNA levels in ring stage Pfgdv1.gfp.dd.T1 parasites grown in the absence of Shld1 as the reference. To quantify stage V gametocytes, Pfs25 RNA levels were determined using RT-qPCR with Pfs25 specific primers. As a positive control uninfected human blood was mixed with purified NF54 stage V gametocytes and then added to PAXgene tubes or NucleoZOL to replicate the RNA isolation described above for the 2016 or 2017 samples. Stage V parasite density per microliter was calculated using a hemocytometer to quantify the cell concentration and Giemsa-stained thin smears to determine the stage V gametocytemia. The control RNA was serially diluted and used as a RT-qPCR template to generate a standard curve for Pfs25 ( Supplementary  Fig. 1). Using the slope of the standard curve to convert Pfs25 C T to stage V parasites, we then calculated the number of stage V gametocytes per microliter of blood in each sample (Fig. 1f). Similarly, to quantify Pfs25 transcript levels, a region of Pfs25 (24-128 bp) containing the RT-qPCR site was PCR amplified (primers PF3D7_1031000-Fw to PF3D7_1031000-1RvCL, Supplementary Table 5), inserted into a Topo-TA plasmid and used to transform chemically competent One Shot® TOPO E. coli. A single colony was isolated, the plasmid purified, and the insertion of a single copy of Pfs25 confirmed by sequencing. The plasmid concentration was quantified by absorbance at 260 using a Qubit 3.0 Fluorometer (Thermo Fisher Scientific) and the plasmid copy number per microliter was determined using an online web-tool (https://cels.uri.edu/gsc/cndna.html). The plasmid was serial diluted and used as a qPCR template to generate a standard curve correlating Pfs25 C T to transcript number. The results were used to calculate the number of Pfs25 transcripts per microliter of blood in each sample (Fig. 1f). The number Pfs25 transcripts/µl was then divided by the number of ring stage parasites per microliter of blood measured from the Giemsa-stained D0 thick smear to determine the % sexual stage parasitemia (Fig. 1g).
Plasma metabolomic analysis. Plasma from 40 samples with D0 parasitemia > 0.35%, 10 with high GCR (5-66%) and 10 with undetectable to low GCR (0-0.3%) from each year (Supplementary Table 4) and three uninfected control samples were selected for evaluation using the AbsoluteIDQ p180 assay (Biocrates, Inc., Washington, DC) to quantify 180 metabolites according to the manufacturer's protocol at the Duke University's Proteomic Core Facility. The metabolites are detected using a using a triple quadrupole tandem mass spectrometer and included staple isotope labeled internal standards. Acylcarnitines, glycerophospholipids, and sphingolipids are analyzed by introducing the sample using a Flow Injection Analysis method with a single point calibration to determine concentration. Amino acids and biogenic amines were analyzed by performing by ultra-high pressure LC-MS/MS using a reversed phase analytical column for analyte separation. A seven point calibration curve was used to quantify the amino acids and biogenic amines.   Table 5). The PCR amplicon and p1605-GFP-FKBP-int plasmid 69 were digested with Xho1 and AvrII. Prior to ligation with T4 ligase the digested plasmid was treated with alkaline phosphatase for 1 h. Electrocompetent DH10B E. coli were transformed with the ligation reaction and grown overnight on LB plates containing ampicillin (100 µg/ml). Ampicillin-resistant colonies were selected, screened for insert by restriction enzyme digestion and sequenced before being used for parasite transformation. Established protocols 70 were used for transformation (100-150 µg plasmid DNA per transformation) and WR99210 (2.5 nM)resistant parasites were obtained from two independent transformations. After 3 weeks off drug, WR99210 (Jacobus Pharmaceuticals, Plainsboro, NJ) was reapplied and the parasites were cloned by limiting dilution. The presence of a singlecrossover chromosomal integration was assessed by PCR amplification using primers listed in Supplementary Table 5.
Flow cytometric analysis. Cells were resuspended at 0.01% hematocrit (~1,000 cell/µl) in buffer A (154 mM NaCl, 9.27 mM glucose, 10 mM Tris-HCl, pH 7.4) with 500 nM SYTO 59 and incubated at RT in the dark for 20 to 30 min prior to analysis on an Accuri C6 flow cytometer (BD Biosciences, Franklin Lakes, NJ). The forward and side scatter signals were used to select the total RBC population and then the SYTO 59 signal (Ex 640 nm/Em 675 ± 25 nm) was used to identify P. falciparum infected RBCs and monitor DNA replication during asexual development.
Quantitative reverse transcriptase-polymerase chain reaction. Samples collected at the indicated developmental stages were preserved in NucleoZOL (Macherey-Nagel) and RNA was isolated from the aqueous phase using the RNeasy Micro kit (Qiagen) according to the manufacturer's instructions. In addition to DNase treatment during RNA isolation on the Qiagen micro columns, purified RNA (1 µg) was treated with gDNA Wipeout Buffer before conversion to cDNA using SuperScript® VILO™ (Thermo Fisher Scientific). Reverse transcriptase minus controls were used to confirm the absence of genomic DNA. cDNA was used as a template for RT-qPCR (QuantStudio3 with Qcapture software 2.9.13, Applied Biosystems, Foster City, CA) with the indicated primers (Supplementary Table 5) and Fast SYBR Green PCR Master Mix (Applied Biosystems) using the following conditions, 20 sec activation at 95°C, 40 cycles of 1 sec at 95°C and 20 sec at 60°C. All samples were run in triplicate and tested for both the gene of interest and the control constitutive gene, arginyl-tRNA synthetase (PF3D7_1218600) 31 , on the same plate. The results were analyzed using QuantStudio TM Design & Analysis Software (Applied Biosystems) and the ΔC T values determined by subtracting the mean cycle threshold (C T ) value for arginyl-tRNA synthetase from the mean target gene C T . To compare gene expression profiles, the relative quantity (2 −ΔΔC T ) was calculated for each gene using RNA from Pfgdv1.gfp.dd.T1 parasites grown in the absence of Shld1 as the reference. The efficiency of the primers was tested by serial dilution and ranged from 87 to 111%.
GDV1 protein-dosing. Pfgdv1.gfp.dd parasite were MACs/sorbitol synchronized and Shld1 concentrations ranging from 0 to 1500 nM added at ring stage. For the immunoblot schizonts harvested 14 h later, treated with 0.05% saponin to remove the RBCs and then extracted in 8 M urea/5% SDS. The immunoblot was probed with anti-GFP mAb (#11-814-460-001, Roche, Indianapolis, IN) followed by horseradish peroxidase-labeled anti-mouse Ig and visualized using the Pierce SuperSignal West Dura kit (Thermo Fisher Scientific) and an ImageQuant LAS 4000 (GE Healthcare, Piscataway, NJ). The blot was then reprobed with rabbit anti-Histone 3 (anti-H3) antibody (#Ab1791, Abcam, Cambridge, MA), followed by alkaline phosphatase labeled secondary Ig and visualized using 5-bromo-4-chloro-3-indolyl-phosphate (BCIP) and nitro blue tetrazolium (NBT) (Thermo Fisher Scientific). The density of the upper and lower bands was determined after background subtraction and normalization to the anti-histone 3 antibody signal using Image J software. For expression profiling, Shld1 was added to ring stage parasites and at 46 ± 2 hpi they were harvested by centrifugation and resuspended in NucleoZOL for RNA preparation and RT-qPCR analysis. The remaining cultures were treated with NAG and monitored over the next 6 days for gametocyte production. The GCR was calculated by dividing the day 6 gametocytemia by the ring stage parasitemia at the time of RNA collection.
Genetic diversity. Genomic DNA was extracted from two 3 mm punches of dried blood spot (D0) using QIAamp DNA Mini Kit (Qiagen), eluted in 120 μl of distilled sterile water and used to analyze gdv1 and msp2 genotypes. For gdv1 a 613 bp section of gdv1 (bp 207-819) containing the target SNP (NGS_SNP. Pf3D7_09_v3.1378602) 29 was PCR amplified using primers, gdv1-outer Fw and gdv1-seq Rv (Supplementary Table 5). The PCR reaction mix contained 1 × AccuPrime ™ PCR Buffer II (15 mM MgCl2), 200 nM of each primer, and 0.4 unit of AccuPrime™ Taq DNA Polymerase (Thermo Fisher Scientific) in addition to 2 μl of genomic DNA (gDNA) template and consisted of initial denaturation at 94°C for 4 min, followed by 40-50 cycles at 94°C for 1 min; 38°C for 50 sec, and 60°C for 2 min; with final extension at 72°C for 8 min. The PCR product was purified using a PCR clean-up kit (Macherey-Nagel) and sequenced using primers gdv1-seq Fw and gdv1-seq Rv (Supplementary Table 5). For msp2 a nested PCR was used to distinguish the two major allelic families (FC27 and 3D7). First, the central polymorphic region was PCR amplified using family specific primers, msp2.Outer. M2-OF and msp2.Outer.M2-OR 66 (Supplementary Table 5) and reaction conditions described above for gdv1. Next, 0.5 μl of the outer PCR product was used as template in a PCR with primer msp2 3D7/FC27.S1fw and either msp2 FC27.M5rev or msp2 3D7.N5rev. PCR products were separated using 2% ethidium bromidestained agarose gels and visualized under UV illumination.
Statistical analysis. STATA v14 36 was used for the Pearson correlation and ANOVA analysis, as well as the independent regression analysis to evaluate the association of the clinical parameters and the ln transformed parasite data (D0 parasitemia, gametocytemia and gametocyte conversion rates on D4 and D8) 71 . To determine whether the parameters were normally distributed, histograms and normal quantile plots were used and those parameters that were not, D0 parasitemia, gametocytemia and GCR on D4 and D8, were ln transformed and retested. The metabolomics data for the high and low GCR cohorts were compared by unpaired, two-tailed Student's t-tests and corrected for multiple comparison FDR < 0.05 (Microsoft Excel 2013) 72 . The LysoPC ROC curve was calculated using SPSS v24 73 . ANOVA analysis followed by a post test for multiple comparisons (Prism GraphPad v7.05) 71 was used to compare the clinical parameters, including LysoPC levels in the high and low GCR samples in 2016 and 2017. Gene expression levels in the high and low GCR cohort samples were performed in duplicate and the 2 −ΔΔC T calculated using 18s rRNA or Sbp1 as the endogenous control and RNA levels in ring stage Pfgdv1.gfp.dd.T1 parasites grown in the absence of Shld1 as the reference. Differences between the 2 −ΔΔC T values for the high and low cohorts were analyzed by ANOVA using a Kruskal-Wallis test followed a Dunn's multiple comparison test (Prism GraphPad v7.05). RT-qPCR analysis comparing gene expression in Pfgdv1.gfp.dd parasites with and without Shld1 as well as 3D7.G def parasites were performed in triplicate in at least 2 separate experiments and analyzed by two way or one way ANOVA, respectively, with a Bonferroni post-test to correct for multiple comparisons (Prism GraphPad v7.05). Fisher's exact test was used to evaluate the frequency of gdv1 and msp2 alleles in the 2016 and 2017 high and low GCR samples and the Kruskal-Wallis test followed a Dunn's multiple comparison test was used to compare the GCR of the gdv1 and msp2 alleles (Prism GraphPad v7.05). The four samples, two from each cohort, with both gdv1 alleles and the 11 samples, 5 from the high and 6 from the low GCR cohort, with both msp2 alleles were included in the analysis. All statistical tests were two sided and P values > 0.05 were considered non-significant (ns), P values ≤ 0.05 = *, P values ≤ 0.01 = **, P values ≤ 0.001 = ***, P values ≤ 0.0001 = ****.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.