Title: A Drug Repurposing Approach Reveals Targetable Epigenetic Pathways in Plasmodium vivax Hypnozoites

: Radical cure of Plasmodium vivax malaria must include elimination of quiescent 10 ‘hypnozoite’ forms in the liver; however, the only FDA-approved treatments are contraindicated in many vulnerable populations. To identify new drugs and drug targets, we screened the Repurposing, Focused Rescue, and Accelerated Medchem library against P. vivax liver stages and identified the DNA methyltransferase inhibitors hydralazine and cadralazine as active against hypnozoites. We then used bisulfite sequencing and immunostaining to identify cytosine 15 modifications in the infectious stage (sporozoites) and liver stages, respectively. A subsequent screen of epigenetic inhibitors revealed hypnozoites are broadly sensitive to histone acetyltransferase and methyltransferase inhibitors, indicating that several epigenetic mechanisms are likely modulating hypnozoite persistence. Our data present an avenue for the discovery and development of improved radical cure antimalarials. 20 One-Sentence Summary: A drug repurposing screen reveals antihypertension drugs are active against P. vivax hypnozoites and epigenetic mechanisms play a role in hypnozoite quiescence.


Main Text:
Controlling malaria caused by Plasmodium vivax is complicated by the ability of P. vivax sporozoites, the infectious stage inoculated by mosquitoes, to invade hepatocytes and become quiescent (1).These quiescent 'hypnozoites' persist, undetectable, for months or even years before resuming growth and initiating a 'relapse' blood stage infection, leading to subsequent transmission back to mosquitoes (2).Hypnozoites are refractory to all antimalarials except the 8aminoquinoline class, which cannot be administered to pregnant women or glucose-6-phosphate dehydrogenase-deficient individuals and are ineffective in persons with specific cytochrome P450 genotypes (3).New antimalarial drug discovery and development with a target chemoprofile for killing hypnozoites, termed radical cure, has only recently been made possible with the introduction of cell-based phenotypic screening platforms featuring a monolayer of hepatocytes infected with sporozoites, a portion of which go on to form hypnozoites (4).Protein target-based approaches for hypnozonticidal drug discovery are in their infancy as the transcriptome of hypnozoites has only recently been reported and robust methods for routine in vitro propagation and genetic manipulation of blood stages of P. vivax are still underdeveloped (5,6).
To address the lack of radical cure drug leads and targets, we used our advanced P. vivax liver stage platform to screen the Repurposing, Focused Rescue, and Accelerated Medchem (ReFRAME) library (7).This library consists of approximately 12,000 developmental, approved, and discontinued drugs with the expectation that the repurposing of compounds with some optimization or regulatory success could expedite the decade-long path drugs typically progress through from discovery to licensure (8).To accomplish this screen, we assembled an international collaboration with laboratories in malaria-endemic countries whereby vivax-malaria patient blood was collected and fed to mosquitoes to produce sporozoites for infecting primary human hepatocytes (PHH) in screening assays performed on-site.As expected, few hits with anti-hypnozoite activity were detected since discovery of inhibitors for non-proliferating cells is notoriously difficult (9,10).Interestingly, two structurally related compounds used for treating hypertension, hydralazine and cadralazine, were found effective at killing hypnozoites.As these inhibitors have been shown to modulate DNA methylation (11), we next pursued and confirmed the existence of cytosine modifications in P. vivax sporozoite and liver stages.To further investigate the role of epigenetics in hypnozoites, we screened an epigenetic inhibitor library using an improved version of the screening platform.Hypnozoites were found to be susceptible to several classes of epigenetic inhibitors, including six different histone deacetylase inhibitors and two inhibitors with targets regulating histone methylation.In lieu of traditional molecular biology methods for understanding this pathogen, our chemical biology approach reveals multiple, druggable pathways in P. vivax hypnozoites and sheds light on some of the processes underpinning their quiescence.

ReFRAME library screening cascade, hit identification, and confirmation
Hypnozoites become insensitive to most antimalarials after five days in culture, indicating they mature following hepatocyte infection (12,13).Importantly, development of hits with radical cure activity in vivo requires the screening against mature hypnozoites in vitro (14).While our 8day P. vivax liver stage platform has been used for screening smaller libraries against mature hypnozoites (13), the size of the ReFRAME library presented a logistical challenge.We anticipated dozens of P. vivax cases, each with a unique genetic background, would be needed to produce the sporozoites required to screen the 40 microtiter plates containing the library, thus the library was split such that plates were screened at Shoklo Malaria Research Unit (SMRU) in Thailand and the Institute Pasteur of Cambodia (IPC).Ultimately, 36 P. vivax cases from either site were needed to complete the primary screen over the course of 18 months (Fig. 1A and fig.S1).
From our analysis of primary screen activity, we noted several hydrazinophthalazine vasodilators were potentially active (fig.S1).We then selected 72 compounds, including 10 hydrazinophthalazine analogs, for confirmation of activity and potency determination in a doseresponse format, as well as counterscreening for additional antimalarial activity and cytotoxicity.Of the twelve hits which exhibited selective hypnozonticidal activity, cadralazine displayed the best combination of potency (pEC50 = 6.33 ± 0.33) and maximal inhibition near 100% (Fig. 1, B  and C, and fig.S2).To ensure hits were not merely specific to our platform, select hits were communicated to the Novartis Institute for Tropical Diseases (NITD), where the activities of hydralazine and cadralazine were independently confirmed using a P. vivax case from southern Thailand and separate batches of compounds (Fig. 1C and fig.S3).
Currently, the gold-standard for in vivo anti-relapse efficacy is rhesus macaques infected with Plasmodium cynomolgi, a zoonotic, relapsing species closely related to P. vivax (15).To further pursue drug repurposing of hydrazinophthalazines, we sought to confirm and measure the potency of cadralazine and other ReFRAME hits against P. cynomolgi B strain hypnozoites in vitro using an 8-day assay featuring primary simian hepatocytes (PSH) at NITD.Surprisingly, hydralazine and cadralazine were found inactive when tested in three different PSH donors (Fig. 1C and fig.S3).This negative result was later confirmed in an 8-day, simianized version of the platform at the University of Georgia (UGA) using the P. cynomolgi Rossan strain infected into two different PSH lots (Fig. 1C).Furthermore, hydralazine and cadralazine were not identified as hits in any of the 112 screens of the ReFRAME published to date (ReFRAMEdb.org),suggesting these compounds are specific to the P. vivax liver stage assay.

Immunofluorescent detection of DNA methylation in P. vivax and P. cynomolgi liver stages
Hydrazinophthalazines have been shown to involve direct inhibition of DNA methyltransferases (16) and hydralazine has also been used to study DNA methylation in the Plasmodium falciparum asexual blood stages (17).Investigations into blood-stage parasites methylation have identified the presence of low levels of 5-methylcytosine (5mC), 5-hydroxmethylcystosine (5hmC), and under-characterized 5hmC-like marks throughout the genome (17,18).To confirm the possible mechanism of hydrazinophthalazines on hypnozoites, we first used immunofluorescence staining with commercial anti-5mC and anti-5hmC monoclonal antibodies to identify methylation patterns in P. vivax liver schizonts and hypnozoites at 6 days postinfection.We found clear evidence of 5mC, but not 5hmC, in both schizonts and hypnozoites, morphologically consistent with the presence of 5mC in the parasite's nucleus (7) (Fig. 2A and figs.S4-S6).Due to expected 5mC and 5hmC signals from host hepatic nuclei, we opted to use high-content imaging (HCI) as an unbiased approach for quantifying 5mC signal within parasites.Image masks were generated to quantify the area of 5mC or 5hmC stain within each parasite (fig.S7), the values of which were then plotted as stain area per hypnozoite or per schizont (Fig. 2B).While some evidence of 5hmC-positive forms did appear from this analysis, the net area per parasite was found significantly lower than that quantified for 5mC-stained forms, indicating the relative level of 5mC marks predominate that of 5hmC.
Given the different susceptibility of P. cynomolgi hypnozoites to hydrazinophthalazines as compared to P. vivax, we performed HCI analysis of 5mC-and 5hmC-stained P. cynomolgi M/B-strain liver schizonts and hypnozoites at 8 days post-infection.Like P. vivax, we found both P. cynomolgi liver schizonts and hypnozoites are positive for 5mC, but not 5hmC.However, the 5mC stain morphology and intensity were relatively lower in P. cynomolgi hypnozoites versus P. vivax hypnozoites, indicating the possibility of different roles of methylation in the epigenetic network of these two species (fig.S8).

Detection of cytosine modifications in P. vivax and P. cynomolgi sporozoites using liquid chromatography-tandem mass spectrometry and bisulfite sequencing
We next sought to confirm the presence of cytosine methylation in the P. vivax and P. cynomolgi genomes using mass spectrometry and bisulfite sequencing.Initially, we assessed if these methodologies were plausible on P. vivax liver stages, but concluded that, without an available single live-cell approach, sequencing coverage of the parasite's genome would be overwhelmed by the genomic material from the host cell and neighboring uninfected hepatocytes following the harvesting of liver stage cultures (19).We therefore collected sufficient genomic material from P. vivax and P. cynomolgi sporozoites to analyze the nucleoside mixture arising from the enzymatic digestion of genomic DNA by liquid chromatography-tandem mass spectrometry as well as for detection of DNA methyltransferase (DNMT) activity using a commercial in vitro DNA methylation assay (17).While we detected 5mC and DNMT activity in Plasmodiumenriched samples with these approaches, possible contamination by the mosquito's microbiota could not be excluded (fig.S9).We next analyzed DNA methylation loci at single-nucleotide resolution using bisulfite sequencing of 3×10 7 P. vivax sporozoites, generated from three different cases, as well as 3×10 7 P. cynomolgi sporozoites (Fig. 3, A and B).A total of 161 and 147 million high-quality reads were sequenced for P. vivax and P. cynomolgi samples, respectively (fig.S9).The average 5mC level detected across all cytosines was 0.49% and 0.39% for P. vivax and P. cynomolgi, respectively.These percentages are comparable to the 0.58% methylation level detected in P. falciparum blood stages (17), but likely underestimate methylated loci considering the coverage we achieved (see methods).
We then monitored the distribution of detected 5mC along the P. vivax and P. cynomolgi chromosomes (figs.S10 and S11) and observed a stable methylation level throughout, including in telomeric and sub-telomeric regions.We further examined the context of genome-wide methylations and, similar to what we previously observed in P. falciparum (17), methylation was detected as asymmetrical, with CHH (where H can be any nucleotide but G) at 69.5% and 70.5%, CG at 16% and 15.7%, and CHG at 14.3% and 13.64%, for P. vivax and P. cynomolgi, respectively (Fig. 3C).We then measured the proportion of 5mC in the various compartments of gene bodies (exons, the introns, promoters, and terminators) as well as strand-specificity (Fig. 3,  D and E).We observed a slightly increased distribution of 5mC in promoters and exons compared to the intronic region, as well as in the template versus non-template strand, in P. vivax and P. cynomolgi.These results were consistent with previous data obtained in P. falciparum and in plants (17).Such a strand specificity of DNA methylation patterns can affect the affinity of the RNA polymerase II and impact transcription.We therefore investigated the relationship between DNA methylation and transcription and examined the methylation level in the various compartments of gene bodies and compared it to the mRNA levels measured by RNA-seq in P. vivax sporozoites (20).We found a trend between methylation and mRNA abundance in the proximal promoter regions and the beginning of the gene bodies, with highly-expressed genes appearing hypomethylated and weakly-expressed genes hypermethylated (Fig. 3F).These results suggest that methylation level in proximal promoter regions as well as in the first exon of the genes may affect, at least partially, gene expression in malaria parasites.While these data will need to be further validated and linked to hypnozoite formation at a single-cell level, we have determined that 5mC is present at a low level in P. vivax and P. cynomolgi sporozoites and could control liver stage development and hypnozoite quiescence.

Assay improvements and epigenetic inhibitor library screen
Upon completion of the ReFRAME and previously reported screens (13), we retroactively analyzed the screening platform's performance and developed an improved protocol to address two shortcomings.First, we hypothesized that false negatives could arise when a true-positive compound results in nonviable forms which persist in culture and are counted during HCI; and that extending the assay endpoint by four days would allow more time for attenuated forms to be cleared from the culture (13).Second, because the ReFRAME primary screen was completed using two lots of PHH, we noted a few lot-specific results, such as differences in potency of the monensin control (fig.S12).We therefore added the phase I metabolism inhibitor 1aminobenzotriazole to culture media on treatment days to minimize the effect of lot-specific hepatic metabolism on assay results (21) (fig.S13).Following development of the improved in vitro assay, this 12-day protocol was used to reconfirm twelve ReFRAME hits (fig.S14), resulting in confirmation of poziotinib--a primary screen hit that was previously confirmed in the NITD P. vivax assay only (Fig. 1B and fig.S3).Given the apparent importance of epigenetic mechanisms in hypnozoites, we obtained a commercially-available epigenetic inhibitor library and screened it against P. vivax liver stages at SMRU and IPC using the improved protocol (fig.S15).Confirmation of hits in dose-response assays resulted in selective potency for 11 epigenetic inhibitors targeting five different epigenetic mechanisms (Table 1).

Discussion
Herein we demonstrate several significant advances that are prerequisites for radical cure antimalarial drug discovery and development, including the first report of screening a mediumsized compound library against mature hypnozoites as well as detection of novel hits with mechanisms unrelated to that of 8-aminoquinolines.Despite the success of the original screening platform protocol, the improvements described herein should be considered for other platforms grappling with the biological complexity of metabolically-active primary hepatocytes or using a phenotypic readout without a validated hypnozoite viability marker.This report also provides an opportunity to compare radical cure hits assayed against both P. vivax and P. cynomolgi hypnozoites.While the P. cynomolgi-infected rhesus macaque system has long been used as a model system for studying malaria relapse, the only class of compounds available to assess the predictive value of this model for chemotherapeutics are the 8-aminoquinolines, which lack a distinct parasite target (22)(23)(24)(25).We found mixed results, with hydrazinophthalazines negative, and poziotinib positive, against P. cynomolgi hypnozoites.While further studies are needed to confirm if the target of any hit in this report is parasite-or host-directed, this would indicate there is sufficient diversity in gene expression, structural biology, or mechanisms of hepatic quiescence between P. cynomolgi and P. vivax hypnozoites that some hits will be speciesspecific.However, this result might also be attributed to differential metabolism in human and monkey hepatocytes (26).Regardless, these results highlight that in vivo P. vivax relapse models should be further developed and validated, as those hits lacking activity in P. cynomolgi might be abandoned for no other reason than the inability to demonstrate in vivo efficacy prior to first-inhuman studies (27).This report also adds to the broader discussion surrounding the successes and challenges of drug repurposing (28).Of the hits we identified, colforsin daropate, rhodamine 123, and poziotinib are used for cancer indications and, like the hydrazinophthalazines, have known human targets, implying that selectivity and off-target toxicity are hurdles to be addressed if repurposing for radical cure is to be successful.While direct repositioning of a known drug as a safe treatment for a new indication is the ideal outcome, it is also challenging.However, the identification of hits that can serve as advanced starting points for further optimization has potential for reducing the time and cost involved in developing an efficacious therapy.Alternatively, identification of tool compounds that delve into and reveal novel aspects of the biology of a disease can also promote further drug discovery approaches.As such, this report provides significant chemical biology leads into the essential nature of hypnozoite quiescence.Epigenetic control of pathogenic dormancy via DNA methylation has been described for cancer cells (29) and tuberculosis (30); and is a critical process in plants, which share evolutionary traits with Plasmodium (31).DNA methylation in the genus Plasmodium was first described in P. falciparum blood stages (17), and one type of modification has been demonstrated as positively associated with continuous gene expression (18).As technologies improve, it will be critical to validate the importance of such modifications in live stages, including hypnozoites.
Several additional chemical biology leads were revealed in our screens.Five hydroxamic acidcontaining inhibitors (panobinostat, AR42, abexinostat, givinostat and practinostat) and one natural product (raddeanin A) targeting histone deacetylase were found to selectively kill hypnozoites.Furthermore, the hit MI2, targeting the interaction between menin, a global regulator of gene expression, and MLL, a DNA-binding protein that methylates histone H3 lysine 4 (32), and the hit cyproheptadine, targeting SET-domain-containing lysine methyltransferase (33), highlight the interplay between DNA methylation and histone modifications (34) and are further evidence histone methylation regulates hypnozoites (20,35).Other hits, including 666-15, targeting the transcription factor cAMP response element-binding protein (36), and the kinase inhibitors cerdulatinib and CCT241736, suggest transcription factors and kinase signaling may also play a role in establishing, maintaining, or exiting quiescence.Similarly, the cancer drug poziotinib inhibits HER2, a tyrosine protein kinase associated with the downregulation of apoptosis and metastasis (37).As we recently reported that host apoptotic pathways are downregulated in P. vivax-infected hepatocytes, poziotinib could act by increasing apoptotic pathways in infected host cells (19).The hit MS-0735, an analog of our previously reported hypnozonticidal hit MMV018983 (13), is a ribonucleotide-reductase (RNR) inhibitor used as an antiviral.Needed for producing deoxyribonucleosides for DNA synthesis, RNR is a peculiar mechanism for nonreplicating hypnozoites, however, it has been reported that RNR is critical for DNA damage repair (38) and is expressed in P. vivax liver schizonts and hypnozoites (19).Ongoing efforts to make available disruptive methods for studying P. vivax, including genetic manipulation and in vivo models for relapse, will enable further validation of these leads and a more comprehensive understanding of the mechanisms of hypnozoite quiescence.
We are grateful to Calibr's Compound Management and High Throughput Screening Groups for their assistance with this project.HCI data from drug studies was produced by the Biomedical Microscopy Core at UGA, supported by the Georgia Research Alliance.SMRU is part of the Mahidol Oxford Research Unit, supported by the Wellcome Trust of Great Britain (#220211).Material has been reviewed by the Walter Reed Army Institute of Research.There is no objection to its presentation and/or publication.The opinions or assertions contained herein are the private views of the author, and are not to be construed as official, or as reflecting true views of the Department of the Army or the Department of Defense.This publication includes data generated at the University of California, San Diego IGM Genomics Center utilizing an Illumina NovaSeq 6000 that was purchased with funding from a National Institutes of Health SIG grant   non-template strand.(F) The previously reported mRNA abundance of P. vivax sporozoites was retrieved (20) and genes ranked.The 5mC levels in 5' flanking regions, gene bodies, and 3' flanking regions were placed into five bins and are shown for highly expressed (90th percentile, left) and weakly expressed (10th percentile, right) genes.Red: template strand, blue: non template strand.A3).Pharmacokinetic studies were conducted at WuXi AppTec Co., Ltd., in accordance with the WuXi IACUC standard animal procedures along with the IACUC guidelines that are in compliance with the Animal Welfare Act (40).

ReFRAME primary screen against P. vivax liver stages
The complete, step-by-step protocol for the P. vivax liver stage assay is published (41).In summary, two days after assay plates (Greiner Bio-One cat 781956) were seeded with PHH, sporozoites were dissected from mosquito salivary glands and allowed to infect cultures.The ReFRAME library was screened using the original, 8-day radical cure assay, in which developing liver schizonts and mature, PI4Ki-insensitive hypnozoites were treated on days 5-7 post-infection (7,13).On treatment days, a pintool was used to transfer 40 nL of compounds from the source plates to the assay plates.A single PHH lot, UBV, was first used for screening, however, once all available cryovials were used, screening was completed with lot BGW (fig.S1).Screening was initiated at SMRU until a second screening site was established at IPC, where all unfinished source plates were shipped and assayed.Some plates were assayed more than once in order to obtain a single run with a sufficient Z' factor of > 0.0 or two moderate-quality runs allowing for identification of reproducibly-active wells (fig.S1).Quantification of PvLS growth was performed by fixing and staining cultures with recombinant mouse-anti P. vivax Upregulated in Infectious Sporozoites 4 (rPvUIS4) (42).followed by high-content imaging and analysis using an ImageXpress Micro (Molecular Devices) or Lionheart FX (Agilent).Hypnozoites were classified as forms of less than 125 μm 2 growth area.
Normalization, hit selection, and dose-response confirmation in P. vivax liver stage assays Primary screening data were imported into Genedata Screener, Version 15.0.1-Standard and normalized to DMSO (neutral) and inhibitor (monensin) control-treated wells (neutral controls minus inhibitors).For four plates where the monensin control failed due to solubility issues combined with PHH lot variability (fig.S12), data were normalized using the Robust Z-Score method, which calculates for each well the Robust Z-Score (number of standard deviations off the median) based on the statistics of the compound wells per plate.Genedata multiplicative pattern correction was applied to adjust for plate edge effects.Sixty-two most active (≥ 67% normalized inhibition of hypnozoite numbers) and non-toxic (≤ 40% host cell toxicity) compounds and 10 hydrazinophthalazines were selected for reconfirmation in an 8-point 1:3 dose response following the 8-day protocol with PHH lot BGW using a dose-response of monensin and nigericin as redundant positive controls.Once hydralazine and cadralazine were identified as reconfirmed hits, commercially available batches of powder were obtained (budralazine, Chemcruz cat sc-504334 batch D3019, cadralazine, Chemcruz cat sc-500641 batch B2417, and hydralazine, Selleckchem cat s2562 batch S256202) and used for additional reconfirmation runs using the same 8-day protocol (Fig. 1 and fig.S2).
Hit confirmation in P. cynomolgi liver stage assays at UGA P. cynomolgi assays at UGA were performed using the step-by-step protocol for the P. vivax liver stage assay (41) with a few modifications.A Japanese macaque (Macaca fuscata) was intravenously infected with P. cynomolgi Rossan strain cryopreserved ring stage parasites (43) and allowed to reach patency.When parasitemia reached approximately 5,000 parasites per microliter, An. dirus mosquitoes were fed directly on the infected animal over a period of three to four days.The blood-fed mosquitoes were then checked for infection six to eight days by dissecting and staining midguts with 2% mercurochrome to detect oocysts.Two experiments were performed, one with PSH lot CWP, and one with PSH lot NPI.Two days after assay plates (Greiner Bio-One cat 781956) were seeded with 20,000 live PSH per well, sporozoites were dissected from mosquito salivary glands at day 16 post-bloodmeal and allowed to infect cultures.Hits were confirmed using the same 8-day radical cure assay.On treatment days, a pin tool was used to transfer 40 nL of compounds from the source plates to the assay plates.Quantification of P. cynomolgi liver stage growth was performed by fixing and staining cultures with 100 ng/mL mouse monoclonal antibody 13.3 (anti-GAPDH) obtained from The European Malaria Reagent Repository (http://www.malariaresearch.eu)followed by high-content imaging and analysis using an ImageXpress Micro (Molecular Devices).Hypnozoites were classified as forms of less than 105 μm 2 growth area.
Hit confirmation in P. cynomolgi and P. vivax liver stage assays at NITD Lots of both PSH and PHH were obtained from BioIVT.Hepatocytes were seeded at 22,000 cells per well in a 384-well plate (Corning cat 356667).Prior to and during the infection, the hepatocytes were cultured in BioIVT CP Medium (cat Z990003) with the addition of 1% PSN (Gibco cat 15640055) and 0.1 % gentamicin in the case of P. vivax.Two days post-seeding, the hepatocytes were infected with sporozoites dissected from the salivary glands of An. dirus mosquitoes.Sporozoites were collected in RPMI 1640 (KD Medical cat CUS-0645).Hepatocytes were infected with 10,000 sporozoites per well and spun for 5 minutes at 200 × g.Once the sporozoites were removed after 24 h of incubation, the culture media was exchanged to include 5% PSN in the case of P. cynomolgi.On days 4, 5, 6, and 7 post-infection, the hepatocytes received fresh compound addition in media.The cells were fixed on day 8 using 4% paraformaldehyde.
Liver stage parasites were detected by immunofluorescence assay.Hepatocytes were permeabilized for 1 h at room temperature in blocking buffer consisting of 2% Bovine Serum Albumin (Millipore Sigma cat A2153) and 0.2% Triton X-100 (Millipore Sigma cat 648466) in 1× PBS (Gibco cat 20012-027).For P. cynomolgi staining, the two in-house primary antibodies used were mouse anti-PcUIS4 monoclonal at 10 ng/mL, and rabbit anti-PcHSP70 polyclonal at 200 ng/mL.For P. vivax staining, rabbit anti-PvMIF was used at 1:1000 (44).The primary antibodies were diluted in blocking buffer and incubated overnight at 4 °C.Hepatocytes were washed thrice with 1× PBS and then incubated with secondary antibodies (Invitrogen cat A11013, RRID: AB_2534080 and A11036, RRID: AB_10563566) used at a 1:1000 dilution and Hoechst 33342 (Invitrogen cat H3570) used at 2 μg/mL for 2 h at room temperature.After the incubation, the hepatocytes were washed 3 times with 1× PBS and were stored in 50 μL per well of 1× PBS prior to imaging on an ImageXpress Micro (Molecular Devices).
Confirmed hit counterscreens: P. falciparum asexual blood stage at Calibr The SYBR Green I-based parasite proliferation assay (45) was used to determine the activity of compounds against the asexual blood stage of P. falciparum strain Dd2-HLH, a transgenic line expressing firefly luciferase (46).Briefly, acoustic compound transfer (Labcyte Echo 550) was used to prepare assay ready plates to which parasites in assay medium were added and incubated with compounds for 72 h.SYBR Green I in lysis buffer was used as detection reagent.
Fluorescence signal was read on the PHERAstar FSX plate reader (BMG Labtech).Compounds were tested in technical triplicates on different assay plates across three biological replicates performed on different days.Data were uploaded to Genedata Screener, Version 16.0.3-Standardand normalized to DMSO (neutral) and inhibitor control-treated wells (neutral controls minus inhibitors), with 1.25 µM dihydroartemisinin used as a positive control.Dose curves (thirteen point, 1:3 dilution series) were fitted with the four parameter Hill Equation.
Confirmed hit counterscreens: P. falciparum asexual blood stage at UGA Budralazine, cadralazine and hydralazine (same catalog and batches as above) were tested using the [ 3 H]-hypoxanthine drug susceptibility assay as previously described, with some modifications (47).Strain W2 (48,49) was grown in continuous culture using RPMI 1640 media containing 10% heat-inactivated type A+ human plasma, sodium bicarbonate (2.4 g/L), HEPES (5.94 g/L), and 4% washed human type A+ erythrocytes.Cultures were gassed with a 90% N2, 5% O2, and 5% CO2 mixture and incubated at 37 °C.Cultures were sorbitol synchronized to achieve >70% ring stage parasites (50).Assay were started by establishing a 0.5−0.7%parasitemia and 1.5% hematocrit in complete media.Assays were performed in 96-well plates with a volume of 90 μL/well of parasitized erythrocytes and 10 μL/well of 10× test compound.Dihydroartemisinin was plated as a positive control and DMSO as a negative control.Assay plates were incubated in the abovementioned gas mixture at 37 °C for 48 h; then, 3 H-hypoxanthine (185 MBq, Perkin Elmer cat NET177005MC) was added, and plates were incubated for another 24 h.After 72 h of incubation, the assay plates were frozen at −80 °C.Plates were allowed to thaw at room temperature before well contents were collected onto filtermats using a plate harvester (PerkinElmer).A Micro Beta liquid scintillation counter (PerkinElmer) was used to quantify radiation (counts-per-minute) representing relative parasite growth.Values were normalized to controls and plotted using CDD Vault.Potency values represent means of at least two independent experiments.
Confirmed hit counterscreens: P. cynomolgi asexual blood stage at UGA Budralazine, cadralazine and hydralazine (same catalog and batches as above) were tested against P. cynomolgi DC strain using the [ 3 H]-hypoxanthine drug susceptibility assay as previously described, with some modifications (47).P. cynomolgi was grown in continuous culture using RPMI 1640 +GlutaMAX media containing 20% heat-inactivated rhesus serum, hypoxanthine (32 mg/L), HEPES (7.15 g/L), additional glucose (2g/L), and 5% washed rhesus erythrocytes.Cultures were incubated at 37°C under mixed gas conditions of 90% N2, 5% O2, and 5% CO2.Schizonts were synchronized over a 60/20 Percoll gradient to achieve >90% late-stage parasites.Assays were started the following day when ring-stage parasites were present.Parasites were prepped for assay by establishing 0.5% ring-stage parasitemia and 2% hematocrit in complete media without hypoxanthine.Assays were performed in 96-well plates with a volume of 90 μL/well of parasitized erythrocytes and 10 μL/well of 10× test compounds.Compounds were plated from a starting concentration of 5 μM in an 11-point 1:2 dilution series and tested in duplicate.Uninfected RBCs were plated as a positive control and DMSO was used as a negative control.[ 3 H]-hypoxanthine (185 MBq, Perkin Elmer cat NET177005MC) was then added to all wells and plates were incubated under the previously-mentioned conditions for 72 h.After 72 h the assay plates were frozen at -80 °C.Plates were thawed the following day at room temperature and well contents were collected onto filtermats using a plate harvester (PerkinElmer).A Micro Beta liquid scintillation counter (PerkinElmer) was used to quantify radiation (counts-per-minute) representing relative parasite growth.Values were normalized to controls and plotted using CDD Vault.Potency values represent means of at least two independent experiments.
Compounds were tested in technical triplicates on different assay plates across three biological replicates performed on different days.Data were uploaded to Genedata Screener, Version 16.0.3-Standardand normalized to DMSO (neutral) and inhibitor control-treated wells (neutral controls minus inhibitors), with 1 µM KAF156 used as a positive control.Dose curves (thirteen point, 1:3 dilution series) were fitted with the four parameter Hill Equation.
Immunofluorescent staining of methyl-cytosine modifications in P. vivax liver stages Sporozoites from three different P. vivax cases were infected into PHH lot BGW at day 2 postseed (for case 1) or day 3 post-seed (for cases 2 and 3) in 384-well plates (Greiner Bio-One cat 781956) using the same methods for initiating P. vivax liver stage screening assays described above.Cultures were fixed at day 6 post-infection and stained with rPvUIS4 and Hoechst 33342 as previously described (41,42).Cultures were then stained for either with rabbit anti-5mC monoclonal antibody (clone RM231, ThermoFisher Scientific cat MA5-24694, RRID: AB_2665309) or rabbit anti-5hmC monoclonal antibody (clone RM236, ThermoFisher Scientific cat MA5-24695, RRID: AB_2665308) using methods adapted from those previously described (18).In summary, cultures were re-permeablilized with 0.1% (v/v) Triton-X 100 for 20 min at room temperature and then washed thrice with 1× PBS.Chromatin was then denatured with 4 N HCl for 30 min at room temperature and washed thrice with 1× PBS.The denaturing reaction was then neutralized with 100 mM Tris (pH 8.0) for 10 min at room temperature and washed thrice with 1× PBS.Cultures were then quenched with 50 mM NH4Cl for 10 min at room temperature and washed thrice with 1× PBS.Cultures were then blocked with 0.1% (v/v) Tween 20 and 2% (w/v) Bovine Serum Albumin for 10 min at room temperature and washed thrice with PBS.
Cultures were then stained with either antibody diluted to 10 μg/mL in PBS overnight at 4 °C and washed thrice with 1× PBS.Cultures were then stained with 10 μg/mL Texas Red-conjugated, goat anti-rabbit IgG secondary antibody (ThermoFisher Scientific, cat T-2767, RRID: AB_2556776) overnight at 4 °C and washed thrice with 1× PBS.For a negative stain control, a separate set of infected wells were prepared as above and stained with secondary antibody only (2' control, Fig. 2B).High resolution images of individual PvLS parasites and PHH nuclei were obtained by capturing 8 planes in the Z dimension using a 100× objective on Deltavision Core (GE Healthcare Life Sciences) and deconvoluted using softWoRx (GE Healthcare Life Sciences) (Fig. 2A and figs.S4, S5, and S6).An ImageXpress Micro high content imager was used to quantify methyl-cytosine modifications for the entire population of PvLS parasites from each case.
A 20× objective was used to capture 25 fields of view from each well (covering the entire growth area) of the 384-well plate.Using the associated MetaXpress high content analysis software, the rPvUIS4 stain from each PvLS parasite was used to define parasite objects, and the 5mC or 5hmC staining of host cell nuclei were used to define positive methyl-cytosine modification objects.The 2-dimensional area of intersection of both objects was then quantified for each parasite (Fig. 2B,     S7).
Immunofluorescent staining of methyl-cytosine modifications in P. cynomolgi liver stages A Japanese macaque (Macaca fuscata) was intravenously infected with P. cynomolgi M/B strain ( 53) and allowed to reach patency before skin feeding to An. dirus mosquitoes as described above.
One round of mosquito dissection and culture infection was performed with PSH lot NPI.Two days after assay plates (Greiner Bio-One cat 781956) were seeded with 20,000 primary simian hepatocytes per well, sporozoites were dissected from mosquito salivary glands at day 16 postbloodmeal and allowed to infect cultures.Cultures were fixed on day 8 post-infection and stained for 5mC and 5hmC as described above.An ImageXpress Micro high content imager was used to quantify methyl-cytosine modifications for the entire population of P. cynomolgi liver stage parasites.A 20× objective was used to capture 25 fields of view from each well (covering the entire growth area) of the 384-well plate.Using the associated MetaXpress high content analysis software, the GAPDH stain from each liver stage parasite was used to define parasite objects, and the 5mC or 5hmC staining of host cell nuclei were used to define positive methyl-cytosine modification objects.The 2-dimensional area of intersection of both objects was then quantified for each parasite (fig.S8).

Collection of P. vivax and P. cynomolgi sporozoites for methyl-cytosine characterization
For quantification of methyl-cytosine modification levels by mass spec, sporozoites from 3 different P. vivax cases, numbering 18.7×10 6 from case 1, 10×10 6 from case 2, and 14.7×10 6 from case 3, were dissected from infected An. dirus mosquitoes at IPC as previously described ( 41) and cryopreserved as previously described (54).For quantification of DMNT activity from nuclear extracts, sporozoites from 2 different P. vivax cases, numbering 21×10 6 from case 1 and 20×10 6 from case 2, were similarly dissected and cryopreserved.To serve as a negative control, salivary glands from uninfected mosquitoes at IPC were similarly dissected and cryopreserved.A total of 4.8×10 6 sporozoites for mass spec and 34.1×10 6 sporozoites for DNMT activity assays were also collected from An. dirus mosquitoes infected from feeding on a rhesus macaque infected with P.
cynomolgi M/B strain at ENPRC and cryopreserved as described above.To serve as a negative control, salivary glands and ovaries from uninfected mosquitoes at ENPRC were similarly dissected and cryopreserved.For mapping of methyl-cytosine modifications by bisulfite sequencing, sporozoites from three different P. vivax cases, numbering 9.8×10 6 from case 1, 12.3×10 6 from case 2, and 15.1×10 6 from case 3, were dissected from infected An. dirus mosquitoes at IPC and cryopreserved as described above.A total of 5.3×10 6 sporozoites were also collected from An. dirus mosquitoes infected from feeding on a rhesus macaque infected with P.
cynomolgi M/B strain at ENPRC and cryopreserved as described above.Frozen sporozoites and salivary glands were shipped from IPC and ENPRC to University of California, Riverside on dry ice.
Parasite pellets were lysed with 100 µl lysis buffer (20 mM Tris, pH 8.1, 20 mM EDTA, 400 mM NaCl, 1% SDS and 20 mg/ml proteinase K) and incubated at 55 0 C overnight.Saturated solution of NaCl (0.5× volume of reaction mixture) was subsequently added to the digestion mixture and incubated at 55 °C for another 15 min.The samples were centrifuged at 14500 RCF for 30 min at 4 °C and the supernatant removed to a 1.5-mL microcentrifuge, genomic DNA (gDNA) was then precipitated with 2× volume of 100% chilled ethanol and resuspended in 95 μL water.Samples were then treated with 3 μL of 10 mg/mL RNase A and 2 μL of 25 units/μL RNase T1 and incubated overnight at 37 °C.gDNA was then extracted by chloroform/isoamyl alcohol solution, precipitated again with 100% chilled ethanol, and washed with 70% ethanol.The gDNA pellets were then dissolved in nuclease-free water.One μg of gDNA was enzymatically digested into mononucleosides using nuclease P1 and alkaline phosphatase.Enzymes in the digestion mixture were removed by chloroform extraction.The resulting aqueous layer was dried by using a Speedvac and the dried residues were subsequently reconstituted in doubly distilled water.Approximately 5 ng of the DNA digestion mixture was injected for LC-MS/MS/MS analyses for quantifications of 5mC, 5hmC and dG.An LTQ XL linear ion-trap mass spectrometer equipped with a nano electrospray ionization source and coupled with an EASY-nLC II system (Thermo Fisher Scientific) was used for the LC-MS/MS/MS experiments.The amounts of 5mC, 5hmC and dG (in moles) in the nucleoside mixtures were calculated from area ratios of peaks found in the selected-ion chromatograms for the analytes over their corresponding isotope-labeled standards, the amounts of the labeled standards added (in moles), and the calibration curves.The final levels of 5mC and 5hmC, in terms of percentages of dG, were calculated by comparing the moles of 5mC and 5hmC relative to those of dG.

Extraction of nuclear protein extracts
Cryopreserved sporozoites, or parasites extracted from red blood cells by saponin lysis, were resuspended in 1 ml of cytoplasmic lysis buffer (20mM HEPES pH 7.9, 10 mM KCl, 1mM EDTA, 1mM EGTA, 1mM dithiothreitol (DTT), 0.5 mM AEBSF, 0.65% Igepal, 1× Roche complete protease inhibitor cocktail) and incubated for 10 min on ice.Nuclei were separated from cytoplasmic fraction by 10 min of centrifugation at 1500 RCF followed by two washes with cytoplasmic lysis buffer and one time wash with ice cold 1x PBS.Nuclei pellets were resuspended in 100 µl of nuclei lysis buffer (20 mM HEPES pH 7.9, 0.1 M NaCl, 1mM EDTA, 1 mM EGTA, 1mM DTT, 25 % glycerol, 0.5 mM AEBSF, 1× Roche complete protease inhibitor cocktail) for 20 min at 4 °C with rotation.Nuclear extracts were cleared by 10 min of centrifugation at 6000 RCF.Protein concentration of nuclear extract was quantified by BCA assay and DNMT assays were performed immediately after estimation of protein concentration.

DNA methyltransferase assay
DNA methyltransferase (DNMT) activity of nuclear extracts from P. cynomolgi sporozoites, P. vivax sporozoites, and uninfected mosquito salivary glands were measured using the Epiquik DNMT activity/inhibition assay ultra-kit (cat P-3010) following the manufacturer's instructions.Purified bacterial DNMT enzyme was used as a positive control.A blank control was used to subtract the residual background values.Each reaction was performed in duplicate.DNMT activity was measured in relative unit fluorescence per h per mg of protein for 10 min at 1 min intervals.
Bisulfite conversion and library preparation P. cynomolgi and P. vivax sporozoites were lysed using 100 µl of lysis buffer containing 20mM Tris (pH8.1),20 mM EDTA, 400 mM NaCl, 1 % SDS (w/v) for 30 min at room temperature followed by addition of 20 µl of proteinase K (20 mg/ml) to the pellet and incubated at 55 °C overnight.The gDNA mixture was purified with phenol-chloroform followed by chloroform.Precipitation of gDNA was performed using chilled ethanol and treated with RNase A followed by another round of ethanol precipitation.50 ng of unmethylated lambda DNA was added as a control to each sample before bisulfite conversion of the DNA.500 ng of gDNA of each sample was used for the bisulfite conversion following the manufacturer's instructions (Epitect fast bisulfite conversion kit, Qiagen cat59824).Libraries from bisulfite converted DNA were prepared using the Accel-NGS methyl-Seq DNA library kit (Swift biosciences cat 30024).Libraries were generated following the manufacturer's instructions and DNA was cleaned through SPRI select beads (Beckman Coulter).Libraries were sequenced using the NOVASeq platform.

DNA methylation analysis
Four sets of reads for P. vivax and P. cynomolgi were analyzed.Read qualities were checked with FastQC v0.11.8.FastQC indicated the presence of adapter contamination and overrepresented kmers.As a result, (1) the first 9-14 base pairs were trimmed and (2) reads with overrepresented kmers were discarded (see fig.S9 for summary statistics after the cleaning step).Reads were mapped against the corresponding reference genomes downloaded from PlasmoDB (namely, PlasmoDB-48_Pfalciparum3D7, PlasmoDB-48_PcynomolgiB and PlasmoDB-48_PvivaxP01) using Bismark v0.22.2 with default parameters.To determine the bisulfite conversion rate, reads were also mapped against the lambda phage (see fig.S9 for the conversion rate).Alignment files for the replicates were merged together using Samtools v1.9.Read methylation levels were obtained using Bismark v0.22.2 with default parameters (see fig.S9).
A cytosine in the genome was considered methylated if (1) the number of reads covering that cytosine was higher than a given threshold (10 for P. falciparum, 5 for P. vivax and 3 for P. cynomolgi) and (2) the ratio of methylated reads over all reads covering a cytosine was higher than a given threshold (we chose 0.1 for this second threshold).Genome-wide cytosine density and methylated cytosine density in Fig. 3A and B were calculated in 1 kbp non-overlapping sliding windows using a custom script (available at https://github.com/salehsereshki/pyMalaria).The distribution of CG, CHG, CHH methylation in Fig. 3C and D were obtained by computing the number of methylated cytosines in each context over all the methylated cytosines.For the methylation analyses in genes in Fig. 3E, (1) 500 bp flanking regions and gene body were split into five bins, (2) methylation levels were averaged across all the genes using a custom script (available at the https://github.com/salehsereshki/pyMalaria).To study the correlation between cytosine methylation and gene expression, the same gene body computation was done for the 10% high and low expressed genes using a previously reported P. vivax transcriptome (20).These plots are represented in Fig. 3F.
Additional ReFRAME hit confirmation using an improved P. vivax liver stage assay Twelve hits were re-confirmed using the 12-day radical cure assay, implementing three assay improvements (41).First, 100 μM 1-aminobenzotriazole (ABT, Caymen Chem cat 15252) was added to media on treatment days to reduce hepatic metabolism.Second, the assay endpoint was extended 4 days to allow for nonviable liver stage forms to be cleared from cultures and therefore not be quantified during high-content imaging.Third, nigericin replaced monensin as the positive ionophore control.Confirmation was performed with one independent experiment for all compounds except cadralazine, which was confirmed in four independent experiments.

Epigenetic inhibitor library screen against P. vivax liver stages
The Epigenetic Inhibitor library (Targetmol, cat L1200), containing 773 compounds at 10 mM, was purchased and re-plated in pintool-ready 384-well source plates with 200 μM nigericin and DMSO control wells.The library was screened using the 12-day radical cure assay noted above.
The 24 hits exhibiting the highest inhibition against hypnozoites were replated in a dose-response for confirmation of activity in a 12-day radical cure assay as described above.Confirmation was performed with two independent experiments.

Assessment of effect of ABT on hepatic cytochrome P450 3A4 activity
Two experiments were performed, one on uninduced PHHs, and another on rifampicin-induced PHHs (BioIVT, lot BGW).Cells were thawed and 18,000 live cells/well were seeded into collagen-coated 384-well plates as described above.Media was exchanged every-other-day until day 7 post-seed when media exchange included a dilution series of ABT.One hour after addition of ABT, cytochrome P450 3A4 activity (CYP3A4) was measured using a luciferin-IPA kit (Promega cat V9001) following the lytic protocol with 3 μM IPA.Lysed well contents were transferred to a white 384-well luminometer plate (Greiner Bio-One cat 201106) before reading on a Spectramax i3X (Molecular Devices) with a 1 sec integration time.In the second experiment, cells were similarly seeded and cultured before addition of 25 μM rifampicin (MP Biomedial cat BP2679-250), or an equivalent v/v DMSO vehicle control, in media on days 4 and 6.At day 7 post-seed, CYP3A4 activity was measured following addition of ABT as above.The fold change was calculated between induced and uninduced wells at each ABT dilution point.

Pharmacokinetics of cadralazine in rhesus macaques
Because cadralazine was found substantially more potent against hypnozoites than hydralazine (Fig. 1B), it was selected for a rhesus macaque pharmacokinetic study in which plasma levels were measured over 24 h following an oral dose of 1 mg/kg, which was calculated to be well-tolerated, and 30 mg/kg, which was calculated to likely cause drug-induced hypotension.Plasma exposure following the 30 mg/kg dose was found to cover for several hours the in vitro EC90, and without noticeable side effects, indicating further studies were warranted to assess in vivo efficacy (fig.S16).
Immunofluorescent staining of 5mC and 5hmC in P. vivax blood stage parasites An immunofluorescent staining approach has been used to detect both 5mC and 5hmC in P. falciparum blood stage parasites (18), thus we sought to confirm these marks in P. vivax blood stages.P. vivax blood samples were collected between 2017 and 2019 by active and passive case detection from individuals residing in Mondolkiri, Eastern Cambodia.The presence of P. vivax was determined using an RDT (CareStartTM Malaria Pf/pan RDTs, Accesbio) or microscopy and monoinfections were confirmed by qPCR using species-specific primers (55).Venous blood used was collected in lithium heparin tubes and immediately processed on-site in a mobile laboratory.Leukocytes were depleted using NWF filters (56).The leukocyte depleted parasitized red blood cells were cryopreserved using glycerolyte 57 solution (Baxter) and immediately stored in liquid nitrogen (57).Blood isolates were thawed by addition of 12%, then 1.6%, and then 0.9% (w/v) NaCl solution followed by heparin treatment for 10 min at 37 °C.Blood stage parasites were then purified from thawed isolates using a KCl-Percoll density gradient (58) followed by a wash with RPMI and two washes with 1× PBS.Parasites were then fixed with 3% (v/v) paraformaldehyde and 0.01% (v/v) glutaraldehyde in 1× PBS for 1 h at 4 °C.After fixation, blood stage parasites were permeabilized, denatured, neutralized, quenched, and blocked as described above.Staining for 5mC and 5hmC was carried out as described above except the primary and secondary antibodies were diluted to 1 μg/mL instead of 10 μg/mL.Parasites were stained with 10 μg/mL Hoechst 33342 for 30 min at room temperature and then washed twice with 1× PBS after staining.Parasites were mounted on a coverslip and imaged with a 100× objective on a Leica DM250.
While we did detect 5mC and 5hmC methylation in residual human white blood cells, we could not confirm positive 5mC or 5hmC staining in P. vivax blood stage parasites from these isolates (fig.S17).These negative results could be due to one or more factors.First, while P. falciparum blood stage cultures can reach parasitemias above 10%, P. vivax blood stages cannot be propagated in vitro and the parasitemia of isolates is typically just above the level of detection.Second, P. vivax blood stage isolates were cryopreserved before staining and the stability of DNA methylation after cryopreservation is unknown.Third, the hydrochloric acid treatment needed to denature chromatin during the stain protocol causes red cells to aggregate, thereby making finding and imaging P. vivax blood stages difficult.cynomolgi liver forms assayed at NITD in primary simian hepatocyte (PSH) lots NDO, NPI, XXJ infected with one batch of P. cynomolgi sporozoites.Cytotoxicity (pCC50) against PSH was measured using nuclei counts.Maduramicin is a positive control with activity against P. cynomolgi hypnozoites (A,B) Heat maps represent red as more potent and blue as inactive at highest dose tested.

Fig. 1 .
Fig. 1.Hypnozonticidal hit detection and confirmation.(A) Index chart depicting the primary screen of the ReFRAME library against P. vivax hypnozoites in an 8-day assay.Hypnozoite counts were normalized by mean quantity per well for each plate (Z factor).Teal: library, black: DMSO, red: 1 μM monensin.(B) Primary screen hits were confirmed by dose-response in 8-day P. vivax liver stage assays and counterscreened against P. berghei liver schizonts, P. falciparum asexual blood stages (strain Dd2), HEK293T, and HepG2.Values represent pEC50 or pCC50 ± SD of all independent experiments (n=2-6) for which a pEC50 or pCC50 was obtained.(C) Doseresponse curves for cadralazine against P. vivax and P. cynomolgi liver forms in 8-day assays at the IPC, UGA, and NITD.(B,C) Heat maps represent red as more potent and blue as inactive at highest dose tested.Asterisk (*) indicates only one independent experiment resulted in a calculated pEC50 or pCC50.(C) All replicate wells were plotted together from all independent experiments (n=3 for P. vivax at IPC, n=1 for P. vivax at NITD, n=2 for P. cynomolgi at UGA, and n=4 for P. cynomolgi at NITD), bars represent SEM.

Fig. 3 .
Fig. 3. Density of cytosine and methylated cytosine (5mC) in sporozoites.(A) CG content of chromosome 1 for P. vivax and P. cynomolgi.The total number of cytosines was quantified on each strand using 1 kbp long non-overlapping windows.(B) The total number of methylated cytosines was quantified on each strand using 1 kbp long non-overlapping windows.(C) The number of 5mC present in all possible contexts (CG, CHG, and CHH) quantified throughout the genome of P. vivax and P. cynomolgi.(D) Repartitioned 5mC quantity within different compartments of the genome in P. vivax and P. cynomolgi.(E) Strand-specificity of 5mC for all genes in P. vivax and P. cynomolgi.Flanking regions and gene bodies were divided into five bins and the methylation level of each bin was averaged among all genes.Red: template strand, blue:

Fig. S2 .
Fig. S2.Dose-response confirmation and counterscreens of primary screen hits and analogs.(A) Primary screen hits and structurally-or mechanistically-related compounds were tested by dose-response in 8-day P. vivax liver stage assays at Institute Pasteur of Cambodia and counterscreened against P. berghei liver schizonts, P. falciparum asexual blood stages, P. cynomolgi asexual blood stages, HEK293T, and HepG2.Values represent pEC50 or pCC50 ± SD of all independent experiments (n=2-6) for which a pEC50 or pCC50 was obtained.An asterisk (*) indicates only one independent experiment resulted in a calculated pEC50 or pCC50.(B) Structures of hits which confirmed to be active against P. vivax hypnozoites in dose-response

Fig. S3 .
Fig. S3.Select ReFRAME hits confirmed at Novartis Institute for Tropical Diseases (NITD).(A) Dose-response curves for hydralazine and poziotinib against P. vivax liver forms assayed at NITD.All replicate wells were plotted together from a single independent experiment, bars represent SEM.(B) Potency data (pEC50) for ReFRAME hits against P.

Fig. S4 .
Fig. S4.Cytosine modifications in P. vivax liver forms, full panels from Case 1 (also shown in Fig. 2).(A) Immunofluorescent imaging of a 5mC-positive P. vivax hypnozoite (top) and schizont (bottom) at day 6 post-infection.(B) Immunofluorescent imaging of a 5hmC-negative P. vivax hypnozoite (top) and schizont (bottom) at day 7 post-infection.Yellow arrows indicate autofluorescence in the blue channel associated with cell debris above the hepatocyte monolayer.White arrows indicate hepatocyte nuclei which are dimly-stained with Hoechst 33342 and positive for 5mC or 5hmC.Purple arrows indicate 5mC-positive foci within the parasite.Bars represent 20 µm.

Fig. S7 .
Fig. S7.High-content analysis of cytosine modifications and P. vivax liver stage population metrics.(A) Masks used to quantify parasite area and 5mC or 5hmC signal, i) raw image taken with a 20x objective, ii) mask for P. vivax liver stages, iii) mask for 5mC or 5hmC signal, iv) intersection of parasite mask (light blue) and 5mC or 5hmC signal mask (yellow), leading to quantified area of signal per form.(B) Histogram of growth area all parasites quantified for Case 1, 2, and 3 in Fig. 2. Hypnozoites were classified as forms with an area 125µm 2 and smaller.

Fig. S8 .
Fig. S8.Cytosine modifications in P. cynomolgi liver forms.(A) Immunofluorescent imaging of a 5mC-positive P. cynomolgi hypnozoite (top) and schizont (bottom) at day 8 post-infection.(B) Immunofluorescent imaging of a 5hmC-negative P. cynomolgi hypnozoite (top) and schizont (bottom) at day 8 post-infection.Yellow arrows indicate autofluorescence in the blue channel associated with cell debris above the hepatocyte monolayer.White arrows indicate hepatocyte nuclei which are dimly-stained with Hoechst 33342 and positive for 5mC or 5hmC.Purple arrows indicate 5mC-positive foci within the parasite.Bars represent 20 µm.(C) High-content quantification of 5mC or 5hmC stain area within hypnozoites or schizonts.Significance determined using Kurskal-Wallis tests for hypnozoites and schizonts, with Dunn's multiple comparisons, ****p <.0001, ns = not significant.Line, box and whiskers represent median, upper and lower quartiles, and minimum-to-maximum values, respectively, of all hypnozoites (124 ≤ n ≤ 240) or all schizonts (7 ≤ n ≤ 26) in culture, 2' indicates a secondary stain only control.

Fig. S9 .
Fig. S9.Measurement of DNA methylation and DNA methyltransferase (DNMT) in P. vivax and P. cynomolgi sporozoites.(A) Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of 5mC or 5hmC from enzymatically-digested gDNA from P. vivax sporozoites, P. cynomolgi sporozoites, and P. falciparum blood stage parasites, as well as negative controls including uninfected mosquito salivary glands and ovaries from the same colony of mosquitoes used to generate the respective sporozoites.Bars represent SD of two independent experiments.(B) DNMT activity measured from nuclear extracts of P. vivax sporozoites, P. cynomolgi sporozoites, and uninfected mosquito salivary glands using the Epiquick DNMT activity assay.Data are from a single experiment.(C) Summary statistics of

Fig. S10 .
Fig. S10.Cytosine and methylation density plots for P. vivax sporozoites.(A) CG content of chromosome 1 to 14 (Chr 1-14).The total number of cytosines quantified on each strand using 1 kbp long non-overlapping windows.(B) The total number of methylated cytosines quantified on each strand using 1 kbp long non-overlapping windows.

Fig. S11 .
Fig. S11.Cytosine and methylation density plots for P. cynomolgi sporozoites.(A) CG content of chromosome 1 to 14 (Chr 1-14).The total number of cytosines quantified on each strand using 1 kbp long non-overlapping windows.(B) The total number of methylated cytosines quantified on each strand using 1 kbp long non-overlapping windows.

Fig. S13 .
Fig. S13.Characterization of primary human hepatocyte (PHH) metabolism following 1aminobenzotriazole (ABT) treatment.(A) PHH lot BGW were seeded into 384-well plates and cultured for 7 days before addition of a dilution series of ABT in media.Cytochrome P450 3A4 activity (CYP3A4) was measured using luciferin-IPA (Promega).RLU: relative luminescence units.Bars represent SD of quadruplicate wells.Data are representative of two independent experiments.(B) PHH lot BGW were cultured in 384-well plates before addition of 25 μM rifampicin in media on days 4 and 6 to induce CYP3A4 expression.At day 7 post-seed, CYP3A4 activity was measured by adding luciferin-IPA and a dilution series of ABT in media.Fold change was calculated based on matching uninduced controls.Data are from one independent experiment.(C) PHH lot BGW were seeded in 384-well plates and cultured for 7 days before treatment with 100 μM ABT for 1h, followed by addition of substrates for 1 h and collection for analysis by mass spectrometry.Data are combined from two independent experiments, bars represent SD of all replicates.Significance determined by student's t tests, ****p <.0001, ***p <.001, **p <.01, ns, not significant.

Fig. S15 .
Fig. S15.Epigenetic Inhibitor library screen and hits.(A) Index chart of an epigenetic inhibitor library screened against P. vivax hypnozoites in a v3 (12-day + ABT) assay.Teal circle: library, black square: DMSO, pink triangle: 200 nM nigericin.(B) Structures of epigenetic inhibitor library hits which were confirmed to be active against P. vivax hypnozoites in dose-

Fig. S16 .
Fig. S16.Pharmacokinetics of cadralazine in nonhuman primates.Mean plasma concentration of cadralazine was measured in three males rhesus macaques after oral dosing.Plasma was collected following a 1 mg/kg dose, and again following a 30 mg/kg dose.Bars represent SD.The approximate IC50 and IC90 from P. vivax hypnozoite assays are indicated.

Fig. S17 .
Fig. S17.Cytosine modification in P. vivax blood stages.(A) P. vivax blood stages from patient isolates appeared negative when stained with 5mC.A white blood cell positive for 5mC serves as a stain control.(B) P. vivax blood stages from patient isolates appeared negative when stained with 5hmC.A white blood cell positive for 5hmC serves as a stain control.Bars represent 10 µm.

Table 1
Source plates were made from the master library at Calibr at Scripps Research such that 3-5 μL of 10 mM solution was added to each well of a sterile, conical-bottom 384-well plate (Greiner Bio-One cat 784261).Most compounds were diluted in DMSO, however, a subset was diluted in water due to limited DMSO solubility.Plates were sealed and shipped on dry ice to SMRU and IPC and stored at -20 °C prior to use.Column 24 of each plate was filled with 5 μL DMSO to serve as negative control wells.Control compounds included 1 mM monensin (positive control for hypnozoite and schizont activity), 1 (39)3)irmed hits from an epigenetic inhibitor library screened against P. vivax liver stages.Mean pEC50 or pCC50 and SD from two independent experiments.HDAC: histone deacetylase.CREB: cAMP response element-binding protein.FLT3: fms-like tyrosine kinase 3. 5 SYK: spleen tyrosine kinase.JAK: Janus kinase.SETD: SET domain containing histone lysine methyltransferase.mM the phosphatidylinositol 4-kinase inhibitor (PI4Ki) KDU691 or MMV390048 (positive control for schizont activity), 1 mM atovaquone (negative control for radical cure activity) and 10 mM tafenoquine (clinically relevant control for hypnozoite activity)(7,13).Ethical Approval for Human Subjects and Animal UseThe Thai human subjects protocols for this study were approved by the Institutional Ethics Committee of the Thai Ministry of Public Health and the Oxford Tropical Medicine Ethical Committee (TMEC 14-016 and OxTREC 40-14).The Cambodian human subjects protocols for this study were approved by the Cambodian National Ethics Committee for Health Research (100NECHR, 104NHECR, 111NECHR, 113NHECR and 237NHECR).Protocols conformed to the Helsinki Declaration on Ethical Principles for Medical Research Involving Human Subjects(39)and informed written consent was obtained for all volunteers or legal guardians.P. cynomolgi sporozoites were generated at Emory National Primate Research Center (ENPRC) using procedures approved by the Emory University Institutional Animal Care and Use Committee (PROTO201900110), as well as at UGA using procedures approved by UGA's Institutional Animal Care and Use Committee (A2020 03-002-Y3-A15).P. cynomolgi sporozoites were also produced at the Armed Forces Research Institute of Medical Science under an IACUC-approved animal use protocol in an AAALAC International -accredited facility with a Public Health Services Animal Welfare Assurance and in compliance with the Animal Welfare Act and other federal statutes and regulations relating to laboratory animals.P. berghei sporozoites were generated by the Sporocore at UGA using procedures approved by UGA's Institutional Animal Care and Use Committee (A2016 06-010-Y1-A0 and A2020 01-013-Y2-