In silico prediction of AP2-EXP competitors

To identify competitors of DNA binding by an ApiAP2 protein, we used AP2-EXP (PF3D7_1466400), which is the only ApiAP2 protein whose AP2 domain is structurally characterized (PDB accession: 3IGM) [40]. In silico molecular docking was run on AUTODOCK [41] using over ten thousand small molecules from both the Tres Cantos Antimalarial Set (TCAMS) [42] and the Drug Bank [43] against AP2-EXP. Docking hits were prioritized based on the free energy of interaction and proximity to amino acids necessary for DNA contact (S1A Fig). In some cases, the compound that generated a high docking score was not available for purchase, so alternative choices with >0.9 Tanimoto similarity score were identified, computationally docked, and used for further evaluation (All compounds listed in S1B Fig). These docking simulations resulted in a final list of nine top scoring predicted competitors of DNA binding by AP2-EXP, and each compound was assigned an alphabetical identifier A-I (Figs 1A, S2A, and S2B) for use in this study. We determined that all nine compounds kill asexual P. falciparum parasites at micromolar concentrations (range 11.33–198.90μM) in a growth inhibition assay (Figs 1A and S3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Nine putative competitors of DNA binding by AP2-EXP were selected by an in-silico screen and tested in vivo and in vitro.

A) Each compound that was prioritized based on the in-silico screen was assigned an identifier (A-I) and tested for anti-Plasmodium activity in a 48-hour growth inhibition assay against asexual blood stage P. falciparum. B) Each putative AP2-EXP competitor was added to an EMSA containing the AP2-EXP AP2 domain. Activity against AP2-EXP will result in a loss of the shifted DNA probe. DMSO vehicle was used as a control for normal DNA binding by AP2-EXP. 150 fmoles of DNA probe, 125ng of AP2-EXP, and 300μM of each compound were used for each lane. C) Chemical structures of Compounds A, B, C, and I. Three of the four compounds that prevent AP2-EXP DNA binding in vitro (Compounds A, B, and C) have a benzoxazole core moiety, denoted by a red box.

https://doi.org/10.1371/journal.ppat.1010887.g001

Four compounds have activity against AP2-EXP in vitro

The nine predicted competitors were tested to assess their ability to disrupt AP2-EXP DNA binding using a competitive electrophoretic mobility shift assay (EMSA) (S4A Fig, DNA oligonucleotides in S1 Table). Four of the nine compounds (A, B, C and I) (Fig 1B) were able to effectively prevent DNA binding by AP2-EXP in vitro. Three of the four AP2-EXP competitors have a benzoxazole as the core moiety (Compounds A, B, and C), and Compound I is made up of planar rings. However, Compound D does not affect AP2-EXP DNA binding despite having a benzoxazole group, indicating that this is not the sole determinant of activity (Fig 1C). Compounds A, B and C also have the lowest IC₅₀ values measured amongst the original candidates (11.33–18.31μM), while Compound I has the highest recorded IC₅₀ value at 198.90μM (Fig 1A).

To assess whether the observed inhibition of ApiAP2 domain DNA binding was specific to the AP2-EXP AP2 domain or might be pan-ApiAP2, we repeated competitive EMSAs with three different purified P. falciparum AP2 domains: AP2-I Domain 3 (AP2-I D3) [12,18] (S4B Fig), AP2-HS Domain 1 (AP2-HS D1) [12,35] (S4C Fig), and PfSIP2 Domain 1 (PfSIP2 D1) [20] (S4D Fig) (S1 Table). We found that Compounds A, B, C, and I can prevent DNA binding by AP2-I D3 (S5A Fig), Compounds B and I prevent DNA binding by AP2-HS D1 (S5B Fig) and Compounds B, C and I prevent DNA binding by PfSIP2 D1 (S5C Fig).

To assess the relative specificity for ApiAP2 proteins, we performed competitive EMSAs using two non ApiAP2 proteins: the Arabidopsis thaliana AP2/ERF domain protein Ethylene Response Factor 1 [44] (AtERF1) (S4E Fig) and the human High Mobility Group Box (HMGB) domain protein SOX2 (S4F Fig) [45]. Only Compound I was able to compete AtERF1 (S5D Fig). SOX2 was competed by Compound A (S5E Fig). The lack of AtERF1 inhibition is surprising, given the high structural similarity between AP2-EXP and AtERF1 [44] (S5F Fig). SOX2 has no structural similarity to AP2-EXP [46] (S5F Fig). Therefore, Compounds B and C have specificity for ApiAP2 protein domains relative to the two protein domains tested in vitro. Finally, we performed a competitive EMSA against AP2-EXP using antimalarial drugs with known modes of action: chloroquine, artemisinin, and pyrimethamine [2]. None of these drugs affects AP2-EXP DNA binding, providing evidence that the EMSA assay used can differentiate between putative ApiAP2 competitors and antimalarials with different targets (S5G Fig). Pyrimethamine, chloroquine, and artemisinin are diverse in chemical structure (S5H Fig), therefore the assay is not likely to be biased towards a specific class of antimalarial. (EMSA Results Summarized in S2 Table).

Each compound identified in the in-silico screen contains planar rings which could potentially intercalate DNA [47]. To eliminate the possibility that the four ApiAP2 competing compounds interfere non-specifically with DNA binding proteins by intercalating with DNA, we conducted an ethidium bromide exclusion assay [48]. Compared to the positive control major groove intercalator DRAQ5 [48], only Compound I was able to effectively displace ethidium bromide from DNA (S6 Fig). Based on their relative specificity for ApiAP2 proteins compared to Compounds A and I, we focused on Compounds B and C for further study.

Structural analogs of Compound B differ in activity against AP2-EXP in vitro

Our initial competitive EMSA results established that the benzoxazole moiety is overrepresented but not the sole determinant of activity against ApiAP2 proteins (Fig 1B and 1C). Therefore, to elucidate additional factors which influence compound activity in vitro, we tested the four closest available analogs to Compound B from the TCAMS based on similarity score [42] (Designated as Compounds B-1, B-2, B-3 and B-4) against AP2-EXP. Compounds B-1 and B-4 prevent DNA binding by AP2-EXP (Fig 2A) and AP2-I D3 (S7A Fig), while Compounds B-2 and B-3 do not prevent DNA binding by either AP2-EXP or AP2-I D3 (Figs 2A and S7A). Compound B-1 is the least effective against AP2-EXP DNA binding relative to Compounds B and B-4 (Fig 2B). Non-AP2 domain competitor Compounds B-2 and B-3 each have halogen atoms substituted on the benzene ring. Conversely, Compounds B, B-1, and B-4 have methyl or ethyl groups substituted onto the benzene ring and can prevent DNA binding in vitro (Fig 2C). Compound B-1 has both a methyl group and a chlorine atom substitution. If the halogen atom substitution decreases activity against DNA binding, this mix of substitutions is consistent with its lower efficacy against ApiAP2 proteins in vitro relative to Compounds B and B-4. Compounds B, B-1 and B-4 have no difference in DNA intercalation ability measured by ethidium bromide exclusion (S7B Fig). A molecular dynamics simulation of AP2-EXP docked with Compound B and each analogue predicts that Compounds B, B-1, B-2, and B-4 dock stably with AP2-EXP, while non-AP2 competitor Compound B-3 moves away from the protein (S1–S5 Movies). In summary, substituting bulky, electronegative groups onto Compound B can abolish its activity against AP2-EXP in vitro.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Analogues of Compound B prevent AP2-EXP DNA binding with differing efficacy.

A) Compounds B, B-1, B-2, B-3, and B-4 were added to an EMSA with AP2-EXP. Activity against AP2-EXP will result in the loss of the shifted probe. 40 fmoles of DNA probe, 125ng of AP2-EXP, and 300μM of each compound were used for each lane. B) Compounds B, B-1 and B-4 were titrated at 25, 50, and 150μM into an EMSA with AP2-EXP to determine whether there are differences in efficacy against AP2-EXP. DMSO vehicle was used as a control for normal DNA binding. 40 fmoles of DNA probe and 125ng of AP2-EXP were used for each lane. The concentration increase for each compound is indicated by the triangle from left to right. C) Chemical structures of Compounds B, B-1, B-2, B-3, and B-4. Compounds that prevent DNA binding by AP2-EXP are highlighted in black. Each compound has a different substitution pattern on the right-side benzene ring, denoted by a red box. The IC₅₀ against asexual P. falciparum of compounds B-1, B-2, B-3, and B-4 measured by Gamo et al [42] is indicated at the bottom of each identifier. The IC₅₀ of Compound B was determined in this study.

https://doi.org/10.1371/journal.ppat.1010887.g002

Putative ApiAP2 inhibitor Compounds B and C maximally affect P. falciparum trophozoites

To measure the phenotypic defects caused by putative ApiAP2 inhibiting compounds, we determined the parasite killing phenotypes of Compounds B and C against P. falciparum during the IDC. Compound H was used as a control since it does not compete AP2-EXP in vitro (Fig 1B). Each compound was spiked (40μM) into synchronous asexual blood stage P. falciparum parasites at 6 hpi and parasitemia was recorded at 6-hour intervals from 6 to 48 hpi (Fig 3A–3D). Incubation with Compound H caused an observable developmental defect starting at 42 hpi in which fewer parasites entered schizogony (Fig 3D and 3E). In contrast, treatment with either Compound B or C resulted in parasites that similarly failed to progress into mature trophozoites (Fig 3B, 3C and 3E), suggesting that Compounds B and C may share a common mode of action.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Compounds B and C affect P. falciparum in the mid trophozoite stage, coinciding with the maximum expression of AP2-EXP.

A-D) Highly synchronous wild type Pf3D7 parasites at 10% starting parasitemia were spiked with DMSO vehicle control (A), 40μM Compound B (B), 40μM Compound C (C), or 40μM Compound H (D) and parasite staging was recorded at 6-hour intervals throughout the IDC. Each growth assay was performed in technical and biological triplicate. Error bars represent standard deviation of the mean. The color key indicates the asexual blood stage morphologies recorded at each time point. E) Representative Giemsa-stained parasite images corresponding to the growth assay from panels (A-D). F) Highly synchronous asexual blood stage wild type Pf3D7 parasites at 1% starting parasitemia were dosed with 40μM Compound C for 8-hour intervals starting at 6 hpi. Compound C was removed following each dose interval. At 54 hpi, the ratio of reinvaded rings to stalled trophozoites was recorded. DMSO was used as a vehicle control for normal growth. Each assay was performed in biological and technical triplicate. Error bars represent standard deviation of the mean. Significance of the difference between the DMSO control and Compound C dosed values was tested using the Mann Whitney U-test. G) AP2-EXP protein expression in the asexual blood stages ring (R), trophozoite (T), and schizont (S), and Stage III gametocytes (Gam), was probed by western blot against AP2-EXP endogenously tagged with GFP (AP2-EXP::GFP). The expected molecular weight of AP2-EXP::GFP (147kDa) is indicated by an arrow. Wild type Pf3D7 protein was used as a negative control. The full-length western blot is provided in S10B Fig.

https://doi.org/10.1371/journal.ppat.1010887.g003

Due to its commercial availability, we continued with Compound C as the lead compound for phenotypic and molecular characterization. To better determine the timing of the antimalarial action of Compound C, 40μM Compound C was spiked into a highly synchronous asexual blood stage P. falciparum culture for 8-hour intervals and then washed out starting at 6 hpi (S8A Fig). Since the phenotype caused by continuous Compound C exposure manifested as a complete arrest in the early trophozoite stage, resulting in no reinvasion at 48 hpi (Fig 3C and 3E), we used the ratio of reinvaded rings at 54 hpi to trophozoites remaining in the culture at 54 hpi as a metric for Compound C induced growth inhibition. With the exception of 38–46 hpi, each interval of Compound C exposure resulted in a significant decrease in the ring to trophozoite ratio (Fig 3F). The Compound C spike in interval from 22–30 hpi resulted in the lowest ring to trophozoite ratio amongst the intervals tested (0.54, compared to 9.71 in the DMSO control), therefore, the maximum activity of Compound C against asexual blood stage P. falciparum occurs between 22–30 hpi (Figs 3F and S8B).

AP2-EXP expression coincides with the growth phenotype of Compound C

In order to determine whether the observed timing of Compound C activity coincides with the presence of AP2-EXP in vivo, we tracked its expression throughout the IDC. First, we used Selection Linked Integration (SLI) [49] to create a parasite line with endogenously GFP tagged AP2-EXP (AP2-EXP::GFP) (S9A Fig). Using this line, we observed maximum abundance of AP2-EXP::GFP in 30 hpi trophozoites by both western blot and live fluorescence microscopy (Figs 3G, S10A and S10B). AP2-EXP was not detected in maturing Stage III gametocytes (Fig 3G). Maximum protein abundance of AP2-EXP in mid-trophozoites was independently confirmed in a parasite line expressing AP2-EXP tagged with 2XHA and the TetR:DOZI mRNA repression aptamer (AP2-EXP::HA) (Figs S9B and S11). Therefore, the maximum abundance of AP2-EXP at 30 hpi coincides with the timing of action for Compound C.

Compound C disrupts the P. falciparum transcriptome with stage specificity

Since ApiAP2 proteins are predicted to act as transcription factors [14,17,21] we hypothesized that specific changes in transcript abundance should occur when parasite cultures are exposed to Compound C. We cultured parasites in 12μM (0.66xIC₅₀) Compound C or DMSO vehicle and measured RNA abundance at seven time points throughout the asexual blood stage. As a quality control, we checked a set of periodically expressed control genes [50,51] (S12A–S12D Fig) and the total transcriptome (S12E–S12F Fig) of each dataset to monitor the IDC progression. Since 12μM Compound C prevents fewer than 50% of parasites from completing the IDC (S13A and S13B Fig), there is not a fully penetrant perturbation to the representation of all IDC stages upon Compound C incubation. The Spearman Correlation between the total transcriptome of DMSO and Compound C samples deviated most significantly at 30 hpi, (Corr. = -0.16) (Fig 4A) again coinciding with the observed activity of Compound C (Fig 3C, 3E, 3F, Gene Expression values in S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Compound C disrupts the P. falciparum transcriptome specifically at 30 hpi, with bias towards AP2-EXP target genes predicted by ChIP-seq.

A) Highly synchronous wild type Pf3D7 parasites were spiked with 12μM (0.66xIC₅₀) Compound C or DMSO vehicle control at 10 hpi. Total RNA was harvested at 6, 12, 18, 24, 30, 42, and 48 hpi for quantification by DNA microarray. The full transcriptome Spearman Correlation between Compound C and DMSO control spiked parasites was plotted as a correlogram. An asterisk indicates p-value of correlation < 0.05. B) Three replicates of AP2-EXP Chip-seq were collected at 30 hpi using two genetically tagged parasite lines (AP2-EXP::GFP and AP2-EXP::HA). A no epitope control was collected using wild type Pf3D7 parasites and the anti-GFP antibody. Log₂ enrichment of the immunoprecipitate (ChIP) over Input DNA for one replicate of AP2-EXP::GFP, AP2-EXP::HA, and No Epitope Control is plotted relative to the coding sequence start site of AP2-EXP target genes identified in at least 2/3 replicates. CATGCA is the most overrepresented DNA motif within the peaks of AP2-EXP occupancy conserved in at least 2/3 ChIP-seq experiments. C) A volcano plot of the changes in abundance for AP2-EXP target genes at 24–30 hpi in the presence of Compound C. 46/93 detected AP2-EXP target transcripts decrease in abundance by log₂ fold change > 2 (Adjusted p-value cutoff < 0.1). D) GO-term analysis of transcripts that are both AP2-EXP targets and decreased in abundance with respect to Compound C (Bonferroni p-value cutoff < 0.05).

https://doi.org/10.1371/journal.ppat.1010887.g004

Next, we quantified differences in transcript abundance that are unique to the progression from 24 to 30 hpi for Compound C vs. DMSO vehicle control populations using Linear Models for Microarray Data (LIMMA [52]) (Differentially Expressed genes in S3 Table). Overall, 463 RNA transcripts are decreased in abundance by log₂ fold change >2 between 24–30 hpi in the presence of Compound C. Gene ontology (GO) analysis [53] (Full GO Terms in S3 Table) revealed an enrichment for genes that encode proteins important for the parasite to invade or modify red blood cells among decreased transcripts (e.g., RON3, GAP45, RhopH2, RhopH3, MSP1, MSP6). We independently determined differential transcript abundance over the entire time course using the time course specific software Rnits [54] and found high overlap with our LIMMA analysis (Overlaps in S3 Table). Therefore, the maximal perturbation of the transcriptome (Fig 4A), action of Compound C (Fig 3C, 3E and 3F), and maximal abundance of AP2-EXP (Fig 3G) all occur at roughly 30 hpi in the IDC.

AP2-EXP gene targets correlate with differentially abundant transcripts

We used ChIP-seq to determine the genome wide binding occupancy of AP2-EXP. A total of three samples from highly synchronous trophozoites were collected at 30 hpi using both the AP2-EXP::GFP (2 replicates) and AP2-EXP::HA (1 replicate) parasite lines (Peaks in S4 Table). As a quality control, we determined that ChIP recovers intact AP2-EXP (S14A Fig), and a co-immunoprecipitation blot confirmed that AP2-EXP interacts with Histone H3 in the nucleus (S14B Fig). In aggregate, AP2-EXP binds 240 genomic loci, corresponding to 101 total target genes (Genes in S4 Table). Enrichment of AP2-EXP was well conserved between the GFP and HA tagged parasite lines, as indicated by nearly identical coverage in a metagene plot of AP2-EXP target genes (Fig 4B). A ‘no epitope control’ ChIP done on a wild type Pf3D7 parasite line using the anti-GFP antibody detected only 2 peaks, neither of which overlapped with those of AP2-EXP::GFP or AP2-EXP::HA (Fig 4B and S4 Table). AP2-EXP peaks were highly enriched for the known AP2-EXP and PbAP2-Sp DNA motif CATGCA [17,55] (Figs 4B and S15A–S15C).

Out of the 463 transcripts decreased in abundance by log₂ fold change > 2 in the presence of Compound C, 46 are AP2-EXP targets (S3 Table). This represents 50% (46/93) of AP2-EXP target genes detected in the RNA time course (Figs 4C and S16). We found that 7/14 total transcripts decreased in abundance by log₂ fold change > 5, and 16/33 decreased in abundance by log₂ fold change > 4, are AP2-EXP targets (Fig 4C and S3 Table). AP2-EXP targets that are dysregulated at 24–30 hpi have functions related to red blood cell invasion and host cell remodeling (Fig 4D and S3 Table). To assess whether AP2-EXP DNA binding is impacted by Compound C in vivo, parasite cultures were spiked with either 40μM Compound C or DMSO vehicle control at 30 hpi and AP2-EXP occupancy was measured by ChIP-quantitative PCR (ChIP-qPCR). On average, AP2-EXP occupancy was found to decrease at five genomic loci (gap45, sip2, ron3, ralp, and ama1) in the presence of Compound C compared to non AP2-EXP bound control loci, although these results were not statistically significant (S17 Fig and S1 Table).

We then compared AP2-EXP genomic occupancy with several published datasets in order to further evaluate its function as a sequence specific transcription factor. AP2-EXP occupancy is correlated with a nucleosome depleted region (S18A and S18B Fig) [56] and activating chromatin marks [57,58] (S19A and S19B Fig), and is proximal to target gene transcription start sites (TSS) [59] (S20 Fig). The majority of AP2-EXP target genes increase in abundance starting at 32 hpi (S21 Fig) [60]. AP2-EXP target gene function is enriched for invasion and red blood cell modification, (S3 Table) as was found for the genes that decrease in abundance in the presence of Compound C (Fig 4D and S3 Table). While additional targets of Compound C cannot be ruled out, in aggregate, our findings suggest that Compound C inhibits the function of AP2-EXP as a sequence specific transcription factor during P. falciparum asexual blood stage development.

ApiAP2 competitor Compounds B and C are active against mosquito stage Plasmodium parasites

Elucidation of the role of essential P. falciparum ApiAP2 proteins in the IDC is challenging due to the inability to knock out asexual blood stage essential genes by genomic deletion. We attempted to disrupt AP2-EXP protein abundance with the conditional knockdown approaches Knock Sideways [49], TetR:DOZI mRNA repression [61], and glms ribozyme mediated cleavage [62]. Each of these genetic systems failed to mediate protein knockdown or mislocalization (S9A, S9B, S22 and S23A–S23C Figs and S1 Table).

To account for this limitation, we tested the ApiAP2 competitor Compounds B and C against rodent infectious P. berghei parasites due to the well-characterized genetic phenotype of the AP2-EXP orthologue PbAP2-Sp [30,55] as a master regulator of sporogenesis. PbAP2-Sp and AP2-EXP are highly similar at amino acid level, and the DNA contacting amino acids identified via crystallographic characterization of AP2-EXP [40] are conserved between AP2-EXP and PbAP2-Sp (S24 Fig). However, to ensure that PbAP2-Sp DNA binding is similarly inhibited in a competitive EMSA, we first tested the ability of Compounds B, C, and F to prevent PbAP2-Sp from binding DNA. As with AP2-EXP, Compounds B and C block PbAP2-Sp DNA binding in vitro, while Compound F had no effect (Fig 5A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Compounds B and C are active against mosquito stage P. berghei parasites.

A) Compounds B, C and F were added to an EMSA with PbAP2-Sp. Activity against PbAP2-Sp will result in the loss of the shifted probe. 150 fmoles of DNA probe, 125ng of PbAP2-Sp, and 300μM of each compound were used for each lane. DMSO vehicle was used as a control for normal DNA binding by PbAP2-Sp. B) Midgut oocyst counts per mosquito for P. berghei infected mosquitoes on day 14 post infection following compound injection on day 10 post infection. Compound identity is indicated in the legend. Rep. 1 and Rep. 2 correspond to biological replicates one and two, respectively. Significance of difference between DMSO vehicle injected samples and compound injected samples was assessed using the Mann Whitney U-test. Bars indicate which compound injected dataset has been compared to the DMSO control dataset. Error bars represent the standard deviation of the mean. C) Midgut sporozoite counts per five infected mosquitoes on day 14 post infection following compound injection on day 10 post infection. Compound identity is indicated in the legend. Rep. 1 and Rep. 2 correspond to biological replicates one and two, respectively. Significance of difference between DMSO vehicle injected samples and compound injected samples was assessed using the Mann Whitney U-test. Bars indicate which compound injected dataset has been compared to the DMSO control dataset. Error bars represent the standard deviation of the mean.

https://doi.org/10.1371/journal.ppat.1010887.g005

Since these compounds compete PbAP2-Sp in vitro, we injected Compound B or C into the midgut of Anopheles mosquitoes infected with P. berghei parasites to determine their effect on Plasmodium mosquito stage development (S25 Fig). Compound F was used as a non-AP2 domain inhibiting control. Mosquitoes injected with Compounds C or F developed comparable oocyst numbers to the DMSO control (Fig 5B) while Compound B prevented the development of midgut oocysts entirely (Fig 5B). Compound C treated mosquitoes developed 10-fold fewer midgut sporozoites than the control (Fig 5C). Therefore, consistent with their in vitro activity against ApiAP2 proteins, Compounds B and C are strong inhibitors of P. berghei sporozoite development in vivo (parasite counts in S5 Table).

Ethics statement

The human red blood cell samples were sourced ethically and their research use was in accord with the terms of the informed consents under an IRB/EC approved protocol (Protocol title: Prospective Collection of Biological Specimens from Subjects Presenting at Specimen Donation Centers for Research, IRB Tracking Number: 20161665, Sponsor: SERATRIALS, LLC). Written consent was obtained from the participants.

All in vivo studies using laboratory mice were conducted in accordance with the GSK Policy on the Care, Welfare and Treatment of Laboratory Animals and were reviewed by the ethical review process at the institution where the work was performed. Mice were anesthetized using intraperitoneal injection of 150μI of the following mixture: {2 ml of 100 mg/ml Ketaject (Phoenix Pharmaceutical) + 1 ml of 10 mg/ml ACE Promazine (Henry Schein) + 7 ml saline}. Toe pinching was used to check that the mice were appropriately anesthetized. Protocols were approved by the Johns Hopkins University Animal Care and Use Committee (ACUC, Protocol Number: M018H18).

Parasite lines

Parasites were grown at 37°C, 5% O2, 7% CO2 using RPMI1640 media supplemented with hypoxanthine and 0.5% Albumax II (Thermo). The parasite lines used in this study were AP2-EXP::GFP, AP2-EXP::HA, AP2-EXP::glms, and wild type Pf3D7 (Malaria Research and Reference Reagents Repository). The human biological samples were sourced ethically and their research use was in accord with the terms of the informed consents under an IRB/EC approved protocol (Protocol title: Prospective Collection of Biological Specimens from Subjects Presenting at Specimen Donation Centers for Research, IRB Tracking Number: 20161665, Sponsor: SERATRIALS, LLC). Written consent was obtained from the participants.

To create AP2-EXP::GFP the C-Terminal coding region of AP2-EXP was cloned (S1 Table) into the plasmid pSLI::2xFKBP [49] for endogenous tagging by single homologous recombination (S9A Fig). Transgenic parasites with the correct C-Terminal tag were further transfected with pLyn-FRB-mCherry [49] for inducible mislocalization using the same method. Parasites were maintained in media with 2.5nM WR99210. All ChIP and western blot experiments using AP2-EXP::GFP except S23C Fig were collected using AP2-EXP::GFP without the pLyn mislocalization plasmid.

To create AP2-EXP::HA we used the pSN054 vector system [65]. The right homology region (RHR) was amplified by PCR (S1 Table), and the recodonized left homology region (LHR) was synthesized. Synthesized single guide RNA fragments (S1 Table) were cloned into the linearized pSN054 donor vector [65] The parental parasite line used for transfection expresses Cas9 and T7 RNA polymerase [66] (S9B Fig). Cell cultures were maintained in 500nM anhydrotetracycline (aTc, Sigma-Aldrich 37919) and 2.5 mg/mL of Blasticidin S.

To create AP2-EXP::glms::HA, the C-terminal coding region of AP2-EXP (S1 Table) was cloned into the plasmid pSLI::3xHA::glms [67] (S22 Fig). Parasite cultures were maintained in media with 2.5nM WR99210.

All P. falciparum transfections were performed as described [49,68]. All parasite strains were cloned by limiting dilution and genotyped using PCR. AP2-EXP::GFP and AP2-EXP::HA were also genotyped using whole genome sequencing (NCBI SRA: PRJNA818769: www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA818769)

Genomic DNA isolation

Parasite cultures were lysed with 0.1% Saponin in 1xPBS and collected by centrifugation at 1500 RPM, then resuspended in 1xPBS. Genomic DNA was then isolated using the Qiagen DNeasy nucleic acid isolation kit according to the manufacturer’s instructions.

AP2-EXP knockdown assays

For AP2-EXP::GFP + pLyn, synchronous parasites were split into two populations. 250nM Rapalog was added to one parasite group for 48 hours as described [49] while the second group was used as a control. After 48 hours parasite protein was collected in nuclear and cytosolic fractions and AP2-EXP::GFP, Histone H3, and Pf Aldolase were detected by western blot.

For AP2-EXP::HA, parasites were grown routinely in the presence of 500nM aTc. aTc was washed out of one group, while the second group was maintained with aTc as a control for 120 hours before protein harvest. AP2-EXP::HA and Histone H3 were detected by western blot.

For AP2-EXP::glms::HA, synchronous parasites were split into two populations. 5mM glucosamine was added to one parasite group for 72 hours as described [62] while the second group was used as a control. After 72 hours parasite protein was harvested for detection of AP2-EXP::glms::HA and Histone H3 by western blot.

In silico docking screen

The Tres Cantos Antimalarial Set of 13533 small molecules was retrieved from the supplementary material of Gamo et al [42] and downloaded in Spatial Data File (SDF) format from BatchEntrez. 4603 small molecule structures were downloaded from DrugBank [43] version 2.5 in SDF format. Using the SDF library, a 3-dimensional MOL2 library was created with the use of BALLOON and the Merck Molecular Force Field. BALLOON was used to calculate up to 20 conformations using a genetic algorithm and to select the conformation with the lowest energy of conformation. The 3-dimensional MOL2 library was then prepared for AutoDock using the python script "ligand_prepare4.py" from AutoDockTools [69]. The prepared ligands were then saved in PDBQT file format. The python script "ligand_prepare4.py" automatically defines the torsions of the molecule.

The homodimerized AP2-EXP AP2 domain was downloaded from the solved crystal structure (PDB ID 3IGM). AutoDockTools was used to remove water molecules, add hydrogens, and calculate charges. Residues Arg-88 and Asn-118 were found to be flexible using FlexPred and set as flexible residues for docking. The flexible residues were saved as flexible residues in a flexible residue file and the remaining molecule was saved as rigid residues in a rigid residue file as specified in the AutoDock manual [69]. The rigid residues file was used to pre-calculate an energy grid of the macromolecule using autogrid4 [69]. A set of 14 grid map types S, Cl, F,A, Br, N, P, OA, SA, C, I, HD, Ca, NA were calculated for the macromolecule using AutoGrid [69].

From the prepared ligands a docking set was prepared for AutoDock with the macromolecule and pre-calculated energy grids. For each docking set a docking parameter file was prepared, with the Lamarckian Genetic Algorithm (LGA) specified as the search algorithm for AutoDock. To allow for high throughput parallel processing PERL was used on top of AutoDock to manage the dockings. The algorithm was initially set to run with a maximum of 25000 energy evaluations and 20 repeats. Since the calculations are computationally intensive, this setting was used as initial screening of the ligands, to identify a subset of ligands for a more thorough evaluation [69]. The 1000 best hits from the GSK compound evaluation and the DrugBank compound evaluation were selected for a more thorough evaluation. The docking parameters were reconfigured to 250000 energy evaluations and 100 repeats and docking repeated. The docking results were examined with PERL, to provide an automated approach to the interpretation of the docking results. The docking results were evaluated using two different approaches. The top candidates for competition of DNA were selected using a PERL script to filter based on having a geometric center within 10 Å of the location of sense-strand DNA binding. Hits were then filtered based on having a predicted free energy of interaction < -5 kJ/Mol.

IC₅₀ determination

The IC₅₀ of each putative ApiAP2 competing compound was determined using a 48-hour growth inhibition assay as described with minor modifications [70]. Briefly, parasites were seeded at 0.5% parasitemia and 4% hematocrit on a 96 well plate. Serial dilutions of drug or DMSO vehicle control were added in to each well and parasites were incubated in standard culture conditions for 48 hours. The assay was run in biological triplicate. Parasite cultures were frozen at -80°C overnight, then 100μL of culture was added to 100μL 2x Lysis Buffer containing 0.2μL/1mL Sybr Green I (20 mM Tris pH 7.5, 5 mM EDTA, 0.008% saponin, 0.08% Triton X-100). The mixture was incubated at room temperature for three hours, then fluorescence was measured at excitation/emission 585nm/535nm for each well. Growth values were normalized to the vehicle control. IC₅₀ values were calculated and growth inhibition curves were plotted using GraphPad Prism with the following parameters: nonlinear regression, Dose-Response—Special, X is log(concentration).

Recombinant protein expression

Recombinant AP2 domains AP2-EXP, AP2-I D3, AP2-HS D1, and PfSIP2 D1 were overexpressed and purified from BL21 PlysS E. coli as described previously [12]. Recombinant AtERF1 was purified using the same method after cloning the AtERF1 AP2 domain [44] into the pGex4t-1 overexpression vector (S1 Table). The PbAP2-Sp coding sequence was cloned into the pGex4t-1 overexpression vector using the domain boundaries defined previously (S1 Table) [29]. Recombinant protein was quantified using Braford reagent (Pierce). Input, flowthrough, and eluate fractions were analyzed by SDS-PAGE to ensure recovery of the full-length recombinant protein. Recombinant full-length SOX2 protein was purchased from Abcam (ab169843).

Electrophoretic mobility shift assays

EMSAs were run in DNA binding buffer (10mM Tris pH 7.5, 50mM KCl, 1mM DTT, 6mM MgCl2, 60ng/μL Poly DiDC, 65ng BSA). Recombinant proteins were titrated to empirically determine the minimum mass required for DNA binding and this mass (S4 Fig) was used for each EMSA unless otherwise specified. PAGE purified DNA probes with a 5’ biotin ligated on the forward DNA strand, along with an unlabeled complementary stand, were purchased from IDT (S1 Table). DNA probes were double stranded by heating to 95°C, followed by stepwise cooling in Annealing Buffer (10mMTtris-Cl pH7.5, 1mM EDTA, 10mM NaCl). Each recombinant protein was incubated in DNA binding mixture plus competitor compound for 15 minutes prior to addition of the cognate double stranded DNA oligonucleotide. Protein, competitor, and DNA oligonucleotide were incubated together for an additional 5 minutes. The mixtures were separated on a .5x TBE polyacrylamide gel, transferred to a nylon membrane (Amersham) at 50 Volts for 30 minutes, and probed using the Light Shift nucleic acid detection module (Thermo) according to the manufacturer’s protocol. All gels were imaged using a Bio Rad chemiluminescence imager.

Ethidium bromide exclusion assay

10μM of double stranded DNA (sequence TGCATGCA, purchased from IDT) in 0.01mM EDTA, 9.4mM NaCl, 2mM HEPES buffer pH 7.9 was incubated for 10 minutes in the presence of 100nM ethidium bromide. Fluorescence was measured at excitation/emission 546nm/595nm for each well. Following the baseline reading, each putative ApiAP2 competing compound was added. DRAQ5 nucleic acid dye was used as a positive control for knockdown of ethidium bromide fluorescence by DNA intercalation. Ethidium bromide exclusion assays were performed in technical triplicate using a 96 well plate.

Molecular dynamics simulations

Five compounds were prepared for molecular dynamics (MD) simulation: B, B-1, B-2, B-3, and B-4. In preparation for MD, parameters for the compounds were obtained using Antechamber [71–73] with the Generalized Amber Force Field. Starting conformations for each compound bound to AP2-EXP were based on predictions from docking. All complexes were prepared using the tleap module of AmberTools [74] with the protein.ff14SB forcefield [75]. Each complex was solvated in an octahedral box of TIP3P water with a 10 -Å buffer around the protein complex. Na⁺ and Cl^- ions were added to neutralize the protein and achieve physiological conditions. All MD minimizations and simulations were performed using Amber with GPU acceleration [76,77]. First, complexes minimized with 5000 steps each of steepest decent and conjugate gradient minimization with 500 kcal/mol∙Ų restraints on all complex atoms. Restraints were reduced to 100 kcal/mol∙Ų and the minimization protocol was repeated. Restraints were then retained only on the compound for a final minimization step. Following minimization, all complexes were heated from 0 to 300 K using a 100-ps run with constant volume periodic boundaries and 10 kcal/mol∙Ų restraints on all protein and compound atoms. To equilibrate complexes, 10 ns of MD was performed first with 10 kcal/mol∙Ų, then with 1 kcal/mol∙Ų restraints on protein and compound atoms using the NPT ensemble. With kcal/mol∙Ų restraints retained on complexes, 500 ns production simulations were performed. A 2-fs timestep was used and all bonds between heavy atoms and hydrogens were fixed with the SHAKE algorithm [78]. A cut-off distance of 10 Å was used to evaluate long-range electrostatics with Particle Mesh Ewald (PME) and for van der Waals forces. The ‘strip’ and ‘trajout’ commands of the CPPTRAJ module [79] were used to remove solvent atoms and extract 50,000 evenly spaced frames from each simulation for analysis.

AP2 domain competitor phenotyping assays

For Compound B, C, and H phenotyping time courses (Fig 3A-3E), highly synchronous Pf3D7 wild type parasites at 10% starting parasitemia were spiked with 40μM of each drug at 6 hpi. An equivalent volume of DMSO was used as a control. Parasites were morphologically assessed by Giemsa staining and parasitemia was counted every 6 hours until 48 hpi.

For the Compound C exposure interval assay (Figs 3F and S8A) 40μM of Compound C was spiked into highly synchronous Pf3D7 wild type parasites at 1% starting parasitemia for individual 8-hour intervals starting at 6 hpi. An equivalent volume of DMSO vehicle was used as a control for normal growth. At 54 hpi, cultures were counted using Giemsa-stained slides (Raw data in S8B Fig). In Fig 3F, Compound C induced growth inhibition is reported at the ratio of nascent rings to stalled trophozoites in culture at 54 hpi. The Mann Whitney U-test was used to check for significance of difference between each dosage interval and the DMSO vehicle control.

To determine the phenotype caused by continuous exposure of 0.66xIC₅₀ Compound C to wild type Pf3D7 parasites in S13A, and S13B Fig, highly synchronous Pf3D7 parasites at 5% starting parasitemia were spiked with 12μM Compound C, 40μM Compound C, or DMSO vehicle control at 10 hpi. The percentages of mature trophozoites and reinvaded rings were recorded at 24 hpi and 48 hpi, respectively.

For all phenotyping assays, 40μM of compound was used because it was the maximum concentration achievable in culture after resuspending each compound in 100% DMSO. Higher concentrations of compound caused parasite mortality due to the high volume of DMSO required (>1% volume/volume). All assays were performed in biological and technical triplicate. Error bars represent the standard deviation of the mean.

Preparation of RNA for DNA microarray

DNA microarrays were prepared using the protocol described in [50]. After collecting a control timepoint at 6 hpi, parasites were cultured with either 12μM Compound C (0.66xIC₅₀) or DMSO vehicle control. Parasite RNA was then harvested at 12, 18, 24, 30, 42, and 48 hpi. cDNA was synthesized using SuperScript II, hybridized onto Agilent DNA microarrays [50], and scanned using an Axon 4200A Scanner. Agilent Feature Extraction Software version 11.0.1.1 with the protocol GE2-v5_95_Feb07_nospikein was used to extract signal intensities. Raw signal intensities for all microarrays are available in NCBI GEO GSE202876: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE202876

Analysis of DNA microarrays

LIMMA [52] was used to normalize and extract signal intensities. Arrays were normalized using robust splines normalization and within arrays parameters selected. Average expression values were calculated per gene by averaging the log₂ Cy5 (cDNA)/Cy3 (Reference Pool) signal intensity across all probes. Correlation plots for the Compound C vs. DMSO control total transcriptome and control genes were made using the log₂ signal intensities for all detected genes using the R package CorrPlot downloaded from: https://github.com/taiyun/corrplot. The significance of correlation was tested using the cor.mtest function of CorrPlot with confidence level = 0.95. The LIMMA eBayes function was used on average gene abundance data to determine changes in transcript abundance between 24 and 30hpi that occur differentially in DMSO control vs. Compound C dosed parasites. Volcano plots were made with LIMMA eBayes data using the Enhanced Volcano package in R downloaded from: https://github.com/kevinblighe/EnhancedVolcano. The Rnits [54] R package was used to model differences in gene expression across the entire time course between the DMSO control and Compound C spike in parasites. Rnits was run using the parameters: center genes, normalize by intensity, and background filter probes. The Rnits model for differential expression was fit at the gene level. All data from LIMMA, eBayes, and Rnits is provided in S3 Table. Heat maps for control genes and AP2-EXP target genes were made using Java Treeview [80] after means centering of transcript abundance values with Cluster3.0 [81].

ChIP sample preparation

ChIP-seq was performed as described in [19]. Highly synchronous parasites were grown up to 5–10% parasitemia and AP2-EXP ChIP was performed at 30 hpi for all replicates. Immunoprecipitation was performed overnight with .5mg/mL 3F10 anti-HA (Sigma) or .1mg/mL Ab290 (Abcam) anti-GFP.

DNA library preparation

ChIP-seq DNA libraries were prepared as described in [19] using the NEBNext II DNA library kit (New England Biolabs) according to the manufacturer’s instructions. Quality was assessed using an Agilent 2100 Bioanalyzer or TapeStation. Libraries were sequenced using a HiSeq 2500 (Replicate GFP1) or NextSeq 550 (Replicates GFP2 and HA3) Illumina sequencer. AMPure XP beads (Beckmann Coulter) were used to size select and purify DNA between NEBNext II library preparation steps. Whole genome sequencing DNA libraries were prepared using the Illumina Tru-Seq PCR free DNA library kit and sequenced using a HiSeq 2500.

qPCR

ChIP-qPCR samples were collected from 30–35 hpi trophozoites following two hours of Compound C (40μM) or DMSO vehicle control spike in. Primer pairs to be used for ChIP-qPCR were first evaluated to check for 80–110% efficiency using sonicated genomic DNA. RT-qPCR was carried out using Sybr Green Polymerase master mix (Thermo) with the specified primer concentration (S1 Table). The Ct was calculated using SDSv1.4 (Applied Biosystems) software, averaged over technical triplicate. The percent of input per immunoprecipitated DNA fraction was calculated using the delta Ct method. Each assay was performed in biological triplicate, with the exception of the ama1 primer pair, where n = 2. Error bars represent standard deviation of the mean. Data was obtained using an Applied Biosystems 7300 Real-Time PCR Machine. Significance was assessed using an unpaired t-test.

ChIP-seq data analysis

First, reads in each DNA library were trimmed to remove low quality base calls and Illumina adaptor sequences using Trimmomatic [82] with a quality score cutoff of 20. FastQC [83] was used to assess the quality of each DNA library following this step. Reads were then mapped to the Plasmodium falciparum genome version 36 downloaded from PlasmoDB using BWA-Mem [84]. Multiply mapped reads were filtered out using Samtools [85]. Filtered and mapped Input and immunoprecipitate bam files were used for peak calling by MACS2 [86] with the following parameters: effective genome size 20000000, No Model, q 0.01. Peaks called in each ChIP-seq replicate were overlapped using Bedtools [87] Intersect, and peaks that occur in a minimum of 2/3 replicates were used for downstream analysis. For data visualization using IGV,.bam files were converted to log₂ Immunoprecipitate fraction/Input fraction bigwigs using DeepTools [88] BamCompare. Bedtools [87] ClosestBed was used to correlate regions called as peaks of occupancy by MACS2 to the closest gene. Results were filtered based on 1.5kb proximity to the MACS2 peak and strandedness of the putative target gene. For comparison of ChIP-seq library coverages to Transcription Start Sites and Start Codons, DeepTools PlotHeatmap was used. Matrix files underlying PlotHeatmap were created using DeepTools ComputeMatrix. A four-color plot of conserved AP2-EXP DNA binding sites was made using the program cegr-tools four color plot downloaded from https://github.com/seqcode/cegr-tools/tree/master/src/org/seqcode/cegrtools. DNA motifs enriched in AP2-EXP ChIP-seq peaks were determined using the DREME Suite [89]. DNA motifs were sorted from highest to lowest level of conservation to the aggregate AP2-EXP DNA motif using FIMO [90].

For comparison of AP2-EXP peaks with existing post-translational histone marks, chromatin post translational modification datasets [57,91] were downloaded from https://github.com/Daread/plasmodiumExprPrediction [92]. Chromatin reader and nucleosome occupancy datasets [58,93,94] were downloaded from NCBI. ChIP-seq data for BDP1 and HP1 was processed as described above. Nucleosome occupancy data was downloaded from NCBI and processed as described above, with the exception that bam files were turned into bigwig using BamtoBigWig with the—mnase parameter selected. DeepTools ComputeMatrix was used to format the data. DeepTools PlotHeatmap was used to plot coverage of each dataset within 10kb of AP2-EXP peaks. For negative control coverage plots, Bedtools [87] ShuffleBed was used to create random genomic intervals on the same chromosome and the same length as AP2-EXP peaks of occupancy.

For comparison of AP2-EXP peaks identified by Shang et al [17] to this study, AP2-EXP peaks of occupancy in the trophozoite stage were downloaded from https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE184658. Peak summits from each replicate were expanded to 200 nucleotide windows using Bedtools SlopBed [87]. Conserved peaks in 2/2 replicates were identified using Bedtools intersect [87], then AP2-EXP peaks from this study and Shang et al were overlapped using Bedtools Intersect [87].

Western blot

Full parasite protein western blot samples were collected by lysing RBCs with 0.1% saponin and boiling protein in Loading Buffer (50mM Tris-Cl pH 8.0, 20% SDS, 1% Bromophenol Blue). Fractionated parasite protein was prepared as described in [95]. Blots were performed as described [19]. Primary antibodies used were: 1/1000 rat ant-HA (Roche 3F10), 1/1000 mouse anti-GFP (Roche), 1/3000 rabbit anti-aldolase conjugated to HRP (Abcam ab38905), or 1/3000 mouse anti-H3 (Abcam ab10799). Secondary antibody concentrations used were 1/3000 goat anti-rat HRP conjugate (Millipore), 1/3000 goat anti-mouse HRP conjugate, or 1/10,000 (Pierce) goat anti-rabbit HRP conjugate (Millipore). ECL reagent (Pierce) was used to detect HRP signal. Blots were exposed to autoradiography film (VWR) and visualized using an autoradiography developer.

Protein pulldown

Parasite nuclear protein was isolated as described [95]. Following the final centrifugation step, the supernatant was collected and diluted by 1:3 in Dilution Buffer (30% glycerol, 20mM HEPES pH 7.8). GFP tagged AP2-EXP was pulled down using Chromotek anti-GFP or mock immunoprecipitated using Chromotek negative control magnetic beads. Beads were washed twice in Wash Buffer prior to use (20mM HEPES pH 7.4, 250mM NaCl, 1mM EDTA, 1mM TCEP, 0.05% NP-40). Protein and beads were incubated together for 1 hour at 4°C. The beads were then washed twice with Wash Buffer and bound protein was collected in Loading Buffer by boiling at 95°C for 10 minutes.

Fluorescent microscopy

Samples were prepared by incubating packed infected red blood cells with DRAQ5 dye (Thermo) for 15 minutes. Parasites were then washed in 1xPBS to remove excess dye and immediately placed on a glass slide for imaging. Fluorescent microscopy images were acquired using an Olympus Bx61 fluorescent microscope. All images were processed using SlideBook 5.0.

Mosquito stage P. berghei AP2 competitor growth assay

A female Swiss Webster mouse was inoculated with Plasmodium berghei ANKA 2.34

from frozen stock. Once the parasitemia reached 15%, the blood was harvested by heart puncture, washed twice with 1xPBS and resuspended to 10 mL in 1xPBS. 5 female Swiss Webster mice were infected with 500μL of the resuspended blood. 3–4 days after blood passage, exflagellation of the P. berghei gametocytes was assayed. Briefly, a drop of tail vein blood was incubated in RPMI 1640 (Invitrogen) containing 1μM xanthurenic acid (Sigma) for ~12–15 minutes. The mixture was added to a slide and observed under a light microscope at 40x. Ten or more fields were observed. Mice with 0.3–0.7 exflagellations per field were anesthetized and fed on 3-day post emergence Anopheles stephensi mosquitoes. Mosquitoes were maintained at 19°C and fed 10% sucrose. Bloodfed mosquitoes were separated 30 hours post-bloodfeeding. 10 days post-bloodfeeding, 5–10 mosquitoes were dissected and oocysts were counted by mercurochrome staining and light microscopy to ensure P. berghei oocyst development. The remaining mosquitoes were injected by standard mouth pipette technique with small molecule inhibitors (18μM Compound C, 40μM Compound B, 20μM Compound F) or a mixture of PBS-DMSO for control. The surviving injected mosquitoes were dissected 14 days post-bloodfeeding. Oocyst numbers were counted as before. Pooled groups of midguts were ground using a pestle, centrifuged at 7,000 rpm for 5 minutes and resuspended in 20μL 1xPBS to release developing mid-gut sporozoites. Sporozoites were counted on a haemocytometer. Significance of difference between DMSO control and compound injected samples was assessed using the Mann Whitney U-test.