PfGCN5 forms a unique complex that is highly divergent from the SAGA complex

The evolutionarily conserved SAGA complex in eukaryotes is composed of 18–20 subunits, which are organized into several modules including the HAT catalytic core consisting of GCN5, ADA2, ADA3, and Sgf29 [9,10]. By contrast, bioinformatic analysis of the Plasmodium genomes using conserved modular components of the SAGA complexes identified only two ubiquitous subunits, GCN5 and ADA2 [55,63], and a potential Tra1 homolog (PF3D7_1303800) [64], suggesting that the GCN5 complex(es) in these early-branching, parasitic protists might be highly divergent from the SAGA complex. Our recent work, aiming to identify “readers” of modified histones with the H3K4me3 peptide, surprisingly pulled down a putative PfGCN5 complex containing the PfGCN5, PfADA2, and two large proteins containing multiple plant homeodomains (PHDs), namely PfPHD1 and PfPHD2 [62]. To precisely define the GCN5 complex(es) in P. falciparum, we tagged the C-terminus of the endogenous PfGCN5 gene in the 3D7 strain with a PTP tag consisting of a protein C epitope, a tobacco etch virus (TEV) protease cleavage site, and two protein A domains (S1A and S1B Fig), which allows for efficient TAP of protein complexes under native conditions with extremely low backgrounds [65,66]. Integration-specific PCR confirmed the correct integration of the PTP tag at the PfGCN5 locus (S1C Fig). The transgenic parasites showed no noticeable in vitro growth defects (not shown). Western blot analysis using the anti-protein C antibody revealed that PfGCN5::PTP was expressed in all developmental stages of the IDC with the peak protein level in early trophozoites (S1D Fig). Six protein bands were detected, and the band pattern agreed with that detected with an antibody against the PfGCN5 C-terminal fragment [60], confirming proteolytic processing of PfGCN5 (S1D Fig). Live-cell imaging of the green fluorescent protein (GFP)-tagged PfGCN5 parasite line [67] showed that the PfGCN5::GFP protein was expressed throughout the IDC and localized in the nucleus (S1E Fig). Thus, we performed the TAP procedure using nuclear extracts from 10⁹ synchronized trophozoites of the PfGCN5::PTP parasite, which was followed by liquid chromatography and tandem mass spectrometry (LC-MS/MS) for accurate protein identification. The MS data were subjected to Significance Analysis of INTeractome (SAINT) using a threshold of probability above 94% and false discovery rate (FDR) below 1% [68].

Three independent experiments of TAP and LC-MS/MS consistently identified nine proteins (Fig 1A and S1A and S1B Table), presumably representing the core subunits of this PfGCN5 complex (Fig 1A). This is in sharp contrast to the detection of only some abundant cellular proteins in the three replicates of the wild-type (WT) 3D7 control pulldown experiments (S1A Table). Seven proteins identified using the TAP procedure were also present in the PfGCN5-associated proteins identified by a single-step pulldown procedure [62]. In agreement with our earlier work showing interactions between PfGCN5 and PfADA2 [63], these two proteins were among the most enriched proteins in the PfGCN5::PTP pulldown, demonstrating the high efficiency of the TAP procedure and preserved integrity of the complex. Consistent with the recent PfGCN5::GFP pulldown results [62], the PfGCN5 core complex includes two large proteins PfPHD1 (PF3D7_1008100) and PfPHD2 (PF3D7_1433400), each containing four PHD zinc fingers (Figs 1A, 1B, S2A and S2B). PfPHD1 also contains two AT hooks, which are DNA-binding domains with a preference for AT-rich regions [69]. Sequence analysis indicated that these PHDs belong to the PHD superfamily with some containing additional cysteine and histidine residues (called extended PHD, ePHD). The ePHD has been found to bind dsDNA, methylated H3K4, or other TFs [70–73]. Only the fourth PHD in PfPHD1 conforms to the canonical PHDs that bind to H3K4me3/2 [74] (S2C Fig). Our recent study confirmed that this domain indeed preferentially binds H3K4me3/2 [62]. Furthermore, these two proteins were found to harbor large numbers of acetylation sites in our acetylome study [75], indicating that they are substrates of protein lysine acetyltransferases. An AP2-domain family TF (PF3D7_0802100), named PfAP2-LT, which is highly expressed at the late stages of the IDC [33], was consistently identified in all experimental replicates of the pulldown studies. Of note, the interaction between the PfGCN5 N-terminal fragment and AP2-LT has been identified in a genome-wide yeast two-hybrid screen [76]. Further, a histone assembly protein PfNAPS (PF3D7_0919000) was also identified in the PfGCN5 interactome. Finally, in the PfGCN5 core complex are three proteins with unknown functions (PF3D7_1019700, PF3D7_1364400, and PF3D7_1402800), which are conserved in all Plasmodium species (Fig 1A and 1B).

Download:

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

Fig 1. Identification of the PfGCN5 Core Complex in P. falciparum.

(A) Proteins identified from parasite nuclear extracts by IP and by LC-MS/MS. TAP procedure was performed using the PfGCN5::PTP line at the trophozoite stage (three replicates R1 –R3), while single-step IPs with the anti-Myc and anti-GFP beads were done using the PfPHD1::Myc parasite line at the trophozoite stage and PfGCN5::GFP at the ring stage (two replicates, R1, and R2). The wildtype 3D7 was used as IP control. The wildtype 3D7 was used as IP control. The proteomic data were analyzed by SAINT using a threshold of probability >94% and 1% FDR. Nine proteins consistently identified are marked as the PfGCN5 complex core subunits. Gene ID and annotation are shown on the right. (B) Schematic diagrams showing the features (putative domains and protein size) of the core subunits. HAT: histone acetyltransferase enzymatic domain; ADA2: ADA2 binding domain; BrD: Bromodomain; Med15: Med15 domain; ADA2LD: ASA2-like domain; PHD-SF: PHD finger superfamily; ePHD: extended PHD finger; PHD-TAF3: PHD finger homologous to transcription initiation factor TFIID subunit 3 (TAF3); AT: AT hook; TM: transmembrane region; AP2: AP2-domain; cyclin-like: cyclin-like domain; NAP: nucleosome assembly protein. The C-terminals of PfGCN5 and PfPHD1 labeled by the blue line and “X” containing BrD and PHD-TAF3 were deleted for functional studies (see details in S6 Fig). (C) Gel filtration analysis of the PfGCN5 complex. Aliquots of different fractions were used for Western blots with anti-PTP, PfPHD1 and PfPHD2 antibodies to detect full-length of PfGCN5, PfPHD1 and PfPHD2, and for the HAT assay using recombinant histone H3.

https://doi.org/10.1371/journal.ppat.1009351.g001

To confirm that PfPHD1 and PfPHD2 are the core constituents of the PfGCN5 complex, rabbit polyclonal antibodies were generated against peptides of PfPHD1 (aa 3685–3702) and PfPHD2 (aa 5738–5756). Indirect immunofluorescence assay (IFA) showed that the pre-immune sera did not react with parasitized red blood cells (RBCs), whereas the anti-PfPHD1 and -PfPHD2 antibodies detected fluorescent signals in the parasite nuclei (S3A Fig). Nuclear extracts of trophozoites from the PfGCN5::PTP line were subjected to immunoprecipitation (IP) with anti-PfPHD1 and anti-PfPHD2 antibodies, and the immunoprecipitated proteins were analyzed by immunoblotting with the anti-protein C antibodies. In contrast to the control IP with the pre-immune sera, PfGCN5-PTP was only detected in the IP with the anti-PfPHD1 or -PfPHD2 antibodies, confirming co-purification of PfGCN5 with PfPHD1 and PfPHD2 (S3B Fig).

Previous IPs and proteomic analysis with the tagged PfPHD1 and PfPHD2 subunits suggest they may be present in different versions of the PfGCN5 complex [62]. Whereas IP from PfPHD1::3×HA only purified PfPHD1, PfGCN5, ADA2, and PF3D7_1402800, IP from PfPHD2::GFP identified 12 putative subunits with marginal enrichment of PfPHD1 but no pulldown of PF3D7_1402800 [62]. To clarify this discrepancy with our TAP results, we separately tagged two subunits in these two putative versions of the PfGCN5 complexes for reciprocal IP: the PfPHD1 with a C-terminal c-Myc tag (S4 Fig) and PF3D7_1019700 with a C-terminal GFP tag (S5 Fig). Correct integration of the c-Myc tag at the PfPHD1 locus and the GFP tag at the PF3D7_1019700 locus was confirmed by Southern blot analysis and integration-specific PCR, respectively (S4A, S4B, S5A and S5B Figs). Nuclear localization of the PfPHD1::Myc and PF3D7_1019700::GFP proteins was verified by cellular fractionation–Western blot (S4C Fig) and live-cell image analysis (S5C Fig), respectively. Affinity purification of the trophozoite nuclear extracts from the PfPHD1::Myc parasites with the Myc-trap beads followed by LC-MS/MS consistently identified all 9 subunits of the putative PfGCN5 complex, including PfPHD2 (Fig 1A and S1C and S1D Table). Similarly, two IP replicates from the PF3D7_1019700::GFP identified 7 of the 9 core components of the PfGCN5 complex, including both PfPHD1 and PfPHD2 (S5D Fig and S1E and S1F Table). Furthermore, reciprocal IP using agarose conjugated with anti-PfPHD1 antibodies identified PfPHD2 in the pulldown and vice versa (S3C Fig). These results are consistent with the presence of both PfPHD1 and PfPHD2 in the same PfGCN5 complex.

To estimate the size of the PfGCN5 complex, the purified native complex by TAP from the PfGCN5::PTP parasites was subjected to gel filtration, and fractions were assayed for HAT activity and Western blots using anti-PTP, -PfPHD1, and -PfPHD2 antibodies (Fig 1C). The results showed that PfGCN5 and its associated HAT activity, PfPHD1, and PfPHD2 all were detected in fractions 15 and 16, which is compatible with one major PfGCN5 complex in P. falciparum asexual stages. Based on the calibration using molecular mass standards, the size of the complex was approximately 2.3 MDa, which is comparable to the size (2.26 MDa) estimated based on the predicted molecular masses of the 9 core subunits, each presented as a single copy (Fig 1B). Notably, IPs performed using single beads from the PfPHD1::Myc and PF3D7_1019700::GFP parasite lines identified additional proteins associated with chromatin besides the core subunits of the PfGCN5 complex identified by TAP. These proteins include the bromodomain protein 1 (BDP1) [77,78], HP1 [35,44,45], and other AP2 family TFs (Fig 1A and S1 Table), suggesting that the PfGCN5 complex may have broader functions in chromatin biology in the malaria parasite.

The identification of AP2-LT in the core PfGCN5 complex and the presence of other AP2 family members in the IPs suggest that different AP2 TFs may be dynamically associated with the PfGCN5 complex during development, especially since the IP was conducted at the trophozoite stage when AP2-LT was highly expressed. To investigate this possibility, we performed IPs at the ring stage of PfGCN5::GFP parasite with the GFP-trap beads. Except for PF3D7_1364400, the rest of the eight-core subunits of the PfGCN5 complex were identified, suggesting that AP2-LT is a stable core subunit during the asexual development (Fig 1A and S1G and S1H Table).

Domain deletions in PfGCN5 and PfPHD1 cause severe growth defects in parasites

To characterize the function of the PfGCN5 complex in transcription regulation, we attempted to knock out the PfGCN5, PfADA2, PfPHD1, and PfPHD2 genes by double-crossover homologous recombination but were unsuccessful after multiple tries, indicating these genes are essential for parasite survival. This result is consistent with the mutagenesis scores in the genome-wide piggyback transposon mutagenesis study showing the essentiality of these genes [79]. Since PTM-binding domains such as the bromodomain (BrD) and PHD are important for anchoring and holding the respective proteins or complexes to the chromatin, we speculated that deleting these domains might disturb histone modifications without causing lethality to the parasite. Thus, we attempted to delete the BrD and PHD from the C-termini of PfGCN5 and PfPHD1, respectively, using a single-crossover gene disruption strategy, and meanwhile tag the C-termini of these truncated proteins with a GFP tag for sorting parasites with truncated PfGCN5 or PfPHD1 (S6A and S6C Fig). After transfection, the parasites were selected with WR99210, and GFP-positive parasites were cloned by sorting GFP-positive parasites using flow cytometry. Correct integration of the plasmids at the PfGCN5 and the PfPHD1 loci in the parasite genome was confirmed by Southern (S6B and S6D Fig) and Western blots (S6E Fig). Phenotypic analyses of the parasites with the domain deletions in these two proteins revealed that the parasites with PfGCN5 BrD deletion (GCN5-ΔBrD) and parasites with PfPHD1 PHD domain deletion (PHD1-ΔPHD), to the greatest extent, phenocopied each other (Fig 2). Both the GCN5-ΔBrD and PHD1-ΔPHD parasites grew significantly more slowly than the WT parasites; When starting at 0.1% parasitemia, they only reached ~1% parasitemia on day 7 compared to ~10% in WT parasites (Fig 2A). Both the GCN5-ΔBrD and PHD1-ΔPHD parasites had a ~2-fold proliferation rate compared to ~5-fold in the WT parasite. A more detailed analysis of the growth defects in these domain deletion lines showed that mature schizonts in these mutant parasites produced similar numbers of merozoites as the WT parasites (Fig 2B), but these merozoites had substantially reduced efficiency (by almost 60%) in the invasion of RBCs (P < 0.05, paired Wilcoxon test; Fig 2C). In addition, these domain deletion mutants also had a 2–3 h longer IDC than the WT parasites (P < 0.05, paired Wilcoxon test; Fig 2D), and a more extended ring stage (Figs 2E, 2F and S6F). Furthermore, these domain deletion parasites were inclined to produce more gametocytes than WT when the gametocytogenesis was induced by using the established method [80] (P < 0.05, paired Wilcoxon test; Fig 2G).

Download:

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

Fig 2. Growth phenotypes in PfGCN5 and PfPHD1 domain deletion mutants.

(A) Asexual growth rates of WT 3D7, PfGCN5-ΔBrD::GFP, and PfPHD1-ΔPHD::GFP. ** indicate P < 0.01 (ANOVA) at days 5 and 7. The starting parasitemia was 0.1%. (B) The distribution of the number of mature schizonts with a variable number of merozoite. No differences were identified among the three parasite lines (P > 0.05, ANOVA). (C) Invasion rates of the GCN5-ΔBrD::GFP and PHD1-ΔPHD::GFP parasite lines estimated by incubating purified schizonts with RBCs (P < 0.05, paired Wilcoxon test). (D) The duration of the IDC showing significantly increased lengths in GCN5-ΔBrD::GFP and PHD1-ΔPHD::GFP parasite lines (P < 0.05, paired Wilcoxon test). (E, F) Detailed analysis of the IDC showing extended ring stage in the GCN5-ΔBrD::GFP (E) and PHD1-ΔPHD::GFP (F). (G) Gametocytemias at day 6 after induction of gametocytogenesis showing significantly increased gametocytemia in the two domain deletion mutants (P < 0.05, paired Wilcoxon test).

https://doi.org/10.1371/journal.ppat.1009351.g002

Domain deletions in PfGCN5 and PfPHD1 are associated with globally reduced H3K9ac and H3K4me3

P. falciparum has an extensively euchromatic epigenome with a preponderance of the histone marks H3K9ac and H3K4me3 [36,54]. The presence of BrD and PHD in the PfGCN5 complex, which bind to acetylated H3K9/14 and H3K4me3/2 marks, respectively, strongly suggests that both domains may be required for anchoring the PfGCN5 complex to chromosomal regions to reinforce the euchromatic state. To determine the impacts of BrD and PHD deletions on the overall euchromatic histone marks, histones were purified from parasites of different developmental stages, and several histone marks were analyzed by Western blots. Consistent with PfGCN5 being the major HAT mediating H3K9 and H3K14 acetylation, deletion of BrD in PfGCN5 led to a significant reduction of the H3K9ac and H3K14ac levels in the trophozoite stage, corresponding to the time of peak PfGCN5 expression (Fig 3A). Similarly, deletion of the PHD in PfPHD1 also resulted in the reduction of H3K9 and H3K14 acetylation. In comparison, domain deletions in PfGCN5 and PfPHD1 did not cause noticeable changes in H4 tetra-acetylation (at positions H4K5, 8, 14, and 20), which is mediated by another HAT protein, PfMYST [57]. Interestingly, domain deletions in these two subunits of the PfGCN5 complex also resulted in significantly reduced levels of H3K4me3 in trophozoites (Fig 3A), another major euchromatin mark conferred by the PfSET1 histone methyltransferase, highlighting the presence of extensive crosstalk between the two euchromatin marks. This result echoes the findings from studies of the SAGA complexes in model organisms, where GCN5 deletion or the Sgf29 Tudor domain deletion reduced the levels of both H3K9ac and H3K4me3 [17,18,81]. Taken together, these results indicate that both BrD in PfGCN5 and PHD in PfPHD1 are important for anchoring the PfGCN5 complex to maintain the euchromatic histone marks.

Download:

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

Fig 3. Domain deletions affect the abundance and localization of active histone marks and the integrity of the PfGCN5 complex.

(A) The levels of active histone marks in 3D7, PfGCN5-ΔBrD::GFP (ΔBrD), and PfPHD1-ΔPHD::GFP (ΔPHD) parasite lines. Histones were purified from the ring, trophozoite, and schizont stages, and detected by Western blots with specific antibodies against the modified histones H3K9ac, H3K14ac, H3K4me3, and H4Acs. Anti-H3 antibodies were used for loading control. The Western blots were performed in three biological replicates, and the band intensities were determined using a densitometer. The number underneath each band indicates the fold change ± standard deviation between the ratio of the corresponding histone mark normalized to the H3 control. The ratios in 3D7 at the ring stage were set as 1. (B, C) Co-localization of full-length PfGCN5 (GCN5::GFP) (B) or truncated PfGCN5 (GCN5-ΔBrd::GFP) (C) with H3K9ac and DAPI by IFA with anti-GFP and H3K9Ac antibodies. Note the expansion of the truncated PfGCN5-ΔBrD::GFP and H3K9ac beyond the periphery of the euchromatin areas demarcated by DAPI staining. (D) Effects of domain deletions in PfGCN5 and PfPHD1 on complex integrity. Proteins were pulled down from the trophozoite nuclear extracts of the PfGCN5::GFP, GCN5-ΔBrD::GFP, and PHD1-ΔPHD::GFP parasite lines and identified by LC-MS/MS. R1, R2, R3 and R4 indicate individual repeats of the experiment. Shown here are proteins passing the threshold of SAINT (probability >94% and FDR <1%). The nine PfGCN5 complex core subunits were all detected in the IPs of PfGCN5::GFP and GCN5-ΔBrD::GFP, whereas only four of the core subunits were identified in the IPs of PHD1-ΔPHD::GFP.

https://doi.org/10.1371/journal.ppat.1009351.g003

Spatial compartmentalization of chromatin in the nucleus is critical for gene regulation in malaria parasites. The active chromatin marks H3K9ac and the H3K14ac overlap extensively with the DAPI staining (often used to define the euchromatin domains), whereas the heterochromatin mark H3K9me3 mostly occupies the nuclear periphery outside of DAPI [35,82]. To analyze whether these changes in histone modifications were associated with altered spatial organization of the chromatin, we used IFA and live-cell microscopy to observe the nuclear locations of these histone modifications as well as the truncated PfGCN5 and PfPHD1. In the PfGCN5-ΔBrD::GFP parasites, Western blot with the anti-GFP antibodies detected reduced expression of the truncated PfGCN5 protein, while the pattern of the PfGCN5 fragments remained similar (S8A Fig). Live-cell microscopy of the GFP-tagged truncated PfGCN5-ΔBrD::GFP and PfPHD1-ΔPHD::GFP parasites showed that the GFP signals while overlapping largely with the parasite nuclei from DAPI staining (S6G and S6H Fig), were more diffuse than those in the WT parasites (S1E Fig). Consistent results were obtained by IFA with anti-GFP and H3K9ac antibodies in WT and PfGCN5-ΔBrD::GFP parasites (Fig 3B and 3C). In the PfGCN5::GFP parasites, there were high levels of colocalizations between the DAPI and PfGCN5 (r² = 0.83–0.91), and between PfGCN5 and H3K9ac (r² = 0.89–1.0), indicating that PfGCN5 is tightly associated with H3K9ac in the active euchromatin area demarcated by DAPI staining (Fig 3B). However, the levels of the colocalization substantially decreased in the PfGCN5-ΔBrD::GFP parasites (r² = 0.37–0.74) (Fig 3C). Some signals of the truncated PfGCN5 were localized beyond the DAPI area, suggesting that PfGCN5 might have spread to the perinuclear heterochromatic area. A similar pattern was found in the PfPHD1-ΔPHD::GFP parasites (S7A and S7B Fig). Since the reduced H3 acetylation in trophozoites after BrD or PHD domain deletion (Fig 3A) could be due to slower replacement of these histone marks after DNA replication, we compared the abundance of the H3K9ac in trophozoites of these parasite lines with 1N or 2N+ single nucleus. While the H3K9ac signal levels in the 1N trophozoites were similar between the WT and domain deletion parasites, the H3K9ac intensities in the 2N+ trophozoites were significantly lower in GCN5-ΔBrD/PHD1-ΔPHD parasites than the WT (S7C Fig). Collectively, these results indicated that domain deletions in PfGCN5 and PfPHD1 altered the nuclear distribution of the truncated proteins and reduced the levels of euchromatin marks in the parasites.

PHD deletion in PfPHD1 affects the integrity of the PfGCN5 core complex

To determine whether BrD deletion in PfGCN5 and PHD deletion in PfPHD1 affected the integrity of the PfGCN5 complex, we performed IP using the GFP-Trap antibodies with nuclear extracts from the individual domain deletion lines and analyzed the affinity-purified proteins by LC-MS/MS. Four replicates of IP using the PfGCN5::GFP parasites as the positive control consistently detected the core components of the PfGCN5 complex (Fig 3D and S2A and S2B Table). IP with the PfGCN5-ΔBrD::GFP parasites also consistently purified the 9 core subunits of the PfGCN5 complex, suggesting the deletion of the BrD from PfGCN5 did not affect the integrity of the complex (Fig 3D and S2C and S2D Table). However, only four major components of the PfGCN5 complex (PfGCN5, PfADA2, PfPHD1, and PF3D7_1402800) were detected from the PfPHD1-ΔPHD::GFP parasites (Fig 3D and S2E and S2F Table). It is noteworthy that both domain deletions resulted in the disassociation between the PfGCN5 core complex and BDP1, whereas BrD deletion resulted in a tighter association between the PfGCN5 core complex and several chromatin regulators, including chromodomain protein 1 (CHD1) and two AP2 TFs (Figs 1A and 3D), indicating that these domain deletions affected the complex integrity and interactions with other chromatin factors.

BrD and PHD deletions profoundly affect global transcription

Since GCN5-associated complexes facilitate transcription of target genes by bridging transcriptional activator and the preinitiation complex, deletion of domains from subunits of the complex that interact with histone tails weakens the anchoring and retention of the complex, leading to reduced transcriptional activation of the target genes [81,83,84]. To gain a mechanistic understanding of the PfGCN5 complex in transcriptional regulation, we compared the transcriptomes of the WT parasites and parasites with domain deletions. Parasites were highly synchronized from purified schizonts with a 3 h window, and RNA-seq analysis was performed in three biological replicates during the IDC at 10, 20, 30, and 40 h after RBC invasion in the WT parasites and at 10, 23, 33, and 43 h in the parasites with BrD and PHD deletions to more closely match the developmental stages of the WT and domain-deletion parasites based on comparison of their IDC (Fig 2E and 2F). Differential gene expression was analyzed by DESeq2 using a P-adjustment value of <0.01 as the threshold of significance. To rule out potential normalization artifact by DESeq2, which assumes that the median expression level between the two conditions is the same, we also used Transcripts Per Kilobase Million (TPM) to normalize the RNA-seq data. The results showed that the differentially expressed genes detected by the two normalization methods were highly congruent (S3 and S4 Tables).

Compared to the phaseogram of WT parasites displaying a clear cascade-like gene expression pattern [22], PfGCN5 BrD deletion profoundly disturbed the global transcription pattern, causing 3533 (62.6%) genes to be differentially expressed in at least one of the four IDC time points analyzed (Fig 4A and S3 and S4 Tables). Specifically, BrD deletion resulted in the down-regulation of 997, 799, 861, and 902 genes, and up-regulation of 1127, 780, 846, and 368 genes at the ring, early trophozoite, late trophozoite, and schizont stage, respectively (Figs 4B–4D, S8A and S8B). Noticeably, the numbers of up- and down-regulated transcripts were comparable at all stages except at the schizont stage, where 2.5-fold more transcripts were down-regulated than up-regulated (Fig 4D). In comparison, PHD deletion in PfPHD1 caused a similar but more profound disturbance of gene expression during the IDC, with 3870 (68.6%) transcripts being differentially expressed in at least one of the four stages analyzed (Fig 4A), which is congruent with the more substantial disruption of the PfGCN5 complex upon PHD deletion (Fig 4). The PfPHD1 PHD deletion resulted in the down-regulation of 872, 1021, 557, and 787 genes, and up-regulation of 1481, 1266, 648, and 1028 genes at the ring, early trophozoite, late trophozoite and schizont stage, respectively (Figs 4D–4F, S8C and S8D). Of note, only in late trophozoites did PfPHD1 PHD deletion disturb the expression of fewer genes than PfGCN5 Brd deletion (1205 vs. 1707 genes) (Fig 4D).

Download:

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

Fig 4. Global transcriptomic changes up domain deletions in PfGCN5 and PfPHD1.

(A) The phaseograms of transcriptome from the WT 3D7, PfGCN5-ΔBrD::GFP (ΔBrd), PfPHD1-ΔPHD::GFP (ΔPHD) showing the disturbance of the cascade-like gene expression pattern in the deletion mutants at different developmental stages. R, ring; ET, early trophozoite; LT, late trophozoite; S, schizont. (B, C) Volcano plots showing altered gene expression at the ring (B) and schizont (C) stages in PfGCN5-ΔBrD::GFP compared to the WT 3D7. The x-axis indicates log₂ (Fold change) of the transcript level in PfGCN5-ΔBrD::GFP compared to WT 3D7, while the y-axis indicates -log₁₀ of the P values.(D) Number of genes with altered expression at different developmental stages of the IDC in the two domain deletion mutants. The up- and down- regulated genes are labeled in red and blue, respectively. (E, F) Volcano plots showing altered gene expression at the ring (E) and schizont (F) stages in PfPHD1-ΔPHD::GFP compared to the WT 3D7. The x-axis indicates log₂ (Fold change) of the transcript level in PfPHD1-ΔPHD::GFP compared to WT 3D7, while the y-axis indicates log₁₀ of the P values. (G, H) Overlaps of down-regulated (G) and up-regulated (H) genes between the PfGCN5-ΔBrD::GFP and PfPHD1-ΔPHD::GFP parasite lines at different stages. (I, J) The enriched functions or pathways of up-regulated and down-regulated genes upon PfGCN5 BrD (I) and PfPHD PHD deletion determined by GSEA. Each pixel represents the normalized enrichment score (NES). The positive and negative NESs represent up-and down-regulated functions after domain deletions, respectively.

https://doi.org/10.1371/journal.ppat.1009351.g004

With both PfGCN5 and PfPHD1 being integral members of the PfGCN5 complex, the global transcription changes resulting from the domain deletions of these two genes were remarkably similar; their transcriptomes showed a significant correlation between the respective stages with correlation coefficients ranging from 0.67 to 0.82 (S8E Fig). In addition, 44.1–63.8% of the down-regulated genes in each stage were shared between the two domain deletion strains (Fig 4G). In comparison, 36.1–49.2% of the up-regulated genes were shared between the two domain deletion mutants in the ring, early and late trophozoites, but this up-regulated gene repertoire in the two domain deletion strains only shared 18.8% during the schizont stage (Fig 4H).

BrD and PHD deletions affect parasite-specific cellular pathways

To determine whether deletion of PfGCN5 BrD or PfPHD1 PHD affected specific biological processes, Gene Set Enrichment Analysis (GSEA) [85,86] and gene ontology (GO) enrichment analysis (PlasmoDB) were performed on the genes with significantly altered expression (Figs 4I, 4J, S9 and S5 Table). The two analyses identified similar sets of enriched gene categories during the IDC. During the early IDC stages, transcripts associated with cytoadherence, merozoite invasion, DNA replication, and organellar activities were upregulated. In particular, the 60 var genes, encoding the virulent factor PfEMP1 that mediates cytoadherence, were dramatically upregulated in the early stages. At the early trophozoite stage, the overall transcripts of var genes in the WT 3D7 were sharply decreased. In contrast, the var transcripts remained at high levels in the domain deletion parasites (~4-fold higher upon BrD deletion and ~25-fold higher upon PHD deletion) (Fig 5A and 5B and S6 Table). It is noteworthy that many var members were upregulated, albeit var2csa was a major var gene expressed in PfGCN5-ΔBrD, suggesting activation of the overall var gene family. Western blots using antibodies against the conserved cytoplasmic ATS domain of the PfEMP1 proteins detected more complicated expression patterns and higher abundance of the PfEMP1 on the surface of RBC infected by trophozoite-stage parasites upon PfGNC5 BrD and PfPHD1 PHD deletions (Fig 5C and 5D). To further determine whether domain deletion in PfGCN5 and PfPHD1 activated multiple var members in a single infected RBC, we performed single-cell RNA-fluorescent in situ hybridization (FISH) using the Type B var exon 2 as the probe [35], which is predicted to hybridize to 22 Type B var genes. Var genes are clustered into 6–8 foci at the nuclear periphery and colocalized with the “telomere bouquets” [38], while the active var gene is localized to a var-specific expression site [87,88]. Consistent with the mutually exclusive expression of var genes in single cells, the majority of the RNA-FISH positive cells in 3D7 contained one fluorescent spot (mean ± standard deviation, 1.04±0.21, n = 67) indicating expression of one Type B var (Fig 5E). In contrast, 43.9% and 58.2% positive rings had more than one hybridization signal in the PfGCN5-ΔBrD (1.55±0.66, n = 66) and PfPHD1-ΔPHD parasites (1.91±0.98, n = 79), respectively. In addition, 28.7% (45/157) of the WT 3D7 ring-stage parasites showed hybridization, which increased to 34.5% (61/177) and 41.3% (71/172) in the PfGCN5-ΔBrD and PfPHD1-ΔPHD rings, respectively. It is noteworthy that the hybridization signals were mostly localized in areas at the periphery of the DAPI staining. Thus, these results indicate the presence of multiple var expression sites in the PfGCN5-ΔBrD and PfPHD1-ΔPHD parasites.

Download:

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

Fig 5. Virulence gene expression altered upon domain deletions.

(A) The expression of total var genes at different developmental stages. The y-axis indicates the total reads of var genes from RNA-seq. R, ring; ET, early trophozoite; LT, late trophozoite; S, schizont. (B) Pie graphs showing the overall levels of the var gene transcripts in the WT 3D7, PfGCN5-ΔBrD::GFP, and PfPHD1-ΔPHD::GFP parasite lines at the early trophozoite stage. The numbers in parentheses are the total expression of all var genes identified by RNA-seq analysis. (C) Western blot showing PfEMP1 protein levels in the iRBC membranes of the WT 3D7, PfGCN5-ΔBrD::GFP, and PfPHD1-ΔPHD::GFP parasite lines with the anti-ATS antibodies. Aldolase (ALD) expression was used as the loading control. (D) Bar graph showing the overall PfEMP1 protein levels in WT 3D7, PfGCN5-ΔBrD::GFP, and PfPHD1-ΔPHD::GFP parasite lines. Bars indicate means and standard deviations from five replicates. (E) Representative images of RNA FISH analysis showing the single locus of the B-type var gene expression in the WT 3D7 and more than one B-type var gene locus in the two deletion mutants.

https://doi.org/10.1371/journal.ppat.1009351.g005

Conversely, genes involved in the biological processes of translation and transcription were significantly enriched in the down-regulated genes upon BrD or PHD deletion during the early IDC, which was probably responsible for the slowing down of development (Figs 4I, 4J, S9 and S6 Table). During the late IDC (late trophozoites and schizonts), genes involved in RBC invasion were greatly reduced, which is consistent with the phenotype of reduced RBC invasion rates of the PfGCN5-ΔBrD and PfPHD1-ΔPHD merozoites (Figs 4I, 4J, S9 and S6 Table). Of the 86 putative invasion-related genes [89], 76 showed peak expression at the late stages of IDC in WT parasites (S10B Fig). Except for MSRP1, which was up-regulated, 75 genes were significantly down-regulated at the late stages in the PfGCN5-ΔBrD and PfPHD1-ΔPHD parasites (Figs 6A and S10B). In addition, other enriched gene categories such as “exit from host” and cell cycle were also down-regulated in late IDC (Figs 4I, 4J and S9). These data collectively indicate the involvement of the PfGCN5 complex in the regulation of the general cellular processes such as transcription, translation, and organellar function, as well as parasite-specific processes of pathogenesis and host cell invasion.

Download:

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

Fig 6. Down-regulation of invasion-related pathway and alteration of AP2 genes in domain deletion mutants.

(A) Heatmaps displaying down-regulation of genes involved in the invasion of the RBC in the two deletion mutants. * and # indicate the AP2-I and putative AP2-LT target genes, respectively. Ø and Ψ indicate the down-regulated genes upon knockdown of BDP1 and the target genes identified by BDP1-HA ChIP-seq analysis, respectively. (B) Heatmaps showing altered expression of the AP2 genes in the deletion mutants. (C) Putative AP2-LT target genes (with AP2-LT binding motifs) are significantly enriched in the down-regulated genes in PfGCN5-ΔBrD::GFP at the late stages. The numbers of up- and down-regulated genes upon PfGCN5 BrD domain deletion are shown as the red and blue bars at four developmental stages. These altered genes were further classified as the putative target genes of AP2-LT and the remaining genes. R, ring; ET, early trophozoite; LT, late trophozoite; S, schizont. ***, P <0.001 (Fisher’s exact test).

https://doi.org/10.1371/journal.ppat.1009351.g006

Potential coordination of the PfGCN5 complex by AP2 transcription factors and BDP1

The profound effects of PfGCN5 BrD/PfPHD1 PHD deletion on particular pathways such as invasion and cell adhesion suggest that these coordinated changes may involve the participation of specific TFs of the ApiAP2 domain family [90,91]. The consistent pulldown with the PfGCN5 complex of the AP2-LT that is expressed abundantly in late stages of the IDC and occasionally other AP2 proteins (Figs 1A and 3D) [62] suggests coordination of chromatin modification with transcription activators/repressors. To this end, we analyzed the transcriptional changes of all 27 ApiAP2 TFs after BrD or PHD deletion (Figs 6B and S10C). Indeed, the cascade of AP2 transcriptions was substantially disturbed; some AP2-TFs such as AP2-SP2 and AP2-O3 were activated at stages when they are normally silenced in WT parasites, whereas some (e.g., AP2-SIP2, AP2-O, AP2-I, and AP2-LT) were down-regulated at the stages when they are supposed to be active (Figs 6B and S10C and S6 Table). PfSIP2 is associated with the chromosomal end clusters and is required for heterochromatin formation and genome integrity, including the silencing of subtelomeric var genes [92]. Its down-regulation upon BrD and PHD domain deletion may influence the organization of the subtelomeric heterochromatin, resulting in the overall de-repression of var genes (Figs 6B and S10C). The master regulator of gametocytogenesis AP2-G [93,94] was consistently up-regulated upon Brd and PHD deletions (Fig 6B), which agrees with the increased gametocytogenesis detected in the mutant parasite lines (Fig 2G). The down-regulation of AP2-I and AP2-LT might partially explain the down-regulation of the invasion-related gene upon domain deletions. Of the 75 invasion-related genes down-regulated upon BrD or PHD deletion, 19 (mostly associated with the rhoptry) are targets of AP2-I, and 33 were predicted targets of AP2-LT (Fig 6A and S6 Table) [33,78]. The AP2-LT subunit of the PfGCN5 complex is predicted to bind to 986 genes with the motif sequence ACACA [33]. Analysis of the genes altered upon BrD deletion in PfGCN5 and PHD domain deletion in PfPHD1 revealed that genes down-regulated in late trophozoite and schizont stages were significantly enriched in those containing the AP2-LT binding motif (Figs 6C and S10D). This finding is consistent with the decreased recruitment of the PfGCN5 complex in the deletion mutants by AP2-LT to genes with AP2-LT binding motifs, resulting in extensive down-regulation of the gene categories during the late-stage development (Fig 6B). In contrast, genes up-regulated at the ring and trophozoite stages were significantly enriched in those containing the AP2-LT binding motif (Figs 6C and S10D), in agreement with the up-regulation of AP2-LT at the early stages upon domain deletion (Fig 6B). Thus, the unique presence of AP2 TFs in the GCN5 complexes in apicomplexan parasites suggests that the GCN5 complexes are specifically recruited to regulate the expression of certain clusters of genes [11].

The parasites with PfGCN5/PfPHD1 domain deletion displayed a similar invasion phenotype with the BDP1 knockdown parasites [77]. Of the down-regulated invasion genes upon BrD and PHD domain deletions, 24 and 34 are BDP1 targets identified by BDP1 knockdown and BDP1 chromosome immunoprecipitation (ChIP)-seq analysis, respectively (Fig 6A). BDP1 regulates these invasion-related genes by binding to the acetylated H3 at their TSSs. Thus, the disassociation between the PfGCN5 complex and BDP1 upon domain deletions (Fig 3D) may underlie such a phenotypic similarity. Additionally, BDP1 knockdown resulted in the down-regulation of 47 genes at the later stage of IDC (TP5) [77] and a significant proportion of these genes were also down-regulated in the two domain deletion mutants at the schizont stage (Fig 6D and 6E, p <0.001, Fisher’s exact test, compared to the expectation from the whole genome scale). Furthermore, the potential target genes of BDP1 and AP2-I identified by ChIP-seq also showed substantial overlaps with genes down-regulated at the schizont stage upon BrD and PHD domain deletions (Fig 6F and 6G). These data highlight the connection between the recruitment of the PfGCN5 complex to specific gene groups by the AP2 domain TFs and the BrD-containing reader complex in orchestrating transcription regulation.

BrD and PHD deletions broadly alter chromatin structure

The above transcriptomic analyses showed that the expression of many genes was disturbed in the PfGNC5-ΔBrD and PfPHD1-ΔPHD parasites; genes that are normally active were down-regulated, whereas genes supposed to be silent were active at the wrong time during the IDC. Since epigenetic regulation of gene expression in P. falciparum is most evident in the heterochromatic regions [27], while gene expression from the euchromatic regions correlates positively with the chromatin accessibility [52], we compared the chromatin status and accessibility of genes with altered expression upon PfGCN5 and PfPHD1 domain deletions. We first compared the up- and down-regulated genes with the accessibility of their promoters previously determined using the published assay for transposase accessible chromatin sequencing (ATAC-seq) [52]. Based on the value of ATAC-seq peak in the promoter (chromatin openness level), the down-regulated genes upon BrD or PHD deletion had more open chromatin structure at their promoters in the WT parasites, whereas the up-regulated genes had less open promoters in the WT parasites (Fig 7A). Conversely, when the genes with altered expression were compared with the published HP1 occupancy signals, the up-regulated genes upon BrD or PHD deletion are significantly more often associated with the heterochromatin loci that are normally enriched with HP1 and repressed during the IDC [95] (Fig 7B and S7 Table). This group of genes includes many variant gene families (var, rifin and stevor) and AP2-G (Figs 5B and 6B and S7 Table). In addition, genes specific for sexual-stage development, which usually are silent during the IDC, are also among the genes that were up-regulated upon BrD or PHD deletion but have low accessibility in their promoters during the IDC in the WT parasites. BrD deletion led to significant up-regulation of 353 gametocyte- and 401 ookinete-specific genes, respectively (Figs 7C, 7D, S10E and S8 Table), and many were up-regulated at the ring stage. Similarly, PHD deletion caused up-regulation of 403 gametocyte- and 401 ookinete-specific genes, respectively (Figs 7C, 7D, S10E and S8 Table). Among them, 151 gametocyte- and 199 ookinete-specific genes are shared between both deletional mutants. Taken together, both domain deletions similarly affected chromatin structure and led to the activation of genes involved in sexual development.

Download:

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

Fig 7. Correlation of genes showing altered expression in domain deletion mutants with promoter accessibility and chromatin states.

(A, B) Changed levels of gene expression in PfGCN5-ΔBrd and PfPHD1-ΔPHD are negatively correlated with the accessibility of the promoters (from the ATAC-seq analysis, The y-axis shows that the ratio of ATAC-seq peak (scaled per million reads) vs genomic DNA library) (A), but are positively correlated with the heterochromatin state (represented by the HP1 occupancy) (B). U, upregulation; D, downregulation. *, P <0.05; ****, P < 0.0001 (Wilcoxon rank-sum test). (C) Heatmaps displaying the transcriptional activation of gametocyte and ookinete genes in the domain deletion mutants. R, ring; ET, early trophozoite; LT, late trophozoite; S, schizont. (D) Overlaps of activated gametocyte- or ookinete-specific genes between PfGCN5-ΔBrd and PfPHD1-ΔPHD.

https://doi.org/10.1371/journal.ppat.1009351.g007

To verify that the PfGCN5 complex directly regulated the accessibility and chromatin state of the promoters, we selected three genes to evaluate the recruitment of the complex by ChIP with antibodies against the tags fused to PfGCN5 or PfPHD followed by qPCR analysis, and the results were compared to those with respective control antibodies (S11 Fig). Compared to the WT parasites, the truncated PfGCN5-ΔBrD and PfPHD1-ΔPHD were significantly depleted at the PfMSP1 promoter in schizonts, whereas they were enriched at the ring stage (S11 Fig). Their dynamic associations with the PfMSP1 promoter were correlated with the down- and up-regulation of PfMSP1 at the schizont and ring stages, respectively, in these two domain deletion lines. Consistently, the PfGCN5-ΔBrD and PfPHD1-ΔPHD proteins were significantly enriched at the promoter of the var2csa gene and the gametocyte-specific gene Pfg27/25 at the ring stage (S11B and S11C Fig), which was correlated with significant up-regulation of these two genes in the domain deletion parasites. To determine whether activation of genes located in the heterochromatin after domain deletions were due to the alteration of chromatin states, we selected VAR2CSA and a RIFIN gene, which were significantly up-regulated in domain deletion mutants at the ring stage, for the evaluation of the H3K9me3 and H3K9ac levels by ChIP-qPCR. In general, domain deletions decreased the heterochromatin mark H3K9me3 at the promoters of these genes, whereas domain deletions elevated the H3K9ac mark (S12 Fig).

Parasite culture

The P. falciparum strain 3D7 and its genetically modified clones were cultured at 37°C in a gas mixture of 5% CO₂, 3% O₂ and 92% N₂ with type O⁺ RBCs at 5% hematocrit in RPMI 1640 medium supplemented with 25 mM NaHCO₃, 25 mM HEPES, 50 mg/L hypoxanthine, 0.5% Albumax II and 40 mg/ml gentamicin sulfate [112]. Synchronization of asexual stages was performed by two rounds of sorbitol treatment at the rings stage or by incubation of schizonts with RBCs for 3 h to obtain highly synchronized ring-stage parasites [113].

Genetic manipulation of PfGCN5 and its associated genes

To tag the C-terminus of PfGCN5 with the PTP tag [65,66], the C-terminal PfGCN5 fragment [nucleotides (nt) 3778–4758] was amplified from P. falciparum genomic DNA using primers F1 × R1. For the deletion of the PfGCN5 BrD and the PfPHD1 PHD domain, the PfGCN5 fragment (nt 3286–4044) and the PfPHD1 fragment (nt 3286–4044) were amplified using primers F2 × R2 and F3 × R3, respectively. All amplified fragments were first cloned into a modified pBluescript SK plasmid to fuse with the PTP or GFP and pDT 3’ UTR as described earlier [57,114]. This cassette was then subcloned into pHD22Y at BamHI and NotI sites to produce pHD22Y/PfGCN-PTP, pHD22Y/GCN5-ΔBrD-GFP, and pHD22Y/PHD1-ΔPHD-GFP, respectively. A similar strategy was used to tag PfPHD1 with c-Myc and PF3D7_1019700 with GFP. All primers used for tagging, domain deletion, PCR verification of integration and probe for Southern blot are listed in S9A Table.

Parasite transfection was done using the RBC loading method [115]. Briefly, 100 μg of plasmid were introduced into fresh RBCs by electroporation. Purified schizonts were used to infect the RBCs pre-loaded with the plasmid and selection was done with 2.5 nM of WR99210 for approximately 4 weeks with weekly replenishment of fresh RBCs until resistant parasites appeared. Resistant parasites were subjected to three cycles of drug on-off selection and single clones of parasites with stable integration of the constructs were obtained by limiting dilution [113]. For the parasites transfected with constructs containing a GFP tag, fluorescence-activated cell sorting was employed to clone the GFP-positive parasites. Correct integrations of plasmids into the parasite genome were screened by integration-specific PCR or Southern blot with the digoxigenin (DIG)-labeled probes using an established protocol [80].

Purification of protein complexes

TAP was performed using the PTP-tagged PfGCN5 parasite line according to the published method [66]. Briefly, 10⁹ parasites were lysed in 5 volumes of the hypotonic buffer (10 mM HEPES, pH 7.9, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM DTT, 0.5 mM EDTA) at 4°C for 10 min followed by centrifugation for 20 min at 500 × g. The resultant pellet (nucleus) was further lysed in 5 volumes of PA150 buffer (150 mM KCl, 20 mM Tris-HCl, pH 7.7, 3 mM MgCl₂, 0.5 mM DTT, and 0.1% Tween 20) containing a protease inhibitor cocktail (Roche). The lysate was centrifuged for 10 min at 16,000 × g and the supernatant was incubated with 100 μl (settled volume) of IgG agarose beads (GE Healthcare) at 4°C for 2 h. The beads were washed twice with PA150 and equilibrated twice with the TEV buffer (150 mM KCl, 20 mM Tris-HCl, pH 7.7, 3 mM MgCl₂, 0.5 mM EDTA, 1 mM DTT, and 0.1% Tween 20). To release the PfGCN5 and its associated proteins from the IgG beads, the beads were incubated with 2 ml of TEV buffer containing 150 U of TEV protease and rotated overnight at 4°C. The supernatant was collected, and the beads were rinsed with another 4 ml of the PC150 buffer (150 mM KCl, 20 mM Tris-HCl, pH 7.7, 3 mM MgCl2, 1 mM CaCl₂, 0.1% Tween 20). Then, 7.5 μl of 1 M CaCl₂ were added to titrate the EDTA from the TEV buffer and the combined supernatant was incubated with the anti-protein C beads for 2 h at 4°C. The beads were washed four times with PC150 and eluted with the buffer containing 10 nM EGTA/5 mM EDTA. For the single-step pulldown of GFP-tagged or Myc-tagged protein, GFP- or Myc-trap (Cat# gta-20, RRID:AB_2631357 or Cat# yta-20, RRID:AB_2631369, Chromotek) beads were used with lysates from 10⁹ parasites according to the manufacture’s protocol.

Mass spectrometry

The proteins in the elution were concentrated by Amicon Ultra centrifugal filters (Millipore Sigma) and separated briefly in a 10% Bis-Tris SDS-PAGE gel for 10 min. Proteins in gel were excised, in-gel digested, and analyzed by nano-LC/MS/MS using a Waters NanoAcquity HPLC system interfaced to a Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific) [116]. Peptides were loaded on a trapping column and eluted over a 75 μm analytical column at 350 nL/min. MS and MS/MS were performed at 70,000 FWHM and 17,500 FWHM resolutions, respectively. The fifteen most abundant ions were selected for MS/MS. Parasite proteins were identified by searching the Uniprot P. falciparum protein database (v01/2014). Data were filtered at 1% protein and 0.2% peptide FDR, and at least two unique peptides per protein. Mascot DAT files were parsed into the Scaffold software for validation and filtering to create a non-redundant list per sample. The available mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [117] partner repository with the dataset identifier PXD023389 and 10.6019/PXD023389.

Antibody generation

To generate antibodies against PfPHD1 and PfPHD2, a PfPHD1 peptide (DNGKLQKVDGRKKRRYHK, aa 3685–3702) and a PfPHD2 peptide (DDNVKAEDYKDENNDNDGD, aa 5738–5756) were synthesized and rabbits were immunized with these peptides. After three times immunizations, the antibodies were purified by affinity purification with peptides conjugated to the beads (Proteintech Group).

Gel filtration

To estimate the size of the PfGCN5 complex, nuclear extract from the PfGCN5::PTP was incubated with IgG beads, eluted by TEV protease cleavage as described above, and applied to a Superose 6 gel filtration column (GE Healthcare). Molecular mass standards (Gel Filtration Calibration Kit HMW, GE Healthcare) were run under the same conditions to estimate the size of the complex. The fractions were analyzed by Western blotting using antibodies against the PTP tag, PfPHD1 and PfPHD2, while HAT activity in the fractions was measured as described previously [57].

Growth phenotype analysis

The growth phenotypes of GCN5-ΔBrD::GFP and PHD1-ΔPHD::GFP lines during the IDC were compared with the WT 3D7 parasites as described [57]. To measure cell cycle progression, highly synchronous rings were obtained by incubation of purified schizonts with RBCs for 3 h. Progression of parasites through the IDC was monitored using Giemsa-stained smears every 2 h. Cycle time was determined as the duration between the peak ring parasitemias of two consecutive cycles. To measure parasite proliferation, synchronous cultures after two rounds of consecutive synchronization by sorbitol were initiated at 0.1% rings, and parasitemia was monitored daily for 7 days without replenishment of the RBCs. The number of merozoites produced per schizont was determined from mature segmenters. Three independent biological replications were done for each parasite line. Merozoite invasion assay was performed as described earlier [118]. Briefly, ∼1×10⁵ of purified schizont from the WT, PfGCN5-ΔBrD::GFP and PfPHD1-ΔPHD::GFP lines were mixed with ∼1×10⁷ fresh RBCs, and the parasitemia of culture was determined 24 h later. The invasion rate was calculated as the percentage of merozoites invaded into RBCs assuming that each ruptured schizont releases an average of 10 merozoites using the formula: ring parasitemia × [(RBCs per μl)/(schizonts per μl ×10)]. To measure the gametocyte development, gametocyte induction was conducted by using an established method [80,119], and the gametocytemia was determined by counting gametocytes in Giemsa-stained thin blood smears at the middle developmental stage (stage III).

Histone modifications

To estimate histone modifications in the domain deletion mutants, histones were purified from the WT, GCN5-ΔBrD::GFP and PHD1-ΔPHD::GFP lines [120]. Equal amounts of the histones at each developmental stage were separated by 15% SDS/PAGE and transferred to nitrocellulose membranes. Western blotting was performed using a standard procedure with anti-acetyl histone H3, H3K9Ac (Catalog no. 07–352, RRID:AB_310544, Millipore), anti-tri methyl histone H3, H3K4me3 (catalog no. 07–473, RRID:AB_1977252, Millipore) and anti-acetyl histone H4, H4Ac (catalog no. 06–598, RRID:AB_2295074, Millipore) at 1:1000 dilution as the primary antibodies and horseradish peroxidase-conjugated goat anti-rabbit IgG (diluted at 1:2000) as the secondary antibodies. The detected proteins were visualized using an enhanced chemiluminescence (ECL) kit (Invitrogen).

Immunofluorescence assay (IFA)

IFA was performed as described [121,122]. The parasitized RBCs were washed once with PBS and the cell pellet (∼100 μl) was fixed with 1 ml of 4% (v/v) paraformaldehyde and 0.0075% (v/v) glutaraldehyde in PBS for 30 min followed by 10 min quenching with 50 mM glycine in PBS. Fixed cells were washed twice with PBS and treated with 0.5% (v/v) Triton X-100 in PBS for 10 min. Then, cells were washed twice with PBS and blocked in 3% (v/v) BSA for 1 h at room temperature. The anti-PfPHD1 (1 μg/ml), PfPHD2 antibodies (1 μg/ml), goat anti-GFP (1:2000; ab6673; Abcam, RRID:AB_305643, USA), mouse anti-Myc (ab56; Abcam, USA) and rabbit anti-H3K9ac (1:1000; 06–942, RRID:AB_310308, Millipore, USA) antibodies in PBS containing 1% BSA were added and incubated for another 1.5 h. After washing the cells three times with PBS, FITC-conjugated goat anti-rabbit IgG antibodies (Cat# F6005, RRID:AB_259682, Sigma, USA), Alexa fluor 488-conjugated secondary donkey anti-goat IgG antibody or Alexa fluor 594-conjugated secondary goat anti-rabbit antibody IgG antibody (A32814 RRID:AB_2762838 and R37117 RRID:AB_2556545, Thermo Fisher Scientific, USA) were added at 1:2000 dilution in 3% (v/v) BSA and incubated for 45 min. Nuclear staining was performed by incubating slides with 4’,6-diamidino-2-phenylindole (DAPI, final 0.5 μg/mL; Invitrogen). Images were captured using an epifluorescence microscope (Nikon Eclipse Ni, USA; 100x/1.4 oil immersion lens) and were processed by Adobe Photoshop CS (Adobe Systems Inc. San José, CA). To quantitate co-localizations, images from at least 20 parasites were randomly selected, analyzed by ImageJ (1.52a; http://imagej.nih.gov/ij), and Pearson’s coefficients were calculated.

Transcriptome analysis

To compare the transcriptomes during the IDC among the WT, GCN5-ΔBrD::GFP and PHD1-ΔPHD::GFP lines, RNA-seq was performed. Three replicates of total RNA from parasites at ring, early trophozoite, late trophozoite and schizont stages were harvested by using the ZYMO RNA purification kit, and used to generate the sequencing libraries using the KAPA Stranded mRNA Seq kit for the Illumina sequencing platform according to the manufacturer’s protocol (KAPA biosystems). Libraries were sequenced on an Illumina HiSeq 2500 in the Rapid Run mode using 100 nt single read sequencing. Reads from Illumina sequencing were mapped to the P. falciparum genome sequence (Genedb v3.1) using HISAT2 [123]. FeatureCounts was utilized to calculate the read counts for each transcription in P. falciparum while the differential expression analysis was performed by DESeq2 [124,125] with the criteria of P-adjustment of <0.01 as the cutoff. RNA-seq data were also normalized by TPM to further validate the DEseq2 results. The differential expressed genes between domain deletion and WT are selected according to the following criteria: 1) the absolute fold change of TPM higher than 1 in all three biological replicates, and 2) P-adjustment from the DEseq lower than 0.01. The GO enrichment for up-regulated and down-regulated genes after BrD or PHD deletion was performed on PlasmDB (https://plasmodb.org/plasmo/). GSEA was performed as described [85,86] and the normalized enrichment score (NES) was used to measure the function enriched for a specific expression gene set. RNA-Seq data were submitted to NCBI GEO repository (accession number GSE164070).

Phaseogram of the transcriptomes of P. falciparum IDC

The sine wave model was utilized here to model the gene expression timing [126]. The gene transcription level from RNA-seq was first normalized as TPM (transcripts per million). Only the differential expressed genes in PfGCN5-ΔBrd or PfPHD1-ΔPHD as compared with WT were considered for the analysis. The TPM of each gene E(t) was modeled as (1) where E(t) = [TPM_12h, TPM_24h, TPM_36h, TPM_48h] is the TPM at the t = [12, 24, 36, 48] hours of sample collection, ω is the angular frequency and given by ω = 2π/48, A is the amplitude of the expression profile, and C is the vertical offset of the profile from zero. To identify the parameter α and A, A×sin(ωt−α) are changed to (2)

Then the R command lm was used to fit a linear regression model between E(t) and sin(ωt)+cos(ωt). The fitting coefficient from the lm result indicates A×cos(−α) and A×sin(−α). The α and A were calculated as (3) (4)

The α indicates the horizontal offset of the profile from zero, which is used in the phaseogram to order the gene in the heatmap.

Association between chromatin structure and transcriptomic changes upon domain deletions

To quantify the association between open chromatin accessibility and transcriptional changes upon BrD deletion in PfGCN5 or PHD domain deletion in PfPHD1, we retrieved the ATAC-seq profile showing ATAC-seq peaks upstream the TSSs [52]. Each TSS was assigned to the nearest ATAC-seq peak with a distance restriction lower than 1 kb. The values of chromatin accessibility (ATAC-seq RPM + 0.1)/(gDNA RPM + 0.1) were then compared to the altered expression from the up- or down-regulated genes after domain deletions, where RPM represents the scaled reads per million reads. To investigate the association between the PfHP1 occupancy and transcriptional changes in the domain deletion mutants, the PfHP1 values (ChIP/input ratio) along the coding sequence were downloaded [95] and compared with the altered expression from the up and down-regulated gene after domain deletions.

ChIP quantitative PCR

ChIP-qPCR was performed as described [56,78,95] with some modifications. Synchronized GCN5-ΔBrD::GFP and PHD1-ΔPHD::GFP parasite lines at the ring stage [10–16 h post-invasion (hpi), ~5 × 10⁹ infected RBCs (iRBCs)] and schizont stage (40–46 hpi, ~1.5 × 10⁹ iRBCs) were harvested and crosslinked with paraformaldehyde (1% final concentration; EMS, USA) at 37°C for 15 min with agitation and then immediately neutralized by adding glycine (0.125 M final concentration) on ice for 5 min with agitation. The fixed iRBCs were lysed with saponin (0.06% final concentration; Sigma, USA) on ice for 5–10 min. Parasites were treated with a lysis buffer (10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM DTT, 10 mM Hepes pH 7.9, 1 × Protease inhibitor) and then gently homogenized using a douncer to free nuclei. Pelleted nuclei were suspended in a shearing buffer (0.1% SDS, 5 mM EDTA, 50 mM Tris-HCl, pH 8.1, 1X Protease inhibitor) [95]. Sonication was performed using a rod bioruptor (Microson ultrasonic cell disruptor, Misonix, Inc. USA) at high power for 20 cycles of 30 sec ON/30 sec OFF, resulting in sheared chromatin of approximately 100–1000 bps. 50 μl of input samples was set aside before the remaining chromatin was diluted in incubation buffer (0.01% SDS, 1.5% Triton X-100, 0.5 mM EDTA, 200 mM NaCl, 5 mM Tris-HCl, pH 8.1). The chromatin (75 μl/400 ng) was incubated with goat anti-GFP (ab6673; Abcam), mouse anti-Myc (IgG2a monoclonal, ab56; Abcam), and their respective control antibodies (goat IgG, ab37373, mouse IgG2a, ab18413, Abcam) overnight at 4°C while rotating followed by the addition of 20 μl of agarose beads for 1 h. Beads were then washed for 5 min at 4°C while rotating with the following: buffer 1 (0.1% SDS, 1% Triton X-100, 150 mM NaCl, 2 mM EDTA, 20 mM Tris HCl, pH 8.1); buffer 2 (0.1% SDS, 1% Triton X-100, 500 mM NaCl, 2 mM EDTA, 20 mM Tris HCl, pH 8.1), buffer 3 (250 mM LiCl, 1% NP-40, 1% Na-deoxycholate, 1 mM EDTA, 10 mM Tris HCl, pH 8.1) and finally twice with buffer 4 (10 mM EDTA, 10 mM Tris HCl, pH 8). The immunoprecipitated (IPed) chromatin was eluted with the elution buffer (1% SDS, 0.1M NaHCO₃) at room temperature for 15 min with rotation. The eluted chromatin and input samples were reverse cross-linked in 10% SDS, 1 M NaHCO_3, 5 M NaCl, 10% Triton X-100 at 45°C overnight while shaking and purified by the phenol:chloroform method. For qPCR, the concentration of immunoprecipitated gDNA was determined by Qubit dsDNA Broad-Range Assay Kit (Invitrogen, USA), and 10 ng per well in triplicate were used for qPCR using the FastStart Universal SYBR Green Master [Rox] (Sigma-Aldrich, USA) as described [56]. Primer pairs targeting 5′UTRs were designed to amplify fragments less than 200 bp (S9B Table). Fold enrichment relative to constitutively expressed reference gene seryl-tRNA synthetase (stRNA, PF3D7_0717700) was calculated by using the 2^−ΔΔCt method [127]. The fold changes of binding enrichment were calculated using a formula: 2^{−[(IP Ct-target − IP Ct-stRNA)−(control IP Ct-target–control IP Ct-stRNA)]} for each primer set targeting specific promoter regions.

RNA fluorescent in-situ hybridization (FISH)

RNA FISH was performed as described [35]. Briefly, purified ring-stage parasites were lysed with saponin and released parasites fixed in suspension with ice-cold 4% paraformaldehyde. Parasites were then deposited on Teflon-coated microscope slides and hybridized with denatured var probes at 42°C for at least 16 h. All the FISH probes were PCR amplified from genomic DNA using the primers listed in S9B Table. The slides were then washed three times in 2×SSC at 42°C. Finally, the slides were incubated with streptavidin-488 antibody at room temperature for 30 min. Images were taken using a Nikon ECLIPSE E600 epifluorescence microscope. NIS Elements 3.0 software was used for acquisition and ImageJ for composition.

Statistical analysis

For all experiments, three or more independent biological replicates were performed. The results are presented as mean ± SD. Results are regarded significant if P < 0.05 as established by ANOVA, Fisher’s exact test, paired Mann Whitney U test or paired Wilcoxon test, and the respective analysis was shown in the figure legends. To analyze the schizont numbers containing different numbers of merozoites, a χ² goodness of fit test was first used to evaluate if the number of schizonts that contain a certain number of merozoites was independent of the parasite lines. Then the proportions of schizonts with a certain number of merozoites were compared among these cell lines based on ANOVA for each merozoite number.