Identification of pfap2-g transcription start sites (TSSs)

To guide the design of the experiments to assess HC nucleation, ensuring that fragments containing the pfap2-g (PF3D7_1222600) TSSs were included, we first mapped the TSSs of this gene using 5’ Rapid Amplification of cDNA Ends (5’RACE). This was performed with the E5 line [39] (a 3D7 subclone with a high sexual conversion rate) and the transgenic E5-PfAP2-G-DD line, which has a destabilization domain appended to PfAP2-G and shows unusually high sexual conversion rates when the stabilizing ligand Shield-1 (Shld1) is added [39,43]. In both parasite lines, we consistently identified two TSSs blocks in the pfap2-g upstream region, Block 1 and Block 2, located ~1,900 and ~1,600 bp from the start codon, respectively (Fig 1). Of note, previous genome-wide studies on transcription initiation did not identify any TSSs for pfap2-g [44,45], likely because the gene was only expressed in a small fraction of the cells in the parasite lines used.

Download:

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

Fig 1. Identification of pfap2-g transcription start sites (TSSs).

Position relative to the start codon and frequency (number of clones) of the TSSs identified by the 5’RACE analysis of pfap2-g in the E5 and E5-AP2-G-DD (E5-DD) lines. The E5-DD line was cultured in the presence of Shld1. TSSs were identified from experiments in which PCR was performed with either LA Taq or KAPA DNA polymerases, as indicated. The position of the two main TSS blocks (B1 and B2) and the region containing the previously described PfAP2-G binding motifs is indicated in the schematic (to scale) below.

https://doi.org/10.1371/journal.ppat.1013137.g001

To validate the predicted TSSs identified by 5’RACE, we performed a reverse transcriptase coupled to quantitative PCR (RT-qPCR) analysis using primers located upstream, between or downstream of the two blocks. This analysis confirmed that essentially no transcripts contained sequences upstream of Block 1 and the majority of transcripts contained sequences located between the two blocks (S1 Fig). The irregular pattern observed in the transcript levels measured at different positions may be explained by the occurrence of different mRNA species or by technical limitations associated with the extreme AT-richness of this region.

Identification of regions of the pfap2-g locus with capacity to nucleate HC de novo

To identify specific regions of the pfap2-g locus able to nucleate HC de novo, we designed a screening approach based on the integration of pfap2-g fragments into the PF3D7_1144400 locus [16], which is euchromatic (S2A Fig) and encodes a gene dispensable in asexual blood stages [46]. The resulting transgenic lines were analyzed using H3K9me3 chromatin immunoprecipitation (ChIP) coupled to quantitative PCR (ChIP-qPCR) to evaluate HC levels at the site of integration. The analysis of each transgenic line also included heterochromatic and euchromatic genes used as ChIP-qPCR positive and negative controls, respectively (S2A Fig).

To detect HC formation within the ectopically integrated fragments without interference from the endogenous locus and to prevent unwanted plasmid integration via recombination with the endogenous pfap2-g gene, we first deleted the upstream regulatory region and part of the coding sequence (CDS) of the endogenous pfap2-g gene (positions -3,522 to +992 bp from the start codon) in the 3D7-A subclone 1.2B. This subclone does not form gametocytes due to a non-sense mutation [47] in gdv-1, an upstream activator of pfap2-g [48]. The resulting parasite line, which we termed Δ5’ap2-g, was subsequently edited to integrate the DNA fragments under analysis into the PF3D7_1144400 locus (Fig 2A). The parasite lines used in this study are described in S2B Fig.

Download:

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

Fig 2. Identification of regions of the pfap2-g locus that can nucleate HC de novo.

(A) Left: Strategy for the deletion of the upstream region and part of the CDS of pfap2-g using CRISPR/Cas9 (Δ5’ap2-g line). The scissors indicate the positions targeted by single guide RNAs (sgRNAs). Homology regions (HR1 and HR2) are indicated as boxes. A non-functional gfp (gfp*) gene was used to replace the deleted region. Right: Strategy for the integration of different fragments into the euchromatic locus PF3D7_1144400 using CRISPR/Cas9. Fx refers to the fragment to be integrated. Black arrowheads indicate the position of the primers used for diagnostic PCR, and colored arrowheads the position of the primers used for ChIP-qPCR analyses. (B) Schematic (to scale) of the position of the ~ 1 kb pfap2-g, ama1 and var (PF3D7_1240300) fragments tested for HC nucleation capacity. The transgenic line with the var fragment was generated with plasmids containing the yfcu marker but, for the experiments presented here, it was analyzed before negative selection, which makes it analogous to the other transgenic lines. Square brackets indicate the region of the endogenous pfap2-g locus that was deleted. The previously described position of the HC domains in pfap2-g and the var gene is indicated (HC) [16]. The position of the TSSs and the region containing the PfAP2-G binding motifs is also indicated. (C) H3K9me3 ChIP-qPCR analysis of the transgenic lines with the integrated fragments (position of primers shown in panel A). Values are the % of the DNA recovered in the H3K9me3 IP (% input) at each position relative to the average % input in the positive control heterochromatic genes var PF3D7_1240300 (or mspdbl2 for the 1240300 transgenic line) and clag3.2 (“HC genes”). “EU genes” are euchromatic genes actin I (actI) and serine-tRNA ligase (serrs) used as negative controls (average of the two is shown). The dashed line indicates the coverage in the positive control heterochromatic genes. Data are presented as the average and s.d. of two biological replicates. (D) qPCR analysis of the number of copies of the PF3D7_1144400 HRs (5’HR and 3’HR) and the fragments under analysis in the gDNA of the transgenic lines. Copy number was calculated relative to 1.2B (wild type) gDNA. The dashed lines indicate the expected value for single, correct integration. For fragments of which the endogenous copy was deleted (i.e., pfap2-g fragments), the expected value is 1. For fragments of which the endogenous copy is present in the genome, the expected value is 2.

https://doi.org/10.1371/journal.ppat.1013137.g002

We first analyzed by ChIP-qPCR five transgenic lines in which we integrated overlapping ~1 kb fragments (F0, F1, F2, F3 or F4) of the pfap2-g upstream region and beginning of the CDS (Fig 2B), which are heterochromatic in the endogenous pfap2-g locus [15–18]. The sequence of fragments F1-F4 falls within the region of the endogenous pfap2-g locus that was deleted in the Δ5’ap2-g line, whereas part of the F0 sequence was still present at the endogenous locus (Fig 2B). Fragments from the CDS (F0 and F1) and, to a lesser extent, from the proximal upstream region (F2) promoted HC formation, although with a lower coverage than in the ChIP-qPCR positive controls (Figs 2C and S2C). HC was detected within the fragment itself and in the adjacent regions of the PF3D7_1144400 locus, indicating that the newly formed HC was able to spread from the nucleation site into flanking regions. In contrast, fragments from the distal part of the pfap2-g upstream region (F3 and F4) did not promote HC formation, as evidenced by similar H3K9me3 levels as in the ChIP-qPCR negative controls (Fig 2C). Additionally, integration of a fragment of similar size from the upstream region of the euchromatic gene ama1 did not promote HC formation, whereas integration of a fragment from the heterochromatic var gene PF3D7_1240300 resulted in prominent HC formation, in this case with a coverage similar to the ChIP-qPCR positive controls (Fig 2B and 2C). The euchromatic or heterochromatic state of the endogenous ama1 and var genes was confirmed by ChIP-qPCR (S2A Fig). To further dissect the sequences that can nucleate HC, we generated additional transgenic lines with shorter fragments of the pfap2-g F0 and F1 regions integrated, but clear HC formation was not observed for any of them (S3 Fig). We cannot exclude the possibility that HC was formed in a small fraction of the parasites with these or the F3 or F4 fragments integrated, which would not be detectable in a bulk analysis that reflects the average H3K9me3 occupancy in the population, but this would correspond to highly inefficient nucleation.

Unexpectedly, qPCR and Southern blot analysis of genomic DNA (gDNA) of the transgenic lines revealed integration of multiple copies of the fragment in all cases except for the lines with the pfap2-g F4 or var gene ~1 kb fragments (Figs 2D and S4) and two of the lines with shorter fragments (S3 Fig). Although the qPCR results are compatible with either integration of concatemers (containing multiple copies of the fragments) or presence of episomal copies of the donor plasmid, the complex pattern of bands observed in the Southern Blot analysis (S4 Fig) suggests integration of concatemers. While we cannot exclude the possibility that episomal copies of the plasmids were also present and accounted for part of the HC signal detected, HC formation at the F0 and F1 fragments integrated in the genome is demonstrated by the detection of H3K9me3 with the external primer pairs 5’ext. and 3’ext., which recognize sequences adjacent to the site of integration that are not present in the transfection plasmids (Fig 2A and 2C). While H3K9me3 coverage at these positions was lower than in the homology regions or the fragment itself, it was clearly higher than in ChIP-qPCR euchromatic controls. On the other hand, we cannot rule out the possibility that integration of concatemers containing multiple copies of the fragments affects their ability to nucleate HC formation, although the different ability of different fragments to form HC is suggestive of sequence specificity.

Long fragments are needed for efficient HC nucleation by pfap2-g sequences

To characterize the potential effect of integration of multiple copies of the fragments on de novo HC nucleation, the pfap2-g fragments F0 and F1 were integrated at the PF3D7_1144400 locus using donor plasmids that carried the yfcu negative selection marker. This enabled elimination of parasites carrying either episomal plasmid or concatemer integration. ChIP-qPCR analysis of the new transgenic lines containing single copies of fragments F0 or F1 revealed absence of HC at the integrated fragments (Fig 3A and 3B).

Download:

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

Fig 3. Effect of fragment size on HC nucleation.

(A) H3K9me3 ChIP-qPCR analysis, as in Fig 2C, of transgenic lines carrying single copies of pfap2-g fragments F0 and F1 (ap2-g F0-S and ap2-g F1-S). Data are presented as the average and s.d. of two biological replicates. (B) qPCR analysis of copy number, as in Fig 2D. (C) Schematic (to scale) of the position of the ~ 2 kb pfap2-g and ama1 fragments. Square brackets indicate the region of the endogenous pfap2-g locus that was deleted. The position of the TSSs and the region containing the PfAP2-G binding motifs is indicated. (D) H3K9me3 ChIP-qPCR analysis, as in panel A, of transgenic lines carrying single copies of ~2 kb pfap2-g and ama1 fragments. (E) qPCR analysis of copy number, as in panel B. Additionally, presence of the plasmid backbone was also assessed (using a reference curve constructed with gDNA of the W4-2 parasite line [74], with one integrated plasmid copy).

https://doi.org/10.1371/journal.ppat.1013137.g003

Formation of HC only in parasite lines with multiple copies of F0 or F1 may be explained by the larger total size of putative HC-promoting DNA or by sequence repetitiveness (occurrence of the same sequence more than one time), as repetitiveness itself may promote HC formation. In Drosophila and other organisms, presence of multiple copies of a transgene is sufficient to trigger HC-based silencing [49,50]. To distinguish between these possibilities, we assessed HC formation at the PF3D7_1144400 locus after integration of a single copy of longer fragments (~2 kb) from the pfap2-g gene (Fig 3C–E). This revealed formation of HC with a single copy of the pfap2-g F1 + F0 fragment, including the 5’ end of the CDS and a small part of the upstream region, albeit at lower levels than in ChIP-qPCR positive controls (Fig 3D). H3K9me3 signal using primers for the flanking regions (5’ext. and 3’ext.), which was clearly observed, demonstrates that HC was formed at the fragment correctly integrated in the PF3D7_1144400 locus. In contrast, no HC was detected after integration of the F2 + F1 pfap2-g fragment or a ~ 2 kb fragment of the euchromatic ama1 gene. These results indicate that single copies of fragments of the pfap2-g locus can nucleate de novo HC formation in a previously euchromatic region, without requiring repetitiveness.

HC nucleation by pfap2-g fragments occurs in only a subset of parasites

H3K9me3 coverage at the HC nucleation-positive pfap2-g fragments integrated at the PF3D7_1144400 locus was lower than in ChIP-qPCR positive HC controls (Figs 2C and 3D). To determine whether this reflects that HC is formed in only a subset of the parasites, we generated subclones of the transgenic line carrying the pfap2-g F1 fragment (multiple copies). In two of the subclones, no HC was observed at the site of fragment integration, whereas in two other subclones (3E and 9B) HC formation was apparent (Fig 4A). HC-positive and -negative subclones contained a similar number of copies of the integrated fragment (Fig 4B). Likewise, analysis of four subclones of the transgenic line with a single copy of the pfap2-g F1 + F0 fragment integrated showed that HC formed in two of the subclones (A3 and D3) at levels similar to the ChIP-qPCR positive controls, but was absent in the other two (Fig 4C and 4D). These results confirm that the intermediate levels of H3K9me3 observed in transgenic lines with HC-positive pfap2-g fragments reflect population heterogeneity.

Download:

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

Fig 4. Cell to cell heterogeneity in HC nucleation by pfap2-g fragments.

(A) H3K9me3 ChIP-qPCR analysis, as in Fig 2C, of subclones of the ap2-g F1 transgenic line. Data is from a single biological replicate. (B) qPCR analysis of copy number, as in Fig 2D. (C) H3K9me3 ChIP-qPCR analysis, as in panel A, of subclones of the ap2-g F1 + F0 transgenic line. Data are presented as the average and s.d. of two biological replicates. (D) qPCR analysis of copy number, as in Fig 3E. (E) H3K9me3 (top) and H3K9ac (bottom) ChIP-seq profiles at the PF3D7_1144400 locus in the parental 1.2B line (wild type). Block arrows are annotated genes (start codon to stop codon). Data is from a previously published study [16]. (F) H3K9me3 (top) and H3K9ac (bottom) ChIP-seq profile at the edited PF3D7_1144400 locus in the 3E subclone of the ap2-g F1 transgenic line. Reads were aligned against the edited genome. The vertical dashed lines mark the position of the integrated fragment. (G) H3K9me3 ChIP-seq profile at the edited PF3D7_1144400 locus, as in panel E, in the A3 and D3 subclones of the ap2-g F1 + F0 transgenic line.

https://doi.org/10.1371/journal.ppat.1013137.g004

Additionally, to map the extension of the HC formed at the PF3D7_1144400 locus after integration of nucleation-positive pfap2-g fragments, we assessed the distribution of H3K9me3 using ChIP-seq. While HC was absent from the wild-type locus [16] (Fig 4E), in HC-positive subclones with the F1 or the F1 + F0 integrated fragments HC spanned over most of the PF3D7_1144400 CDS, such that the limits of the H3K9me3-enriched domain roughly coincided with the beginning and end of the gene (Fig 4F and 4G). In the pfap2-g F1 fragment 3E subclone, we also analyzed the distribution of H3K9ac, which occurred at very low levels across the PF3D7_1144400 CDS in wild type parasites [16] (Fig 4E) and was depleted throughout the full heterochromatic region in the 3E subclone (Fig 4F).

HC formation by sequences from other CVGs

Next, we tested the ability of sequences from other CVGs to nucleate HC ectopically. Following the same approach as for pfap2-g, we integrated at the euchromatic locus PF3D7_1144400 fragments from the mspdbl2 (PF3D7_1036300) locus, which is silenced in the majority of parasites and located in a small internal HC island [16,18,40,41] (S2A Fig), similar to pfap2-g, and from three var gene loci. These included two subtelomeric var genes from group B, PF3D7_0733000 and PF3D7_1041300, and a var gene located in a large internal HC island, PF3D7_1240300, from group B/C [51]. We used plasmids containing the yfcu negative selection marker to obtain parasites with single copy integrations.

None of the mspdbl2 fragments tested, including ~1 kb and ~2 kb fragments from the coding and upstream sequences, formed HC at the ectopic site (Fig 5A and 5B). In contrast, all ~ 1 kb fragments of var genes (including upstream regions and in some cases also the beginning of the CDS) robustly formed HC, generally at levels similar to the ChIP-qPCR positive control (Fig 5C and 5D). All transgenic lines with mspdbl2 or var fragments had a single copy of the fragment integrated, with the exception of the line with the PF3D7_1041300 prom+CDS fragment (Fig 5E). However, subclones of this transgenic line with a single copy of the fragment also nucleated HC efficiently, and a small deletion occurring in some parasites with the PF3D7_1240300 fragment did not affect HC nucleation (S5 Fig). Together, these results show that var gene sequences can nucleate HC efficiently, whereas some pfap2-g sequences can nucleate HC with lower efficiency and mspdbl2 sequences cannot.

Download:

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

Fig 5. HC nucleation by sequences from other CVGs.

(A) Schematic (to scale) of the position of mspdbl2 fragments tested for HC nucleation capacity. The previously described position of the HC domain at this locus is indicated (HC) [16]. (B) H3K9me3 ChIP-qPCR analysis, as in Fig 2C, of transgenic lines carrying mspdbl2 fragments. Data are presented as the average and s.d. of two biological replicates. (C) Schematic (to scale) of the position of fragments from different var genes tested for HC nucleation capacity: 0733000 (from PF3D7_0733000), 1041300 (from PF3D7_1041300) and 1240300 (from PF3D7_1240300). The previously described position of the HC domains at these loci is indicated (HC) [16]. (D) H3K9me3 ChIP-qPCR analysis, as in panel B, of transgenic lines carrying var fragments. (E) qPCR analysis of copy number, as in Fig 2D.

https://doi.org/10.1371/journal.ppat.1013137.g005

Analysis of the role of different regions of pfap2-g in HC maintenance

The H3K9me3 ChIP-seq analysis of Δ5’ap2-g-derived transgenic lines, in which the upstream region and part of the CDS of the pfap2-g locus were deleted (Fig 2A), revealed that HC was absent from the remaining part of the endogenous pfap2-g locus (Fig 6A). This unexpected observation indicates that the region deleted is necessary to maintain HC at the full pfap2-g locus. In contrast, ChIP-seq analysis of the E5-Δap2-g line, in which most of the pfap2-g CDS but not the upstream region was deleted [39], showed that this line maintained HC in the remaining part of the locus (Fig 6A). These results suggest that specific parts of the pfap2-g locus are needed for HC maintenance, which prompted us to investigate the determinants of HC maintenance in more detail.

Download:

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

Fig 6. Analysis of the role of different regions of pfap2-g in HC maintenance.

(A) H3K9me3 ChIP-seq profiles at the pfap2-g locus in wt (1.2B) and transgenic parasite lines. In experiments with transgenic lines, reads were aligned against the respective edited genome. Block arrows are annotated genes (start codon to stop codon). (B) Left: Schematic (to scale) of the pfap2-g regions deleted to assess their role in HC maintenance. The numbered vertical lines at the top indicate the position of the primer pairs used for ChIP-qPCR analysis. The position of the TSSs and the PfAP2-G binding region is shown. Right: strategy for the deletion of pfap2-g regions using CRISPR/Cas9. The scissors indicate the positions targeted by the sgRNAs. A non-functional gfp gene (gfp*) was used to replace the deleted region. In this example, part of the upstream region and CDS were deleted, but in other cases, different regions were eliminated. Black and grey arrowheads represent the primers used to assess the number of integrated copies. Primers for gfp* were also used in the ChIP-qPCR analyses. (C) H3K9me3 ChIP-qPCR analysis of the parental wt 1.2B parasite line, from which transgenic lines with deletions at the pfap2-g locus were derived. Values are the % input at each position relative to the % input in the positive control heterochromatic genes, as in Fig 2C. Numbers in the x axis are the primer pairs shown in panel B. The dashed line indicates the coverage in the positive control heterochromatic genes. Data are presented as the average and s.d. of two biological replicates. (D) H3K9me3 ChIP-qPCR analysis, as in panel C, of the ΔCDS and ΔU transgenic lines. (E) H3K9me3 ChIP-qPCR analysis, as in panel C, of the ΔU5’ transgenic line and the 10A and 12H subclones of the ΔU3’ transgenic line.

https://doi.org/10.1371/journal.ppat.1013137.g006

We systematically tested the impact of deleting different regions of the pfap2-g locus on the presence of HC at the rest of the locus. The deleted regions were replaced by a non-functional gfp gene (gfp*) using CRISPR/Cas9, and HC levels were determined using H3K9me3 ChIP-qPCR with a collection of primers at different positions of the locus (Fig 6B). The ΔCDS transgenic line, in which the deleted region spanned most of the pfap2-g CDS, had H3K9me3 levels at the rest of the locus similar to the wild type 1.2B line (Fig 6C and 6D), consistent with the ChIP-seq results for the E5-Δap2-g line (Fig 6A). In contrast, deletion of the pfap2-g upstream region (ΔU) resulted in complete HC loss from the full locus (Fig 6D), similar to the Δ5’ap2-g-derived lines (Fig 6A). Together, these results confirm that the pfap2-g upstream region but not the CDS is essential for the maintenance of HC.

To narrow down the region needed for HC maintenance, we generated the ΔU5’ and ΔU3’ lines, in which we removed the distal or the proximal part of the pfap2-g upstream region, respectively. While the distal part was not required for HC maintenance at the rest of the locus, deletion of the proximal part (ΔU3’) affected HC levels. One ΔU3’ subclone had HC at the rest of the locus, whereas another subclone completely lost it (Fig 6E). These results suggest that the proximal part of the pfap2-g upstream region plays an important role for HC maintenance, although it is not absolutely essential. We also generated and analyzed several additional transgenic lines, in which different parts of the locus were deleted. The analysis of these additional lines supported the conclusion that the proximal upstream region is important for HC maintenance whereas the other regions are dispensable (Figs 7A and S6). The regions required for HC maintenance showed only a small overlap with those capable of nucleation (Fig 7B). Of note, the majority of these transgenic lines had a single copy of the gfp* gene integrated. In a few cases, plasmid concatemers were integrated in some of the parasites (ΔU5’, ΔF1 and ΔF2–3), but analysis of subclones showed that the increased separation between the regions upstream and downstream of the deletion in lines with concatemer integration did not impair HC maintenance (Figs 6, 7 and S6).

Download:

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

Fig 7. Summary of HC distribution in parasite lines with different deletions at the pfap2-g locus.

(A) Schematic (to scale) of HC distribution, as determined by ChIP-seq or ChIP-qPCR, at the pfap2-g locus after deletion of different regions. The deleted region in each transgenic line is indicated by square brackets. TSSs are represented by black boxes. Symbols + and – indicate presence or absence of HC, respectively, whereas + /– indicates partial HC coverage (compared to the wt 1.2B line) in a bulk parasite population or occurrence of HC in some subclones but not others. “1 or 2 copies” refers to the number of gfp* copies integrated. (B) Schematic (to scale) of the relative position of regions of the pfap2-g locus capable of nucleating HC or needed for HC maintenance. The position of the TSSs and the PfAP2-G binding sites is shown.

https://doi.org/10.1371/journal.ppat.1013137.g007

Reestablishment of HC at the pfap2-g locus is inefficient

Our experiments based on ectopic fragment integration revealed that pfap2-g fragments promote de novo HC nucleation inefficiently. This made us predict that if a parasite line had the endogenous pfap2-g locus in an euchromatic state, it would take several generations for HC to form again. To test this hypothesis, we used the E5-AP2-G-DD parasite line [39], in which the PfAP2-G protein is fused to a C-terminal FKBP destabilization domain (DD) that, in the absence of the Shld1 ligand, drives protein degradation. The upstream regulatory region of the gene is intact. In this parasite line, when cultured without Shld1, absence of HC at the pfap2-g locus would not result in massive sexual conversion and therefore would not pose a fitness cost. Indeed, the E5-AP2-G-DD line shows an unusually high sexual conversion rate when Shld1 is added [39,43], consistent with accumulation of parasites with pfap2-g in an euchromatic state during culture without Shld1.

We observed sexual conversion rates of ~45% after a first PfAP2-G stabilization with Shld1 for one cycle, but subsequent stabilization two cycles later resulted in a much lower conversion rate (Fig 8A). After induction with Shld1 for one cycle, culturing in the absence of Shld1 for up to 60 generations led to progressive, partial recovery of high sexual conversion rates when Shld1 was added again (S7 Fig). Together, these results suggest that in cultures always maintained without Shld1, parasites with the pfap2-g locus in an euchromatic state progressively accumulated, because this did not pose a fitness cost, and these parasites underwent sexual conversion upon Shld1 addition. The remaining parasites had HC at the pfap2-g locus and displayed a regular conversion rate. Consistent with this model, subclones of the E5-AP2-G-DD line (obtained from cultures not exposed to Shld1) had dramatically different sexual conversion rates when Shld1 was added, ranging from <5% to >50% (Fig 8B). As expected, higher sexual conversion was associated with higher pfap2-g transcript levels, which increased in the presence of Shld1 as a consequence of the PfAP2-G transcriptional feedback loop [39,43,52,53] (Fig 8C).

Download:

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

Fig 8. HC dynamics at the pfap2-g locus in the E5-AP2-G-DD line.

(A) Sexual conversion rate of the E5-AP2-G-DD line after exposing cultures to Shld1 for the first time (1^st Shld1 add.) and after re-exposing them to Shld1 following 2 growth cycles without Shld1 (2^nd Shld1 add.). Data are presented as the average and s.d. of two biological replicates. (B) Sexual conversion rates after Shld1 addition for the parental E5-AP2-G-DD line and its subclones (sc) B11 and D7. Data are presented as the average and s.d. of three biological replicates. (C) Expression of pfap2-g and ama1 (the latter used as a control), determined by RT-qPCR, in E5-AP2-G-DD subclones B11 and D7 at the mature schizont stage (ML10-treated cultures) in the absence or presence of Shld1. Values are normalized by expression of serrs. Data are presented as the average and s.d. of two biological replicates. (D) H3K9me3 ChIP-seq profiles at the pfap2-g locus in the E5-AP2-G-DD line and subclones B11 and D7. Reads were aligned against the edited genome. Block arrows are annotated genes (start codon to stop codon). (E) Quantification of H3K9me3 ChIP-seq coverage at the pfap2-g, var gene PF3D7_1240300, clag3.2 and mspdbl2 loci in the parental E5-AP2-G-DD line and its subclones B11 and D7 (no Shld1 added). Average coverage was calculated at the region from -1 kb to +0.5 kb from the start codon of each gene. (F) Model for heterochromatin dynamics at the pfap2-g locus and sexual conversion rate (after addition of Shld1) in the E5-AP2-G-DD line and its subclones B11 and D7. Higher number of gametocytes indicates higher sexual conversion rate. Red and green circles represent parasites with heterochromatic or euchromatic pfap2-g, respectively. Parasites with euchromatic pfap2-g convert into gametocytes upon Shld1 addition, whereas in parasites with heterochromatic pfap2-g, the gene is activated and drives sexual conversion in only a small fraction of the parasites at each cycle, as in wild type parasites. After a first treatment of the E5-AP2-G-DD line with Shld1, only parasites with heterochromatic pfap2-g continued asexual growth. A second treatment with Shld1 resulted in low levels of sexual conversion.

https://doi.org/10.1371/journal.ppat.1013137.g008

ChIP-seq analysis of cultures maintained without Shld1 revealed a lower H3K9me3 coverage at the pfap2-g locus in the parental E5-AP2-G-DD line and a subclone with high sexual conversion (B11) compared with a subclone with low conversion (D7) (Fig 8D–F). The H3K9me3 coverage fold-change at the pfap2-g locus was among the top differences in the full genome (14^th largest D7 vs B11 fold-change, permutation test p value = 0.030), with other large differences corresponding mainly to genes involved in antigenic variation (e.g., var, rif) that are well-known to show HC occupancy variation between subclones [16] (S1 Dataset). This is consistent with the gene being in an euchromatic state in the approximately 50% of parasites that convert upon Shld1 addition in the parental E5-AP2-G-DD line and the B11 subclone. No subclone with 100% conversion upon Shld1 addition was identified, suggesting that when a subclone originated from a single parasite with euchromatic pfap2-g, HC formed in some parasites during the time between subcloning and ChIP-seq analysis (24 generations), albeit at low speed consistent with inefficient nucleation. The reduced H3K9me3 coverage in the high conversion subclone was homogeneous across the full pfap2-g locus, in contrast with changes upon activation of the gene during natural sexual conversion that involve HC depletion only at the upstream region [30,54]. Together, these experiments suggest that the pfap2-g locus can be in an euchromatic state when the protein is non-functional, both the euchromatic and the heterochromatic states are heritable, and spontaneous reestablishment of HC occurs but is inefficient and requires multiple generations (Fig 8F). However, we acknowledge that while this model is consistent with the results of the experiments with the E5-AP2-G-DD line, we do not provide direct evidence for the occurrence of the pfap2-g locus in a fully euchromatic state in some parasites or for slow, progressive re-establishment of HC, because we could not analyze subclones at earlier time points.

Parasite cultures and sexual conversion assays

The 3D7 subclones E5 (derived from the 3D7-B stock) and 1.2B (derived from the 3D7-A stock) and the transgenic lines E5-PfAP2-G-DD and E5-Δap2-g have been previously described [39,67]. Parasites were cultured in B+ erythrocytes at 3% hematocrit under standard conditions, with RPMI-1640-based culture medium supplemented with 0.5% Albumax II (Invitrogen). Erythrocytes were obtained from the ‘Banc de Sang i Teixits’ (Barcelona) after ethical approval by the Hospital Clínic (Barcelona) ethics committee. Routine synchronization was carried out using 5% D-sorbitol lysis to eliminate late asexual stages (trophozoites and schizonts). For the analysis of transcripts by RT-qPCR to validate TSSs, cultures were tightly synchronized by purification of mature schizonts using 63% Percoll (Sigma-Aldrich) gradients followed by sorbitol lysis 5 h later [68].

For the stabilization of PfAP2-G in the E5-AP2-G-DD line (experiments in Fig 8), 0.5 μM AquaShield-1 (Shld1, Cheminpharma) was added to cultures ~10 h after sorbitol synchronization (ring/ early trophozoite stage). To measure sexual conversion rates, at the following cycle after stabilization cultures at the ring stage were treated with 50 mM N-acetylglucosamine (Sigma-Aldrich, A3286) for ≥4 days to eliminate asexual parasites. Light microscopy quantification of Giemsa-stained smears was used to determine sexual conversion rates, calculated as the gametocytemia at day ≥ 4 divided by the initial parasitemia before adding N-acetylglucosamine [43]. For the experiments to measure pfap2-g transcript levels (Fig 8C), cultures at the trophozoite stage were treated with ML10 for 22 h before collecting RNA, as previously described [43], to have essentially all parasites at the same mature schizont stage in both subclones.

5’RACE and RT-qPCR analysis

TSSs were identified using the FirstChoice 5′ RLM-RACE kit (Thermo Fisher) following the manufacturer’s instructions, as previously described [20]. In brief, total RNA from non-synchronized E5 and E5-PfAP2-G-DD (treated with Shld1) cultures was extracted using the TRIzol (Invitrogen) method [68]. Next, RNA was treated with calf intestinal alkaline phosphatase to remove free 5’-phosphate groups from non-mRNA molecules. Then, samples were treated with the tobacco acid pyrophosphatase to remove the 5’-cap structure of full-length mRNAs, allowing the ligation of an RNA adapter with T4 RNA ligase. We used a redesigned RNA adapter (S1 Table) because the one provided with the kit resulted in artefacts that incorporated exogenous sequences of another chromosome into the pfap2-g PCR-amplified product. cDNA synthesis was performed by reverse transcription using random hexamers followed by nested PCR using pfap2-g specific primers and primers specific for the RACE adapter (S1 Table). Nested PCR was performed either with the LA Taq DNA Polymerase (Takara), with 25 cycles of amplification for each PCR reaction and an annealing temperature of 60ºC, or with the KAPA HiFi PCR Kit (KAPA Biosystems, KK2103), with 20 cycles of amplification in the first PCR and 30 in the second, and an annealing temperature of 62ºC. PCR products were cloned into the pCR-2.1 plasmid using the TA Cloning Kit (Invitrogen) and sequenced with M13 forward and reverse primers.

For RT-qPCR, RNA was extracted using the TRIzol method (Invitrogen), DNAse I-treated (Qiagen), purified using the RNeasy MinElute Cleanup kit (Qiagen) and reverse-transcribed using the AMV Reverse Transcription Kit (Promega) with a mixture of random primers and oligo (dT). Transcript abundance was measured by real-time quantitative PCR (qPCR) in triplicate wells using the standard curve method, with the Power SYBR Green Master Mix (Applied Biosystems) and a 7900HT Fast Real-Time PCR System, as previously described [68]. Using the standard curve method, results are not affected by primer efficiency. However, all primer pairs had an efficiency >80%. Transcript levels of the serine-tRNA ligase (serrs, PF3D7_0717700) were used for normalization. All primers used for RT-qPCR analysis are described in S1 Table.

Plasmids

The pL7-KO-ap2g plasmid (S8A Fig) used to delete the pfap2-g upstream region and part of the CDS (Δ5’ap2-g line) was derived from the pL6-egfp-yfcu plasmid [69]. The original HR2 was removed by digestion with EcoRI and NcoI, followed by treatment with T4 DNA polymerase (NEB) to generate blunt ends and religation with T4 DNA ligase (Roche). The pfap2-g HR1 (positions -3929 to -3521 bp relative to the start codon) was PCR amplified and cloned into NotI/PstI sites, disrupting the yfcu cassette. The pfap2-g HR2 (positions +991 to +1973 bp) was amplified and cloned into SpeI/AflII sites, replacing the original HR1. To add the sgRNA, two annealed oligonucleotides containing the guide sequence (recognizing a sequence located at the pfap2-g -3503 to -3522 bp position) were cloned into a BtgZI site using the In-Fusion HD Cloning Kit (Clontech). Last, an incomplete gfp sequence (gfp*), lacking the start codon, was amplified and cloned between HR1 and HR2, removing the remaining part of the yfcu cassette, using PstI/SpeI sites. To generate plasmid pDC2-Cas9-hDHFR-ap2g-3’ (S8A Fig) expressing a second sgRNA in addition to Cas9 and the hdhfr selectable marker, a guide RNA covering positions +921 to +940 bp was cloned into a BbsI site of plasmid pDC2-Cas9-hDHFR [70] using the InFusion system.

For the integration of the fragments into the PF3D7_1144400 locus, pL1144400-Fx plasmids (“x” indicates each different fragment; S4A Fig) were generated from the pL6-egfp-yfcu plasmid [69]. The original HR2 was removed as described above. The PF3D7_1144400 HR1 (positions +1144 to +1562 bp from the start codon) was amplified with primers incorporating NcoI + SpeI sites in the reverse primer and a SacII site in the forward primer. The PF3D7_1144400 HR2 (positions +1646 to +2199 bp) was amplified incorporating SpeI and AflII sites, and cloned together with HR1 into SacII/AflII sites, replacing the original HR1. The original hdfhr and sgRNA expression cassettes were removed using restriction sites AatII and AflII, blunt ending and religation. The yfcu cassette was removed using restriction sites NotI and SacII, blunt ending and religation. However, for all fragments except ~1Kb fragments in Fig 2 and shorter fragments (S3 Fig), the selection cassette was reintroduced using the same restriction sites (pL1144400-Fx-yFCU plasmids) (S9 Fig). Fragments (Fx) from different genes were PCR amplified and cloned into NcoI/SpeI sites with the primers listed in S1 Table. To generate the plasmid pDC2-Cas9-hDHFRyFCU-1144400 (S4A and S9 Figs), used for transfections with pL1144400-Fx and pL1144400-Fx-yFCU plasmids, a guide RNA (covering positions +1566 to +1585 bp from the PF3D7_1144400 start codon) was cloned into a BbsI site of plasmid pDC2-Cas9-hDHFRyFCU [71] using the InFusion system.

To generate the plasmid pL-ap2g_Del1_GFP_yfcu, used to obtain the ΔATG-1 line, the original HR1, ama1 fragment and HR2 regions were removed from the pL11444000-ama1-yFCU plasmid (S9 Fig) by digestion with SacII and EcoRI. The HR1 and HR2 were cloned by InFusion, introducing HindIII and SpeI restriction sites between them to clone the gfp* sequence by ligation (S10A Fig). For the ΔATG-2, ΔCDS, ΔU, ΔU5’, ΔU3’ and ΔF2–3’-A lines, HRs for each construct were amplified, introducing AflII and HindIII restriction sites between HR1 and HR2, and cloned into SacII/NcoI sites of a plasmid derived from pL6-eGFP-yFCU [69] using the InFusion system. Digestion of the plasmid with SacII and NcoI removed the original HRs and the hdhfr sequence. Next, a different sgRNA for each transgenic line was cloned into a BtgZI site. Finally, the gfp* sequence was cloned into the AflII/HindIII sites between HR1 and HR2 by ligation, resulting in pL-7-ap2g_Delx_GFP_yfcu plasmids (where “x” refers to each fragment deleted) (S10A Fig). Finally, a second sgRNA used for each transgenic line was cloned into a BbsI site of plasmid pDC2-Cas9-hDHFR [70], resulting in pDC2-Cas9-hDHFR-sgRNA-Delx plasmids (S10A Fig and S1 Table).

For deletion ΔF1 and to generate a second transgenic line with the deletion ΔF2–3’ (ΔF2–3’-B line), we followed an alternative three-plasmid strategy (S10B Fig). Donor plasmids pUC19-KO-Fx were generated by simultaneously cloning the respective HR1 and HR2 for each construct and the gfp* sequence into a BamHI site of the pUC19 vector, using the InFusion system. Guides were cloned into the pDC2-Cas9-hDHFR plasmid as described above.

All guides were designed and selected using the EuPaGDT web-based tool [72]. PCR amplification from P. falciparum gDNA was performed using LA Taq DNA Polymerase (Takara). For plasmid cloning and amplification, we used Escherichia coli DH5α or Stellar Competent Cells (Clontech) for difficult cloning. Oligonucleotides were purchased from Integrated DNA Technologies (IDT). All primers and oligonucleotides are described in S1 Table.

Generation of transgenic lines

Unless otherwise stated, transfections were performed according to standard procedures [73], electroporating cultures at the ring stage with a BioRad GenePulser Xcell electroporator. Starting 15 h after transfection, cultures were selected for 4 days with 10 nM WR99210 (Jacobus Pharmaceuticals). When parasites were observed, for transfections with donor plasmids containing the yfcu negative selection marker, cultures where selected with 1 µM 5-fluorocytosine (5-FC; clinical grade Ancotil, Mylan N.V.) for 8 days, to remove parasites carrying episomal copies of the plasmids or integration of concatemers containing the full plasmid. Note that in the case of the var 1240300 line, cultures were analyzed without 5-FC selection (for comparison with transgenic lines generated with plasmids lacking the yfcu selection marker, Fig 2), and after 5-FC selection (Fig 5). Diagnostic PCR to assess correct edition was performed using the LA Taq DNA Polymerase (Takara), typically with primers external to the region edited. Diagnostic qPCR using the standard curve method was performed to assess the presence of plasmid regions not intended to be integrated and to determine the relative copy number of HRs and integrated fragments. Analysis by qPCR was performed in triplicate wells as described above for RT-qPCR. We used gDNA from the W4-2 (containing a single copy of the hdhfr gene and plasmid backbone) or the wt 3D7-A parasite lines [74] for the standard curve included in each plate for each primer pair. In all cases, copy number was determined relative to a single copy gene (serrs). The primers used for diagnostic PCR and qPCR are described in S1 Table. Additionally, for the parasite lines containing integrated fragments F1 or F1 + F0 analyzed by ChIP-seq, we used the input reads to characterize their genome. Alignment against the edited genomes confirmed correct integration at the PF3D7_1144400 locus, with multiple PF3D7_1144400/ pfap2-g hybrid reads spanning the integration junctions. These parasite lines were subcloned after transfection, and hence are genetically homogeneous, excluding the possibility that the fragments integrated elsewhere in a subset of the parasites.

To obtain the Δ5’ap2-g line, 1.2B cultures were transfected with 60 µg of plasmid pDC2-Cas9-HDHFR-ap2g-3’ and 12 µg of pL7-KO-ap2g linearized using a ScaI site located in the backbone of the plasmid. Subclones were obtained by limiting dilution and assessed by diagnostic PCR (with primers external from the edited region) and qPCR analysis to assess presence of plasmid sequences not intended to be integrated (hdhfr marker). The H11 subclone was selected for further work because it contained no episomal copy of the transfection plasmids or multiple integrations (S8B and S8C Fig).

For the integration of fragments at the PF3D7_1144400 locus, Δ5’ap2-g (H11 subclone) cultures were transfected with 60 µg of pDC2-Cas9-yFCU-hDHFR-1144400 and 12 µg of either pL1144400-Fx (S4 Fig) or pL1144400-Fx-yFCU (S9 Fig), linearized using the ScaI or the XmnI restriction sites.

To generate transgenic lines with deletions at the pfap2-g locus, 1.2B cultures were transfected with 60 µg of pDC2-Cas9-yFCU-hDHFR-sgRNA-Delx and 12 µg of either pL-ap2g_Del1_GFP_yfcu (ΔATG-1, the only transgenic line that was generated using only one guide RNA) or pL-7-ap2g_Delx_GFP_yfcu (ΔATG-2, ΔCDS, ΔU and ΔU5’) plasmids, linearized with XmnI, ScaI or PvuI (S10A Fig). The exception were the lines carrying the ΔF2–3’-B and ΔF1 deletions, which were generated using a three-plasmids transfection with 12 µg of pUC19-KO-Fx (linearized using XmnI) and 60 µg of each of two different pDC2-Cas9-hDHFR-sgRNA-Delx plasmids containing different guides (S10B Fig). This was followed by subcloning by limiting dilution, when required, to obtain pure populations of edited parasites. In two cases in which we could not obtain edited parasites using regular transfections (ΔU3’ and ΔF2–3-A, the latter is equivalent to ΔF2–3-B, with the same F2-3 region deleted, but generated differently), we used Nucleofactor transfection of schizonts. For this, cultures were tightly synchronized to a 2 h age window using Percoll purification followed by sorbitol lysis 2 hours later. When most parasites reached the mature schizont stage, a new Percoll purification was performed, followed by transfection using the AMAXA P3 primary cell 4D Nucleofector X Kit L (Lonza), according to manufacturer’s instructions, with 60 µg of pDC2-Cas9-hDHFR-gRNA-Delx plasmid and 60 µg of pL-7-ap2g_Delx_GFP_yfcu, linearized with ScaI or PvuI.

Southern blot

Southern blot was performed according to standard procedures [74]. gDNA was digested with the restriction enzymes AatII and HincII, plus AflII to cleave episomal plasmids. Digestions were resolved in 1.25% agarose gels and transferred to Amersham Hybond N+ nylon membranes (GE Healthcare) for detection with a specific probe labelled with α-³²P dATP (Perkin Elmer). The primers used to amplify the probe were the same as the ones used for the amplification of the HR1 of PF3D7_1144400 (S1 Table). Hybridization and washes were performed at 62°C.

ChIP-qPCR and ChIP-seq

Chromatin extraction from cultures at the late trophozoite/schizont stage was performed as previously described, using the MAGnify Chromatin Immunoprecipitation System (Life Technologies) [30,47,64], with minor modifications. Briefly, for ChIP-qPCR, formaldehyde-crosslinked samples were sonicated using a Bioruptor Plus sonication device (Diagenode) at high power, with 6 cycles of 30 seconds ON/ 30 seconds OFF followed by centrifugation and recovery of the aqueous phase, which was sonicated for 2 additional cycles of 30 seconds ON/ 30 seconds OFF to obtain sheared chromatin with a size around 400–500 bp. Immunoprecipitation was performed overnight at 4ºC with 0.5 µg of chromatin and 1 µg of antibodies against H3K9me3 (Diagenode), previously coupled to protein A/G magnetic beads provided in the MAGnify kit. Washing, de-crosslinking and elution were performed following the kit recommendations. Eluted DNA was diluted 1:2 for further downstream analysis by qPCR. For qPCR analysis, samples were analyzed in triplicate wells using the standard curve method as described above for RT-qPCR, with a standard curve in each plate for each primer pair (primers described in S1 Table). The standard curve method [68] prevents inaccuracies related with low primer efficiency, but all primer pairs had an efficiency >80%.

For ChIP-seq [47], samples were sonicated using an M220 sonicator (Covaris) at 10% duty factor, 200 cycles per burst, 140 W of peak incident power for 10 min. Immunoprecipitation was performed overnight at 4ºC with 2 µg of chromatin and 4 µg of antibodies against H3K9me3 (Diagenode) or H3K9ac (Diagenode), previously coupled to protein A/G magnetic beads provided in the MAGnify kit. Washing, de-crosslinking and elution were performed following the kit recommendations, but avoiding high temperatures that may result in denaturation of extremely AT-rich intergenic regions: de-crosslinking, proteinase K treatment and elution were performed at 45ºC (for 2 h, overnight, and 1.5 h, respectively).

Libraries for Illumina sequencing were prepared from 5 ng of immunoprecipitated DNA using a protocol adapted for extremely AT-rich genomes [45,47]. In brief, after end-repair and addition of 3′ A-overhangs, NEBNext Multiplex Oligos for Illumina (NEB, E7500) were ligated. Purification steps were performed with Agencourt AMPure XP beads (Beckman Coulter, A63880). Libraries were amplified using the KAPA HiFi PCR Kit (KAPA Biosystems, KK2103) in KAPA HiFi Fidelity Buffer with the following conditions: 95°C for 3 min; 9 cycles at 98°C for 20 s and 62°C for 2 min 30 s; and 62 °C for 5 min. Amplified libraries were purified using 0.9X AMPure XP beads to remove adapter dimers. The library size was analysed in a 4200 TapeStation System (Agilent Technologies). We obtained 12–16 million 125 bp paired-end reads per sample using the HiSeq2500 or the NextSeq 500 System (Illumina).

The analysis of ChIP-seq data was performed as previously described [16,30]. After initial processing, the reads where aligned to the reference 3D7 genome modified when required to include the alterations introduced after genome editing (i.e., deletion of the pfap2-g upstream region and part of the CDS, deletion of the CDS, integration of fragments at the PF3D7_1144400 locus or integration of a plasmid at the pfap2-g locus to add a C-terminal DD tag). After full processing, data was visualized using IGV (v2.4.10) [75]. To calculate genome-wide HC occupancy at genes, H3K9me3 coverage was quantified at the region -1000 to + 500 bp from the start codon of each gene, as previously reported [16,30]. To calculate the statistical significance of the changes in HC coverage at the pfap2-g locus, we used a permutation test. For this, we generated a null distribution for the coverage fold-change between D7 and B11 in all clonally variant genes [61] by generating 10,000 permutations of the fold-change values and compared the observed ranking of each gene with the ranking in the null distribution.