A heat-shock response regulated by the PfAP2-HS transcription factor protects human malaria parasites from febrile temperatures.

Periodic fever is a characteristic clinical feature of human malaria, but how parasites survive febrile episodes is not known. Although the genomes of Plasmodium species encode a full set of chaperones, they lack the conserved eukaryotic transcription factor HSF1, which activates the expression of chaperones following heat shock. Here, we show that PfAP2-HS, a transcription factor in the ApiAP2 family, regulates the protective heat-shock response in Plasmodium falciparum. PfAP2-HS activates the transcription of hsp70-1 and hsp90 at elevated temperatures. The main binding site of PfAP2-HS in the entire genome coincides with a tandem G-box DNA motif in the hsp70-1 promoter. Engineered parasites lacking PfAP2-HS have reduced heat-shock survival and severe growth defects at 37 °C but not at 35 °C. Parasites lacking PfAP2-HS also have increased sensitivity to imbalances in protein homeostasis (proteostasis) produced by artemisinin, the frontline antimalarial drug, or the proteasome inhibitor epoxomicin. We propose that PfAP2-HS contributes to the maintenance of proteostasis under basal conditions and upregulates specific chaperone-encoding genes at febrile temperatures to protect the parasite against protein damage.

INTRODUCTION

A temperature increase of only a few degrees Celsius above the optimal growth temperature of any organism causes aberrant protein folding and aggregation, which contributes to an imbalance in proteostasis that can lead to cell-cycle arrest or cell death[1]. To counteract the effect of high temperatures and other proteotoxic conditions, cells have a well-characterized heat-shock response that induces expression of molecular chaperones that aid protein refolding and prevent non-specific protein aggregation[1,2]. In most eukaryotes, the immediate upregulation of chaperone-encoding genes during heat-shock depends on the conserved transcription factor HSF1, whereas other transcriptional changes during thermal stress are driven by different transcription factors[3-6].

The human response to blood-stage infection with malaria parasites involves periodic fever episodes, which are the hallmark of clinical malaria[7-9]. Fever is an important part of the human innate immune response, and it may contribute to reducing the total parasite burden[8-10]. Infection with P. falciparum, which causes the most severe forms of human malaria, results in fevers that typically occur on alternate days (tertian fever). Tertian fever reflects the ~48 h duration of the asexual intraerythrocytic development cycle (IDC), during which parasites progress through the ring, trophozoite and multinucleated schizont stages. Fever episodes are triggered by schizont rupture and release of invasive merozoites[8,9]. In vitro, febrile temperatures inhibit parasite growth, with maximal effect on trophozoites and schizonts[9,11,12], and also induce conversion of asexual parasites into sexual forms that mediate transmission to mosquitoes[13].

Despite the importance of the heat-shock response for human malaria parasite survival during host fever episodes, the regulation of the heat-shock response has not been characterized in these organisms. The genomes of Plasmodium spp. lack an ortholog of HSF1, but they encode the main chaperone families described in other organisms[14], and several specific P. falciparum chaperones have been shown to be essential for heat-shock survival [15-19]. Phosphatidylinositol 3-phosphate and apicoplast-targeted pathways are also essential for parasite survival under thermal stress[18,19]. Transcript levels of over three hundred genes are altered at febrile temperatures[20], but how these transcriptional changes are regulated is not known. Here we set out to characterise the regulation of the protective response of P. falciparum to increased temperature.

RESULTS

A nonsense mutation in pfap2-hs is associated with low survival from heat-shock

To understand the molecular basis of heat-shock resistance in P. falciparum, we first analysed parasite lines that had been previously selected with periodic heat-shock for five consecutive rounds of the IDC[21] (Fig. 1a). Although the parental parasite line (3D7-A) appeared to have lost the ability to withstand heat-shock (~30% survival to a 3 h heat-shock at 41.5°C) during growth in vitro, it re-adapted to heat-shock pressure (>75% heat-shock survival) in only three generations[21], suggesting that this line contained a selectable subpopulation of parasites resistant to heat-shock. In order to evaluate whether the heat-shock resistance phenotype of 3D7-A had a genetic or an epigenetic basis, we first analysed the transcriptome across the full IDC under basal conditions (no heat-shock) of two independently selected lines (3D7-A-HS r1 and r2) and non-selected cultures maintained in parallel (3D7-A r1 and r2). This analysis failed to identify any basal transcript level differences that could explain the heat-shock resistance phenotypes (Extended Data Fig. 1). Therefore, we sequenced the genomes of these lines, which revealed a novel single nucleotide polymorphism (SNP) that was predominant in non-selected cultures, but virtually absent after heat-shock selection (Fig. 1b-c, Supplementary Table 1). The mutation is also absent from two other 3D7 stocks in our laboratory and from the 3D7 reference genome, indicating that it arose spontaneously in the 3D7-A stock during culture. This SNP results in a premature STOP codon (Q3417X) in the gene PF3D7_1342900, which encodes a putative transcription factor of the ApiAP2 family[22-24] that we termed PfAP2-HS. PfAP2-HS has three AP2 domains (D1-D3), and the Q3417X mutation results in a truncated protein that lacks D3 (PfAP2-HSΔD3) (Fig. 1d). This result indicates that adaptation of 3D7-A to heat-shock involved selection of parasites expressing full-length PfAP2-HS, consistent with a role for this protein in the heat-shock response. In support of this idea, the first AP2 domain (D1) of PfAP2-HS was previously reported to recognize in vitro a DNA motif termed G-box[23], which is enriched in the upstream region of some heat-shock protein (HSP) chaperone genes[25].

Fig 1.

Mutations in PfAP2-HS and sensitivity to heat-shock.

a, Schematic of the parasite lines used in this study. Colours indicate wild type PfAP2-HS or truncated forms lacking AP2 domain 3 (ΔD3), the three AP2 domains (ΔD1–3), or virtually the full protein (KO). Parasite lines shown with a colour gradient consist of a mixture of individual parasites expressing different versions of the protein. An asterisk indicates a heat-shock (HS) sensitive phenotype, and r1 and r2 are independent replicates of the selection of 3D7-A with periodic heat-shock (3D7-A-HS r1 and r2 are the selected lines, whereas 3D7-A r1 and r2 are controls maintained in parallel at 37°C). b, Proportion of Illumina reads with (Alt) or without (Ref) a nonsense mutation in pfap2-hs in two independently selected heat-shock-adapted cultures (3D7-A-HS r1 and r2) and their controls (3D7-A r1 and r2). c, Sanger sequencing confirmation of the mutation (in the r1 replicate, representative of r1 and r2). d, Schematic of wild-type PfAP2-HS, PfAP2-HS_ΔD3 and ΔPfAP2-HS. The position of the AP2 domains is indicated (D1–3). e, Survival of tightly synchronized cultures exposed to heat-shock at different ages (in h post-invasion, hpi) for two heat-shock-sensitive (3D7-A r2 and 10G) and two heat-shock-resistant (3D7-A-HS r2 and 1.2B) lines (mean of n=2 independent biological replicates). f, Heat-shock survival at the trophozoite stage of 3D7-A subclones carrying or not the Q3417X mutation (mean of n=2 independent biological replicates). g, Heat-shock survival of tightly synchronized cultures of parasite lines expressing wild-type or mutated PfAP2-HS. Values are the mean and s.e.m. of n=5 (lines of 3D7 origin) or n=3 (HB3 and D10 lines) independent biological replicates. P values were calculated using a two-sided unpaired t-test.

Extended Data Fig. 1

Transcriptional analysis of heatshock-adapted and control lines.

a, Microarray-based transcriptomic comparison across the asexual blood cycle of 3D7-A cultures adapted to heat-shock after five cycles of selection with a 3 h heat-shock at 41.5°C at the trophozoite stage (3D7-A-HS r1 and r2) and control parasite lines maintained in parallel without heat-shock (3D7-A r1 and r2). Values are the log2 of the maximum expression fold-change (from the average of all lines compared) across a time interval corresponding to half the length of the asexual cycle, calculated using the aMAFC score as previously described[21]. Genes with a >1.5-foldchange in expression in two independent 3D7-A heat-shock-adapted lines (3D7-A-HS r1 and r2) relative to their respective controls (3D7-A r1 and r2) are shown. Data for parasite lines 10G (heat-shock-sensitive subclone), 1.2B (heat-shock-resistant subclone) and 3D7-A (right panel) is from RoviraGraells et al.[21]. b, Time-course expression of genes in panel a that showed a concordant change in expression between heat-shock-adapted and control cultures, and between the heat-shock-resistant subclone 1.2B and the heat-shock-sensitive subclone 10G. Based on the predicted function of the three genes, clag2 was considered the most plausible candidate to play a role in heat-shock resistance. c, Expression of clag2 is neither necessary nor sufficient for heat-shock resistance. RT-qPCR analysis of clag2 transcript levels (normalized against rhoph2) in schizonts of heatshock sensitive (S) and heat-shock resistant (R) 3D7-A subclones (see Fig. 1f), and of the heatshock-adapted and control lines (n=1 biological replicates).

To test the involvement of PfAP2-HS in heat-shock resistance, we used a heat-shock survival assay with a 3 h heat-shock at 41.5°C[21] at the mature trophozoite stage, because maximal survival differences between heat-shock sensitive and resistant parasite lines were observed when exposing parasites at this stage (Fig. 1e). The analysis of a collection of 3D7-A subclones revealed that all subclones with the Q3417X mutation (e.g., 10G subclone) have a heat-shock-sensitive phenotype, whereas subclones with the wild-type allele (e.g., 10E subclone) have a heat-shock resistant phenotype (Fig. 1a,f).

Deletion of PfAP2-HS reduces survival from heat-shock

To further characterize PfAP2-HS, we sought to disrupt the entire gene using CRISPR-Cas9 technology. After several unsuccessful attempts with different 3D7 subclones at 37°C (the physiological temperature for P. falciparum), we reasoned that PfAP2-HS may play a role in regulating the expression of chaperones under basal conditions, in addition to being necessary for heat-shock survival. Therefore, we attempted to knock out the gene in cultures maintained at 35°C, as mild hypothermia is expected to reduce protein unfolding and favour proteome integrity[26,27]. Indeed, at 35°C, knockout of pfap2-hs was readily achieved in both the heat-shock-resistant 10E and the heat-shock-sensitive 10G subclones of 3D7-A (10E_Δpfap2-hs and 10G_Δpfap2-hs lines) (Fig. 1a,d, Extended Data Fig. 2a). Deletion of pfap2-hs resulted in severely increased sensitivity to heat-shock, with a level of heat-shock survival below that of parasites expressing PfAP2-HSΔD3. Deletion of the gene in two additional parasite lines of unrelated genetic background, HB3 and D10, also resulted in a major reduction in heat-shock survival (Fig. 1g).

Extended Data Fig. 2

Generation and characterization of transgenic parasite lines edited at the pfap2-hs locus.

a, Schematic of the CRISPR/Cas9 strategy used to knockout pfap2-hs, using two guide RNAs. b-d, Tagging of endogenous PfAP2-HS using CRISPRCas9 technology. The tags used were a C-terminal 3xHA (b), a C-terminal eYFP (c) and an N-terminal eYFP (d). e, C-terminal tagging of endogenous PfAP2-HS by single homologous recombination with a tag consisting of a 2xHA epitope and an FKBP destabilization domain (DD domain). In all panels, the position of the primers used for analytical PCR (arrowheads), guide RNA and AP2 domains (blue vertical bars) is indicated. The electrophoresis images at the right are the analytical PCR validation of the genetic edition (single genomic DNA extraction and PCR analysis), showing correct edition and absence of wild-type locus in all cases except for the 8A subclone of 1.2B-ddFKBP (8A and 12E are subclones obtained after drug cycling). The bar charts at the right show the level of survival (mean of n=2 independent biological replicates) of sorbitol-synchronized cultures of the transgenic lines upon heat-shock (HS) exposure at the trophozoite stage, with heat-shock-resistant (10E) and heatshock-sensitive (10G, expressing PfAP2-HSΔD3) subclones as controls. Addition of a C-terminal eYFP or HA-FKBP tag did not affect growth at 37°C but resulted in high heat-shock sensitivity, similar to the 10G line. In contrast, C-terminal addition of the smaller 3xHA tag or addition of an N-terminal eYFP did not affect growth at 37°C or heat-shock sensitivity. In all cases, tagged PfAP2-HS was not detectable by immunofluorescence or Western blot analysis, probably as a consequence of the very low abundance of this transcription factor (see proteomic data in www.PlasmoDB.org).

The PfAP2-HS-driven transcriptional response to heat-shock is extremely compact

To define the PfAP2-HS-dependent and independent heat-shock response, we carried out a time-course transcriptome analysis of the 10E (wild-type PfAP2-HS), 10G (PfAP2-HSΔD3) and 10E_Δpfap2-hs lines during and after heat-shock (Fig. 2a, Extended Data Fig. 3a, Supplementary Table 2). Hierarchical clustering based on changes in cultures exposed to heat-shock compared with control cultures maintained in parallel without heat-shock revealed one cluster of transcripts (cluster I) that are rapidly increased during heat-shock in 10E but not in 10G or 10E_Δpfap2-hs. Cluster I comprises only three genes: a gene of unknown function (PF3D7_1421800), the cytoplasmic hsp70 (hsp70–1; PF3D7_0818900) and hsp90 (PF3D7_0708400) (Fig. 2a). The regulatory regions of these two chaperone-encoding genes contain the best two matches in the full genome for a tandem G-box[23,25] (Extended Data Fig. 3b). While hundreds of genes in the P. falciparum genome contain a single G-box, only hsp70–1 and hsp90 showed PfAP2-HS-dependent activation during heat-shock, suggesting that the tandem G-box arrangement may be needed for activation by PfAP2-HS. The strongest transcriptional response to heat-shock was observed for hsp70–1 (~16-fold increase versus ~4-fold for hsp90).

Fig 2.

Global transcriptional alterations in parasites exposed to heat-shock.

a, Hierarchical clustering of genes with altered transcript levels (≥4 fold-change at any of the time points analysed) during (1.5 and 3 h) or 2 h after finishing (2 h post) heat-shock (HS). Values are the log2 of the expression fold-change in heat-shock versus control cultures. 13 genes had values out of the range displayed (actual range: −4.78 to +4.93). For each cluster, mean values (with 95% confidence interval) for the genes in the cluster and representative enriched GO terms are shown. Columns at the left indicate annotation as chaperone[14], presence of the G-box[23] or tandem G-box (TdGbox) in the upstream region, and log2 fold-change after heat-shock in a previous study[20] (Oakley). b, Venn diagrams of the genes altered upon heat-shock in the three parasite lines. c, Pearson correlation of the genome-wide transcript levels of each culture versus the most similar time point of a high-density time-course reference transcriptome[28]. d, Age progression during the assay, statistically estimated[61] from the transcriptomic data. e, ChIP-seq analysis of HA-tagged PfAP2-HS, representative of n=5 and n=3 independent biological replicates for 35°C and heat-shock, respectively. The log2-transformed ChIP/input ratio at the hsp70–1 and hsp90 loci is shown. The position of the tandem G-box is indicated.

Extended Data Fig. 3

Validation of the transcriptomic changes upon heat-shock and distribution of the tandem G-box motif.

a, RT-qPCR analysis of transcript levels (normalized against serine--tRNA ligase) of the genes selected for validation, using biological samples independent from the samples used for microarray analysis. Values are the average of triplicate reactions. The log2 expression fold-change [heat-shock (HS) relative to control (35°C) conditions] for these genes in the microarray analysis (Fig. 2a) is shown in a heatmap to facilitate comparison. b, Genes in the P. falciparum genome containing tandem arrangements (maximum distance between the two: 9 nucleotides) of the Gbox [(A/G)NGGGG(C/A)] motif in their regulatory regions (defined as the 2 kb upstream of the start codon or until the neighbour gene, when it is closer). The sequence of the G-box in each gene is shown in blue, and the level of concordance with the consensus G-box motif is expressed as the P value of the match (determined using the FIMO v5.0.5 function in the MEME suite). Expression changes upon heat-shock for these genes are shown as in panel a.

To validate the observation that rapid activation of the cluster I genes upon heat-shock depends on PfAP2-HS and requires its D3, we analysed the heat-shock response of pfap2-hs knockout parasite lines of different genetic backgrounds and several 3D7-A mutant subclones expressing PfAP2-HSΔD3 (Extended Data Fig. 4). In all knockout and mutant lines examined, the hsp70–1 response to heat-shock was delayed and of reduced magnitude. These experiments also confirmed that the hsp90 response is weaker than the hsp70–1 response, and is delayed in PfAP2-HS mutants.

Extended Data Fig. 4

Changes in hsp70–1, hsp90 and PF3D7_1421800 transcript levels in parasites lacking the entire PfAP2-HS or D3.

Fold-increase in transcript levels (determined by RT-qPCR, normalized against serine--tRNA ligase) during and after heat-shock (HS) starting at 33–35 (a) or 30–35 (b) hpi, relative to cultures maintained in parallel without heat-shock (37 or 35°C). In panel a, values for three individual 3D7-A subclones carrying or not the Q3417X mutation are shown as dotted lines, whereas the average of the three subclones is shown as a continuous line. In panel b, the mean of n=2 independent biological replicates is shown.

PfAP2-HS-independent transcriptome alterations induced by heat-shock

Genes in other clusters (II-VI) of our transcriptomic analyses showed changes in expression during heat-shock that were independent of PfAP2-HS; these changes were more pronounced in the mutant than in the wild type lines (Fig. 2a). Indeed, more genes with altered transcript levels were identified in the heat-shock sensitive 10G and 10E_Δpfap2-hs lines than in 10E (Fig. 2b). Furthermore, the alterations in clusters II-VI transcripts persisted 2 h after heat-shock in both heat-shock-sensitive lines, whereas in 10E the majority of transcripts returned to basal levels (Fig. 2a). This suggests that many of these altered transcripts reflect unresolved cell damage or death. In 10E, the rapid PfAP2-HS-dependent response may protect cells from damage and enable rapid recovery from heat shock, thus limiting (in magnitude and duration) the changes in the expression of clusters II-VI genes that reflect cell damage. Indeed, after heat-shock the transcriptome of the 10G and 10E_Δpfap2-hs lines showed a more pronounced deviation from a reference transcriptome[28] than 10E (Fig. 2c). Global transcriptional analysis also revealed that heat-shock resulted in delayed IDC progression, again more pronounced in 10G and 10E_Δpfap2-hs than in 10E (Fig. 2d).

In addition to genes reflecting cell damage, clusters II-VI likely include some genes that participate in the PfAP2-HS-independent heat-shock response. In particular, clusters V-VI include several chaperone-encoding genes upregulated during heat-shock, although at a later time point than cluster I genes (Fig. 2a). However, the expression of the majority of known P. falciparum chaperones[14] was not altered by heat-shock and, except for cluster I genes, the alterations occurred mainly in the mutant lines (Extended Data Fig. 5). To provide a clearer view of the wild-type heat-shock response, we analysed changes upon heat-shock in the 10E line alone. Overall, there was generally good concordance with the genes and processes altered upon heat-shock described in a previous study using non-synchronized cultures[20] (Extended Data Fig. 6, Supplementary Table 3). Altogether, we conclude that a number of genes are up or downregulated during heat-shock and some may contribute to heat-shock protection through PfAP2-HS-independent responses, but in the absence of the rapid PfAP2-HS-dependent activation of cluster I genes these responses are insufficient to ensure parasite survival at febrile temperatures.

Extended Data Fig. 5

Transcript level changes upon heat-shock in chaperone-encoding genes.

Log2 expression foldchange [heat-shock (HS) relative to control (35°C) conditions, as in Fig. 2a] for all chaperone-encoding genes described by Pavithra and colleagues[14]. Columns at the left indicate presence of the G-box[23] or tandem G-box (TdGbox) in the upstream region, and log2 fold-change during heat-shock in a previous study[20] (Oakley).

Extended Data Fig. 6

Transcriptomic characterization of the heatshock response in parasites expressing complete PfAP2-HS (10E line).

Log2 expression fold-change [heat-shock (HS) relative to control (35°C) conditions] in the wild-type 10E line determined by microarray analysis. Genes with a fold-change ≥2 at any of the time points analysed are shown. The mean log2 expression fold-change (with 95% confidence interval) and representative enriched GO terms are shown for each cluster. Columns at the left indicate fold-change during heatshock in a previous study[20] (Oakley), and annotation as chaperone[14]. Ten genes had values out of the range displayed (actual range: −3.89 to +4.03).

Genome-wide analysis of PfAP2-HS binding sites

To determine the genome-wide occupancy of PfAP2-HS, we analysed a parasite line expressing endogenous HA-tagged PfAP2-HS (Extended Data Fig. 2b) using chromatin immunoprecipitation followed by sequencing (ChIP-seq). The main binding site of PfAP2-HS coincides with the position of the tandem G-box in the upstream region of hsp70–1 (Fig. 2e, Extended Data Fig. 7, Supplementary Table 4). This is the only binding site with a median fold-enrichment >10 (ChIP versus input) that was consistently detected in five independent ChIP-seq biological replicates, revealing an extremely restricted distribution of PfAP2-HS binding. Similar enrichment was observed in control and heat-shock conditions using ChIP-seq and ChIP-qPCR (Extended Data Fig. 7), which suggests that PfAP2-HS binds constitutively to this site and is activated in situ by heat-shock. This is reminiscent of yeast HSF1, which binds the promoter of hsp70 and most of its other target promoters under both basal and heat-shock conditions[5]. Enrichment for PfAP2-HS at the hsp90 promoter also coincided with the position of the G-box but was weaker and a significant peak at this position was called in only one of the replicates (Fig. 2e, Extended Data Fig. 7, Supplementary Table 4). No enrichment was observed at the promoter of cluster I gene PF3D7_1421800, which lacks a G-box motif. The only other site consistently enriched for PfAP2-HS binding, albeit at much lower levels than hsp70–1, was the small nucleolar RNA snoR04 (PF3D7_0510900) locus (Extended Data Fig. 7), which also lacks a G-box and was not upregulated during heat-shock.

Extended Data Fig. 7

ChIP analysis of the chromosomal distribution of PfAP2-HS.

a, ChIP-seq analysis of HA-tagged PfAP2-HS. Number of reads of ChIP (IP) and input (in) tracks, and log2-transformed ChIP/input ratio tracks (IP/in) for five independent biological replicates [three including heat-shock (HS) and 37°C conditions, two including only the 37°C condition]. Snapshots are shown for the three genes in cluster I (Fig. 2a) and snoR04. Binding at the hsp70–1 and hsp90 promoters coincides with the position of a tandem G-box motif, whereas PF3D7_1421800 and snoR04 lack a G-box. b, Peaks present in ≥3 out of 5 replicate ChIP-seq experiments (37°C) or ≥2 out of 3 replicate experiments (heat-shock) and with a MACS score >100 in each positive replicate. c, ChIP-qPCR analysis of HA-tagged PfAP2-HS binding at selected loci, in cultures exposed to heat-shock (HS) or control (37°C) conditions (mean and s.e.m. of % input in n=3 independent biological replicates). No significant difference (P<0.05) was observed between 37°C and heat-shock using a two-sided unpaired t-test.

Growth defects under basal conditions in parasite lines mutated for PfAP2-HS

Both knockout lines of 3D7 origin (10E_Δpfap2-hs and 10G_Δpfap2-hs) showed severe temperature-dependent growth defects in the absence of heat-shock. They grew at similar rates to the parental lines at 35°C, but their growth was markedly reduced at 37°C or 37.5°C (Fig. 3a). The D10_Δpfap2-hs line also had clearly reduced growth at 37°C compared to 35°C, whereas the HB3_Δpfap2-hs line did not (Fig. 3b). Both 3D7 Δpfap2-hs lines also showed a reduced number of merozoites per schizont, especially at 37°C or 37.5°C (Fig. 3c), which partly explains the reduced growth rate. Additionally, even at 35°C, the duration of the IDC was ~4 h longer in both knockout lines (Fig. 3d), which is reminiscent of the slower life cycle progression observed in parasites under proteotoxic stress[29]. In contrast, no growth rate or life cycle duration differences were observed between the 10G (PfAP2-HSΔD3) and 10E lines, indicating that D3 is necessary for heat-shock survival but not for growth under nonstress (37°C) conditions (Fig. 3a,c-d). Normal growth at 37°C but low heat-shock survival was also observed in transgenic lines in which bulky C-terminal tags were added to the C-terminus of endogenous PfAP2-HS, suggesting interference of the tag with the function of D3, located only 18 amino acids from the end of the protein (Fig. 1a, Extended Data Fig. 2).

Fig 3.

Phenotypes of parasite lines lacking PfAP2-HS.

a, Growth rate of Δpfap2-hs and parental lines of 3D7 genetic background at different temperatures (mean and s.e.m. of n=4 independent biological replicates). P values were calculated using a two-sided unpaired t-test (10E_Δpfap2-hs: 37 vs. 35°C, P=2.3 ×10−3; 37.5 vs. 35°C, P=1.7 ×10-4. 10G_Δpfap2-hs: 37 vs. 35°C, P=0.011; 37.5 vs. 35°C, P=0.001). Only significant P values (P<0.05) are shown. b, Same as in panel a for parasite lines of HB3 and D10 genetic background (mean and s.e.m. of n=4 independent biological replicates). c, Number of merozoites per schizont (median and quartiles box with 10–90 percentile whiskers). Values were obtained from 100 schizonts for each parasite line and condition. d, Duration of the asexual blood cycle. The cumulative percent of new rings formed at each time point is shown (mean of n=2 independent biological replicates).

Genome-wide sequence analysis has previously found that nonsense mutations arise in pfap2-hs during adaptation to culture conditions[30,31], but are virtually absent from clinical isolates (in the www.malariagen.net/data/pf3k-5 dataset[32], only one out of >2,500 isolates carries a high-confidence SNP resulting in a premature stop codon). The lack of mutations observed in clinical isolates suggests that there is a selection against loss-of-function mutations in PfAP2-HS during human infections, where parasites are frequently exposed to febrile conditions. We exposed a culture-adapted isolate in which ~50% of the parasites carried a mutation that results in truncation of PfAP2-HS before its first AP2 domain[30] (monoclonal Gambian Line 1, PfAP2-HS_ΔD1–3) to one round of heat-shock (41.5 °C, 3 h), and found that at the next generation only ~20% of the parasites carried the mutation. In control cultures maintained in parallel without heat-shock, the frequency of the mutation remained stable (Fig. 4a-b). This result indicates strong selection by heat-shock against parasites carrying the PfAP2-HS truncation. In contrast, there was relatively weak selection against mutants during culture either at 35°C or 37°C, as the prevalence of the mutation only decreased from ~50% to ~20% after culturing for 23 generations at either temperature (Fig. 4c). Consistent with these results, a subclone carrying the mutation (1H) was more sensitive to heat-shock than a wild-type subclone (4E), but both showed no measurable difference in growth at 35°C or 37°C (Fig. 4d-f). Together, these results indicate that PfAP2-HS is essential for heat-shock survival in all the genetic backgrounds tested. However, PfAP2-HS is necessary for normal progression through the IDC at 37°C in only specific genetic backgrounds (i.e., 3D7 and D10), whereas in others (HB3 and the Gambian isolate) it is not essential.

Fig 4.

Characterisation of a cultured-adapted field isolate with mutations in pfap2-hs.

a, Schematic of wild-type PfAP2-HS and PfAP2-HS_ΔD1–3 occurring in Line 1 from The Gambia after culture adaptation (C to G mutation at codon 931, S931X). The position of the AP2 domains is indicated (D1–3). b, Frequency of the mutation (as determined by Sanger sequencing) in culture-adapted Line 1 before (Pre) and after (Post) performing a heat-shock (HS) at the trophozoite stage and culturing for an additional cycle (mean of n=2 independent biological replicates). c, Frequency of the mutation during culture at different temperatures. Day 0 is when the frozen stock from The Gambia (culture-adapted for 91 days) was placed back in culture. d, Sanger sequencing determination of the presence or absence of the mutation at codon 931 in Line 1 subclones 4E and 1H. e, Heat-shock survival of tightly synchronised 4E and 1H cultures (mean and s.e.m. of n=4 independent biological replicates). The P value was calculated using a two-sided unpaired t-test. f, Growth rate of 4E and 1H at different temperatures (mean and s.e.m. of n=5 independent biological replicates). No significant difference (P<0.05) was observed between growth at 35°C and 37°C using a two-sided unpaired t-test.

Transcriptional alterations in parasites lacking PfAP2-HS under basal conditions

To gain insight on the molecular basis of the growth defects of some of the knockout lines, we compared the trophozoite transcriptome of 10E_Δpfap2-hs with that of the parental 10E under basal (no heat-shock) conditions. This revealed only a small set of genes with a ≥2 fold-decrease in transcript levels, which included hsp70–1, the direct PfAP2-HS target snoR04 RNA and several genes mainly involved in ribosome formation. Transcript levels for hsp90 were also reduced (<2 fold-decrease) in the knockout line (Extended Data Fig. 8a-b, Supplementary Table 2). Reduced hsp70–1 and hsp90 transcript levels under basal conditions in 10E_Δpfap2-hs mature trophozoites were independently confirmed by reverse transcription–quantitative PCR (RT-qPCR), and also observed at the late ring stage and in the knockout lines of D10 and HB3 genetic background (Extended Data Fig. 8c). These results indicate that PfAP2-HS contributes to regulating the basal expression of the same chaperone-encoding genes that it activates upon heat-shock, among a few other genes. Together with the observation that the growth defect of the pfap2-hs knockout lines is attenuated at 35°C, this suggests that knockout parasites have reduced proteostasis capacity, such that at 37°C they are at the edge of proteostasis collapse. Parasite lines that can grow normally at 37°C in spite of PfAP2-HS deletions including the three AP2 domains therefore must have alternative pathways active to ensure basal proteostasis. We hypothesise that mutant parasites expressing truncated PfAP2-HS are frequently selected under culture conditions because the truncations do not pose a fitness cost at 37°C in the lines in which they appear, and they prevent unnecessary activation of the heat-shock response, which can be detrimental[33], by unintended mild stress that may occur during culture.

Extended Data Fig. 8

Transcriptional changes associated with PfAP2HS deletion under basal (no heat-shock) conditions.

a, Changes in transcript levels in the absence of heat-shock for genes with an average expression fold-change >2 between 10E_Δpfap2-hs and 10E. Values are the log2 of the average expression fold-change relative to 10E across the time period compared (~27–30.5 hpi). Genes artificially modified or introduced in the knockout line, which serve as controls, are shown at the bottom (their values are out of the range displayed). The column at the left indicates the presence of the G-box[23]. b, Expression plots for selected genes under basal conditions. Expression values are plotted against statistically-estimated parasite age, expressed in h post-invasion (hpi). Grey shading marks the interval used to calculate the average expression fold-change. c, RT-qPCR analysis of hsp70–1 and hsp90 transcript levels in pfap2-hs knockout (KO) lines compared to their wild type (WT) controls (the parental line for each knockout line) under basal conditions. Expression values are normalized against serine--tRNA ligase, and expressed as the fold-change (FC) in the knockout versus control lines. The mean of n=3 (10E, 30–35 hpi) or n=2 (others) independent biological replicates is shown.

PfAP2-HS-deficient parasites show increased sensitivity to artemisinin

Artemisinins are potent antimalarial drugs that kill parasites by causing general protein damage[34,35]. Resistance to artemisinin is associated with mutations in the Kelch13 protein[36,37] and involves cellular stress response pathways such as the ubiquitin/proteasome system and the ER-based unfolded protein response (UPR)[34,35,38,39]. Since PfAP2-HS regulates the expression of key chaperones, we tested the sensitivity of PfAP2-HS-deficient lines to dihydroartemisinin (DHA), the active metabolite of artemisinins. In all four different genetic backgrounds (3D7, D10, HB3 and Gambian isolate), knockout of pfap2-hs (or truncation before D1) resulted in higher sensitivity to a pulse of DHA than in lines with full PfAP2-HS, both at the ring or the trophozoite stage, whereas 10G showed increased sensitivity only when exposed at the trophozoite stage (Fig. 5a). These results indicate that deletion of PfAP2-HS renders parasites more sensitive to chemical proteotoxic stress, in addition to heat-shock, likely as a consequence of basal defects in cellular proteostasis. We reasoned that if parasites lacking the PfAP2-HS protein bear constitutive proteome defects, they should have low tolerance to disruption of other factors involved in proteostasis maintenance. Indeed, the 10E_Δpfap2-hs line was more sensitive to the proteasome inhibitor epoxomicin than the parental 10E line or the 10G line (Fig. 5b). Furthermore, after heat-shock, there was more accumulation of polyubiquitinated proteins in the knockout line than in 10E or 10G, reflecting higher levels of unresolved protein damage (Extended Data Fig. 9a). We also assessed the links between the PfAP2-HS-driven heat-shock response and the other main cell stress response pathway, the UPR. Using phosphorylation of eIF2α as a UPR marker, we found that the UPR does not depend on PfAP2-HS and is not directly activated by heat-shock, because the marker was significantly elevated after heat-shock only in the knockout line (Extended Data Fig. 9b).

Fig 5.

Sensitivity of parasites lacking PfAP2-HS to proteotoxic conditions.

a, Survival (%) after a 3 h dihydroartemisinin (DHA) pulse at the ring or trophozoite (troph.) stage. Values are the mean and s.e.m. of n=3 (3D7-A and Line 1 genetic backgrounds) or mean of n=2 (HB3 and D10 genetic backgrounds) independent biological replicates. Mean IC50 for each line is shown (same colour code as the plots). b, Survival (%) after a 3 h epoxomicin pulse at the trophozoite stage. Values are the mean and s.e.m. of n=4 (100 nM) or n=3 (150 nM) independent biological replicates. In all panels, P values were calculated using a two-sided unpaired t-test (only for experiments with n≥3). Only significant P values (P<0.05) are shown.

Extended Data Fig. 9

Analysis of proteome stress and unfolded protein response (UPR) markers in pfap2-hs mutants.

a-b, Western blot analysis (representative of n=4) of polyubiquitinated proteins (Ub) (a) or phosphorylated eIF2α (eIF2α-P) (b) immediately after a 3 h heat-shock (3 h HS) and 2 h later (2 h post HS). Histone H3 is a loading control. DHA was used as a positive control, as it is a known inducer of the UPR[29,38]. The Log2 of histone H3-normalized signal in heat-shock or DHA-treated cultures versus control cultures is shown at the bottom (mean and s.e.m. of n=4, except for the DHA control mean of n=2 independent biological replicates). P values were calculated using a two-sided unpaired t-test. Only significant P values (P<0.05) are shown. The position of molecular weight markers is shown (in kDa).

DISCUSSION

Our results show that the PfAP2-HS transcription factor is bound to the tandem G-box DNA motif in the promoter of the chaperone-encoding gene hsp70–1 and in response to febrile temperatures rapidly upregulates the expression of this gene and, to a lesser extent, hsp90 (Fig. 6a). Binding of PfAP2-HS to the G-box is mediated by D1[23], but rapid activation of hsp70–1 and hsp90 during heat-shock requires D3, which is not capable of binding DNA in vitro[23] and likely participates in protein-protein interactions or dimerization within the cell[40]. Other components of the protein folding machinery necessary for heat-shock survival[15-17,19] are either constitutively expressed or induced later, but the rapid PfAP2-HS-driven response is essential to avoid irreversible damage. Importantly, parasites lacking either the entire PfAP2-HS or its D3 cannot survive heat-shock.

Fig 6.

Model of the P. falciparum heat-shock response and phylogenetic analysis of AP2-HS.

a, The P. falciparum heat-shock response involves rapid upregulation of the expression of a very restricted set of chaperones by PfAP2-HS. The PF3D7_1421800 gene (in brackets) shows PfAP2-HS-dependent increased transcript levels upon heat-shock, but PfAP2-HS binding was not detected in its promoter, and it lacks a G-box. The main defects associated with PfAP2-HS deletion or truncation, under heat-shock or basal conditions, are listed. b, Phylogenetic analysis of the protein sequence of AP2-HS orthologs in Plasmodium spp. c, Schematic of the domain structure of AP2-HS orthologs in Plasmodium spp. The position of the AP2 domains (D1–3) is based on domains identified in PlasmoDB release 50, except for those marked with an asterisk that were annotated manually based on sequence alignments.

Although the sequence and domain organization of PfAP2-HS does not show any similarity with HSF1, the conserved master regulator of the heat-shock response in most eukaryotes, from yeast to mammals[3,6], it serves an analogous role. HSF1 regulates a compact transcriptional program that includes the hsp70 and hsp90 genes[4,5]. In yeast, the only essential role of this transcription factor is activating hsp70 and hsp905, the same chaperone-encoding genes activated by PfAP2-HS during heat-shock. In addition to its role in the protective heat-shock response, PfAP2-HS is essential for growth at 37°C in some P. falciparum genetic backgrounds. The function of PfAP2-HS under basal conditions is independent of its D3. Several lines of evidence suggest that the role for PfAP2-HS under basal conditions involves proteostasis maintenance (Fig. 6a), similar to yeast HSF1[5]. In other organisms, the heat-shock response mediates protection against different types of proteotoxic stress, in addition to high temperature[1,3]. Here we report that parasites lacking PfAP2-HS have increased sensitivity to artemisinin, and future research will be needed to establish the precise role of the P. falciparum heat-shock response in protection against different types of stress. We note that orthologs of pfap2-hs are present in all Plasmodium spp. analysed (Fig. 6b-c and Extended Data Fig. 10), including murine Plasmodium species that do not induce host fever. This observation suggests that, at least in these species, the heat-shock response regulated by AP2-HS may play a role in protection against different conditions.

Extended Data Fig. 10

Sequence alignment of the three AP2 domains (D1-D3) present in AP2-HS orthologs in Plasmodium spp.

Dots indicate identity with the amino acid in the first sequence.

Finally, while several ApiAP2 transcription factors regulate life cycle transitions in malaria parasites[24,41-43], PfAP2-HS controls a protective response to a within-host environmental challenge. Our findings that the PfAP2-HS transcription factor regulates the activation of a protective heat-shock response settles the long-standing question of whether malaria parasites can respond to changes in within-host environmental conditions with specific transcriptional responses[44].

ONLINE METHODS

Parasite cultures.

The 3D7-A stock of the clonal P. falciparum line 3D7[45], the 3D7-A subclones 10G, 1.2B, 10E, 4D, 6D, 1.2F, W4–1, W4–2, W4–4 and W4–5[46,47], the HB3B[48] (mosquito and chimpanzee-passaged HB3, provided by Osamu Kaneko, Ehime University, Japan) and D10[49] (provided by Robin F. Anders, La Trobe University, Australia) clonal parasite lines, and the culture-adapted Line 1 from The Gambia[30] have been previously described. The heat-shock-selected lines 3D7-A-HS r1 and r2 were derived from 3D7-A by exposing cultures to a 3 h heat-shock (41.5°C) at the trophozoite stage for five consecutive cycles (each replicate, r1 and r2, is a fully independent selection from the 3D7-A stock), and the 3D7-A r1 and r2 lines are cultures maintained in parallel at 37°C without heat-shock[21]. Parasites were cultured in B+ erythrocytes at a 3 % haematocrit under standard culture conditions in RPMI-based media containing Albumax II (without human serum), in a 5% CO2, 3% O2, balance N2 atmosphere (except for cultures for ChIP-seq experiments, in which O+ erythrocytes were used). Regular synchronization was performed using 5 % sorbitol lysis, whereas tight synchronization (1, 2 or 5 h age window) was achieved by Percoll purification followed by sorbitol treatment 1, 2 or 5 h later. All cultures were regularly maintained at 37°C, with the exception of the pfap2-hs knockout lines that were maintained at 35°C. For experiments performed in parallel with the knockout lines and other parasite lines, all cultures were maintained at 35°C for at least one cycle before the experiment.

Generation of transgenic parasite lines.

We used two single guide RNAs (hereafter referred to as sgRNA or guide) to knockout pfap2-hs (11,577 bp) using the CRISPR/Cas9 system[50] (Extended Data Fig. 2a, Supplementary Table 5). One guide targets a sequence near the 5’ end of the gene (position 866–885 from the start codon) whereas the other recognizes a sequence near the 3’ end (positions 11,486–11,505). The 5’ guide was cloned into a modified pL6-egfp donor plasmid[50] in which the yfcu cassette had been removed by digestion with NotI and SacII, end blunting and re-ligation. 5’ and 3’ homology regions (HR1: positions −2 to 808 of the gene; HR2: positions +11,520 of the gene to 490 bp after the STOP codon) were also cloned in this plasmid, flanking the hdhfr expression cassette, to generate plasmid pL7-pfap2hs_KO_sgRNA5’. The 3’ guide was cloned into a modified version of the pDC2-Cas9-U6-hdhfr[51] plasmid, in which we previously removed the hdhfr expression cassette by digesting with NcoI and SacII, end blunting and re-ligation, and replaced the BbsI guide cloning site by a BtgZI site. The resulting plasmid was named pDC2_wo/hdhfr_pfap2hs_sgRNA3’. All guides were cloned using the In-Fusion system (Takara) as described[50], whereas homology regions were PCR-amplified from genomic DNA and cloned by ligation using restriction sites SpeI and AflII (HR1), and EcoRI and NcoI (HR2).

For constructs aimed at C-terminal tagging of pfap2-hs using CRISPR/Cas9 (10E_pfap2-hs_eYFP-Cterm and 10E_pfap2-hs_3xHA-Cterm lines) we used a guide corresponding to positions 11,508–11,527 of the gene (Extended Data Fig. 2b-c, Supplementary Table 5). The guide was cloned in the pDC2-Cas9-U6-hdhfr[51] plasmid to obtain pDC2_pfap2hs_sgRNA-C. The donor plasmid for tagging with eYFP (pHR-C_pfap2hs_eYFP) was based on plasmid pHRap2g-eYFP[52], with the pfap2-g homology regions and hsp90 3’ sequence replaced by pfap2-hs homology regions. The 5’ homology region (HR1) was generated with a PCR-amplified fragment spanning from nucleotide 10,964 to the sequence of the guide (recodonized), and a 47 bp fragment (generated by annealing two complementary oligonucleotides) consisting of a recodonized version of the remaining nucleotides to the end of the gene. The two fragments were cloned simultaneously, using the In-Fusion system, into SpeI-BglII sites. The 3’ homology region (HR2) was a PCR fragment spanning position +1 to +590 after the pfap2-hs STOP codon. It was cloned into XhoI-AatII restriction sites. The donor plasmid for 3xHA C-terminal tagging (pHR-C_pfap2hs_3xHA_hsp90–3’) was also based on plasmid pHRap2g-eYFP[52], with the eYFP coding sequence replaced by the 3xHA sequence (amplified from plasmid pHH1inv-pfap2-g-HAx341) and the same homology regions as in plasmid pHR-C_pfap2hs_eYFP (but HR2 was cloned, using the In-Fusion system, into EcoRI-AatII sites, because in this construct the hsp90 3’ region in pHRap2g-eYFP was maintained).

For N-terminal tagging (10E_pfap2-hs_eYFP-Nterm line), we cloned a guide targeting pfap2-hs positions 73–92 in the pDC2-Cas9-U6-hDHFRyFCU[53] plasmid to obtain plasmid pDC2_pfap2hs_sgRNA-N (Extended Data Fig. 2d, Supplementary Table 5). The donor plasmid (pfap2hs_HR-N_eYFP) consisted of a 5’ homology region (HR1) including positions −366 to −1 relative to the pfap2-hs start codon, the eYFP gene and an in frame 3’ homology region (HR2) spanning positions 4–756 of the gene (excluding the ATG) in which the nucleotides up to the position of the guide were recodonized. HR1 and HR2 were cloned into SacII-NcoI and SpeI-EcoRI sites, respectively. HR2 was amplified in two steps using a nested PCR approach to add the recodonized sequences. The eYFP fragment (PCR-amplified from plasmid pHR-C_pfap2hs_eYFP) was cloned using the In-Fusion system into SpeI/NcoI sites.

To tag PfAP2-HS with a 2xHA-ddFKBP domain tag (1.2B_ pfap2-hs_ddFKBP line) we used a single homologous recombination approach (Extended Data Fig. 2e). To generate the pfap2hs_HA-ddFKBP plasmid, we replaced the pfap2-g homology region in plasmid PfAP2-G-ddFKBP[41] by a PCR-amplified fragment including positions 9,551–11,574 of pfap2-hs in frame with the tag. The fragment was cloned using restriction sites NotI and XhoI. All oligonucleotides used to generate the plasmids are described in Supplementary Table 5. The relevant parts of all plasmids (i.e., the new sequences incorporated) were sequenced before transfection.

Transfections were performed by electroporation of ring stage cultures with 100 μg of plasmid (HA-ddFKBP tagging) or with a mixture of 12 μg linearized donor plasmid and 60 μg of circular Cas9 plasmid (CRISPR-Cas9 system). Linearization was achieved by digestion with the PvuI restriction enzyme (cleaving the amp resistance gene of the donor plasmid). Transfected cultures were selected with 10 nM WR99210 for four days as previously described[53] (transfections using the CRISPR-Cas9 system), or with continuous WR99210 pressure until parasites were observed, followed by 3 off/on drug cycles and subcloning by limiting dilution (transfections with the pfap2hs_HA-ddFKBP plasmid). In all cases, to assess correct integration we used analytical PCR of genomic DNA (Extended Data Fig. 2) with specific primers (Supplementary Table 5).

Heat-shock resistance assay.

Heat-shock was always performed on cultures at the mature trophozoite stage unless otherwise indicated. To measure survival to heat-shock, cultures were tightly synchronized to a defined age window, diluted to 1% parasitaemia, split in two identical petri dishes (heat-shock and control) maintained in independent air-tight incubation chambers, and exposed to heat-shock when the majority of parasites were at the mature trophozoite stage (typically ~30–35 h post-invasion, hpi; Δpfap2-hs lines were tightly synchronized 3 h earlier than the other lines but exposed to heat-shock in parallel to account for their slower IDC progression). The exception was experiments to screen many subclones (i.e., Fig. 1f) or to characterize transgenic parasite lines (i.e., Extended Data Fig. 2), in which cultures were only sorbitol-synchronized and heat-shock performed ~20–25 h after sorbitol lysis (mature trophozoite stage). For heat-shock, the full incubation chamber was transferred to an incubator at 41.5°C for 3 h, and then placed back to 37 or 35°C (the latter temperature was used for all lines in experiments including the pfap2-hs knockout lines). The chamber with the control cultures was always maintained at 37 or 35°C. After reinvasion (typically ~60–70 h after synchronization to ensure that all parasites had completed the cycle, including parasites subjected to heat-shock that show delayed progression through the IDC), parasitaemia of control and heat-shock-exposed cultures was measured by flow cytometry using a FACScalibur flow cytometer (Becton Dickinson) and SYTO 11 to stain nucleic acids (Supplementary Fig. 1), as previously described[54].

Phenotypic characterization.

To determine the growth rate (increase in parasitaemia between consecutive cycles) at different temperatures, the parasitaemia of sorbitol-synchronized cultures was adjusted to 1% and then accurately determined by flow cytometry. Cultures where then split between two or three dishes maintained in parallel in incubators at the different temperatures tested. Parasitaemia was again determined by flow cytometry at the next cycle to determine the growth rate. To measure the duration of the IDC (at 35°C) in the different parasite lines we used a recently described method based on synchronization to a 1 h age window achieved by Percoll-purification of schizonts followed by sorbitol lysis 1 h later[54]. The determination of the number of merozoites per fully mature schizont was based on light microscopy analysis of Giemsa-stained smears from Percoll-purified schizonts[54]. DHA (Sigma no. D7439) or epoxomicin (Selleckchem no. S7038) sensitivity was measured after exposing tightly synchronized cultures (1% parasitaemia) at the ring (10–15 hpi, DHA only) or trophozoite (30–35 hpi, DHA or epoxomicin) stage to a 3 h pulse of the compounds at different concentrations (DHA, 2.5, 5, 10, 20 or 200 nM, which is lower than the ~700 nM plasma concentration after patient treatment that kills the vast majority of sensitive parasites[35]; epoxomicin, 100 or 150 nM, which is higher than the reported 7.7 nM IC50 after exposing parasites for 50 h[55] and similar to the concentration used in previous studies with a 3 h pulse[29,38]). Parasitaemia was measured by light microscopy analysis of Giemsa-stained smears at the next cycle (typically 70–80 h after Percoll-sorbitol synchronization). For these experiments, the Δpfap2-hs lines were tightly synchronized 3 h earlier than the other lines but exposed to DHA or epoxomicin in parallel (13–18 or 33–38 hpi), to account for their slower IDC progression. For DHA experiments, drug concentrations were log transformed and percent survival data were fit to sigmoidal dose-response curves to calculate the IC50 values using GraphPad Prism.

Transcriptional analysis by RT-qPCR.

RNA from tightly synchronized cultures exposed to heat-shock and their controls was obtained using the Trizol method, DNAse-treated and purified essentially as described. Reverse transcription and qPCR analysis of cDNAs were also performed as described before[56,57]. In brief, a mixture of random primers and oligo (dT) were used for reverse transcription, and for qPCR we used the PowerSYBR Green Master Mix (Applied biosystems) and the standard curve method (each plate included a standard curve for each primer pair). All primers used are listed in Supplementary Table 5. Unless otherwise indicated, transcript levels were normalized against serine--tRNA ligase (PF3D7_0717700), which shows relatively stable expression throughout the IDC.

Transcriptomic analysis using microarrays.

To compare the transcriptome of control and heat-shock-adapted 3D7-A parasite lines across the IDC we used previously described two-colour long oligonucleotide-based glass microarrays[21]. RNA was obtained from tightly synchronized cultures (5 h age window) at 8–13, 16–21, 24–29, 32–37 and 40–45 hpi. All samples (Cy5-labeled) were hybridized together with a reference pool (Cy3-labeled) consisting of a mixture of equal amounts of cDNA from rings, trophozoites and schizonts from control and heat-shock-adapted lines. Comparative genome hybridization was used to identify potential transcript level differences attributable to genetic deletions or duplications. 5,142 genes produced useful data. Sample preparation, microarray hybridization and data analysis were performed essentially as described[21].

To analyse the transcriptome of 10E, 10G and 10E_Δpfap2-hs parasite lines under control and heat-shock conditions, we used two-colour long oligonucleotide-based custom Agilent microarrays[58]. The microarray design was based on Agilent design AMADID no. 037237[58,59], but we modified it as previously described (new design AMADID no. 084561)[60]. RNA was obtained from cultures synchronized to a 5 h age window at a ~2.5% parasitaemia. Given the slower IDC progression of 10E_Δpfap2-hs, cultures of this parasite line were synchronized to 0–5 hpi 3 h earlier than 10E and 10G cultures, such that at the time of starting heat-shock (in parallel for all lines) all cultures were approximately at the same stage of IDC progression. Heat-shock was started at 30–35 hpi (33–38 hpi for the 10E_Δpfap2-hs line) and samples collected before, during and after heat-shock as indicated. RNA was prepared using the Trizol method. Sample preparation and microarray hybridization were performed essentially as described[59]. All samples (Cy5-labeled) were hybridized together with a reference pool (Cy3-labeled) consisting of a mixture of equal amounts of cDNA from rings, trophozoites and schizonts from 3D7-A. Microarray images were obtained using a DNA Microarray Scanner (no. G2505C, Agilent Technologies) located in a low ozone area, and initial data processing was performed using the GE2 _1105_Oct12 extraction protocol in Agilent Feature Extraction 11.5.1.1 software.

Agilent microarray data was analysed using Bioconductor in an R environment (R version 3.5.3). For each individual microarray, we calculated Cy3 and Cy5 background signal as the median of the 100 lowest signal probes for each channel, and probes with both Cy3 and Cy5 signals below three times the array background were excluded. Gene level log2(Cy5/Cy3) values, statistical estimation of parasite age[61] and estimation of average expression fold-differences across a time interval (for the comparison between parasite lines in the absence of heat-shock) were performed as described[21]. The log2 of the expression fold-change upon heat-shock was calculated, for each gene and time point, as the log2(Cy5/Cy3) in the heat-shock-exposed sample minus the log2(Cy5/Cy3) in the control sample at the same parasite age, calculated using linear interpolation in the log2(Cy5/Cy3) versus estimated age plot. Visual inspection was used to exclude from further analysis genes with apparent artefacts. Genes missing data for ≥2 time points (or ≥1 for the comparison between parasite lines in the absence of heat-shock across a time interval), or with values within the lowest 15th percentile of expression intensity (Cy5 sample channel) in all samples, were also excluded from further analysis. 4,964 genes produced useful data.

To assess the level of similarity between the transcriptome of our samples and a reference non-stressed transcriptome with high temporal resolution (HB3 line)[28] we calculated the Pearson correlation between each sample and the time point of the reference transcriptome with which it has higher similarity. Heatmaps and hierarchical clustering based on Spearman (Fig. 2) or Pearson (Extended Data Fig. 6) correlation were generated using Multiple Experiment Viewer (MeV) 4.9[62]. Expression trend plots for each cluster were generated using ggplot2, with LOESS smoothing, and Venn diagrams using the eulerr package (both in an R environment). Motif analysis (5 to 8 bp) was performed using MEME 5.0.3 software. Functional enrichment analysis using GO terms annotated in PlasmoDB release 43 was performed using Ontologizer 2.1 software[63] with the topology-elim method[64]. Gene set enrichment analysis (GSEA) was performed using GSEA v3.0 Preranked[65].

Whole-genome sequencing analysis, analysis of publicly available genome sequences from field isolates and phylogenetic analysis.

To sequence the full genome of control and heat-shock-adapted 3D7-A lines (two biological replicates), we used PCR-free whole-genome Illumina sequencing. In brief, genomic DNA was sheared to ~150–400 bp fragments using a Covaris S220 ultrasonicator and analysed using an Agilent 2100 Bioanalyzer. For library preparation we used the NEBNext DNA Library Prep Master Mix Set for Illumina (no. E6040S) using specific paired-end TruSeq Illumina adaptors for each sample. After quality check by qPCR, we obtained >6 million 150 bp paired reads for each sample using an Illumina MiSeq sequencing system. After checking reads quality (FastQC algorithm) and trimming adaptors (Cutadapt algorithm), sequence reads were mapped to the PlasmodDB P. falciparum 3D7 reference genome release 24 (https://plasmodb.org/plasmo/) using the Bowtie2 local alignment algorithm and alignment quality was assessed using the QualiMap platform. Average genome coverage was 76 to 98-fold. To identify SNPs and small indels we followed the Genome Analysis Toolkit (GATK) best practices workflow, using SAMtools, PicardTools and GATK algorithms. Variant calling was performed using GATK-UnifiedGenotyper. Variants with low calling quality (Phred QUAL<20) and low read depth (DP<10) were filtered out using GATK-VariantFiltration, and only variants present in both biological replicates were considered. Differences in SNP/indel frequency between control and heat-shock-adapted lines were calculated for each SNP/indel using Microsoft Excel, and those showing <25% difference in any of the two replicates were filtered out. Genome Browse (Golden Helix) was used to visualize alignments and variants.

For the analysis of publicly available genome sequences, we used the Pf3K Project (2016): pilot data release 5 (www.malariagen.net/data/pf3k-5) containing the sequence of >2,500 field isolates. Only SNPs that passed all quality filters and did not fall within a region with multiple large insertions and deletions were considered to be high-confidence. Using these criteria, a single high-confidence polymorphism (occurring in a single isolate) was identified at the pfap2-hs gene (producing the C3168X mutation that results in a truncated PfAP2-HS protein that lacks D3).

For sequence alignment and construction of the phylogenetic tree (with the Neighbor-Joining method) we used Clustal Omega[66], with default parameters. From the tree without distance corrections obtained, the cladogram was generated using FigTree 1.4.4.

ChIP experiments and data analysis.

For ChIP experiments, synchronous 50 ml cultures at 2.5 to 5% parasitemia were harvested at the mid trophozoite stage. For replicates in which ChIP was performed in parallel under heat-shock and control conditions, cultures were split off from a single parent flask at the mid trophozoite stage. Control flasks were immediately returned to 37°C whereas heat-shock flasks were maintained at 41.5°C for 3 h before harvesting for ChIP analysis.

ChIP followed by qPCR or Illumina sequencing was performed as described[67] using the 3F10 rat anti-HA antibody (1:500; Roche no. 11867423001) to immunoprecipitate HA tagged AP2-HS, with the following modification: total chromatin was diluted 5-fold in dilution buffer following sonication. The Illumina HiSeq system was used to obtain 125 bp paired-end (replicates 1–3) or 150 bp single-end (replicates 4–5) reads.

ChIP-seq data analysis was performed essentially as described[67]. In brief, after trimming, quality control, mapping the remaining reads to the Plasmodium falciparum genome (PlasmoDB release 28) using BWA-MEM and filtering duplicated reads, peak calling was performed using MACS2[68] with a q-value cut-off of 0.01. Conversion to log2 coverage of immunoprecipitate/input was performed using DeepTools BamCompare, selecting the paired end parameter for all tools when analyzing experiments including control and heat-shock conditions. Overlapping intervals within called peaks for each dataset were determined using Bedtools MultiIntersect. The closest annotated gene coding sequence for each called peak was determined using Bedtools ClosestBed. To visualize aligned data, we used the Integrative Genomics Viewer (IGV).

ChIP samples were analysed by qPCR in triplicate wells with primers described in Supplementary Table 5. All primer pairs were confirmed to have between 80 and 110% efficiency using sheared genomic DNA as a template control. The percent input was calculated using the formula 100*2(Ct adjusted input – Ct IP).

Western blot.

Synchronized cultures at the mature trophozoite stage were exposed to a regular 3 h heat-shock or to a 1.5 h DHA pulse (10 or 100 nM, used as positive control for a condition known to produce proteotoxic stress and induce the UPR)[29,38]. Parasites were obtained using saponin lysis (0.15% w/v saponin) and pellets solubilized in 1x SDS-PAGE loading buffer with 4% β-mercaptoethanol and boiled at 95°C for 5 min. Proteins were resolved by SDS-PAGE on 4–20% TGX Mini-PROTEAN gels (Bio-rad) and transferred to nitrocellulose membranes (Bio-rad). After blocking with 5% (w/v) bovine serum albumin (Biowest) in TBS-T (0.1% Tween 20 in tris buffered saline), membranes were incubated at 4°C overnight with the following primary antibodies: rabbit anti-ubiquitin (1:1,000; Cell Signaling Technology no. 3933), rabbit anti-phospho-eIF2α (1:1,000; Cell Signaling Technology no. 3398) and rabbit anti-histone H3 (1:1,000; Cell Signaling Technology no. 9715). After incubation with a goat anti-rabbit IgG-peroxidase (1:5,000; Millipore no. AP307P) secondary antibody, peroxidase was detected using the Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) in an ImageQuant LAS 4000 imaging system. To control for equal loading, parts of the membranes corresponding to different molecular weight ranges were separately hybridized with different antibodies. Signal quantification was performed using ImageJ.

Statistical analysis.

Statistical analysis was performed using Microsoft Excel and GraphPad Prism. P values were calculated using a two-tailed t-test (equal variance). No adjustment for multiple comparisons was made. Only significant P values (P<0.05) are shown in the figures. No statistical analysis was performed for experiments involving only two replicates. In all cases, n indicates independent biological replicates (i.e., samples were obtained from independent cultures).

Genetic changes between 3D7-A-HS and control 3D7A lines identified by whole genome sequencing.

Frequency of SNPs and indels (% of reads containing the mutation) in two heat-shock-adapted 3D7-A lines (3D7-A-HS r1 and r2) and their respective controls (3D7-A r1 and r2).

Global transcriptional alterations in the 10E, 10G and 10E-Δpfap2-hs parasite lines exposed to heat-shock.

Log2 of the expression fold-change in cultures exposed to heatshock (HS) versus control (35°C, no heat-shock) cultures, during (1.5 and 3 h) or after (2 h post) heatshock, as in Fig. 2a. Only genes with a ≥4 fold-change (at any of the time points analysed) are shown. Genes are ordered by hierarchical clustering. Annotation as chaperone[14], presence of the G-box[23] or tandem Gbox (TdGbox) in the upstream region, log2 foldchange after heat-shock in a previous study[20] (Oakley) and GO terms are provided for each gene. For each cluster, significantly enriched motifs and enriched GO terms are shown. Gene set enrichment analysis (GSEA) of the comparative transcriptomic analysis between 10E and 10E_Δpfap2-hs lines under basal conditions is also included.

Log2 expression foldchange in cultures exposed to heat-shock (HS) relative to control cultures not exposed to heath-shock (35°C) in the wild-type 10E line, as in Extended Data Fig. 6. Only genes with a fold-change ≥2 at any of the time points analysed are shown. Annotation as chaperone[14], presence of the G-box[23] or tandem Gbox (TdG-box) in the upstream region, log2 foldchange during heat-shock in a previous study[20] (Oakley) and GO terms annotation is shown for each gene. For each cluster, significantly enriched GO terms are shown.

PfAP2-HS ChIP-seq peaks in cultures exposed to heat-shock or under control conditions.

List of peaks with a MACS2 Score greater than 100 in any of the replicates, with values for each individual replicate, and peaks identified in each individual sample with a qvalue<0.01.

List of oligonucleotides used in this study.