BLASTP analysis was used to identify the closest human and Plasmodium berghei orthologues of the PfeIF2α kinases. Catalytic domains of the putative PfeIF2α kinases as defined by the alignment of P. falciparum kinases 10 were aligned with the four human eIF2α kinases and other P. falciparum and human sequences that were selected to represent all kinase subfamilies. The sequences were aligned using the HMMER package against a profile generated from our previous kinome analysis 10. After removal of gaps and positions with a low quality of alignment, alternate phylogenies generated with the neighbour joining method were visualized using NeighbourNet implemented on SplitsTree version 4 24.

BLASTP searches of PlasmoDB using metazoan eIF2α sequences as queries identified PF07_0117 as the P. falciparum homologue of eIF2α, which was then confirmed by reciprocal analysis. Alignment of these sequences was performed using ClustalW.

Expression of recombinant GST fusion proteins was induced in E. coli (strain BL21, codon plus) with 0.25 mM Isopropyl Thiogalactoside (IPTG). After induction, bacteria were grown at 16°C overnight and the resulting bacterial pellets were stored at -20°C until use. All subsequent work was done on ice, centrifugation steps at 4°C. Protein extraction was performed by digestion of bacterial pellets for 5 min. with lysozyme (Sigma), followed by 10 min. in lysis buffer (1 × PBS, 2 mM ethylenediaminetetraacetic acid (EDTA), 1 mM dithiothreitol (DTT), 0.5% Triton ×100, 1 mM Phenyl Methyl Sulphonate (PMSF), Benzamidine Hydrochloride Hydrate (BHH), 1× complete cocktail protease inhibitors (Roche)). Bacterial lysates were sonicated at 20% amplitude (Bioblock Scientific, Vibracell 72405), 5 × 15 sec. pulses/15 sec. rest, and cleared by centrifugation 13000 g, 15 min. GST-fusion proteins were purified by incubation of cleared lysates on glutathione agarose beads (Sigma) for 2 hours, followed by four washes with lysis buffer and eluted for 20 min. in elution buffer (Tris 40 mM, pH8.7, 75 mM NaCl, 15 mM reduced glutathione). Protein concentration was monitored using the Bradford assay (Biorad reagent). Kinase assays were carried out immediately after purification.

Kinase reactions (30 μl) were carried out in a standard kinase buffer containing 20 mM Tris-HCl, pH 7.5, 20 mM MgCl₂, 2 mM MnCl₂, phosphatase inhibitors; 10 mM NaF, 10 mM β-glycerophosphate, 10 μM ATP and 0.1 MBq [γ-³²P] ATP, using 2 μg recombinant kinase, and 10 μg non-physiological substrate (α-casein, β-casein), or recombinant GST-PfeIF2α. Reactions were allowed to proceed for 30 min. at 30°C and stopped by addition of reducing Laemmli buffer, 3 minutes, 100°C. Samples were separated by SDS-PAGE and phosphorylation of kinase substrates assessed by autoradiography of the dried gels.

A gene disruption plasmid was produced for PF14_0423 in the plasmid pCAM-BSD 26 that contains the gene conferring resistance to blasticidin. The oligonucleotide pair GGGG

GGATCC

CGCCGGCG

Plasmodium falciparum clone 3D7 was cultured as previously described 27. In brief, asexual cultures were maintained in complete RPMI at a haematocrit of 5%, between 0.5% and 10% parasitaemia. Asexual growth cycle was analyzed by flow cytometric assessment of DNA content as previously described 28. Gametocytogenesis was induced as described previously 29; briefly, gametocyte cultures were set up at 0.5–0.7% parasitaemia in 6% haematocrit (using human blood not more than 7 days after the bleed), in an initial volume of 15 ml in 75 cm² flasks. Cultures were maintained for 4–5 days until 8–10% parasitaemia was reached and parasites appeared stressed, after which the volume was increased to 25 ml. For each clone a mixture of day 14 and day 17 gametocyte cultures were fed to Anopheles gambiae, through membrane feeders as described 29. Female mosquitoes were dissected 10 days post-infection and midguts examined by light microscopy for presence of oocysts. Sporozoite invasion of salivary glands was assessed by removal of salivary glands 16 days post-infection and examination by light microscopy. DNA was extracted from oocyst-positive midguts using previously published methods 30. Fisher's exact test was used to compare infection prevalence between oocyst and sporozoite stages, where appropriate.

Parasite pellets were obtained by saponin lysis: erythrocytes were centrifuged at 1300 g for 2 min. at room temperature, washed in an equal volume of Phosphate Buffered Saline (PBS), pH 7.5, and centrifuged at 1300 g for 2 min. at 4°C. Erythrocytes were lysed on ice by resuspension and repeated pipetting in 0.15% saponin in PBS. The PBS volume was then increased and parasites recovered by centrifugation at 5500 g for 5 min., at 4°C. After two washes in PBS, the parasite pellets were stored at -80°C.

Plasmodium falciparum 3D7 parasites and clonal lines of pfeik1^- and pfeik2^- parasites were synchronized to the late ring stage, cultured in complete RPMI at 2% haematocrit, and grown to approximately 8 – 10% parasitaemia. The parasites were washed two times in 1× PBS, equally partitioned and washed with either complete RPMI or RPMI medium lacking amino acids, after which, the parasites were re-plated in their respective medium. The plates were incubated at 37°C with 5% CO₂ for 5 hours. After 5 hours, one culture maintained in amino acid free medium was supplemented with complete RPMI, and re-incubated at 37°C for an additional 45 minutes. Post-incubation, the parasites were isolated by tetanolysin (List Biological) treatment, washed with 1× PBS buffer containing Complete™ protease inhibitor cocktail (Roche), 2 mM NaF, and 2 mM Na₃VO₄. Samples were resuspended in 2× SDS-Laemmli buffer. Parasite proteins were resolved by SDS-PAGE and transferred to nitrocellulose for immunoblotting.

Rabbit anti-phosphorylated eIF2α (Ser 51) was purchased from Cell Signaling Technology (Danvers, MA). Rat anti-BiP was acquired from the Malaria Research and Reference Reagent Resource Center (ATCC, Manassas, VA). Secondary antibodies used were conjugated with horseradish peroxidase (HRP). For immunoblotting, nitrocellulose membranes were blocked with 5% BSA in TBS-0.1% Tween 20 (TBST) for 1 hour at room temperature. Rabbit anti-phosphorylated eIF2α (Ser 51) was diluted 1:1000 in TBST. Rat anti-BiP was diluted 1:10,000 in TBST. Respective secondary antibodies were diluted 1:20,000. Bound antibodies were detected with Western Lightning™ Chemiluminescence reagent (Perkin Elmer).

To obtain genomic DNA, parasite pellets were resuspended in PBS and treated with 150 μg/ml proteinase K and 2% SDS at 55°C for 4 hours. The DNA was extracted using phenol/chloroform/isoamyl alcohol (25:24:1), and precipitated in ethanol with 0.3 M sodium acetate at -20°C. Restriction digests were carried out with HindIII. Probes were labelled with alkaline phosphatase using the Gene Images AlkPhos Direct Labelling kit (Amersham).

BLASTP searches of PlasmoDB using metazoan eIF2α sequences were used to identify PF07_0117 as the P. falciparum orthologue, which was confirmed by reciprocal analysis. The alignment of P. falciparum eIF2α with sequences from T. gondii, human, rice and E. cuniculi is shown in Figure 1A. Overall, the P. falciparum sequence shares ~70% identity with T. gondii eIF2α and ~50%, ~40% and ~28% with the orthologues in humans, rice and E. cuniculi, respectively. Importantly, the serine that is targeted for phosphorylation is conserved in all species. Furthermore, eIF2α contacts the kinase through a large number of residues that interact with the surface of the kinase domain. These residues are also conserved in most species, as are residues that protect the regulatory serine from the activity of other kinases 31 (Figure 1A); interestingly, several of these are not conserved in the E. cuniculi orthologue, which is consistent with the absence of eIF2α kinases in this organism 21.

The presence of the target serine residues, and of residues which in other species are involved in interaction with eIF2α kinases, suggests that PfeIF2α may be regulated by phosphorylation under stress conditions. To test this hypothesis, cultured intraerythrocytic parasites were starved of amino acids, and the phosphorylation status of PfeIF2α was monitored by western blot using an antibody that specifically recognizes the phosphorylated form (Ser51) of human eIF2α, reasoning that the high level of sequence conservation between the human and plasmodial sequences would allow cross-reaction of the antibody (Figure 1B). Indeed, the antibody recognized the expected 37-kDa band in parasite extract, and the intensity of the signal was considerably stronger in the lane containing extracts from parasites that had been stressed by amino-acid starvation than in extracts from unstressed parasites, despite equal quantities of the eIF2α factor (as quantitated with a non-phosphodependent antibody). Furthermore, this effect was removed by restoring the amino acids in the culture medium. This demonstrates that the P. falciparum equivalent residue of human eIF2α Ser51 is phosphorylated in response to starvation.

Bioinformatics approaches were then used to identify P. falciparum protein kinase(s) potentially responsible for this response. An analysis of the complete complement of P. falciparum protein kinases 10 identified a distinct phylogenetic cluster of three sequences, PF14_0423, PFA0380w and PFF1370, the latter of which (called PfPK4) had previously been characterized as an eIF2α kinase 22. Reciprocal BLASTP analysis using the putative catalytic domains as queries confirmed the homology of these three genes with the eIF2α kinase family. A Hidden Markov Model (HMM) was used to generate an alignment of the three P. falciparum sequences with those of human eIF2α kinases; sequences of kinases from other families were included as outgroups. The resulting alignment was used to generate a phylogenetic tree (Figure 2A), which clearly shows that the three P. falciparum genes cluster with the eIF2α kinases, as opposed to other kinase families, confirming their relatedness to this family. Interestingly PfeIK1 (PF14_0423), on which the present study focuses, clusters most closely with GCN2, which is suggestive of a role in response to nutrient levels.

The PF14_0423 gene model proposed in PlasmoDB 32 predicts a single intron that falls close to the 5' end of the sequence so that the kinase domain is encoded entirely within the second exon. All the residues that are required for catalytic activity 33 are present in the kinase domain, suggesting the gene encodes an active enzyme. The sequence shares the feature of insertions within the catalytic domain with other eIF2α kinases 34 (Figures 2B and 2C). Three of the human eIF2α kinases have N-terminal extensions containing regulatory domains; the fourth, GCN2, has extensions on either side of the kinase domain (reviewed in 35). As PfeIK1 has extensions on both sides of the catalytic domain, it is most similar to GCN2 not only in the sequence of its catalytic domain, as the phylogenetic tree (Figure 2A) demonstrates, but also in overall structure (Figure 1C). Furthermore, the C-terminal extension of PfeIK1 contains an "anti-codon binding" domain (Superfamily entry SSF52954) that may mediate binding to uncharged tRNAs, a function that is performed in GCN2 by the HisRS domain present in the C-terminal extension (Figure 1C) 18. This adds weight to the possibility that PfeIK1 is involved in the response to amino acid starvation, like GCN2. The other functional domains present in the GCN2 extensions were not recognisable in PfeIK1.

In order to verify that the pfeik1 gene encodes a functional kinase, the catalytic domain was expressed as a GST fusion protein in E. coli. A recombinant protein of the expected size (76 kDa) was obtained and purified for use in kinase assays. The protein appeared as a doublet in most preparations, with both bands reacting with an anti-GST antibody. Kinase assays were performed with α- or β-casein as substrates, in the presence or absence of GST-PfeIK1 (Figure 3A). A weak signal was detectable with β-casein on the autoradiogram even in the absence of the kinase, indicating a low level of contaminating kinase activity in the substrate itself. This signal was much stronger in the presence of GST-PfeIK1, and a signal was also observed with α-casein, which was not labelled in the absence of the kinase. Furthermore, a signal at a size matching that of the upper band in the GST-PfeIK1 doublet was also seen, indicating possible autophosphorylation, an established property of at least some mammalian eIF2α kinases, including GCN2 34363738. GCN2 autophosphorylation occurs on two threonine residues in the activation loop 36, only one of which conserved in PfeIK1 (Figure 2B). Autophosphorylation was more clearly seen in the absence of any exogenous substrate (Figure 3B). The possible functional relevance of PfeIK1 autophosphorylation remains to be determined. Taken together, these data suggest that PfeIK1 possesses catalytic activity. To ensure that the signals were not due to co-purified activities from the bacterial extract, the assays were repeated using a catalytically inactive mutant (Lys458→Met) of GST-PfeIK1. These reactions yielded an identical pattern as the reaction containing no recombinant kinase (Figure 3A), confirming that the phosphorylation of the caseins is due to GST-PfeIFK1, and that the recombinant kinase can autophosphorylate.

In order to establish whether PfeIK1 is an eIF2α kinase as predicted, its activity was tested towards recombinant P. falciparum eIF2α expressed as a 64 kDa GST fusion. Figure 3B (left lane) shows that GST-PfeIK1 can phosphorylate wild-type GST-PfeIF2α. The signal appears very weak, which may be explained by the fact that the recombinant kinase contains only the catalytic domain and may not mimic the enzyme in a fully activated, physiological status. Indeed, an activation mechanism for GCN2 has been proposed 37, in which a conformational alteration of the so-called "hinge region" of the catalytic domain is induced by uncharged tRNA binding to the HisRS domain, which would favour productive binding of ATP to the active site. Such a positive effect of the regulatory domain would not be possible with GST-PfeIK1, since it contains only the catalytic domain.

Consistent with the hypothesis that PfeIK1 may regulate translation through PfeIF2α phosphorylation, mutation of the predicted target for phosphorylation in the substrate (Ser59→Ala) prevents labelling with the recombinant enzyme (Figure 3B).

Microarray data available in PlasmoDB 139 indicate that pfeik1 is expressed in asexual parasites; it can be hypothesized that the kinase plays a role in the parasite's stress response, and may therefore (i) not be essential for the asexual cycle, and (ii) be involved in regulation of gametocytogenesis, similar to the function of a eIF2α kinase in T. gondii stage transition from tachyzoite to bradyzoite. P. falciparum clones that do not express PfeIK1 were generated to test these hypotheses. The strategy used to disrupt expression of the kinase relies on single cross-over homologous recombination, and has been used successfully to knock-out other P. falciparum protein kinase genes 4041. Briefly, a plasmid based on the pCAM-BSD vector 26 containing a cassette conferring resistance to blasticidin and an insert comprising the central region of the PfeIK1 catalytic domain, was transferred by electroporation into asexual parasites of the 3D7 clone. Homologous recombination is expected to generate a pseudo-diploid locus in which neither of the two truncated copies encodes a functional kinase: the 5' copy lacks an essential glutamate residue in subdomain VIII and all downstream sequence including the 3'UTR; the 3' copy lacks the both the promoter region and the essential ATP orientation motif in subdomain I (Figure 4A).

Blasticidin-resistant parasite populations were obtained and shown by PCR analysis to contain parasites whose pfeik1 locus was disrupted. Clonal lines deriving from two independent transfection experiments were established by limiting dilution, and their genotypes were analysed by PCR (Figure 4B). The amplicon corresponding to the wild-type locus was not detected in clones C1 and C8 (lane 1), but was observed in wild-type parasites (lane 5). In contrast, PCR products that are diagnostic of both the 5' (lanes 3 & 7) and 3' (lanes 4 & 8) boundaries of the integrated plasmid were amplified from C8, but not 3D7 parasites (lanes 11 & 12). The C1 and C8 clones also yielded a signal with primers that are specific for the transfection plasmid, and detect retained episomes or integrated concatemers. Integration was verified by Southern blot analysis of HindIII-digested genomic DNA (Figures 4C and 4D); the 12 kb band containing the wild-type locus is replaced in clones C1 and C8 by the expected two bands (10.4 kb and 6.8 kb) resulting from integration. The remaining 5.3 kb band is derived from linearized plasmid, or from digestion of concatemers of plasmid (which may or may not be integrated into the chromosome). These results confirm that the pfeik1 locus was indeed disrupted in clones C1 and C8, and demonstrate PfeIK1 is not required for completion of the asexual cycle in in vitro cultures. Additionally, asexual parasite cultures were synchronized and carefully monitored through several life cycles; samples were taken every 30 minutes and assessed for DNA content by flow cytometry 42. No significant difference was observed in asexual cycle duration of the parental 3D7 clone and that of pfeik1^- parasites; cycle times of 49.0 h +/- 0.5 and 49.2 h +/- 0.7, respectively, were measured (Figure 5).

To determine whether pfeik1^- parasites were defective in responding to amino acid-limitation, we cultured these parasites in RPMI medium containing either all or no amino acids and assayed for eIF2α phosphorylation through western blot analysis (Figure 6). We observed that pfeik1^- parasites were unable to modulate the phosphorylation state of eIF2α in response to changing amino acid conditions, in direct contrast to wild-type parental clone 3D7. A further control was provided by performing the assay using a parasite clone lacking PfeIK2, another enzyme related to eIF2α kinases (see Figure 2A; a full characterisation of PfeIK2 and pfeik2^- parasite clones is to be published elsewhere). The pfeik2^- parasites, which were generated using the same strategy as that described here for pfeik1 and were therefore also resistant to blasticidin, readily phosphorylated eIF2α in amino acid starvation conditions, like wild-type 3D7 parasites. This demonstrates that the abolition of eIF2α phosphorylation observed in pfeik1^- parasites is not due to non-specific effects resulting from the genetic manipulations performed to obtain the mutant clones. Taken together, these data identify PfeIK1 as a crucial regulator of amino acid starvation stress response of intra-erythrocytic parasites.

The pfeik1^- parasites were able to differentiate into gametocytes (data not shown). Further, qualitative results showed that pfeik1^- male gametocytes were competent to differentiate into gametes (in vitro exflagellation). To investigate whether PfeIK1 plays an essential role in subsequent life cycle stages, mosquitoes were fed with cultures of pfeik1^- gametocytes. The numbers of oocysts associated with midguts dissected 10 days post-feeding, and the numbers of mosquitoes with sporozoite-positive salivary glands 16 days post-feeding, were then determined. This revealed that the complete sexual cycle can occur in the absence of PfeIK1, resulting in formation of oocysts and sporozoites (Table 1). Infection rates and median numbers of oocysts per infected mosquito are low relative to what is routinely observed in transmission experiments with the wild-type clone 3D7. However, this is to be expected from parasites that have been kept in continuous culture for a long period of time; in the present case it had taken ~7 months in culture to obtain knockout clones suitable for mosquito infection experiments. Circumstantial evidence that low infection levels are not a direct consequence of pfeik1 disruption is provided by the observation that our control for these experiments (mock-transfected 3D7 that had been cultured for the same duration, in parallel to the pfeik1^- parasites), had completely lost the ability to produce gametocytes and therefore infect mosquitoes. Importantly, to verify that the parasites infecting the mosquitoes had not reverted to the wild-type genotype, midguts from infected mosquitoes were collected 10 days post-feeding, from which total DNA was extracted and used in nested PCR experiments. The wild-type locus could be amplified from mosquitoes infected with wild-type 3D7 parasites, but not from those infected with pfeik1^- C8 parasites (Figure 7, lower panel, lanes 1, 3, 5). Conversely, the amplicon diagnostic of the 3' boundary of the integrated plasmid could only be amplified from midguts of pfeik1^- C8-infected mosquitoes, but not from mosquitoes infected with wild-type parasites (lanes 2, 4, 8).

On the basis of the similarities between PfeIK1 and GCN2, we hypothesized that PfeIK1 is involved in modulating the response to amino acid starvation depicted in Figure 1B. That this is indeed the case was demonstrated through a reverse genetics approach: parasites lacking PfeIK1 do not phosphorylate eIF2α in response to amino-acid depletion (Figure 6). Future work will determine the impact of PfeIK1 activation on both the rate of translation and the possible selection of specific messages that are translated under stress conditions. Overall, the data presented here suggest that eIF2α phosphorylation in response to amino-acid starvation is not essential to parasite survival during the erythrocytic asexual cycle (at least in an in vitro cultivation context), or for completion of sporogony.

Commitment to gametocytogenesis has been proposed to be linked to stress response, and eIF2α might possibly be involved in this process. At first sight, the data presented here suggest that PfeIK1 does not regulate gametocytogenesis, since pfeik1^- parasite are able to undergo sexual development. However, caution must be exercised, as compensatory mechanisms can be at play in knock-out parasites. Indeed, in a similar situation concerning another protein kinase family, it was observed that disruption of the gene encoding one of the two P. falciparum mitogen-activated protein kinases (MAPKs), pfmap-1, does not cause any detectable phenotype, but that pfmap-1^- parasites overexpress the second parasite MAPK, Pfmap-2 40. A similar compensation mechanism may operate between the three PfeIKs represented in the parasite kinome (Figure 2A). Even though compensatory mechanisms to permit sexual differentiation are presumably less likely to occur than those allowing the survival of asexual parasites (because of the absence of a true selection pressure), it cannot be formally excluded that PfeIK1 plays a role in gametocytogenesis in a wild-type parasite background. Investigating this possibility will require inducible and/or multiple knock-outs and the availability of mono-specific antibodies to monitor the levels of each PfeIK in parasites lacking one of them.