Main

The life cycle of Plasmodium is extraordinarily complex, requiring specialized protein expression for life in both invertebrate and vertebrate host environments, for intracellular and extracellular survival, for invasion of multiple cell types, and for evasion of host immune responses. Interventional strategies including anti-malarial vaccines and drugs will be most effective if targeted at specific parasite life stages and/or specific proteins expressed at these stages. The genomes of P. falciparum1 and P. yoelii yoelii2 are now completed and offer the promise of identifying new and effective drug and vaccine targets.

Functional genomics has fundamentally changed the traditional gene-by-gene approach of the pre-genomic era by capitalizing on the success of genome sequencing efforts. DNA microarrays have been successfully used to study differential gene expression in the abundant blood stages of the Plasmodium parasite3,4. However, transcriptional analysis by DNA microarrays generally requires microgram quantities of RNA and has been restricted to stages that can be cultivated in vitro, limiting current large-scale gene expression analyses to the blood stages of P. falciparum. As several key stages of the parasite life cycle, in particular the pre-erythrocytic stages, are not readily accessible to study, and as differential gene expression is in fact a surrogate for protein expression, global proteomic analyses offer a unique means of determining not only protein expression, but also subcellular localization and post-translational modifications.

We report here a comprehensive view of the protein complements isolated from sporozoites (the infectious form injected by the mosquito), merozoites (the invasive stage of the erythrocytes), trophozoites (the form multiplying in erythrocytes), and gametocytes (sexual stages) of the human malaria parasite P. falciparum. These proteomes were analysed by multidimensional protein identification technology (MudPIT), which combines in-line, high-resolution liquid chromatography and tandem mass spectrometry5. Two levels of control were implemented to differentiate parasite from host proteins. By using combined host–parasite sequence databases and noninfected controls, 2,415 parasite proteins were confidently identified out of thousands of host proteins; that is, 46% of all gene products were detected in four stages of the Plasmodium life cycle (Supplementary Table 1).

Comparative proteomics throughout the life cycle

The sporozoite proteome appeared markedly different from the other stages (Table 1). Almost half (49%) of the sporozoite proteins were unique to this stage, which shared an average of 25% of its proteins with any other stage. On the other hand, trophozoites, merozoites and gametocytes had between 20% and 33% unique proteins, and they shared between 39% and 56% of their proteins. Consequently, only 152 proteins (6%) were common to all four stages. Those common proteins were mostly housekeeping proteins such as ribosomal proteins, transcription factors, histones and cytoskeletal proteins (Supplementary Table 1). Proteins were sorted into main functional classes based on the Munich Information Centre for Protein Sequences (MIPS) catalogue6, with some adaptations for classes specific to the parasite, such as cell surface and apical organelle proteins (Fig. 1). When considering the annotated proteins in the database, some marked differences appeared between sporozoites and blood stages (Fig. 1). Although great care was taken to ensure that the results reflect the state of the parasite in the host, a portion of the data set may reflect the parasite's response to different purification treatments. However, the stage-specific detection of known protein markers at each stage established the relevance of our data set.

Table 1 Comparative summary of the protein lists for each stage
Figure 1: Functional profiles of expressed proteins.
figure 1

Proteins identified in each stage are plotted as a function of their broad functional classification as defined by the MIPS catalogue6. To avoid redundancy, only one class was assigned per protein. The complete protein list is given in Supplementary Table 1.

The merozoite proteome

Merozoites are released from an infected erythrocyte, and after a short period in the plasma, bind to and invade new erythrocytes. Proteins on the surface and in the apical organelles of the merozoite mediate cell recognition and invasion in an active process involving an actin-myosin motor. Four putative components of the invasion motor7, merozoite cap protein-1 (MCP1), actin, myosin A, and myosin A tail domain interacting protein (MTIP), were abundant merozoite proteins (Supplementary Table 2). Abundant merozoite surface proteins (MSPs) such as MSP1 and MSP2 are linked by a glycosylphosphatidyl (GPI) anchor to the membrane, and both have been implicated in immune evasion (reviewed in ref. 8). A second family of peripheral membrane proteins, represented by MSP3 and MSP6, was also detected (Fig. 2a), although these proteins are largely soluble proteins of the parasitophorous vacuole, which are released on schizont rupture. Other vacuolar proteins, such as the acidic basic repeat antigen (ABRA) and serine repeat antigen (SERA), were detected in the merozoite fraction, but some such as S-antigen9 were not (Supplementary Table 2). Notably, MSP8 and a related MSP8-like protein were only identified in sporozoites (Fig. 2a). Some MSPs are diverse in sequence and may be extensively modified by proteolysis; these features, together with the association of a variety of peripheral and soluble proteins, provide for a complex surface architecture.

Figure 2: Expression patterns of known stage-specific proteins.
figure 2

a, Cell surface, organelle, and secreted proteins are plotted as a function of their known subcellular localization. b, stevor, var and rif polymorphic surface variants are plotted as a function of the chromosome encoding their genes. The matrices are colour-coded by sequence coverage measured in each stage (proteins not detected in a stage are represented by black squares). Locus names associated with these proteins are listed in Supplementary Table 2. Spz, sporozoite; mrz, merozoite; tpz, trophozoite; gmt, gametocyte.

Many apical organellar proteins, in the micronemes and rhoptries, have a single transmembrane domain. Among these proteins, apical membrane antigen 1 (AMA1) and MAEBL were found in both sporozoite and merozoite preparations (Fig. 2a). Erythrocyte-binding antigens (EBA), such as EBA 175 and EBA 140/BAEBL, were found only in the merozoite and trophozoite fractions. Of note, the reticulocyte-binding protein (PfRH) family (PFD0110w, MAL13P1.176, PF13_01998, PFL2520w and PFD1150c), which has similarity with the Py235 family of P. y. yoelii rhoptry proteins and the Plasmodium vivax reticulocyte-binding proteins, was not detected in the merozoite fraction. Some PfRH proteins were, however, detected in sporozoites (Fig. 2a), including RH3, which is a transcribed pseudogene in blood stages10. Components of the low molecular mass rhoptry complex, the rhoptry-associated proteins (RAP) 1, 2 and 3, were all found in merozoites. RAP1 was also detected in sporozoites. The high molecular mass rhoptry protein complex (RhopH), together with ring-infected erythrocyte surface antigen (RESA), which is a component of dense granules, is transferred intact to new erythrocytes at or after invasion and may contribute to the host cell remodelling process. RhopH1, RhopH2 (PFI1445w; Ling, I. T., et al., unpublished data) and RhopH3 were found in the merozoite proteome. RhopH1 (PFC0120w/PFC0110w) has been shown to be a member of the cyto-adherence linked asexual gene family (CLAG)11; however, the presence of CLAG9 in the merozoite fraction (Fig. 2a) suggests that CLAG9 may also be a RhopH protein, casting some doubt on the proposed role for this protein in cyto-adherence12.

The trophozoite proteome

After erythrocyte invasion the parasite modifies the host cell. The principal modifications during the initial trophozoite phase (lasting about 30 h) allow the parasite to transport molecules in and out of the cell, to prepare the surface of the red blood cell to mediate cyto-adherence, and to digest the cytoplasmic contents, particularly haemoglobin, in its food vacuole. In the next phase of schizogony (the final 18 h of the asexual development in the blood cell), nuclear division is followed by merozoite formation and release.

Knob-associated histidine-rich protein (KAHRP) and erythrocyte membrane proteins 2 and 3 (EMP2 and -3) bind to the erythrocyte cytoskeleton (Fig. 2a). Of the proteins of the parasitophorous vacuole and the tubovesicular membrane structure extending into the cytoplasm of the red blood cell, three (the skeleton-binding protein 1, and exported proteins EXP1 and EXP2) were represented by peptides (Fig. 2a); although a fourth (Sar1 homologue, small GTP-binding protein; PFD0810w) was not. It is likely that one or more of the hypothetical proteins detected only in the trophozoite sample are involved in these unusual structures.

Digestion of haemoglobin is a major parasite catabolic process13. Members of the plasmepsin family (aspartic proteinases; PF14_0075 to PF14_0078)14, falcipain family (cysteine proteinases; PF11_0161, PF11_0162 and PF11_0165)15, and falcilysin (a metallopeptidase; PF13_0322)16 implicated in this process were all clearly identified (Supplementary Table 1). Several proteases expressed in the merozoite and trophozoite fractions, and not involved in haemoglobin digestion, may be important in parasite release at the end of schizogony, invasion of the new cell, or merozoite protein processing. Possible candidates for this mechanism include cysteine proteinases of the falcipain and SERA families, or subtilisins such as SUB1 and SUB2, both located in apical organelles (Fig. 2a).

The gametocyte proteome

Stage V gametocytes are dimorphic, with a male:female ratio of 1:4. They are arrested in the cell cycle until they enter the mosquito where development is induced within minutes to form the male and female gametes. Gametocyte structure reflects these ensuing fates; that is, the female has abundant ribosomes and endoplasmic reticulum/vesicular network to re-initiate translation, whereas the male is largely devoid of ribosomes and is terminally differentiated17.

Gametocyte-specific transcription factors, RNA-binding proteins, and gametocyte-specific proteins involved in the regulation of messenger RNA processing (particularly splicing factors, RNA helicases, RNA-binding proteins, ribonucleoproteins (RNPs) and small nuclear ribonucleoprotein particles (snRNPS)) were highly represented in the gametocyte proteome (Supplementary Table 1). Transcription in the terminally differentiated gametocytes is ‘suppressed’, but the female gametocytes contain mRNAs encoding gamete/zygote/ookinete surface antigens (for example, P25/28) that are subject to post-transcriptional control; this control is released rapidly during gamete development17. Ribosomal proteins were largely represented: 82% of known small subunit (SSU) proteins and 69% of known large subunit (LSU) proteins were detected in gametocytes compared to 94% and 82%, respectively, from all stages examined (Supplementary Table 1). We suggest that this reflects the accumulation of ribosomes in the female gametocyte to accommodate for the sudden increase in protein synthesis required during gametogenesis and early zygote development.

Other protein groupings highly represented in the gametocyte were in the cell cycle/DNA processing and energy classes (Fig. 1). The former is consistent with the biological observation that the mature gametocyte is arrested in G0 of the cell cycle and will require a full complement of pre-existing cell cycle regulatory cascades to respond, within seconds, to the gametogenesis stimuli (that is, xanthurenic acid and a drop in temperature)18. Metabolic pathways of the malaria parasite may be stage-specific, with asexual blood stage parasites dependent on glycolysis and conversion of pyruvate to lactate (l-lactate dehydrogenase) for energy. In the gametocyte and sporozoite preparations, peptides from enzymes involved in the mitochondrial tricarboxylic acid (TCA) cycle and oxidative phosphorylation were identified (Table 2). This observation suggests that gametocytes have fully functional mitochondria as a pre-adaptation to life in the mosquito, as suggested by morphological and biochemical studies19 and their sensitivity to anti-malarials attacking respiration (primaquine and artimesinin-based products)17. It will be interesting to observe whether other mosquito and liver stages, which show similar drug sensitivities, express the same metabolic proteome.

Table 2 Examples on enzymes in stage-specific metabolic pathways

Cell surface proteins (Fig. 1) included most of the known surface antigens (Fig. 2a and Supplementary Table 2). However, Pfs35 and a sexual stage-specific kinase (PF13_0258) were not detected. Nevertheless the cultured gametocytes analysed in this study expressed a specific repertoire of rifin and PfEMP1 proteins (Fig. 2b and Supplementary Table 2). Together these observations suggest that the gametocyte, which is very long-lived in the red blood cell (that is, 9–12 days compared with 2 days for the pathogenic asexual parasites), expresses a limited repertoire of the highly polymorphic families of surface antigens so widely represented in the asexual parasites.

The sporozoite proteome

Sporozoites are injected by the mosquito during ingestion of a blood meal. Although, they are in the blood stream for only minutes, sporozoites probably require mechanisms to evade the host humoral immune system in order for at least a fraction of the thousands of sporozoites injected by the mosquito to survive the hostile environment in the blood and successfully invade hepatocytes.

The main class of annotated sporozoite proteins identified was cell surface and organelle proteins (Fig. 1). Sporozoites are an invasive stage and possess the apical complex machinery involved in host cell invasion. As observed in the analysis of the P. y. yoelii sporozoite transcriptome20, actin and myosin were found in the motile sporozoites (Supplementary Table 2). Many proteins associated with rhoptry, micronemes and dense granules were detected (Fig. 2a). Among the proteins found were known markers of the sporozoite stage, such as the circumsporozoite protein (CSP) and sporozoite surface protein 2 (SSP2; also known as TRAP), both present in large quantities at the sporozoite surface (Fig. 2a). Peptides derived from CTRP (circumsporozoite protein and thrombospondin-related adhesive protein (TRAP)-related protein), an ookinete cell surface protein involved in recognition and/or motility21, were detected in the sporozoite fractions (Supplementary Table 1).

Most surprisingly, peptides derived from multiple var (coding for PfEMP1) and rif genes were identified in the sporozoite samples. PfEMP1 and rifins are coded for by large multigene families (var and rif)22,23 and are present on the surface of the infected red blood cell. No peptides derived from rif genes were identified in the trophozoite sample, whereas sporozoites expressed 21 different rifins and 25 PfEMP1 isoforms (Fig. 2b); that is, a total of 14% of the rif genes and 33% of the var genes encoded by the genome. Furthermore, very little overlap was observed between stages: only ten PfEMP1 and two rifin isoforms expressed in sporozoites were found in other stages. Whereas in the blood stream the asexual stage parasites undergo asexual multiplication and therefore have an opportunity to undergo antigenic ‘switching’ of the variant antigen genes, the non-replicative sporozoites may not have this opportunity. Expressing such a polymorphic array of var (PfEMP1) and rif genes could be part of a sporozoite survival mechanism.

Chromosomal clusters encoding co-expressed proteins

The distinct proteomes of each stage of the Plasmodium life cycle suggested that there is a highly coordinated expression of Plasmodium genes involved in common processes. Co-expression groups are a widespread phenomenon in eukaryotes, where mRNA array analyses have been used to establish gene expression profiles. Analysis of co-regulated gene groups facilitates both searching for regulatory motifs common to co-regulated genes, and predicting protein function on the basis of the ‘guilt by association’ model. Furthermore, mRNA analyses in Saccharomyces cerevisiae24 and Homo sapiens25,26 have demonstrated that co-regulated genes do not map to random locations in the genome but are in fact frequently organized into gene clusters on a chromosome. Gene clustering in Plasmodium species has been demonstrated. Ordered arrays of genes involved in virulence and antigenic variation (for example, var, vir and rif genes) are located in the subtelomeric regions of the chromosomes27,28.

To determine whether gene clustering exists along the entire P. falciparum genome, genes whose protein products were detected in our analysis were mapped onto all 14 chromosomes in a stage-dependent manner (Fig. 3a). The 2,415 proteins identified represented an average of 45% of the open reading frames (ORFs) predicted per chromosome. The number of protein hits by chromosome was similar for all stages: sporozoite, merozoite, trophozoite and gametocyte protein lists constituting 19.7%, 15.8%, 19.5% and 21.6% of the predicted ORFs per chromosomes, respectively. Groups of three or more consecutive loci whose protein products were detected in a particular stage were defined as chromosomal clusters encoding co-expressed proteins (Fig. 3b). On the basis of this definition a total of 98 clusters containing 3 loci, 32 clusters containing 4 loci, 5 clusters containing 5 loci, and 3 clusters containing 6 loci were identified (Supplementary Table 3). For each chromosome, the frequency of finding clusters encoding co-expressed proteins containing 3–6 adjacent loci markedly exceeded the probability of finding such clusters by chance (see the footnote of Supplementary Table 3 for details on the probability calculation). Therefore, chromosomal clusters encoding co-expressed proteins were prevalent in the P. falciparum genome.

Figure 3: Distribution of expressed proteins by chromosome.
figure 3

a, For each stage, genes whose products were detected (coloured vertical bars) are plotted in the order they appear on their chromosome (grey boxes). b, Groups of at least three consecutive expressed genes are defined as chromosomal clusters of co-expressed proteins. Examples of such clusters, circled in b, are specified in Table 3 and the complete description of the 138 clusters can be found in Supplementary Table 3.

Functionally related genes have been shown to cluster in the S. cerevisiae24 and human genomes26. This phenomenon also occurs in P. falciparum. A total of 138 clusters encoding co-expressed proteins were identified and 67 of them (49%) contained at least two loci that have been functionally annotated. Of these 67 clusters, 30 contained at least two loci whose annotation clearly indicates that the proteins are functionally related. For example, clusters on chromosomes 3, 5 and 10 contained ribosomal proteins, proteins involved in protein modification, and proteins involved in nucleotide metabolism, respectively (Table 3). Chromosome 14 contained a cluster of four aspartic proteases co-expressed in all of the blood stages (Table 3). This cluster was not detected in sporozoites, where no haemoglobin degradation is expected to occur. Interestingly, whereas the falcipain gene cluster on chromosome 11 appeared in our analysis as a cluster of co-expressed proteins (Supplementary Table 3), the SERA gene cluster on chromosome 2, coding for proteins that share a papain-like sequence motif29, did not. Of the ten sporozoite-specific clusters, five involved var and rif genes, such as the rif cluster located in the subtelomeric domain of chromosome 14 (Table 3). On the basis of their presence in clusters encoding co-expressed proteins, we were able to suggest functional roles for 24 proteins annotated as hypothetical in the P. falciparum genome (Supplementary Table 3). For example, a gametocyte-specific cluster on chromosome 13 encoded two transmission-blocking antigens (Pfs48/45 and Pfs47) and a hypothetical protein, PF13_0246, which might be a gametocyte surface protein. Two clusters on chromosomes 2 and 11 were highly specific to the trophozoite stage (Table 3). Each of these clusters contained well-known secreted and surface proteins, namely KAHRP, PfEMP3, antigen 332, and RESA, all of which have been implicated in knob formation. The highly coordinated expression of these genes makes the three hypothetical proteins listed in these trophozoite-specific gene clusters possible candidates for involvement in cyto-adherence.

Table 3 Examples of chromosomal gene clusters encoding co-expressed proteins

Discussion

Although sample handling is a principal consideration when studying pathogens, the expression of large numbers of previously identified proteins was consistent with their published expression profiles, validating our data set as a meaningful sampling of each stage's proteome. This is a particularly important aspect of our analysis as 65% of the 5,276 genes encoded by the P. falciparum genome are annotated as hypothetical1, and of the 2,415 expressed proteins we identified, 51% are hypothetical proteins (Supplementary Table 1). Our results confirmed that these hypothetical ORFs predicted by gene modelling algorithms were indeed coding regions. Furthermore, from all four stages analysed, we identified 439 proteins predicted to have at least one transmembrane segment or a GPI addition signal (18% of the data set) and 304 soluble proteins with a signal sequence; that is, potentially secreted or located to organelles. Well over half of the secreted proteins and integral membrane proteins detected were annotated as hypothetical (Supplementary Table 4). The obvious interest in this class of proteins is that, with no homology to known proteins, they represent potential Plasmodium-specific proteins and may provide targets for new drug and vaccine development.

Our comprehensive large-scale analysis of protein expression showed that most surface proteins are more widely expressed than initially thought. In particular, the var and rif genes, which were thought to be involved in immune evasion only in the blood stage, have now been shown to be expressed in apparently large and varied numbers at the sporozoite stage. These surface proteins might be involved in general interaction processes with host cells and/or immune evasion. An alternative hypothesis is that stage-specific regulation is not as exact as previously thought.

One mechanism of protein expression control that contributes to stage specificity in P. falciparum arises from the chromosomal clustering of genes encoding co-expressed proteins. The clusters described in this study demonstrate a widespread high order of chromosomal organization in P. falciparum and probably correspond to regions of open chromatin allowing for co-regulated gene expression. The high (A + T) content of the P. falciparum genome makes the identification of regulatory sequences such as promoters and enhancers challenging31,32. Focusing analyses on stage-specific and multi-stage clusters will facilitate finding stage-specific and general cis-acting sequences in the Plasmodium genome and will help decipher gene expression regulation during the parasite life cycle.

The malaria parasite is a complex multi-stage organism, which has co-evolved in mosquitoes and vertebrates for millions of years. Designing drugs or vaccines that substantially and persistently interrupt the life cycle of this complex parasite will require a comprehensive understanding of its biology. The P. falciparum genome sequence and comparative proteomics approaches may initiate new strategies for controlling the devastating disease caused by this parasite.

Methods

Parasite material

Plasmodium falciparum clone 3D7 (Oxford) was used throughout. Sporozoites were initially isolated from the salivary glands of Anopheles stephansi mosquitoes, 14 days after infection, by centrifugation in a Renograffin 60 gradient, as described33. Four sporozoite samples were used as is. A fifth sample underwent an additional purification step on Dynabeads M-450 Epoxy coupled to NFS1 (an anti-P. falciparum CS protein monoclonal antibody)34 according to the manufacturer's instructions (Dynal). Trophozoite-infected erythrocytes from synchronized cultures were purified on 70% Percoll-alanine30, and the trophozoites released from the erythrocytes35. Of the of 260 parasitized erythrocytes counted by Giemsa-stained thin-blood film, 100% were identified as trophozoites. Merozoites were prepared essentially as described in ref. 36, using highly synchronized schizonts and purifying the merozoites by passage through membrane filters. Starting with synchronized asexual parasites grown in suspension culture as described37,38, gametocytes were prepared by daily media changes of static cultures at 37 °C. When there were very few mature asexual stages present, gametocyte-infected erythrocytes were collected from the 52.5%/45% and 45%/30% interfaces of a Percoll gradient39. The gametocytes consisted mostly of stage IV and V parasites with minor contamination (<3%) from mixed asexual stage parasites. Finally, cellular debris from the upper bodies of parasite-free A. stephansi and non-infected human erythrocytes were used as controls for sporozoites and blood-stage parasites, respectively. Every effort was made to minimize enzymatic activity and protein degradation during sampling, and the subsequent isolation of the parasites; however, we cannot exclude that some of the differences in protein profiles that we observe between the different life-cycle stages may be a consequence of the sample-handling procedures.

Cell lysis

Five sporozoite, four merozoite, four trophozoite and three gametocyte preparations were lysed, digested and analysed independently. Cell pellets were first diluted ten times in 100 mM Tris-HCl pH 8.5, and incubated in ice for 1 h. After centrifugation at 18,000 g for 30 min, supernatants were set aside and microsomal membrane pellets were washed in 0.1 M sodium carbonate, pH 11.6. Soluble and insoluble protein fractions were separated by centrifugation at 18,000 g for 30 min. Supernatants obtained from both centrifugation steps were either combined (sporozoites, trophozoites and merozoites) or digested and analysed independently (gametocytes).

Peptide generation and analysis

The method follows that of Washburn et al.5, with the exception that Tris(2-carboxyethyl)phosphine hydrochloride (TCEP-HCl; Pierce) was used to reduce urea-denatured proteins. Peptide mixtures were analysed through MudPIT as described5.

Protein sequence databases

The P. falciparum database contained 5,283 protein sequences. Spectra resulting from contaminant mosquito and erythrocyte peptides had to be taken into account in the sporozoite and blood-stage samples, respectively. Tandem mass spectrometry (MS/MS) data sets from blood stages were therefore searched against a database containing both P. falciparum protein sequences and 24,006 ORFs from the human, mouse and rat RefSeq NCBI databases. At the date of the searches, the Anopheles gambiae genome was not available. The NCBI database contained 922 Anopheles and 313 Aedes proteins, which were combined to the 14,335 ORFs of the NCBI Drosophila melanogaster40 database to create a control diptera database. Finally, these databases were complemented with a set of 172 known protein contaminants, such as proteases, bovine serum albumin and human keratins.

MS/MS data set analysis

The SEQUEST algorithm was used to match MS/MS spectra to peptides in the sequence databases41. To account for carboxyamidomethylation, MS/MS data sets were searched with a relative molecular mass of 57,000 (Mr, 57K) added to the average molecular mass of cysteines. Peptide hits were filtered and sorted with DTASelect42. Spectra/peptide matches were only retained if they were at least half-tryptic (Lys or Arg at either end of the identified peptide) and with minimum cross-correlation scores (XCorr) of 1.8 for +1, 2.5 for +2, and 3.5 for +3 spectra and DeltaCn (top match's XCorr minus the second-best match's XCorr divided by the top match's XCorr) of 0.08. Peptide hits were deemed unambiguous only if they were not found in non-infected controls and were uniquely assigned to parasite proteins by searching against combined parasite–host databases. Finally, for low coverage loci, peptide/spectrum matches were visually assessed on two main criteria: any given MS/MS spectrum had to be clearly above the baseline noise, and both b and y ion series had to show continuity. The Contrast tool42 was used to compare and merge protein lists from replicate sample runs and to compare the proteomes established for the four stages.