Several hundred parasite proteins predicted to be exported to the erythrocyte [8]–[10] presumably underlie the molecular basis of erythrocyte remodeling in order to make this host cell a suitable environment for intracellular parasite growth. Most are ascribed as ‘hypothetical’ proteins, and since genetic manipulation of this parasite remains limited, their role in infection is poorly understood. Nonetheless, non-essential proteins involved in antigen export to the erythrocyte as well as stabilization/destabilization of the erythrocyte cytoskeleton [7],[11] have been increasingly amenable to study. Prevalent molecular and genetic analyses have thus assessed their contribution to antigen export and/or function at the host membrane [12],[13]. In contrast very little is understood at the molecular/genetic level about the biogenesis of the TVN since genes underlying it are expected to be essential for infection and cannot be knocked out.

Of the 81 genes that changed with PPMP treatment, 39 were up-regulated and 42 down-regulated. We then limited the list by focusing on gene products that met two criteria: those that contained a host-targeting signal and were conserved between human and rodent malaria parasites (Figure S1, [8],[15]). We reasoned that these proteins were likely to reflect essential functions of parasite remodeling preserved throughout the genus. An intersection of these genes with those that show PPMP-induced changes in transcriptional profiles relative to the trophozoite stage yielded two conserved genes (PFD0495c and PFC0435w; Figure 1D) of which one (PFD0495c) was up-regulated (double asterisk Figure 1Di). A biological replicate of the array experiment confirmed that these two gene transcripts were consistently modulated by PPMP treatment (Figure 1Dii). Nine other genes from the Hiller secretome and eight others from the Sergeant exportome were unchanged. In van Ooij et al. [15], we show that PFC0435w is a TVN junction protein, suggesting that our microarray data may be predictive of TVN function. To further investigate this we examined PFD0495c. It is one of 31 upregulated genes in PPMP-treated parasites whose change in expression cannot be due to developmental delay because then the transcript level would be lower than the mock-treated control (see Figure 1Eii). It is important to note that PFD0495c was recognized as a conserved protein (Figure S2) by two, independent secretome predictions, and was confirmed by RT-PCR to be specifically up-regulated by PPMP (Figure 1E). To the best of our knowledge this is the first study that has used a microarray readout from perturbation of malaria parasites to identify a candidate gene and follow up on its validation.

To establish export of PFD0495c to the erythrocyte we utilized piggyBac (a type II transposon element from the lepidopteran Trichoplusia ni, that specifically excises and integrates at TTAA target sites) to randomly insert a tagged copy of pfd0495c-gfp in the genome [15],[16] (also shown for reference in Figure 2A). Integrants detected after 11 days of drug selection were cloned. Clone 1, which was selected for further characterization, showed a single site of insertion between PFL1425w and PFL1430c (Figure 2B). This insertion site is not expected to influence export. Further clone 1 showed no significant defect on in vitro parasite growth compared to either the uncloned population or parent 3D7 parasites (Figure 2C, Figure S3, and Protocol S1).

Examination of pfd0495c-gfp parasites by fluorescence microscopy revealed fluorescence at the erythrocyte membrane (Figure 2Aiii and Aiv, arrow) and in a few intraerythrocytic structures (arrowhead) indicating that the protein was exported. However a high level of fluorescence was also detected in the parasite. To investigate whether this was a consequence of transgene expression, we used antibodies to the C-terminus of PFD0495c to examine the distribution of the native protein. As shown in Figure 3A, the antibodies recognized a single band of ~130 kDa in infected (the predicted size of the protein is ~100 kDa, but its apparent decrease in mobility could be due to the fact that it is transmembrane protein) but not uninfected erythrocytes. Indirect immunofluorescence microscopy revealed a high level of fluorescence associated with largely vesicular elements (Figure 3Bi, small arrowhead) localized in the intraerythrocytic space. A few tubular elements were also detected in the periphery of the erythrocyte (Figure 3Bi, arrow). A significant amount of the green fluorescence exported to the erythrocyte was distinct from the Maurer's clefts, although some was located adjacent to clefts (Figure 3Bii and Biii). Finally a significant amount of fluorescence was detected within the parasite. The precise significance of high levels of protein localization in both the parasite as well as the erythrocyte is unknown. However, the major consequence of transgene expression appears to be displacement of the tagged protein from the intraerythrocytic vesicles to the erythrocyte membrane, suggesting it may influence the dynamics of protein distribution between intraerythrocytic compartments and the erythrocyte membrane.

Transcription of pfd0495c-gfp is driven by the cam promoter, which is largely constitutive with peak transcription at the trophozoite and schizont stages [17],[18]. In contrast endogenous pfd0495c shows peak transcription at the schizont and early ring stages, although transcriptional activity remains detectable in late rings and trophozoites [17],[18] albeit at lower levels. We examined protein levels of PFD0495c in both 3D7 wildtype and transgenic parasites to determine how these lines differed in protein expression and their response to PPMP treatment. Although antibodies developed to the C-terminus of PFD0495c were clearly specific, they were not of sufficient titer to detect varying levels of protein throughout the asexual life cycle (not shown). We therefore developed high titer antibodies to the N-terminal repeat region of PFD0495c which also recognized one major protein band of the expected size of ~130 kDa in 3D7 late stage schizonts (Figure 3C). Notably, this band is not detected in ring stage parasites. In the transgenic line antibodies to GFP and PFD0495c detect a protein of ~180 kDa, which presumably corresponds to the fusion protein, in both late trophozoites and schizonts (Figure 3D and E). This is consistent with the major timing of transcript production driven by the cam promoter [17],[18]. Endogenous PFD0495c is also detected in the transgenic line at the schizont stage (Figure 3E). These data suggest expression of endogenous PFD0495c remains the same in transgenic and wild type parasites. The former express significant levels of fusion protein earlier (in trophozoite forms) than the endogenous protein. However, absolute levels of the fusion protein detected are small compared to amounts of endogenous protein. Further PPMP treatment appears to decrease protein expression in wild type and transgenic parasites. Hence, the modest three-fold increase in transcription from PPMP treatment (Figure 1D) does not result in a detectable increase in protein. One explanation is that pfd0495c is also post-transcriptionally regulated.

The lack of a detectable protein product for endogenous pfd0495c in ring stages (Figure 3C) even while these stages are known to be transcriptionally active (Figure S4A, B) was surprising. It is possible that either the RNA is not efficiently translated or there is high turnover of protein resulting in low levels of endogenous PFD0495c in both wild type and transgenic parasites at ring stages (Figure S4C). Hence the transcriptional/translational profile of PFD0495c is clearly distinct from a gene like RESA that is transcribed and synthesized in schizonts and carried over by daughter merozoites into the newly formed ring. The PFD0495c transcriptional profile is also distinct from MSP1, another schizont stage merozoite protein internalized upon invasion. Rather it closely mimics pfhrpii, a bonafide ring stage transcript, suggesting that PFD0495c is indeed transcribed (but not efficiently translated) in rings. How this transcription is linked to PPMP responsiveness is not understood. One explanation is that there are low levels of endogenous protein in 3D7 wild type parasites at almost all stages (visible upon longer exposures of the blots in Figure S4C).

Western blots in Figure 3 revealed that transgenic parasites express the fusion protein a little earlier (by the trophozoite stage) in the life cycle. Further transgenic but not wild type parasites treated with 5 µM PPMP for 24 h, display two pfd0495c products (transgene and endogenous) in schizont stages. Since PFD0495c associates with vesicular structures within the erythrocyte cytoplasm and its transcript is upregulated when TVN development is blocked by PPMP, we examined the TVN of transgenic parasites expressing pfd0495c-gfp and compared it to 3D7. We found that while 3D7 displayed characteristic tubules (Figure 4Ai small arrows), pfd0495c-gfp expressing parasites contained large membrane loops (Figure 4Aii), suggesting the transgene may influence TVN organization and/or function. The TVN imports lipids and raft proteins from the infected erythrocyte surface, a function that can be can be blocked by preventing TVN-tubule development with PPMP treatment [3]. As shown (Figure 4Bi) infected erythrocytes expressing the non-specific transgene pfhrpii-gfp (or 3D7 parasites; not shown) internalize a membrane-impermeable endocytic lipid marker FM4-64 [19] via TVN tubules, and this import can be blocked by short-term (30 min) treatment with PPMP (Figure 4Bii). However, clone 1 expressing a second copy of pfd0495c continues to internalize FM4-64 despite treatment with PPMP (Figure 4Biv vs. Bii). After long term (24h) treatment with PPMP labeling of all intraerythrocytic membranes with a membrane-permeable lipid, BODIPY-TR-ceramide, revealed elevated accumulation of numerous membrane loops in the erythrocyte cytoplasm in clone 1-infected cells, compared to residual intraerythrocytic structures seen in their 3D7 counterparts (Figure 4Ci versus Cii and see Figure S5). This difference in TVN organization was not due to changes in parasite sphingomyelin synthase activity inhibited by PPMP (Figure S3) or transfection per se (Figure 4Bi–ii; [20]). In addition, the effects of PPMP on parasite growth in the parent line and multiple transgenic clones were completely reversible upon washing out drug (Figure S3). This suggests that membrane accumulation seen in Figure 4Cii is not due to non-specific degeneration of intraerythrocytic structures as a consequence of parasite death but due to the presence of a second copy of pfd0495c in the genome. Together these data support the idea that PFD0495c changes lipid transport at the infected erythrocyte membrane. This lipid can be delivered to the parasite independent of sphingomyelin synthase activity and modulates tubular assembly in the network (Figure 4Biv, 4Cii). However, transgenic parasites appear to be less sensitive to the effects of PPMP compared to 3D7. Although they are decreased in size compared to untreated counterparts, they are slightly larger than treated 3D7 and consistently show two nuclei rather than just one nucleus seen in treated 3D7 (Figure 4Cii versus 4Ci). This may be explained by the fact that transgenic parasites express higher levels of PFD0495c protein in presence of PPMP.

Next we were interested in determining whether PFD0495c is essential for TVN development. It is one of only 11 conserved, exported proteins encoded by syntenic genes in Plasmodium genomes. We attempted to delete nine of the 11 genes in P. berghei but were unsuccessful [15], suggesting that they may be essential genes. Thus independent evidence that PFD0495c contributed to the TVN formation/function would likely not be obtained by undertaking knock out experiments. We therefore chose an alternate approach of loading cargo into resealed erythrocytes to monitor for effects on parasite invasion or growth. The principle is that by loading dominant-negative recombinant fragments of the protein of interest into ghosted erythrocytes and infecting, endogenous protein interactions will be disrupted, which will have effects on parasite fitness. Only proteins that function within the erythrocyte cytosol may be considered, and the domains that function within the erythrocyte cytosol must be identified. This approach has been employed to identify known cytoplasmic determinants of both host and parasite origin needed for parasite invasion [7],[21] and has recently been optimized to support parasite invasion and growth to the same degree as normal erythrocytes [21]. To extend it to investigation of a protein of unknown function, we needed to identify which domain of the protein was exposed to the erythrocyte cytoplasm. As shown in Figure 5, selective permeabilization of the erythrocyte membrane by tetanolysin, which leaves vacuolar and intraerythrocytic structures intact, suggested that the C-terminus of PFD0495c is cytoplasmically oriented in the erythrocyte of a late trophozoite/schizont-infected cell (Figure 5A and B).

To assess the functional importance of this domain during infection, we introduced 50 µM of a recombinant form of the C-terminal region of PFD0495c fused to glutathione S-transferase (GST-PFD0495cCterm; Figure S6) into the cytoplasm of resealed erythrocyte ghosts [21] (Figure 5Ci). Introduction of GST-PFD0495cCterm had no significant effect on invasion (measured as ring formation), but was clearly inhibitory to trophozoite growth. GST alone or a recombinant fusion of a parasite protein domain (PfSBPCterm) known to be exposed to the erythrocyte cytoplasm but not required for infection [12] had no effect (Figure 5 Cii). In addition, an N-terminal fragment (GST-Nterm) that is thought to be lumenal has no effect on parasite growth (Figure 5 Cii), which further validates the topology. High micromolar concentrations of recombinant fragments used in these experiments have been used in other studies at concentrations of peptides and protein domains (40–80 µM; [7],[21]) that inhibit functions of known host and parasite proteins in resealed ghosts. Further the cytoplasmic C-terminus of both SBP and PFD0495c are highly polar, a feature which is also shared by predicted cytoplasmic regions of many intraerythrocytic parasite proteins [22]. Thus inhibition of parasite growth seen by GST-PFD0495cCterm appears to be a sequence-specific effect, likely due to its interference with the actions of the endogenous protein and not due to non-specific effects of charge.

The finding that microarray outputs of malaria parasites may be predictive for function is somewhat contrary to the idea that has emerged from prior work that the parasite is hard-wired and does not respond to environmental cues [25]. One reason for the discrepancy may be that most studies investigate transcriptional change in response to drugs that kill the parasite. The secondary effects of death may confound analysis by generating noise that is difficult to filter out. In contrast PPMP treatments for periods used in these studies (24 h) arrest parasites at the early trophozoite stage but do not kill them. Additionally, because secretome gene products are at the host-pathogen interface and function between the intracellular parasite and the blood, they may be more sensitive to environmental cues. The fact that PFD0495c protein levels did not increase concomitantly with PPMP up-regulation of transcription may be due to several reasons. One explanation for this discrepancy is that the increase in transcript levels is too modest to be able to detect an increase in protein. A second possibility is that the high molecular weight oligomeric forms inhibit detection of protein at the correct molecular weight. A third alternative is that pfd0495c messenger RNA is degraded faster than protein can be made. In fact, transcription of EVP-1 is detected throughout the lifecycle, dipping during the late trophozoite stage and peaking at late schizogony (Figure 1Eii; [17],[18]). Shock and colleagues recently undertook global analysis of steady-state mRNA levels and determined that the half-life of a transcript is shortest in ring stages and longest in schizont stages [26]. The half-life of EVP-1 is 7 min in rings and 92 min in schizonts [26]. Perhaps a longer half-life is required for efficient translation of evp-1 message. Lastly, higher protein turn-over could offer another explanation. Since Westerns measure steady-state levels, it is possible that more protein is synthesized but also rapidly degraded.

Notably, although pfd0495c is up-regulated with PPMP treatment, schizont stage genes, such as MSP-1, are not. This suggests that our treated parasite population is well synchronized and not contaminated with parasites that are beyond the PPMP block. In fact, MSP-1 is down-regulated with PPMP treatment indicating that the parasites do not progress through schizogony. Although PPMP-induced changes are relatively modest, they were predictive for TVN function of our candidate genes. The use of microarrays as predictive tools is enormously beneficial in human malaria parasites that are difficult to manipulate genetically.

Our studies provide a road map to rapidly move from in silico predictions to initial annotation of these proteins in infection (Figure S7). piggyBac provides a rapid system to produce transgenic parasites with which to examine changes in intraerythrocytic transport functions in antigen export and lipid/nutrient import. Producing these parasites with such speed enables detailed characterization of sites of integration as well as other characteristics to ensure that they are suitable for analysis of a given transport phenotype. Finally for genes likely to be essential, our study provides an alternative approach to assessing functional importance of their gene products for invasion or intracellular parasite growth. This method relies on the ability to disrupt endogenous protein interactions by loading dominant-negative recombinant forms into resealed erythrocyte ghosts. The use of small peptide domains (of 30–90 amino acids) in GST fusions circumvents difficulties in expressing recombinant forms of P. falciparum proteins without codon optimization and is consistent with published data that small protein domains of P. falciparum are efficiently expressed as soluble protein in E. coli [27]. Larger, more insoluble domains can be truncated into smaller domains to identify specific inhibitory portions. Truncations in conjunction with secondary structure prediction and production of overlapping fragments can assist in production of optimal fragments (Murphy and Haldar, unpublished). In principle this strategy may be applicable to over 50% of parasite genes involved in pathogen-host interactions where over half have unknown functions [28].

Our studies of EVP1 suggest that its associated vesicles are distinct from Maurer's clefts known to function in protein export to the erythrocyte. The distribution of EVP1 is also distinct from PVM proteins such as PfEXP1 and ETRAMPs that are seen in membrane extensions off the PVM. In this context the intraerythrocytic compartment to which EVP1 localizes may define a hitherto poorly defined membrane intermediate in the infected erythrocyte. Our initial localization studies suggest the bulk of these vesicles are not at the erythrocyte periphery, although the gene has an elevated rate of polymorphisms [29]–[31], suggesting it is under host pressure. Further expression of a second copy of the gene can change the dynamics of localization, suggesting function in a novel endovesiculation pathway or transbilayer lipid movement at the infected erythrocyte membrane. We speculate that the physiological substrate of EVP-1 could be a variety of lipids, including sphingolipids, although a definitive substrate remains to be identified. Nevertheless changing lipid import properties at the infected erythrocyte membrane does not circumvent the block induced by PPMP (although the parasites are arrested at a state of two rather than one nucleus), suggesting additional factors also contribute to proper TVN and membrane import. We can begin to piece together the sequence of events that build and maintain the TVN. Sphingomyelin synthesis is a critical early event. If its activity is inhibited early in the asexual cycle, then the TVN is not properly assembled and nutrients are not imported. Several pieces of evidence support the idea that EVP1 is required later in the cycle. The predominant form of the protein is detected in later stages and not in ring stages. The resealed ghosted erythrocyte experiments demonstrate that the dominant-negative GST recombinant form of the C-terminus inhibits parasite maturation not invasion. Finally the transgenic line retains sensitivity to PPMP but has altered TVN properties. Since the TVN is assembled from the late ring stage to early schizogony, the data unexpectedly reveal that vesicular import from the erythrocyte membrane at its later stages of assembly may be important for the maintenance of this network. Hence the TVN remains a dynamic structure even though its tubules appear as relatively immobile structures in the erythrocyte cytoplasm.

Protocol S1

Figure S1

Figure S2

Figure S3

Figure S4

Figure S5

Figure S6

Figure S7