Digestive-vacuole genesis and endocytic processes in the early intraerythrocytic stages of Plasmodium falciparum

Plasmodium falciparum causes ~1 million malaria-related deaths annually (Snow et al., 2005). Disease pathology is associated with the parasite's life cycle within a parasitophorous vacuole (PV) inside the red blood cells (RBCs) of its human host. As it develops, the intraerythrocytic parasite degrades ~75% of the host haemoglobin (Francis et al., 1997; Loria et al., 1999; Rosenthal and Meshnick, 1996). This provides a source of amino acids and creates space for growth, as well as generating osmolytes that prevent premature host cell lysis (Lew et al., 2003).

Early studies of the feeding process in P. falciparum-infected RBCs indicated that uptake of host cytoplasm is most active in trophozoites and involves morphologically distinct endocytic structures, termed cytostomes, at the parasite surface (Aikawa et al., 1966). The resulting endocytic invaginations are surrounded by both the parasite plasma membrane (PPM) and the PV membrane (PVM) (Slomianny, 1990). The double-membrane-bound vesicles that bud from the cytostome were proposed to migrate to and fuse with an acidic digestive vacuole (DV). Single-membrane-bound vesicles were assumed to be delivered into the DV lumen, followed by release and degradation of the haemoglobin (Yayon et al., 1984).

Recent studies have re-examined the process of haemoglobin uptake in infected RBCs but have led to somewhat conflicting conclusions. A study using serial thin-section electron microscopy (EM) (Elliott et al., 2008) described four distinct pathways for haemoglobin uptake by P. falciparum. These authors proposed that ring-stage parasites fold around a ‘gulp’ of host cell cytoplasm as the first step in the biogenesis of the DV. They suggested that small cytostome-derived haemoglobin-containing vesicles and tubules continue the uptake of haemoglobin as the parasite matures. In more mature parasites they also described additional cytostome-independent endocytic structures called phagosomes (Elliott et al., 2008).

Another study examined the ultrastructure of the endocytic apparatus and concluded that haemoglobin uptake occurs by a vesicle-independent process (Lazarus et al., 2008). These authors reported that cytostomal invaginations elongate to form tubes that appose the DV but remain connected to the parasite surface and open to the RBC cytoplasm. They suggested that the tubules pinch off from the parasite surface and simultaneously undergo fusion with the DV to release their contents into the DV lumen (Lazarus et al., 2008).

Another area of debate is the point at which haemoglobin digestion is initiated. Some authors have argued that haemoglobin digestion commences only after cytostomal vesicles fuse with the DV (Olliaro and Goldberg, 1995; Slomianny and Prensier, 1990). Others suggest that haemoglobin digestion is initiated en route to the DV, and that the DV is merely a dumping site for haemozoin crystals (Hempelmann et al., 2003).

In this work we have re-examined DV genesis and estimated the pH of the DV and the endocytic compartments by following the uptake of pH-sensitive probes. We have used live-cell imaging and photobleaching to examine the dynamics and connectivity of different compartments. We also used electron tomography to generate high-resolution images of sections of parasite-infected RBCs. We have combined this with serial sectioning to obtain tomographic reconstructions of whole parasites at different stages of development. We compared the data from these two imaging modalities to provide novel insights into the feeding process.

Human RBCs that have been lysed and resealed at high haematocrit retain the ability to support the growth of P. falciparum (Dluzewski et al., 1983) and can be used to trap fluorescent markers (Goodyer et al., 1997; Krogstad et al., 1985). We used an optimised ratio of lysis buffer to RBC volume (2.5:1) to minimise the loss of RBC cytoplasm while achieving a relatively homogenous population of resealed RBCs containing the pH-sensitive probe, SNARF-1-dextran (Whitaker et al., 1991). The SNARF-1 chromophore undergoes an acid-base transition from a yellow- to a red-emitting form that can be monitored using a single excitation wavelength (Bassnett et al., 1990; Seksek and Bolard, 1996; van Erp et al., 1991). The resealed RBCs retain 20-45% of their haemoglobin (corresponding to 4-9 mM monomeric haemoglobin) and contain 0.1-0.2 mol% of the fluorescent dextran with respect to haemoglobin, corresponding to an intracellular concentration of 10-20 μM and a trapping efficiency of 30-60%. The resealed RBCs were invaded with 54±5% of the efficiency of control RBCs, in agreement with previous data (Frankland et al., 2006). Parasites developing in resealed RBCs complete the cycle and form merozoites that are capable of reinvading further resealed RBCs, although the efficiency of invasion in the second round is reduced (data not shown).

We constructed a SNARF-1 pH calibration curve by suspending resealed RBC in buffers of different pH in the presence of the ionophore, carbonyl cyanide 3-chlorophenylhydrazone (CCCP) (Allen and Kirk, 2004; Klonis et al., 2007). CCCP has previously been used to equilibrate compartments in P. falciparum-infected RBCs to external buffers of known pH (Allen and Kirk, 2004; Klonis et al., 2007). Fluorescence images were collected for resealed RBCs with different levels of trapped fluorophore on three separate occasions, and the ratio R_ry (red/yellow fluorescence intensity) was used as an indicator of pH (Fig. 1A). The relatively small standard deviation (s.d.) values indicate that any differences in SNARF-1 uptake and haemoglobin levels do not affect the R_ry values. SNARF-1-dextran in buffered solutions gave a similar calibration curve (data not shown). The decrease in R_ry values with decreasing pH is adequately described by a simple acid-base transition with an apparent pKa of 7.2, which is similar to the reported value of ~7.5 (Owen, 1992; Whitaker et al., 1991). Given its response profile, SNARF-1 can be used to distinguish between acidic and neutral compartments, but cannot provide accurate values at pH values less than 6 (R_ry<0.9).

In order to visualise the spatial variation of R_ry values (and hence pH) within a cell, we generated R_ry images in HSB colour space, where the R_ry value at each pixel is mapped to a particular colour in the hue channel and its corresponding total intensity (I_r + I_y) is placed in the brightness channel. This generated the look-up table shown in Fig. 1B, where neutral to basic compartments (R_ry values >1.2) appear blue-purple whereas compartments with pH values <6 (R_ry values<0.8) appear green. This is illustrated by the R_ry images of resealed RBC resuspended at pH 5.5 and 7.5 in the presence of CCCP (Fig. 1B). We examined the effect of CCCP treatment on SNARF-1 that has been accumulated into the DV of trophozoite-stage parasites (Fig. 1C). In untreated cells, the DV appears green, consistent with a DV pH value <6 (Fig. 1Ci), whereas in cells suspended in PBS at pH 7.5 in the presence of CCCP, the DV appears purple (Fig. 1Cii).

To examine the earliest events associated with the uptake of the RBC cytoplasm, we purified tightly synchronised (~6 hour window) parasites (D10 strain) at the schizont stage (40-48 hours) and used them to invade resealed RBC containing SNARF-1-dextran. After 20 to 24 hours (i.e. 12-24 hours after invasion) we transferred them to a microscope chamber that was gassed and maintained at 37°C (Fig. 2). The cells appear to be viable for ~10 hours, however we imaged them within 2 hours. By analysing differential interference contrast (DIC) images, we chose parasites that had no visible haemozoin (Fig. 2A-D). Based on the estimated time after invasion and the absence of visible pigment, we define these as mid-ring-stage parasites. In some infected RBCs (Fig. 2A), the parasite is observed as a dark region in the fluorescence intensity images (second column), with no evidence for endocytic compartments. The images shown are average fluorescence intensity z-section projections taken through the parasite, confirming that this is not a consequence of an internalised compartment being out of the plane of focus.

In other mid-ring-stage parasites, punctate structures containing SNARF-1-dextran were observed near the parasite periphery (Fig. 2B-D, yellow arrowheads). The SNARF-1 concentration within these compartments is comparable with that in the host cell cytoplasm, suggesting that they represent early endocytic events. They are observed as regions of altered R_ry values in the ratiometric images (Fig. 2) indicating they are acidic in nature; however, analysis of the fluorescence signal in this compartment was complicated by contributions from the host cell. To better observe these compartments, we selectively photobleached the SNARF-1 fluorescence in the host compartment by parking the laser in a region distal to the parasite (indicated by ‘x’) and illuminating with an unattenuated laser for 10 seconds. This results in the loss of fluorescence throughout this compartment (Fig. 2, middle and right panels) (Klonis et al., 2002), without affecting the fluorescence intensity of any SNARF-1 that has been internalised by the parasite. This increases the contrast of the acidic compartments and confirms that they are located within the parasite and are not in direct communication with the RBC cytoplasm. Indeed they were often revealed as multiple puncta that appear green in the post-bleach R_ry images. The average R_ry of these compartments was 0.85±0.07 (n=24; four experiments) corresponding to a pH of <6. In no examples did we observe endocytic compartments that exhibited a neutral pH. By contrast, the R_ry values in the RBC cytoplasm of uninfected and infected erythrocytes were 1.43±0.19 (n=29; four experiments) and 1.40±0.16 (n=42; four experiments), corresponding to pH values of 7.2±0.3 and 7.14±0.25, respectively. We also analysed another parasite strain (3D7) and examined parasites developing in resealed RBCs containing TMR-dextran or recombinant GFP as reporters (supplementary material Fig. S1). We observed equivalent structures with these different reporters.

In many ring-stage parasites, a large fluorophore-labelled feature (1-2 μm in diameter) appeared to lie within the parasite (Fig. 2B-D, green arrows) and was also evident in the DIC images. The fluorophore appeared to be at a similar concentration to the bulk of the RBC cytoplasm and the R_ry value (in the pre-bleach image) was similar to that of the RBC cytoplasm. The structure could correspond to a proposed macropinocytic structure call the ‘big gulp’, which was recently described (Elliott et al., 2008). Interestingly, photobleaching of the host cell fluorescence depleted the fluorescence associated with these spherical structures to the same degree (Fig. 2B-D, middle and right panels) demonstrating that this structure remained connected to the RBC cytoplasm. We analysed 100 infected RBCs before the appearance of haemozoin pigment and found punctate peripheral structures in 94% of cells and larger spherical structures in 63% of cells.

DIC images were used to identify a second population of parasites containing small dark puncta; these probably represent the first detectable haemozoin particles (Fig. 2E-F). Based on this criterion, we define these as late-ring-stage parasites. Many of the cells possessed more than one pigment-containing structure or a combination of pigmented and non-pigmented structures. These compartments often showed brighter fluorescence than the host cytoplasm, indicating a concentration event. The internalised compartments were invariably acidic with a pH <6 (R_ry=0.86±0.09; n=28; four experiments). The low pH, the accumulation of the host cytoplasmic components, and the presence of haemozoin suggest that haemoglobin digestion is initiated in several independent pre-DV compartments in these early stage parasites. Application of a laser pulse to one of the peripheral structures (Fig. 2I, yellow arrow) results in selective photobleaching of this compartment, confirming the presence of several independent compartments.

The cells shown in Fig. 2G,H contained more prominent pigment in the DIC image and probably represent the next stage of DV formation (i.e. the early trophozoite stage). The haemozoin was confined to one region of the parasite and was associated with internalised acidic structures that were considerably brighter than the surrounding host cell cytoplasm. In Fig. 2G, the endocytic compartments might be in the process of coalescing into a single site within the parasite, whereas in Fig. 2H they appear to have coalesced. Other larger features were observed within the parasites at these stages (Fig. 2E-H, green arrows), which appear to represent invaginations of the host cell cytoplasm. In all cases, photobleaching measurements demonstrate that these features remain connected to host cell cytoplasm.

We examined the distribution of SNARF-1-dextran in the endocytic compartments of mature parasitised RBCs (i.e. 32-40 hours after invasion). In these cells, SNARF-1 was considerably more concentrated within the parasite compared with the host cytoplasm. The probe accumulated in the major haemozoin-containing compartment in the parasite cytoplasm, i.e. in the mature DV (Fig. 3, red arrows). In approximately 70% of the mature-stage parasites, additional acidified compartments were observed within the parasite cytoplasm (Fig. 3, yellow arrows). These structures usually did not contain any visible haemozoin, although in some cases (Fig. 3C), a small amount of haemozoin was visible. The structures varied in size but were always smaller than the DV. These could represent endocytic structures en route to the DV, or separate digestive compartments. Both the DV and the extra-DV compartments appeared green in the R_ry images, indicating that they are acidified.

It should be noted that the DV compartment is very susceptible to photo-induced damage. At higher illumination levels or upon continuous illumination the R_ry value for this compartment increased during imaging (data not shown). Illumination-induced changes in DV pH have been reported previously for another pH-sensitive probe (Wissing et al., 2002). Light-induced generation of damaging hydroxyl radicals during bleaching might be exacerbated by the presence of haemoglobin breakdown products in the DV. We used the lowest possible levels of illumination and used highly sensitive APD detectors to minimise any light-induced artefacts in the data presented here. None the less, it is difficult to ascribe a reliable pH to the DV compartment. By contrast, the extra-DV compartment is less susceptible to photo-induced damage (as is the case for the acidic compartments in ring-stage parasites). This permitted a more reliable estimate of R_ry (0.82±0.04; n=125; 12 experiments), corresponding to a pH <6. Application of a laser pulse to an extra-DV compartment completely ablated the fluorescence associated with this compartment but did not affect the intensity of the DV (Fig. 3A) and vice versa (Fig. 3B,C). The same result was obtained in a total of 75 cells, indicating that the lumens of these compartments are not connected.

Plasmepsin II is a protease involved in haemoglobin digestion and represents a marker for compartments where haemoglobin digestion is occurring. We examined plasmepsin-II-GFP transfectants (Klemba et al., 2004) developing in TMR-dextran-containing RBCs (Fig. 4A-D). In mid-ring-stage parasites (Fig. 4A), plasmepsin-II-GFP was co-located with TMR-dextran in the peripheral puncta (Fig. 4A, yellow arrows), indicating that plasmepsin II is delivered to endocytic compartments early in parasite development and would be available to initiate haemoglobin digestion. By contrast, the larger spherical profile structures (Fig. 4B, white arrow) had no associated plasmepsin-II-GFP. In more mature-stage parasites, plasmepsin-II-GFP and TMR-dextran were co-located in the DV (Fig. 4C,D, yellow arrows) and in some extra-DV and endocytic compartments (Fig. 4C,D, red arrowheads). However, there are some peripheral plasmepsin-II-GFP compartments that are not associated with accumulation of TMR-dextran (Fig. 4C, blue arrows). These structures might represent regions of the parasite surface where cytostomal invaginations are forming, but which have not yet budded from the PPM.

We also imaged cells labelled with LysoSensor Blue, a weak base that accumulates in acidic compartments (Lin et al., 2001; Wissing et al., 2002). We found that low levels of the probe and low level illumination were needed to prevent alkalinisation of smaller compartments and to avoid photo-induced damage. Under optimised imaging conditions, LysoSensor Blue accumulated in the punctate peripheral compartments in ring-stage-infected RBCs (Fig. 4E,F, depicted in red) and overlapped with the SNARF-1 fluorescence (Fig. 4E,F, depicted in green). It is not present in the larger spherical profile structures (Fig. 4E, arrow). In mature-stage parasites, LysoSensor Blue accumulated in the DV where its fluorescence overlapped with the SNARF-1 fluorescence (Fig. 4G,H). The SNARF-1-containing extra-DV compartments were also labelled with LysoSensor Blue, confirming that these compartments are acidic (Fig. 4H, yellow arrow). Interestingly, there were also some compartments in the cytoplasm of mature-stage parasites that were weakly labelled with LysoSensor Blue, but not with SNARF-1 (Fig. 4H, blue arrow). These might represent additional endocytic or secretory compartments (e.g. rhoptries) that are separate from the haemoglobin-uptake pathway.

Electron tomography is an invaluable tool for obtaining 3D information from samples at high resolution (Hanssen et al., 2008) and can be combined with serial sectioning techniques to permit high-resolution imaging of organelles, and even whole cells (Lucic et al., 2008; Noske et al., 2008). Sections are examined in an EM operating at moderate to high voltage with a tiltable stage. The images are aligned and tomographic reconstructions of the sample are generated computationally using segmentation tool. We carried out an electron tomographic analysis of serial sections (7-15 sections of ~300 nm) of infected RBCs (K1 strain) at six different stages of development.

Individual virtual sections (7 nm) from within tomograms of whole parasites at early development stages are presented in Fig. 5A,E,I,M. (See translations through entire data sets for Fig. 5E and M in supplementary material Movies 1 and 2.) The parasite-limiting membrane and selected features within the parasite were rendered to reveal details of the 3D organisation. Rotations of the rendered data for B,F,J,N are presented in supplementary material Movies 3-6. The parasite surface is depicted in translucent cream, the cytostome as a yellow ring. Attached cytostomal invaginations are represented in translucent yellow and the nucleus is in gold. The surface is rendered in opaque gold in the right panels.

At the earliest stage examined (Fig. 5A-D; parasite volume, 2.60 fl), very few intracellular features were apparent. The parasite exhibited a cupped-hand morphology with several finger-like extensions (Fig. 5B, red arrows), but there was no evidence for endocytic vesicles. Another parasite (Fig. 5E-H; parasite volume, 2.54 fl) exhibited a flatter shape and the characteristic cytostome was associated with a cytostomal invagination (Fig. 5E,F, red arrows). The cytostome comprised an electron-dense ring with an internal diameter of ~90 nm (outer diameter ~190 nm) that encircled the neck of invaginations of the PPM and PVM (membrane-limited opening of ~55 nm). Additional sections through cytostomes (Fig. 5Q,R) reveal the complex ultrastructure of this feature. The endocytic flasks remained connected to the RBC cytoplasm through the open mouth of the cytostome ring. A number of small vesicular structures near the parasite periphery (Fig. 5F-H, depicted in orange) are likely to be derived from cytostomal invaginations. These contained small crystals (Fig. 5E, orange; also see translation through the tomogram in supplementary material Movie 1). This indicates that haemoglobin digestion and haemozoin formation have been initiated. These structures probably represent the acidic peripheral compartments observed in SNARF-1-dextran-labelled ring-stage parasites.

A slightly larger volume parasite (Fig. 5I-L; 2.65 fl) had a cupped profile (Fig. 5L, supplementary material Movie 5). This cell contained several vesicles with microcrystalline contents (Fig. 5I, orange) that accumulated in one region of the cell. In a more mature parasite (Fig. 5M-P; 3.0 fl), the cup-shape was even more pronounced (see Fig. 5P). Two cytostomes were observed at the surface of the parasite (Fig. 5O, arrowheads) and a series of vesicles and a larger flattened structure containing small crystalline structures were present in one region of the parasite (Fig. 5M-O, orange, blue). The larger compartment might represent the initial stage of coalescence of the pigment-containing vesicles into a single DV. The double-membrane-bound haemoglobin-containing structure (asterisk) is likely to be equivalent to structure referred to as the ‘big gulp’ (Elliott et al., 2008). In single sections it appeared to be inside the parasite, however, a global analysis revealed that this invagination remained connected to the RBC cytoplasm (see supplementary material Movies 2 and 6). A region of a late-ring-to early trophozoite-infected RBC was examined (Fig. 5S,T). This cell had a structure (depicted in pink) that contained haemoglobin, as well as a crystalline feature (black arrowhead). This might represent a haemoglobin-containing vesicle where haemozoin formation is being initiated.

We also generated individual virtual sections (7 nm) and rendered tomograms of trophozoite-infected RBCs (Fig. 6, supplementary material Movies 7-9). In a mid-trophozoite parasite (Fig. 6A-D; 12.1 fl; two nuclei), a centrally located DV, containing large crystals of haemozoin, was surrounded by a single membrane. The DV was folded and wrapped around a region of parasite cytoplasm (Fig. 6A,C, blue outline/rendering). An additional compartment, adjacent to the DV, contained microcrystallites of haemozoin (Fig. 6B-D, orange outline, black arrow). This compartment is probably equivalent to one of the acidified extra-DV compartments observed in the SNARF-1-dextran-labelled trophozoites. The cell had three cytostomal invaginations (Fig. 6C-D, yellow rings with attached translucent yellow pouches), one of which was visible in the virtual section (Fig. 6A, arrow). The trophozoite stage sees the genesis of small compartments comprising a clear lumen around an electron-dense body (Fig. 6, pink outline/rendering). These are likely to be acidocalcisomes (Docampo et al., 2005).

In a more mature-stage parasite (21.7 fl, four nuclei), a large DV was surrounded by a single membrane (Fig. 6E-H, blue). Characteristic electron-dense cytostomes, again with an internal diameter of ~90 nm, and associated endocytic flasks were observed at the parasite surface (depicted in translucent yellow; yellow arrow in Fig. 6E). A cytostome also constricted a bi-lobal haemoglobin-containing feature that appeared to be inside the parasite (Fig. 6F,G, yellow arrows). This type of compartment might also contribute to the extra-DV compartments observed in SNARF-1-dextran-labelled trophozoites; however, the structure was opposed to the PPM at the interface between two physical EM sections, so we cannot rule out the possibility that it retains some connectivity with the PPM. Interestingly, we observed a haemoglobin-containing vesicle located within the DV (Fig. 6E, square). A virtual section though a tilted plane suggests that the DV is engulfing the haemoglobin-containing vesicle (see inset). The vesicle contained an electron-lucent body with the appearance of a haemozoin crystal.

Two recent studies of DV genesis and haemoglobin uptake processes used serial thin section electron microscopy (Elliott et al., 2008; Lazarus et al., 2008). However, because of the huge amount of work involved in this very challenging technique, these studies were restricted to reconstruction of a limited cell depth or of a few whole cells. Moreover the resolution in the z-plane is determined by the section thickness (i.e. ~70 nm), which limits the ultrastructural information. In this work, we used electron tomography to study the digestive apparatus. Electron tomography permits analysis of 300-400 nm sections with enhanced z-plane resolution and has been used to identify novel features of the P. falciparum exomembrane system (Hanssen et al., 2008; Henrich et al., 2009). We implemented methods for collecting data from multiple sections and used programs for tiling and stitching tomographic data (Hanssen et al., 2009; Noske et al., 2008) to generate whole parasite images; this allowed us to re-examine the 3D ultrastructure of the digestive apparatus at different stages of parasite development.

We compared the electron tomography data with images of live parasites in resealed RBCs containing a fluorescent tracer. Resealed RBCs have been validated previously for use in studies of P. falciparum invasion and growth (Dluzewski et al., 1985; Rangachari et al., 1987) and for trapping inhibitors in the RBC cytoplasm (Frankland et al., 2006; Tilley et al., 1990). In this work, we resealed RBCs to contain SNARF-1-dextran, using it as a convenient, chemically stable, non-toxic, pH-sensitive marker of endocytic processes. We optimised temperature and gas control in the microscope sample chamber to allow imaging of viable parasites. This assured us that any observations have physiological relevance.

Very early stage parasites developing in SNARF-1-labelled RBCs showed no endocytic compartments; however, at the mid-ring stage, the fluorescent marker was observed in acidified peripheral compartments. Electron tomography revealed cytostomal invaginations that remained connected to the RBC cytoplasm, as well as several small peripherally located vesicular compartments containing microcrystals of haemozoin. We suggest that these cytostome-mediated endocytic events represent the first step in the genesis of the haemoglobin-degrading apparatus of P. falciparum.

In late-ring-stage parasites, the SNARF-1 signal in the endocytic structures was often more intense than in the RBC cytoplasm. As SNARF-1 fluorescence intensity decreases at low pH (Bassnett et al., 1990), this indicates that endocytosis of the host compartment is followed by a process that concentrates the contents of the endocytic compartment. Excess lipids in the cytostomal vesicles might be consumed by internalisation and degradation, providing a catalyst for haemozoin formation (Bendrat et al., 1995; Jackson et al., 2004). Electron tomography confirms the accumulation of haemozoin crystals in late-ring-stage parasites. Moreover, a GFP chimera of plasmepsin II was delivered to the early endocytic structures, indicating that conditions for initiation of haemoglobin degradation are met. The data suggest that following budding of cytostomal vesicles, the integrity of the inner membrane is rapidly lost, allowing mixing of proteases and haemoglobin in an acidified compartment; this would initiate haemoglobin digestion and haemozoin formation. Indeed, we observed examples of haemoglobin-containing endocytic compartments apparently undergoing the first stages of crystal formation. Our observation of haemoglobin digestion in mid-ring-stage parasites contrasts with the often-quoted view that haemoglobin degradation is only initiated in more mature-stage parasites.

It is interesting to consider how the pH of the ring-stage endocytic structures might be controlled. The membrane potential across the PPM is maintained by the electrogenic extrusion of protons via a V-type proton pump (Allen and Kirk, 2004; Mikkelsen et al., 1982). Acidification of the DV is thought to involve the same V-type proton-ATPase, as well as a proton-pyrophosphatase (Saliba et al., 2003). It appears likely that proton pumps are transferred to the DV via the same endocytic process that drives haemoglobin uptake. Thus, endocytosis of a region of the PPM would capture proton pumps leading to the immediate acidification of the endocytic compartments once it has budded from the PPM. The fact that we did not observe any neutral compartments that are wholly within the parasite is consistent with their rapid and obligate acidification.

A larger fluorescent feature, with a spherical appearance, that is associated with most of the SNARF-1-labelled ring-stage parasites, is likely to be equivalent to the recently described ‘big gulp’ (Elliott et al., 2008). Our electron tomography studies reveal that the ring-stage parasite adopts a pronounced cup shape, and we propose that this invagination gives rise to the feature of spherical appearance in the confocal microscopy images. Indeed, early ultrastructural studies showed that shortly after invasion the ovoid-shaped merozoite flattens into a thin disc, then invaginates to form a cup shape (Bannister et al., 2000; Bannister et al., 2004; Langreth et al., 1978). However, we found that this structure is not acidified and remains connected to the RBC cytoplasm. Moreover it co-exists with the peripheral structures in which haemoglobin digestion is occurring. Taken together, our live-cell imaging and electron tomography data argue against the ‘big gulp’ as the progenitor of the DV and instead indicate the formation of several cytostome-derived acidified compartments as the earliest event in DV genesis, as originally postulated by Aikawa and colleagues (Aikawa et al., 1966).

An interesting question is what determines the cup shape of the ring parasites. Shortly after invasion, material is released from the dense granules into the PV (Aikawa et al., 1990), followed by the elaboration of different membrane structures in the RBC cytoplasm (Lanzer et al., 2006; Tilley et al., 2008). The membrane vesicles involved in protein export (Marti et al., 2004) would be expected to expand the parasite surface; however, the high concentration of haemoglobin in the RBC cytoplasm might prevent a significant increase in parasite volume. So the cup shape might simply reflect the need to increase surface area without a significant increase in volume. The increased parasite surface area might also provide material that can be endocytosed via the cytostomes. Indeed, elongation of the endocytic flask to accommodate the expanding surface area of the parasite could lead to strain at the constricted neck of the flask and promote budding into the parasite cytoplasm.

As the parasite matures the peripheral structures coalesce into a central acidic DV that is marked by the presence of prominent haemozoin crystals. Another recent study (Dluzewski et al., 2008) also concluded that small crystal-containing vacuoles coalesce as the parasite matures to form a single DV. The trophozoite-stage parasites have 3-4 cytostomes with attached endocytic invaginations that remain open to the RBC cytoplasm. A recent study suggested that the cytostomal invaginations do not detach and migrate to the DV but extend till they contact and are engulfed by the DV (Lazarus et al., 2008). By contrast, we observed (in the SNARF-1-labelled cells) acidified extra-DV compartments in the parasite cytoplasm (~70% of the trophozoite-stage parasites examined). Many of these were tightly juxtaposed to the DV and they might have been assumed to be part of the DV in previous studies. However, selective photobleaching experiments showed that they are indeed separate structures. Plasmepsin-II-GFP was present in these acidified structures and electron tomography revealed examples of compartments adjacent to the DV that contained haemozoin crystals or small haemozoin crystals in the presence of haemoglobin. These structures are presumably derived from large cytostomal invaginations and appear to represent relatively long-lived extra-DV compartments. We also observed a vesicle that appeared to contain both haemoglobin and an early haemozoin crystal being engulfed by the DV. This suggests that the DV can phagocytose haemoglobin-containing structures as well as fusing with them. Taken together, our data suggest that following endocytosis, haemoglobin digestion can be initiated before, during or after delivery of the endocytic structures to the DV.

Our data provide clear evidence for the budding of cytostome-derived vesicles into the parasite cytoplasm and support previous studies showing that this is a major pathway for the uptake of haemoglobin in both ring-stage and trophozoite-stage parasites. We also observed a large invaginated structure bisected by a cytostome that might represent an example of a phagosome (Elliott et al., 2008). At this stage, we cannot exclude a role for an additional cytostome-independent pathway in the trophozoite stage. Identification of the protein components of the cytostome and generation of transfectants expressing GFP chimeras of these proteins would make it feasible to distinguish cytostomal invaginations from phagotrophs in live cells and would help to resolve this question.

We found that the pH of the pre-DV and extra-DV compartments was less than 6. This is consistent with other recent estimates of DV pH (Klonis et al., 2007; Kuhn et al., 2007). The pH optima for the DV plasmepsins, histo-aspartic protease (HAP) and the falcipains are in the range pH 5-6 (Banerjee et al., 2002; Goldberg et al., 1990; Rosenthal, 2004) and the pH optimum for β-haematin formation is between 3 and 6 (Egan et al., 2001; Slater and Cerami, 1992). Thus our data indicate that haemoglobin digestion and haemozoin will occur in both ring-stage and trophozoite-stage parasites, and in both the DV, extra-DV and pre-DV compartments.

In a previous study, we labelled plasmepsin-II-GFP transfectants with LysoSensor Blue (Klonis et al., 2007) and found that a number of plasmepsin-II-GFP-labelled features were not acidified. We concluded that these were endocytic compartments that were not acidified; however, our current data indicate that they are more likely to represent nascent cytostomal invaginations that have not yet lost the connection to the RBC cytoplasm. This is consistent with a previous immuno-EM study showing plasmepsin II located at nascent cytostomal invaginations (Klemba et al., 2004).

Our finding that the pre-DV compartments in the ring-stage parasites are acidified and that haemoglobin degradation is initiated at this stage has implications for the action of quinoline antimalarials. Quinoline antimalarials accumulate in acidic compartments and inhibit haemozoin formation (Geary et al., 1990; Hawley et al., 1996; Tilley et al., 2006). Our data suggest that both ring and trophozoite stages would accumulate chloroquine in acidic compartments. Although some studies suggest maximal chloroquine sensitivity at the trophozoite stage (Hoppe et al., 2004; Yayon et al., 1983), others find that ring stages are the most sensitive (Orjih, 1997; Zhang et al., 1986); yet others suggest that both stages are sensitive but that ring stages exhibit a ‘delayed death’ effect (Gligorijevic et al., 2008; Yayon et al., 1983). Our data indicate that chloroquine would accumulate in ring-infected RBCs and would inhibit the early stages of haemozoin formation.

The data also have implications for the action of endoperoxide antimalarials. Endoperoxides are activated by the haem that is released during haemoglobin breakdown [or by reduced iron that is derived from haem (Krishna et al., 2004; Weissbuch and Leiserowitz, 2008)]. Recent reports suggest that artemisinin targets the DV and associated lipid bodies (del Pilar Crespo et al., 2008; Hartwig et al., 2009) and also inhibits endocytosis of haemoglobin (Hoppe et al., 2004). Our data suggest that the release of haem from haemoglobin would be initiated in the ring stage of parasite growth and would be available as a source of an iron-based activator. This might well underlie the finding that artemisinin is highly active against ring-stage parasites (Skinner et al., 1996).

The live-cell imaging and multi-section electron tomography approaches used in this study represent important technical advances. In combination with data from previous studies, these approaches provide new insights and resolve some controversies regarding the endocytic processes at the different stages of parasite development. This is illustrated in the model presented in Fig. 7. The data suggest that uptake of haemoglobin commences early in the intraerythrocytic development of the parasite. Haemoglobin digestion is initiated in multiple pre-DV compartments that coalesce in the early trophozoite stage to form a central DV. The data show that cytostome-dependent endocytosis is the mechanism for uptake of haemoglobin the ring stage, and is probably the principle pathway in the trophozoite stage. Following budding of endocytic vesicles, haemoglobin digestion can occur either before, during or after fusion with the DV. These findings are important in understanding the mode of action of different classes of antimalarial drugs.

RBCs were lysed in 2.5 volumes of ice-cold 5 mM phosphate, pH 7.5, 1 mM Mg-ATP (Sigma) in the presence of 50 μM tetamethylrhodamine-dextran (TMR-dextran; 10 kDa; Invitrogen) or 5′(and 6′)-carboxy-10-dimethylamino-3-hydroxy-spiro[7H-benzo[c]xanthene-7,1′(3H)-isobenzofuran]-3′-one (SNARF-1-dextran; 10 kDa; Invitrogen) or recombinant His-tagged enhanced GFP (pEGFP-N1, Clontech) and resealed as described previously (Frankland et al., 2006; Krogstad et al., 1985). The levels of haemoglobin and incorporated SNARF-1-dextran were determined spectrophotometrically. P. falciparum (D10, 3D7, K1 strains and plasmepsin-II-GFP transfectants) was cultured and isolated as described previously (Klemba et al., 2004; Klonis et al., 2007).

Live-cell imaging of SNARF-1-dextran-containing resealed RBC was performed at 37°C using a ×100 oil immersion objective (1.4 NA) on a Zeiss LSM 510 inverted confocal microscope equipped with an incubation chamber and heated stage under an atmosphere of 5% CO₂ in air. Glass coverslips (Menzel-Glaser, #1.5) were cleaned with 10% NaOH, 60% ethanol, before being mounted in a Sykes-Moore culture chamber (Bellco Glass, NJ). The chamber was coated with concanavalin A (0.05 mg/ml, Sigma) then rinsed with RPMI. SNARF-1 was excited at 514 nm and the yellow (580-635 nm) and red (>635 nm) fluorescence detected using the avalanche photodiode detectors (APDs) of a Confocor3 system. Spot photobleaching was performed as described previously (Klonis et al., 2002). For pH calibration, resealed cells were suspended in MES-buffer saline (20 mM MES and 150 mM NaCI, pH 5.5 and 6.0), PBS saline (20 mM PBS and 150 mM NaCl, pH 6.5, 7.0, 7.5, 8.0) and TRIS saline (20 mM TRIS and −150 mM NaCl, pH 9.0) in the presence of carbonyl cyanide m-chlorophenylhydrazone (CCCP; Sigma; 10 μM). GFP and TMR-dextran were excited at 488 nm and 543 nm using a Leica TCS SP2 confocal microscope, while cells co-labelled with SNARF-1 and LysoSensor Blue-192 (1 μM at 37°C for 20 minutes; Invitrogen) were imaged using the FITC and DAPI filter cubes of an Olympus IX81 microscope.

Images were analysed using NIH ImageJ (http://rsb.info.nih.gov/ij/). We define the ratio R_ry as I_r/I_y where the I_r and I_y correspond to the fluorescence intensities in the red and yellow channels. R_ry images were constructed in HSB colour space by mapping the R_ry value at each pixel to a particular hue value and placing the total intensity (I_r + I_y) in the brightness channel. The brightness channel was often linearly adjusted to increase the contrast. Pixels that are saturated (values of 255) are coloured cyan in the R_ry images and were not analysed when calculating R_ry values. Quantification of R_ry values was performed using data from individual sections.

Infected RBCs were prepared for electron tomography as described previously (Hanssen et al., 2009). For electron tomography, 300 nm sections were cut, and collected as serial sections. Each section was layered with fiducials, stained with uranyl acetate and lead citrate and observed on a Tecnai G² F30 (FEI Company) transmission EM (Bio21 Institute, Melbourne). Tilt series were acquired using Xplore 3D (FEI Company). Tomograms were recorded between −69 and +69° at 1.5° intervals for the first axis and 3° for the second axis and aligned with IMOD (Kremer et al., 1996; Mastronarde, 1997). Segmentation models were generated with 3dmod (http://bio3d.colorado.edu/). Individual tomograms were stacked using the join programs from the eTomo GUI interface of the IMOD package. The models were corrected by a factor of 1.5 in the z direction to account for shrinkage of the resin during exposure to the electron beam.

We thank the National Health and Medical Research Council of Australia and the Australian Research Council. We thank Mike Klemba, Virginia Tech, USA, for supplying the plasmepsin-II-GFP transfectants. We thank Samantha Deed for expert technical assistance.