Lipid droplet dynamics are essential for the development of the malaria parasite Plasmodium falciparum

Plasmodium falciparum, the deadliest species of malaria parasites infecting humans, has a complex life cycle which requires transition between two intrinsically different hosts (human and mosquito). Hence, the parasite faces the challenges of having to navigate growth, proliferation, transmission and sexual reproduction in vastly different host environments. Each of these cellular processes relies on coordinated lipid access and metabolism mechanisms, with the parasite either synthesising essential lipids de novo or acquiring them from its host environment (Shears et al., 2015; Sherman, 1979; Shunmugam et al., 2022). During the asexual blood stage cycle, which is solely responsible for pathogenesis in the human host, the parasite invades a mature erythrocyte (Maier et al., 2019). Once inside, the parasite grows from a ring stage, via a trophozoite stage to a schizont stage. During the last stage, the parasite undergoes an unusual process of asexual multiplication called schizogony. Whereas most eukaryotes divide by binary fission, Plasmodium undergo a closed mitosis with rounds of DNA replication and nuclear division without cell division (cytokinesis), which leads to multinucleated cells (Francia and Striepen, 2014; Voß et al., 2023). Schizogony is the ‘mode’ of multiplication which occurs in both the liver and the blood stage of the parasite life cycle in the mammalian host. During intraerythrocytic schizogony, the parasite eventually segments into 8–32 individual daughter merozoites. After egress from the infected cell, each merozoite will invade a new erythrocyte to continue the asexual blood cell cycle. In order to provide building blocks for the fast growth and multiplication of P. falciparum within each 48 h asexual replication round, the parasite requires a ready supply of lipids to synthesise new cellular membranes. At the same time P. falciparum needs to prevent excess lipid accumulation to counteract potential lipid toxicity.

Previous studies have shown that there is a drastic change in lipid composition throughout the asexual blood stage cycle. In particular, the levels of the neutral lipid species diacylglycerol (DAG) and triacylglycerol (TAG) were elevated compared to that in uninfected erythrocytes (Gulati et al., 2015; Ridgway et al., 2022; Tran et al., 2016), whereas the levels of the other major group of neutral lipids (cholesterol esters; CE) were reduced to almost undetectable levels (Tran et al., 2016; Vielemeyer et al., 2004).

Given that neutral lipids serve as crucial repositories of energy and building material in eukaryotic cells (Olzmann and Carvalho, 2019), the remarkable enrichment of DAGs and TAGs in P. falciparum might be reflected in the dynamics of the major lipid storage organelles – lipid droplets (LDs). Despite early recognition at the end of the 19th century (Altmann, 1894), LDs were long considered as being inert fat particles (Coleman, 2020). In more recent studies, LDs have been shown to be very dynamic organelles that are crucial for lipid homeostasis in many different cell types (Farese and Walther, 2009; Fujimoto and Parton, 2011; Pol et al., 2014). They have a distinct structure, consisting of a hydrophobic core composed mainly of the neutral lipids TAG and CE, surrounded by a monolayer of phospholipids containing embedded proteins. Given that the enzymes for the last steps of TAG and CE synthesis reside in the endoplasmic reticulum (ER), the ER is regarded as key for the formation of LDs (Coleman and Lee, 2004; Fujimoto and Parton, 2011). Similarly, several cellular organelles are in need of neutral lipids that are contained within LDs and hence, interactions of LDs with other organelles can be observed (Murphy et al., 2009; Rakotonirina-Ricquebourg et al., 2022; Schuldiner and Bohnert, 2017). The role of LDs goes beyond being a mere lipid storage compartment; they have been implicated in a range of additional functions including the degradation of hydrophobic proteins (Fujimoto and Ohsaki, 2006) and haem detoxification in P. falciparum (Hoang et al., 2010).

Given the significant growth and changes occurring during the maturation of the parasites, we hypothesise that size, composition and dynamics of LDs vary significantly across the P. falciparum intraerythrocytic cell cycle.

Although fluorescence microscopy and lipidomic studies have explored the general distribution of neutral lipids in asexual stages (Gulati et al., 2015; Palacpac et al., 2004a; Tran et al., 2016; Vielemeyer et al., 2004), LD dynamics in the malaria parasite remain underexplored. Here, we applied a combination of advanced light microscopy techniques and volume imaging using focused ion beam scanning electron microscopy (FIB-SEM) in combination with inhibitors of essential enzymes in neutral lipid metabolism to characterise the dynamics and importance of LDs in the asexual life cycle stages of the parasite.

In order to characterise the content of LDs, we made use of a particular trait of the lipophilic fluorescent dye Nile Red [9-(diethylamino)-5H-benzo[a]phenoxazin-5-one], which displays altered sensitivity of the dye depending on the lipid environment (Greenspan et al., 1985). Previous studies in other organisms have shown that an excitation/emission profile of 506 nm/550 nm allows for observation of mainly non-polar head groups, whereas an excitation/emission profile of 598 nm/>610 nm shows a higher specificity for polar head groups (Diaz et al., 2008; Gao et al., 2014; Greenspan et al., 1985). Hence, signals appearing in the emission >610 nm range (termed magenta emission) indicate polar lipids [e.g. phospholipids (PLs)], whereas the signals observed in the <590 nm emission range (termed yellow emission) represent neutral lipids (e.g. acyl-glycerols and/or esterified cholesterol).

In order to gain information on the lipid distribution of RBCs infected (iRBCs) with different asexual blood cycle stages of P. falciparum, we stained the cells with Nile Red (Fig. 1A). RBCs infected with ring and trophozoite stages exhibited a few small intense puncta in the parasite cytoplasm at yellow emission. These puncta, which were reminiscent of LDs, increase in number and size in schizont stages and were often in close proximity to the digestive vacuole (DV) (as indicated by the dark hemozoin crystals observed in the DIC images). When the polar lipid excitation/emission (magenta) settings were applied, we observed a diffuse overall staining of the cytoplasm, likely representing intracellular membranes of a vast array of organelles, which are mainly composed of PLs (Fig. 1A). Most of the puncta we observed in the yellow emission also appeared under these imaging conditions. However, their intensity was faint and, particularly in later stages of the parasite development, these could no longer be distinguished from the overall fluorescence signal within the parasite. Interestingly, smaller puncta appeared around the periphery of the segmented schizont (Fig. 1A and Fig. S1, yellow arrowheads), indicating that the content of these puncta can easily be distinguished from LDs by applying these differences in spectral analysis settings.

The data in Fig. 1 (supported by Fig. S1) indicate that the spectral properties of Nile Red can be exploited to examine the intracellular distribution of different lipid classes across the intraerythrocytic cycle. This enabled us to distinguish between neutral lipid (NL)-containing LDs and other lipid-rich structures in the parasite, whereas traditional Nile Red spectral analysis (Fig. S1 and Fig. 1A, red panel) obscures the differential signals. This is particularly obvious in the schizont stages, where the overall amount of lipids increases due to requirements of new membranes around developing daughter merozoites (Fig. 1A, red and magenta).

To quantify the observed changes in number and size of NL-enriched LDs during the asexual cycle of the parasite, we first measured the intensity of the overall Nile Red signal (traditional, non-differentiating settings) in iRBCs across the different asexual cycle stages (Fig. 1B). To this end, we tightly synchronized parasite cultures and defined the analysed stages as follows: ring stages contained one nucleus (no haemozoin visible), trophozoites contained one nucleus (haemozoin clearly visible), early schizonts displayed up to five nuclei whereas mid-schizonts contained six or more nuclei, but did not contain fully segmented merozoites. The late schizont stages clearly displayed segmented daughter merozoites. The integrated intensity increased steadily as the parasite matures, reflecting the increase in lipid abundance during intraerythrocytic development. Strikingly, when employing the split excitation/emission experiment and analysing the yellow fluorescence signal from Z-stack projections of the cells, the average number of LDs observed per cell increases as the parasite matures, from being completely absent in ring stages to having, on average, over two LDs in the mid-schizont stage. However, the average number of LDs drops significantly in late schizont parasites (Fig. 1C). Using this methodology, we also determined the integrated intensity of the observed LDs and found that, similar to the number of LDs/cell, the intensity of the LDs also increased as the parasite matures, however, with a significant drop in intensity in the late schizont stages (Fig. 1D). Taken together, we observed a steady increase in LD number and LD fluorescence intensity during the growth phases intraerythrocytic cycle, peaking in mid-schizont stages but declining rapidly in late schizont stages. The decrease in late schizont stages indicates that NLs might be recruited from the storage LDs for assembly in merozoite membranes.

To increase the spatial resolution and to eliminate the discrepancies observed with the differential analysis of lipid groups, we employed osmium tetroxide staining and electron microscopy to visualise LDs. Osmium tetroxide preserves lipids throughout the preparation process for electron microscopical investigation and preferentially binds to unsaturated bonds in fatty acids, imparting electron density that is observable via electron microscopy (Adams et al., 1967; Hayes et al., 1963). A correlation between electron density of LDs and NLs content was demonstrated previously, allowing ready identification of LDs at an ultrastructural level (Cheng et al., 2009; Fujimoto et al., 2013). Using an osmium-thiocarbohydrazide-osmium (OTO) staining method (Tapia et al., 2012), we were able to identify high electron density material in droplet-like organelles throughout the examined asexual stages on two-dimensional (2D) electron microscopy sections in that we designated as LDs (Fig. 2A–D).

To overcome the limitations of 2D imaging in determining the exact number and/or size of LDs in any given parasite we analysed, we exploited a focused ion beam-scanning electron microscopy (FIB-SEM) volume imaging technique, henceforth referred to as three-dimensional (3D) FIB-SEM imaging. Here, LDs from asexual stages were segmented from 3D volume data, identified based on their high contrast threshold and morphology (Fig. 2A–D,A′–D′; Movies 1 and 2). We subsequently determined numbers and volumes of LDs from these 3D data sets to identify potential differences in LD numbers and/or volumes in the various asexual life cycle stages (Fig. 2E,F). Whereas no high-density LDs were observed in ring or early trophozoite stages, multiple LDs were identified in the parasite cytoplasm in trophozoite stages, increasing in number in early and, more significantly, in the mid-schizont stage (Fig. 2E). These numbers correspond with the number of LDs in our confocal analysis (Fig. 1A,C,D), supporting the notion that NL-enriched LDs increase from trophozoites to mid-schizont stage.

High-resolution 3D volume data sets furthermore permit the quantification of additional parameters. We subsequently determined the average volume of LDs in our 3D volume data sets (Fig. 2F). We found that the relative LD volume-to-parasite area increased almost 10-fold in mid-schizont stages when compared to LD volumes in trophozoite stages (Fig. 2F). However, the average volume of LDs decreased again at late schizont stages (Fig. 2F), corroborating our confocal microscopy findings of a very small pool of NL available in late schizont stages (Fig. 1A, lowest panel, 1D). Notably, LDs were only observed in the residual body, closely associated with the remnant DV in the analysed late schizont stages (Fig. 2D, arrowhead). Again, this is in line with our confocal data (Fig. 1), where NL-enriched LDs were also sometimes found in close proximity to the DV. Rhoptries, apically localised organelles that first appear in daughter merozoites during late schizogony, also contained electron-dense material (Fig. 2D, red arrowheads) but with a much lower density than LDs, and can therefore be clearly distinguished via this method. This indicates that the small puncta found at the periphery of late schizonts by fluorescence microscopy (Fig. 1A) most likely represent rhoptries. Of note, no high electron density (NL-enriched) droplets were detected in merozoites within the late stage schizonts, suggesting that merozoites do not contain LDs. Taken together, high-resolution 3D FIB-SEM permits accurate quantification of tempo-spatial LD dynamics during the 48 h P. falciparum growth cycle inside erythrocytes.

Membrane contact sites (MCS) refer to regions where the membranes of two organelles come into close proximity without merging (Prinz et al., 2020; Scorrano et al., 2019). These contact sites facilitate direct communication and exchange of lipids and other molecules between organelles, contributing to various cellular functions. In the FIB-SEM data, the tight interaction of LDs and other cellular organelles appear as high electron density regions that connect LDs with various organelles (Figs S2, S3).

We set out to determine whether LDs have preferential contact sites with particular organelles and also whether these preferences might differ across the asexual cycle stages. This quantitative approach could potentially offer clues about LD functions as parasites develop within their host cell. LDs were often in close contact with other cellular organelles (sometimes with several at the same time) (Fig. 3). We observed frequent contact sites between LDs and the ER across all studied intraerythrocytic stages, and with the nucleus in trophozoites and early- and mid-schizonts (a combined 43–53% of all observed interactions). In trophozoites and the various studied schizont stages, contact sites of LDs with the DV were also commonly observed, with 27–36% of all observed MCS occurring between these two organelles. Similarly, contact between LDs and the mitochondrion occurred in 7–18% of cases, and between LDs and invaginations of the parasite plasma membrane in 7–18% of analysed LDs. These invaginations resembled cytostomes, the sites at which haemoglobin and other host cell material is taken up by parasites (Fig. 3A–D,A′–D′; Movie 2).

The ER is known to be an LD biosynthesis site in mammalian cells and a range of other eukaryotes (Choudhary and Schneiter, 2021; Fujimoto and Parton, 2011; Hugenroth and Bohnert, 2020; Jacquier et al., 2011). Interestingly, when studying LD–ER contact sites through our 3D data sets, we frequently observed small LDs that appeared to be budding from the ER, which could represent sites of LD formation (Fig. S2A,A′,B,B′). To our knowledge, signatures of LD biogenesis have not been shown before in P. falciparum. The consistent occurrence of contact sites with the mitochondrion, DV, ER and uptake sites throughout the asexual erythrocytic life cycle (Fig. 3E; Figs S2, S3) suggests that LDs contribute to numerous aspects of the biology and development of the parasite.

We notice an apparent absence of LD contact sites with the nucleus in late schizont stages (Fig. 3E). However, given that the ER has its origin around the nuclear envelope and stays intricately linked with the nucleus, it is difficult to distinguish between LDs that might provide lipids for the growth of the nuclear envelope and those associated with the nascent ER close to the nucleus. In late schizont stages, the majority of LDs were associated with the DV and ER in the residual body – a structure that contains parasite material, such as hemozoin, that has not been incorporated into daughter merozoites after segmentation (Rudlaff et al., 2020). This might indicate that remnants of organelles, including ‘unused’ LDs, end up in this disposal site.

Together, we observed dynamic interactions of the LDs with several organelles during the course of intra-erythrocytic growth, which is indicative of central roles for organelle functions.

TAG hydrolysis is a universal lipid metabolic process whereby TAGs are broken down into glycerol and fatty acids (FAs). Orlistat, a commercially available anti-obesity drug, irreversibly inhibits the lipase responsible for this breakdown and has been demonstrated to inhibit malaria parasite growth (Yuan et al., 2011) (Fig. 4A). This was shown to be a consequence of elevated TAG levels in late schizont stages, resulting in a subsequent inhibition of merozoite egress from the host RBC (Gulati et al., 2015).

We hypothesised that the TAGs present in LDs might be contributing to merozoite formation, which could explain the reduction of LD number and volume in segmenting schizonts. We reasoned that, if our hypothesis is correct, inhibiting TAG catabolism through the addition of Orlistat would inhibit merozoite formation, and that adding Orlistat after the depletion of LDs in segmenting schizonts would not lead to an impairment in merozoite formation. To this end, we added 60 µM (8×IC₅₀ concentration) of Orlistat to synchronised parasites at 8, 20, 32 and 44 h after invasion and determined their appearance at 60 h, compared to a DMSO control (Fig. 4B,C). When Orlistat was added between 8 and 32 h after invasion, the parasites progressed to the schizont stage where they stalled, unable to complete merozoite formation, eventually resulting in death of the parasites, as observed at the 84 h time point. In contrast, when the drug was added at 44 h into the cycle, we observed that >40% of parasites were able to survive, resulting in abundant rings at 60 h into the assay (Fig. 4B,C). These findings corroborate the results of Gulati et al. (2015) indicating that TAG hydrolysis and FA release is crucial at the time point of merozoite membrane formation.

Given that we had observed a decrease in the volume of NL-enriched LDs from mid to late schizont stages (Fig. 2F), we wanted to test whether LDs instead accumulate in the presence of Orlistat in late schizont stages. The split emission fluorescence microscopy data of Nile Red stained parasites (yellow emission) showed a delay in nuclear division when Orlistat was added between 8 and 32 h post invasion (Fig. 4B,D). Nuclear division occurred eventually, yet the majority of parasites remained arrested at the mid-schizont stage, failing to successfully undergo full merozoite formation (Fig. 4B,D, 44 h and 60 h). Although the control cells showed the typical membrane pattern surrounding each individual merozoite at 44 h (late schizont stage, Fig. 4E), treated cells displayed a much higher number of intensely fluorescing large puncta (Fig. 4D). These puncta had accumulated non-polar lipids (likely TAGs, Fig. 4E), which was typical for the LDs that were still present at 60 h in the presence of Orlistat, except when the drug was added after merozoite formation as indicated above (Fig. 4E, lowest panel). These results support the hypothesis that TAG is consumed from LDs in support of the schizogony of the parasite. The observations indicate that lipids derived from TAG hydrolysis in LDs (i.e. FA release from LDs) are essential for merozoite development.

The majority of TAG is synthesised at the ER membrane through the glycerol 3-phosphate (Kennedy) pathway (Lehner and Kuksis, 1996), and the resulting TAG is then packaged into LDs. LDs might undergo expansion through the activity of TAG synthesis enzymes (Olzmann and Carvalho, 2019). One of the crucial enzymes involved in TAG biosynthesis is diacylglycerol acyltransferase (DGAT) (Bell and Coleman, 1980). This enzyme catalyses the formation of an ester linkage between fatty acyl CoA and the free hydroxyl group of DAG to synthesise TAG (Fig. 5A). In Plasmodium, TAG synthesis occurs mainly during the late trophozoite stage (Palacpac et al., 2004a; Vielemeyer et al., 2004). This is the stage at which we start to observe prominent LDs (Figs 1 and 2).

Xanthohumol (XN), a prenylated chalcone from Homulus lupulus, has been shown to inhibit both the DGAT1 and DGAT2 enzymes of mammalian and yeast cells (Inokoshi et al., 2009; Tabata et al., 1997). In P. falciparum, XN inhibited the in vitro replication of the parasite and interfered with the process of haemin degradation (Frölich et al., 2005). We hypothesised that XN inhibition of the parasite DGAT enzyme would impair the development of LDs. To determine the effects of XN on LD development, we added XN at different time points post invasion, and performed Nile Red labelling to visualise LDs (Fig. 5). In contrast to our results with Orlistat, we observed immediate arrest of parasite development, irrespective of the time point of XN addition (Fig. 5B,C). Interestingly, we observed an absence of LDs when XN was added prior to the onset of LD formation in late trophozoites, confirming our hypothesis that XN interferes with the formation of LDs (Fig. 5D,E). Small amounts of NLs surrounding the nucleus in the pyknotic parasite cells were observed, resembling the pattern observed during the untreated control ring stages. However, no obvious LD accumulation was observed in cells treated earlier than 32 h post infection (Fig. 5E). These results support the hypothesis that XN interferes with TAG synthesis and storage, which seems to be essential for development of the parasite through the asexual part of its life cycle. When treated at 32 or 44 h after invasion, TAG-containing LDs seemed to form (Fig. 5D,E) but parasite development was still impaired, with mostly pyknotic parasites observed at the 60 h time point (Fig. 5C), which could be due to another mode of action of XN.

Previous reports have shown that Plasmodium contains only one DGAT enzyme, which most closely resembles the DGAT1 enzyme of mammalian cells (Palacpac et al., 2004b; Vial and Ben Mamoun, 2005; Vielemeyer et al., 2004). In order to identify whether XN affects the parasite by interfering with the parasite DGAT or haemin degradation, we repeated the experiment with T863 (Fig. 6A). T863 is a potent, selective inhibitor of DGAT1 that acts on the acyl-CoA binding site of DGAT1 and inhibits DGAT1-mediated TAG formation in cells (Cao et al., 2011). When added to ring stage parasites, T863 arrested parasite development in a similar rapid fashion to what was seen with XN (Fig. 6B) leading to the formation of pyknotic cells (Fig. 6C) without any significant accumulation of LDs throughout the assay (Fig. 6D). However, in contrast to XN, when T863 was added at 20 or 32 h after invasion, parasites seemed to develop and divide, although slower than the control cells and with only ∼25% cell survival at the 60 h time point (Fig. 6B,C). When T863 was added at 44 h post invasion, the parasites seemed to develop similarly to the ones treated with Orlistat, with the survival rate (∼50%) being very similar to what was seen with this treatment as well (Fig. 6B,C). In contrast to Orlistat treatment this late in the cycle though, no LDs could be observed in our split emission and ratio analysis (Fig. 6D,E). This suggests that the inhibitory effect of T863, which selectively targets DGAT-1, prevents LD formation when added at an early stage, where TAG is not abundant yet. However, when added at a stage where LDs are already present (32 h post invasion), the parasites can still develop for a while, utilising the available TAG for membrane formation. Depending on the window of synchronisation, this explains the >50% survival rate when observed at 60 h. For parasites that were earlier in the cycle when the inhibitor was added, it is likely that not enough TAG could be accumulated in the presence of T863, hence the amount of TAG was not sufficient for merozoite membrane formation and subsequent merozoite release. These results also indicate that the effect of XN in later stage parasites is independent of DGAT inhibition (and might be a reflection of insufficient haemin digestion or lipid toxity due to non-segregated lipids in the cytoplasm).

To substantiate our light microscopical observations on the effects of Orlistat, XN and T863 on LDs, we exposed highly synchronised infected erythrocytes to each of the individual drugs at 32 h post invasion. We then sampled and prepared cells for transmission electron microscopy (TEM) analysis at 44 h and 60 h post invasion.

Whereas control parasites revealed fully formed merozoites ready for egress in most cells observed at 44 h post invasion (Fig. 7A, left panel), the inhibitor-treated cells all displayed multiple parasite nuclei but none of them had progressed to segmentation into individual merozoites within the host erythrocyte (Fig. 7B–D). In Orlistat-treated cells, multiple large LDs were observed at 44 h post invasion, with some of them closely associated with the DV (Fig. 7B, yellow arrowheads). Cells treated with XN displayed large (mostly single) LDs at 44 h post invasion, which were clearly identifiable by their electron-dense content (Fig. 7C). In contrast, cells treated with the inhibitor T863 displayed some globular organelles without a discernible double membrane; these were reminiscent of LDs, but with a much lower electron density (Fig. 7D). These observations corroborate our findings from the split emission experiments. Cells treated with Orlistat and XN displayed multinucleated parasites interspersed with non-polar lipid-containing LDs when treated with the inhibitors at 32 h post invasion and observed at 44 h. In contrast, T863-treated cells displayed LDs depleted of non-polar lipids (Figs 4,5,6D).

In order to study the fate of the cells after inhibitor treatment into the next erythrocytic cycle, cells were also sampled at 60 h post invasion (Fig. 7, right panel). Whereas control cells developed into rings as expected (Fig. 7A), Orlistat- and XN-treated cells exhibited plasma membrane disintegration, and membrane whorls, blebs and aggregations were visible in the cytoplasm, all hallmarks of cellular degradation. In contrast, when cells were treated with T863 at 32 h post invasion, parasite development was delayed (no rings observed at 60 h post invasion) but segmentation of the parasite into daughter merozoites was clearly visible (Fig. 7D).

The ultrastructural analysis of cells treated with the three different inhibitors support our light microscopy results. In the absence of TAG hydrolysis (Orlistat treatment) or decreased TAG synthesis and storage capacity (XN treatment), cell segmentation is prevented, leading to parasite death. In contrast, in cells treated with T863 at this later stage in the erythrocytic cycle, when TAG-enriched LDs would have already formed, the inhibitor delayed parasite development but did not prevent segmentation, suggesting that the inhibitory effect of T863 mainly targets early TAG synthesis.

Overall, the data from our inhibitor studies suggest that the NLs in parasite LDs mainly comprise TAGs, and that interference with TAG synthesis, storage or hydrolysis results in detrimental effects for the parasite. This highlights the fundamental importance of TAG metabolism for parasite survival.

Applying two novel methodological approaches to the study of LDs in P. falciparum – namely emission splitting of the Nile Red signal and 3D FIB-SEM – together with the temporal resolution of different developmental stages, we were able to reveal the detailed dynamic of LD synthesis, growth and utilisation during the asexual replication cycle of the malaria parasite in the human host.

NLs accumulate in LDs until the mid-schizont stage before these storage compartments are utilised to meet the demand for NLs during the formation of merozoites in segmenting schizonts. Although a constant increase in NLs from rings to late schizonts is implied in our data that utilised traditional Nile Red imaging conditions (excitation: 559 nm; emission: 636 nm) (Fig. 1; Fig. S1), the refined analysis using the split emission approach revealed a more differentiated picture – both LD numbers and NL content decreased at the late schizont stage, indicating that NLs, such as acylglycerols and CEs, are hydrolysed to provide building blocks for membrane formation. While the parasite accumulates lipids for later use from a nutrient-rich host environment, LDs prevent the cytosol from being overloaded by excess free lipids and hence avoid lipid toxicity.

In addition to using the emission splitting system that differentiates polar from non-polar lipids in the parasite cell, we employed 3D FIB-SEM. In this protocol, LD content is intensely stained with heavy metals, making LDs easily distinguishable from other organelles in the cellular context. The ultrastructural data corroborated our light microscopy findings where number and volume of NL-enriched LDs increased as the parasites developed to the mid-schizont stage. Whereas the number of LDs decreased only slightly from mid to late schizonts, the volume of LDs decreased dramatically to <25% of the mid-schizont level. This result clearly demonstrates that light microscopy analysis alone, due to its limited resolution, especially in a very small organism like P. falciparum (< 5 µm), will not disclose the full picture of organellar dynamics. The detailed ultrastructural analysis also revealed that the small LDs remaining in late schizont stages were mainly associated with the residual body. This indicates that small remnants of LDs might be left over for disposal once the merozoites are released from the host cell to start a new intraerythrocytic cycle. The non-utilisation and surplus of lipids that are discarded in the DV might seem to be contradictory to the ‘economy of nature’. However, our experiments are conducted under conditions where parasite access to lipids and lipid precursors is not limited and hence the possibility exists that fewer surplus lipids might be found in a more lipid-restricted environment. In perspective, these techniques can now be applied to clinical samples, in order to study LD dynamics during human infections.

Taken together, split emission analysis of Nile Red-stained P. falciparum-iRBC in combination with analysing LDs at ultrastructural resolution in 3D revealed significant changes in the number and size of LDs in the parasite's asexual cycle.

Biogenesis and subsequent dynamics of LDs rely on close contact of LDs with various organelles. Interrogation of our 3D data revealed that most LDs were in close contact with various organelles throughout the asexual life cycle via membrane contact sites (MCS). These sites are regions where membranes of two organelles form contact sites that enable the transfer of lipids, metabolites and ions (Scorrano et al., 2019).

LD formation starts with synthesis of the molecules of the organelle LP core, which is mediated by enzymes located primarily on the ER (Pol et al., 2014; Watkins et al., 2007). Unsurprisingly, the most abundant LD contact sites in our study were with the ER and the nucleus (the nuclear envelope being contiguous with the ER).

In mid-schizont stages, we often observed an accumulation of LDs that were much bigger in size compared to LDs in earlier stages (Figs 1A, 2C, 3C). Interestingly, we observed potential LD budding sites from the ER in multiple parasite stages (Fig. S2A,B,F), suggesting continuous LD biogenesis during the growth of the parasite. These might be able to fuse to form larger LDs in schizont stages (Fig. S2D, green arrowhead).

We observed a drop in combined ER–LD and nucleus–LD contact sites in late schizont stages, in line with the shift from accumulation to usage of LD lipids. In late stage schizonts, no LD interactions with nuclei were observed. In these stages, the nuclei are segregated to newly formed merozoites and hence spatially separated from the few remaining LDs that are found in the residual body (Fig. 3E).

In P. falciparum, LDs have been implicated in the delivery of NL to the DV, where they are thought to play a role in haemozoin formation, a process that detoxifies haem that is released during haemoglobin digestion in the DV of the parasite (Ambele and Egan, 2012; Jackson et al., 2004). Hence, the high percentage of observed LD–DV contacts in our study supports the proclaimed role of NLs in haem detoxification, a process that is essential for parasite survival.

We observed an increase in LD–mitochondria contact sites from early- to mid-schizonts. In many cell types, LDs deliver FA-containing NLs to fuel the energy generating fatty acid oxidation of mitochondria (Rakotonirina-Ricquebourg et al., 2022). This is unlikely to be the case in P. falciparum given that the parasite lacks a functional β-oxidation pathway in its mitochondria (van Dooren et al., 2006). However, delivery of energy in form of ATP from mitochondria to LDs has been reported to fuel FA activation in LDs of mammalian adipocytes. The high demand for energy to activate and access the lipids stored in LDs might explain the increase in LD–mitochondria contact in mid-schizont stages.

The multifaceted interaction of LDs with various organelles underscores the versatile nature of LDs in orchestrating cellular functions. The quantification of contact points between organelles might indicate the relative importance of the LD–organelle interaction for a particular function; however, this can also be obscured by spatial constraints. Further investigations into proteins involved in the molecular mechanisms could reveal exciting new insights into specific functions associated with these contact sites.

Our study also provides more detailed insights into the composition of LDs. Our light microscopy data represents for the first time a detailed quantitative analysis of neutral lipids throughout the asexual life cycle stages by employing a methodology that allows separation of lipid species due to their polarity (Greenspan et al., 1985; Greenspan and Fowler, 1985).

Traditional Nile Red analyses indicate a steady increase of NLs throughout the development of the parasites (Fig. S1). By separating the signal for polar lipids (magenta) and the signal for NL (yellow), we could show that NLs accumulate in LDs until the mid-schizont stage. In contrast, polar lipids continuously increase and are mostly associated with parasite membranes (Fig. 1).

To distinguish between synthesis and utilisation of the lipids contained in LDs in the parasite, we combined the split emission approach with the use of specific inhibitors to disrupt lipid metabolism in LDs.

TAGs have been identified as one of the major NL classes in P. falciparum LDs (Vielemeyer et al., 2004), and FAs released from TAG hydrolysis can serve as a source of FAs for various metabolic processes (Watt and Spriet, 2010). Using the TAG hydrolysis inhibitor Orlistat, we identified the contribution of LDs to meeting these demands during merozoite formation. When Orlistat was added between 20 and 32 h of development, the parasites continued to develop, but failed to produce proper merozoites, which were also not released. The NL ratio profile clearly shows an accumulation of NL-containing LDs in late schizont stages, which is not observed in the controls. These findings indicate that TAG is the major NL found in LD of P. falciparum, and highlight the importance of TAG utilisation during schizogony, where the development of organellar and plasma membranes of individual merozoites heavily relies on the availability of membrane building blocks. Furthermore, our results offer a plausible explanation for the severe growth phenotype associated with Orlistat exposure described by Gulati et al. (2015).

In addition to the dependence on free FAs for membrane biogenesis, the parasite relies on controlling DAG levels during the early development stages, with excess DAG being converted into TAG for storage, in order to maintain lipid homeostasis and to avoid lipotoxicity (Sheokand et al., 2023). DGATs are key enzymes that esterify DAGs with acyl-CoA to generate TAGs at the ER or on LDs (Yen et al., 2008). Exposing parasites to the DGAT inhibitors XN and T863 resulted in developmental arrests and death of the parasites even in the early stages of the intraerythrocytic cycle, arguing that this phenotype is associated with lipid toxicity due to excess DAGs in the cytoplasm rather than a lack of TAGs in the LD. In addition, disruption of TAG synthesis might also lead to DAG accumulation in the ER, which can cause severe defects in the endomembrane system, as has been shown in in other eukaryotic cells (Li et al., 2020).

Our inhibitor assays clearly demonstrate the functional duality and emphasise the indispensability of LDs in P. falciparum throughout the asexual developmental stages of the parasite (Fig. 8); on the one hand, conversion of DAG into TAG and its storage during times of abundant nutrient supply, and on the other hand, LD depletion along with TAG hydrolysis, which are crucial for parasite survival and highlight the vulnerability of these parasites when either of these metabolic functions is disturbed.

Disrupting the function of LDs has been implicated in many different human diseases, including type 2 diabetes, cancer, and neurological diseases like Alzheimer's and Parkinson's disease (Murphy, 2012). We are only at the start of understanding LD biology in P. falciparum, but their demonstrated importance in the survival of the parasite might open new avenues for therapies.

The P. falciparum parasite lines used for all experiments are 3D7 wild-type parasites (Walter Reed Army Institute for Research, Silver Spring, MD, USA). Asexual blood stage parasites (ABS) were cultured following established procedures (Maier and Rug, 2013). Human type O+ RBCs and human serum provided by the Australian Red Cross blood bank were approved for experimental use by the Australian National University ethics committee agreement 2017/351 and the Australian Red Cross agreement 23-02 ACT-03. The cultures were maintained at 4% haematocrit resuspended in complete culture medium: RPMI 1640-HEPES with Glutamax medium (Thermo Fisher Scientific) supplemented with 10 mM glucose (Sigma-Aldrich), 480 µM hypoxanthine (Sigma-Aldrich), 20 µg/ml gentamicin (Gibco), 2.5% (v/v) human serum (Australian Red Cross Life Blood) and 0.375% (w/v) Albumax II (Thermo Fisher Scientific). The cultures were incubated at 37°C under microaerophilic conditions (1% O₂, 5% CO₂, 94% N₂). In order to achieve synchronised parasite cultures, 5% w/v D-sorbitol treatment was applied 2 days prior to experiments (Lambros and Vanderberg, 1979).

Images were acquired using a confocal microscope (Zeiss LSM800) with a 63× oil immersion objective lens (NA 1.4). Cultured parasites were fixed with 2% paraformaldehyde (EMS) and subsequently stained with Nile Red (Sigma-Aldrich; 4 µg/ml) for NLs and Hoechst 33582 (Sigma-Aldrich; 1 µg/m) for nucleic acid staining, which allowed the different specified stages to be distinguished. Hoechst 33582 fluorescence was detected at 352 nm excitation and 455 nm emission, while Nile Red fluorescence imaging was initially performed with standard settings at 559 nm excitation and 636 nm emission. Throughout each experiment depicted in Fig. S1 and Fig. 2, images were consistently acquired under identical conditions. Z-stack imaging was employed to cover 3D volumes of entire parasites, and the final images are presented as maximum intensity projections. Quantification of Nile Red fluorescence was performed by using ImageJ (https://imagej.net/). The parasite area was selected based on an intensity threshold of Nile Red, and the relevant area was measured by using ‘analyse particles’. The mean fluorescence intensity (MFI) of Nile Red was measured based on maximum intensity projection images. The integrated intensity was calculated by multiplying the MFI by the area occupied by the parasite (Shihan et al., 2021).

To separate various lipid groups, a different excitation and emission regime was applied after Nile Red staining of iRBCs. A Leica Stellaris 8 confocal microscope with detailed spectral separation capabilities was used for this analysis. Nile Red-stained cells were separately imaged with a 63× oil immersion objective lens (NA 1.4) in two channels – one for visualising non-polar head groups like TAGs, with an excitation wavelength setting at 506 nm and capturing emission at 520–590 nm, and another for visualising polar head groups, like PLs, with excitation wavelength of 598 nm and emission range from 610 to 750 nm (Diaz et al., 2008; Greenspan et al., 1985).

Samples for FIB-SEM analysis were prepared using the osmium-thiocarbohydrazide-osmium (OTO) method (Rudlaff et al., 2020; Tapia et al., 2012) with some modification. Trophozoites (∼5% parasitemia) were magnet purified, eluted from the column and then incubated in complete culture medium for 2 h (Ribaut et al., 2008). The magnet-purified cells underwent a brief wash with PBS and were subsequently fixed using 2.5% glutaraldehyde (aqueous glutaraldehyde EM grade, EMS) and 2% paraformaldehyde (aqueous paraformaldehyde EM grade, EMS) in PBS. The fixed parasites were washed with PBS, then post-fixed with 2% osmium tetroxide and 1.5% potassium ferrocyanide in PBS for 1 h. Next, the samples were washed in milliQ water, incubated in thiocarbohydrazide (TCH) for 20 min, followed by a second incubation in 2% osmium tetroxide for 30 min after washing with milliQ water. The heavily fixed cells were incubated overnight in 1% uranyl acetate in milliQ water, then washed in milliQ water with subsequent dehydration using increasing concentrations of ethanol and two washes with 100% acetone in the final dehydration step. Following dehydration, samples were incubated in a gradual concentration of EPON 812 resin in acetone. The cell pellets were polymerised after overnight incubation in 100% resin at 60°C. The resin embedded samples were mounted on SEM stubs using a mixture of super glue and silver paste and trimmed to a small sample surface area (1 mm²). Subsequently, the trimmed block was sectioned with an ultra-microtome (Leica EM UC7) to expose the region of interest (ROI) and the appropriate cell density of the exposed ROI was confirmed with light microscopy prior to the FIB-SEM experiment. Finally, to improve stability under FIB-SEM conditions, the samples were sputter coated with a ∼10 nm thick layer of gold.

A region with high cell density, additionally confirmed via a secondary electron image, was selected for processing in a FIB-SEM (Zeiss Crossbeam 550). Initially, a platinum layer of ∼1 µm thickness was deposited on top of selected regions of interest (ROIs) using the Crossbeam's gas injection system. Subsequently, trapezoid shaped trenches (30 µm×40 µm) were created with an accelerating voltage of 30 kV and a current of 30 nA to expose ROIs for imaging and also prevent redeposition during the milling process. The selected areas were polished with a 30 kV ion beam at a current of 7 nA prior to serial imaging. Finally, the serial milling was set at 30 kV and a current of 700pA with 10 nm slice thickness. Cross sectional images were obtained by detecting backscattered electrons with the energy-selective back-scatter detector (ESB) using an electron accelerating voltage of 1.5 kV, a current of 2 nA and an 8µs dwell time, with a frame resolution of 3072×2304. The FIB-SEM experiment was conducted three times with three biological and technical replicates each. Each raw dataset comprised an average of 800 images with an ∼10 nm³ voxel size.

All processing, visualisation and analysis were conducted using the Microscopy Image Browser (MIB) (Belevich et al., 2016) and Dragonfly 2020.2 [Object Research Systems (ORS) Inc, Montreal, Canada, 2020]. Images of each run were aligned using cross correlation, and denoising was performed with a DNN filter. A Gaussian filter was applied prior to segmentation. The data analysis focussed on single infected cells, encompassing entire cells. A total of 9 trophozoites, 10 early schizonts, 11 mid-schizonts and 3 late schizonts were used for quantification. Life cycle stages were defined as follows: ring stage cells contained a single nucleus (no DV observed), trophozoite stage cells contained a single nucleus (DV observed), early schizont cells displayed 2 to 5 nuclei, mid-schizont stages contained 6 or more nuclei (without fully segmented merozoites), and late schizont cells contained more than 20 nuclei with fully segmented merozoites. Manual or semi-automatic graphcut segmentation in MIB and Dragonfly were used for 3D segmentation and rendering.

To quantify the relative volume of LDs in different life cycle stages, the LD volume per parasite was calculated as the ratio of LD volume over the volume of the whole parasite using segmented models from renderings in MIB and Dragonfly.

Highly synchronised iRBCs at ∼10% parasitemia were cultured using standard protocols (Maier and Rug, 2013). Inhibitors, including 60 µM Orlistat, 40 µM XN and 200 µM T863, as well 0.1% DMSO (control), were added to the cultures at 32 h post invasion. Cells were collected at 44 h (12 h after inhibitor addition) and at 60 h (24 h after inhibitor addition) for TEM analysis. TEM samples were prepared the same way as for the FIB-SEM samples, using the OTO method (Rudlaff et al., 2020; Tapia et al., 2012). However, the resin-embedded samples were sectioned using an ultra-microtome (Leica EM UC7) to obtain 70 nm ultrathin sections for imaging in a JEOL F200 TEM at 200 kV. The images were captured with a Gatan Rio16 camera.

Except where otherwise specified, data were analysed in Prism 8 using one-way ANOVA with Tukey's multiple comparison post test.

The authors acknowledge expert technical support by Daryl Webb, Angus Rae, and Chung-Han Tsai at the Centre for Advanced Microscopy, Australian National University, a facility within the Microscopy Australia (ROR: 042mm0k03) network, enabled by NCRIS and university support.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.262162.reviewer-comments.pdf