Multimodal analysis of Plasmodium knowlesi‐infected erythrocytes reveals large invaginations, swelling of the host cell, and rheological defects

Abstract The simian parasite Plasmodium knowlesi causes severe and fatal malaria infections in humans, but the process of host cell remodelling that underpins the pathology of this zoonotic parasite is only poorly understood. We have used serial block‐face scanning electron microscopy to explore the topography of P. knowlesi‐infected red blood cells (RBCs) at different stages of asexual development. The parasite elaborates large flattened cisternae (Sinton Mulligan's clefts) and tubular vesicles in the host cell cytoplasm, as well as parasitophorous vacuole membrane bulges and blebs, and caveolar structures at the RBC membrane. Large invaginations of host RBC cytoplasm are formed early in development, both from classical cytostomal structures and from larger stabilised pores. Although degradation of haemoglobin is observed in multiple disconnected digestive vacuoles, the persistence of large invaginations during development suggests inefficient consumption of the host cell cytoplasm. The parasite eventually occupies ~40% of the host RBC volume, inducing a 20% increase in volume of the host RBC and an 11% decrease in the surface area to volume ratio, which collectively decreases the ability of the P. knowlesi‐infected RBCs to enter small capillaries of a human erythrocyte microchannel analyser. Ektacytometry reveals a markedly decreased deformability, whereas correlative light microscopy/scanning electron microscopy and python‐based skeleton analysis (Skan) reveal modifications to the surface of infected RBCs that underpin these physical changes. We show that P. knowlesi‐infected RBCs are refractory to treatment with sorbitol lysis but are hypersensitive to hypotonic lysis. The observed physical changes in the host RBCs may underpin the pathology observed in patients infected with P. knowlesi.


| INTRODUCTION
Plasmodium knowlesi is a parasite of long-tailed macaques (Macaca fascicularis) that causes zoonotic infections in humans (Chin, Contacos, Collins, Jeter, & Alpert, 1968). It is the most common cause of malaria in East Malaysia, and the mosquito vectors for P. knowlesi are present throughout Southeast Asia, where an estimated 500 million people are at risk of infection (Barber, Rajahram, Grigg, William, & Anstey, 2017;Shearer et al., 2016;William et al., 2014).
The blood stages of P. knowlesi are responsible for disease pathology, and high parasitemia infections can cause severe malaria in adult humans at a rate similar to that of Plasmodium falciparum (Barber et al., 2013;Cox-Singh & Singh, 2008). Anaemia and thrombocytopenia (low platelet count) are frequently observed, along with impaired renal and liver function (Barber et al., 2011;Daneshvar et al., 2009;Singh & Daneshvar, 2013).
The potential for trapping or sequestration of mature stage P. knowlesi-infected RBCs in sites away from the peripheral circulation is indicated by the reported decrease in the percentage parasitemia in peripheral blood in P. knowlesi infections in rhesus macaques as the parasites mature from ring to schizont stage (Miller, Fremount, & Luse, 1971). The same study showed accumulation of schizont-infected RBCs in the hepatic sinusoids and in the venules of the small intestine.
At higher parasitemia levels, parasitised RBCs were found in other organs, including cerebral capillaries and venules (Miller et al., 1971).
These early studies indicate that occlusion of small capillaries by trapped P. knowlesi-infected RBCs could underpin the pathology of P. knowlesi in human infections; however, the mechanism of trapping remains unclear.
P. knowlesi has been successfully adapted for sustained culture in human RBCs (Gruring et al., 2014;Moon et al., 2013), providing an important model for this emerging zoonotic pathogen. This provides an opportunity to undertake a detailed ultrastructural and biomechanical analysis of the development of P. knowlesi in human RBCs. Using a multimodal approach, we observed a marked accumulation of haemoglobin-containing invaginations, novel deposits at the RBCs membrane, substantial host cell swelling, and altered rheological properties, which may collectively contribute to the pathology of P. knowlesi-infected human RBCs.

| RESULTS
We found that P. knowlesi A1 strain developing in human RBCs in culture exhibits an intraerythrocytic cycle of ∼32-hr duration, in reasonable agreement with a previous report (Moon et al., 2013). Using a magnet purification/invasion/magnet depletion regime (see methods), we synchronised P. knowlesi cultures to windows of 7 hr and collected samples at different stages of development. We analysed samples by conventional thin section transmission electron microscopy (TEM) and serial block-face scanning electron microscopy (SBF-SEM; Denk & Horstmann, 2004;Sakaguchi, Miyazaki, Fujioka, Kaneko, & Murata, 2016; Figures 1-3). We generated SBF-SEM 3D volumes at~50-nm resolution. The cellular features are revealed in individual SBF-SEM "sections," and rendering these features provides a detailed topographic map of cellular organisation. The following features were identified: RBC membrane (RBC; red), parasitophorous vacuole (PV) membrane (PVM, transparent yellow), invaginations (yellow), PV bulges (fawn), nucleus (N, pink), digestive vacuoles (DV; green), Sinton Mulligan's clefts (white), PV blebs (red), lipid bodies (deep blue), and tubular vesicles (TV; cyan). See Videos S1-S12 for translations through the individual sections and rotations of the rendered models (two example cells for each stage are shown).
2.1 | 3D ultrastructure of the ring and trophozoite stages of P. knowlesi In the earliest stages of development (0 to 7-hr postinvasion), the parasite adopts a bowl shape, with invagination of regions of host cell cytoplasm already apparent in some cells . At this stage, Sinton Mulligan's clefts and TV are already evident in the RBC cytoplasm (rendered respectively in white and cyan, Figure 1c,d). The parasite has an average volume of 5 ± 1 fl ( Figure 2b) and a relatively high surface area to volume ratio ( Figure S1f). The organisation of the different structures is best appreciated in Videos S1 and S2. We note that the infected RBC volume is equivalent to that of uninfected RBCs (~68 fl; Figure 2f).
Upon maturation to the early trophozoite stage (7-to 14-hr postinvasion), a more flattened parasite shape is observed, and the invaginations of the host RBC into the parasite become more elabo-    (i-l) 14 to 21-hr postinvasion. Parasitophorous vacuole bulges (PVB) have formed (black arrows) and a pore in (j) indicates active haemoglobin uptake. (k-l) Larger DVs (green) accumulate next to the invaginations. Colours: PVM, pale yellow; nuclei, pink; invagination, yellow; DV, green; RBC, translucent red; clefts, white; TV, cyan. Panels on the far right represent 90 o rotations of the preceding panels (see Videos S1-S6 for translations and rotations of these cells)

| Merozoites undergo morphology changes upon release from the schizont
The schizont merozoites are quite elongated (length 3.6 ± 0.3 μm, width 1.2 ± 0.3 μm) and exhibit an average volume of 3.8 ± 0.2 fl (n = 18 merozoites from two schizonts with ruptured PVM), whereas the free merozoites are shorter (length 3.0 ± 0.2 μm, width 1.5 ± 0.1 μm) with an average volume of 3.4 ± 0.5 fl (n = 11). Individual ultrastructural features of merozoites are difficult to discern at the level of resolution afforded by SBF-SEM. We therefore undertook electron tomography of multiple sections of schizonts and free merozoites (see Figure 4b,e,f for virtual sections from the joined stacks).
Merozoites are characterised by a large elongated nucleus and prominent apical organelles. A pair of flask-shaped rhoptries with the typical neck and bulb features are observed, along with additional electron-dense features (blue-green) that are likely dense granules.
The apicoplast and mitochondrion are closely associated. Stacked membranes (Memb) are evident in the merozoite cytoplasm. The organisation of the different structures is best appreciated in Videos S13 and S14.

| Uptake of the host cell cytoplasm involves cytostomal invaginations and phagotrophs
We surveyed structures involved in the uptake of host cell cytoplasm at different stages of development. As described above, soon after  The DVs fuse into one remnant body at late schizont stage. (e) Invagination size reaches a maximum at trophozoite stage. No haemoglobin uptake occurs in late schizont stage. (f) The infected red blood cell (RBC) volume increases with parasite age. Data represent mean values and standard errors. Unpaired t test; *p < 0.1; **p < 0.01; ****p < 0.0001 diameter of~90 nm and outer diameter of~120 nm. Other (often larger volume) invaginations are connected to the bulk RBC cytoplasm via an opening with a diameter of~400 nm. This is inconsistent with a cytostomal structure and is likely equivalent to the~400 nm openings that are observed in thin sections (Figure 5o,p).
Both the smaller and larger invaginations remain in physical continuum with the host RBC cytoplasm, as indicated by equivalent densities (level of heavy metal binding) of the connected compartments.
Moreover, when the infected RBCs are treated with equinatoxin II (EqtII), a pore-forming toxin that selectively breaches the RBC

| The exomembrane system comprises a range of structural features
We made a detailed examination of the membrane features that P. knowlesi elaborates in the RBC cytoplasm, using SBF-SEM and electron tomography ( Figure 6). 10 to 15 narrow, slit-like Sinton Mulligan's clefts are present from the earliest stages following invasion (clefts;~1-μm diameter; Figure 6a,b rendered in white; Figure S1c).
The clefts are morphologically similar to, but larger than, the~500nm Maurer's clefts of P. falciparum .  Figure S1d). We did not observe extensive networks of TV as has been reported for P. falciparum (Behari & Haldar, 1994)

| The host RBC membrane skeleton is stretched and decorated with deposits
We established a correlative imaging method to probe the cytoplasmic surface of the host RBC. Late trophozoite stage P. knowlesi-infected RBCs (~30% parasitemia) were immobilised onto marked glass slides that had been functionalised with (3-aminopropyl)triethoxysilane and cross linked to the RBC binding lectin, phytohemagglutinin (Shi et al., 2013). The cells were labelled with the nucleic acid-binding dye, Syto61, to identify infected RBCs (Figure 7a), before application of a stream of hypotonic buffer, leaving remnant RBC membranes bound to the glass slide.
The sheared membranes were fixed, dehydrated, and gold coated, and secondary and back-scattered electrons were detected using high-resolution SEM. The spectrin-actin network appears as bright (raised) skeletal elements over dark patches of background (Figure 7 c,d,e,f). Measurement of skeleton parameters from SEM images was

| P. knowlesi-infected RBCs are swollen and rigidified
Our SBF-SEM data suggest that the total volume of the infected RBC increases during intraerythrocytic development of P. knowlesi. Therefore, we sought a complementary method to measure the surface area and volume of live P. knowlesi-infected RBCs taken directly from culture. We made use of a human erythrocyte microchannel analyser (HEMA) microfluidics device (Figure 8a,c), which can trap hundreds of RBCs and conform them into a defined regular geometry, permitting very accurate analysis of RBC surface area and volume (Gifford et al., 2003;Herricks, Antia, & Rathod, 2009;Lelliott et al., 2017).
Using the HEMA, we estimated that the average volume of an uninfected RBC is 88 ± 11 fl, and the average surface area is 147 ± 13 μm 2 (Figure 8b; Figure    uninfected RBCs that were fixed, dehydrated, gold-coated, and imaged using SEM. Caveolae were identified by labelling with PvPHIST-81 antibody followed by AlexaFluor647 staining, as overlaid in (d), appearing as bright electron-scattering puncta. (e,f, insets) Light microscopy images before shearing of the RBCs whose membrane are depicted below. (h,i) High-resolution SEM images of the external surface of (h) uninfected and (i) infected RBCs. (g) Average skeleton network branch distances were estimated using Skan. Representative 2.8 × 1 μm sections of external (whole cell) membranes imaged with backscatter SEM are shown for (h) uninfected and (i) infected RBCs FIGURE 8 Rigidity, volume, and permeability properties of Plasmodium knowlesi and Plasmodium falciparum-infected red blood cells (RBCs). (a,c) Images of the wedge-shaped channels in the human erythrocyte microchannel analyser (HEMA) microdevice with trapped uninfected RBCs (URBCs) and P. knowlesi-infected RBCs (Pk-IRBCs; a) or P. falciparum (Pf-IRBCs; c). (b,d) Mean values for volume, surface area and surface area to volume ratio for URBCs and Pk-IRBCs (b) and Pf-IRBCs (d). Ektacytometry profiles (0-20 Pa shear stress) for uninfected RBCs (squares) and RBCs infected with P. knowlesi (e) and P falciparum (f) trophozoites (97% 22-to 26-hr and 99% 30-40-hr postinvasion, respectively, circles). (g) The average EI ± SEM at 3 Pa (n = 3, minimum parasitemia 88%; 22-32 h and 30-40 h post invasion for P. knowlesi and P. falciparum, respectively). (h,i) RBCs infected with P. knowlesi and P. falciparum trophozoites were subjected (10 min) to increasing concentrations of sorbitol under iso-osmotic conditions (h, n = 3) or buffers with different osmolarities (i, n = 4 and 3, respectively). Survival was measured by flow cytometry following Syto61labelling. Error bars indicate standard error of the mean (s.e.m.) 2.8 | P. knowlesi and P. falciparum-infected RBCs exhibit differentially modified permeability properties P. falciparum induces new permeability pathways (NPPs) in the host RBC membrane that are thought to play important roles in the delivery of nutrients and the efflux of waste (Lew, Tiffert, & Ginsburg, 2003;Nguitragool et al., 2011). Consequently, P. falciparum-infected RBCs are susceptible to swelling and lysis upon exposure to 5% sorbitol (a substrate for the NPP). To further investigate the basis for the swelling of mature stage P. knowlesi-infected RBCs, we examined the susceptibility to lysis upon exposure to solutes that can pass through the NPP. Although P. falciparum-infected RBCs are rapidly lysed, P. knowlesi-infected RBCs withstand resuspension in 5% sorbitol ( Figure 8h), indicating that P. knowlesi lacks NPP or has NPPs with a different substrate specificity to those of P. falciparum.
We also examined the susceptibility of P. knowlesi-infected RBCs to hypo-osmotic lysis. P. falciparum and P. knowlesi-infected RBCs were resuspended in buffers at different osmolarities. P. knowlesi-infected RBCs exhibited a higher sensitivity to hypo-osmotic lysis (Figure 8i), consistent with the observed swelling of mature stage P. knowlesiinfected RBCs.

| DISCUSSION
Here, we present the first detailed ultrastructural examination of P. knowlesi at different stages of development in mature human RBCs.
We examined the cleft-like structures, resembling P. falciparum Maurer's clefts and TV features that wrap around regions of host cell cytoplasm, similar to the P. falciparum TVN. They are assembled shortly after invasion, indicating that their function is required at all stages of intraerythrocytic development. Some of the TV structures are closely associated with the PVM and with invaginations of the RBC cytoplasm into the parasite cytoplasm, suggesting that they may play a role in generating the phagotrophic structures involved in the uptake of host cell cytoplasm. In the mature P. knowlesi trophozoite, structures with an electron-lucent lumen bulge and bleb from the PVM. These structures are reminiscent of structures that have been observed in P. falciparum-infected RBCs (Aikawa, Uni, Andrutis, & Howard, 1986;Atkinson & Aikawa, 1990) and may represent sites where secreted proteins accumulate prior to export (McMillan et al., 2013).
Nuclear division is initiated~22 hr after invasion, producing up to 12 daughter merozoites. As division continues, the nuclei elongate, and apical organelles accumulate at the tip of the daughter cell, which faces towards the cell periphery. The mother cell adopts a crenated aspect as the plasma membrane develops around the individual merozoites, and the DV and other cellular contents contract to a central RB.
Each of the daughter merozoites remains connected to the RB via a thin neck of cytoplasm until the final stages of development. In very mature schizonts, the PVM is breached. This is consistent with studies in P. falciparum suggesting that the PVM is permeabilised prior to rupture of the infected RBC (Hale et al., 2017).

Uptake of host cell haemoglobin occurs soon after invasion
and involves both classical cytostome-mediated endocytosis and an alternative phagocytic mechanism. The cytostomes exhibit an inner diameter of~90 nm and outer diameter of~120 nm, equivalent to that of P. falciparum cytostomes (Abu Bakar, Klonis, Hanssen, Chan, & Tilley, 2010;Lazarus, Schneider, & Taraschi, 2008). In addition, nonclassical pores of~400 nm open into larger invaginations.
Although P. falciparum mainly employs cytostome-dependent uptake of host cell cytoplasm, large cytostome-independent invaginations, called phagotrophs, are occasionally observed in trophozoite stage parasites (Abu Bakar et al., 2010;Elliott et al., 2008;Hanssen et al., 2011;Lazarus et al., 2008). P. knowlesi appears to utilise phagotrophlike structures as an important component of the haemoglobin uptake process. The persistence of very large invaginations from the early ring stage to the early schizont stage suggests that the process for budding of these structures is relatively inefficient in P. knowlesi, potentially underpinning the volume control problems (see below).
The remarkable deformability and durability of the plasma membrane of RBCs is essential for survival in circulation and the ability to navigate small capillaries and splenic sinuses (Mohandas & Gallagher, 2008). We found that the ability of P. knowlesi trophozoite-infected RBCs to elongate in response to a shear force was markedly compromised compared with uninfected RBCs. This is consistent with a recent study using ektacytometry and micropipette aspiration, which revealed increased rigidity of infected and uninfected RBCs in blood samples collected from adult patients with severe knowlesi malaria, reaching levels similar to those observed for patients with severe falciparum malaria (Barber et al., 2018).
Knob deposition provides a major contribution to increased membrane rigidity in P. falciparum-infected RBCs (Glenister, Coppel, Cowman, Mohandas, & Cooke, 2002;Zhang et al., 2015). P. knowlesiinfected RBCs lack knobs, and it is interesting to consider the molecular basis for the decreased deformability. SEM imaging of the cytoplasmic surface of infected RBC membranes revealed proteinaceous deposits. Similarly, SEM of whole cells revealed raised "pimples" that showed a similar distribution to the deposits at the cytoplasmic surface. These protein deposits may strengthen and stiffen the horizontal linkages within the RBC membrane skeleton, which may in turn constrain the flexibility of the skeleton (Dearnley et al., 2016).
The host RBC surface area increases by~11% upon P. knowlesiinfection, and it is interesting to consider the origin of the extra material. SEM of whole trophozoite P. knowlesi-infected RBCs reveals the presence of caveolar pits, whereas electron tomography revealed small vesicles apparently budding from the Sinton-Mulligan's clefts.
Fusion of vesicles with the RBC membrane may deposit sufficient phospholipids to drive the surface area expansion. We observed only a modest (4.5%) increase in membrane skeleton network branch length in P. knowlesi-infected RBC relative to uninfected RBCs. The degree of stretching of the membrane skeleton is less dramatic than that reported for P. falciparum trophozoite-infected RBC, which has been estimated to be from 8% (Nunez-Iglesias et al., 2018) to~50% (Shi et al., 2013). The stretching of the membrane skeleton is insufficient to account for the increased surface area. It is possible that reorganisation of higher-order spectrin oligomers (Nans et al., 2011) could provide material for the observed surface area increase.
Intraerythrocytic malaria parasites face the challenge of expanding their own volume while limiting the swelling of the host cell. In P. falciparum, volume control is complicated by the induction of NPPs that play roles in the delivery of nutrients and in the efflux of waste products (Lew et al., 2003;Martin & Kirk, 2007;Nguitragool et al., 2011;Saliba et al., 2006). P. falciparum tackles the problem of volume control by tightly coupling parasite growth and haemoglobin digestion  and by effluxing amino acids (Lew et al., 2003). P. falciparum schizonts occupy~43% of the uninfected RBC volume , but we found that the P. falciparum-infected RBC volume changes little. This finding agrees with several previous studies of infected RBC volume (Esposito et al., 2010;Hanssen et al., 2012;Herricks et al., 2011), although another study (Waldecker et al., 2017) found that P. falciparuminfected RBCs become more spherical. By contrast, mature P. knowlesi-infected RBCs swell to~120% of the volume of uninfected RBCs. P. knowlesi-infected RBCs are not susceptible to sorbitol lysis, suggesting that they lack NPP-although, further work is needed to confirm this finding. Instead, we suggest that the swelling may be due to inefficient degradation of host cell haemoglobin, as indicated by the persistence of the large invaginations. The baseline swelling of P. knowlesi-infected RBCs may explain their increased sensitivity to hypo-osmotic lysis. Taken together, our data suggest that P. knowlesi lacks efficient volume control mechanisms, which might be exploited by the use of drugs that target osmotic regulators (Dennis, Lehane, Ridgway, Holleran, & Kirk, 2018;Rottmann et al., 2010).
Decreased surface area to volume ratio is a major parameter leading to splenic entrapment (Safeukui et al., 2012). Accordingly, swelling of P. knowlesi-infected RBC could be responsible for sequestration of mature stage P. knowlesi-infected RBCs in small capillaries. Autopsy of a fatal knowlesi malaria case revealed brain capillary congestion (Menezes et al., 2012). Similarly, impaired microcirculatory flow has been observed in P. knowlesi infections of rhesus macaques-unnatural hosts who develop severe and fatal disease (Knisely et al., 1964). A recent study suggested that impaired deformability of both P. knowlesi-infected RBCs and bystander uninfected RBCs may contribute to microvascular accumulation, impaired organ perfusion, and anaemia (Barber et al., 2018). Our data indicate that swelling of infected RBCs may also contribute to impaired rheology.
Mature stage P. knowlesi-infected RBCs were enriched from a culture by magnetic separation (Fivelman et al., 2007). The purified mature stages were resuspended at~60% parasitemia with fresh RBCs and incubated for 7 hr with shaking. Mature stages were further depleted by a second passage through the magnet and an aliquot was taken as the 0 to 7-hr window (~15% parasitemia). Additional aliquots were maintained in culture for a further 7 hr, then magnet purified to generate the 14 to 21-hr window (~60% parasitemia) or returned to culture for a further 7 or 14 hr to generate the 21 to 28-hr and 28-to 35-hr windows.

| Cell sorbitol sensitivity and osmotic fragility analysis
To assess comparative sorbitol sensitivity, P. knowlesi-and P. falciparum-infected RBCs were incubated with iso-osmotic buffers of increasing sorbitol concentrations for 10 min at 37°C. For osmotic fragility analysis, cells were pelleted and resuspended in buffers of decreasing osmolarity, where 300 mOsm is physiological. Cells were centrifuged at 800 g for 90 s, washed, and labelled with Syto-61 (Fu, Tilley, Kenny, & Klonis, 2010). Flow cytometry readouts were corrected for background signal from uninfected RBCs and normalised to the respective infected RBC control, which was taken as 100% survival.

| Preparation of sheared membranes
Coverslips were sequentially treated with (3-Aminopropyl) triethoxysilane, bis(sulfosuccimidyl) suberate crosslinker, and finally incubated with the ligand erythroagglutinating phytohemagglutinin (Shi et al., 2013). Parasites were enriched from culture at the required age of development and immobilised on the functionalised glass slides via a 2-3 min incubation at room temperature (RT). Reference images of cells stained for RNA using 1 μM Syto 61 in phosphate buffered saline (PBS) for 20 min were collected before shearing to facilitate identification of P. knowlesi-infected RBCs. Bound RBCs were allowed to stand for 30 min prior to shearing with~30 ml of 5P8-10 hypotonic buffer (5-mM Na 2 HPO 4 /NaH 2 PO 4 , 10-mM NaCl, pH 8) applied at a glancing angle across the bound cells using a 30 ml syringe with a 21-G needle. The membrane discs were immediately fixed with 2.5% glutaraldehyde for 1.5 hr before dehydration in a series of ethanol: water mixtures and drying in air from 100% ethanol.
Membranes were washed in PBS before incubation with a goat anti-rabbit AlexaFluor647 secondary antibody (1:500) in 3% BSA for 1 hr. To facilitate correlative imaging, the RBC membranes were labelled with mouse anti-α-spectrin (Abcam) and goat anti-mouse AlexaFluor488 (1:500, 1 hr each) in 3% BSA. A final PBS wash was applied before imaging at 100× magnification (widefield fluorescence, DeltaVision Elite). After imaging, this sample was fixed with glutaraldehyde and dehydrated before critical point drying (Leica EM CPD300).
Dried samples were gold coated on the rotating mount of a Dynavac SC100 sputter coating instrument for 35 s using a 25 mA current, measuring~0.2 nm thickness on the quartz crystal microbalance. The coating procedure was optimised to prevent undercoating or overcoating, which compromises skeleton tracing.
SEM images were recorded using the Everhart-Thornley Detector (ETD) detector of an FEI Teneo instrument (in "Optiplan" mode) with a working distance of 5 mm, a beam current of 50 pA and a 2 kV accelerating voltage. For skeleton analysis, multiple images in each of multiple cells were recorded at 200,000-250,000 magnification.

| Microfluidics-based analysis of cellular dimensions
HEMA (Gifford et al., 2003) microfluidic devices were fabricated using established procedures (Lake et al., 2015) at the Melbourne Centre for Nanofabrication and bound to clean coverslips after plasma treatment to ensure tight bonding (Lelliott et al., 2017). P. knowlesi-infected RBCs were introduced into the channels at a flow pressure of 180-250 Pa, and trapped cells were imaged under flow (Gifford et al., 2003). Measurements of cell position within the channels were performed using a custom plugin written in ImageJ (NIH Image/ImageJ). Code written in R was applied to calculate volume, surface area, and surface area to volume ratio (SA:V) from measurements of cell position using the known geometry of the channels. (Plugin and R code are available upon request).

| Serial block-face scanning electron microscopy
The reduced osmium-thiocarbohydrazide-osmium staining method was described previously Parkyn Schneider et al., 2017). In short, parasite pellets were fixed with 2.5% glutaraldehyde in PBS for 1 hr at 4°C. Agarose-embedded cells were stained in ferrocyanide-reduced osmium tetroxide in 0.15 M cacodylate buffer for 1 hr on ice. After washing, the cells were incubated with freshly prepared 1% thiocarbonhydrazide solution in H 2 O for 20 min at RT.
The cells were then further stained with 2% osmium tetroxide in H 2 O for 30 min at RT. Subsequently, the cells were en-bloc stained with 1% uranyl acetate overnight followed by Walton's lead aspartate for 30 min at 60°C. The cells were dehydrated in a graded series of ethanol-H 2 O, followed by progressive infiltration with EPON resin.
After polymerisation, a 200 × 200 × 200 μm resin block was trimmed using an ultramicrotome (Leica EM UC7, Leica Microsystems). The block was mounted on a microtome stub using silver glue and further cleaned by diamond knife after gold coating. The serial images (every 50 nm) were collected using a SBF-SEM, equipped with an in-chamber diamond knife (Teneo VolumeScope, FEI Company), using back scattered electron signals at 3 kV, under low vacuum conditions. The pixel size of each image in the stack was 5 nm.

| SBF-SEM image analysis and volume rendering
Serial sections were contrasted and aligned using IMOD software (Boulder Laboratory for 3D Electron Microscopy of Cells). The regions of interest were segmented and reconstructed into 3D models, using semi-automatic methods described previously (Maiorca et al., 2012).
Briefly, the pixel size of the raw images was binned to 20 nm in the xy direction, then, images were subjected to bandpass filter and Gaussian smoothing before a threshold value was manually selected from each reconstructed tomogram (Parkyn Schneider et al., 2017). Cellular components were rendered as individual objects, which were separated by rough bounding areas drawn manually on 2D sections.
The components were labelled with different colours.
Manual segmentation was used for rendering some features where a more accurate volume and surface area estimate were required. The IMOD slicer tool was used to orient image stacks at the maximum length or width of the object of interest, and the relative distances from the ends were measured using open contours in IMOD.
The correct positioning of the open contours was confirmed on 3D models.

| Transmission electron microscopy and electron tomography
Permeabilization with EqtII was performed as described previously (Jackson et al., 2007). Briefly, cells were prefixed with 2% paraformaldehyde in PBS for 10 min at RT, followed by incubation of 47 ng/μl EqtII for 6 min. The cells were pelleted at 1,400 g for 2 min. For conventional staining, the cells were fixed with 2.5% glutaraldehyde in PBS for 1 hr at 4°C. Cells were preembedded in agarose and fixed in

| Quantification and statistical analysis
Statistical analyses were performed using the GraphPad Prism 7 software package. p values were calculated using unpaired t tests, with details provided in the figure legends. For quantification of the membrane skeleton branch length, the program Skan was applied to unmodified SEM images with a crop range of 20 pixels, offset of 0.075, Gaussian smoothing radius 0.1, and a threshold radius of 5 × 10 −8 . Three separate experiments were performed, analysing a total of 29 uninfected RBCs and 45 infected RBCs. To overlay fluorescence and SEM data, the "landmark correspondences" plugin in FIJI was used to align features between the data sets (minimum of 10 points per image pair) before merging the channels.