Identification of a Stomatin Orthologue in Vacuoles Induced in Human Erythrocytes by Malaria Parasites

When the human malaria parasite Plasmodium falciparum infects erythrocytes, proteins associated with host-derived detergent-resistant membrane (DRM) rafts are selectively recruited into the newly formed vacuole, but parasite proteins that contribute to raft-based vacuole development are unknown. In mammalian cells, DRM-associated integral membrane proteins such as caveolin-1 and flotillin-1 that form oligomers have been linked to the formation of DRM-based invaginations called caveolae. Here we show that the P. falciparum genome does not encode caveolins or flotillins but does contain an orthologue of human band 7 stomatin, a protein known to oligomerize, associate with non-caveolar DRMs and is distantly related to flotillins. Stomatins are members of a large protein family conserved in evolution and P. falciparum (Pf) stomatin appears to be a prokaryotic-like molecule. Evidence is presented that it associates with DRMs and may oligomerize, suggesting that these features are conserved in the stomatin family. Further, Pfstomatin is an integral membrane protein concentrated at the apical end of extracellular parasites, where it co-localizes with invasion-associated rhoptry organelles. A resident rhoptry protein, RhopH2 also resides in DRMs. This provides the first evidence that rhoptries of an apicomplexan parasite contain DRM rafts. Further, when the parasite invades erythrocytes, rhoptry Pfstomatin and RhopH2 are inserted into the newly formed vacuole. Thus, like caveolin-1 and flotillin-1, a stomatin may also associate with non-clathrin coated, DRM-enriched vacuoles. We propose a new model of invasion and vacuole formation involving DRM-based interactions of both host and parasite molecules.

Plasmodium falciparum is an apicomplexan parasite that causes the most virulent form of human malaria. During the blood stages of infection, it invades mature human erythrocytes and develops within a parasitophorous vacuolar membrane (PVM). 1 These host cells are enucleated, have no intracellular organelles and under steady-state conditions have no dedicated machinery for endocytosis of their plasma membrane (1). Yet electron microscopy studies suggest that the nascent vacuole induced by Plasmodium is formed by an invagination of the erythrocyte membrane (2). Although major erythrocyte proteins are excluded from the PVM, proteins that reside in buoyant cholesterol-rich detergent-resistant membrane (DRM) rafts concentrate in the vacuole (3,4). In addition, mild depletion of erythrocyte cholesterol that has no significant effect on major erythrocyte membrane function but disrupts DRM rafts, blocks malarial infection (5). Emerging evidence suggests that DRM rafts play a role in membrane curvature (6 -8). Together these studies imply that PVM formation may involve recruitment of erythrocyte DRM components induced by P. falciparum (9). However, parasite molecules and mechanisms contributing to raft-based vacuole biogenesis remain unknown.
Malarial infection of the erythrocyte is initiated by the extracellular merozoite stage. As the merozoite enters, it appears to release content from apical secretory organelles into the junction of invasion: specifically, contents of organelles called rhoptries are thought to be delivered to the newly forming vacuole. This is seen in both Plasmodium as well as the related apicomplexan parasite Toxoplasma gondii (reviewed in Refs. 10 and 11). In P. falciparum, studies have suggested that the vacuolar membrane is derived mostly from lipids of the apical organelles (12,13). In T. gondii the current model suggests that the newly formed PVM is extensively modified by fusion of rhoptry-derived secretory vesicles (14). Thus, the contents of the rhoptries are thought to play a critical role in formation of the PVM, but direct evidence that rhoptry proteins insert into the new vacuolar bilayer and insight into their membrane properties have been lacking. Further, although homologues of apical proteins that contribute to plasmodial entry are found in T. gondii (reviewed in Ref. 11), none appear to be related to known proteins identified in other eukaryotic or prokaryotic cells. Thus, other than host cell adhesion, it has been difficult to gain insight into the membrane properties of apicomplexan apical organellar proteins, especially those of the rhoptries that may contribute to vacuole biogenesis.
The best-defined, raft-based plasma membrane invaginations in mammalian cells are caveolae (reviewed in Ref. 15). These structures are non-clathrin-coated, implicated in a variety of signaling processes and have been shown to be specialized forms of DRM rafts (reviewed in Ref. 15). A prominent resident DRM raft protein of caveolae is caveolin-1, which is integrally associated with the cytoplasmic face of caveolae (16 -18) and appears to play a role in caveolae lipid organization (19). Overexpression of this protein is sufficient to induce caveolae in cells (20,21), presumably due to its ability to both associate with DRM rafts and oligomerize (22)(23)(24), thereby providing a scaffold to nucleate raft-based invagination. The protein flotillin-1, which is DRM-associated and forms oligomers has also been shown to induce caveolae formation in cells that do not express caveolin-1 (25). Band 7 stomatin is a homologue of flotillin-1 (26), forms large oligomers (27), but is a component of non-caveolar DRM rafts (5,28). (Note: flotillin-1 is also detected in non-caveolar DRM rafts, Ref. 29). Sequence comparisons indicate that band 7 stomatin is a member of a large protein family of both eukaryotic and prokaryotic proteins (26). It acquired the name stomatin because of the apparent association of erythrocyte band 7 protein with stomatocytosis, a disease characterized by increased permeability to monovalent cations in red blood cells (30). However, despite its name, stomatin is not linked to stomatocytosis (31). Further, although multiple stomatin orthologues have been identified and linked to the regulation of channels in Caenorhabditis elegans, the precise function of stomatins is still unknown (32)(33)(34)(35)(36)(37). Recent, cumulative evidence suggests that band 7 stomatin may modulate membrane functions especially those in DRM rafts (reviewed in Ref. 38), but whether DRM association is a conserved property of the stomatin family and whether stomatins reside at the cytoplasmic face of endovacuolar membranes to nucleate rafts remains unknown.
Here we show that the P. falciparum genome encodes an orthologue of band 7 stomatin (designated Pfstomatin), but not of caveolins or flotillins. Pfstomatin is most similar to stomatinlike protein-2 (SLP-2) (39), a prokaryotic-like orthologue of band 7 stomatin that is also expressed in mammalian erythrocytes. Both Pfstomatin and SLP-2, like stomatin, associate with DRM rafts. In the extracellular merozoite stage, Pfstomatin co-localizes with rhoptries. Along with another rhotpry DRM protein it is delivered to the cytoplasmic face of the PVM induced by infection. On the basis of these results we propose a new model of PVM biogenesis where movement of DRM raft proteins from parasite rhoptries as well as DRM-associated host erythrocyte proteins underlie vacuolar development in the erythrocyte.

EXPERIMENTAL PROCEDURES
Culturing and Synchronization of P. falciparum-The P. falciparum 3D7 line was cultured as previously described (40,41). When necessary, parasite growth was synchronized to within a 4-h window by using a 65% Percoll gradient (Amersham Biosciences) and sorbitol (42).
Cloning of Pfstomatin-The 5Ј and 3Ј oligonucleotides ATGAACAC-TTTTTTTTTAAGTAGATTTAG and CTCGAGGTTGTTCATATCAGA-ATGAATTTG were used in a PCR reaction to amplify the complete open reading frame of Pfstomatin (accession number NP_473293) from a cDNA library made from asexual parasites. The PCR reaction was performed using PWO enzyme (Roche Applied Science), and initiated with five cycles of 1 min at 94°C, 1 min at 45°C, and 1 min at 72°C to allow for annealing of highly A/T rich primers; followed by 25 cycles of 1 min at 94°C, 1 min at 61°C, and 1 min 45 s at 72°C. PCR products were purified by agarose gel electrophoresis and cloned into pGEM-T cloning vector (Invitrogen).
Phylogenetic and Sequence Analysis-Blast and PSI-Blast searches against the non-redundant protein sequence database, using Pfstomatin as a query, were used to identify orthologues of Pfstomatin. For phylogenetic analyses, alignments of DNA or protein sequences were performed using ClustalW software (www.embnet.org), followed by manual editing. Using the PAUP program (Sinauer Associates, Sunderland, MA), maximum parsimony with the branch and bound algorithm was used as a tree searching method in Fig. 1D, and distance (minimum evolution) was used in Fig. 1B. The statistical reliability of the branches was established by bootstrapping using 1000 repetitions. The TreeView program (taxonomy.zoology.gla.ac.uk/rod/treeview.html) was used to visualize the unrooted trees. Sequences were checked for mitochondrial targeting signals using MitoPro II (www.mips.biochem. mpg.de/cgi-bin/proj/medgen/mitofilter).
Production and Purification of a Recombinant Fragment of Pfstomatin-To produce a recombinant Pfstomatin C-terminal fragment, the full-length DNA sequence was amplified and directionally cloned into pET29c (to produce a protein that contains an N-terminal S-tag and a C-terminal hexahistidine tag) using 5Ј-end NcoI and 3Ј-end XhoI, and transformed into the BL21 Codon Plus (DE3) RIL cell line (Strategene). For protein expression, the transformed cells were grown in LB plus kanamycin (OD 600 ϭ 0.6). Isopropyl-␤-D-thiogalactopyranoside (IPTG) (Eppendorf) was added to the culture to a final concentration of 1 mM, and cells were incubated for 4 -6 h at 37°C.
Protein purification was performed using the Xpress System (Invitrogen) according to the manufacturer's protocol. Briefly, cells were harvested, lysed in guanidinium lysis buffer (6 M guanidine hydrochloride, 20 mM sodium phosphate, 500 mM sodium chloride, pH 7.8), sonicated, and spun to remove insoluble debris. The supernatant was added to a pre-equilibrated nickel column. The column was washed multiple times with denaturing wash buffers (8 M urea, 20 mM sodium phosphate, 500 mM sodium chloride) at pH 8.0, pH 6.0, and pH 5.3. Recombinant Pfstomatin was eluted from the column using wash buffer at pH 4.0. The Bradford assay was used to measure the protein concentration of all fractions. Colloidal Coomassie staining of a 12% reducing SDS-PAGE gel was used to establish protein purity of peak fractions (Fig.  2Ai). Depending on the preparation, 90 -95% of the total purified protein corresponded to an 18 kDa C-terminal fragment containing the last 161 residues of the 374 amino acid protein, as confirmed by protein sequencing (performed at the Medical College of Wisconsin). The Nterminal residue was methionine 214, which is marked with a blue dot in Fig. 1A, suggesting the presence of an internal initiation of translation start site. The remaining 5-10% of protein corresponded to fulllength Pfstomatin (see 49-kDa band in Fig. 2Ai).
To test whether Pfstomatin forms disulfide-linked oligomers, a preparation of the recombinant protein was subjected to a 12% SDS-PAGE gel in the presence and absence of 100 mM dithiothreiotol. The Western blot was probed with S-protein, which forms a high affinity interaction with the S-tag (Novagen), and only detects the full-length protein (Fig.  2Aii). The Western blot was probed with a peroxidase-conjugated secondary antibody and processed using ECL Western blotting reagents (Amersham Biosciences).
Production of Anti-Pfstomatin Polyclonal Antibody-Polyclonal rat serum was produced against preparations of recombinant Pfstomatin (that contained 95% of the 18-kDa C-terminal fragment and 5% fulllength protein, see above and Fig. 2Ai). Rodents were inoculated four times, at 21-day intervals with 50 g of protein mixed with Ribi MPLϩTDM adjuvant (Ribi Immunochemical). Seven days following the last inoculation rodents were exsanguinated. An IgG fraction was obtained by dilution of sera 1:1 into wash/binding buffer (20 mM sodium phosphate, 150 mM NaCl, pH 7.4) and affinity purification using a protein G-agarose column. After six washes in binding/wash buffer, bound antibody was eluted with 4 ml of 100 mM glycine-HCl, pH 3.0 and neutralized by addition of 1 M Tris-HCl, pH 9.0. Purified rat antibody was used at a dilution of 1:250 in Western blots and at a dilution of 1:400 in indirect immunofluorescence assays.
Fluorescence Microscopy-Indirect immunofluorescence assays were carried out by a modification of procedures described by Lauer et al. (4). Briefly, cells were washed three times in RPMI 1640 and then resuspended at 1 ϫ 10 7 cells/ml, and allowed to adhere to coverslips precoated with 0.1% poly-L-lysine in 5% sodium bicarbonate for 30 min at room temperature. After three 5-min washes in PBS, the cells were fixed in 1% or 2% formaldehyde in PBS for 10 min, permeabilized with 0.05% saponin in PBS or 0.5% Triton X-100 in PBS, and blocked with 0.2% fish skin gelatin in PBS. Cells were probed with the relevant primary antibodies (diluted 1:100 to 1:8000) and secondary antibodies in blocking solution. Mouse monoclonal anti-merozoite surface protein 1 (MSP-1) was supplied by MR4, and mouse monoclonal anti-EXP1 was kindly supplied by Dr. J. McBride. In studies using Mito Tracker (Molecular Probes) cells were stained with 150 nM Mito Tracker for 45 min before fixation. For apicoplast localization, 3D7 cells transfected with pHRPACPGFP were used (43). For all cells, parasite nuclei (blue) were stained by treatment with 10 g/ml of Hoechst 33342 (Molecular Probes) for 10 min.
High resolution image capture and processing was done as previously described (43) with an Olympus IX inverted fluorescent microscope and a Photometrix cooled CCD camera driven by Delta Vision software (Applied Precision Inc., Seattle, WA).
Isolation of DRM Rafts-DRM rafts were isolated as previously de-scribed (4,5). Briefly, vacuolar parasites freed from the red cell membrane or uninfected erythrocytes were extracted on ice in 2 ml of 1% Triton X-100 in TBS (25 mM Tris-HCl, 150 mM NaCl pH 7.4) for 30 min in the presence of protease inhibitor mixture (Roche Applied Science) containing a broad spectrum of serine, cysteine, and metalloprotease inhibitors as well as caspain and 1 mM EDTA. This extract was mixed with an equal volume of 80% sucrose and overlaid with 5 ml of 35% sucrose in TBS and 2 ml of 5% sucrose in TBS. The gradient was subjected to ultracentrifugation at 200,000 ϫ g in a Beckman SW41 rotor for 3 h at 4°C. After centrifugation, the top 1 ml (corresponding to 5% sucrose) was collected as fraction 1, and sequential fractions 2 to 6 were collected as 2-ml aliquots. The Bradford assay was used to measure protein concentration and, where necessary, proteins were concentrated using acetone precipitation. Samples, containing between 5 and 80 g of protein were subjected to SDS-PAGE using 12% gels under reducing conditions. For blotting the following primary antibodies were used: rat polyclonal anti-Pfstomatin, rabbit polyclonal anti-ERD2 (44), rabbit polyclonal anti-SLP-2 (39), rabbit polyclonal anti-RhopH2 (45), and rabbit polyclonal anti-MSP-1 produced against recombinant 19kDa MSP-1 (MR4). Finally, peroxidase-conjugated secondary antibodies were used to probe the blot and membranes were processed using ECL Western blotting reagents (Amersham Biosciences).
Membrane Association of Pfstomatin-Synchronized late stage parasites (5 ϫ 10 7 parasites at 36 -48 h, see above) expressing cytosolic GFP (46) were washed twice in PBS and treated in 0.01% saponin for 10 min to release parasites from the host erythrocyte membrane. Intact parasites were recovered by centrifugation, washed twice in PBS, and lysed in 100 volumes of water in the presence of protease inhibitor mixture at 4°C. These lysates were supplemented with Na 2 CO 3 , NaCl, or water, such that they were at a final concentration of 100 mM Na 2 CO 3 , pH 11.3, or 0.5 M NaCl, or in water alone. Integral membrane proteins do not dissociate from the membrane even after a wash in Na 2 CO 3 pH 11.3, while peripheral membrane proteins are released from the membrane after a wash in Na 2 CO 3 pH 11.3 but not after a wash in NaCl. These samples were subjected to centrifugation at 214,000 ϫ g for 1 h to collect the membrane fractions. Pellets and supernatants were solubilized in SDS sample buffer, subjected to 12% reducing SDS-PAGE, and probed in Western blots with rat anti-Pfstomatin and rabbit polyclonal anti-green fluorescent protein (GFP) (Clontech), followed by peroxidase-conjugated secondary antibodies.
Reconstitution of Recombinant Pfstomatin Fragment into Proteoliposomes-Liposomes were prepared with dioleoyl L-␣-phosphatidylcholine (DOPC) (Sigma) or 1:1 molar ratios of DOPC and cholesterol (Sigma). After dissolving in chloroform and drying in N 2 , the lipids were resuspended in M/N buffer (20 mM Mes, 140 mM NaCl, pH 6.3) (47). This solution was sonicated in a water bath for 30 min at room temperature (temperatures were maintained above 23°C and below 30°C). M/N buffer was added to make a final solution of 2 mol of lipid in 2 ml, and the preparation was subjected to centrifugation at 3000 ϫ g for 10 min. The supernatant (1 ml) containing liposomes was mixed with recombinant Pfstomatin (see above) at 250 g/ml previously dialyzed into M/N buffer, 5 M urea, and 50 mM n-octyl glucoside (Sigma). A total of 10 g of protein was slowly added to the lipid solution during continuous mixing over 10 min. This preparation was mixed with an equal volume of 80% sucrose/M/N buffer and overlaid with 3 ml of 35% sucrose in M/N buffer, 3 ml of 27% sucrose in M/N buffer, and 2 ml of 10% sucrose in M/N buffer. The sample was subjected to centrifugation at 200,000 ϫ g for 4 h. Fractions (1 ml) were collected, subjected to acetone precipitation and processed for a Western blot. The blot was probed using mouse monoclonal anti-histidine antibodies (Oncogene) and peroxidase-conjugated secondary antibody and processed using ECL Western blotting reagents (Amersham Biosciences).
Selective Permeabilization of the Infected Red Cell Membrane Using Tetanolysin-Tetanolysin is a bacterial toxin that can be used to selectively permeabilize cellular plasma membranes (48). Infected red cells (0 -8-h post-invasion) were placed on poly-L-lysine-coated coverslips as described above. After 30 min they were treated with 40 l of 50 units/ml of tetanolysin (List Biological Laboratories) for 30 min at 37°C. Control cells were treated with tetanolysin as well as 40 l of 0.01% saponin in PBS for 10 min at room temperature. Cells were washed twice in PBS for 5 min, blocked in 0.2% fish skin gelatin (Sigma) for 30 min at 37°C, and incubated with the appropriate primary antibodies for 30 min at 37°C. As a control for PVM permeabilization, cells were also probed with antibodies to MSP-1. Cells were washed 10 times for 2 min, fixed in 2% formaldehyde in PBS for 10 min, and quenched in NH 4 Cl for 10 min. Cells were washed three times in PBS for 5 min, permeabilized in 0.05% saponin in PBS for 20 min at room temperature, and incubated with 1:200 dilution of the appropriate secondary anti-bodies for 1 h at room temperature. Finally cell nuclei were stained with Hoechst 33342, and samples were subjected to high resolution fluorescence microscopy.

Identification of a Stomatin Orthologue in the P. falciparum
Genome-A BLASTP search of the P. falciparum genome (49), using human band 7 stomatin as query, identified a single malarial orthologue (accession no. NP_473293). Pfstomatin contains the stomatin signature sequence (39) that is conserved in evolution (outlined by a green box in Fig. 1A), and is recognized as part of the Pfam Band 7/stomatin family with a Evalue of 1.1e Ϫ78 . The phylogenetic analysis in Fig. 1B confirmed that Pfstomatin is indeed a member of the stomatin family. It is distantly related to flotillin, prohibitin, and HflC/K (all these proteins are members of the SFPH superfamily, named after its members stomatin, flotillin, prohibitin, and HflK/C, Ref. 26). No orthologues of caveolin or flotillin and HflK/C were found. However, two putative prohibitin genes (accession nos. NP_704264 and NP_700618) are present in the P. falciparum genome.
Human band 7 stomatin may associate with membranes via a hairpin loop that enters the bilayer but does not span it, so both its N and C termini are exposed at the cytoplasmic face of the membrane (50). The region of the protein inserted into the membrane in band 7 stomatin is outlined by a red box in Fig.  1A. This integral membrane domain is conserved among the close homologues of human band 7 stomatin, such as murine stomatin and C. elegans MEC-2 (51), but is absent in SLP-2 (39). However, Pfstomatin is predicted by TMpred to have an integral membrane domain in a different location than band 7 stomatin, but in the same relative location as its Arabidopsis thaliana orthologue (red box in Fig. 1A). Thus, in principle, Pfstomatin might also form a hairpin loop that enters the bilayer. Pfstomatin does contain an unusually long N-terminal sequence and in this regard is unique (Fig. 1A).
Pfstomatin Is a Close Orthologue of Bacterial and Plant Stomatins-Analysis of the similarity between the full-length amino acid sequences of Pfstomatin, human band 7 stomatin, and orthologues in A. thaliana and Rickettsia conorii suggested that Pfstomatin is more closely related to plant and bacterial proteins than human band 7 stomatin (Fig. 1C). It does however have a close similarity with mammalian SLP-2. Consistent with this, the ClustalW alignment of these sequences in Fig. 1A shows that the four residues where Pfstomatin differs from the stomatin signature sequence are common to all other sequences used in the alignment (marked with asterisks in Fig. 1A).
A phylogenetic tree using the DNA sequence of 15 stomatins further confirmed that Pfstomatin, SLP-2, plant and bacterial stomatins are close orthologues (Fig. 1D). A BLASTP search using Pfstomatin against the non-redundant database reveals more than 100 orthologues with E-values higher then 2e Ϫ29 . These orthologues are present in dozens of species including members of Archae, Eubacteria, and lower and higher eukaryotes including plants, and frequently more than one member of the stomatin family is expressed. The presence of a single stomatin gene in P. falciparum suggests an indispensable function for this protein in this organism.
Development of Pfstomatin-specific Antibodies, Expression of Pfstomatin in Infected Erythrocytes, and Evidence for Oligomerization-Polyclonal antibodies were raised against preparations of recombinant Pfstomatin, which contain 95% of an 18-kDa C-terminal fragment and 5% of the full-length protein as determined by colloidal Coomassie staining (see "Experimental Procedures" and Fig. 2Ai). The N-terminal residues on the 18-kDa fragment were determined by protein sequencing and correspond to methionine 214 (marked with blue dot in Fig.  1A), followed by RYEIRDIILPVNIKNAMEKQAEA. This fragment likely arises due to initiation from an internal methio-nine. Internal initiation is not uncommon during expression of recombinant plasmodial proteins because the high A-T content of the encoding genes results in biased amino acid compositions (52). As indicated earlier (see "Experimental Procedures"), the preparation of Pfstomatin shown in Fig. 2Ai was used for antibody production. Since SLP-2 and band 7 are homologues of Pfstomatin, we considered the possibility that antibody raised to Pfstomatin might cross-react with these red cell homologues. We therefore tested for cross-reaction of anti-Pfstomatin with uninfected red cells (data not shown). Only one of three anti-Pfstomatin sera did not cross-react with SLP-2 or band 7. We presume the specificity of the single antiserum was due to the increased antigenicity of the parasite specific sequences resident in Pfstomatin. Since we were interested in using a reagent that recognized Pfstomatin but not SLP-2 or band 7, the desired antiserum was processed to purify the IgG fraction and then characterized (see "Experimental Procedures"). The purified antiserum recognized a 43-kDa band present in lysates prepared from infected but not uninfected erythrocytes (Fig. 2Aii). This was as expected since predicted molecular mass of Pfstomatin is 43 kDa and the antiserum recognized neither human band 7 stomatin nor human SLP-2. Pre-immune serum did not cross-react with infected or uninfected red cells (data not shown). These data suggest that the antibody specifically recognized Pfstomatin in infected erythrocytes. Furthermore, the antibody also recognized low levels of a 75-kDa protein and larger proteins in infected cells. These bands may reflect low levels of oligomers. Recombinant Pfstomatin can also form homo-oligomers in vitro: oligomerization may be enhanced in the absence of reducing agent (Fig. 2B), similar to the formation of disulfide cross-linked oligomers of SLP-2 that are present in erythrocytes (39).
The intracellular parasite develops through distinct ring (0 -24 h), trophozoite (24 -36 h), and schizont stages (36 -48 h), and multiple parasite proteins show stage-specific expression. We therefore examined the steady-state levels of Pfstomatin expression, at different stages of intraerythrocytic growth. As shown in Fig. 2C, a Western blot normalized for equal number of infected erythrocytes, clearly shows elevation of Pfstomatin at 36 h and 42 h post-infection. This suggests that the trophozoite and schizont stages contain highest levels of this protein per infected erythrocyte; hence these stages were used in subsequent assays.
Association of Pfstomatin and SLP-2 in DRM Rafts-It has been shown that human band 7 stomatin is present in DRM rafts (5,28). However its membrane-spanning domain is not conserved in Pfstomatin or the other stomatins (Fig. 1A). SLP-2 lacks a transmembrane domain and is peripherally associated with membranes (39). Further, because of the involvement of cholesterol, DRM rafts are not expected to be a feature of prokaryotic membranes. However, we were interested to determine whether prokaryotic-like stomatins such as Pfstomatin and SLP-2 that are expressed in eukaryotic cells associate with DRM rafts, to gain insight into the extent of conservation of this membrane-binding property in the stomatin family.
In order to isolate DRM rafts from infected erythrocytes, membrane fractions from purified, late stage schizont-infected erythrocytes were extracted with 1% Triton X-100 in Trisbuffered saline at 4°C, subjected to a sucrose gradient, and then six fractions were collected (see "Experimental Procedures"). Proteins associated with DRM rafts were found in floating fraction 2 (and sometimes in fractions 3 and 4) while non-DRM raft associated proteins were exclusively in the loading zone fractions 5 and 6. As shown in Fig. 3Ai, Pfstomatin was detected in fraction 2 (as well as in 3 and 4). In contrast, PfERD2 a parasite Golgi membrane-associated protein remained in the loading zone, confirming that membrane associated but non-DRM raft-associated proteins do not float on this sucrose gradient (Fig. 3Aii). SLP-2 was detected in DRM-floating fractions prepared from uninfected erythrocytes (Fig. 3B) as was flotillin-2 (a marker for DRM rafts in red cells; data not  3. Pfstomatin and SLP-2 are present in DRM rafts. A, Western blots of fractions isolated from sucrose gradients that were loaded with parasites extracted with 1% Triton X-100 to separate DRM rafts (see "Experimental Procedures") probed with antibodies to Pfstomatin (i) or PfERD2 (ii). Fraction 2 (and in some cases fractions 3 and 4) contain DRM rafts; fractions 5 and 6 contain material remaining in the loading zone. B, Western blot of fractions isolated from a sucrose gradient that was loaded with uninfected erythrocytes extracted with 1% Triton X-100 to separate DRM rafts (see "Experimental Procedures"), probed with antibodies to SLP-2 and human flotillin-2 (not shown).
shown). These results suggest that although Pfstomatin and SLP-2 are prokaryotic-like, and their membrane anchors are distinct from those of band 7, they nonetheless partition into DRM rafts.
Membrane Association of Pfstomatin-Human band 7 stomatin and its close orthologues have an integral membrane domain (as discussed above) (53). However, this is not a property shared by all stomatins, as illustrated by the peripheral association of SLP-2 (39) (see also red boxes in Fig. 1A). In contrast, for Pfstomatin the program TMpred predicts an integral membrane domain with a score of 588 (scores above 500 are significant; the domain is marked with a red box in Fig. 1A). Thus, to test the nature of Pfstomatin's membrane association infected cell lysates were treated with 100 mM Na 2 CO 3 , pH 11.3, 0.5 M NaCl, or water (see "Experimental Procedures"). The lysates were prepared from transgenic parasites expressing soluble green fluorescent protein (GFP), since GFP provides a convenient cytosolic marker (see "Experimental Procedures"). Equivalent amounts of cell membrane and soluble fractions analyzed for Pfstomatin and GFP content by Western blots (Fig. 4A, i and ii, respectively) revealed that greater than 90% of the Pfstomatin signal remains associated with the membrane fraction after carbonate treatment, as expected for an integral membrane protein. In contrast, over 95% of the cytosolic GFP was present in the soluble fraction showing that the cells were disrupted before the various treatments. Hence, these data suggest that endogenous Pfstomatin is an integral membrane protein in infected erythrocytes.
Band 7 stomatin is similar in overall structure to caveolin-1 (38), a recombinant form of which reconstitutes into lipid vesicles in a cholesterol-dependent manner (47): this property is thought to underlie an important feature of DRM raft association in caveolin-1. However, whether the association of stomatins with membrane also requires cholesterol remains unknown. The primary limitation in testing this has been the difficulty in producing recombinant band 7 stomatin. Since we were able to produce a fragment of Pfstomatin as recombinant protein, we investigated membrane association properties of this fragment. To do this, we mixed the histidine-tagged recombinant C-terminal fragment of Pfstomatin with dioleoyl L-␣phosphatidylcholine (DOPC) liposomes in the presence and absence of cholesterol (the N-terminal residue of the fragment is marked by a blue dot above the sequence in Fig. 1A) (see "Experimental Procedures"). The liposome-protein mix was subjected to sucrose gradient centrifugation, in which liposomes float to the low density fractions at the top of the gradient, while unbound protein remains in the high density fraction in the loading zone (see "Experimental Procedures"). As shown in Fig. 4Bi, Western blotting revealed all of the recombinant protein to be at the top of the gradient, in the low density fractions 2 and 3 when liposomes without cholesterol were examined. Pfstomatin also bound quantitatively to liposomes containing a 1:1 M ratio of DOPC:cholesterol (Fig. 4Bii). These results suggest that the C-terminal fragment of Pfstomatin can post translationally associate with membranes. This association is peripheral (data not shown) but not dependent on cholesterol, suggesting that residence in DRM rafts may be determined by other features of Pfstomatin.

Localization of Pfstomatin to Apical Organelles in Schizonts and Merozoites: Detection of the Apical Rhoptry Protein
RhopH2 in DRM Rafts-In silico analyses of the N termini of SLP-2 and Pfstomatin (see "Experimental Procedures") suggested that these sequences may function as mitochondrialtargeting signals. However, SLP-2 resides at the erythrocyte plasma membrane (39), suggesting that for stomatins, the leader sequence alone may not determine cellular location. To determine the location of Pfstomatin relative to mitochondria, infected red cells were co-stained with MitoTracker and anti-Pfstomatin in indirect immunofluorescence assays (Fig. 5A, i-iii, MitoTracker in red and Pfstomatin in green) (see "Experimental Procedures"). There was no overlap between the red and green signal, showing that despite its leader, Pfstomatin does not localize to the mitochondrion. In addition to the mitochondrion, P. falciparum contains a chloroplast-like organelle called the apicoplast (54). In a related apicomplexan parasite T. gondii, there are similarities between the mitochondrial and plastid-targeting signals (55,56). However, as shown in Fig 5A, iv-vi, in infected erythrocytes Pfstomatin failed to co-localize with the GFP targeted to the apicoplast. These data strongly suggest that Pfstomatin does not reside in either the mitochondrion or the apicoplast. Nonetheless, it shows a punctuate distribution at the apical end of the parasite, suggesting an organellar location. Further, there is a high degree of co-localization between Pfstomatin and the resident rhoptry protein RhopH2 in merozoites (present in late schizonts as well as free in the extracellular milieu) (Fig. 5A, vii-ix). Thus, it appears that Pfstomatin is localized at the apical end of the merozoite, likely in the rhoptries.
While there is evidence of host DRM rafts in the PVM of P. falciparum-infected erythrocytes (4), Pfstomatin is the first reported DRM raft protein in parasite apical organelles. Since RhopH2 co-localizes with Pfstomatin, we tested whether RhopH2 also associates with DRM rafts. As shown in Fig.  5Bi, RhopH2 was detected in buoyant DRM fractions 2-4 isolated from schizont-infected erythrocytes. These data argue that rhoptries contain DRM rafts. Moreover, the parasite plasma membrane (PPM) marker MSP-1 is also found in DRM rafts (Fig.  5Bii). This agrees with recent results showing that MSP-1 is present in DRM domains on the merozoite surface (57).
In Newly Formed Rings, Pfstomatin and RhopH2 Are Inserted into the PVM and Exposed on its Cytoplasmic Face-Multiple components of merozoite apical organelles are trans- ferred to the parasitophorous vacuole of rings during or immediately after invasion (reviewed in Ref. 10). Further, release of apical contents is thought to be important in the establishment of the nascent vacuole. To explore whether Pfstomatin undergoes a similar release, we performed an indirect immunofluorescence assay (see "Experimental Procedures"), where early rings were co-stained with antibodies to Pfstomatin and the PVM-marker EXP-1 (Fig. 6A, i-iii). As shown, there is overlap of both markers (yellow) in early rings consistent with redistribution of Pfstomatin to the periphery of rings. As previously shown (45), RhopH2 also redistributes to the periphery of newly formed ring stage parasites (Fig. 6B, i-iii).
Although rhoptry proteins are released into the nascent vacuole (reviewed in Ref. 10), whether they are inserted into the vacuolar bilayer to modify membrane properties has not been firmly established. We were interested in determining whether the peripheral location of RhopH2 and/or Pfstomatin in rings, reflects their delivery to the cytoplasmic face of the vacuole, to provide definitive evidence that these proteins are inserted into the PVM. Therefore, we develop a method to selectively permeabilize the erythrocyte membrane without permeabilizing the PVM. This was achieved by treating ring-infected erythrocytes with the bacterial toxin tetanolysin (48) (see "Experimental Procedures"). Antibodies added to these cells may access proteins exposed on the cytoplasmic leaflet of the PVM but not in the lumen of the vacuole or in the parasite plasma membrane (PPM). In contrast, in the presence of saponin, antibodies can access both the PVM and PPM.
As shown in Fig. 6C, i-iii and D, i-iii, in tetanolysin-treated cells, antibodies to Pfstomatin and RhopH2 display a prominent peripheral distribution of the protein associated with the vacuolar parasite, while those to MSP-1, a major component of the PPM (58), fail to access their antigen. In contrast, when the PVM is permeabilized with saponin, cells can be stained for MSP-1, Pfstomatin, and RhopH2 (Fig. 6E, i-iii and F, i-iii). A, indirect immunofluorescence assays of: (i-iii) free merozoites probed with antibodies to Pfstomatin (i; green), and Mito Tracker (ii; red); the relative lack of co-localization of these markers is shown in the merged panel iii. iv-vi intracellular schizont-stage parasites expressing GFP in the apicoplast were probed with antibodies to Pfstomatin (iv; red), GFP (v; green); the relative lack of co-localization of these markers is shown in the merged panel vi. vii-ix free merozoites probed with antibodies to: Pfstomatin (vii; green), RhopH2 (viii; red); the extent of co-localization is shown in the merged panel ix. All images shown are single optical sections. The nucleus (blue) was stained with Hoechst 33342. Scale bars correspond to 2 m. B, purified schizonts were subjected to DRM analysis and Western blots probed with antibodies to RhopH2 (i), MSP-1 (ii), or PfERD2 (not shown here but note that the membraneassociated protein ERD2 remains in the loading zone as shown in Fig. 3Aii). These data show that parasite DRM proteins from the rhoptries insert into the vacuolar bilayer and proteins like Pfstomatin and RhopH2 are present on the cytoplasmic face of the PVM. Since both Pfstomatin and RhopH2 appear to interact with large protein complexes (this report and Cooper et al., Ref. 59, respectively), their cytoplasmic location suggests that they may provide the site of nucleation for protein scaffolds in DRM rafts in the PVM (see model in Fig. 7). DISCUSSION Although many mammalian cells express multiple stomatins, our studies found that the lower eukaryote P. falciparum contains a single stomatin orthologue. Pfstomatin is prokaryotic-like, but it has a close orthologue in SLP-2 in mammalian cells. The large evolutionary distance between the human paralogues, band 7 stomatin and SLP-2, suggests that stomatins are divided into at least two functional groups and that humans have retained proteins from both. These data, in addition to the localization of Pfstomatin to the apical organelles, suggest that the presence of a prokaryotic-like stomatin in P. falciparum is not a result of the endosymbiotic events that spawned the mitochondrion or the apicoplast, but retention of the molecule through evolution.
Caveolin-1, flotillin-1 and band 7 stomatin share overall similarity in a hairpin structure where the membrane-spanning domain is thought to function as a hook that embeds the protein in the cytoplasmic face of the bilayer without threading through to the extracellular face (38). Previous studies have suggested that loss of the C-terminal domain of band 7 stomatin results in decrease or loss of membrane association (27). In accord, we find that the (non-membrane spanning), C-terminal fragment of Pfstomatin can bind to DOPC liposomes with very high efficiency, possibly due to exposure of hydrophobic residues that facilitate association with the bilayer. This fragment shares a high degree of sequence conservation with SLP-2; thus it is possible that the C-terminal domain of SLP-2 may underlie its membrane association while the predicted membrane-spanning domain of Pfstomatin may confer its integral membrane association with the bilayer. Our results also suggest that the C-terminal domains of Pfstomatin (and by sequence similarity the same region of human SLP-2) may allow post-translational association with lipid bilayers in a cholesterol-independent manner. Finally, since SLP-2 is known to be peripherally associated with erythrocyte plasma membrane, the presence of stomatins in DRM rafts is not invariably dependent on an overt membrane spanning domain or integral membrane proteins within the bilayer. Together the data suggest that there may be multiple mechanisms to confer membrane and DRM raft association of stomatins in a wide range of organisms.
Pfstomatin localizes to the apical organelles of P. falciparum. This location is intriguing because these structures are implicated in the infection of host cells by Plasmodia as well as by the related apicomplexan Toxoplasma gondii (an orthologue of Pfstomatin can be identified in the T. gondii database). Importantly, translocation of an apical Pfstomatin across the newly formed vacuolar membrane shows that this protein may possess a cytoplasmic topology (see model in Fig. 7) equivalent to that of caveolin-1 and flotillin-1 that induce the formation of non-clathrin-coated caveolae (see model in Fig. 7). Both caveolin-1 and flotillin-1 are thought to induce raft nucleation by their ability to associate with DRM rafts and oligomerize, thus supporting larger domains with specialized raft-associated properties. Recently it has been proposed that DRM association reflects lipid-binding properties of proteins and proposed capacity to form lipid shells, which in turn may play an important role in targeting to rafts (8). Our data show that Pfstomatin is in DRMs and may have high affinity to lipids. In addition, Pfstomatin may oligomerize and previous studies have shown that RhopH2 resides in a complex oligomeric state (59) with other parasite proteins. Thus, Pfstomatin, RhopH2 and possibly additional oligomeric rhoptry DRM raft proteins may underlie nucleation and/or stabilization of raft domains in vacuole biogenesis.
In the invasion of apicomplexan parasites, resident proteins of the micronemes are secreted to the parasite surface and mediate one or more steps of attachment to the host plasma membrane (reviewed in Ref. 11). Secretion from the rhoptries (that may be cholesterol-rich, Ref. 60) into the ensuing vacuole is proposed to modify the vacuole, but the membrane characteristics of the secreted material and the kinetics of parasite protein export into the PVM are not well understood. In studies of P. falciparum, Dluzewski et al. (12,13) have argued that the membrane of the newly formed vacuole is derived mostly from apical organelles. In contrast, Ward et al. (61) suggest that it comes from the erythrocyte bilayer. We show here that integral DRM raft-associated Pfstomatin and RhopH2 are translocated across the parasitophorous vacuole and exposed on its cytoplasmic face. Our previous work has demonstrated that erythrocyte DRM raft proteins are recruited into the cholesterol-rich vacuole at the time of parasite entry and that disruption of erythrocyte DRM rafts blocks infection (4). Thus, we propose that the establishment of the plasmodial vacuole may be linked to multiple DRM raft-based interactions of both host and parasite FIG. 7. Proposed model for recruitment of parasite and host DRM rafts during PVM formation. The PVM contains proteins originating from rhoptry DRM rafts as well as from erythrocyte DRM rafts. Moreover, Pfstomatin localizes to the cytoplasmic face of the vacuolar DRM rafts, with a topology consistent for that required for raft-induced plasma membrane invagination in cells. The cytoplasmic location of RhopH2 suggests that additional parasite protein complexes may contribute to raft-based vacuole biogenesis in infected erythrocytes. In addition, DRM rafts are present not only on the PVM but also on the PPM, suggesting that establishment of P. falciparum erythrocyte infection may be linked to multiple DRM rafts based within the PVM as well as possibly between the PPM and PVM. molecules (model in Fig. 7) and this may be applicable to other apicomplexan vacuoles.