The Plastid of Toxoplasma gondii Is Divided by Association with the Centrosomes

The Apicomplexa constitute a large and diverse protozoan phylum of obligate intracellular parasites responsible for many diseases of humans and livestock, including malaria, toxoplasmosis, cryptosporidiosis, and coccidiosis. The apical complex that gives this phylum its name consists of an assemblage of cytoskeletal elements (Nichols and Chiappino 1987; Morrissette et al. 1997) and secretory organelles (Carruthers and Sibley 1997). These structures appear to be synthesized de novo during each cell cycle and are thought to be involved in host cell invasion and establishment of an intracellular parasitophorous vacuole. Most apicomplexans are only capable of replicating within the parasitophorous vacuole, until they emerge to infect a new host cell (Roos et al. 1999a).

Various members of the Apicomplexa have recently been found to harbor another distinctive organelle just apical to the nucleus, the “apicoplast” (McFadden et al. 1996; Köhler et al. 1997; Lang-Unnasch et al. 1998; Zhu et al. 2000). Unlike organelles of the apical complex, the apicoplast contains its own genome, a 35-kb circular DNA that has been sequenced from both Plasmodium falciparum and Toxoplasma gondii (Wilson et al. 1996; Köhler et al. 1997). The apicoplast is essential for parasite survival (Fichera and Roos 1997; McConkey et al. 1997), but the critical function(s) of this organelle are not yet known.

Although well-established phylogenies based on morphological and molecular characters group the Apicomplexa with ciliates and dinoflagellates in the superphylum Alveolata (Gajadhar et al. 1991; Cavalier-Smith 1993), phylogenetic analyses of apicoplast ribosomal genes, elongation factor A (tufA), and organization of the organellar genome reveal a plastid origin (Wilson et al. 1996; Köhler et al. 1997; McFadden et al. 1997; Denny et al. 1998). The algal origins of the apicoplast are attributable to its evolution by secondary endosymbiosis (Palmer and Delwiche 1996): an ancient ancestor of the apicomplexa engulfed a eukaryotic alga and retained the algal plastid (Roos et al. 1999b). This model is consistent with the presence of multiple membranes surrounding the apicoplast (Köhler et al. 1997; Hopkins et al. 1999; McFadden and Roos 1999)—remnants of the endosomal membrane, the algal cell membrane, and the two plastid membranes. Origin of the apicoplast by lateral genetic transfer is also consistent with the targeting of nuclear-encoded proteins to the organelle via the general secretory pathway (Schwartzbach et al. 1998; Roos et al. 1999b; DeRocher et al. 2000; Waller et al. 2000), a mechanism very different from chloroplast targeting in algae and plants (Schatz and Dobberstein 1996).

In this study, we have exploited the ease of genetic manipulation in T. gondii and the availability of in vivo fluorescent markers to study division of the apicoplast and the segregation of this essential organelle and its DNA between daughter cells. Our results indicate that replication of the apicoplast genome and division of this organelle is significantly different from plastid division in plants and algae: apicoplast division is not autonomous (as observed in plants), but is instead inextricably coupled to cell division. Apicomplexan parasites replicate by assembling daughter cells on a complex scaffolding of cytoskeletal elements and flattened membrane vesicles (the “inner membrane complex;” Morrissette et al. 1997). Daughter cell assembly occurs either after every cycle of DNA replication, producing two daughters at a time (endodyogeny; Scholtyseck and Piekarski 1965; Ogino and Yoneda 1966; Sheffield and Melton 1968), or after multiple rounds of DNA replication, producing multiple daughters (schizogony; Frenkel et al. 1970; Müller 1975; Roos et al. 1999a; Speer and Dubey 1999; Waller et al. 2000). These parasites appear to have solved the problem of replication and faithful segregation of a single organelle into two or more daughter parasites by linking apicoplast replication to the machinery already in place for nuclear division: the centrosome/mitotic spindle.

RH strain T. gondii tachyzoites and transgenic lines derived from this strain were maintained by serial passage in primary human foreskin fibroblast cultures (HFF), grown at 5% CO₂ in bicarbonate-buffered Dulbecco's modified Eagle's medium (GIBCO BRL), supplemented with 10% heat inactivated newborn bovine serum (Hyclone) and penicillin/streptomycin/gentamycin (Roos et al. 1994). The culture medium was replaced with modified Eagle's medium containing 1% dialyzed fetal bovine serum (GIBCO BRL) before parasite infection. Dinitroaniline herbicides (synthesized and provided by Dr. J.W. Benbow, Lehigh University, Bethlehem, PA) were stored as 10 mM stocks in DMSO and used at a final concentration of 1 μM.

All fluorescent protein expression plasmids used in this study are related to plasmid ptubP30-GFP/sag chloramphenicol acetyl transferase (CAT) (Striepen et al. 1998). In this construct (based on Bluescript plasmid pKS+; Stratagene), a BglII site (compatible with BamHI) separates the 5′ untranslated region of α-tubulin (Nagel and Boothroyd 1988) from the P30 signal sequence (Burg et al. 1988), an AvrII site separates P30 sequences from the green fluorescent protein (GFP) coding sequence, a PstI site (compatible with NsiI) separates GFP from the 3′ untranslated region of DHFR-TS (Roos 1993), and a NotI site separates this domain from pKS+ and the selectable marker sagCATsag (Kim et al. 1993). Plasmids encoding the yellow fluorescent protein (YFP, a derivative of Aequoria victoria GFP; Miller et al. 1999), cyan fluorescent protein (CFP, a derivative of Aequoria victoria GFP; Miller et al. 1999), and red fluorescent protein (RFP; from Discosoma sp.; Matz et al. 1999) were obtained from CLONTECH Laboratories, Inc., and used as templates to amplify the respective coding regions.

Reporters targeted to the apicoplast were engineered as in-frame COOH-terminal fluorescent protein fusions, using PCR primers to introduce a 5′ AvrII site immediately upstream of the ATG initiation and a 3′ NsiI site immediately downstream of the stop codon (CFP/YFP sense primer 5′-acgtCCTAGGatggtgagcaagggcgaggagc-3′, antisense primer 5′-cagtATGCATtacttgtacagctcgtccatgccg-3′; RFP sense primer: 5′-agctCCTAGGatggtgcgctcctccaagaacg-3′, antisense primer 5′-gactATGCATctacaggaacaggtggtggcgg-3′). Amplification products were cloned as AvrII/PstI fragments in place of GFP in vector pdhfrCAT-GFP (Striepen et al. 1998), excised together with dhfr 3′ UTR using AvrII/NotI, and introduced in place of GFP in apicoplast targeting vectors tubulin acyl carrier protein (ACP)-GFP/sagCAT (Waller et al. 1998) or tubFNR-YFP/sagCAT (Vollmer et al. 2000). A parasite line stably expressing plasmid pgraPCNA-GFP (exploiting the GRA1 promoter) was provided by Drs. M. Guerini and M.W. White (Montana State University, Bozeman, MT; Guerini et al. 2000).

β-Tubulin proteins were engineered to contain a COOH-terminal fluorescent domain using primers 5′-actgAGATCTaaaatgagagaaatcgtccacgttc-3′ and 5′-cagtCCTAGGcgcgccttcctctgcaccc-3′, and ligating the resulting β-tubulin PCR product in place of P30 sequences in the plasmids described above (BglII/AvrII). Tubulin proteins with a fluorescent NH₂-terminal domain were constructed by amplifying α- or β-tubulin sequences (Nagel and Boothroyd 1988) from cDNA clones (α-tubulin sense primer 5′-atcgCCTAGGatgagagaggttatcagcatcc-3′, antisense primer 5′-gcatCTGCAGttagtactcgtcaccatagccc-3′; β-tubulin sense primer 5′-actgCCTAGGatgagagaatcgtccacgttcag-3′, antisense primer 5′-gcatCTGCAGttagtactcgtcaccatagccc-3′), introducing these AvrII/PstI fragments in place of GFP in plasmid ptubP30GFP, and replacing P30 sequences with YFP (sense primer 5′-acgtcAGATCTaaaatggtgagcaagggcgaggagc-3′, antisense primer 5′-cagtCCTAGGcttgtacagctcgtccatgccg-3′).

Parasite transfection was performed by electroporation, as previously described (Roos et al. 1994). Briefly, 10⁷ purified parasites were resuspended in cytomix [(mM) 120 KCl, 0.15 CaCl₂, 10 K₂HPO₄/KH₂PO₄, pH 7.6, 25 Hepes, pH 7.6, 2 EDTA, 5 MgCl₂, 2 ATP, and 5 glutathione] containing 50 μg sterilized plasmid DNA, and electroporated in a 2-mm gap cuvette with a 1.5 kEV pulse at a resistance setting of 24 Ω using a BTX 600 electroporation system. Electroporated parasites were used to infect HFF cells grown on coverslips (for microscopy) or in T25 flasks (for drug selection). Stable transformants were selected in 20 μM chloramphenicol (Kim et al. 1993); the relatively low level of resistance provided by CAT in T. gondii parasites selects for multi-copy plasmid integration (Striepen et al. 1998), enhancing expression levels.

For microscopy, HFF cells were grown to confluence on sterilized coverslips in six-well plates. Cultures were infected with 5 × 10⁵ parasites and examined 16–36 h after infection. For imaging of native fluorescent proteins, coverslips were mounted in PBS or medium without further treatment. GFP-expressing parasites were imaged using an Axiovert microscope (Carl Zeiss, Inc.) and a FITC filter set (450–480-nm excitation/515–565-nm emission). RFP was detected using a Texas red filter set (530–585-nm excitation/615-long pass emission). To record registration-free images of YFP/CFP expression, we used a single emission filter (505–555 nm; Chroma) and dichroic mirror for both fluors, introducing specific filters for CFP (399–429 nm) or YFP (480–495 nm) into the excitation path on a slider. Images were collected using an interline chip cooled CCD camera (Orca 9545; Hamamatsu).

For time-lapse video microscopy, cells were grown in 3ΔT dishes in which the culture surface consists of a #1 glass coverslip (Bioptechs). Monolayers were infected with transgenic parasites and cultured for 24 h before imaging. During imaging, the standard infection medium was supplemented with Hepes, pH 7.6 (GIBCO BRL) to a final concentration of 10 mM, permitting culture at ambient [CO₂]. The temperature of the dish and the 100× oil immersion lens was maintained at 37°C for the duration of the experiment using stage and objective heaters (Bioptechs). Electronic shutters (Ludl) were introduced into both light paths, permitting illumination only during imaging, to reduce photobleaching. Shutter and camera control, image contrast, and image overlays were performed using Openlab software (Improvision) and Photoshop (Adobe Systems, Inc.).

For immunofluorescence, parasite-infected cells were fixed in 3% paraformaldehyde and permeabilized with 0.25% Triton X-100 in PBS. Anti–GFP antibodies (Clontech) were used at 1:500 (rabbit polyclonal) or 1:250 (mouse monoclonal), and detected using appropriate FITC-conjugated anti–immunoglobulin (1:200; Sigma-Aldrich). Anti–centrin antibody 26-14.1 (provided by Dr. J.L. Salisbury, Mayo Clinic, Rochester, MN) was used at 1:200 to label centrosomes, and anti–NET1 mAb 45.15 (provided by Dr. G.E. Ward, University of Vermont, Burlington, VT) was used at 1:1,000 to label the T. gondii inner membrane complex. Appropriate rhodamine-conjugated anti–immunoglobulins (1:200; Cappel) were used in double-labeling assays. Cells were mounted on slides using Gelmount (Fisher Scientific).

To stain plastid DNA, cells were fixed, permeabilized, incubated 10 min in 1 μg/ml DAPI (Molecular Probes) in PBS, washed to remove unbound dye, and mounted. Stained preparations were viewed on an Axiovert or a DMIRBE microscope (Leica) using Openlab software or an Olympus BX50 confocal system running Fluoview software (Olympus). To remove background haze and reveal greater detail in parasites double-labeled with DAPI and anti-GFP, serial 0.25-μm optical sections were collected and the resulting image stacks were deconvolved using a multiple neighbor algorithm. Deconvolved images were then used to reconstruct images depicting the entire cell using Openlab software.

Infected cells were fixed directly in the culture dish by replacing the culture medium with a freshly prepared solution of 1% glutaraldehyde (8% stock; Electron Microscope Sciences) and 1% OsO₄ in 50 mM phosphate buffer at pH 6.3. Samples were fixed 45 min on ice, rinsed in distilled water to remove excess phosphate, stained overnight in 0.5% uranyl acetate, dehydrated, and embedded in Epon directly in the culture dish. Thin sections were cut using a diamond knife, mounted on uncoated grids, and stained with uranyl acetate and lead citrate before examination with a Phillips 200 electron microscope.

A quicktime movie of Fig. 2,Fig. 2 (below). Images of GFP-labeled plastids were taken by fluorescence microscopy (green) and merged with phase-contrast images depicting the outline of the parasite and host cell at the same time point. Additional movie and image files are available at http://webs.cb.uga.edu/~striepen. is available. Infected cells were time lapse imaged as detailed in

We have previously shown that fusion of GFP to a nuclear-encoded apicoplast protein (the acyl carrier protein of a type II fatty acyl synthase) permits direct visualization of the apicoplast in living parasites (Waller et al. 1998). Most parasite cells contain a single plastid (Fig. 1 A, arrow), confirming earlier work where the plastid genome was detected by in situ hybridization (Köhler et al. 1997). At low frequency, however, we encountered relatively large parasites harboring two apicoplasts (Fig. 1 A, double-headed arrow). We also noted that the shape of the apicoplast can be quite variable: in addition to the ellipsoid organelle described in electron microscopic studies, dumbbell and U-shaped plastids were also observed (Fig. 1 B). These distorted apicoplasts can be more than three times longer than the ellipsoid form (up to 3 μm from end to end), and they appear synchronously in every parasite within a given parasitophorous vacuole, suggesting a link to the ∼8-h cell cycle (which is synchronous for all T. gondii parasites within the vacuole; Fichera et al. 1995).

To evaluate whether the U-shaped and dumbbell forms of the apicoplast might be intermediates in plastid division, we performed time-lapse video imaging on a clonal line stably expressing ACP-GFP (see Materials and Methods for further details). Fig. 2,t = 0 are all siblings within a single vacuole, and each harbors an elongated apicoplast (contrast with the ovoid apicoplasts in the parasites within a separate vacuole; Fig. 2, top right). 2 min into this time series, all four apicoplasts are bent in half, with their ends pointing towards the apical end of the parasite. 2-min later (4' image), these apicoplasts have clearly divided into two daughter organelles within a single mother cell. 10 min after plastid division (14' image), the parasites initiate cytokinesis, resulting in the formation of eight daughter parasites (clearly visible by 22'). Plastid division as well as cytokinesis was synchronized in all four parasites of the vacuole. Division of the apicoplast consistently took place just before the emergence of the two daughter parasites, arguing that plastid replication might be linked to parasite cell division. Furthermore, it appears that the ends of the organelle are not free but pulled by some cellular structure resulting in the division of the organelle. shows a representative series of images, additional images and movies are available. The four parasites in the center of this image at

The apicoplast genome is a 35 kb DNA circle (Wilson et al. 1996; see also http://www.sas.upenn.edu/~jkissing/ toxomap.html), present at approximately six copies per cell in T. gondii tachyzoites (Fichera and Roos 1997). As shown in Fig. 3 A, this genome (in addition to the nuclear DNA) can be visualized in fixed and permeabilized cells by staining with intercalating dyes such as DAPI or Hoechst 33258 (Köhler et al. 1997). Staining of apicoplast DNA reveals a spot smaller than the organelle itself, however, suggesting concentration of the apicoplast genome as a intra-organellar nucleoid (Fig. 3 A). This finding is reminiscent of multicopy chloroplast genomes, which associate with as yet uncharacterized proteins to form multiple nucleoids (Kuroiwa et al. 1998; Cannon et al. 1999). Some parasites exhibit two closely associated apicoplast nucleoids of equal intensity (Fig. 3 B), and double labeling of the plastid genome (with DAPI) and it's lumen (with anti–GFP antibody) reveals that the two closely associated DAPI-labeled spots are invariably associated with elongated apicoplasts, and always localize to the ends of the organelle (Fig. 3, C–H).

To more precisely characterize the timing and context of plastid division, we labeled both the nucleus and plastid in living parasites, taking advantage of a proliferating cell nuclear antigen 1 (PCNA)–GFP fusion protein to label the T. gondii nucleus (Guerini et al. 2000). For these experiments, the plastid was labeled using a RFP reporter (a native fluorescent molecule that is entirely distinct from GFP; Matz et al. 1999) fused to Ferredoxin/NADPH Reductase (FNR; Vollmer et al. 2000). The FNR cDNA sequence predicts a bipartite apicoplast-targeting domain (Roos et al. 1999b), and FNR-RFP shows identical labeling to ACP-GFP in double transfection experiments (not shown). A stable transgenic parasite line expressing PCNA-GFP was transiently transfected using plasmid ptubFNR-RFP/sagCAT. Living parasites were imaged using GFP and RFP filter sets, and the images were collected, colored, and merged.

As shown in Fig. 4 A, 90% of all parasites in the transfected population exhibit a single central nucleus with a well-defined nucleolus, as is typical for G1 and S-phase T. gondii tachyzoites (M.W. White, unpublished observation). These parasites invariably contain a single ellipsoid plastid just apical of the nucleus. Nuclear division in T. gondii, which takes place without breakdown of the nuclear envelope, is preceded by migration of the nucleus moves to the basal end (Fig. 4 B, bottom half). Next, the nucleus forms two lobes that extend towards the apical end (Fig. 4 B, top half). Ultimately (over the course of ∼20 min), two fully separated daughter nuclei appear (Fig. 4 C; movies and time-lapse images of this process are available at http://webs.cb.uga.edu/~striepen/pcnatl2min.html). Division of the apicoplast occurs in close synchrony with nuclear division in these double-labeled parasites: U- and dumbbell-shaped plastids form slightly before the appearance of lobulated nuclei, and daughter apicoplasts separate before division of the nucleus (Fig. 4 B). The apicoplast remains closely associated with the apical end of the nucleus throughout this process (Fig. 4B and Fig. C).

As noted above, apicomplexan parasites replicate by assembling daughter parasites upon an inner membrane complex within the mother (Scholtyseck and Piekarski 1965; Ogino and Yoneda 1966; Sheffield and Melton 1968). T. gondii tachyzoites divide by endodyogeny, in which two daughters are formed within a single mother. This is a simpler, but fundamentally identical, version of schizogony, in which multiple daughters are assembled simultaneously (e.g., in the formation of Plasmodium merozoites, Eimeria sporozoites, and feline intestinal epithelial forms of Toxoplasma; Frenkel et al. 1970; Müller 1975; Roos et al. 1999a; Speer and Dubey 1999; Waller et al. 2000). We have used several probes to study the relationship between apicoplast division and the developing daughter parasites.

T. gondii parasites expressing ACP-GFP were fixed, permeabilized, and incubated with anti–GFP antibody to label the plastid (Fig. 5B and Fig. E, green) and a monoclonal antibody 45.15 specific for NET1 (also known as IMC1), a putative intermediate filament protein associated with the inner membrane complex (Fig. 5A and Fig. D, red). The ends of dividing plastids are always inserted into the forming daughters and appear to be attached to one side of the structure (Fig. 5C and Fig. F). These observations confirm that the characteristic folding of the apicoplast observed during replication (Fig. 1 and Fig. 2) coincides with bending around the edge of the forming daughter pellicle.

Similar studies were carried out in living parasites by fusing fluorescent proteins to the T. gondii α- and β-tubulin genes (Nagel and Boothroyd 1988), as microtubules form part of both the parasite's pellicle and the mitotic spindle (Nichols and Chiappino 1987; Morrissette et al. 1997). Both NH₂- and COOH-terminal fusions of YFP with β-tubulin showed general cytoplasmic staining without incorporation into any specific structure, presumably due to steric limitations imposed by the incorporation of a bulky GFP tag into microtubules (data not shown). In contrast, YFP–α-tubulin labeled the subpellicular microtubules of both the mother and developing daughter parasites (Fig. 5G and Fig. J). We also observed a spot of intense staining at the very apical end of the parasite (Fig. 5 J, arrowhead), probably attributable to either the conoid or the apical polar rings (Nichols and Chiappino 1987; Morrissette et al. 1997). An additional focus of intense labeling in developing daughter parasites was observed in the nuclear region (Fig. 5 G, double arrow). Double labeling with centrosome- and microtubule-specific antibodies (not shown) reveal that this label represents ends of the intranuclear mitotic spindle.

For in vivo double labeling, a parasite line stably expressing ACP-CFP was transiently transfected with a ptubYFP–α-tubulin. Overlays of α-tubulin and plastid markers (Fig. 5I and Fig. L) show that the ends of the (U- or dumbbell-shaped) dividing apicoplast are inserted into the developing daughter parasites, and once again indicate that the ends of the plastid are always in contact with the daughter's pellicle, immediately adjacent to the intranuclear spindle. Using the bright apical dot (conoid?) as a marker for the extreme apex of developing parasites, the distance between the point of plastid attachment and the apical complex is ∼1 μm (SD ± 0.7 μm; n = 100). Time-lapse movies (not shown) confirm the spatial relationship between the apical complex, the intranuclear spindle, and the plastid attachment site on the inner membrane complex throughout endodyogeny.

To further explore the hypothesis that the ends of the dividing plastid are associated with the mitotic cytoskeleton, ACP-GFP transgenic parasites were stained with polyclonal anticentrin antiserum 26-14.1 (Shaw et al. 2000). Centrins are well-characterized calcium binding proteins found throughout the cytoplasm, but concentrated in the matrix surrounding centrioles (Salisbury 1995). Parasites expressing ACP-GFP were fixed, permeabilized and doubly labeled for GFP (Fig. 6, green) and centrin (red). The anticentrin antiserum always labeled just beyond the ends of the dividing apicoplast, whether dumbbell shaped (Fig. 6 B), U shaped (C), or even after plastid fission (A, arrows). Indeed, images of nondividing parasites (Fig. 6 D) suggest that a close connection between the (nondividing, ovoid) apicoplast and centrosome persists throughout the parasite cell cycle. Intracellular parasites within a single parasitophorous vacuole typically orient with their apical ends pointed outwards; the fact that the ring of green dots in such images lies outside the ring of red dots indicates a more apical localization of the apicoplast during G1 and S phases.

Electron microscopic observation confirms the apicoplast/spindle/centrosome/inner membrane complex association, as shown in Fig. 7. Apicoplasts (P) are readily identified by their multiple membranes (Fig. 7B and Fig. C) in each of the four daughter parasites being assembled within these two T. gondii tachyzoites. In each case, the apicoplast is closely apposed to the inner membrane complex, and immediately posterior to a centriole (white arrows). The intranuclear spindle microtubules are visible in two of the daughter parasites (black arrows), forming conical projections from the nucleus directly adjacent to the apicoplast.

Microtubule organization in T. gondii is sensitive to dinitroaniline herbicides (Stokkermans et al. 1996), which specifically target α-tubulin (Anthony et al. 1999; D.S. Roos, unpublished observation). Treatment of parasites with these drugs does not completely abolish microtubule formation, but affects different microtubule subsets differentially (a remarkable observation given that T. gondii contains only a single gene each for α-, β- and γ-tubulin; Nagel and Boothroyd 1988). Oryzalin treatment completely blocks assembly of subpellicular microtubules, preventing inner membrane complex development and hence daughter parasite assembly; ethalfluralin exhibits a similar but less dramatic effect (Stokkermans et al. 1996; Morrissette and Roos 1998). The formation of intranuclear spindles seems not abolished by dinitroanilines, however (Shaw et al. 2000). After prolonged treatment, parasites resemble “artificial schizonts,” with a swollen lobed nucleus incapable of completing division, and multiple centrosomes and spindles present (Morrissette and Roos 1998; Shaw et al. 2000).

To test the effect of dinitroanilines on apicoplast division, cultures were inoculated with T. gondii tachyzoites stably expressing ACP-GFP, incubated 24 h (by which point most vacuoles contained four or eight parasites), treated with 1 μM oryzalin or ethalfluralin, and examined 20 h later. In the presence of dinitroanilines, parasites form large amorphous bodies, as shown in Fig. 8 A (compare with Fig. 1). Many parasites contain multiple apicoplasts (Fig. 6 A) or large reticulate plastids (Fig. 6B and Fig. D), strikingly similar to the pattern observed in P. falciparum schizonts (Waller et al. 2000). Double labeling with anticentrin antibody reveals the presence of multiple centrosomes, closely associated with the reticulate apicoplast (Fig. 6C and Fig. E).

The apicomplexan plastid is a remarkable organelle, acquired by secondary endosymbiosis from an algal ancestor and maintained throughout the aeons separating Plasmodium from Toxoplasma with little genetic change. Although the apicoplast has lost all photosynthetic function, it is nevertheless an essential organelle (Fichera and Roos 1997; McConkey et al. 1997). Parasites lacking plastid function die upon entry into the next host cell (Fichera et al. 1995; Fichera and Roos 1997; C.Y. He, unpublished observation). Several groups are seeking to identify the critical metabolic function(s) provided by the apicoplast, and current evidence indicates a possible role in the synthesis of both fatty acids and isoprenoids (Waller et al. 1998; Jomaa et al. 1999; McFadden and Roos 1999). Both T. gondii and P. falciparum harbor a single apicoplast (Fig. 1; Köhler et al. 1997; Waller et al. 2000). How is this essential organelle efficiently replicated and segregated during cell division?

Perhaps the best-studied example of plastid replication occurs in the mesophyll cells of Arabidopsis thaliana. Each cell contains numerous chloroplasts, which divide autonomously many times during the growth phase of a mesophyll cell (Pyke 1999). Plastid replication is independent of mitosis in this system. In contrast, simultaneous visualization of plastid division, nuclear division, daughter cell formation, and cytokinesis (Fig. 2, Fig. 4, and Fig. 5) demonstrate that these processes in T. gondii are intimately coupled (Fig. 9 summarizes our current model of apicoplast replication).

Apicoplast division is not only linked to mitosis in T. gondii, but even appears to use the mitotic machinery. The ends of the dividing apicoplast are invariably associated with the centrosomes (Fig. 6). These findings are supported by electron microscopic evidence (Fig. 7; Ogino and Yoneda 1966; van der Zypen and Piekarski 1967; Sheffield and Melton 1968; Speer and Dubey 1999). Centrosome association could explain how precise plastid distribution is ensured even in large schizonts, where hundreds of progeny must be assembled at once. It is tempting to speculate that the entire daughter cell is built around the centriole.

Chloroplast division in plants and algae is thought to require a fission ring (Kuroiwa et al. 1998; Pyke 1999) containing the tubulin homologue FtsZ known to be associated with bacterial cell division. The A. thaliana genome encodes at least two genes exhibiting strong similarity to FtsZ, and one of these genes carries a putative plastid leader sequence (Lutkenhaus and Addinall 1997; Osteryoung et al. 1998). One model suggests that a combination of stromal and cytoplasmic FtsZ rings are required for chloroplast division (Kuroiwa et al. 1998; Osteryoung et al. 1998). An algal mitochondrial FtsZ has also been identified, suggesting a conserved mechanism of bacterial origin for plastid and mitochondrial division in algae (Beech et al. 2000). To date, however, we have not been able to identify a candidate FtsZ gene from the P. falciparum genome database (Gardner 1999), and have not been able to amplify any FtsZ-related sequences from either T. gondii or P. falciparum DNA using degenerate PCR primers based on conserved FtsZ domains. Furthermore, targeting of the A. thaliana plastid-targeted FtsZ1-1 to the T. gondii apicoplast (as a GFP fusion) produced no distinctive suborganellar localization and failed to affect apicoplast division (M.J. Crawford and D.S. Roos, unpublished observations). As electron microscopic studies show no evidence for a fission ring in apicomplexan parasites, it is possible that these parasites are able to divide the apicoplast through a combination of force generated by the mitotic spindle and growth of the daughter pellicle.

In higher plants, plastid genome copy number may reach >100 per organelle, and this DNA is concentrated in multiple nuceloids (Kuroiwa et al. 1998). Current models favor partitioning via attachment to the plastid envelope, although the recently discovered complexity of bacterial nucleoid division suggests that this might be an oversimplified view (Kuroiwa et al. 1998; Jensen and Shapiro 1999). We have shown that the apicoplast genome is organized in a single nucleoid body, and divides early during organellar replication (Fig. 3 and Fig. 9), via association with the ends of the organelle. It is not clear whether the nucleoid (located within the four membranes of the apicoplast) and the centrosome (outside this organelle) identify the ends of the apicoplast independently, or if they are physically linked. Whether DNA segregation to the ends of the apicoplast is directly dependent on centrosome association is currently under investigation, as such an association would provide a simple explanation for the problem of precise DNA segregation, particularly in the large reticulate plastid of schizonts. Association of organellar DNA with the cellular division machinery is not without precedent: segregation of the mitochondrial genome in Trypansoma brucei has been shown to be associated with basal body movements (Robinson and Gull 1991).

How has this mechanism of plastid division, very different from that found in plants, evolved? One attractive model speculates that it may be a consequence of the endosymbiotic origins of the apicoplast. One presumes that the proto-apicoplast was an alga that took up residence within the parasite's endosomal compartment. Because sorting endosomes are tightly associated with centrioles (Ren et al. 1998), the machinery that couples the sorting endosome to the centrosome could have been harnessed for apicoplast replication. Functional compartmentalization of the apicoplast within the endomembrane system is consistent with observations that proteins target to the apicoplast via the secretory pathway (Roos et al. 1999b; DeRocher et al. 2000; Waller et al. 2000).

We thank Drs. H.L. Compton for help with confocal microscopy, J.W. Benbow for dinitroaniline synthesis, J.L. Salisbury and G.E. Ward for antibodies, and M. Guerini and M.W. White for the PCNA1-GFP parasite line. Cynthia He, Hu Ke, and Qinghao Xu provided helpful discussions, and Drs. Kojo Mensa-Wilmot and Mark Farmer provided critical reading of the manuscript.

This work was supported by grants from the National Institutes of Health, a Burroughs Wellcome Scholar in Molecular Parasitology award to D.S. Roos, a fellowship from Deutsche Forschungsgemeinschaft, and support from Merck Research Laboratories to B. Striepen.