Toxoplasma gondii ADP-ribosylation Factor 1 Mediates Enhanced Release of Constitutively Secreted Dense Granule Proteins*

Toxoplasma gondii dense granules are morphologically similar to dense matrix granules in specialized secretory cells, yet are secreted in a constitutive, calcium-independent fashion. We previously demonstrated that secretion of dense granule proteins in permeabilized parasites was augmented by the non-hydrolyzable GTP analogue guanosine 5′-3-O-(thio)triphosphate (GTPγS) (Chaturvedi, S., Qi, H., Coleman, D. L., Hanson, P., Rodriguez, A., and Joiner, K. A. (1998) J. Biol. Chem. 274, 2424–2431). As now demonstrated by pharmacological and electron microscopic approaches, GTPγS enhanced release of dense granule proteins in the permeabilized cell system. To investigate the role of ADP-ribosylation factor 1 (ARF1) in this process, a cDNA encoding T. gondii ARF1 (TgARF1) was isolated. Endogenous and transgenic TgARF1 localized to the Golgi of T. gondii, but not to dense granules. An epitope-tagged mutant of TgARF1 predicted to be impaired in GTP hydrolysis (Q71L) partially dispersed the Golgi signal, with localization to scattered vesicles, whereas a mutant impaired in nucleotide binding (T31N) was cytosolic in location. Both mutants caused partial dispersion of a Golgi/trans-Golgi network marker. TgARF1 mutants inhibited delivery of the secretory reporter,Escherichia coli alkaline phosphatase, to dense granules, precluding an in vivo assessment of the role of TgARF1 in release of intact dense granules. To circumvent this limitation, recombinant TgARF1 was purified using two separate approaches, and used in the permeabilized cell assay. TgARF1 protein purified on a Cibacron G3 column and able to bind GTP stimulated dense granule secretion in the permeabilized cell secretion assay. These results are the first to show that ARF1 can augment release of constitutively secreted vesicles at the target membrane.

Remarkable progress has been made recently in understand-ing the protein machinery responsible for secretory events in mammalian and yeast cells (1). The components involved in both constitutive secretion as well as regulated release in secretory cells are broadly conserved evolutionarily (2). Although this suggests that the same should be true in protozoan parasites, the presence of unusual secretory organelles in these organisms confounds this simple hypothesis. By the same argument, parasites provide a unique means to explore selected issues in regulation of secretion from preformed organelles (reviewed in Ref. 3).
We have used this logic to explore the organization of the secretory pathway in the protozoan parasite Toxoplasma gondii. This parasite is an obligate intracellular pathogen that resides in the host cell within a specialized compartment, the parasitophorous vacuole (PV), 1 which is separated from the host cell cytoplasm by the parasitophorous vacuole membrane (PVM). The success of infection depends on the ability of the parasite to modify the PV and the PVM by secreting proteins to the extra-parasite environment (reviewed in Ref. 4). Thus, secretion is a critical feature of parasite survival.
to dense granules (8), rather than to more conventional constitutive secretory vesicles, containing surface proteins of the parasite such as SAG1, making the dense granule pathway highly unique. Because constitutively secreted dense granule proteins are packaged in a distinctive organelle, T. gondii provides a more tractable system for studying control of vesicle docking and fusion at the plasma membrane than is the case in many other cells.
For this reason, we previously developed a permeabilized cell secretion assay, using the pore forming protein streptolysin O (9), to explore the machinery mediating the unusual exocytic process for dense granules. Release was not only constitutive, but also calcium-independent (9, 10), an important distinction when compared to release of dense matrix granules in mammalian cells. Addition to permeabilized parasites of hamster NSF and bovine ␣-SNAP increased the secretion of the stably transfected dense granule secretory reporter ␤-lactamase (BLA). In contrast, bovine Rab GDP dissociation inhibitor blocked BLA secretion, suggesting altogether that the NSF/ SNAP/SNARE/Rab machinery participates in dense granule release. The non-hydrolyzable GTP analogue GTP␥S significantly enhanced BLA secretion in the presence of an ATP regenerating system (ARS), indicating that one or more GTP binding proteins were implicated in dense granules exocytosis.
To pursue this last point, we have now characterized the role of the GTP binding protein ADP-ribosylation factor-1 (ARF1) in dense granule release. ARF proteins, which are broadly conserved, were first characterized by their ability to activate the cholera toxin-mediated ADP-ribosylation of the ␣-subunit of the heterotrimeric G proteins (11). ARFS exist in two stages, bound to guanidine nucleotides: the GDP-bound "off" form, which is cytosolic, and the GTP-bound active form, which interacts with Golgi membranes and phospholipid vesicles. A major function of ARF1 is to drive, in a GTP-dependent cycle, the assembly of sets of cytosolic coat proteins onto Golgi membranes facilitating vesicle budding and conferring the timing for the release of coat proteins to the cytosol (12,13). ARF1 was initially described to recruit the COPI coat involved in traffic between the ER and Golgi. More recent data demonstrate that ARF1 plays a role in generating clathrin-coated vesicles, by first recruiting adaptor complexes on vesicles at the trans-Golgi network (TGN) and immature secretory vesicles (14 -17). Additional effects of ARF1 on the cytoskeleton are also described (18).
Our interests for the current work relate to the role of ARF1 in secretion from the trans-Golgi network. In mammalian cells, ARF1 stimulates release of nascent secretory vesicles from the trans-Golgi network (19), acting through phospholipase D (20). A role of ARF1 in regulated exocytosis in mammalian cells has also been identified (21,22). In this situation, ARF1 enhances the regulated secretion of preformed granules, release of which is induced physiologically by a rise in intracellular calcium following secretagogue addition. Although no evidence has previously existed to support a role for ARF1 in the docking and fusion of constitutively secreted vesicles, we provide such evidence in the current paper. The finding that ARF1 can facilitate docking and fusion of constitutively secreted dense granules was made possible by the unusual features of the T. gondii secretory system.

Cloning of ARF1
An expressed sequence tag (EST) clone encoding a fragment of an ARF1 homologue (W66111/gi:374337) was identified from the T. gondii dbEST sequence data base. The ARF1 open reading frame in this sequence appeared to be interrupted by an unknown fragment. To subclone a full-length ARF1 cDNA, T. gondii RNA was reverse-transcribed with Superscript II Reverse Transcriptase (Life Technologies, Inc., Gaithersburg, MD) for use as the PCR template. A 3Ј-RACE PCR strategy was employed in which the 5Ј-end (sense strand) oligonucleotide primers TARFUT (5Ј-AAA CAC GCG TCC TCT CTC TGC AAG  CGA-3Ј) and TARF1 (5Ј-GGG CAC CAT GGG TTT GAG CGT CAG C-3Ј) were used in sequential PCR reactions with the 3Ј-(antisense) oligo-dT 17 polyadenylation site anchor primer. In the first round, TARFUT and oligo-dT 17 primers were used for cDNA amplification with Taq DNA polymerase (Roche Molecular Biochemicals, Branchburg, NJ). The products were purified on QIAquick PCR columns (Qiagen, Valencia, CA) and used as template in a second round PCR reaction with primers TARF1 and oligo-dT 17 . The resulting products were again purified on QIAquick columns and subcloned into pGEM-T (Promega, Madison, WI). To identify 3Ј-RACE PCR product clones from the resulting colonies, an internal ARF1 probe was generated by PCR from T. gondii cDNA using the known EST sequence with primers TARF1 and TARF2 (5Ј-GCG ATC TCT GTC GTT GCT GT-3Ј). The single PCR product was agarose gel-purified and labeled by random oligonucleotide priming with [ 32 P]dCTP and the Rediprime II kit (Amersham Pharmacia Biotech, Piscataway, NJ). Colony hybridization with this probe identified several clones which contained ϳ0.8-kbp inserts, and sequencing analysis confirmed the presence of a full ARF1 homologue open reading frame (ORF) and 3Ј-untranslated region. For expression in the parasite, an HA epitope tag with a BglII cloning site was engineered for targeting of the 3Ј-antisense end of the ARF1 ORF in primer ARFHAR (5Ј-ACT AGA TCT AAG CGT AGT CTG GGA CGT CGT ATG GGT AAT CGA TGT TTT TCT GCG CAA G-3Ј). The TgARF1 RACE clone ORF was PCR-amplified with TARF1, encoding an NcoI site, and ARFHAR, and subcloned into the NcoI-BglII cloning site of pNTPRab11, utilizing the T. gondii NTPase3 promoter cassette for expression. 2 The TgARF1 cDNA sequence has been deposited in GenBank (accession number AF227524).

Mutagenesis of TgARF1
Site-directed mutagenesis of TgARF1-HA was performed using a two-independent PCR amplification approach followed by triple ligation. For TgARFQ71L, the 237-bp PCR product from primers TARF1 and ARFQ71L (5Ј-GAA GTA GTG GCG CCA CAG AGG ACG AAT CTT GTC CAG TCC ACC G-3Ј) was digested with NcoI and NarI. A 333-bp TgARF1 fragment was prepared by digestion with NarI and BglII, and the two products were subcloned in vector pNTPRab11 following removal of the Rab11 open reading frame with digestion by NcoI and BglII. A plasmid encoding the untagged TgARF1T31N mutation was obtained from Kristen Hager and David Roos (University of Pennsylvania) and used for amplification with TARF1 and ARFHAR, digested with NcoI and BglII, and subcloned into pHXNTPRab11 as above. TgARF1 lacking the 17 N-terminal residues (TgARF1d17-HA) (19,23,24) was constructed in the same vector. Plasmid sequences were verified by dideoxynucleotide sequencing at the WM Keck Sequencing Center, Yale University School of Medicine.

Overexpression and Purification of Recombinant TgARF1-HA Protein
The TgARF1 coding sequence with a C-terminal epitope tag (HA) was subcloned into pET-24d for overexpression in BL21(DE3) E. coli cells, which do not express ␤-lactamase (BLA). This precludes contamination of purified TgARF1-HA with BLA, which is essential for subsequent assays monitoring BLA release from stably transfected T. gondii. Even highly purified preparations of human ARF1 and ARF1 mutants (obtained from D. Shields, New York, NY) generated in standard vectors have sufficient residual BLA contamination to invalidate secretion assays. 3 TgARF1 was inserted into the pET vector between the restriction sites NcoI (at the initiating Met of ARF) and BamHI. Expression of the protein was induced with isopropylthio-␤-galactoside (IPTG, 1 mM) for 4 h at 37°C and bacterial cells were collected by centrifugation. TgARF1-HA was solubilized from the bacterial pellet using a French press at 8,000 p.s.i. in buffer A (50 mM Tris, 2 mM EDTA, 1 mM DDT and 1 mM phenylmethylsulfonyl fluoride). The lysate was clarified by centrifugation at 18,000 ϫ g. Two different protocols were used to purify recombinant TgARF1. First, the protocol described for human and bovine ARF1 (25) was used, consisting of a DEAE-Sephacel column as a first step. The second purification step consisted of a gel filtration column, AcA 54 Ultrogel, equilibrated in buffer B (10 mM potassium phosphate, pH 7.4, 1 mM EDTA, 100 mM NaCl, 1 mM DTT). The column was developed in the same buffer at a flow rate of 19 ml/h; 2.5-ml fractions were collected. Pooled fractions containing ARF1-HA from the Ultrogel column were concentrated using a Centricon-3 concentrator (Amicon, Beverly, MA). Samples were kept at Ϫ80°C. Second, a purification protocol for soluble TgARF1-HA was developed using as a first step a dye column, Cibacron G3, equilibrated in buffer A. Subsequent wash with 1 M NaCl was performed, and the elution of the active protein was performed with 1.2 M NaCl. Fractions containing TgARF1-HA were pooled and concentrated using Centricon-3 concentrators (Amicon). The second step was a gel filtration column, AcA 54 Ultrogel, as described above. Purified recombinant TgARF1-HA was frozen immediately at Ϫ80°C.

SDS-PAGE and Immunoblotting
The expression of recombinant protein TgARF1-HA was monitored using SDS-PAGE, and immunoblots were performed as previously described (8). Immunoblots were developed using an anti-HA monoclonal antibody (1:1000) or a goat anti-human ARF1 (1:500), followed by goat anti-mouse or rabbit anti-goat IgG-horseradish peroxidase conjugation (1:2000) and analysis using the ECL detection system (Amersham Pharmacia Biotech, UK).

Parasites
T. gondii tachyzoites were maintained by serial passage in monolayers of either African Green Monkey (Vero) cells or human foreskin fibroblasts (HFF) grown in modified Eagle's minimal medium or ␣-minimal essential medium, respectively, supplemented with 7.5% fetal bovine serum. The RH strain and a stable transgenic clone of the RH strain, expressing the soluble foreign secretion reporter E. coli ␤-lactamase (BLA) were described previously (8). The stable transgenic clone of the RH strain, expressing bacterial alkaline phosphatase fused with the LDL receptor (BAP-LDLR) is as previously described (26). For experiments with extracellular parasites, infected cells were scraped, and parasites were isolated by two passages through a 27-gauge needle.

Parasite Permeabilization with SLO and Cytosol Depletion
Permeabilization of extracellular parasites with SLO was performed using a protocol described earlier (9). Assessment of permeabilization was done by staining SLO permeabilized parasites with 4 g/ml propidium iodide for 5 min at room temperature. The percentage of positive nuclear staining was quantitated by fluorescence microscopy.

Immunofluorescence Assay (IFA)
Use of HA Epitope Tag to Localize the Transgenes in Transiently Transfected Parasites-The detailed technique is described in Karsten et al. (8). Briefly, confluent HFF cell monolayers (12-mm coverslips) were infected with transiently transfected parasites. Transient transfection was performed by electroporation. After 16 -24 h, cells were fixed and permeabilized with 3% paraformaldehyde in phosphate-buffered saline and 0.1% Triton X-100 and incubated with anti-HA monoclonal antibody (Babco, Richmond, CA) (1:200), followed by FITC or tetramethyl rhodamine-conjugated goat anti-mouse IgG (1:500). Coverslips were mounted in Mowiol and observed with an epifluorescence microscope. Images were captured with a charge-coupled device camera.
Localization of Stably Expressed BAP-LDLR-Confluent monolayers of HFF cell were infected with a Toxoplasma stable line expressing BAP-LDLR (26). Cells were fixed and permeabilized as described earlier after 16 -24 h of infection. As previously described (8), a purified rabbit anti-BAP polyclonal was used as the first antibody (1:1000) and a goat anti-rabbit rhodamine-conjugated was used as secondary antibody (1:500).
Anti-TgARF1 Antibody and Localization of Endogenous TgARF1-Purified recombinant ARF1-HA was sent to Cocalico Biologicals, Inc. for production of a polyclonal rabbit antiserum. Localization of endogenous TgARF1 in T. gondii was done by IFA. HFF cells were infected with tachyzoites (RH) for 16 -24 h. After fixation and permeabilization as described above, cells were stained with the anti-TgARF1-HA antibody (1:100) followed by FITC-conjugated goat anti-rabbit antibody (1:500).

Secretion Assay and Enzymatic Assay for BLA
Secretion assays were performed as described in a previous study (9). In brief, extracellular parasites (5 ϫ 10 7 parasites/ml) expressing the soluble secretory reporter BLA were incubated with various reagents in modified potassium acetate buffer (115 mM potassium acetate, 2.5 mM MgCl 2 , 10 mM glucose, 25 mM HEPES, pH 7.2) at 37°C for 30 min. After the incubation, parasites were centrifuged at 760 ϫ g for 10 min at 4°C. Supernatant was further centrifuged at 7000 ϫ g for 10 min at 4°C.
Determination of BLA in parasite supernatants was done by the method described earlier (9) with minor modifications. 40 l of parasite supernatant (5 ϫ 10 7 parasites/ml) were added to each well of 96-well plate. 160 l of nitrocefin mix (0.2 mM nitrocefin, 0.25 mg/ml bovine serum albumin, 50 mM potassium phosphate, pH 7.0) was added to develop the reaction. Samples were incubated for 20 min at 25°C. Plates were read at 492 nm in an enzyme-linked immunosorbent assay reader.

Electron Microscopy
Immunocytochemistry and Electron Microscopy-Infected Vero cell monolayers were washed twice with phosphate-buffered saline and fixed with 8% paraformaldehyde in 0.25 M Hepes, pH 7.4, for 2 days at 4°C. Sections were obtained in the Yale Center for Cell Imaging using a Leica Ultracut microtome with fetal calf serum cryoattachment at Ϫ108°C. Immunolabeling was performed with rabbit anti-TgARF1 antisera (1:50) followed by protein A-gold (1:70, from the laboratory of J. Slot, Utrecht, Holland). The sections were contrasted with neutral uranyl acetate (2%), infiltrated with methyl cellulose (1.8%) and uranyl acetate (0.5%), air-dried, and examined with a Philips 410 transmission electron microscope.
Tannic Acid Fixation-Extracellular parasites were permeabilized with SLO as described previously (9). GTP␥S and tannic acid arrest of secretion is similar to the methodology described in Newman et al. (29). For 8 min at 37°C, permeabilized parasites were incubated with or without 100 M GTP␥S, 2% tannic acid, and an ATP-regenerating system containing 2 mM ATP. Between washes, samples were fixed sequentially with 2% glutaraldehyde and 1% osmium tretroxide in 100 mM sodium cacodylate buffer. Fixed samples were processed for Epon embedding and sectioned for transmission electron microscopy using standard protocols.

GTP Binding Assay
Binding of [ 35 S]GTP␥S to TgARF1-HA was determined essentially as described previously (11). Purified recombinant TgARF1-HA or TgARF1 mutants were incubated at 30°C with 1 M [ 35 S]GTP␥S (2.8 ϫ 10 6 cpm) in 100 l of 2 mM Tris-HCl (pH 8), 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM EDTA, and 1 mM DTT. After 2 h at 30°C, samples were filtered through nitrocellulose filters. The filters were washed three times with the same buffer. Radioactivity retained by the nitrocellulose filters was quantified with a liquid scintillation analyzer (Packard 2500 TR).

RESULTS
Augmented Dense Granule Protein Release with GTP␥S Reflects Increased Dense Granule Docking and Fusion-We have previously argued that soluble proteins are delivered to and retained quantitatively in T. gondii dense granules, via a bulk flow pathway (8). Because this result was not expected, we considered at the time the alternative that a portion of dense granule proteins was released via a budding process from forming dense granules, analogous to the situation in pancreatic ␤-cells (27). Failure to block dense granule protein release using Brefeldin A was interpreted at the time to provide evidence that all release was from preformed dense granules. However, it has recently been demonstrated that Brefeldin A can augment release of endosomes containing soluble proteins previously routed through immature secretory granules (28). To determine if this process was operative in T. gondii, parasites were treated with wortmannin, which blocks trafficking of lysosomal proenzymes to endosomes, by inhibiting AP-1/clathrin-coated vesicles formation from immature secretory granules (28). No inhibition of either BLA or GRA3 secretion was observed in intact parasites (Table I). This result supports the original argument that soluble proteins in T. gondii are delivered and retained quantitatively within dense granules.
Because this did not address the influence of GTP binding proteins on DG protein release, we next asked whether there were any qualitative or quantitative differences in the morphology of DG exocytosis, in the presence or absence of GTP␥S, using the permeabilized cell system (9). We took advantage of a previously described approach, involving fixation with tannic acid (29), to allow the capture of multiple DG docking/fusion events. As shown in Figs. 1 (A and B) and 2, DG docking and fusion events were readily detected and quantitated (see morphologic definitions for docking and fusion in the legend of Fig.  2) in the presence of tannic acid, ARS, and GTP␥S. In untreated parasites, the total percentage of dense granules that were docked or fused was Ͻ2%. In the presence of SLO and tannic acid alone, 29% of all dense granules (n ϭ 119 analyzed) were docked and 4% were fusing. With SLO, tannic acid, ARS, and GTP␥S, 24% of dense granules (n ϭ 101 analyzed) were docked and 16% were fusing. Of note, no budding of coated vesicles from DG, nor dense granule protein coats, were noted in either the presence (Figs. 1 and 2) or absence (not shown) of GTP␥S. Formation of coated vesicles at the anterior nuclear envelope, and at other locations within cell, was readily visualized in the presence of GTP␥S (Fig. 1C). The later events in dense granule fusion and luminal content release were readily captured using tannic acid fixation (29) (Fig. 2, D-F).
Altogether, these results suggest that a T. gondii GTP binding protein enhances exocytosis of preformed dense granules in a calcium-independent fashion. We therefore turned our atten-tion to the role of the T. gondii GTP binding protein, ARF1, in this process.
Cloning and Sequence Analysis of ARF1-The Toxoplasma data base of expressed sequence tags (EST) revealed a cDNA sequence (W66111) with partial homology to human ARF1. Using a PCR RACE approach, we amplified from parasite cDNA the full-length T. gondii ARF1 homologue. Sequence analysis of cDNA clones revealed an open reading frame encoding a polypeptide of 183 amino acids, with a predicted molecular mass of 21 kDa and an isoelectric point of 6.8. The protein sequence contains prototypical GTP and Mg 2ϩ binding sites and switch regions (Fig. 3) characteristic of RAS-related proteins, and an additional C-terminal signature sequence for the ADP-ribosylation factor family. BLASTN and BLASTP analyses of the sequence data bases through NCBI revealed strongest homology to ARF1 proteins of plants, animals, yeast, and protozoa, and we named the T. gondii homologue TgARF1 accordingly. The highly conserved switch regions of the ARF1 subfamily serve in part as effector binding sites for phospholipase D. Although the T. gondii sequence varies little in these conserved domains from those in the well characterized mammalian and yeast sequences, a Ser-83 residue adjacent to the switch 2 domain is at a site conferring differential stimulation of phospholipase D1 between yeast and mammalian ARF1 (30).
TgARF1-HA Localizes to the Golgi/TGN by Immunofluorescence and Immunoelectron Microscopy-We examined the localization of TgARF1 in T. gondii. A C-terminal HA epitope tag was added, and the TgARF1-HA construct was transiently expressed in T. gondii. Staining with an anti-HA antibody revealed that TgARF1-HA localized predominantly to a region anterior to the nucleus, consistent with the Golgi/TGN of the parasite (Fig. 4A). No staining was apparent on dense gran- Steps in dense granule exocytosis. Dense granules in extracellular parasites treated as described in the legend to Fig. 1 were classified in three stages. Preformed dense granules (A) were not associated with the cortical cisternae and had an intact delimiting membrane (arrowhead). Docking dense granules (B and C) were in contact with the cortical cisternae, often via membranous threads (unlabeled arrow in B), with no or minimal discontinuity in the cortical cisternae. An electron dense band was commonly observed at the junction of the docked DG and the cortical cisternae (arrow in C). Fusing dense granules (D, E) were associated with a discontinuity in the cortical cisternae. Contents from the dense granule lumen are secreted into the extracellular space, occasionally with evagination of the plasma membrane (E), and loss of clear definition of the delimiting membrane surrounding the dense granule. The secreted contents often have a membranous appearance (F). Magnification, ϫ 30,000. These events were assessed quantitatively, and the results are provided in the text. ules. This localization pattern was confirmed with rabbit antiserum to TgARF1, raised against the recombinant protein expressed in E. coli (Fig. 4D). Brefeldin A, which disrupts the T. gondii Golgi (31), induced a redistribution of TgARF1 to the nuclear envelope and ER (not shown). These results are all concordant with the known localization of ARF1 to the Golgi complex in mammalian cells. This result was confirmed by two separate approaches. First, thin section cryoimmunoelectron microscopy was done using the rabbit anti-TgARF1 antiserum. As shown in Fig. 5, TgARF1 localized to the Golgi stacks. No TgARF1 staining of dense granules was observed. Second, a stable line of T. gondii expressing bacterial alkaline phosphatase fused with the LDL receptor (BAP-LDLR) was transiently transfected with TgARF1-HA. As previously assessed by immunoelectron microscopy, BAP-LDLR localizes to the Golgi/TGN of the parasite (26). Transiently transfected TgARF1-HA co-localized with BAP-LDLR in the Golgi region (Fig. 6A), confirming localization of TgARF1-HA to the Golgi/TGN of the parasite.
Expression and Localization of TgARF1 Mutants-We generated two mutants of TgARF1, which are impaired in the GTP binding or hydrolysis cycle. By analogy to the Ras-Q61L mutation, which inhibits GTP hydrolysis but not binding, a Q71L mutation was introduced in T. gondii TgARF1-HA (25) and transiently expressed in the organism. The Q71L mutant localized to the Golgi as well as to dispersed punctate structures (Fig. 4B) in comparison to the discrete Golgi/TGN structure observed with wild type TgARF1-HA. Thus, the Q71L mutant appears to behave as in mammalian cells, where this mutant has distinct effects on the structure and/or function of ER, Golgi apparatus, and endocytic pathway (32). We also generated a T31N mutant of T. gondii TgARF1, a mutant predicted to be defective in GTP binding (19). Following transient transfection of the epitope-tagged protein, the T31N mutant was diffusely distributed throughout the parasite cytosol (Fig. 4C), likely due the inability to bind GTP. Together, these results demonstrate that the localization of TgARF1 within the parasite is coupled to the GTP cycle, as has been described for ARF1 in other systems (reviewed in Ref. 33).
Effect of TgARF1 Mutants on Organelle Structure-The effect of overexpression of TgARF1Q71L-HA on the localization of BAP-LDLR was assessed. TgARF1Q71L-HA was transiently transfected into the BAP-LDLR stable line. In contrast to wildtype TgARF1-HA, this mutant altered the localization of BAP-LDLR, consistent with partial disruption of the Golgi/TGN (Fig. 6B), as shown above. Mechanistically, this alteration in BAP-LDLR localization may be due to both an enhancement of retrograde transport and to augmented anterograde flow from the TGN. A similar partial dispersion of the BAP-LDLR signal was seen with the a T31N mutant (not shown).
Effects of TgARF1 Mutants on Localization of Dense Granule Proteins-We assessed the effects of TgARF1 mutants on transport of proteins to dense granules. Native TgARF1, and TgARF1 mutants ARF1T31N and ARF1Q71L, were transiently overexpressed in a parasite clone stably expressing the soluble dense granule secretory reporter BAP (8) or in wild type parasites. As shown in Fig. 7, in parasites transfected with TgARF1T31N, discrete labeling of BAP (Fig. 7B) or GRA3 (Fig.  7K) in dense granules was reduced and replaced by a reticular endoplasmic reticulum-like staining pattern. Similarly, in parasites transfected with the TgARF1Q71L mutant, labeling of dense granules for BAP (Fig. 7E) or for GRA3 (Fig. 7N) was partially replaced by perinuclear and posterior reticular staining. Less GRA3 was detected at the parasitophorous vacuole membrane, with both the TgARF1T31N (Fig. 7K) and TgARF1Q71L (Fig. 7N) mutants. Staining for GRA3 in the wild type TgARF-1 transfectants was variably altered, with mild effects in most cells (Fig. 7H). These results suggest that one predominant effect of both Q71L and T31N mutants is to block delivery of soluble proteins to dense granules and to induce accumulation of the secretory proteins in the ER. This observation precluded an assessment of ARF1 effects on post-Golgi release of dense granule proteins.
Expression and Purification of Recombinant TgARF1 Protein-We therefore opted to analyze the effects of TgARF1 on dense granule secretion in the permeabilized cell assay, using purified TgARF1-HA protein. TgARF1-HA was expressed in E. coli using the pET system. Induction with 1 mM IPTG of BL21(DE3) cells carrying the TgARF1-HA/pET-24d construct resulted in a time-dependent increase in the accumulation of TgARF1-HA protein, which continued for at least 120 min (not shown). The induced protein migrated at ϳ21 kDa on SDS-PAGE.
Previous work overexpressing human and bovine ARF1 in E. coli has shown that recombinant ARF1 proteins are easily solubilized from the bacterial pellet with TX-100 and lysozyme treatment or using a French press cell. Our results confirmed these facts. Recombinant TgARF1-HA was solubilized from IPTG-induced bacterial cell pellets (Fig. 8, lane 2). Different protocols were attempted to purify recombinant TgARF1-HA. First, the protocol described for human and bovine ARF1 (25) was used. This protocol consists of a DEAE-Sephacel column as a first step, where typically almost all of the bacterial proteins are absorbed, while ARF1 is not retained. In our experiments, recombinant TgARF1-HA was adsorbed to the DEAE matrix along with the bacterial proteins (not shown), necessitating FIG. 4. Localization of wild type TgARF1 and TgARF1 mutants by IFA. Tachyzoites (RH) were transiently transfected with either wild type TgARF1 or TgARF1 mutants, both of which were tagged at the C terminus with an HA epitope (A-C). Monolayers of HFF cells were infected for 18 h and fixed with paraformaldehyde and permeabilized with Triton X-100. IFA images were obtained staining with an HA monoclonal antibody, followed by FITC-goat anti-mouse. Wild type TgARF1 and mutants were expressed under the control of T. gondii NTPase promoter. ARF-1-HA localizes to the Golgi/TGN region (A). The Q71L-HA mutant is more dispersed in the parasite, with small punctate structures in addition to Golgi region staining (B). The T31N mutant is diffusely distributed throughout the parasite cytosol (C). Immunofluorescence localization of endogenous ARF-1, detected with rabbit anti-TgARF1 (D). Corresponding phase contrast images are shown. further purification on an AcA 54 Ultrogel column (not shown). Second, we used an immunoaffinity column, containing monoclonal anti-HA antibodies cross-linked to a Sepharose matrix (Babco, Richmond, CA). Even though the antibody recognized TgARF1-HA in immunoblots, the recombinant TgARF1-HA was not absorbed to the column under the various conditions tested (not shown). Finally, purification of the soluble TgARF1-HA was done using a Cibacron G3 column, equilibrated with buffer A. Cibracon G3 is a triazine dye column, used as an affinity resin for binding nucleotide-dependent enzymes. TgARF1-HA remained adsorbed to this column, and was not present in the column flow (Fig. 8, lane 3), even after a 1 M NaCl wash (not shown). TgARF1 was eluted with 1.2 M NaCl (Fig. 8, lane 4). The second purification step consisted of a gel filtration column, AcA 54 Ultrogel, equilibrated in buffer B (10 mM potassium phosphate, pH 7.4, 1 mM EDTA, 100 mM NaCl, 1 mM DTT). The column was developed in the same buffer at a flow rate of 19 ml/h, and 2.5-ml fractions were collected. Pooled fractions containing purified TgARF1-HA from the Ultrogel column are shown in Fig. 8, lane 5. This band was determined to be epitope-tagged TgARF1-HA by specific recognition (not shown) with the monoclonal anti-HA antibody, the correct molecular size, and the absence of the same band in control cells, carrying the pET24-d plasmid without the coding sequence for TgARF1 (Fig. 8, lane 1).
We tested the ability of purified recombinant TgARF1-HA to bind GTP␥S. TgARF1-HA purified on the Cibacron G3 column bound 35 [S]GTP␥S (Fig. 9, lanes 1 and 2). In contrast, TgARF1-HA purified by the standard protocol using DEAE chromatography did not bind 35 [S]GTP␥S (Fig. 9, lane 3). Despite multiple attempts, it was not possible to purify either the TgARF1Q71L-HA, TgARF1T31N-HA, or TgARF1d17-HA mutants by Cibacron G3 chromatography, nor on other dye resin columns. Although all of these proteins were successfully purified by the standard DEAE method (not shown), the purified proteins lacked GTP binding activity (Fig. 9, lanes 4 and 5), likely explaining the failure of the proteins to bind to dye resin columns. Control samples (containing no protein or bovine serum albumin) did not bind the nucleotide (Fig. 9, lane 6).
Recombinant TgARF1 Stimulates BLA Release from Permeabilized Parasites-Finally, we examined the influence of recombinant TgARF1-HA on the release of the secretory reporter BLA in permeabilized T. gondii. Extracellular tachyzoites were permeabilized with 1 unit/ml streptolysin O, cytosol was depleted, and BLA release was measured, as described previously (9) and under "Materials and Methods." Three separate experiments are illustrated in Fig. 10, A, B, and C. Slightly different experimental variables were tested in the three experiments. ARS alone but not GTP␥S alone augmented BLA release slightly, as previously reported (9). Treatment with 100 g/ml TgARF1-HA (purified on Cibacron G3 and stored at Ϫ80°C) resulted in 1.4-fold (Fig. 10A, lanes 2 and 4), 1.8-fold (B, lanes  2 and 4), or 2.0-fold (C, lanes 3 and 4) enhancement of release of the secretory reporter BLA from permeabilized parasites in the presence (but not the absence) of ARS. Addition of GTP␥S to ARS and TgARF1-HA did not substantially augment BLA release (Fig. 10A, lanes 4 and 5). Addition of TgARF-1 to GTP␥S-treated parasites, in the presence of ARS, did not substantially augment BLA release in comparison to GTP␥S plus ARS alone (Fig. 10B, lanes 4 and 5). Treatment of non-permeabilized cells with recombinant TgARF1-HA did not result in stimulation of secretion, either in the absence or presence of ARS and GTP␥S (Fig. 10A, lanes 6 and 7). TgARF1-HA purified by DEAE chromatography, and lacking GTP binding activity, did not stimulate BLA release from permeabilized parasites (not shown). These results illustrate that TgARF-1 augments release of dense granule proteins, and that the magnitude of the effect is similar to that of GTP␥S alone or in combination with GTP␥S. DISCUSSION Our data suggest that T. gondii TgARF1 stimulates post-Golgi secretion from preformed DG. This represents a specialized function for TgARF1 in this apicomplexan parasite. A related function has been described for ARF1 in selected secretory cells, in which ARF1 mediates the release of preformed granules in the presence of GTP␥S (21,22). Nonetheless, a fundamental difference between our system and release observed in mast cells, chromaffin cells, or PC12 cells is that DG release in the parasite is a constitutive process, which is not triggered by calcium. Hence, our results have direct relevance to constitutive secretion. Moreover, we have been able to identify an effect of TgARF1 at the final step of constitutive vesicle release because of the unique features of the T. gondii system.
In mast cells, ARF1-mediated enhancement of granule secretion is thought to occur via activation of phospholipase D. PLD hydrolyses phosphatidylcholine to generate PA and choline. The conversion of phosphatidylcholine to PA alters the lipid bilayer properties, replacing a non-fusogenic phospholipid with a fusogenic one (34), potentially stimulating preformed granules to fuse and release their contents to the extracellular environment. In addition, PA generated by PLD regulates a phosphatidylinositol 5-kinase, resulting in enhanced synthesis of phosphatidylinositol 4,5-bisphosphate (PIP 2 ). Elevated lev- els of PIP 2 stimulate mast cell exocytosis. In contrast, in both chromaffin and PC12 cells, ARF1 stimulates granule release in a process not dependent on PLD activation. It is of note that ethanol, which diverts PA to PE and blocks PIP 2 generation in mammalian cells, also inhibits dense granule secretion in T. gondii (35). This result suggests that TgARF1 augments DG release in a PLD-dependent fashion. Successful purification of functionally active TgARF1d17-HA capable of binding GTP␥S would have allowed a direct test of this hypothesis (19,23,24) in the permeabilized cell system, but this was not possible (Fig. 9).
There are several non-mutually exclusive alternatives to the scenario presented above. It is possible that TgARF1 is involved in a triggered component of dense granule release. In this scenario, the secretion of preformed dense granules measured in the permeabilized cell assay in the absence of GTP␥S would correspond to the constitutive component of DG release and would not be dependent on ARF1 activity. The putative burst of DG release following invasion would be an ARF1-dependent process, analogous to the situation already described in selected secretory cells (21,22). Nonetheless, the relationship of T. gondii dense granules to dense core secretory granules in mammalian cells is not clear, and there are many unusual features of the T. gondii system. The process is calcium-independent, and no physiological trigger has been identified for dense granule release (9, 10). No immature secretory granule precursors (36) are visible in T. gondii. Soluble pro-teins are routed by the bulk flow pathway quantitatively to dense granules (8). The presence of putative membranous material within the dense granule matrix (Fig. 2) is also unique. Altogether, these features suggest that T. gondii dense granules, despite their morphologic appearance, are evolutionarily distinct from dense core granules in mammalian secretory cells. As another alternative, ARF1 may augment release of a sub-population of dense granules, although no convincing evidence yet exists for dense granule heterogeneity. TgARF1 may also enhance formation of nascent secretory vesicles at the trans-Golgi network (19,20), in addition to or even instead of augmenting release of intact DG. Finally, different pathways may be triggered by GTP␥S and by TgARF1 in the permeabilized cell system. In particular, the effects on GTP␥S on release of intact dense granules could be mediated either via a Rab protein or another member of the ARF family, such as ARF6 (34). Although all of these possibilities exist, and several processes may operate concurrently, the morphology, kinetics, and pharmacological inhibition profile suggest at a minimum that TgARF1 is enhancing constitutive release of preformed dense granules.
TgARF1 was not detected on the dense granule membrane, either by immunofluorescence or immunoelectron microscopy. This does not obviate a role for TgARF1 in mediating release of preformed dense granules, especially if a TgARF1-regulated phospholipase D activity localizes to the cortical cisternae or plasma membrane (37). Furthermore, translocation of ARF1 to TgARF1T31N localization is predominately cytosolic with minor Golgi association (arrows in A, J). Transiently expressed TgARF1-HA is concentrated in the Golgi and cytosol (G). BAP and GRA3 proteins label dense granules in untransfected parasites (small arrows in B, E, H, K). Transport of BAP (B, E) or GRA3 (H, K, N) to dense granules is partially blocked, and the dense granule proteins accumulate in an reticular ER-like pattern throughout the ARF1T31N-transfected parasites (arrows in B, K) and ARF1Q71L parasites (arrows in E, N). Additionally, less secreted GRA3 signal is detected at the parasitophorous vacuolar membrane in parasites expressing ARF1T31N and ARF1Q71L than in untransfected parasites (long arrows in K, N). Transiently overexpressed ARF1HA induces only partial accumulation of GRA3 in the Golgi region (H). secretory granule membranes of rat parotid acinar cells, specifically in the presence of GTP␥S, has been previously demonstrated (38), and such a process may also occur in T. gondii.
Our data suggest that TgARF1 is also involved in maintaining the structure of the T. gondii Golgi and TGN. In mammalian cells, ARF1 supports the GTP-dependent association of COPI with Golgi membranes (39 -41). Localization of T. gondii TgARF1 to the Golgi/TGN is consistent with such a function in vivo in the parasite. Although it is apparent that transfected TgARF1 affects delivery of secretory reporters to dense granules (Fig. 7), likely by effects on the Golgi, this is not likely to be the case in the permeabilized cell assay.
The functions for ARF proteins in the secretory and endocytic pathways of mammalian cells are now quite broad. Although it is assumed that TgARF1 will have most analogous functions in T. gondii, this is yet to be formally established. The parasite has a Golgi-ER retrieval system that is reminiscent of the process in higher eukaryotes (42) and has components of a COP1 coat (31), confirming the expectation from the morphologic consequences of Brefeldin A treatment. T. gondii expresses the AP-1 adaptor complex, and sorts proteins in an AP-1-dependent fashion, suggesting that TgARF1 will also participate in this step (26). 4 AP-3-dependent sorting is also ARF1dependent (43,44), and circumstantial evidence (26) argues in favor of an AP-3-dependent sorting pathway in Toxoplasma. We expect that the unusual features of the T. gondii secretory and endocytic pathways may allow further insights into ARF1 functions, as these pathways are explored in the parasite.