A Single Malaria Merozoite Serine Protease Mediates Shedding of Multiple Surface Proteins by Juxtamembrane Cleavage*

Erythrocyte invasion by the malaria merozoite is accompanied by the regulated discharge of apically located secretory organelles called micronemes. Plasmodium falciparum apical membrane antigen-1 (PfAMA-1), which plays an indispensable role in invasion, translocates from micronemes onto the parasite surface and is proteolytically shed in a soluble form during invasion. We have previously proposed, on the basis of incomplete mass spectrometric mapping data, that PfAMA-1 shedding results from cleavage at two alternative positions. We now show conclusively that the PfAMA-1 ectodomain is shed from the merozoite solely as a result of cleavage at a single site, just 29 residues away from the predicted transmembrane-spanning sequence. Remarkably, this cleavage is mediated by the same membrane-bound parasite serine protease as that responsible for shedding of the merozoite surface protein-1 (MSP-1) complex, an abundant, glycosylphosphatidylinositol-anchored multiprotein complex. Processing of MSP-1 is essential for invasion. Our results indicate the presence on the merozoite surface of a multifunctional serine sheddase with a broad substrate specificity. We further demonstrate that translocation and shedding of PfAMA-1 is an actin-independent process.

Apicomplexan parasites possess specialized zoite stages primarily designed to invade host cells. Most zoites are extracellular only transiently, but this relatively vulnerable stage in the parasite life cycle is widely considered a target for drug or vaccine-based interventions aimed at controlling parasite replication. Invasion is accompanied by dramatic proteolytic modification of zoite surface proteins. These proteins broadly divide into two classes on the basis of their trafficking pathway. Resident surface proteins are sorted directly to the plasma membrane of the intracellular developing zoite and are often characterized by the presence of a glycosylphosphatidylinositol (GPI) 1 membrane anchor, or by being peripherally associated with a GPI-anchored protein. A well studied example of this class is merozoite surface protein-1 (MSP-1). Synthesized as a large precursor protein, it exists on the free Plasmodium falciparum merozoite as a complex of four proteolytic fragments, additionally associated with protein products of two distinct genes (1). The entire complex is held on the parasite surface via interactions with a single, GPI-anchored polypeptide derived from the C-terminal end of the MSP-1 precursor. At erythrocyte invasion this is proteolytically cleaved in a process called secondary processing (reviewed in Ref. 2) to form two products, MSP-1 33 and MSP-1 19 . MSP-1 33 is shed quantitatively, still bound to the rest of the complex, whereas MSP-1 19 , which consists of two epidermal growth factor (epidermal growth factor)-like domains, remains on the invading parasite surface. The protease responsible is a membrane-bound calciumdependent merozoite serine protease. MSP-1 is a conserved Plasmodium molecule, and the processing step is similarly conserved. Upon successful invasion it always goes to completion. When processing is inhibited, invasion cannot occur, making it a potential target for protease inhibitor-based drug development.
Invasion by apicomplexan zoites also involves regulated release of secretory organelles within the anterior domain of the parasite, called rhoptries, micronemes, and dense granules. Several microneme proteins translocate onto the parasite surface, where they form a second class of induced or transiently expressed surface proteins. They are thought to act in many cases as adhesive ligands promoting the interaction between parasite and target host cell (3,4). There is accumulating evidence that for productive invasion, these adhesins must eventually be released from the parasite surface by regulated proteolysis (see, e.g., Ref. 5). Apical membrane antigen-1 (AMA-1) was first identified as a target of monoclonal antibodies (mAb), which prevent erythrocyte invasion by merozoites of the simian malaria parasite Plasmodium knowlesi (6). Homologues of AMA-1 are present in all species of Plasmodium examined, as well as in other apicomplexans. Unsuccessful attempts to disrupt the gene in Plasmodium blood stages (7) and Toxoplasma gondii asexual stages (8) indicate that it has an essential function in this part of the parasite life cycle. Replacement of the P. falciparum ama-1 gene with the divergent Plasmodium chabaudi homologue produced transgenic parasites which invaded mouse erythrocytes more efficiently than wild type P. falciparum (7). These observations, together with the now numerous reports of in vivo protection or inhibition of host cell invasion by antibodies (see, e.g., Ref. 9) or by peptides (10) that bind to the AMA-1 ectodomain, implicate AMA-1 in invasion. All Plasmodium AMA-1 sequences contain 16 completely conserved cysteine residues (Fig. 1), most of which are also positionally conserved in the T. gondii AMA-1 (11). The disulfide bonding pattern between these cysteines (12) and limited structural analysis (13) suggests that the AMA-1 ectodomain contains three domains, each stabilized by intradomain disulfide bonds. AMA-1 is secreted from micronemes onto the parasite surface and shed in a soluble form following parasite release from the host cell (8, 11, 14 -17). This entire process involves at least two sequential proteolytic processing events. In the case of P. falciparum AMA-1 (PfAMA-1), transport to its initial apical location is quickly followed by the loss of an N-terminal pro-sequence, which extends from the secretory signal cleavage site to residue Ser 96 (16,17). The truncated protein (but not the full-length precursor), still in possession of its transmembrane (TM) anchoring sequence, is then translocated onto the parasite surface in the form of a 66-kDa species called PfAMA-1 66 , and is finally quantitatively shed as a result of further proteolysis. On the basis of incomplete mass spectrometric peptide mapping data, we previously proposed that the shed protein took two forms: one (called PfAMA-1 48 in reference to its apparent molecular mass of 48 kDa on SDS-PAGE) that was concluded to represent the bulk of the ectodomain, being the product of a membrane-proximal cleavage of PfAMA-1 66 ; and a second form called PfAMA-1 44 , which was thought to represent a product of cleavage of PfAMA-1 66 at some unidentified point between domains II and III (17).
We now report the results of a detailed extension of this earlier study. Our new data provide a definitive description of proteolytic processing of PfAMA-1, cast new light on the mechanism by which PfAMA-1 is translocated onto the merozoite surface, and reveal a previously unsuspected link between shedding of PfAMA-1 and secondary processing of MSP-1.
A proprietary combinatorial library of 2672 random drug-like compounds was a kind gift of Evotec OAI (Abingdon, Oxford, United Kingdom (UK)). Compounds were screened for inhibition of MSP-1 secondary processing at a concentration of 100 M, using the Western blot-based assay described previously (18). This identified two compounds, 239f and 2312f, which inhibit both MSP-1 processing and erythrocyte invasion in vitro with IC 50 values in the low micromolar range (data not shown).
Synthesis of the peptidyl chloromethylketone N-benzyloxycarbonyl-Gly-Met-Leu-CH 2 Cl (Z-GML-CH 2 Cl) was by the mixed anhydride method (19); complete details of the synthesis and characterization will be provided elsewhere. 2 Design of this compound was based on the sequence flanking the secondary processing site within the P. falciparum MSP-1 (20). The compound inhibits both MSP-1 secondary processing and erythrocyte invasion with an IC 50 value of ϳ30 M. It is also a potent inhibitor of subtilisin BPNЈ and chymotrypsin, but does not significantly inhibit elastase, trypsin or the P. falciparum subtilisinlike protease PfSUB-1 (21) at the highest concentrations tested (500 M). 3 Parasites and Antibodies-Culture and synchronization of P. falciparum (clone 3D7) was as previously described (18). Routinely, parasites were maintained in complete medium containing the serum substitute Albumax, although where indicated protein-free medium was employed. Mature schizonts were isolated by centrifugation on cushions of 63% isotonic Percoll, and either biosynthetically radiolabeled with [ 35 S]methionine/cysteine (22) or used for preparation of naturally released merozoites for use in processing assays as previously described (18). Purification of mAb 4G2dc1, which recognizes a conformational epitope within the PfAMA-1 ectodomain (23), was as described previously (17). The human mAb X509, which recognizes an epitope within the 3D7 MSP-1 33 , has been described (20). The RAP-1-specific mAb 209.3 was a kind gift of Dr. A. Holder (National Institute of Medical Research, London, UK). A mouse antiserum was raised against reduced, alkylated recombinant PfAMA-1 (9) using standard protocols.
Purification and Immunoprecipitation of Shed PfAMA-1-Shed forms of PfAMA-1 were purified from parasite culture supernatants. Briefly, Percoll-enriched mature schizonts (unlabeled or biosynthetically radiolabeled with [ 35 S]methionine/cysteine) were recultured for up to 11 h in the presence or absence of fresh erythrocytes to allow merozoite release and reinvasion. Culture supernatants (referred to as schizont rupture supernatants) were then harvested and clarified by filtration through a 0.22-m filter. Immunoprecipitation of radiolabeled material and analysis by SDS-PAGE and fluorography was as described previously (22). Quantitation of radioactivity associated with selected bands on dried gels was performed with a Storm 860 PhosphorImager, using ImageQuant ® version 5.0 software (Molecular Dynamics). All immunoprecipitations used mAb 4G2dc1 coupled to cyanogen bromideactivated Sepharose 4B Fast Flow (Amersham Biosciences). An affinity column for larger scale purification was produced as described previously (17). Clarified culture supernatant (up to 3 liters at a time) was applied to the column at a flow rate of 1 ml min Ϫ1 at 4°C. The column was washed with 20 mM Tris-HCl, 500 mM NaCl, pH 7.6, 5 mM CHAPS in 20 mM Tris-HCl, 150 mM NaCl, pH 7.6, and finally with the same buffer without CHAPS. Bound proteins were eluted with 0.2 M glycine-HCl, 0.15 M NaCl, pH 2.7, and further purified by gel filtration on a Superdex 200 HR 10/30 column equilibrated in 20 mM Tris-HCl, pH 8.2, 150 mM NaCl. Alternatively, protein eluted from the affinity column was applied to a Vydac 4.6 mm ϫ 15-cm C 4 reversed-phase high pressure liquid chromatography (RP-HPLC) column equilibrated in 0.1% (v/v) trifluoroacetic acid, and eluted at 1 ml min Ϫ1 with a 0Ϫ54% (v/v) gradient of acetonitrile in 0.1% (v/v) trifluoroacetic acid over 60 min, monitoring elution at 215 nm. Eluate samples were dried under vacuum for further analysis by RP-HPLC, SDS-PAGE, or mass spectrometry.
For re-application of samples to the RP-HPLC column following reduction and alkylation, dried samples from the first round of RP-HPLC purification were taken up in 100 l of 8 M urea in 5 mM Tris-HCl, pH 8.2, 50 mM DTT, and incubated at room temperature for 30 min. Iodoacetamide was then added to 100 mM and incubation continued for an additional 20 min at 37°C in the dark. Samples were made up to 1 ml with 0.2% (v/v) trifluoroacetic acid in water, re-applied to the C 4 RP-HPLC column, and eluted using the protocol described above.
Electrospray Mass Spectrometry (ESI-MS)-RP-HPLC-purified protein was dried under vacuum. Where previously reduced and alkylated, samples were simply taken up directly into 60% (v/v) acetonitrile, 0.1% (v/v) formic acid and submitted to mass spectrometry. Where not previously reduced, protein was taken up in 6.25% (v/v) acetonitrile in 5 mM NH 4 HCO 3 , supplemented with DTT to 50 mM, and incubated for 2.5 h at room temperature to allow complete reduction. Samples were then acidified with 0.1% (v/v) formic acid and analyzed on a Platform single quadrupole mass spectrometer (Micromass, Altringham, UK) as described previously (17).

Proteolytic Digestion and Matrix-assisted Laser Desorption/Ionization Time-of-flight (MALDI-TOF) Mass Spectrometry-Purified
PfAMA-1 44 and PfAMA-1 48 proteins purified in the absence of prior reduction were solubilized directly into SDS-PAGE sample buffer containing 0.1 M DTT, alkylated with iodoacetamide, subjected to SDS-PAGE, and then submitted to in-gel proteolytic digestion as described previously (17). Alternatively, dried peaks from RP-HPLC fractionation of reduced and alkylated protein were taken up directly into protease digestion buffer. Digestion with Lys-C was performed in 5 mM NH 4 HCO 3, using the enzyme at 2.5-5.0 ng l Ϫ1 , whereas digestion with Glu-C was in 5 mM sodium phosphate, pH 7.8, using the enzyme at 10 ng l Ϫ1 . Where required, digestion buffers were prepared using 50% (v/v) 18 O water. After overnight digestion at 32°C, the digestion supernatant was acidified by the addition of trifluoroacetic acid to 0.4% (v/v) before analysis on a Reflex-III MALDI-TOF mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany) as described previously (17).
Analytical Gel Filtration Chromatography-A Superdex 200 HR 10/30 column equilibrated in 20 mM Tris-HCl, pH 8.2, 150 mM NaCl was calibrated with marker proteins ferritin (440 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and cytochrome c (12.4 kDa). Blue dextran 2000 was used to determine the void volume (V o ) of the column. Schizont rupture supernatants were obtained from schizonts cultured for 4 h in protein-free medium. Freshly harvested supernatants were concentrated 20-fold by centrifugation in a 12-kDa molecular mass cut-off Vectaspin Micro ultrafiltration unit (Whatman), and 50 l applied at once to the gel filtration column. The column was eluted at 0.4 ml min Ϫ1 . Fractions of 0.4 ml were collected and assayed for the presence of PfAMA-1-derived fragments by Western blot using the mouse anti-PfAMA-1 antiserum.
Immunofluorescence-Percoll-enriched mature schizonts were recultured for 1 h in fresh medium containing cytochalasin D (final concentration 2 M, added from a 2 mM stock solution in Me 2 SO) or Me 2 SO only, to allow merozoite release. Residual intact schizonts and released merozoites were then pelleted by centrifugation (6500 ϫ g, 1 min at room temperature). Thin films were air-dried, fixed in ice-cold acetone, and probed sequentially with mAb 4G2dc1, fluorescein isothiocyanateconjugated sheep anti-mouse IgG (Sigma), and the RAP-1-specific mAb 209.3 which had been directly conjugated to Texas Red as described previously (24). Slides were dipped in a solution of 4,6-diamidino-2phenylindole in phosphate-buffered saline, washed, and mounted in Citifluor (Citifluor UKC Chemical Laboratories, Canterbury, UK). The cells were visualized on an integrated DeltaVision system (Applied Precision) including an Olympus IX70 inverted microscope with a 100ϫ objective. Images were captured with a Micromax air-cooled CCD camera (Princeton Instruments, Trenton, NJ) on a Silicon Graphics Iris work station, using Softworx image acquisition and deconvolution software.

RESULTS
The C Terminus of PfAMA-1 44 Lies within Domain III of PfAMA-1-In previous attempts to determine the precise site(s) at which PfAMA-1 is cleaved to release it from the merozoite, shed fragments were isolated and subjected to proteolytic digestion in the presence of 50% (v/v) 18 O water. Under these conditions, peptide products of digestion incorporate both 18 O and 16 O at their terminal ␣-carboxyl group as a result of serine protease-mediated hydrolysis (25). Such peptides possess a distinctive isotope pattern in MALDI-TOF mass spectrometry and are distinguishable from any peptide derived from the extreme C terminus of the polypeptide substrate, which should exhibit a normal 16 O isotope spectrum. No peptide of the latter type was evident in 18 O-containing chymotryptic or tryptic digests of PfAMA-1 44 and PfAMA-1 48 (17), suggesting that the C-terminal products of digestion by these proteases were too small or not sufficiently basic to be detected by MALDI-TOF.
We extended this approach, using an expanded range of proteolytic enzymes. When affinity-purified PfAMA-1 44 was digested with endoproteinase Glu-C in the presence of 50% (v/v) 18 O water, MALDI-TOF analysis (data not shown) now readily identified a single unlabeled species of m/z 858.52. This can only correspond to the predicted Glu-C peptide with the sequence S 458 KRIKLN 464 (calculated m/z 858.55). This result was unexpected; the sequence lies just downstream of Cys 443 , the first cysteine of domain III (Fig. 1), and therefore indicated that, rather than lying between domains II and III as previously concluded, the C terminus of PfAMA-1 44 44 and PfAMA-1 48 showed that the difference in molecular mass between the two species is reduction-dependent ( Fig. 2A); in the absence of DTT, the two proteins co-migrate. This is consistent with PfAMA-1 44 being disulfide-bound to another species, which, together with PfAMA-1 44 , approximates the total mass of PfAMA-1 48 . Under reducing conditions, this peptide would dissociate and migrate at the solvent front on SDS-PAGE and so would not be readily detectable.
To identify it, and to better analyze the structure of the shed  (12) and Nair et al. (13) and suggest that the ectodomain comprises three independently folded domains. The downward arrow indicates the position of the C terminus of PfAMA-1 44 (this work), whereas the upward arrow indicates the C terminus of PfAMA-1 48 , i.e. the membrane-proximal site where PfAMA-1 is cleaved to completely release it from the merozoite surface (this work). Secretory signal and predicted TM domains are shown in dark gray, and the pro-sequence region is hatched. B, primary sequence of 3D7 PfAMA-1. Sequence encompassed by PfAMA-1 66 (17) is shaded, the first Cys residue of domain III is double underlined, and the Glu-C peptide SKRIKLN forming the C terminus of PfAMA-1 44 (this work) is boxed. Positions indicated by arrows are as in A, and the sequence of the small polypeptide that is shed disulfidebound to PfAMA-1 44 lies between the arrows. Secretory signal and predicted TM sequences are underlined.
species, larger amounts of protein were isolated on a mAb 4G2dc1 affinity column (Fig. 2B, left-hand lane), then further purified by gel filtration or RP-HPLC. The PfAMA-1 44 and PfAMA-1 48 species co-migrated precisely on both the gel filtration column (Fig. 2C) and on the RP-HPLC column (Fig. 2D), where the elution profile was characterized by a single major peak with a retention time of 43.5 min. A sample of the gel filtration peak (fraction 39 in Fig. 2C) was reduced and analyzed directly by ESI-MS. Only two species were detectable (Fig. 3), one of molecular mass 48,336.85 Ϯ 5.17 Da, probably representing PfAMA-1 48 (the major component of the mixture on SDS-PAGE), and the second of molecular mass 6040.97 Ϯ 1.19 Da. Identical results were obtained upon ESI-MS analysis of the RP-HPLC peak (data not shown). Comparison of the high molecular mass signal with the predicted 3D7 PfAMA-1 sequence suggests that, if it corresponds to PfAMA-1 48 , the species extends from its previously determined N-terminal residue of Ile 97 through to Thr 517 (the calculated mass of this sequence is 48,341.75 Da). The lower mass signal corresponds closely to the predicted mass of Asp 465 -Thr 517 of the PfAMA-1 sequence (predicted m/z of the reduced, non-alkylated sequence is 6038.87), suggesting that this is the sequence of the peptide hypothesized to be disulfide-bonded to PfAMA-1 44 . PfAMA-1 44 was not detected, probably because of the fact that this is always present at substantially lower levels than PfAMA-1 48 , as noted previously (17), and its charge envelope was likely obscured by that of the more abundant species. This issue is returned to below.
A further sample of the RP-HPLC-purified PfAMA-1 44 and PfAMA-1 48 mixture was taken up in 8 M urea, reduced and alkylated to dissociate any disulfide-dependent interactions, then re-applied to the C 4 RP-HPLC column. Three peaks were now evident, with retention times of 31.6, 43.6, and 45.5 min (Fig. 4A). SDS-PAGE of the major peak (data not shown) dem-onstrated that it contained both PfAMA-1 44 and PfAMA-1 48 , presumably in fully alkylated form. The minor peak at 43.6 min was also found on SDS-PAGE (data not shown) to contain both PfAMA-1 44  analyzed by MALDI-TOF (Fig. 4B). A strong but simple spectrum was obtained, comprising only seven major ions. All derived from sequence between Asp 465 and Thr 517 of PfAMA-1 (see Fig. 1). Most significantly, only one ion of m/z 1119.496, corresponding to 509 CVERRAEVT 517 (the predicted m/z of the carbamidomethylated form of this peptide is 1119.558) lacked an 18 O component to its spectrum, showing conclusively that this peptide was derived from the C terminus of the fragment. Post-source decay fragmentation of this peptide in the MALDI-TOF mass spectrometer (Fig. 4C) confirmed its identity beyond doubt.
The primary structure of PfAMA-1 44 and PfAMA-1 48 was finally confirmed by comparative MALDI-TOF analysis of Lys-C digests of SDS-PAGE purified, reduced, and alkylated protein, performed in the presence of 18 O water (Fig. 5). Each spectrum contained just a single species that was not labeled with 18 O, and in each case this was absent from the other spectrum. In the PfAMA-1 44 digest, the unlabeled ion at m/z 643.470 corresponds to 460 RIKLN 464 (calculated m/z 643.426), and is consistent with the Glu-C data discussed above in assigning the C-terminal residue of PfAMA-1 44 as Asn 464 . In the case of the PfAMA-1 48 digest, the unlabeled ion at m/z 1119.511 again corresponds to 509 CVERRAEVT 517 , demonstrating unequivocally that the C terminus of PfAMA-1 48 is identical to that of the small polypeptide associated with PfAMA-1 44 .
PfAMA-1 48 and PfAMA-1 44 Are Shed in a Monomeric Form and at a Constant Molar Ratio of Approximately 2:1-The above data support the model proposed above, but do not give any information as to the higher order status of the shed fragments. The T. gondii microneme protein TgMIC2 is shed tightly bound to a distinct protein called TgM2AP (26), and several Plasmodium rhoptry proteins form multimeric complexes (see, e.g., Ref. 27). If the same were true of PfAMA-1, such noncovalent interactions could be sensitive to the conditions to which the protein was exposed during purification (e.g. low pH) and so might not be evident using these procedures. We explored this possibility by analytical gel filtration chromatography of material, which had not been exposed to any potentially dissociating conditions. Freshly harvested schizont FIG. 4. Identification and C-terminal sequence of the small polypeptide disulfide-bonded to PfAMA-1 44 . A, PfAMA-1 44 and PfAMA-1 48 purified by RP-HPLC under non-reducing conditions (the major peak in Fig. 2D) was reduced and alkylated then re-applied to the RP-HPLC column. A novel peak at 31.6 min (arrowed) was dried down and digested with Lys-C in 50% (v/v) 18 O water. B, complete MALDI-TOF mass spectrum of the Lys-C digest. The most intense ions are labeled with their m/z values, and peaks unambiguously derived from PfAMA-1 are also labeled with their amino acid sequence. Insets show magnified views of the arrowed peaks. Only the peak corresponding to 509 CVERRAEVT 517 exhibits an unlabeled 16 O isotope spectrum. C, postsource decay sequencing of the peptide corresponding to 509 CVER-RAEVT 517 . The fragmentation spectrum is shown, with prominent b and y ions labeled. The inset shows positions at which fragmentation occurred to produce these ions. rupture supernatants were concentrated and applied directly to a calibrated Superdex column and the position of elution of the PfAMA-1 fragments established by Western blot analysis. Fig. 6 shows that PfAMA-1 44 and PfAMA-1 48 co-eluted from the column at an elution volume consistent with a mass of ϳ46.5 kDa. This is close to the mobility of the proteins on nonreducing SDS-PAGE ( Fig. 2A), and indicates that both forms of the shed protein, both PfAMA-1 48 , and PfAMA-1 44 linked to its cognate C-terminal peptide, exist substantially or exclusively as monomers in solution and are not released as part of a complex with any other proteins.
PfAMA-1 48 is always more abundant than PfAMA-1 44 in schizont rupture supernatants. To estimate the stoichiometry of the relationship, the radioactivity associated with immunoprecipitated, biosynthetically radiolabeled proteins separated on reducing SDS-PAGE was quantified by PhosphorImager analysis (data not shown). In 10 independent sets of counts, the mean ratio between the counts associated with the PfAMA-1 48 and PfAMA-1 44 species was found to be 2.72 Ϯ 0.29. The sequence encompassed by PfAMA-1 48 (Ile 97 -Thr 517 ) contains 10 Met and 16 Cys residues, whereas that encompassed by PfAMA-1 44 (Ile 97 -Asn 464 ) contains 9 Met and 11 Cys residues. Assuming that the specific activities of the [ 35 S]methionine and [ 35 S]cysteine in Pro-mix are similar, and that both radioisotopes are incorporated with equal efficiency at all positions during biosynthetic labeling, our measurements suggest that the molar ratio between the two species is ϳ2.1:1. Our inability to detect PfAMA-1 44 in direct ESI-MS analysis of purified PfAMA-1 44 and PfAMA-1 48 (Fig. 3) probably reflects this stoichiometry.
Shedding of PfAMA-1 Is Mediated by a Parasite Protease Which Is Indistinguishable from That Responsible for Secondary Processing of MSP-1-In initial experiments focused on the provenance and nature of the protease responsible for cleaving PfAMA-1 from the merozoite surface, a suspension of metabolically radiolabeled mature schizonts (ϳ85% parasitemia) was dispensed into 24-well plates and supplemented with increasing amounts of fresh erythrocytes such that the final ratio of host cells to schizonts in individual wells varied from ϳ0.15:1 to 1000:1. Following a period of culture to allow merozoite release, supernatants were analyzed by immunoprecipitation. The total number of erythrocytes present had no detectable influence on the degree of PfAMA-1 shedding (data not shown).
In further experiments, the extent or specificity of shedding was found to be unaffected by the presence or absence of human serum or the serum substitute Albumax (data not shown). Our data are consistent with shedding being mediated by a parasite protease.
To characterize this activity, we examined the sensitivity of PfAMA-1 processing in culture to a range of protease inhibitors or cytochalasin B and D, inhibitors of actin polymerization. As shown in Fig. 7A, most of the compounds under test did not discernibly affect shedding of the PfAMA-1 44 and PfAMA-1 48 species. Shedding was slightly reduced in the presence of ALLN and ALLM, and more substantially inhibited in the presence of TLCK, TPCK, and the serine protease inhibitor AEBSF, suggesting that these compounds directly affect PfAMA-1 processing. However, microscopic examination of these cultures following the incubation step revealed that the observed inhibition was likely the result of a direct effect on schizont rupture, because in all these cases the cultures contained a significant number of morphologically abnormal, residual unruptured schizonts as compared with control cultures (data not shown). PfAMA-1 shedding was unaffected by the presence of 2 M cytochalasin B or D (note that preliminary experiments showed that invasion is completely blocked at 2 M amounts of either cytochalasin). In the presence of chymostatin, or the chelating agents EDTA and EGTA, schizont rupture occurred normally; here, however, not only was shedding of PfAMA-1 44 and PfAMA-1 48 reduced, much of the shed protein took the form of a higher molecular mass species of ϳ52 kDa, indicating aberrant processing. To further explore this, we examined the effect of EGTA in combination with additional calcium. Fig. 7B (lane 4) shows that merozoite release into culture medium supplemented with EGTA plus equimolar levels of CaCl 2 resulted in the normal pattern of PfAMA-1 processing and shedding. In contrast, supplementing harvested supernatants with CaCl 2 , following the period of culture and merozoite release into medium containing EGTA, did not reverse the effect of EGTA (Fig. 7B, lane 5). These results indicate that the observed effect of EGTA was directly caused by its chelating activity, and show that, once shed, the ϳ52-kDa PfAMA-1 fragment produced in the presence of EGTA cannot subsequently be converted to PfAMA-1 44 and PfAMA-1 48 by the addition of excess calcium. This is consistent with the protease mediating formation of these latter species being active only at the merozoite surface. Inhibiting its activity still allowed PfAMA-1 to be shed, albeit less efficiently, as a result of an alternative cleavage at a different site. To identify this site, large amounts of the ϳ52-kDa form were purified from schizont rupture supernatants containing EGTA (Fig. 2B, right-hand  lane), then analyzed by in-gel digestion and MALDI-TOF, or ESI-MS. Despite extensive efforts, it was not possible to establish the mass or C terminus of this species (data not shown).
To better characterize the protease that releases PfAMA-1 from the merozoite surface, we turned to an assay originally developed for studying processing of MSP-1 (18,20). Purified merozoites in a calcium-containing buffer were incubated at 37°C in the presence of a range of protease inhibitors. PfAMA-1 shedding was then assessed by Western blot analysis of the merozoite supernatants. Fig. 7C (upper panel) shows that only a subset of inhibitors (the sulfhydryl-reactive compound pHMB, the serine protease inhibitors PMSF and DCI, and the chelating agent EDTA) had any effect on PfAMA-1 shedding. Interestingly, these reagents have been previously shown to inhibit secondary processing of MSP-1 (18,20). Reprobing the same blots with mAb X509, which recognizes the shed MSP-1 33 product of MSP-1 secondary processing, confirmed this (Fig. 7C, lower panel). Particularly striking was the observation that the relative sensitivity of shedding of both MSP-1 and AMA-1 to the irreversible serine protease inhibitors PMSF, DCI, and AEBSF was similar; at the concentrations used, PMSF potently inhibited both processes, DCI had an intermediate effect, and AEBSF was completely ineffective. This led us to explore the effects on processing of a number of other more selective inhibitors (Fig. 7D). These included a peptidyl chloromethylketone, Z-GML-CH 2 Cl, which was designed and synthesized to be an inhibitor of MSP-1 secondary processing, and two compounds, 239f and 2312f, selected from a large compound library on the basis of their inhibition of MSP-1 processing. In every case, shedding of PfAMA-1 and shedding of MSP-1 were equally susceptible to the compounds under test. Collectively, our results strongly suggest that shedding of both proteins is mediated by the same protease.
Translocation of PfAMA-1 Is Not Sensitive to Cytochalasin D-Cytochalasin blocks invasion by malarial merozoites, arresting it at the point of tight junction formation (28,29) and supporting the widely held view of a role for an actinomyosin motor in propelling the parasite into the host cell. Cytochalasin also impairs anterior-to-posterior translocation of a number of microneme proteins (see, e.g., Refs. 30 and 31). The absence of any effect of cytochalasin on shedding of PfAMA-1 in the experiments described above led us to explore its effects on mobilization of the protein onto the merozoite surface. Schizonts were cultured in medium containing invasion-inhibitory concentrations of cytochalasin D for 1 h to allow merozoite release. Merozoites released over this period were then acetone-fixed and examined by immunofluorescence using mAb 4G2dc1 to determine the subcellular localization of PfAMA-1. As shown in Fig. 8, two distinct patterns of distribution of PfAMA-1 on FIG. 7. The protease responsible for shedding of PfAMA-1 is indistinguishable from that which mediates secondary processing of MSP-1. A, immunoprecipitation of shed PfAMA-1 fragments from schizont rupture supernatants following culture in the presence of the indicated compounds; supernatants were harvested at once (start) or after 5 h of culture. Me 2 SO (0.1% v/v) and methanol (1% v/v) acted as solvent-only controls. B, shedding of PfAMA-1 upon rupture of biosynthetically radiolabeled schizonts was determined as in A. Culture medium used was supplemented with 10 mM EGTA alone, or additionally with 10 mM CaCl 2 , which was added either together with the EGTA prior to the 5-h culture period (lane 4), or following the period of culture in EGTA-containing medium (lane 5). The ϳ52-kDa species shed in the presence of excess chelating agent is arrowed. C, processing of PfAMA-1 and MSP-1 in isolated merozoites. Shed fragments of PfAMA-1, or the MSP-1 33 product of secondary processing, were detected in merozoite supernatants by Western blot using the anti-PfAMA-1 antiserum or mAb X509, respectively; supernatants were either harvested at once (start) or after a 1-h incubation at 37°C in the presence of the indicated compounds. D, further processing assays, performed as in C. Z-GML-CH 2 Cl but not another chloromethylketone, TLCK, inhibits processing of both MSP-1 and PfAMA-1. Similarly, compounds 239f and 2312f but not two structurally related compounds, 247h and 2411e, inhibit processing of both MSP-1 and PfAMA-1. Solvent alone had no effect on either processing activity (data not shown). Structures of compounds 239f and 2312f are shown. individual merozoites were observed, consisting of either a solely apical staining pattern with the entire fluorescence signal concentrated in a discrete zone (sometimes appearing as two punctate dots) just forward of the rhoptries, or alternatively an exclusively circumferential pattern consistent with a merozoite surface location. Merozoites exhibiting both patterns were present in the same thin films. However, the presence of cytochalasin D during merozoite release had no detectable effect on the proportion of free merozoites (ϳ90% in all cases) that exhibited a circumferential pattern of mAb 4G2dc1 reactivity. Our results indicate that PfAMA-1 translocation onto the merozoite surface is not dependent upon an actinomyosinbased mechanism. DISCUSSION We have presented incontrovertible evidence that shedding of PfAMA-1 from the merozoite surface is mediated exclusively by proteolytic cleavage at a single site, Thr 517 , just 29 residues N-terminal to the start of the predicted TM sequence (Fig. 9). A constant proportion (approximately one third) of the shed molecules also contain an internal proteolytic cleavage at Asn 464 within domain III, resulting in the formation of PfAMA-1 44 , which remains bound to its cognate C-terminal fragment of domain III via a single disulfide bond. The reproducibility of this is striking, but its biological significance is unclear. It is also unclear whether both cleavage events are mediated by the same protease; certainly, there is no great sequence similarity surrounding the two cleavage sites. However, as we have previously shown (17), once shed there is no further conversion of PfAMA-1 48 to PfAMA-1 44 , indicating that, like the juxtamembrane cleavage, the cleavage within domain III that produces PfAMA-1 44 takes place exclusively at the parasite surface.
PfAMA-1 shedding is sensitive to precisely the same subset of serine protease inhibitors and chelating agents previously shown to inhibit secondary processing and shedding of the GPI-anchored surface molecule MSP-1 and its associated complex. Furthermore, the activity is equally sensitive to pHMB and a peptidyl chloromethylketone designed to target the MSP-1 processing protease, as well as to two compounds selected from a screen of a large, structurally diverse compound library. It seems most unlikely that shedding of the MSP-1 complex and PfAMA-1 are mediated by distinct merozoite surface proteases with indistinguishable physiochemical characteristics; much more likely is the presence at the merozoite surface of a single enzyme that is responsible for shedding of both. The most obvious argument against this is the lack of any obvious similarity between the MSP-1 secondary processing site and that flanking Thr 517 in PfAMA-1, and the quite different anchoring structures of the two proteins. However, numerous studies on mammalian cell surface proteins that are subject to the action of "sheddases" (usually but not always zinc metalloproteases) indicate that substrate recognition is not dependent primarily on the amino acid sequence flanking the juxtamembrane cleavage site; more important is thought to be the presence of an 8 -10-residue-long stretch of conformationally unrestrained peptide surrounding the target bond, and its distance from the membrane (e.g. Ref. 32). Our data clearly implicate a serine protease and not a metalloprotease in shedding of PfAMA-1 and MSP-1, but the distance from the membrane at which PfAMA-1 is cleaved, 29 residues, is typical of sheddases. MSP-1 secondary processing occurs at a site ϳ96 residues upstream of the GPI anchoring sequence; however, as has been pointed out previously (2,33), the unusual U-shaped structure of the tandem epidermal growth factor-like motif that comprises MSP-1 19 means that the secondary cleavage site probably lies in very close proximity to the GPI anchor, and is therefore ideally situated for interaction with a membranebound protease. Despite considerable sequence diversity within the MSP-1 19 sequence across Plasmodium species, its overall fold is conserved (34 -36). It is also functionally conserved; gene targeting studies have shown that the P. falciparum MSP-1 19 sequence can be replaced with the divergent P. chabaudi sequence without any effects on invasion or secondary processing (37). It is conceivable that its compact structure serves mainly to orient the cleavage site at an appropriate distance from the membrane to allow protease binding. Interestingly, Brossier et al. (5) recently showed that proteolytic shedding of the T. gondii adhesin TgMIC2 is sensitive to mutations in a basic residue that lies 11 positions N-terminal to the TM sequence. This residue is conserved in a number of microneme proteins, leading the authors to suggest that the protease responsible, called MPP1 (38), may mediate shedding of several microneme proteins. Our data from P. falciparum support this concept and, additionally, raise the exciting possibility that resident surface proteins too may be shed by the same enzyme.
Inhibition of normal PfAMA-1 shedding invoked inefficient cleavage at an alternative site, presumably by a distinct protease. The resulting ϳ52-kDa fragment was larger than that shed under normal conditions, indicating that it resulted from cleavage at a site closer to, or even within, the TM sequence. We were unable to identify this alternative site. A recent report has suggested that some microneme proteins may be shed via cleavage at an intramembrane site (39). Our data indicate the presence of such an activity at the merozoite surface, but it clearly is not responsible for the bulk of PfAMA-1 shedding under physiological conditions.
There is now substantial evidence that movement of microneme components across the surface of apicomplexan zoites is controlled by interactions with a sub-plasmalemmal actinomyosin motor (see, e.g., Ref. 40). Exocytosis per se of microneme FIG. 9. New model for translocation and proteolytic processing of PfAMA-1. The schematic is based on the present work and that of Howell et al. (17). Following translocation onto the merozoite surface by an actin-independent mechanism, PfAMA-1 66 is shed in two alternative forms; in both cases, release is mediated by a common juxtamembrane cleavage at Thr 517 . The intact form of the shed molecule exists as PfAMA-1 48 . A proportion (ϳ33%) of the shed molecules also contain an additional cleavage at Asn 464 within domain III, producing PfAMA-1 44 , which remains bound via a single disulfide bond to the small polypeptide comprising the remainder of domain III, and shown in this work to correspond to Asp 465 -Thr 517 . Disulfide bonds are represented by thin lines.
proteins is not necessarily coupled to their translocation; Bumstead and Tomley (31), for example, showed that cytochalasin D did not reduce secretion of the Eimeria tenella tachyzoite protein EtMIC2 in a soluble form, but did prevent its capping over the parasite surface. We found that cytochalasin D at concentrations that completely block erythrocyte invasion affected neither shedding of PfAMA-1 nor its movement onto the merozoite surface. It is conceivable that, given the small dimensions of the P. falciparum merozoite, PfAMA-1 may redistribute spontaneously upon secretion by simple diffusion in the merozoite plasma membrane. Alternatively, its translocation may be via interactions with other motor proteins (41). Discharge of many microneme proteins is induced only upon contact with the appropriate host cell, and it has been suggested that this may be a mechanism of reducing exposure of these functionally critical molecules to host antibodies (42). Relocalization of PfAMA-1 upon merozoite release, in contrast, appears to be relatively unregulated. This may contribute to its accessibility to invasion-inhibitory antibodies.
Fraser et al. (43) have reported that domains I and II, but not the complete ectodomain, of the Plasmodium yoelii AMA-1 exhibit erythrocyte binding activity, leading these authors to suggest that a truncated form of AMA-1 may play an adhesive role during invasion. Our findings are difficult to reconcile with this model, showing as they do that, under physiological conditions either on the merozoite surface or in solution, authentic PfAMA-1 never exists in a form that lacks domain III. We have consistently failed to demonstrate that the shed forms of PfAMA-1 possess any erythrocyte binding activity (17). We have also been unable to demonstrate any capacity for a correctly folded recombinant PfAMA-1 48 molecule (9) to bind to human erythrocytes, either in solution or when immobilized at high density on plastic or nitrocellulose, or when expressed at the surface of COS-7 cells with its cognate TM and cytoplasmic regions. 4 Our gel filtration and immunoprecipitation data unambiguously demonstrate that PfAMA-1 is shed in a monomeric form, uncomplexed with any distinct proteins. This suggests that, whatever the function of PfAMA-1, it does not involve high affinity interactions with other malarial proteins. However, we cannot rule out the possibility that the protein may exist in a multimeric form or be engaged in interactions with other merozoite proteins, only while occupying the merozoite surface, and that these interactions might be disrupted upon shedding.
In summary, we have demonstrated the presence at the merozoite surface of a serine protease activity capable of mediating the shedding of at least two distinct, functionally indispensable proteins by cleaving them at juxtamembrane sites that share no obvious sequence similarity. The molecular identification of this protease may be aided by the recently completed sequence of the P. falciparum genome (44), which has demonstrated that the genome encodes no serine proteases of the chymotrypsin-like superfamily, and contains only three subtilisin-like serine protease genes. 3 PfAMA-1 is of great interest as a potential malaria vaccine candidate. Nair et al. (13) have reported that antibodies against domain III can prevent erythrocyte invasion in vitro. It is conceivable that, as in the case of antibodies against MSP-1 19 (45), these antibodies may function by inhibiting the processing and shedding of PfAMA-1.