The disulfide bond structure of Plasmodium apical membrane antigen-1.

Apical membrane antigen-1 (AMA-1) of Plasmodium falciparum is one of the leading asexual blood stage antigens being considered for inclusion in a malaria vaccine. The ability of this molecule to induce a protective immune response has been shown to be dependent upon a conformation stabilized by disulfide bonds. In this study we have utilized the reversed-phase high performance liquid chromatography of dithiothreitol-reduced and nonreduced tryptic digests of Plasmodium chabaudi AMA-1 secreted from baculovirus-infected insect cells, in conjunction with N-terminal sequencing and electrospray-ionization mass spectrometry, to identify and assign disulfide-linked peptides. All 16 cysteine residues that are conserved in all known sequences of AMA-1 are incorporated into intramolecular disulfide bonds. Six of the eight bonds have been assigned unequivocally, whereas the two unassigned disulfide bonds connect two Cys-Xaa-Cys sequences separated by 14 residues. The eight disulfide bonds fall into three nonoverlapping groups that define three possible subdomains within the AMA-1 ectodomain. Although the pattern of disulfide bonds within subdomain III has not been fully elucidated, one of only two possible linkage patterns closely resembles the cystine knot motif found in growth factors. Sites of amino acid substitutions in AMA-1 that are well separated in the primary sequence are clustered by the disulfide bonds in subdomains II and III. These findings are consistent with the conclusion that these amino acid substitutions are defining conformational disulfide bond-dependent epitopes that are recognized by protective immune responses.

undergoes asexual reproduction, producing a new generation of merozoites, which become directly accessible to immune attack at the time of schizont rupture. As a consequence, there is a great deal of interest in the identification and characterization of merozoite surface antigens as potential vaccine candidates (1). One such protein is the apical membrane antigen-1 (AMA-1) 1 (2)(3)(4)(5)(6)(7)(8), a minor surface antigen that is synthesized in mature, segmenting schizonts during the final 4 h of erythrocytic development (9). Initially, AMA-1 appears to be located in the electron-dense neck of the rhoptries, but after schizont rupture a processed form of the protein spreads circumferentially around the surface of mature merozoites (9).
Although the biological function of AMA-1 is unknown, its location and stage specificity suggest that it may be involved in the process of erythrocyte invasion (9,10). A monoclonal antibody raised against native PK66, the Plasmodium knowlesi homologue of AMA-1, and Fab fragments of this antibody prevented P. knowlesi merozoites from invading rhesus erythrocytes in vitro (11). Furthermore, passive transfer of AMA-1specific polyclonal antibodies into Plasmodium chabaudiinfected mice prevented lethal parasitemias. 2 Protection against parasite challenge can also be induced by active immunization with AMA-1. For example, immunization of rhesus monkeys with the native PK66, affinity purified from infected rhesus erythrocytes, partially protected against blood stage challenge (10). Recombinant AMA-1 expressed using baculovirus-infected insect cells has protected immunized monkeys and mice against simian and rodent parasites, respectively (12). 2 More recently, mice were protected against infection with P. chabaudi by immunization with the presumed ectodomain of P. chabaudi AMA-1 expressed in Escherichia coli. This fragment of AMA-1 (AMA-1B) lacked the putative transmembrane and C-terminal cytoplasmic domains of the full-length polypeptide (residues 481-503 and 503-558, respectively). Protection was induced with antigen that had been refolded in vitro with the formation of intramolecular disulfide bonds but not by antigen that had been reduced and alkylated. 2 Full sequences have been reported for the AMA-1 genes of 11 isolates of Plasmodium falciparum (3,4,13), two strains of P. knowlesi (14), and strains of Plasmodium vivax (7), Plasmodium fragile (5), Plasmodium cynomolgi (8), and P. chabaudi (6). Unlike many other asexual stage antigens, AMA-1 lacks immunodominant blocks of repetitive sequence and has the * This work was supported by the Cooperative Research Centre for Vaccine Technology, the United Nations Development Program/World Bank/World Health Organization Special Programme for Research and Training in Tropical Diseases and National Health and Medical Research Council (Australia). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ structural characteristics of a Type I integral membrane protein. The ectodomain of AMA-1B contains 16 cysteine residues, which are conserved within all known sequences of AMA-1. Because a conformation stabilized by disulfide bonds is necessary for AMA-1 to induce a protective immune response, we have sought to determine the disulfide bond topology in P. chabaudi AMA-1 secreted from baculovirus-infected insect cells. For this we used a combination of reversed-phase HPLC peptide mapping under reducing and nonreducing conditions and subsequent N-terminal sequence and mass analysis of the disulfide-linked peptides.

MATERIALS AND METHODS
Expression of P. chabaudi AMA-1 Using Baculovirus-The methods used for cloning AMA-1 sequences in baculovirus and expression in Sf9 insect cells will be described in detail elsewhere. 2 Briefly, DNA encoding AMA-1B, comprising amino acid residues 1-499 of the 558-residue full-length polypeptide (GenBank accession number U49743), was amplified by polymerase chain reaction from P. chabaudi adami (DS strain) DNA and cloned into the transfer vector pAcRP23. Plasmid DNA produced in E. coli mixed with DNA from wild-type baculovirus was used to transfect Sf9 insect cells. Recombinant baculovirus was selected by plaque purification and then amplified in Sf9 insect cell cultures. Cultures expressing AMA-1B were detected by Western blot analyses using polyclonal antibodies raised to recombinant P. falciparum AMA-1 expressed in E. coli.
SDS-PAGE and Immunoblot Analysis-Baculovirus-expressed P. chabaudi (strain DS) AMA-1B samples were electrophoresed under reducing conditions on 10% polyacrylamide/SDS gels (SDS-PAGE) (15). For immunoblots, proteins were electrophoretically transferred from SDS-PAGE gels onto nitrocellulose filters (16). The filters were incubated in 5% (w/v) nonfat milk powder in phosphate-buffered saline and reacted with polyclonal rabbit antibodies raised against refolded P. chabaudi DS AMA-1B expressed in E. coli. The filters were further incubated with 125 I-labeled protein A and autoradiographed.
Purification of AMA-1B-Crude AMA-1B supernatant from baculovirus-infected Sf9 insect cells was dialyzed overnight at 4°C against 0.5 M NaCl/20 mM Tris-HCl buffer, pH 7.4, with continuous stirring and then centrifuged at 14,500 ϫ g for 20 min at 4°C. Supernatant (250 ml) was pumped at 3 ml/min onto 45 ml of concanavalin A-Sepharose (Pharmacia Biotech Inc., Uppsala, Sweden) packed into a Pharmacia C26/40 C-type column using a Bio-Rad (Richmond, CA) automated Econosystem operated at 4°C. The column was then washed with 10 column volumes of 0.5 M NaCl/20 mM Tris-HCl, pH 7.4. Bound proteins were eluted from the column with the same buffer containing 0.5 M methyl-␣-D-mannopyranoside (Sigma, St. Louis, MO). Fractions containing AMA-1B were identified by enzyme-linked immunosorbent assay utilizing rabbit antiserum raised against refolded recombinant AMA-1B expressed in E. coli. These fractions were then pooled and dialyzed against 0.02% (v/v) Tween 20 (Pierce, Rockford, IL) 20 mM Tris-HCl buffer, pH 8.3.
Anion-exchange chromatography was performed using a Pharmacia FPLC unit. Buffer A comprised 20 mM Tris-HCl, pH 8.3, whereas buffer B comprised 20 mM Tris-HCl, pH 8.3, 500 mM NaCl. The dialyzed solution of AMA-1B (50 ml) was loaded through a superloop, at a flow rate of 3 ml/min onto a Pharmacia 5-ml HiTrap-Q column. The column was washed in buffer A (5 column volumes), and bound proteins were eluted using the gradient profile: 0 -6% buffer B for 0 -10 min; 6 -12% buffer B for 10 -20 min; 12-100% buffer B for 20 -80 min at a flow rate of 3 ml/min. Fractions containing AMA-1B were identified by enzymelinked immunosorbent assay as above.
Reversed-phase HPLC was performed using a Hewlett-Packard (Waldbronn, FRG) 1050 modular HPLC consisting of an on-line degasser, a piston pump, a rheodyne injector, and a diode-array detector. Instrument control, data acquisition, and evaluation were performed using Hewlett-Packard HPLC 3D Chemstation software run on a Hewlett-Packard Vectra 486/33VL personal computer. Buffer A comprised 0.05% (v/v) trifluoracetic acid (HPLC/Spectro grade, Pierce) in Milli-Q grade water (Millipore, Bedford, MA), whereas buffer B comprised 0.05% (v/v) trifluoroacetic acid and 60% (v/v) acetonitrile (Chro-mAR HPLC grade, Malinckrodt, Paris, KY) in Milli-Q grade water. Pooled ion-exchange fractions containing AMA-1B were loaded directly onto a Brownlee (Applied Biosystems, Santa Clara, CA) RP-300 Aquapore C8 4.6-mm inner diameter ϫ 100-mm column in the presence of buffer A. Bound proteins were eluted using a linear gradient of 0 -100% buffer B over 45 min at a flow rate of 1 ml/min. The peak fraction containing AMA-1B was identified by enzyme-linked immunosorbent assay and immunoblot analysis using the rabbit antiserum to AMA-1B. Edman degradation of purified recombinant AMA-1B revealed a single polypeptide with an N-terminal sequence consistent with that expected for AMA-1B. 3 Enzymatic Deglycosylation of AMA-1B-Baculovirus-expressed AMA-1B was deglycosylated using a recombinant N-glycanase enzyme (EC 3.5.1.52 and 3.2.2.18) from Genzyme (Cambridge, MA). An aliquot (100 l) of anion-exchange purified AMA-1B at pH 8.3 was adjusted to a final concentration of 0.5% (w/v) SDS and 50 mM ␤-mercaptoethanol, and the protein was denatured by boiling for 5 min. An aliquot (20 l) of this solution was mixed with 5 l of 7.5% (v/v) Triton X-100 (United Technologies/Packard, Downers Grove, IL), 0.38 units of N-glycanase enzyme (1.5 l) and made up to 30 l with Milli Q grade water (3.5 l). The solution was incubated overnight at 37°C. The reaction was then stopped by the addition of reducing SDS-PAGE sample buffer (15). Human apotransferrin (Sigma) was used as a positive control and was prepared using 10 l from a 2 mg/ml solution in a similar manner described for AMA-1B. Negative controls, comprising either AMA-1B or human apotransferrin, were incubated overnight in digest buffer at 37°C without the N-glycanase enzyme. Samples were analyzed using SDS-PAGE and immunoblot analysis as described above.
Tryptic Digestion of AMA-1B-The RP-HPLC peak fraction containing AMA-1B was reduced to half volume using a Labconco (Kansas City, MO) Centrivap concentrator. The solution was then buffered to pH 6.5 using a 1:10 dilution of 1 M sodium phosphate, pH 6.5. Sequencing grade modified trypsin (Promega, Madison, WI) in 50 mM acetic acid was added at an enzyme/protein ratio of 1:25 (w/w), and the solution was incubated at 37°C for 5 h. A second aliquot of trypsin (1:25 w/w) was then added, and the mixture was incubated for a further 14 h at 37°C. Digestion was stopped by either snap freezing to Ϫ70°C or RP-HPLC.
RP-HPLC Analysis of AMA-1B Tryptic Fragments-Narrow bore RP-HPLC of DTT-reduced and nonreduced AMA-1B tryptic digests was used to identify the location of disulfide-linked peptides. RP-HPLC was conducted using the Hewlett-Packard 1050 system and buffers described above. An aliquot containing 2.5 g of tryptic digest was mixed with an equal volume of 6 M guanidinium chloride, 0.2 M Tris-HCl, pH 8.4, 0.005 M EDTA and then reduced by the addition of DTT (final concentration, 25 mM) for 60 min at 45°C. A tryptic fingerprint for the reduced digest was then obtained by eluting the peptide fragments from a C18 column (2.1-mm inner diameter ϫ 250 mm, Vydac (The Separations Group, Hesparia, CA)) using a 0 -100% buffer B linear gradient over 90 min at a flow rate of 0.1 ml/min. The profile obtained was compared with the tryptic fingerprint obtained for a similar amount of nonreduced digest eluted under identical chromatographic conditions. The disappearance of a peak in the nonreduced fingerprint following reduction indicated the presence of a disulfide-linked peptide. The remainder of the nonreduced tryptic digest mixture (120 g) was fractionated under similar conditions using a 0 -100% buffer B linear gradient over 90 min and a Vydac C18 (4.6 mm-inner diameter ϫ 250 mm) column operated at a flow rate of 1 ml/min. Disulfide-containing peptide fractions were collected and subjected to Edman degradation and electrospray-ionization mass-spectrometric analysis.
RP-HPLC fractionation of the thermolysin digest was performed at 45°C. Buffer A comprised 0.1% (v/v) aqueous trifluoroacetic acid, whereas buffer B comprised aqueous 60% (v/v) acetonitrile containing 0.1% (v/v) trifluoroacetic acid. The thermolysin digest mixture was fractionated on a C8 column (2.1-mm inner diameter ϫ 100 mm, Brownlee RP-300) using a 0 -100% buffer B linear gradient over 12 min at a flow rate of 0.5 ml/min (17). Thermolysin peak fractions 9 and 17 were rechromatographed in a second dimension RP-HPLC buffer system to resolve minor peptide contaminants. In this system, buffer A comprised aqueous 154 mM NaCl, and buffer B comprised 60% aqueous acetonitrile containing 154 mM NaCl. Peptides were eluted from a Brownlee RP-300 column using a 0 -100% buffer B linear gradient over 60 min at a flow rate of 0.1 ml/min.
Capillary Column RP-HPLC/Electrospray-Ionization Mass Spec-trometry-Peptides were separated by on-line rapid reversed-phase capillary HPLC coupled to a triple quadrupole mass spectrometer (Finnigan-MAT model TSQ-700, San Jose, CA) equipped with a Finnigan electrospray ionization source configured as described elsewhere (18,19). The reversed-phase capillary column used in this study (150 mm ϫ 0.2-mm inner diameter, Vydac C18) was fabricated as described elsewhere (20,21) and developed with a linear 15-min gradient at 1.6 l/min from 0 to 90% buffer B, where buffer A was 0.1 M aqueous acetic acid, and buffer B was acetonitrile. The electrospray ionization needle was operated at Ϫ4.5 kV. The sheath liquid was 70% methanol/0.1 M acetic acid delivered at 3 l/min via a syringe drive (Harvard Apparatus, South Natick, MA). Nitrogen sheath and auxiliary gases obtained from a boiling liquid nitrogen source were supplied at 20 p.s.i. and 15 units (arbitrary value), respectively. The heated capillary was set at 150°C. Mass spectra were collected every 3 s in centroid mode over the m/z range 300-2000. Peptide molecular masses were calculated using Finnigan BIOMASS software. The PEPMAP component of this software was used to generate a theoretical tryptic digest of the AMA-1B amino acid sequence. This program also performs an iterative combination of all tryptic fragments containing Cys residues, enabling prediction of the sequences and molecular masses for the corresponding peptides linked through a single disulfide bond. N-terminal Sequence Analysis-N-terminal amino acid sequencing of peptides was performed by automated Edman degradation using a Hewlett-Packard (model G1005A) protein sequencer operating with the routine 3.0 sequencer program. A Hewlett-Packard model HP1090 M liquid chromatograph was used for phenylthiohydantoin amino acid analysis.

Purification of AMA-1 Ectodomain (AMA-1B)-Recombinant
AMA-1B secreted by baculovirus-infected insect cells was purified by affinity chromatography on concanavalin A, followed by anion-exchange and reversed-phase chromatographies. The final RP-HPLC step resolved four peaks (data not shown). The dominant peak was found to correspond to AMA-1B, which was identified by its reactivity on immunoblots with anti-AMA-1B antibodies and by Edman degradation analysis. N-terminal sequence analysis (data not shown) revealed that the signal sequence was cleaved from the mature exported protein between Cys 21 and Ser 22 of the predicted primary translation product.
The AMA-1B isolated by RP-HPLC was a closely migrating doublet when analyzed by SDS-PAGE (Fig. 1a, lane 1). Prior incubation of AMA-1B with N-glycanase generated a single species of M r 57,000 (Fig. 1b, lane 3) smaller than either component of the original doublet (Fig. 1b, lane 4). Thus the observed heterogeneity in RP-HPLC-purified AMA-1B was apparently due to differential glycosylation rather than the existence of sequence variants.
RP-HPLC Analysis of AMA-1B Tryptic Digest-Enzymatic cleavage of AMA-1B with trypsin under slightly acidic condi- tions to reduce the potential for disulfide bond interchange was performed to generate individual disulfide-linked peptides. To identify disulfide-linked peptides, analytical RP-HPLC analyses were performed on identical portions of the tryptic digest with and without DTT reduction. A comparison of the two tryptic fingerprints revealed a number of peptide peaks where chromatographic behavior altered upon reduction (Fig. 2).
Mass Spectrometric and N-terminal Sequence Analyses of RP-HPLC Fractionated AMA-1B Tryptic Peptides-The remainder of the digest (120 g) was fractionated using a larger diameter RP-HPLC column (Fig. 3). Peak fractions containing potential disulfide-linked peptides in the preparative chromatogram were identified by comparing the profiles of the analyt-ical and preparative chromatograms (compare Figs. 2 and 3). These fractions were subjected to RP-HPLC/electrospray-ionization mass spectrometry analysis and Edman degradation to establish cysteine connectivities. The disulfide bond assignments for baculovirus-expressed P. chabaudi AMA-1B obtained using this strategy are summarized in Table I.
Mass spectrometric analyses of fractions thought to contain disulfide-linked peptides (see asterisk-labeled peaks in Figs. 2 and 3) identified three molecular ions with masses equivalent to tryptic fragments predicted by PEPMAP software. The peptide mass found within peak fraction 34 (1592.2 Da) was consistent with peptide Ser 278 -Arg 284 and peptide Cys 354 -Lys 359 linked by a disulfide bond involving Cys 282 and Cys 354 .

TABLE I N-terminal sequence and electrospray-ionization mass spectrometric data used to determine the disulfide bonds within P. chabaudi AMA-1B
Tryptic peptides are labeled according to their corresponding peak fraction shown in Fig. 3. Peptides from thermolysin subdigestion of tryptic peptide 58 are labelled according to their peak fraction numbers in the first and second dimension RP-HPLC (Fig. 4) a Numbers refer to relative position of cysteines in amino acid sequence. b One-letter abbreviation used for amino acids; peptide yields (picomoles), calculated from the values for phenylthiohydantoin amino acids in the first cycle of the Edman degradation, are shown in parenthesis. Sample load: 18% of each tryptic peptide fraction and 100% of each thermolytic peptide fraction was taken for analysis. c X indicates a glycosylated Asn residue. d This is a partial sequence and consequently does not correspond to the observed molecular ion mass. The calculated mass was determined by extending this sequence C-terminally to the next tryptic cleavage site (i.e. Lys 393 and Lys 453 ). e Multiple ion series consistent with the presence of carbohydrate.

FIG. 3. Preparative scale RP-HPLC purification of P. chabaudi adami (DS strain) AMA-1B tryptic digest.
The remaining digest (120 g) was fractionated by elution from a Vydac C18 (4.6-mm inner diameter ϫ 250 mm) column at a flow rate of 1.0 ml/min using a linear 90-min gradient from 0 to 100%B. Buffer A was 0.05% (v/v) trifluoroacetic acid, and buffer B was 0.05% (v/v) trifluoroacetic acid and aqueous 60% (v/v) acetonitrile. Peptide fractions found to contain disulfide bonded peptides by Edman degradation and electrospray-ionization mass spectrometry are indicated by an asterisk. Peak numbers correspond to those given in Table I. (See "Materials and Methods" for further details).
The observed mass of 2290.3 Da for the peptide in peak fraction 42 corresponded to the single tryptic peptide (Tyr 207 -Lys 225 ) with an internal disulfide bond between Cys 208 and Cys 220 . The observed mass of the peptide in peak fraction 46 (2708.6 Da) was consistent with tryptic peptides Phe 257 -Lys 270 and Lys 360 -Lys 368 disulfide-linked through Cys 265 and Cys 363 . Edman degradation confirmed the predicted sequences for the peptides involved in these three disulfide bonds.
The remaining disulfide-linked peptides that were not predicted by PEPMAP software because they were either glycosylated (peak 25) or aberrant cleavage products (peak 56 and 58) were identified using a combination of mass analysis and Edman degradation.
A broad peak eluting at 69 min in the analytical RP-HPLC was lost with DTT reduction (Fig. 2). There were two well resolved peaks (56 and 58) in the corresponding position on the preparative chromatogram (Fig. 3). Edman degradation of the peak fraction 56 established that two peptides comprising residues Glu 149 -Lys 164 and Thr 191 -Tyr 196 were linked by a disulfide bond involving Cys 162 and Cys 192 . The C-terminal Tyr on peptide Thr 191 -Tyr 196 indicated a chymotrypsin-like cleavage that was responsible for the failure of the PEPMAP program to predict this pair of disulfide bonded peptides. The calculated mass of these two disulfide bonded peptides (2557.04 Da) was in close agreement with the observed molecular ion mass (2556.9 Da).
Edman degradation of fraction 58 revealed two peptide sequences commencing at Asn 369 and Ile 420 , respectively (see Table I). The sequence of the peptide commencing with Ile 420 contained two tryptic sites resistant to cleavage (Lys 426 and Lys 430 ), and the observed molecular ion mass of 6611.5 Da was consistent with a fraction containing peptides with trypsinresistant sites. The calculated mass (6612.52 Da) of two peptides Asn 369 -Lys 393 (Lys 393 being the first potential tryptic site C-terminal to Asn 369 ) and Ile 420 -Lys 453 (Lys 453 being the first potential tryptic site C-terminal to the trypsin-resistant sites Lys 426 and Lys 430 ) linked by three disulfide bonds was in close agreement with the observed molecular ion mass (6611.5 Da). Because it was not possible to determine the connectivities of the cysteine residues in the two peptides in fraction 58, this fraction was digested further using thermolysin (see below).
Of the cysteines not assigned to the disulfide bonds discussed above, one (Cys 247 ) was within a predicted tryptic peptide (Glu 241 -Lys 250 ), which also contained a tryptophan residue (Trp 243 ). A candidate peak fraction (25) containing this peptide was identified by examining the chromatogram monitored at 280 nm (not shown), and Edman degradation established that the tryptophan-containing peptide (Glu 241 -Lys 250 ) was linked to peptide Cys 94 -Lys 99 by a disulfide bond between Cys 94 and Cys 247 .
The observed mass for fraction 25 was more than twice that of the calculated mass of the two disulfide bonded peptides in this fraction (Table I). An inspection of the AMA-1B sequence revealed that Asn 249 , a potential site for N-glycosylation (Asn-Xaa-Ser), is immediately followed by a site for tryptic cleavage (Lys 250 ). A blank cycle 9 of the Edman degradation of 25 was consistent with glycosylation of Asn 249 . Mass analysis of peak fraction 25 revealed ions of double and triple charge with a calculated mass of 3704.3 Da. The observed ion series showed a sequential loss of 162 Da for the nine most distal carbohydrate residues followed by two further consecutive losses of 203 Da to give the expected molecular ion mass of 1841.10 Da for the nonglycosylated disulfide-linked peptide. The loss of 162 and 203 Da is consistent with the loss of hexose-like (Fru, Gal, Glc, and Man) and N-acetylhexosamine-like (GalNAc or Glc-NAc) residues, respectively (22).
Thermolysin Subdigestion of Peak Fraction 58 -The thermolysin digest of fraction 58 was fractionated using rapid narrow bore RP-HPLC (Fig. 4a). Peak fractions 9 and 17 were found by electrospray-ionization mass spectrometry to contain masses corresponding to fragments of the tryptic core peptide. An additional dimension of RP-HPLC, using a sodium chloride/ acetonitrile solvent system, to remove minor contaminants, generated peak fractions 2 and 3 from 9 and 17, respectively (Fig. 4, c and b, respectively). Peak fraction 2 from Fig. 4b was found by Edman degradation to contain the sequence for peptide Leu 376 -Lys 393 disulfide-linked through Cys 389 to Cys 443 of peptide Leu 438 -Phe 445 . The observed mass of these disulfide bonded peptides (2952.5 Da) was in excellent agreement with the calculated value (2953.27 Da). These data are consistent with the postulated C terminus of peptide Asn 369 -Lys 393 discussed in the previous section.
Peak fraction 3 shown in Fig. 4c was found by Edman degradation to contain peptide Ile 422 -Gln 437 linked through two disulfide bonds to peptide Phe 446 -Lys 453 . The unique mass of peptide 3 (2729.7 Da) was consistent with peptides Ile 422 - FIG. 4. First and second dimension RP-HPLC purification of thermolytic fragments from tryptic peptide 58. a, narrow bore RP-HPLC elution profile of tryptic fraction 58 (Fig. 3) after subdigestion with thermolysin. The thermolytic digest was fractionated on a Brownlee RP-300 C8 column (2.1-mm inner diameter ϫ 100 mm) using a linear gradient from 0 to 100% buffer B over 12 min at a flow rate of 0.5 ml/min. Buffer A was 0.05% (v/v) trifluoroacetic acid and buffer B was 0.05% (v/v) trifluoroacetic acid and 60% (v/v) aqueous acetonitrile. b, second dimension narrow bore RP-HPLC of peak fraction 9 shown in a. The peptide mixture was fractionated using a Brownlee RP-300 C8 column (2.1-mm inner diameter ϫ 100 mm) and a linear gradient of 0 to 100% buffer B over 60 min at a flow rate of 0.1 ml/min. Buffer A was aqueous 154 mM NaCl, and buffer B was aqueous 60% (v/v) acetonitrile containing 154 mM NaCl. c, second dimension narrow bore RP-HPLC of peak 17 shown in a. RP-HPLC elution conditions were identical to those described for b. Peak 2 (b) and peak 3 (c) were analyzed by Edman degradation and electrospray-ionization mass spectrometry. The resulting sequence data are reported in Table I

DISCUSSION
The disulfide bond-dependent conformation of AMA-1 is essential for inducing a protective immune response against malaria in animal models. 2 To better understand the specificity of the protective immune response induced by immunization with AMA-1, we have sought to determine the pattern of disulfide bonds in this important vaccine candidate. The parasite antigen is present in low abundance and is only expressed over a narrow time window during the cycle of asexual development in the host erythrocyte, and for this reason we have used recombinant AMA-1 produced using the baculovirus expression system for the disulfide bond assignments. The recombinant antigen used lacked the cytoplasmic and putative transmembrane domains but retained all of the putative ectodomain sequence including all 16 conserved cysteine residues. The cloned DNA construct included the sequence encoding the AMA-1 signal for secretion and the antigen with the signal peptide cleaved was recovered from the medium in which the transfected Sf9 insect cells were cultured. This form of AMA-1 was highly antigenic when reacted with the serum of hyperimmune mice. For this reason, and because the AMA-1 ectodomain was efficiently transported through the secretory pathway of the insect cell and was recovered from the culture supernatant in a glycosylated and stable form, we assumed that the protein was folded with the correct disulfide bonding pattern.
The N-terminal sequencing and mass spectrometric analyses  6. a, a schematic diagram showing the relationship between the sites of amino acid substitution occurring among P. chabaudi adami strains DS and 556KA and P. chabaudi chabaudi strain CB and the disulfide bonds within AMA-1. Sequence data were obtained from footnote. 3 Note that P. chabaudi chabaudi CB AMA-1 has only been partially sequenced (residues 51-377). The sequence shown is that for P. chabaudi DS AMA-1 ectodomain (residues 86 -479). b, a schematic diagram showing the relationship between the sites of amino acid substitution occurring among eleven P. falciparum isolates obtained from Ref. 4. The sequence shown is that of P. falciparum 3D7 AMA-1 ectodomain (residues 141-538). Black circles represent disulfide-linked Cys residues, and shaded circles represent the location of amino acid substitutions within the primary sequence. The disulfide bond arrangement between the two Cys-Xaa-Cys motifs in domain III has not been resolved; however, the connectivities portrayed are considered the most likely. of tryptic and thermolytic peptides identified eight intramolecular disulfide bonds involving all 16 conserved cysteine residues within AMA-1. Structural analysis of peptides from those peaks identified as containing disulfide bonded peptides in the nonreducing chromatographic profile revealed no heterogeneity in the cysteine connectivity pattern. However, an ambiguity in the cysteine connectivities remains to be resolved: it is clear that Cys 431 and Cys 433 link with Cys 448 and Cys 450 , but it is not clear whether 431 links with 448 and 433 with 450 or vice versa. This ambiguity is difficult to resolve, because these four cysteines are found in two Cys-Xaa-Cys sequences lacking obvious chemical or enzymatic cleavage sites.
The arrangement of the disulfide bonds (Fig. 5) suggests that the ectodomain of AMA-1 may be composed of three subdomains characterized by three, two, and three disulfide bonds with masses of approximately 19, 13, and 13 kDa, respectively. Searches of available protein and DNA sequence data bases have failed to identify other proteins with any significant sequence relationship to AMA-1. However, the putative subdomain structure for AMA-1 enables the disulfide bond architecture to be compared with that of other proteins. This has not been fruitful for the subdomains I and II, but although the three disulfide bonds in subdomain III have not been fully resolved, one of the two linkage patterns possible closely resembles the cystine knot motif found in growth factors (23,24). In such structures two of the disulfide bonds are involved in localized cyclization of the sequence, via two Cys-Xaa N -Cys motifs, whereas, the third disulfide bond passes through the ring structure. Also common to these cystine knot structures are extended segments of twisted anti-parallel ␤-strand, which lie between the first to second and fourth to fifth cysteines within the motif (25). Structural analyses of a series of AMA-1 homologues, using PredictProtein Server (26 -28), indicated a high probability of ␤-strand in analogous locations of the cystine knot-like structure of AMA-1 (data not shown).
The distribution of mutations within the sequence of AMA-1 is not uniform. The shaded circles in Fig. 6a show the location of amino acid substitutions that occur among the AMA-1 sequences of the P. chabaudi adami strains DS and 556KA and P. chabaudi chabaudi strain CB. 3 Each of the putative subdomains defined by the disulfide bond connectivities exhibits some sequence diversity; however, subdomain I is the most diverse containing 58% of all sites where amino acid substitutions occur in the ectodomain of P. chabaudi AMA-1. A second, smaller hypervariable region also exists around the eighth cysteine residue (Cys 282 ) in subdomain II. The relationship between sites of amino acid substitution and the disulfide bonds differs among the subdomains. In subdomain I, the most diverse of the subdomains, the majority of the amino acid substitutions occur between the first and third cysteine residues with the disulfide bonded "core" of this subdomain relatively free of amino acid substitutions (Fig. 6a). However, in subdomains II and III (Fig. 6a) amino acid substitutions that are well separated from each other within the primary sequence, are close to cysteine residues, and in the folded protein will be clustered as a result of disulfide bond formation. A similar relationship between the pattern of mutations and disulfide bonds is apparent in P. falciparum AMA-1 (Fig. 6b) with the clustering of the amino acid substitutions in subdomains II and III, as a result of the disulfide bond formation, being more pronounced than that seen in P. chabaudi AMA-1. Thus, these amino acid substitutions may define conformational, disulfide bond-dependent epitopes that are recognized by protective immune responses.
Consistent with this conclusion, we have recently established that some of the disulfide bonded tryptic fragments are recognized by antibodies induced by P. chabaudi infection in mice. 4 These antibodies react with refolded AMA-1 but not the reduced and alkylated antigen. We are currently producing the putative subdomains in an E. coli expression system and synthesizing disulfide bonded peptides based on the cysteine connectivity pattern within AMA-1 in an effort to establish which region(s) of the molecule are most relevant for the generation of a protective immune response.