Induction of proinflammatory cytokine responses by glycosylphosphatidylinositols (GPIs) of intraerythrocytic Plasmodium falciparum is believed to contribute to malaria pathogenesis. In this study, we purified the GPIs of P. falciparum to homogeneity and determined their structures by biochemical degradations and mass spectrometry. The parasite GPIs differ from those of the host in that they contain palmitic (major) and myristic (minor) acids at C-2 of inositol, predominantly C18:0 and C18:1 at sn-1 and sn-2, respectively, and do not contain additional phosphoethanolamine substitution in their core glycan structures. The purified parasite GPIs can induce tumor necrosis factor α release from macrophages. We also report a new finding that adults who have resistance to clinical malaria contain high levels of persistent anti-GPI antibodies, whereas susceptible children lack or have low levels of short-lived antibody response. Individuals who were not exposed to the malaria parasite completely lack anti-GPI antibodies. Absence of a persistent anti-GPI antibody response correlated with malaria-specific anemia and fever, suggesting that anti-GPI antibodies provide protection against clinical malaria. The antibodies are mainly directed against the acylated phosphoinositol portion of GPIs. These results are likely to be valuable in studies aimed at the evaluation of chemically defined structures for toxicity versus immunogenicity with implications for the development of GPI-based therapies or vaccines.
The rapid spread of drug-resistant Plasmodium falciparum and the lack of an effective malaria vaccine present an urgent need for alternative approaches to prevent malaria infection and pathogenesis 1. It is becoming increasingly clear that complete protection against malaria infection and illness requires multifactorial immunity 2. Currently, most vaccine efforts are aimed at immunity against infection (antiparasitic) by targeting the parasite proteins 2,3,4,5. However, it has long been suggested that immunity against severe malaria is partly antiparasitic and partly antitoxic (toxic effects in response to parasite factors). The majority of the adults in malaria endemic areas have resistance to severe malaria. However, most children <4 yr of age are susceptible despite exposure to high malaria transmission, which can produce high levels of antibodies against protein antigens including merozoite surface protein (MSP)1-1, erythrocyte membrane antigen (EBA)-175, and apical membrane antigen 1 (AMA)-1. Although antibody responses against parasite proteins correlate with protection against parasitemia (Branch, O.H., unpublished results), resistance to malaria illness is independent of parasitemia levels. This agrees with the resistance of adults and older children to malaria pathology even though they can develop significant parasitemia 6; conversely, severe illness can occur at relatively low-density parasitemias independent of antibody response against parasite proteins 7,8,9. The factors associated with the resistance to clinical disease (antidisease immunity) have not been established; understanding these would lead to alternative approaches for malaria control. In this regard, parasite glycosylphosphatidylinositols (GPIs) appear to offer new opportunities.
GPIs are a distinct class of glycolipids found ubiquitously in eukaryotic cells and implicated in several biological responses 10,11,12. GPIs are particularly abundant in parasites, where they are found as free lipids and attached to proteins. In intraerythrocytic P. falciparum, GPIs represent the major glycoconjugates. Several functionally important parasite proteins, including MSP-1, MSP-2, and MSP-4, are anchored to the cell membranes through GPI moieties 13,14,15,16,17. Recently, we have shown that P. falciparum synthesizes GPIs in a developmental stage–specific manner and that GPI biosynthesis is crucial for the development and survival of the parasite 18. The enzyme specificity of some key steps of parasite GPI biosynthesis differs significantly from those of the host, suggesting the possibility of targeting the parasite GPI structures for the development of antiparasitic drugs. However, detailed structures of parasite GPIs have not been determined. Although the structures of glycan cores have been established using metabolically labeled GPIs 19,20, details regarding the nature of various acyl residues and other possible substitutents were not clear 21. Determination of a detailed structure requires isolation of pure GPIs which, in the case of P. falciparum, is a challenge due to the difficulty in obtaining adequate amounts of parasites free of host cell components. In this study, we were able to successfully purify P. falciparum GPIs to homogeneity and establish their structures.
It has long been believed that malaria pathology is due to factors endogenously produced in response to parasite toxins. Several studies have shown that malaria pathology is at least in part due to parasite toxic factors that can induce TNF-α and other cytokines, which could then lead to clinical effects including fever, hypoglycemia, dyserythropoiesis, and vascular damage in the lungs and brain 22,23. This agrees with the elevated levels of TNF-α in patients with lethal cerebral malaria 24 and the ability of anti–TNF-α antibodies to prevent lethal cerebral pathology in mice 25.
P. falciparum GPIs have been identified as malaria pathogenicity factors based on their ability to induce inflammatory cytokines in macrophages and endothelial cells and cause symptoms reminiscent of acute malaria infection in experimental animals 26,27,28,29. Schofield et al. 26 have shown that parasite fractions enriched with GPIs can induce TNF-α and IL-1 in macrophages; in mice, GPIs can cause transient pyrexia, hypoglycemia, lethal cachexia, and even death in d-galactosamine (GalN)-sensitized animals. Schofield et al. have also shown that P. falciparum GPIs exert toxic effects through the expression of TNF-α, IL-1, inducible nitric oxide synthase (iNOS), and endothelial cell adhesion molecules by activating nuclear factor κB transcription factors 27,28,29. As P. falciparum–infected erythrocytes are sequestered in specific organs, the local high concentrations of toxic responses to the parasite GPIs can affect vital physiologic functions and cause severe illness. Recent studies have shown that GPIs from a Trypanosoma cruzi mucin can also induce proinflammatory cytokines 30. The antagonists of GPI-mediated signaling and a monoclonal antibody against P. falciparum GPIs can block the induction of toxic responses 27,28,29, suggesting that GPI-based therapy is possible.
Because P. falciparum GPIs are pathogenicity factors, we hypothesized that adults in malaria endemic areas should have GPI-specific protective immunity. We tested this hypothesis by analyzing the anti-GPI antibody response in sera from a longitudinal cohort study and in sera of a large group of adults from Western Kenya. The data demonstrate for the first time that people living in malaria endemic areas elicit a parasite GPI–specific IgG response in an age-dependent manner; although adults and older children have high levels of antibodies, malaria-susceptible children either lack or have only very low levels of short-lived antibodies. The results also suggest the involvement of anti-GPI antibodies in protection against malaria pathogenesis.
Materials And Methods
Human blood and serum were purchased from Interstate Blood Bank. RPMI 1640, DME, and cell culture reagents were from Life Technologies. Gelatin, bee venom phospholipase A2 (1,800 U/mg), standard phospholipids, and saponin were from Sigma-Aldrich. Silica Gel 60 high performance thin-layer chromatography (HPTLC) plates were from either EM Science or Whatman. Pronase was purchased from Calbiochem. Aspergillus saitoi α-mannosidase (400 mU/mg) and jack bean α-mannosidase (30 U/mg) were from Oxford Glycosystems. Poly(isobutyl methacrylate) was procured from Polysciences, Inc. Horseradish peroxidase (HRP)-conjugated goat anti–human IgG (H and L chains) and 2,2′-azino-bis(3-ethylbenz-thiazoline-6-sulfonic acid) (ABTS) reagent were from Kirkegaard & Perry Laboratories. Microtiter plates were from Dynex Technologies. [6-3H]glucosamine (GlcN; 23 Ci/mmol), and [3H]arachidonic (202 Ci/mmol), [3H]myristic (55 Ci/mmol), and [3H]oleic acids (8 Ci/mmol) were from Amersham Pharmacia Biotech. [3H]Linoleic (60 Ci/mmol), [3H]palmitic (60 Ci/mmol), [3H]palmitoleic (60 Ci/mmol), and [3H]stearic acids (60 Ci/mmol) were from American Radiolabeled Chemicals. 125I-labeled goat anti–human IgG (8.66 μCi/μg) and En3Hance were from NEN Life Science Products. Murine macrophages (J774A.1) were from American Type Culture Collection. HPLC grade solvents were used throughout the study.
Cell Culture and Isolation of Parasites.
Intraerythrocytic P. falciparum (FCR-3 strain) was cultured in RPMI 1640 using human O-type erythrocytes and 10% O-positive human serum at 3–4% hematocrit 18,19. Cultures were routinely synchronized with 5% sorbitol 31 and tested for mycoplasma 32; only contamination-free cultures were used.
Cultures with 20–30% parasitemia were harvested at midschizont stage. Packed erythrocytes (20 ml) suspended in 180 ml of 0.65% gelatin in culture medium were incubated at 37°C for 30 min 33. The enriched infected erythrocytes (70–80% parasitemia) in the supernatant were recovered and lysed with 0.015% saponin in Trager's buffer 34. The suspension was passed through a 26-gauge needle to solubilize the erythrocyte debris and then centrifuged. The parasites were suspended in the above buffer, layered on 5% BSA, and centrifuged 35. The packed parasites (∼4 ml) were washed with the buffer and stored at −80°C until used.
Metabolic Labeling of the GPIs.
Parasites were radiolabeled in medium containing 5 mM glucose and [3H]GlcN (50 μCi/ml; reference 19). Labeling with 3H–fatty acids (50 μCi/ml) was performed in medium containing 2% serum and 20 mM glucose.
Isolation of GPIs.
All procedures were carried out using acid-washed, siliconized glassware to prevent surface adsorption of GPIs, high quality solvents, and sterile water and buffers. Parasites (10 ml) were lyophilized and extracted three times with 50 ml of chloroform/methanol (2:1, vol/vol) to remove nonglycosylated lipids. The pellet was extracted five times with 50 ml of CMW (chloroform/methanol/water, 10:10:3 [vol/vol/vol]), dried, and then partitioned between water and water-saturated 1-butanol. The organic layer was dried and the residue was extracted with 80% aqueous 1-propanol. Finally, GPIs were purified by HPLC and HPTLC.
To isolate GPIs that are linked to parasite proteins, the delipidated parasite pellet was digested with pronase (50 U/ml) in 50 ml of 100 mM NH4HCO3, 1 mM CaCl2, pH 8.0, at 37°C for 24 h. The released GPIs (designated as amino acid–linked GPIs) were extracted with water-saturated 1-butanol, washed with water, dried, and purified by HPLC. Control erythrocyte membrane debris, saponin-lysate, and ghosts were similarly extracted and fractionated.
HPLC Purification of GPIs.
The GPIs (∼10 μg plus 400,000 cpm of [3H]GlcN-labeled GPIs) were chromatographed on a C4 reversed phase Supelcosil LC-304 HPLC column (4.6 × 250 mm, 5 μm particle size; Supelco) using a linear gradient of 20–60% aqueous 1-propanol containing 0.1% TFA over a period of 80 min and held for 30 min at a flow rate of 0.5 ml/min 36. Fractions (1.0 ml) were collected, and elution of GPIs was monitored by measuring radioactivity. Aliquots (0.5 μl) were also assayed by ELISA with Kenyan adult sera. Glycolipids extracted from control erythrocyte membrane debris, total lysate, and ghosts were similarly chromatographed, and fractions were analyzed for reactivity with Kenyan sera.
Purification and Analysis of GPIs by HPTLC.
The HPLC-purified GPIs (5 μg) were applied onto 10 × 10–cm plates as continuous streaks. Parallel spots with [3H]GlcN-labeled GPIs (50,000 cpm) were used for monitoring GPI bands by fluorography using En3Hance 18,19. The plates were developed with CMW (10:10:2.5, vol/vol/vol). GPIs from the plates were extracted with CMW (10:10:3, vol/vol/vol), dried, dissolved in water-saturated 1-butanol, and washed with water.
In separate experiments, HPTLC plates were scraped (0.5-cm-width fractions), and GPIs were extracted and analyzed by ELISA using Kenyan adult sera. Glycolipids from control erythrocyte membrane debris, total lysate, and ghosts were similarly analyzed.
HPLC- and HPTLC-purified, free and amino acid–linked GPIs (4–5 μg each), and hydrofluoric acid (HF)-released inositol-acylated carbohydrate moiety (2–3 μg), and diacylglycerol moiety (1–2 μg) were treated with 8 M anhydrous methanolic KOH for 2 h at 120°C. After acidification with cold dilute HCl, the fatty acid methyl esters were extracted with chloroform. Phospholipase A2–released fatty acids were esterified with methyl iodide in DMSO/solid NaOH. Gas chromatography–mass spectrometry (GC-MS) was performed on a ThermoQuest GCQplus ion trap electron ionization quadrupole mass spectrometer (Finnigan) using DB-5 capillary column (30 m × 0.25 mm interior diameter with 0.25-μm film; Alltech). The temperature was held at 75°C for 1 min, raised to 150°C at 1°C/min, and then to 325°C at 8°C/min. GC retention times and mass spectra were compared against standards.
The free and amino acid–linked GPIs (∼2 μg each) were hydrolyzed with 2.5 M TFA at 100°C for 5 h, and the hydrolysates were analyzed for mannose by HPLC 37.
Matrix-assisted laser-desorption/ionization time of flight mass spectrometry analysis was performed using a Kratos analytical MALDI-4 mass spectrometer equipped with a nitrogen laser at 20-kV accelerating voltage 38. Spectra were acquired with a time-delayed extraction and were the average of 50 laser shots. The matrix was saturated with α-cyano-4-hydroxycinnamic acid in 50% ethanol. The mass accuracy was within 2 daltons.
TNF-α Induction by GPIs.
Murine macrophages (1.5 × 106 cells/well in 96-well plates) in DMEM and 10% fetal bovine serum were stimulated with 0.03–0.5 μM HPTLC-purified, free and amino acid–linked GPIs (see Fig. 1 B, lanes 2 and 4, respectively). TNF-α in the culture supernatants was measured by a quantitative sandwich ELISA using the TNF-α estimation kit (R&D Systems). Experiments were performed in triplicate with three separate batches of GPI preparations.
HPLC-purified, free GPIs dissolved in methanol were coated (0.25–32 ng/well) onto 96-well microtiter plates. After evaporation of methanol at 37°C, the wells were blocked with 0.5% casein in 50 mM Tris-buffered saline, pH 7.4 (TBS-casein), and incubated with serially diluted sera (1:100 to 1:64,000) in TBS-casein containing 0.05% Tween 20. Initially, randomly selected 20 Kenyan and 2 USA adult sera were assayed. The bound antibodies were measured by HRP-conjugated goat anti–human IgG and ABTS substrate. All other sera were analyzed at 1:200 or 1:400 dilution by coating GPIs at 0.5–2 ng/well.
Competitive Inhibition ELISA.
Sera were diluted 1:200 in TBS-casein containing 0.05% Tween 20 and incubated at room temperature for 30 min with various concentrations of phosphatidylinositols (PIs), phosphoglyceride (PG), cardiolipin (CL), or GPI, and then ELISA was performed as described above.
The HPLC-purified, free and amino acid–linked GPIs (100 ng each) were chromatographed on HPTLC plates. The plates were soaked in 0.1% poly(isobutyl methacrylate) in hexane, dried, and incubated for 2 h at room temperature with TBS containing 1% BSA and then for 2 h in 1:100 diluted sera 39. The plates were washed with cold phosphate-buffered saline, incubated with 125I-labeled goat anti–human IgG (5 μCi/ml) for 1 h, washed, dried, and exposed to x-ray film.
Nitrous Acid Treatment.
The HPLC-purified, free GPIs (20 μg) suspended in 150 μl 0.2 M NaOAc, pH 3.8, were treated with 150 μl 1 M NaNO2 at room temperature for 24 h 40. The released PI moieties were extracted with water-saturated 1-butanol. The glycan moieties were recovered by chromatography on a Bio-Gel P-4 column (1 × 90 cm) in 100 mM pyridine, 100 mM HOAc, pH 5.2. These were used to assess seroreactivity.
Treatment with HF.
The HPLC-purified, free GPIs (2 μg plus 500,000 cpm of [3H]GlcN-labeled GPIs) were treated with 50% aqueous HF (50 μl) in an ice bath for 48 h 40. The reaction mixture was neutralized with LiOH and extracted with water-saturated 1-butanol. The organic layer was washed with water, dried, and the carbohydrate moiety was purified by HPLC.
Treatment with Mannosidase.
Treatment with Phospholipase A2.
HPLC-purified GPIs (2 μg) in 100 μl 100 mM Tris-HCl, 10 mM CaCl2, pH 7.5, were treated with bee venom phospholipase A2 (2,400 U/ml) at 37°C for 18 h. The digest was extracted with water-saturated 1-butanol and purified by HPLC; the retention time of lyso-GPIs is 75.5 min.
Analysis of Cohort Sera for Anti-GPI Antibodies.
Sera were obtained from subjects of a longitudinal malaria project performed in a holoendemic, rural region of Western Kenya within the context of the Asembo Bay Cohort Project (ABCP ). Studies done in the subjects' households show that from March to May the entomological inoculation rate (EIR) was ∼30 infected bites per person per month, whereas during the drier months the EIR was ∼4. Pregnant mothers were identified and, after written consent, they, their delivered infants, and older children living in the same household were enrolled. The mothers and children were followed every 2 wk for ∼4 yr, and clinical parameters, axillary temperature, and illness were recorded. Blood samples were taken monthly and parasite and hemoglobin density were measured. More than 90% of detected parasitemias contained P. falciparum species.
Sera of 100 7–8-yr-old children and 50 nonpregnant adults, collected during February through August 1993, were analyzed at two time points, spaced 1–3 mo apart. For data analysis, we selected two time points for the young children, in the same manner as the older individuals, when they were ∼0.5, 1.5, 2.5, and 3.5 yr old. The study was approved by the Centers for Disease Control and Prevention and Kenyan Institutional Review Boards.
Analysis of the Relationship between Serum Anti-GPI Antibodies and Resistance to Malaria Pathogenesis.
Analysis for correlation of anti-GPI antibody response with either hemoglobin or temperature as the dependent variable was done with the general linear model, Generalized Estimating Equations 41. The variables considered in the analyses were (a) anti-GPI antibody responder category (positive, intermittent, or negative), (b) parasite density, log10(n + 1/μl), (c) age, and (d) antimalaria drug treatment. We controlled for the repeated measures on each patient. The estimated effect of positive and negative antibody responder category are given relative to that of the intermittent antibody responder category.
Isolation and Purification of GPIs from Intraerythrocytic P. falciparum.
As contamination with glycolipids from host erythrocytes and mycoplasma is a major concern in isolating GPIs from intraerythrocytic P. falciparum, we used rigorous protocols. First, parasite cultures were routinely tested for mycoplasma, and parasite-infected erythrocytes were enriched from cultures with 25–30 to 70–80% parasitemia to minimize erythrocyte components. Second, the parasites were released from infected erythrocytes by mild saponin lysis, washed, and purified by density centrifugation to remove erythrocyte membrane debris; metabolically [3H]GlcN-labeled cultures revealed that this procedure did not release parasite GPIs. These procedures eliminated all of the erythrocyte components. Third, the purified parasites were differentially extracted to remove most nonglycosylated lipids, and the extract containing the free GPIs (not linked to proteins) was subjected to solvent fractionation to eliminate soluble hemozoin and other colored components. Finally, the GPIs were purified by successive fractionation using HPLC and HPTLC. The GPIs that are linked to parasite proteins were isolated after exhaustive digestion of the delipidated parasites with pronase, and are referred to as amino acid–linked GPIs. The yields of the HPLC-purified free and amino acid–linked GPIs were 100–120 and 25–30 μg per 10 ml of packed purified parasites, respectively, based on the mannose content.
The parasite GPIs were purified by HPLC using a C4 reversed phase column, which separated matured GPIs from GPI intermediates (Fig. 1 A). The elution of GPIs was monitored by radioactivity, which overlapped with the reactivity of the GPIs to human sera containing anti-GPI antibodies (Fig. 1 A). The compounds isolated from the extracts of erythrocyte debris, lysates, and ghosts were fractionated in parallel with the purification of parasite GPIs by HPLC (Fig. 1 A) and HPTLC and analyzed as controls. These did not react with Kenyan adult sera (Fig. 1 A), suggesting that the activity is specific to parasite GPIs.
The GPIs were further purified by HPTLC; they migrated as broad major bands and in some instances separated into two major bands (Fig. 1 B). The heterogeneity of GPIs is due to the variation in the fatty acid composition of individual GPIs (see below). The HPTLC-purified GPIs (Fig. 1 B, lanes 2 and 4) contained mannose and GlcN in molar ratios of ∼4:1 and were susceptible to nitrous acid, HF, and alkali (not shown). Fatty acid compositional analysis of HPLC-purified GPIs showed the presence of myristic (∼4%), palmitic (∼33%), stearic (∼26%), oleic (∼28%), cis-vaccenic (∼2%), linoleic (∼3%), arachidic (∼2%), and behenic (∼1%) acids, and 1–2% of unidentified acids. The presence of cis-vaccenic acid in the GPIs agrees with the ability of P. falciparum to synthesize two isomeric C18:1 acids (oleic and cis-vaccenic acids ).
Structural Analysis of P. falciparum GPIs.
The structures of the purified P. falciparum GPIs were determined by specific chemical and enzymatic degradation studies using [3H]GlcN- and 3H–fatty acid-labeled GPIs and by mass spectrometry.
The [3H]GlcN-labeled GPIs were characterized by subjecting them to various degradative procedures. Treatment with jack bean α-mannosidase quantitatively converted the [3H]GlcN-labeled GPIs into species with three mannose residues, similar to the results recently reported 18, suggesting that the distal fourth mannose residue does not contain any substituent. The [3H]GlcN-labeled GPIs were dephosphorylated with HF and then deaminated and reduced with NaBH4. The neutral glycan core, thus obtained, on HPTLC migrated as a single band with a mobility relative to the solvent front (Rf) value identical to previously characterized Man4-2,5-anhydromannitol (AHM) 19. Treatment of the glycan core with A. saitoi α-mannosidase shifted the Rf value to that of authentic Man2-AHM, and the product of jack bean α-mannosidase digestion comigrated with authentic AHM (not shown; 19). These results are consistent with the previously established structure of Manα1-2Manα1-2Manα1-6Manα1-4GlcN for the glycan core of P. falciparum GPIs 19,20.
The mass spectrum of the purified, free GPIs (upper half portion of the HPTLC band in Fig. 1 B, lane 2) contained prominent molecular ions (M-H)− at m/z 2,006.3 and 2,034.3 and minor ions at m/z 1,978.3, 2,062.3, and 2,090.3 (Fig. 2 A). The observed difference of 28 mass units (in some fractions 26 units) between the consecutive molecular ions suggests that the GPIs contain mixtures of homologous fatty acids differing by two carbons. The proportions of the GPI species varied considerably depending on the regions of HPTLC band analyzed (compare Fig. 2A and Fig. B), suggesting heterogeneity with respect to acyl substitutents. Mass spectrometry of the control materials did not show molecular ions comparable to the parasite GPIs (not shown). These results show that the purified compounds are GPIs of the parasite, not of erythrocytes.
Treatment of free GPIs with phospholipase A2 resulted in the marked loss of (M-H)− ions of intact GPIs and the appearance of a new set of ions, each 264.5 mass units lower (Fig. 2 B), indicating that the GPIs contain a C18:1 acyl substituent at the sn-2 position. The spectra of some GPI fractions treated with phospholipase A2 also contained significant levels of ions 262.5 mass units lower than the parent ions, suggesting the presence of GPIs with a C18:2 acid at sn-2. GC-MS analysis of phospholipase A2–released fatty acids demonstrated the presence of predominantly oleic acid with minor amounts of cis-vaccenic and linoleic acids. These data together with the total fatty acid composition indicate that the parasite GPIs contain mainly oleic (∼85%), cis-vaccenic (∼6%), and linoleic (∼9%) acids at sn-2.
As reported previously 21, the GPIs were sensitive to GPI-specific phospholipase D, but completely resistant to PI–phospholipase C (data not shown), suggesting the presence of an acyl substituent on C-2 of inositol. To identify this substituent, the GPIs were treated with HF and the released inositol-acylated carbohydrate moiety was purified by HPLC as a single symmetrical peak with a retention time of 47 min. Fatty acid compositional analysis indicated that the HF-released inositol-acylated carbohydrate moiety contained palmitic acid (∼90%) and myritsic acid (∼10%). Positive ion–mode mass spectrum showed (M+Na+) ions at m/z 1,224.4 and 1,252.7 (Fig. 2 C), which were assigned to Man4-GlcN-inositol with myristoyl and palmitoyl substitution, respectively, on inositol. The signals at m/z 1,228.7 and 1,266.8 represent (M+H+) and (M+K+) ions of the carbohydrate with palmitoyl substitution. Of several mass spectra recorded, some (not shown) gave a reasonable signal at m/z 1,200.4, which was assigned to the (M+H+) ion of the carbohydrate with myristoyl substitution. A significant portion of the HF-released glycan moieties retained a phosphate ester group as indicated by the negative ion mode spectrum, which showed molecular ions at m/z 1,280.0 and 1,308.1 (Fig. 2 D). The phosphate ester group that survived HF treatment is likely the one linked to inositol because the adjacent protonated GlcN residue will render the hydrolysis of the phosphate ester bond to be kinetically slower. These results established that the inositol residue is substituted with palmitate (∼90%) and myristate (∼10%). The diacylglycerol moiety released on treatment of GPI with HF contained predominantly C18:0 and C18:1 (two isomers), minor proportions of C14:0, C16:0, C18:2, C20:0, and C22:0, and unidentified acids. As C18:1 and C18:2 are present at sn-2 (see above), the remainders of the above acids should be located at sn-1.
The mass spectra of the HPTLC-purified, amino acid–linked GPIs (fraction I in Fig. 1 B, lane 4) containing one amino acid after pronase digestion showed prominent molecular ions that are 87 mass units higher than those of free GPIs (compare Fig. 2E and Fig. F, with A), indicating that these GPIs contain a Ser residue. GPIs with attached Asn were also evident from (M−H)− ions at m/z 2120.3, 2148.3, and 2176.3 in the mass spectrum (not shown) of fraction II (Fig. 1 B, lane 4). This agrees with the predicted amino acids in GPI-anchored surface proteins 13,14,15, MSP-1, MSP-2, and MSP-4 (Fig. 2 G). Phospholipase A2 treatment caused reduction by 264.5 mass units in the m/z of each of the molecular ions (Fig. 2 F), confirming the presence of predominantly C18:1 acyl residues at sn-2.
The acyl substituents in the GPIs were also studied by metabolic labeling with [3H]myristic, [3H]palmitic, [3H]palmitoleic, [3H]stearic, [3H]oleic, [3H]linoleic, and [3H]arachidonic acids. All except arachidonate were incorporated into the parasite GPIs (Fig. 3 and data not shown). Treatment of the GPIs labeled with [3H]myristic and [3H]palmitic acids with phospholipase A2 showed the presence of incorporated label only in the lyso-GPIs; the released fatty acids were not radioactive (Fig. 3). Upon treatment with HF, the radioactivity was found to be present on both the diacylglycerol and inositol carbohydrate moieties (not shown). In the case of [3H]stearic acid, the radioactivity was incorporated only to the diacylglycerol moiety (not shown). These results agree with the presence of saturated fatty acids at sn-1 and at C-2 of inositol, but not at sn-2, established by mass spectrometry. The data also suggest that the parasite does not incorporate saturated fatty acids as unsaturated acids to the GPIs they synthesize or vice versa. Treatment of [3H]palmitoleic, [3H]oleic, and [3H]linoleic acid–labeled GPIs with phospholipase A2 released >90% of the incorporated activity as free acid; the lyso-GPIs contained negligible radioactivity (Fig. 3). Treatment with HF showed that most of the incorporated radioactivity was on the diacylglycerol moiety, and the inositol carbohydrate moiety was not radioactive (not shown). These results agree with the presence of unsaturated fatty acyl residues at sn-2 as indicated by mass spectrometry of phospholipase A2–treated GPIs. The incorporation of [3H]palmitoleic acid to sn-2 is most likely after its conversion to cis-vaccenic acid; this agrees with the presence of cis-vaccenic acid at sn-2 of the GPIs (see above) and the ability of the parasite to synthesize this acid 42.
Based on the above data, the structure shown in Fig. 4 is proposed for P. falciparum GPIs.
TNF-α–inducing Activity of P. falciparum GPIs.
Previous studies have shown that P. falciparum GPIs can induce the expression of proinflammatory cytokines in macrophages 27,28,29. To determine if highly purified GPIs can induce biologic activity, cultured murine macrophages were incubated with 0.03–0.5 μM purified GPIs, and the amount of TNF-α in the culture medium was measured. Consistent with a previous finding 26, the purified parasite GPIs induced TNF-α (30–450 pg/ml) in a dose-dependent manner (not shown).
Analysis for the Natural Anti-GPI Antibody in Humans.
To examine whether people living in malaria endemic areas express naturally elicited anti-GPI antibodies, we analyzed sera from 300 adults in Western Kenya and 50 USA adults not exposed to malaria. In ELISAs performed with a coating of 2 ng/well of HPLC-purified, free GPIs, all Kenyan adult sera exhibited moderate to high levels of GPI-specific antibodies, predominantly IgGs, whereas all USA adults completely lacked anti-GPI antibodies (Fig. 5). The anti-GPI antibody activity of Kenyan sera was dose dependent and saturable (Fig. 5). There was no significant difference in levels of GPI-specific antibody reactivity between HPLC-purified free and amino acid–linked GPIs, indicating that the immunoreactivity is not due to contamination by parasite proteins. More than 85% of Kenyan adult sera also contained lower but significant levels of GPI-specific IgM antibodies (Naik, R.S., unpublished results). The erythrocyte membrane debris, saponin lysate, and ghosts were nonreactive to the infected Kenyan sera (data not shown).
When ELISA was performed using several commercially available phospholipids, PIs from bovine liver and soybean that lack acyl substituent on inositol, PGs, and CL, Kenyan adult sera showed a low level of activity: ∼5–15% of that observed with P. falciparum GPIs. At all coating concentrations tested (2–50 ng/well), these compounds showed similar low levels of antibody reactivity. Prior incubation of Kenyan sera with 2.5–20 ng/ml of PIs, PGs, or CL, and then ELISA using plates coated with HPLC-purified free GPIs showed in all cases ∼5–15% lower antibody binding activity, irrespective of the concentrations of lipid used (Fig. 6). However, prior incubation with the parasite GPIs inhibited binding by up to 75% in a dose-dependent manner (Fig. 6). Together, these results suggest that Kenyan adult sera have low levels of reactivity to common acylated phosphoglycerols. This is not surprising because of the polyclonal nature of GPI-specific antibodies, which are expected to contain antibodies to the phosphoglycerol portion of GPIs that are common to this class of molecules.
TLC immunoblotting of the HPTLC-purified GPIs confirmed the specificity of the seroreactivity: adult Kenyan sera reacted with the GPIs, whereas control sera from USA adults were nonreactive (Fig. 7). Furthermore, the Kenyan adult sera were nonreactive to bovine liver and soybean PIs that lack an acyl substituent on inositol, PGs, and CL (Fig. 7, and data not shown); the apparent lack of reactivity to these lipids on TLC plates could be due to a low level of sensitivity. Furthermore, treatment of the GPIs with HNO2 shifted the immunoreactivity to the position of the PI moiety, suggesting that this portion of the molecule is antigenic (Fig. 7).
The identity of the antigenic part of GPIs was further confirmed by inhibition of seroreactivity to intact GPIs using carbohydrate and lipid moieties of GPIs isolated after HNO2 fragmentation. The carbohydrate moiety inhibited antibody binding by only ∼5% at a 10-fold higher concentration compared with that of the coated, intact GPI. The lipid moiety, at the same coating concentration as that of intact GPIs, showed >70% antibody-binding activity, indicating that the PI moiety is the antigenic structure. These results suggest that acylated inositol is the major moiety that recognizes the naturally elicited anti-GPI IgGs.
Anti-GPI Antibody Response and Acquired Resistance to Malaria Pathogenesis.
To determine whether the susceptibility of young children in malaria endemic areas is related to the absence of GPI-specific antibodies, sera taken every month after birth for 4 yr from a cohort of 48 children were analyzed 8. The results were compared with those of sera from 100 siblings (7–8 yr) and 50 nonpregnant mothers (20–25 yr) (Fig. 8 A). Malaria parasitemia and clinical parameters (hemoglobin and fever) were followed every 2 wk in this community-based cohort. Whereas a case-control study design compares extremes, this nonbiased design determined if an association between anti-GPI antibodies and malaria pathogenesis was detectable at a population level. We carefully controlled for P. falciparum transmission, antimalaria drug treatment, and age. By sampling the same child over 4 yr, we found that, within an individual, anti-GPI antibody responses correlate with protection against malaria-attributable hemoglobin loss and febrile illness (see below).
We performed ELISA on sera from each child on all monthly samples and for siblings and adults at two time points spaced 1–3 mo apart to determine the anti-GPI antibody level and the persistence of the antibody response. For comparison and correlation analysis, results of four time points (0.5, 1.5, 2.5, and 3.5 yr old), based on the sampling scheme employed for the 7–8-yr-olds and the mothers, were used for children (Fig. 8 A). The data show that ∼50% of the children under 2 yr lacked anti-GPI antibodies (negative responders, defined in the legend to Fig. 8), ∼40% had short-lived antibody responses (lasted in the circulation for only 1–3 mo, intermittent responders), and only ∼10% had persistent (long-lived) antibody responses (positive responders) (Fig. 8 A). In contrast, ∼75% of the 7–8-yr-old children and all adults exhibited a persistent anti-GPI antibody response. The level of anti-GPI IgG increased with increasing age from infancy to adulthood, whereas mean hemoglobin density increased and fever (temperature) decreased with age up to 7–8 yr of life (Fig. 8 B). Thus, the presence of long-lived, high levels of anti-GPI antibodies in sera parallels the naturally acquired resistance against malaria pathogenesis.
We found that anti-GPI antibody response, age, and parasite density were independently associated with hemoglobin (P < 0.0147). Age influenced the anti-GPI antibody responder category; however, it was not significant (P < 0.0581), suggesting that the association between the antibody responder category (as defined in the legend to Fig. 8) and hemoglobin was not just a reflection of age. For example, using a general linear model, it was estimated that 6-mo-old children in the positive antibody responder category had 2.37 g/dl of hemoglobin more than 6-mo-olds in the intermittent responder category. Antibody responder category and parasite density were independently associated with temperature (P < 0.0012). 6-mo-olds in the negative antibody responder category had an estimated temperature 0.71°C higher than the individuals in the intermittent antibody responder category. Fig. 9 shows the associations between anti-GPI antibody responder category and malaria-attributable pathology in children 0.5–3.5 yr old. In each parasitemia category, febrile illness increased and hemoglobin level decreased in children without antibodies or with only short-lived antibodies; however, in children with persistent antibodies, febrile illness was lower and hemoglobin level was higher. These results strongly support the hypothesis that circulating anti-GPI antibodies neutralize the toxic effects of parasite GPIs.
This paper reports two major findings. (a) The results presented here establish for the first time that individuals residing in malaria-endemic areas develop a P. falciparum–specific anti-GPI antibody response, whereas people not exposed to the malaria parasite do not have these antibodies; in addition, anti-GPI antibody responses are correlated with protection against malaria-related febrile illness and hemoglobin loss. This agrees with the previous reports describing P. falciparum GPIs as pathogenicity factors based on their ability to induce a spectrum of proinflammatory cytokine responses and cause malaria symptoms. (b) Another important finding is elucidation of complete structures of the parasite GPIs by direct biochemical analysis and mass spectrometry. Thus, our results identified two novel structural features of P. falciparum GPIs: the presence of palmitate and myristate on C-2 of inositol, and the presence of predominantly oleic and minor amounts of cis-vaccenic acid and C18:2 at sn-2.
The purification of P. falciparum GPIs to homogeneity was crucial for structural determination as well as to obtain unambiguous evidence for biologic activity and establish the presence and specificity of naturally elicited antibodies in the sera of people living in malaria endemic areas. The key steps employed for isolation and purification of GPIs were: (a) growing of mycoplasma-free cultures to high levels of parasitemia and enrichment of infected erythrocytes; (b) metabolic labeling of GPIs and the use of human sera containing anti-GPI antibodies to follow purification steps; (c) use of sterile water and buffers and high quality organic solvents to exclude external contamination; and (d) siliconizing glassware to avoid loss due to surface adsorption. The mass spectrometry results presented here show that the purified GPIs are homogeneous.
We determined the structures of P. falciparum GPIs by subjecting the [3H]GlcN- and 3H–fatty acid-labeled GPIs to various standard degradative procedures, including their susceptibility to nitrous acid, HF, alkali, jack bean α-mannosidase, and GPI-specific phospholipase D, by determining fatty acid composition and finally by direct mass spectrometry. The results enabled us to propose the structures shown in Fig. 4 for the parasite GPIs. The structure of the core glycan is the same as determined previously 19,20. However, the acyl substituents are different from those reported previously, based on radiolabeling, with palmitate at both sn-1 and sn-2 and predominantly myristate on inositol 21. As shown in Fig. 4, palmitate is the major acyl substituent with minor proportions of myristate on inositol of the parasite GPIs; GPIs with acylated inositol residues from other sources reported to date contain only palmitate on the inositol residue 12. The parasite GPIs contain unsaturated acyl substituents at sn-2 (major C18:1 and minor C18:2) and predominantly C18:0 and a range of variable size saturated acyl residues at sn-1. With respect to the nature of the acyl residue at sn-2, the parasite GPIs resemble the GPI of T. cruzi trypomastigote mucin that has a potent cytokine-inducing property 30. Another unusual feature of the parasite GPIs is the presence of cis-vaccenic acid at sn-2.
The parasite GPIs differ significantly from those of humans with respect to both the acyl substituents and the carbohydrate moiety 12. The GPI moieties of human erythrocyte proteins (the erythrocytes do not contain detectable levels of free GPIs), CD59, and acetylcholine esterase contain exclusively a C18 alkyl substituent at sn-1, C22:4 at sn-2, and palmitate on inositol; the carbohydrate moieties contain one or two extra phosphoethanolamine as well as βGalNAc residues 12. The GPI of human spleen CD52 contains a diacylglycerol moiety and lacks GalNAc; however, this differs from the parasite GPIs with respect to the type of fatty acid at sn-2 (C22:4, C22:5, and C22:6) and contains phosphoethanolamine on the first mannose residue 12. These structural differences may contribute to the observed naturally elicited immunologic responses against the parasite GPIs in humans.
Although Schofield et al. have shown that P. falciparum GPIs can transduce signals to elicit inflammatory cytokine responses 27,28,29, there have been concerns as to whether the observed activity was due to contamination (parasite, erythrocytes, and/or mycoplasma origin 43). These concerns were based on the observations by some investigators that aqueous buffer extracts of parasite cultures, presumed to have extracted the parasite GPIs upon boiling, could not elicit TNF-α. Thus, they have argued that the cytokine-inducing property of P. falciparum was due to unknown components. However, it should be noted that GPIs can only be extracted with organic solvents. Because of this existing controversy, we tested TNF-α induction by the highly purified parasite GPIs. The results presented in this paper clearly show that the purified P. falciparum GPIs can induce TNF-α in macrophages. This activity is consistent with the recent finding by Almeida et al. that a highly purified GPI moiety of T. cruzi trypomastigote mucin induces TNF-α 30 and confirms the previous finding by Schofield et al. 26. Furthermore, studies by Almeida et al. also show that C18:1 and/or C18:2 acyl substituent at sn-2 in the T. cruzi GPIs is critical for TNF-α–inducing activity 30. As the P. falciparum GPI also contains C18:1 (major) and C18:2 (minor) at sn-2, it is possible that these acyl substituents contribute to the toxic property of the parasite GPIs.
In malaria endemic areas, younger children have the highest risk of developing severe malaria, whereas older children and adults rarely develop severe disease despite repeated exposure and significant parasitemia 23. People in nonmalarious regions completely lack this resistance, suggesting that the protection is due to a parasite-specific response acquired through repeated infections. In the Western Kenyan population studied here, we found that adults have P. falciparum parasitemia >14% of the time, whereas children <4 yr old have parasitemia >60% of the time. The density of parasitemia is lower in adults and, importantly, the level of parasitemia that can be tolerated without causing febrile illness or anemia is higher in adults (Branch, O.H., unpublished). As P. falciparum GPIs are pathogenicity factors 26,27,28,29, the resistance of adults to malaria illness may be related to a GPI-specific protective immunity.
All the Kenyan adult sera analyzed contained high levels of GPI-specific IgGs, whereas all 50 USA adult sera completely lacked such antibodies. The antibody response was highly specific to GPIs and their intermediates; other phospholipids of the parasite showed only low levels of immunoreactivity. Several phospholipids, including PIs, PGs, and CL, showed only 5–15% of the immunoreactivity exhibited by the parasite GPIs, which appears to be due to the polyclonal nature of the antibodies reacting with the common epitopes. Previously, several studies have reported the presence of significant levels of antibodies that bind phospholipids either directly or via binding of serum β2-glycoprotein I 44. In those studies, lipids were coated with several micrograms per well for ELISA. It is known that proteins and antibodies can nonspecifically bind to lipids when coated at high density. In contrast, our study used 0.5–2 ng/well of GPIs for ELISA and 100 ng for TLC immunoblots. Moreover, in previous studies, plates coated with lipid antigens were incubated with sera diluted with buffers without detergent. Under such conditions, we found high levels of nonspecific activity. Thus, our study clearly demonstrates that the identified IgGs are specific to GPIs.
Our results also establish that the PI portion of the GPIs contributes significantly to immunogenicity. The removal of the sn-2 fatty acid from the GPIs did not affect antibody reactivity (Vijaykumar, M., unpublished results). As treatment with HF abolished immunoreactivity, it appears that the acylated inositol phosphate is the immunogenic portion of the molecule. This agrees with the previous finding that antibodies raised against PIs can inhibit the induction of TNF-α by P. falciparum extracts 45. These results are important in that if an sn-2 acyl substituent is indeed required for cytokine-inducing activity, then it may be possible to synthesize nontoxic molecules for therapeutic purposes.
The TNF-α–inducing activity of GPIs and the correlation between the general resistance of adults in endemic areas to malaria pathogenesis and the presence of serum anti-GPI antibody response suggest that the acquired immunity is related significantly to the anti-GPI antibodies. This prediction agrees with the lack of such an antibody response in the majority of children <4 yr old, the risk of children developing severe malaria, and the correlation between the gradual acquisition of the antibody response in an age-dependent manner (>80% of the 7–8-yr-olds having high levels of serum antibody) and protection against malaria with age. Thus, a direct correlation between anti-GPI antibodies and malaria-related pathology could be observed in young children.
High levels of antibodies against several other antigens such as MSP-1, EBA-175, and circumsporozoite protein are also present in adults, and therefore it may be argued that the anti-GPI antibody response is not independent from those antibodies. However, it is important to consider that the anti-GPI antibody response is related to antidisease immunity, whereas antibodies against parasite proteins studied to date are involved in antiparasite immunity (controlling parasite burden). Furthermore, whereas ∼80% of children <2 yr of age either lack or contain very low levels of short-lived anti-GPI antibodies, these children can have high levels of antibodies against MSP-1 and other proteins (8,9; and Branch, O.H., unpublished results). Although there was a correlation between the levels of MSP-1 and EBA-175 antibodies and protection against parasite density, these antibody responses were not related to anti-GPI antibody response, protection against febrile illness, and hemoglobin loss at any given parasite density.
Almost all malaria vaccine development efforts currently being pursued use parasite proteins in a multicomponent formulation, aimed at providing immunity against infection (antiparasitic ). On the other hand, by blocking the toxic effects of the parasite GPIs, a GPI-based vaccine might significantly reduce malaria pathogenesis. The identification of the PI moiety as the functional part of the molecules should significantly simplify any approach towards the development of antidisease measures.
We thank Manonmani Venkatesan for cell culturing, Dr. Margarette Kolczak (Centers for Disease Control and Prevention) for statistical assistance, the Asembo Bay Cohort Project volunteers for their participation in the study, Dr. Vernon Reinhold (University of New Hampshire) for suggestions regarding mass spectrometry results, Dr. D. Koetch (Director, Kenya Medical Research Institute) for providing approval to publish this work, and S.-Y. Chan and J.E. McClellan for GC-MS analyses.
This work was supported by National Institutes of Health grant AI41139 and in part by US Agency for International Development grant HRN 6001-A-00-4010-00, and National Institutes of Health National Center for Research Resources grant P41-RR10888 (to C.E. Costello).
Abbreviations used in this paper: AHM, 2,5-anhydromannitol; CL, cardiolipin; EBA, erythrocyte membrane antigen; GalN, galactosamine; GC-MS, gas chromatography–mass spectrometry; GlcN, glucosamine; GPI, glycosylphosphatidylinositol; HF, hydrofluoric acid; HPTLC, high performance thin-layer chromatography; HRP, horseradish peroxidase; MSP, merozoite surface protein; PG, phosphoglyceride; PI, phosphatidylinositol.
R.S. Naik, O.H. Branch, and A.S. Woods contributed equally to this work.
D.J. Perkins is also associated with the Division of Hematology, Duke University Department of Medicine, Durham, NC 27705.