Patients were enrolled in Cotonou (6° 21' 35'' North, 2° 26' 23'' East) and Allada (6°39'52'' North, 2°09'35'' East) in southern Benin over one malaria season (April to September 2008). Allada is about 40 km north of Cotonou. In southern Benin, malaria is mesoendemic and transmission is perennial (Damien et al., submitted). The transmission level in Cotonou was characterized in 1992 by 33 infecting bites per man per year 21 . Severe malaria mainly occurs during the rainy season (April to mid-July). Cerebral malaria (CM) patients and pregnant women were included in two health facilities in Cotonou, the National Hospital (CNHU, Centre national hospitalo-universitaire) and the main maternity hospital (HOMEL, hôpital mère-enfant de la lagune), respectively. Uncomplicated malaria (UM) patients were enrolled in two schools located at Allada in children more than five years of age attending for school nurseries. Healthy adults (HA) were recruited in the blood donation centre in CNHU. All patients were included after written informed consent from themselves or their guardian. Patients were managed by the medical team of each health facility. An adequate anti-malarial treatment was administered according to the national malaria treatment policy. Ethical clearance was obtained from the Institutional Ethics Committee of the Faculté des Sciences de la Santé at the Université d'Abomey-Calavi in Benin.

Three groups of patients presenting distinct clinical malaria status were enrolled: uncomplicated malaria (UM), cerebral malaria (CM) and pregnancy-associated malaria (PAM). Briefly, UM was defined as the combination of fever (tympanic temperature ≥ 37.8°C) or a history of fever in the preceding 24 hours, confirmed presence of P. falciparum infection, and absence of any severity sign as defined by the WHO 22 . CM was defined by a Blantyre score at diagnosis ≤ 2 combined with a coma duration of six hours at least, and confirmed presence of P. falciparum infection with exclusion of other cause for coma, particularly meningitis. PAM cases were obtained in pregnant women just before delivery. Patients were all screened for malaria infection by rapid diagnostic test for P. falciparum (Malaria Quick test, Cypress Diagnostics, Langdorp, Belgique). A questionnaire was administered to the children' guardian, or the pregnant women to collect social data, clinical signs and treatment received before hospital admission.

Two additional groups of subjects were added as plasma controls, healthy male adults (HA) living in Cotonou, and non-immune healthy French volunteers (NI) living in France and having never been exposed to malaria. Peripheral venous blood sample (5 ml) was collected from all study individuals in vacutainer tubes containing EDTA. Plasmodium falciparum infections were confirmed by Giemsa-stained thick blood films, and parasitaemia was quantified by counting against 200 leucocytes assuming a mean of 8,000 leucocytes per millimeter of blood. Blood group was determined for each blood sample. Plasmas were separated and frozen from all subjects. Parasite isolates from all malaria-infected patients with O blood group were frozen. Blood count and lumbar puncture for meningitis diagnostic were achieved for patients from the CM group.

Infected red blood cells obtained from patients with O blood group were snap-frozen in a glycerolyte solution, as described 23 . Twenty-six P. falciparum isolates from patients (7 in the CM group, 7 in the PAM group, and 12 in the UM group) were frozen at -150°C. Cryopreserved IE were subsequently thawed and matured in vitro for 24-36 hrs 24 . The multiplicity of infection (MOI) was estimated by PCR typing of the polymorphic regions of the genes encoding merozoite surface protein-1 (msp1) and merozoite surface protein-2 (msp2), as described previously 25 . The MOI was estimated as the largest number of alleles observed between the two loci analysis.

A total of 256 plasma samples were collected as follows: 127 samples were collected at the time of malaria diagnosis (D0), 41 from CM patients, 62 from UM patients, and 24 from women with PAM. CM and UM patients were also seen at convalescence (D30), 17 plasma samples were obtained in the CM group, and 62 in the UM group. The high number of lost to follow up in the CM group is mainly explained by the urban localization of the children enrolled, and the difficulty to get them back one month after a severe malaria attack. Two sets of control plasmas were obtained from 44 healthy adults (HA) living in Cotonou, and from six French volunteers never exposed to malaria (NI).

The fluorescence-activated cell sorter (FACS) analysis was used for measurement of specific anti-VSA antibody, as described before 26 . Briefly, thawed parasites were matured to late stages (late trophozoites and schizontes) 24-36 hrs at 37°C in a candle jar, and enriched by exposure to a strong magnetic field (MACS, Miltenyi Biotec, Paris, France). Aliquots of 2 × 10⁵ IRBC in PBS-FCS 2% were labeled with ethidium bromide (0.002 mg/ml) to allow flow cytometric exclusion of remaining uninfected erythrocytes, and with 100 μl of goat anti-human IgG - FITC (IM1627, Beckman Coulter, Nyon, Switzerland). Labeled IRBC were sequentially exposed to 5 μl of plasma. Samples were analysed from a minimum acquisition of 5,000 IRBC using a Facscalibur cytometer (Becton Dickinson). The level of IgG recognizing VSAs was expressed as mean fluorescence intensity (MFI) in the channel for ethidium bromide-labeled IRBC. Nonspecific labeling was evaluated by analysis of uninfected (ethidium bromide-negative) erythrocytes. All plasma samples tested against a particular parasite isolate were processed and analysed in a single assay. NI plasmas samples were tested against all isolates. However, due to limitations in amounts of IRBC available, not all plasma samples were tested against all parasite isolates (Table 1). To be able to compare VSA IgG levels between isolates and plasma samples, the mean plus two standard deviations of log MFI values obtained with the six NI samples was subtracted to all log MFI values obtained with each isolate. The threshold for positivity was defined as two standard deviations above the mean MFI from the six NI samples.

Data were analysed by using the Statview software programme. For analyses performed with data obtained the day of diagnosis, data obtained with homologous plasma/isolate pairs were excluded. The levels of antibody responders were compared using the χ² test. VSA Ab recognition was evaluated by analysis of covariance (ANCOVA), two-factor ANOVA, t-test and Wilcoxon rank test for paired data. Clonality and msp family prevalence were compared between clinical groups by Kruskall-Wallis and Mann-Whitney U-tests. Spearman correlation test was used to estimate MOI and parasite density correlation. Values of P < 0.05 were considered statistically significant.

A total of 127 malaria-patients were enrolled. Three groups of patients were identified as CM (n = 41), UM (n = 62), and PAM (n = 24) patients, as well as one group of 34 malaria-exposed healthy male adults (HA). CM patients were between six months and seven years of age with a mean (± SD) of 3.3 (± 1.7) years. UM patients were aged 9.3 (± 2.1) years, PAM women were aged 24.9 (± 6.4) years, and HA were aged 35.1 (± 8.9) years. In PAM women, gravidity varied between 1 and 5 (mean 1.9 ± 1.2), 13 were primigravidae (PG) and 11 were multigravidae (MG). Parasitaemia was highly diverse between and within groups: CM, median (interquartile range) of 32,653 (2,082 - 64,846) per μl of blood; UM, 3,555 (516 - 16,285)/μl; PAM, 17,391 (400 - 88,379)/μl. Mean tympanic temperature (± SD) at inclusion was of 38.8 (± 1.2) for the CM patients, 37.2 (± .7) in UM patients and 38.0 (± .9) in pregnant women.

The CM patients were all comatose, admitted at hospital for at least two hours and presented a mean Blantyre score (± SD) of 1.8 ± 0.46 (min. 0, max. 2). The coma lasted 2.5 days on average (min. 1, max. 9). Meningitis was excluded for all children. Anaemia characterized the CM group with a mean haemoglobin level of 6.6 g/L (± 1.5) and a mean hematocrit of 20.8 (± 4.5). Erythrocytes count was low with 2.7 (± .6) T/L although the mean globular volume and the mean haemoglobin corpuscular content were normal. Respiratory distress occurred in 4/39 (10.3%) children of the CM group. Among the children suffering of CM, six died and five came out of hospital without medical consent. At that time among these five children, two presented a deep coma, two a normal consciousness and the data was missing for one.

Means (CI95%) for MOI were respectively 1.49 (0.6 - 3.6), 1.61 (0.6 - 3.1) and 1.83 (0.5 - 4.7) for CM, UM and PAM isolates (P = .73 by Kruskall-Wallis test). However, MOI increased with parasitaemia (P = .001, rho = .307). The prevalence rates of the allelic families for all isolates were 65%, 27% and 40% for the K1, MAD20 and RO33 families of the msp1 locus, respectively, and 94% and 7% for the 3D7 and FC27 families of the msp2 locus, respectively. Interestingly, PAM isolates presented a higher frequency for MAD20 and FC27 alleles (48% and 23%, P = .04 and P = .009, respectively), and a lower frequency for 3D7 alleles (76.5%, P = .003) than other clinical groups of isolates. RO33 alleles were more frequent in UM isolates than in other groups (P = .05).

The levels of anti-VSA antibody responders were compared against each type of isolates in plasmas of the different clinical groups of patients at the day of diagnosis. Among CM, UM and HA plasma groups, these levels were different between the three origins of isolates (see Figure 1). CM and UM plasmas presented similar patterns of anti-VSA responses. Both plasma types were greater antibody responders against VSA_UM than against both VSA_CM and VSA_PAM (P < .05 for all, P = .0006 for VSA_UM versus VSA_PAM in UM plasmas). In HA plasmas, antibody responses against VSA_UM were higher than VSA_PAM. The primigravidae (PG) plasmas contained similar antibody responses to the different VSA types. In multigravidae (MG) plasmas, responses were higher to VSA_UM than to VSA_CM. The levels of anti-VSA_PAM antibody responders in MG and PG plasmas were similar. The quantitative results are also available (see Additional file 1), showing all the specific parasite/plasma combinations tested and including the non immune plasma used as controls in the qualitative analysis.

The level of anti-VSA antibodies was analysed at different ages in all plasma groups, except for anti-VSA_PAM Abs that were measured only in UM, CM, and HA plasmas (Figure 2). Anti-VSA_UM antibodies levels remained at baseline level until 6 years of age, then increased until 12. From that time, a plateau was reached and maintained until 54 with a mean MFI value around 3.5. Overall, MFI levels for anti-VSA_UM Abs increased with age (ANCOVA, P < .0001). Anti-VSA_CM antibodies gradually increased until 18 years then decreased, although values did not vary statistically. Mean MFI values remained low for anti-VSA_PAM Abs (max. 1.5) and did not vary with age. VSA_UM and VSA_CM antibody levels were then compared according to three age groups: until 6 years old, between 6 and 18, older than 18. Until 6 years old, levels were similar (t-test). In both other age groups, VSA_UM antibody levels were higher than VSA_CM antibody levels (P = .01, P = .0008, respectively).

First, MFI values obtained in plasma samples the day of diagnosis and one month later were compared. Plasmas from the same patients were tested against the isolate infecting these patients. For UM plasmas and UM isolates, 10 pairs of results were obtained with both D0 and D30 plasma samples for the homologous isolate. As expected, the levels of anti-VSA_UM antibodies had increased at D30 (Wilcoxon rank test, P = .01). For CM plasmas and CM isolates, only two pairs with both MFI at D0 and D30 for the homologous isolate were available and both showed increased D30 values (see Table 2). Afterwards, the paired MFI values obtained at both Day 0 and at Day 30 in UM or CM plasmas against all UM or CM isolates, were compared, excluding homologous isolate-plasma pairs tested before (n = 125 UM, n = 55 CM) (Figure 3). Interestingly, UM plasmas exhibited similar levels of anti-VSA_UM antibodies at D0 and one month later (panel A), whereas CM plasmas exhibited higher anti-VSA_CM antibodies levels at D30 (panel C, P = .002). The same analysis was performed after matching ages in UM and CM groups, i.e. with children aged between five and seven years of age. Results were similar as before, anti-VSA antibodies significantly increased at D30 in the CM plasmas (P = .033), but not in the UM plasmas (P = .206). The same type of analysis was performed to test the potential cross-acquisition of anti-VSA_CM by UM patients after an UM infection, as well as anti-VSA_UM by CM patients after a CM infection. No difference was obtained between the anti-VSA_CM antibody levels displayed at D0 and D30 by UM plasmas (panel B). Conversely, anti-VSA_UM antibodies were higher in CM plasmas at D30 than at D0 (panel D, P = .05). These results show that infection with an isolate of the CM group induced the acquisition of antibodies directed against other CM isolates, but also against UM isolates. No such phenomenon was observed following infection with an isolate of the UM group.