Our epidemiological survey, summarized in Table 1, shows that all individuals studied were exposed to malaria infections throughout the year. A significant proportion of studied individuals reported a prior experience with P. vivax or P. falciparum malaria when compared with individuals that could not recall infections in the past and mentioned that they never had malaria even though they were born in the endemic area (p<0.0001). Among donors with a previous malaria infection(s), years of residence in the endemic area correlated positively with the past months since last malaria episode (Spearman’s r = 0.2411, P<0.0001, n = 245). At the time of blood collection 34 individuals were infected, 25 with P. vivax and 9 with P. falciparum (thus, 74.6% of the infected individuals had P. vivax and 26.4% had P. falciparum), consistent with the current local case distribution data for these two species reported by the Brazilian Ministry of Health [39].

HLA-DRB1* and HLA-DQB1* typing was performed on DNA samples of 276 individuals included in the cohort. Both the number of positive individuals for the HLA-DQB1* and HLA-DRB1* alleles and the frequency of each allele are summarized in Figures 1a and 1b. We found 13 HLA-DRB1* and 5 HLA-DQB1* allelic groups. There were two predominant HLA allelic groups in our studied population, HLA-DRB1*04 (16%) of all HLA-DR genotypes, and HLA-DQB1*03 (38%) of all HLA-DQ genotypes. The HLA-DRB1*09, HLA-DRB1*10 and HLA-DRB1*12 were less frequent in HLA-DRB1* and HLA-DQB1*04 in HLA-DQB1*. As observed in Figure 1c, HLA-DRB1*04 carrier individuals also show a large variety of alleles, with a marked predominance of HLA-DRB1*0411 (28%) over the others (P<0.05).

The prevalence of naturally acquired antibodies specific to the recombinant proteins was determined in plasma of 276 studied individuals (Figure 2). The frequency of responders show that IgG antibody reactivity to PvMSP1₁₉ was 86.7% and the frequency of positive individuals for at least one of the recombinant proteins representing PvMSP-3α and PvMSP-9 sequences was 77% and 76%, respectively. The recombinant protein representing PvMSP1₁₉ was the most recognized of the proteins tested. A substantial variation in terms of recognition was observed for the different regions of PvMSP-3α. The two blocks of repeats (MSP3_RI, 62% and MSP3_RII, 53%) and the C-terminal regions (MSP3_CT, 53%) were significantly more recognized (P<0.01) than the N-terminal region (MSP3_NT, 39%). Similarly, the recombinant representing the two blocks of repeats in MSP9_RIRII (63.7%) were significantly higher when compared to the recombinant representing only the second block of repeats (MSP9_RII, 40.2%; χ² = 10.3; P<0.0013) and the N terminal region (PvMSP9_NT, 33.7%; χ² = 15.2; P<0.0001). Combining data from all recombinant proteins a proportion of the population appears to be non-responders for PvMSP-1 (13%), PvMSP-3α (22.4%) and PvMSP-9 (28.7%) and five individuals were non-responders for all recombinant proteins.

The influence of HLA-DRB1* and HLA-DQB1* alleles and haplotypes on the naturally acquired IgG response to the recombinant proteins representing PvMSP-1, PvMSP-3α and PvMSP-9 was also evaluated (Tables 2 and 3). Although 13% of the population did not present detectable titers of antibodies to PvMSP-1 (MSP1₁₉), genetic restriction to this antigen does not seem to occur, since no association was observed between the HLA DRB1* or DQB1 alleles and the antibody response to MSP1_19. However for the PvMSP-3α and PvMSP-9 recombinants proteins, the IgG responders were associated with the presence of the HLA-DRB1*04 allelic group. The associations were found in IgG responders to the recombinants proteins representing the C-terminal and N-terminal regions of PvMSP-3α (MSP3_CT, P<0.025; MSP3_NT P<0.025) and the two blocks of repeats and the N-terminal region of PvMSP-9 (MSP9_RIRII, P<0.05; MSP9_RII, P<0.05; MSP9_NT, P<0.025). Among the individuals presenting the HLA-DRB1*04 allelic group the frequency of responders to all recombinant proteins was higher in the HLA-DRB1*0411 allele carriers. However, the association was not statistically significant with any specific HLA-DRB1*04 allele (figure 3).

The high frequency of non-responders to the recombinant proteins representing the blocks of repeats of PvMSP-9 (MSP9_RIRII and MSP9_RII) is associated with the presence of the HLA-DRB1*01 allelic group (P<0.001 and P<0.005 respectively) and the recombinant protein representing the full length of PvMSP-3α with the presence of the HLA-DRB1*16 allelic group (MSP3_FL) (P<0.05). Interestingly, this association was not detected when we evaluated the response to the recombinant proteins representing the different regions of PvMSP3.

In the evaluation of the HLA-DQB1* allelic groups we observed an association between HLA-DQB1*03 and responders to the recombinant protein representing the repeats and the N-terminal region of PvMSP-9 (MSP9_RIRII P<0.05; MSP9_NT P<0.05) and the C-terminal region of PvMSP-3α (MSP3_CT, P<0.01). In contrast, a negative association was observed between HLA-DQB1*06 and responders to the C-terminus of PvMSP-3α (MSP3_CT, P<0.025). All HLA associations with specific antibodies against PvMSP-1, PvMSP-3 and PvMSP-9 are summarized in table 4.

The effects of different HLA-DRB1 and HLA-DQB1 allelic groups and time of residence (years) in malaria endemic areas on the level of antibodies to PvMSP-1, PvMSP-3α and PvMSP-9, were also analyzed. The DRB1 and DQB1 alleles that are observed frequently in the population were included in these analyses. Multivariate analyses were performed by using time (years) of residence in endemic area (TREA), time (months) since the last malaria episode (TLI) and past malaria infections (PMI) and DRB1 or DQB1 alleles as independent variables and the reactivity indexes of IgG against all studied recombinant proteins as dependent variables. A significant association was observed only between HLA-DR+TREA and the reactivity index of IgG against PvMSP9-RIRII (F = 1.274, P = 0.0333) and PvMSP9-RII (F = 1.259, P = 0.0401). Interestingly, to all PvMSP-9 antigens, the observed power of association of each independent variable separately was greater in TREA (0.097 to 1.000) than in the HLA allelic groups (0.345 to 0.643). Therefore, the time of exposure is more closely related to the PvMSP-9 antibody levels. In addition the observed power of association was similar between both variables. No significant associations, positive or negative, were observed with HLA-DR+TREA, HLA-DR+PMI, HLA-DR+TLI and antibody levels to PvMSP-3α and PvMSP-1. No significant associations were observed for any HLA-DQ+TREA, HLA-DQ+PMI, HLA-DQ+TLI and PvMSP-1, PvMSP-3α and PvMSP-9 antibody levels.

Since the DRB1*04 allelic group was associated with responders, we also evaluated if the levels of antibodies, represented by the reactivity indexes to the recombinant proteins in the ELISA test, had the influence of these alleles. Comparing the antibody response in carriers and non-carriers of the HLA-DRB1*04 alleles, the levels were significantly higher in HLA-DRB1*04 carriers for the recombinant proteins representing PvMSP-9: MSP9_RIRII (carriers Mean ± SD = 3.23±0.60 vs non-carriers Mean ± SD = 1.95±0.47; df = 254, t = 2.934, P = 0.0037), MSP9_RII (carriers Mean ± SD = 2.49±0.41 vs non-carriers Mean ± SD = 1.41±0.23; df = 188, t = 2.923, P = 0.0045) and PvMSP9_NT (carriers Mean ± SD = 2.01±0.29 vs non-carriers Mean ± SD = 1.37±0.19; df = 124, t = 5.975, P<0.0001). However we did not find any difference in the levels of antibodies between these groups for the PvMSP-1 and PvMSP-3α responders. The antibody levels in individuals homozygous for the HLA-DRB1*04 allelic group did not differ from individuals that carry only one DRB1*04 allelic group (heterozygous) independent of the recombinant protein recognized. Finally, no relationship between a particular HLA haplotype and the frequency of antibody response was observed.

A cross-sectional cohort study was conducted involving 276 individuals (11 to 89 years of age) from communities in the malaria endemic region of Rondônia state, Brazil, where P. vivax malaria accounts for more than 70% of all malaria cases in the last five years [57]. The individuals in the study population have been described elsewhere [16], [58]. Briefly, they consist of rain forest natives as well as migrants from several non-endemic areas of Brazil who have resided in the region for 10 years or more. Samples and survey data were collected in 2007, during the dry months of June-August, coinciding with the period of increased malaria transmission in Rondonia State. Samples from 24 malaria naive individuals living in non-endemic regions and who had never visited malaria transmission zones were obtained from laboratory staff volunteers (Rio de Janeiro, Brazil and Atlanta, USA) and used as controls. Written informed consent was obtained from all adult donors or from parents of donors in the case of minors. The study was reviewed and approved by the Fundação Oswaldo Cruz Ethical Committee and the National Ethical Committee of Brazil.

To evaluate epidemiological factors that may influence the immune response against the recombinant proteins, all donors were interviewed upon informed consent. The survey included questions related to demographics, time of residence in the endemic area, personal and family histories of malaria, use of malaria prophylaxis, presence of malaria symptoms, and personal knowledge of malaria. Survey data was entered into a database created with Epi Info 2007 (Centers for Disease Control and Prevention, Atlanta, GA).

Venous peripheral blood (20 ml) was collected into EDTA tubes, and peripheral blood mononuclear cells (PBMC) were isolated by Ficoll/Hypaque (Pharmacia, Piscataway, NJ) density gradient centrifugation. Plasma was stored at –20°C and thin and thick blood smears of all donors were examined for malaria parasites at 1000× magnification under oil-immersion, all slides were examined by two researchers with expertise in malaria diagnosis. Donors positive for P. vivax and/or P. falciparum at the time of blood collection were subsequently treated per the chemotherapeutic regimen recommended by the Brazilian Ministry of Health.

Genomic DNA was isolated from whole blood drawn in EDTA by using a mixture of 5 ml buffer G2 (QIAamp DNA Blood Midi Kit; Qiagen Inc., Chatsworth, CA, USA) and 95 µl proteinase K (20 mg/ml). After incubation at 50°C for 1 h the DNA was ethanol precipitated, collected with a glass stick and transferred into distilled water. DNA concentration and quality was checked with a NanoDrop ND-1000 spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA). Sequence-specific oligonucleotide probes (SSOPs) were used by Luminex Xmap technology in order to determine the HLA class II allelic groups of studied individuals. Briefly, the system is based on probe arrays bound to color-coded plastic microspheres, and locus-specific biotinylated primers for HLA-DRB1 and HLA-DQB1 loci (LABType, One Lambda Inc, Canoga Park, CA, USA). Biotinylated amplicons were denatured to ssDNA and incubated with DNA complementary probes immobilized on fluorescent coded microspheres (beads) followed by incubation with R-Phycoerythrin conjugated to streptavidin. After hybridization, the samples were analyzed with Luminex Flow Analysis equipment. The HLA Visual 2.0 software (One Lambda, CA) analysis program deduces the HLA-DRB1 and HLA-DQB1 allelic groups. High resolution PCR-SSP typing was also used with the same method to define the DRB1*04 alleles, as these loci have been associated with responders to malaria antigens.

The expression and purification of the recombinants proteins were performed as previously described [16], [58], [59]. Nine recombinant proteins derived from P. vivax (Belem strain) were produced. These include PvMSP-1 sequence representing the 19 kDa C-terminal fragment (MSP1₁₉), PvMSP-3α sequence representing the near full length protein (MSP3_FL), the N-terminal region (MSP3_NT), the first block of repeats (MSP3_RI), the second block of repeats (MSP3R_II), and the C-terminal region (MSP3_CT), and PvMSP-9 sequence representing the N-terminal domain (MSP9_NT), the second block of tandem repeats (MSP9R_1I) and the first and second block of tandem repeats MSP9_RI-RII.

Plasma samples from study participants were screened for the presence of naturally acquired antibodies against the nine recombinant proteins: PvMSP-3α (MSP3_FL, MSP3_RI, MSP3_RII, MSP3_CT, MSP3_NT); PvMSP-9 (MSP9_RIRII, MSP9_RII, MSP9_NT) and PvMSP-1 (PvMSP1₁₉) by ELISA. Briefly, maxisorp 96-well plates (Nunc, Rochester, NY) were coated with PBS containing 2 µg per ml of each recombinant protein. After overnight incubation at 4°C the plates were washed with PBS containing 0.05% Tween 20 (PBS-Tween) and blocked with PBS-Tween containing 5% non-fat dry milk (PBS-Tween-M) for 2 h at 37°C. Individual plasma samples diluted 1∶100 PBS-Tween-M were added in duplicate wells and the plates incubated at room temperature for 1 h. After four washes with PBS-Tween, bound antibodies were detected with peroxidase-conjugated goat anti-human IgG (Sigma, St Louis) followed by o-Phenylenediamine and hydrogen peroxide. The absorbance was read at 492 nm using an ELISA reader (Spectramax 250, Molecular Devices, Sunnyvale, CA). The results for total IgG were expressed as Reactivity Indexes (RI), which were calculated by dividing the mean optical density of tested samples by the mean optical density plus 3 standard deviations of 24 non-exposed control individuals living in non-endemic areas of malaria (OD values higher than 0.089 to PvMSP1; 0,094 to PvMSP-9RIRII; 0,091 to PvMSP9RII; 0,074 to PvMSP9NT; 0,101 to PvMSP3FL; 0,081 to PvMSP3BLI; 0,092 to PvMSP3CT; 0,091 to PvMSP3NT). As positive controls we used five plasmas from exposed native individuals that presented high OD levels for all three proteins. Subjects were scored as positive to the corresponding antigen if the RI was higher than 1.

Survey data was recorded and entered into a database created with Epi Info 2007 (Centers for Disease Control and Prevention, Atlanta, GA). Analyses were done using Predictive Analytics Software, PASW (version 17.0; SPSS Inc., Chicago, IL, USA) and Prism version 5 (GraphPad Software Inc., San Diego, CA). Differences in medians of past malaria infections and time since the last infection for the study population data were tested by non-parametric Mann-Whitney. Student’s t test was used to compare means of normally distributed data of reactivity indexes of IgG antibodies. Differences in gender proportions were evaluated by chi-square (Χ²) test. Allelic groups were grouped by DR status and data were descriptively summarized using frequencies and percentages for all categorical variables. Overall associations of immunologic responses with the alleles from each HLA-DRB1* and HLA-DQB1* loci were evaluated by comparing the allele frequencies between seronegative subjects and seropositive subjects using standard contingency tables. For a locus with K alleles, a 2×K table was constructed by computing allele counts for each of the two groups. Each person contributed two observations to the table (one for each allele). Rare alleles, defined as those with less than five occurrences among subjects, were all pooled into a category labeled “other" for analysis. To evaluate global differences in allele distribution, we performed analyses using simulation methods as implemented in the software PASW. This approach randomly generates new cell counts for contingency tables under the null hypothesis of no association, while keeping the margins of the table fixed. We used an approach that compares each allele versus all others combined, resulting in multiple 2×2 tables, and used the maximum Chi-square statistic from this series of tables as a global test statistic (bipartition). All global tests assume that the two alleles from each subject are independent. Following the global tests of association, we examined individual allele effects on seropositivity using unconditional logistic regression analysis. Regression variables were created for each allele and coded as 0, 1, or 2, according to the number of copies a subject carries. Each allele variable was first included in a separate univariate logistic regression analysis to determine its association with seronegativity, ignoring the effects of other alleles. Next, to verify the influence of haplotypes, multivariate analyses were run that examined the simultaneous effects of multiple alleles at a given locus with seronegativity. Separate models were fit for each of the five loci. The low frequency of many alleles, coupled with data dependency issues, made it impossible to fit all allele count variables from one locus in a single logistic model. Because of this, we used a forward stepwise regression method to choose alleles most associated with antibody response. The significance level for entering an allele variable in the model was 0.05. Variables not included in the final stepwise model were, by default, pooled into a group against which the significant allele variables were compared. To examine the possibility of multiple effects on antibody levels, multivariate regression analysis was also performed by Generalized Linear Models (GLM). The reactivity index of IgG antibodies values were used as dependent continuous variable, HLA alleles as independent variable and time (years) of residence in endemic area (TREA), time (months) since the last malaria episode (TLI) and past malaria infections (PMI) individually as independent covariates. Interaction terms were also included (TREA+HLA; TREA+TLI; TREA+PMI) to test whether the associations between the reactivity index and HLA-DRB1* or HLA-DQB1* were significant. Partial η² (Partial eta-squared) were also used to evaluate the power of association of each variable individually in the model. All variables included in the final locus-specific stepwise models were placed into one overall stepwise model. As before, the significance level for entering a variable in the model was 0.05. All statistical tests were two sided and, unless otherwise specified, HLA analyses were conducted using the PASW software system.