Allele-specific antibodies to Plasmodium vivax merozoite surface protein-1: prevalence and inverse relationship to haemoglobin levels during infection

The dataset analysed consisted of 141 non-complicated P. vivax mono-infected individuals randomly selected from a larger study described elsewhere [14]. Malaria infection was first diagnosed by microscopy examination of thick blood smears and then confirmed by PCR. Briefly, patients were recruited and treated in two different health centres in the Western Brazilian Amazon between February 2006 and January 2008: (1) 77 patients from the Hospital Universitário Júlio Müller in Cuiabá (Mato Grosso State), where active malaria transmission does not occur; (2) 64 patients from the Fundação de Medicina Tropical Dr. Heitor Vieira Dourado in Manaus (Amazonas State), located in a low malaria transmission area. Patients from Cuiabá reported short visits to other areas in the Brazilian endemic region suggesting that they had infected there. Patients from Cuiabá and Manaus were previously exposed to malaria as the median number of reported previous malaria episodes was two. All patients were examined and interviewed by a trained physician who filled in a questionnaire containing demographic, clinical, parasitological and epidemiological questions. Haemoglobin and platelet levels were measured using a blood cell counter (ABX Pentra 90; Horiba Diagnostics, Kyoto, Japan) and parasite densities were determined by examination of 200 fields at l000× magnification under oil immersion. Written informed consent was obtained from all study participants prior to sample collection and ethics approval was obtained from the Ethical Committee Research of Universidade Federal de Minas Gerais (Protocol # ETIC 060/07).

Eleven recombinant proteins corresponding to polymorphic regions (blocks 2 and 10) and conserved portions (block 13) of the PvMSP-1 [9] were used in this study. Five versions of block 2 (isolates Belem, BP29, BP39, BR07 and BP30), five versions of block 10 (isolates BP29, BP39, BP13, BR07 and BP01), and one version of the conserved block 13 (isolate Belem) were selected for the expression of recombinant antigens as previously described [13]. The 11 plasmids used here were kindly provided by Dr. Marcelo U. Ferreira (Universidade de São Paulo, Brazil) and all recombinant proteins were expressed in fusion to the C-terminus of Schistosoma japonicum glutathione S-transferase protein (GST). Briefly, PCR products were cloned into the pCR 2.1-TOPO vector (Invitrogen, Carlsbad, CA, USA) and subcloned into the expression vector pGEX-3X (Amersham Pharmacia Biotech). Recombinant proteins were expressed in transformed Escherichia coli (BL21-DE3) and purified by affinity chromatography as described elsewhere [13, 15]. Fractions were analysed by 12% SDS-PAGE and stained with Coomassie blue. Fractions containing recombinant proteins with a high degree of purity were pooled and extensively dialyzed against PBS. The protein concentration was determined using the BCA kit (Pierce, USA) according to the manufacturer’s recommendations.

The block 10 polymorphic domain of the PvMSP-1 gene was amplified and sequenced according to protocols previously described [13]. The primers PVF7 (5-CCTTAAGAATACCGAGATTTTGCTGAAG-3 [nucleotides 3429–3456]) and PVR3 (5-GCGATTACTTTGTCGTAG-3 [nucleotides 4007–3990]) were used. The sequences were analysed using the programs SEQUENCE SCANNER v1.0 (http://​www.​appliedbiosystem​s.​com) and Bioedit v.7.0.5.3 [16]. Forty-one recovered sequences were aligned to other sequences previously described [9, 13] using CLUSTAL W following the identification and determination of haplotypes (alleles) distribution.

To test the association of each IgG response with haemoglobin concentration and platelet counts, a multiple linear regression approach was applied to the corresponding data considering the latter as the response variables. In particular, a null regression model for each response variable that included only the main effects of putative confounders (gender, age, study site, parasitaemia, and number of previous malaria exposure) was compared to an alternative model that included the same confounding effects plus the effect of the specific IgG response under testing. For the corresponding model comparison, likelihood ratio tests were performed in the corresponding data, where −log₁₀(P value) was used as a measure of the strength of association (i.e., large values of −log₁₀(P value) indicated a strong association between haemoglobin concentration and a given IgG response under testing). To control for multiple testing, the threshold for statistically significant was determined by the Bonferroni correction method [i.e., −log₁₀ (overall significance level/number of tests performed)]. A similar approach was applied to analyse the association between anaemia (haemoglobin <11 g/dl) or thrombocytopaenia (platelet <150,000/mm³) and each IgG response but alternatively using a logistic regression framework due to binary nature of the response variables. To assess the quality of prediction of the fitted models, the area under the receiver-operating characteristic curve (AUC) was calculated for the models providing the strongest association between anaemia and IgG responses; the values of 0.5 and 1 correspond to a random and perfect prediction of the response variable by the model, respectively. In particular, high values of AUC indicated a good agreement between model predictions and the respective data.

All analysis were carried out in the R software version 3.3.0 for Apple Maverick operating system (http://​www.​r-project.​org) using the authors’ own scripts, which are available from the authors upon request. The significance level was set at 5% for the analyses.

Patients presenting non-complicated vivax malaria diagnosed in two different health centres (Cuiabá and Manaus) in the Brazilian Amazon region were enrolled in this study. The median parasitaemia significantly varied between patients from Cuiabá and Manaus (1640 vs. 3713 parasites/µl of blood, respectively; P < 0.0001). However, the groups are reasonably matched for age, gender, number of previous malaria episodes, haemoglobin levels, platelets count and number of white blood cells (Table 1).

Levels (measured as reactivity indexes) and seroprevalence of plasma IgG antibodies to 10 antigens representing allelic forms of PvMSP-1 (BR07, BP29, BP39, BP30 and BEL of block 2 and BR07, BP29, BP39, BP01 and BP13 of block 10) and to one conserved antigen representing block 13 (PvMSP-1₁₉) are showed in Fig. 1a and b, respectively. As expected, the levels of IgG antibodies to PvMSP-1₁₉ were significantly higher than those observed to other polymorphic antigens representing block 2 or block 10 of PvMSP-1 (all P values <0.001) (Fig. 1a). In close agreement with this observation is the high seroprevalence of patients with antibodies against PvMSP-1₁₉ (87.9%) that contrasts with the low seroprevalence of the remaining antigens representing the variable region of the PvMSP-1 molecule. Among those, version BR07 of both block 2 and block 10 were the most recognized variable antigens (42.6 and 36.9%, respectively) (Fig. 1b). The other recombinant variable antigens were poorly recognized by P. vivax patients revealing seroprevalence no greater than 30%. Similar patterns were observed for plasma IgG levels confirming the poor antigenicity of variable antigens of PvMSP-1 as compared to conserved block 13, PvMSP-1₁₉ (Fig. 1a).

Ten out of 141 patients had no detectable IgG response to any antigen of the PvMSP-1 despite their current clinical infection (Fig. 2a). For PvMSP-1 responders, the proportion of patients who seroreacted to the conserved C-terminal PvMSP-1₁₉ antigen was significantly higher than the proportion of patients that seroreacted to any variant antigen of block 2 or block 10. To assess the contribution of PvMSP-1₁₉ determining such high antibody recognition rate, the range of positive responses (against one to 11 antigens tested) among 141 patients with positive IgG responses were analysed. As shown in Fig. 2a, the frequencies of seropositive patients who presented antibodies exclusively to any variant antigens were very low. Further, the overall combined frequency of antibody responses to all block 2 and block 10 variant antigens was significantly lower than that to C-terminal PvMSP-1₁₉, a result that suggests a high natural diversity of past and current P. vivax infections. In line with this observation is the increased total number of seropositive antigens per individual as a function of the self-reported number of previous malaria cases. This high diversity of IgG response profiles is also captured by the combined seroprevalence profiles to block 2 and block 10 recombinant antigens shown in Fig. 2b and c, respectively.

Sequences of the PvMSP-1 gene were obtained from a total number of 41 infected patients. These 41 sequences comprised eight different haplotypes from the block 10 PvMSP-1 gene (see Additional file 1: Figure S1). Interestingly, version BR07, an isolate firstly collected in Porto Velho, the capital of Rondonia state (Western Brazilian Amazonia) in 1995, was the most common sequence among those 41 typed isolates (32%). The amino acid sequences of five haplotypes were identical to those of the antigens used for antibody measurements and were detected in 30 local variant sequences suggesting an adequate panel of antigens used for serology. Four sequences identified as type II, III, VI and VIII which present 100% of similarity to sequences of recombinant antigens BP13, BP29, BP01 and BP39 were recovered at frequencies of 7.3, 14.6, 9.8 and 9.8%, respectively (see Additional file 2: Table S1). Note that 11 isolates (28.6%) share less than 75% of amino acid identity with sequences of the antigens used but had homology with other antigen sequences: haplotypes V and VII showed similarities of up to 90% with Asian isolates and haplotype IV showed a similarity rate of 100% with another Brazilian isolated, BP30 not tested in our study.

A correlation analysis was performed between a pair of IgG antibodies against to 10 variant recombinant proteins among either the overall population (141 patients) or IgG responders for each pair of antigens simultaneously recognized. Most of the correlations provided evidence for weak associations between IgG levels to variants antigens belonging to the same block of PvMSP-1. Cross-reactivity between versions BR07 and both BR29 and BP30 of the block 2 and BP01 and both BP29 and BP39 of the block 10 were detected among IgG responders (Fig. 3). In contrast, those antigenic pairs of block 2 (BR29-BR07) and block 10 (BP01-BP29 and BP01-BP39) showed a moderate proportion of amino acid identity (69.6–77.1%) as previously reported [13]. Interestingly, inter-block positive correlations could be verified either among the overall population or IgG responders. For example, levels of IgG to BELEM version of the block 2 were moderately correlated to IgG levels to BP01 version of block 10. The same pattern was observed for the correlation between BP29 versions from either block 2 and block 10. Negative correlations were also obtained for inter antigenic blocks revealing five putative block 2–block 10 antigens pairs among responders (Fig. 3).

Multiple linear and logistic regression models were then used to determine whether platelets counts, haemoglobin levels or anaemia could be associated with each IgG antibody response to PvMSP-1 adjusting for putative confounding effects (i.e., locality, age, gender, parasitaemia and a total number of previous malaria episodes). Haemoglobin levels were associated with IgG antibody responses to not only the C-terminal of PvMSP-1 domain PvMSP-1₁₉ but also with four among five variant antigens representing block 2 region (BR07, BP29, BP39, and BEL). The strongest association observed with haemoglobin levels refers to the IgG response to block 2 variant antigens, BP39. The respective association would appear to be negative as suggested by the negative estimate for antibody-level coefficient for the respective data (Table 2). None of the antibody responses to five antigens of the variable block 10 were associated with the underlying haemoglobin levels (Fig. 4). With respect to anaemia, there is only a strong association with IgG response to block 2 variant antigens, BP39. In close agreement with the haemoglobin data, this IgG response would appear to increase the odds of anaemia (Table 2). Finally, data suggests no strong association between platelet counts or thrombocytopaenia and any IgG response.