Increased polyclonal immunoglobulin reactivity toward human and bacterial proteins is associated with clinical protection in human Plasmodium infection

78 sera analysed here originated from subjects exposed to transmission in the Brazilian Amazon endemic area, who had variable numbers of reported previous episodes of P. vivax or P. falciparum clinical malaria. The malaria-exposed subjects comprised three distinct groups [see Additional file 1]. The principal study group was derived from a screening of 531 miners living in gold-mining areas in the municipality of Apiacás (AP), Mato Grosso, among whom 99 had been found parasitaemic, presenting with positive Giemsa-stained thick-blood smears [25]. Out of these 99 subjects, only 46 had shown classical malaria symptoms within 72 hours after parasite detection, while the other 53 had remained asymptomatic. Symptoms were mainly headache, anorexia and fever. Included in the present study are 48 of these miners, 24 symptomatic and 24 asymptomatic, who had lived for up to 17 years in Apiacás. There was no significant difference between symptomatic and asymptomatic individuals in their time of residence. According to anamnestic reports, numbers of previous malaria episodes were highly variable, but neither significantly different between the groups. The time elapsed since the respective last episode, however, was significantly longer among the asymptomatics (see Table 1). Within these parasitaemic subgroups, subjects were further distinguished whether or not they were positive for hepatitis B surface antigen (HbSAg) and therefore hepatitis B virus (HBV) carriers. Two further groups included exposed, but aparasitaemic control subjects. The first group (20 subjects) had resided for 10 or more years in Terra Nova Norte (TNN), a small rural community within the endemic region, continuously exposed to malaria. These parasitologically negative, convalescent individuals had been treated for malaria only until two months before the blood collection (Aparasitaemic/Previous malaria). The second group (10 subjects) also lived in TNN, but had no record of previous malarial episodes (Aparasitaemic/No previous malaria). All subjects responded to a questionnaire including information on past malaria and previous treatments. Consent to draw blood was obtained from each individual according to the Fundação Oswaldo Cruz Ethics Committee (MH, November 26^th, 1994) and to the Universidade Federal de Minas Gerais Ethics Committee (April 15^th, 1998). Venous blood samples (20 ml per subject) were drawn in Vacutainer™ (Becton Dickinson, Oxnard, CA) heparinized tubes. Giemsa-stained thick-blood smears were examined at this point for blood parasitaemia. Finally, a third control group consisted of 10 healthy adult volunteers from Belo Horizonte (Minas Gerais) who had never been exposed to malaria transmission or visited endemic regions (Aparasitaemic/Not exposed). IgG and IgM to human brain proteins were assessed in the subjects described above. When measuring IgG reactivity to E. coli proteins, two of the Terra Nova and seven of the Apiacas subjects were not included, but replaced by one Terra Nova and eight Apiacas subjects not assayed for reactivity to brain proteins.

The parasite density was quantified after examination of 200 microscopic fields at 1.000× magnification under oil-immersion. All slides were examined by three well-trained microscopists. Blood parasitaemia was expressed as the number of parasites per 200 leukocytes. Parasitaemia and reactivity to P. falciparum and MSP1-19 in these populations have been analysed and described previously [25].

The assay was done as described [26]. Briefly, protein extracts from human brain and cultured E. coli were run in a discontinuous SDS-PAGE 10% gel (Mighty Small electrophoresis apparatus, Hoefer Scientific Instruments, San Francisco, CA). After eletrophoresis, proteins were transferred onto a nitrocellulose membrane (Schleicher & Schuell, Germany) in a semi-dry system for one hour at 0.8 mA/cm². After overnight blocking of free binding sites in PBS-Tween 0.2%, membranes were incubated for four hours with sera, diluted 1/20, in incubation units fixing membranes in cassettes with 28 independent channels (Miniblot System C-Shell, Immunetics Inc., Cambridge, MA). Whole membranes were then washed and incubated for 90 minutes at room temperature with secondary anti-human IgM or IgG conjugated with alkaline phosphatase (Southern Biotech, Birmingham, AL). Reactivies were revealed with NBT/BCIP (nitroblue-tetrazolium/bromo-chloro-indolyl-phosphate) substrate (Promega, Madison, WI) for three to five minutes, and membranes scanned in 8-bit grayscale and with 600 dpi resolution in a domestic scanner (Apple Color OneScanner). Thereafter, membranes were stained overnight with colloidal gold (Biorad, Hercules, CA) to reveal total protein and again scanned as described.

The method of data processing has been described [26]. Briefly, reactivity profiles from the scanned immunoblot image were adjusted for migration irregularities during electrophoresis, using the scan of the same membrane stained for total protein. Special procedures programmed with the software IgorPro (Wavemetrics, Lake Oswego, OR) on a Macintosh computer (Apple computers, Cupertino, CA) were applied in order to represent each serum sample as a profile on a standardized migration scale, with the optical density (OD) as a function of migration distance. After this rescaling, sections were defined by intervals on the standardized migration scale around reactivity peaks, and, after baseline subtraction, reactivity for each section and serum was quantified as the average OD within a respective section. In this way, each serum can be represented by a vector with a dimension equal to the number of sections, containing the respective reactivity quantitation. In order to make these data commensurate across membranes, they were further normalized by the membrane-wise average reactivity of a unique standard (pool of human IgM or IgG), assayed twice on each membrane.

These vectors were then analysed by Principal Component Analysis (PCA). PCA is a classic method of multivariate analysis designed to describe multidimensional data with high dimensionality through projection onto characteristic subspaces with lower dimensionality. Principal components are defined in a mathematically strict manner as orthogonal axes fitting maximal information in terms of total variance with decreasing proportion and uncorrelated among each other. This includes no information on experimental groups or other particularities, but provides a completely neutral and unbiased description of the data set as such.

The number of bands detected on a respective extract was quantified as the number of sections showing reactivity values two-fold above the average reactivity of the standard used for adjustment. Total reactivity for a given extract was assessed as the average OD over the entire migration scale, from the first to the last section. Only distribution-independent statistics were used: Mann-Whitney rank sum test and Spearman rank correlation. P-values below 0.05 were considered significant.

As an example, Fig. 1 shows one representative out of four total immunoblot membranes on which IgG reactivities to brain proteins were assayed. Most of the reactivity bands appear in samples derived from asymptomatic parasitaemic subjects, with the exception of one symptomatic coinfected with HBV. In Fig. 2, three different (however, correlated) ways to evaluate the reactivity quantitatively are shown : (1) by the number of detected bands; (2) by the summation of total reactivity in terms of averaged optical density over the whole migration scale; (3) by the score of the first principal component calculated from standardized optical densities of all reactivity bands. By any of these criteria, asymptomatic parasite carriers from Apiacas showed more reactivity than parasite-free subjects in all groups with high significance (p < 0.00001). Parasite carriers with symptoms, however, also had less reactivity and, provided that they were not coinfected with HBV, did not significantly differ from parasite-free subjects. Among symptomatics, only HBV-positives showed reactivity levels comparable to the asymptomatics. In the absence of HBV-coinfection, asymptomatic and symptomatic subjects differed with high significance in terms of all three reactivity measures (p = 0.0003, 0.0002 and 0.0001, respectively). Even disregarding the presence of HBV coinfection, this difference was still significant (p = 0.017, 0.008 and 0.005, respectively). Assays of specific anti-parasite reactivity in the same subjects (reported in [25]) are shown in Fig. 2 for comparison. Although showing the same tendency, these specific assays were less discriminatory, considering either only HBV-negative (IgG anti-P. falciparum: p = 0.034; IgG anti-MSP1-19: nonsignificant) or all parasitaemic individuals (p = 0.017 and p = 0.014, respectively).

The first principal component (PCA factor 1) is, by definition, the linear combination of single reactivity measurements which represents a maximum of information about a multivariate dataset in terms of total variance, here 34%. Since this is the most systematic quantitative representation, the analysis will be continued based on principal components. In PCA, factor loads, i. e., coefficients indicating the relative contribution of measured parameters to the respective factor score, can be interpreted according to whether this factor represents a general level difference or a pattern-related property. Here, the factor-1 score contained only positive loads and, thus, represents a modified measure of total reactivity. Factor-1 scores indeed showed similar groupwise distributions as band number or total reactivity and were most discriminating. Among the first few principal components, factor 1 was the only one with interpretable properties.

Further properties of factor-1 scores derived from IgG anti-brain reactivities are shown in Fig. 3, 4, 5. Parasitaemic subjects had provided information on the time they had spent in Apiacas and in malaria-endemic areas in general. When plotted against these exposure times (Fig. 3), factor-1 scores discriminated best between asymptomatic and symptomatic subjects without HBV coinfection who had been exposed for relatively shorter periods. Hence, HBV-free subjects living less than 10 years in Apiacas or less than 20 years in malaria-endemic areas were almost completely separated according to their clinical status by factor-1 scores, but not those with long-term exposure. HBV-coinfected subjects, however, all displayed enhanced scores with no evident relationship to malaria exposure. For instance, both exposure parameters did not differ significamtly between symptomatic and asymptomatic subjects, regardless whether HBV infection is taken into account or not.

Fig. 4 shows the relationship between factor-1 scores and (a) individual parasitaemia and (b) anti-parasite IgG. Within HBV-negative asymptomatic subjects, scores were highest when parasitaemia levels were relatively low, although this association was quantitatively insignificant. A significant correlation was found between factor-1 scores and IgG reactivity to P. falciparum (Spearman rank correlation: +0.57 [p < 0.00001] including parasitaemics and exposed aparasitaemics; +0.37 [p = 0.011] within parasitaemics only), or to PfMSP1-19 (Spearman R, considering parasitaemics and exposed aparasitaemics: +0.42 [p = 0.0002]; not shown). Nevertheless, as can be seen in Fig. 4 for IgG anti-P. falciparum, these correlations are mainly due to the fact that all parameters were relatively enhanced in asymptomatic subjects. Considering asymptomatic and symptomatic individuals separately, correlations are no longer evident. Namely, some HBV-free symptomatics had elevated IgG anti-P. falciparum, but without a parallel increase in factor-1 scores.

Medical records of previous episodes of clinical malaria were available for both parasitaemic and exposed aparasitaemic subjects. Considering subjects of all groups together, factor-1 scores of individuals with evidence for P. falciparum infection in their last clinically diagnosed episode were higher than for those who had had P. vivax malaria (Mann-Whitney test: p < 0.01). Qualitatively, this appeared to be the case for parasite carriers as for aparasitaemic subjects with reported previous malaria (Fig. 5), although a valid statistical evaluation taking subject groups into account is impossible due to the small sample size. Nevertheless, it is interesting that the Plasmodium species presently detected in the parasitaemic subjects, often not identical with the one ascribed to the previous malaria period, was not associated with a similar difference in respect to factor-1 scores.

Patterns of IgG reactivity were also assayed toward an extract of E. coli whole cultures, which provided an independent and completely different set of target antigens, and, consequently, a different set of reactivity bands. Surprisingly, multivariate analysis of these bands yielded a result very similar to that described above for IgG reactivity to human brain proteins (Fig. 6). Although less significantly, the E. coli-derived score shared several properties described above. Analogously, asymptomatic parasitaemics had higher scores for reactivity to E. coli than symptomatics in the absence (p = 0.041), but not in the presence of HBV coinfection. Also similarly to IgG anti-brain, this significance increased when only subjects living less than 10 years in Apiacas were considered (p = 0.002, not shown). As can be seen in Fig. 7, factor-1 scores calculated from anti-E. coli IgG reactivities indeed correlated highly with anti-brain-IgG-derived factor-1 scores (Spearman rank correlation: +0.70; p < 1E-9). Finally, IgG anti-E. coli-derived factor-1 scores were higher in subjects reporting previous infection with P. falciparum, compared to P. vivax infection (p < 0.05; Fig. 8).

When principal components were calculated from IgG reactivities to both brain and E. coli proteins together, the properties described for both respective factor-1 scores fell together in the resulting first principal component (Fig. 9), indicating again that they were highly coincident. The difference between HBV-free asymptomatic and symptomatic parasitaemic subjects reached a higher significance level (p = 0.00005) than for factor-1 scores derived from either extract alone.

IgM reactivities to brain proteins were measured using the same extract as described for IgG (data not shown). Generally, a smaller number of bands was detected. The IgM-derived factor-1 included only positive loads like the IgG-derived one. The scores of both were correlated (Spearman rank correlation: +0.55; p < 1E-6), and, as for IgG, IgM-derived factor-1 scores were higher in asymptomatic than in symptomatic parasitaemic individuals free of HBV. However, IgM-derived factor-1 did not significantly discriminate between them, except when only individuals living less than 10 years in Apiacas were considered (p = 0.013). No significant difference was found either in respect to previously or presently detected Plasmodium species. In contrast to IgG anti-brain or anti-E. coli, within parasitaemics, a remarkable positive correlation existed between IgM-derived factor-1 scores and individual age (Spearman rank correlation: +0.41; p = 0.005), and also with times spent in Apiacas (+0.34; p = 0.03), but not within aparasitaemic subjects. As for IgG anti-brain, there was no significant correlation with blood parasitaemia levels, but with specific IgG anti-P. falciparum (+0.55; p < 1E-5).

Finally, principal components were calculated for IgG and IgM reactivities to human brain proteins together (Fig. 10). Surprisingly, the properties of the respective first factors calculated for each isotype alone did not coincide in a single principal component, as it had occurred in the co-calculation of IgG anti-brain and anti-E. coli reactivities. Instead, the first two principal components (factor-1 and factor-2) of the co-calculation both significantly distinguished between HBV-free asymptomatic and symptomatic parasite carriers (Factor 1: p = 0.013; Factor 2: p = 0.004), and, as above, not between those infected with HBV. Scores of both factors were enhanced in asymptomatic subjects. Since principal components in the same set are by definition uncorrelated, these two factors appear to represent separate effects associated with clinical states.

Factor-1, characterized by positive loads as the above described, was however positively influenced primarily by IgM reactivity, and properties of factor-1 scores were similar to those of the IgM-derived factor-1. Within parasitaemic subjects, scores increased with age (Spearman R: +0.39; p = 0.007), with exposure times in Apiacas and in endemic areas (+0.29 and +0.26, but not significant), and were correlated with anti-P. falciparum IgG (+0.32; p < 0.05). However, like IgG-derived factor 1, scores were also elevated in respect to P. falciparum but not P. vivax pre-exposure (p = 0.01; not shown).

In contrast, factor-2 scores were positively influenced mainly by IgG reactivity, while a number of IgM bands had a negative impact. Thus, factor 2 can be said to represent a certain IgG/IgM relationship. Remarkably, among HBV-free parasite carriers, the highest scores were characteristically found in asymptomatics who had only been exposed to malaria in endemic areas for a limited time. Factor-2 scores showed no significant association with IgG anti-P. falciparum or with pre-exposure to Plasmodium species. Instead, scores were slightly higher in subjects presently infected with P. vivax, but not P. falciparum (p = 0.04; not shown). Neither factor correlated significantly with parasitaemia.

Global IgG reactivity and subtype-specific reactivity have been previously assayed including some subjects examined here [25]. Among parasitaemic subjects, IgG1, IgG2, IgG3 and IgG4 anti-MSP1-19 all correlated significantly with IgM-anti-brain-derived as well as with IgG/IgM-anti-brain-derived factor-1, in particular IgG3 (Spearman R: +0.68 [p < 0.001] and +0.62 [p < 0.01], respectively). These values, however, did not corrrelate significantly with any IgG-derived factor, nor with IgG/IgM-anti-brain-derived factor-2. Only IgG anti-MSP1-19, measured in a larger number of subjects, correlated positively with IgG-anti-brain and IgG-anti-brain/E. coli-derived factor-1 (not shown).