Hospital-derived antibody profiles of malaria patients in Southwest India

The present study was a part of the US National Institutes of Health-sponsored Program Project entitled Malaria Evolution in South Asia-International Center of Excellence for Malaria Research (MESA-ICEMR). Samples were collected at Goa Medical College and Hospital (GMC), Bambolim, Goa, India. The sample collection process was approved by the Institutional Review Boards (IRB) of the Division of Microbiology and Infectious Diseases (DMID) at the US National Institute of Allergy and Infectious Diseases (NIAID), GMC, and the University of Washington (UW). Detailed ethics statements have been previously reported [20]. After sample collection and clearance from the Indian Institute of Technology, Bombay (IITB) Ethics Committee, the samples were transferred from Goa to IITB for microarray analysis. General patient information such as age, gender and time of sample collection, clinical information and results of laboratory tests for diagnosis were provided by the GMC and the UW teams to IITB for 200 patients. Other patient information remained classified to protect patient privacy.

Blood samples were collected from symptomatic malaria positive patients at GMC (Goa, India), year-round for 4 years (2012–2015). Written informed consent was obtained from all the volunteers during enrolment. A detailed description of the study site, enrolment and sample processing protocols has been published elsewhere [20]. Sera from 200 patients (96 P. falciparum, 100 P. vivax, and 4 mixed infections) were analysed by hybridization to protein microarrays. Parasite density (parasites/µL) was determined by microscopy for 199/200 individuals. Patients were further classified based on age (1–10 years, 11–22 years, 23–39 years and 40–65 years), hospitalization status, signs/symptoms of disease severity and time of sample collection (Fig. 1). Classification based on age was performed to segregate the population into young children, adolescents, young adults and older adults. WHO criteria for classification of severe P. falciparum malaria was used to define disease severity for both P. falciparum and P. vivax study groups due to the absence of specific case definitions for P. vivax malaria.

Twenty microlitre (µL) aliquots of sera from 200 patients were transported to IITB in cryovials labelled with unique numerical codes, one for each patient. Antibody reactivity to P. falciparum and P. vivax proteins was studied using a protein microarray (P. falciparum/P. vivax 500) displaying 515 P. vivax and 500 P. falciparum proteins (Antigen Discovery, Inc., Irvine, CA). At ADI, the proteins were expressed as polypeptide fragments and printed from cell-free in vitro transcription translation (IVTT) reactions by the manufacturing team. A few large proteins such as dynein beta chains, rhoptry neck proteins, NOT family proteins, conserved Plasmodium proteins, reticulocyte binding proteins, erythrocyte membrane proteins (PfEMP1) and a few transcription factors were printed as overlapping peptides or exon fragments. The antigens were down-selected from larger arrays (Pf4528 and Pv4441) based on the seroreactivity of different patient populations from other malaria endemic regions [16]. Antigens represent asexual parasite blood stages, preerythrocytic stages, and mixed stage proteins. Accession numbers and description of the polypeptides follow annotations published in PlasmoDB (http://​www.​plasmodb.​org).

Hybridizations were carried out at the Indian Institute of Technology Bombay (IITB). Each slide was composed of eight microarray pads. Multiple slides were used in each large experiment. Highly reactive sera from a single patient was used on one pad of every slide as a positive control. The rest of the microarray pads were used to probe sera from different patients. To track possible variations during array printing on individual pads on slides, batches of slides, or even assay performance during each experiment, select patient sera were probed across all runs and their signal intensities were compared for quality control and reproducibility. To account for non-specific reactivity of the IVTT expression system, each array pad contained 24 “No DNA-control” reaction spots, lacking a DNA template in the plasmid vector (Fig. 2). This ‘background signal’ was used for sample-specific normalization. Signals from immunoglobulin G (IgG) spots and anti-human IgG spots were used as standards to establish the correct PMT and laser power settings for each batch of slides. Select seroreactive proteins were purified and printed on the chip by the manufacturer to serve as positive control spots. Individual serum samples were diluted (1:100) and hybridized to each microarray pad. A biotin-conjugated goat anti-human (IgG) secondary antibody, followed by a fluorescently labelled streptavidin conjugate, were used to probe for IgG bound to antigens on the pads. IgG reactivity was quantified by a microarray scanner (GenePix Pro software) as a unitless relative intensity (Fig. 2). The final hybridization results reported here were performed over 4 days using two different print batches of slides.

In addition to hybridizations and scanning, data analysis was carried out at the Indian Institute of Technology Bombay (IITB). Normalization was performed as previously described [16]. Median normalized signal intensities (MNSI) were used to generate heat maps, while Log2 FOC values were used for statistical tests. Antigens were considered seroreactive if the average signal intensity for a group of different patients (MSI) was greater than a cut-off, defined as the average plus two standard deviations of the reactivity to all 24 IVTT control spots. Patient antibody breadth was defined as the number of seroreactive antigens per individual, per group, based on the criteria mentioned above. Differentially reactive proteins between two groups were determined using the Statistical Analysis for Microarrays (SAM) module in MetaboAnalyst 4.0. The test was performed assuming equal variance between the groups and an adjusted p value < 0.05 was considered statistically significant. Data was graphically represented using GraphPad Prism (Prism v6.0, GraphPad Software Inc., La Jolla CA).

In order to correlate immune responses with aspects of the local clinical manifestations of malaria, the participant data were grouped into two major classes; P. falciparum infected and P. vivax infected based on the Plasmodium species detected by microscopy. Analysis of these two populations would provide information on immune status of a patient with respect to responses to infection by each of the two human malaria parasite species seen. In addition, both study groups (P. falciparum and P. vivax) were sub-divided into individual age distributions as shown in Fig. 1a and b. The overall study population had a median age of 25 years (IQR 20–39 years) and was predominantly male (89.5% and 86% male for P. falciparum and P. vivax groups, respectively, Fig. 1c). This gender bias in seeking hospital care in India has been reported earlier for hospitals in Mumbai and Rourkela [21].

Patients were also sub-divided according to the severity of disease, as immune patterns that correlate with low severity might, in principle, give an indication of disease protecting immunity. According to the clinical information that was documented for each patient, 90/200 patients were hospitalized (in-patients), while 110 participants did not require hospitalization (out-patients) [20]. This number includes patients diagnosed with mixed infections (Fig. 1d). In-patients were further classified into severe (S) and non-severe (NS) groups. Of the 90 in-patients, 64 presented with at least one severe sign/symptom/diseases such as severe anaemia, cerebral malaria, respiratory distress, ARDS, jaundice, convulsions, hypoglycaemia, renal failure and abnormal bleeding. They were classified as patients with severe disease, while in-patients who did not exhibit severe symptoms were non-severe (Fig. 1e). Clinical information on disease severity could not be documented for 107 participants (38 P. falciparum and 66 P. vivax) who were all out-patients. It is important to reiterate that not all hospitalized individuals exhibited severe symptoms.

According to the World Health Organization guidelines for severe P. falciparum malaria, patients with parasite densities > 10% have severe infection, while there is no parasite density threshold for severe P. vivax malaria. A majority of the patients in this study presented with moderate to low parasite densities (< 2% or 100,000/µL) except two with parasite densities of 192,153 (P. falciparum) and 147,809 parasites/µL (P. falciparum + P. vivax), estimated by microscopy. Importantly, there was no correlation between parasite density, hospitalization status and severity, thus making study of immune status important (Additional file 1: Table S1).

Patients were further classified based on the time of sample collection. One hundred and twenty-eight patients presented with symptoms during the peak malaria season (May to November), while a separate group of 68 patients were admitted during the dry season (December to April). In order to eliminate sampling bias, the sample details were kept blind to those handling sera hybridizations or microarray slides until after data analysis.

As detailed in "Methods" section, serum samples from 96 P. falciparum and 100 P. vivax positive patients were probed on a protein array having 500 P. falciparum and 515 P. vivax polypeptides (Antigen Discovery, Inc., Irvine, CA). The magnitude and breadth of the Ab response, determined by the number of features recognized, was greater in case of P. falciparum than P. vivax malaria (Fig. 3a). Two hundred and forty-eight P. falciparum and 73 P. vivax polypeptides were recognized by P. falciparum and P. vivax patient groups, respectively (Fig. 3b, Additional file 2: Table S2). A large number of antigens (33%-P. falciparum and 36%-P. vivax) were found to be seroreactive in 30–40 patients, while only a few antigens (1.2%-P. falciparum and 1.4%-P. vivax) were seroreactive in more than 80 P. falciparum and 90 P. vivax patients. Two P. falciparum antigens and 8 P. vivax antigens showed lowest reactivity among all other seroreactive antigens. These were reactive in 10–20 P. falciparum and 20–30 P. vivax patients, respectively.

The antibody breadth (calculated by the number of seroreactive antigens in a group of individuals) is shown as a box and whisker plot in Fig. 3c. A few patients showed high levels of Ab binding to as many as 404, 398, 397 and 392 out of 500 P. falciparum antigens while one P. vivax patient had high reactivity to 265 out of 515 P. vivax antigens. Despite the large breadth of antibody responses, these patients were not excluded from the study. The gene IDs of the top antigens showing seroreactivity in more than 60% of the patients in both groups are listed in Table 1A and B.

An analysis of the features/functions of the 248 P. falciparum seroreactive antigens indicated that a majority of the seroreactive P. falciparum antigens were conserved Plasmodium proteins (17.9%) (Additional file 3: Fig S1). The next major category was protein binding (16%), followed by nucleic acid binding (6.6%) and host cell surface binding proteins (5.4%), plus others such as Plasmodium exported proteins, MSPs and HSPs. Several enzymes (4.6%), transcription and translation factors (3.5%), proteins involved in ATP-binding (3.9%), ion-binding (3.5%) and proteins involved in ubiquitination (1.5%) also elicited high IgG responses despite their intracellular locations in the nucleus and the cytosol. Antigens with features and functions similar to P. falciparum antigens were also seroreactive in P. vivax patients, along with hypothetical proteins that constituted the major group (33.3%).

It can be observed from heat maps in Fig. 4a that there was also a high level of cross-reactivity between P. falciparum and P. vivax antigens but all of the patterns are not easy to understand. One hundred and seventeen seroreactive P. falciparum antigens (Sr) from P. falciparum patients were also recognized by P. vivax patients (Fig. 4b). These antigens are here forth called cross-reactive antigens (Sr&Cr). Strangely in case of P. vivax antigens, the number recognized by P. falciparum patients was actually more than the antigens recognized by P. vivax patients themselves (Fig. 4b). This indicates that overall immune activity of P. falciparum patients was higher than the P. vivax study group. The cross reactivity with P. vivax antigens was especially high in the few P. falciparum patients who also showed high reactivity to a large number of P. falciparum antigens. While 131 out of 248 P. falciparum antigens were exclusively reactive in the P. falciparum patient group, a similar observation could be made only for 4 out of 73 P. vivax antigens (Fig. 4b). Gene annotations for the data represented in the Venn diagrams are mentioned in Additional file 4: Table S3. A scatter plot of MNSI levels for all P. falciparum and P. vivax antigens in both groups is shown in Fig. 4c to depict the level of crossreactivity observed in the study. Deeper analysis yielded an even more unexpected finding. All antigens that were reactive in both P. falciparum patients and P. vivax patients, shared orthologs with other Plasmodium species (PlasmoDB), including non-human infecting parasites. However, 29 P. falciparum antigens did not have orthologs in P. vivax-Sal1 strain. On the contrary, 98 P. falciparum antigens which were reactive only in P. falciparum group (not crossreactive in P. vivax patients) had orthologs in P. vivax (Additional file 5: Table S4, last tab). Likewise, few crossreactive P. vivax antigens did not share orthologs with P. falciparum-3D7 strain. Data for orthology and synteny are presented in Additional file 5: Table S4.

In order to distinguish between recent and cumulative exposure, patients were further segregated based on age into four sub-groups as mentioned previously (1–10 years-A, 11–22 years-B, 23–39 years-C and 40–65 years-D). The sample size differed in each age group as follows: The number of falciparum patients were 5, 36, 36 and 10, while vivax patients were 9, 30, 34 and 27 patients in age groups A, B, C and D, respectively. Clearly, the study involved some groups with very small numbers, especially children aged 1 to 5 years. Hence, a comparison of seroreactivity between groups was not performed. Moreover, there was no correlation between the number of seroreactive proteins and patient age, when scatter plots depicted antibody breadth for individual patients (Additional file 6: Fig S2). Instead of a statistical analysis between age groups, individual groups were analysed separately to identify top seroreactive antigens (Additional file 7: Table S5). The main aim was to determine if any of the antigens were unique to a particular age group. Based on early work, much from Africa, it is clearly known that children aged 0–5 years are immunologically different from other groups. In India, this is true for some but not all reactive antigens. As anticipated, certain antigens such as HSP40, ACS6, GCVH, among others, were seroreactive only in children. Likewise, several conserved Plasmodium proteins, family of tetratricopeptide repeat proteins, SET1, PK4, GPI1, ApiAP2, CSP, SERA7, PUF1, HSP90, HSP70, PfEMP1 family, and several enzymes like helicases, peptidases and methyltransferases were seroreactive only in older adults (> 40 years). These antigens could represent markers of cumulative exposure. Alternatively, certain 88 P. falciparum antigens were found to be seroreactive in all age-groups; they included transmembrane emp24 domain-containing protein, ApiAP2, several conserved Plasmodium proteins, MSPs, PF70, members of ETRAMP family, liver stage antigens, rhoptry neck proteins, SYN6, PIESP2, members of the PfEMP1 family, enzymes and proteins involved in translation and transcription, rifin and a few cytoskeleton proteins like dynein heavy chain. In case of P. vivax, 41 antigens were seroreactive in all age groups, most of which were hypothetical proteins and merozoite surface proteins, while other hypothetical proteins and other known proteins such as MSPs and HSPs were exclusively seroreactive in older adults (Additional file 7: Table S5).

Malaria positive falciparum and vivax patients were segregated into in-patients and out-patients as mentioned previously. The in-patients showed a significantly higher antibody response compared to out-patients in both P. falciparum (p = 0.002) and P. vivax (p = 0.01) study groups, respectively. SAM analysis using Log2 FOC values revealed a list of five candidate P. falciparum antigens with differential seroreactivity among the two P. falciparum sub-groups sensitive to hospitalization (p < 0.001, Fig. 5a, Table 2). Three antigens were conserved Plasmodium proteins and two others were dynein heavy chain and Plasmodium exported protein (GEXP07). In case of P. vivax, a hypothetical protein and ubiquitin-like protein (p < 0.001, Table 2) showed differential seroreactivity with hospitalized patients.

When in-patients were further classified into two groups, severe (S) and non-severe (NS), non-severe falciparum patients displayed significantly higher Ab levels compared to severe falciparum patients (p < 0.0001). It is possible that some of these antigens could offer protection from severe disease, although many more diverse clinical events will have to be characterized to fully test this hypothesis. Several conserved Plasmodium proteins, MSP 3.8, PHIST a, NOT family protein and a few others were found to be seroreactive only in non-severe patients (p < 0.001, SAM, Fig. 5b, and Table 2).

Goa is characterized by seasonal malaria transmission, yet many clinical cases are also reported during the dry season. To study the dynamics of IgG Ab responses accompanying changes in intensity of malaria incidence and transmission, patients were segregated based on time of the year during sample collection. A marked increase in seroreactivity was observed during the dry season (post-malaria season) in P. falciparum patients (p < 0.001, SAM). Most of these antigens were seroreactive in both seasons, but a few P. falciparum antigens (e.g. TBP, Pf34, MESA and rifin, along with MSPs and conserved Plasmodium proteins), had higher Ab responses in select patients during the dry season, (Fig. 5c and Table 2). The reasons underlying this observation are not known. In the case of P. vivax infections, only a few patients (66%) were seroreactive in both wet and dry seasons and 31% were seroreactive only during peak wet season.