Intermittent preventive treatment with sulfadoxine-pyrimethamine does not modify plasma cytokines and chemokines or intracellular cytokine responses to Plasmodium falciparum in Mozambican Children

The study was conducted at the Centro de Investigação em Saúde da Manhiça, Manhiça District, southern Mozambique. This area is characterized by a perennial malaria transmission mostly due to Plasmodium falciparum. Anopheles funestus is the main vector, and the estimated average number of infective mosquito bites per person per year was 38 (IDRC and INDEPTH network 2002). For children in this area, SP showed a combined (early and late) therapeutic efficacy rate of 83%, with an in vivo parasitological sensitivity of 78.6% at day 14 [67].

The efficacy study was an individually randomized, placebo-controlled trial [57]. Infants were recruited from those attending the EPI clinic to receive dose 2 of DTP/OPV/Hep B between September 2002 and February 2004. Treatment with SP or placebo was administered at 3, 4, 9 months of age alongside the routine vaccinations. Cross-sectional visits were scheduled at 5, 9, 12 and 24 months of age. For the ancillary immunological studies within IPTi, we included the last 501 children recruited in the main trial. At each cross-sectional visit, capillary blood collected by fingerprick was placed into EDTA microtainers (1 ml), to obtain plasma for extracellular cytokines, and into heparin microtainers (0.5 ml) for intracellular cytokine determinations. All immunological assays were performed by personnel who were blinded to the children's study group. Clinical surveillance for malaria morbidity was done through passive case detection. Written informed consent was obtained from all parents or guardians and ethical approval for the protocol was obtained from the ethics review committees of Mozambique and the Hospital Clinic, Barcelona, Spain. The trial registration number is NCT00209795 http://​clinicaltrials.​gov.

For intracellular cytokine staining (ICS) assays, among those recruited, children with the following criteria were selected: (i) having received all 3 doses of SP or placebo, (ii) having a full set of months 5, 9 and 12 blood samples or the least visits missing, and (iii) having an equal representation of SP- and placebo-recipients. Among those, 258 samples were analyzed from month 5, 208 samples from month 9 and 144 samples from month 12. Once month 24 samples were collected, 143 that had the previous set were analyzed.

For luminex assays, those children who came for all 4 visits or the least visits missing were prioritized, and plasmas were selected for analysis when there were sufficient volumes available: 229 children samples from month 5, 221 from month 9, 225 from month 12, and 158 from month 24.

Plasma cytokines and chemokines IL-2, IL-12 (p70), IFN-γ, IL-1β, IL-6, TNF, IL-4, IL-5, IL-8, IL-13, IL-10, IL-7, IL-17, G-CSF, GM-CSF, MCP-1 (MCAF), MIP-1β were determined using a Bio-plex 100 System (Bio-Rad, CA, USA) and following manufacturer's instructions. According to supplier availability, two types of cytokine detection kits were used (10-plex and 17-plex). Most samples (all except those from 26 children) were tested using the 17-plex kit; the first 27 children (visits 5, 9, 12 months) were tested using the 10-plex kit (which did not contain IL-1β, IL-5, IL-7, IL-8, IL-12, IL-13 IL-17, G-CSF, GM-CSF).

Assays were conducted in 96-well Millipore multiscreen filter plates (Millipore, MA, USA), running in duplicate 8 standards and 40 plasmas per plate. Plasma samples previously stored at -80°C were thawed in a water bath at 37°C, and centrifuged at 1,000 g at 4°C for 10 min. Antigen-coated luminex beads were incubated with patient plasma, in conjunction with manufacturer's reagents. Plasma samples were diluted in Bio-Plex human sample diluent as 1:3 volumes (30 μl sample + 90 μl diluent). To each sample well, the following were added in a step wise manner: 50 μl premixed beads (1×), 50 μl of diluted standard or sample, and 25 μl Bio-Plex detection antibody (1×), and 50 μl streptavidin-PE preparation (1×). The mixture was incubated for 30 min in the dark, at room temperature, on a rotating platform. The wells were then washed 3 times to remove unbound excess serum using a vacuum filtration system and a final incubation was done with streptavidin-PE for 10 min at room temperature while shaking at 300 rpm. Beads were finally resuspended in 125 μl assay buffer, and shaken on a microplate shaker at 1,100 rpm at room temperature for 30 sec immediately before reading the plate on the Luminex 100, in accordance with manufacturer's instructions. Calibration was done using the CAL2 high RP1 target value. Concentrations of unknown cytokines were calculated by plotting unknowns against a 5-parameter logistic regression standard curve of 8 points that determine the limits of detection for each cytokine, expressed as pg/ml. The coefficient of variation (%CV, reproducibility between duplicate values) needed to be < 20% to accept the value for analysis. Results were analyzed using the Bio-Rad Bioplex Manager Software v4.0.

CD3+ specific intracellular cytokines IFN-γ, IL-4 and IL-10 were measured by ICS. Heparinized whole blood was aliquoted into two tubes to test in parallel the stimulated and the non-stimulated (negative/background control) samples. These were respectively incubated with 20 μl extract of P. falciparum schizont lysate at 1 × 10⁸schizonts/ml or 20 μl uninfected erythrocyte extract, both in the presence of 1 μg/ml co-stimulatory antibodies to CD28 (clone CD28.2) and CD49d (clone 9F10) (BD Biosciences, San Jose, CA) for 16 h at 37°C and 5% CO₂. To block cytokine secretion, 1 μM Brefeldin A (GolgiPlug™, Pharmingen, San Diego, CA) was added during the final 4 h of incubation. To arrest activation and avoid loss of activated cells by adhesion to polystyrene tubes, 2 mM EDTA/PBS solution was added for 15 min at room temperature, and tubes were vigorously vortexed before and after the incubation. In order to lyse erythocytes and fix lymphocytes, 5 ml 1× FACS Lysing solution were added. These were vortexed, incubated for 10 min at room temperature, and immediately placed at -80°C. Frozen samples were thawed in batches at a later time for parallel processing and staining, hereby avoiding loss of function or increased background staining. Thawing was done by briefly placing samples in a 37°C water bath and immediately washing off lysing solution using 7 ml Cell Wash buffer. Samples with clumping/debris of dead cells were previously filtered through 70 μm cell sieve filters (Cell Strainer BD Falcon), and pellet was resuspended with a pipette in 2 ml permeabilizing solution before vortexing for 10 min at RT. Prior to staining, frozen cells were washed with 8 ml wash buffer and centrifuged at 1,500 rpm for 10 min at room temperature. Supernatant was decanted, residual pellet was resuspended in 500 μl wash buffer, and 100 μl of each sample were transferred to 96 well round bottom microtiter plates. Plates were centrifuged at 2,000 rpm for 5 min, then flicked to discard the supernatants and finally pellets were resuspended by vortexing before adding the staining antibodies. Samples were stained with a pool of fluorescence-conjugated antibodies: 5 μl CD3 PerCP (clone SK7), 7 μl IFN-γ FITC (clone 25723.11), 3 μl IL-4 PE (clone 8D4-8), and 4 μl IL-10 APC (clone JES3-19F1) (BD Biosciences, San Jose, CA). After incubating at room temperature in the dark for 30 min, samples were washed twice and supernatants discarded. Samples were brought up to a total volume of 300 μl by adding Cell Wash buffer before acquisition. Stained samples were acquired on a 4-color FACS Calibur (BD Biosciences) and 50,000-100,000 total events were collected. Isotype controls were included to evaluate autofluorescence and to assist in setting the quadrants. The data were analyzed with CELLQuest software (BD Biosciences). Lymphocytes were gated using FSC vs. SSC plots and then SSC vs. CD3+ T cells plots were used to isolate CD3+ cells (Figure 1). We analyzed samples in which > 1,000 CD3+ events could be acquired, and in which defined lymphocyte and CD3+ cell populations could be identified, thus excluding those with compromised cell recovery that could confound the readings. Each intracellular cytokine was assessed within its respective fluorescence channel (PerCP vs. FITC or PE or APC). Non-stimulated control samples were used to establish the threshold quadrants of background responses that were applied to quantify % of positive cytokine-producing cells in paired antigen-stimulated samples; results were expressed as the percentage of cytokine (IL-4, IL-10, IFN-γ)-producing CD3+ cells. Samples were considered positive if there was a true difference in response proportions (95% confidence interval [CI]) between the response to test antigens (stimulated) and the background responses to the non-stimulated samples: if the CI was entirely below 0.05%, the response was negative; if the CI was entirely above 0.05%, the response was positive; if the CI overlapped with 0.05%, the response was indeterminate.

Malaria infection was defined as the presence of asexual P. falciparum parasites of any density in a blood smear measured by microscopy and/or by PCR [59]. A clinical malaria episode was defined as a positive blood smear plus an axillary temperature of 37.5°C or history of fever within the prior 24 h. The sensitivity and specificity of these definitions are 100% and 84%, respectively, in infants and 100% and 79.4% in children 1-4 years old [68]. To estimate the incidences, the time at risk was calculated as the number of person years at risk (PYAR; episodes per 365.25 days) since the beginning of the time at risk until the end of follow-up, migration, death or withdrawn consent, whichever occurred first. Children did not contribute to the time at risk or the clinical malaria cases during the lag periods (an arbitrary lag of 21 days was applied after a case of clinical malaria) to account for the effective prophylactic time after antimalarial drug treatment. Months were defined as 30.4 days.

A Shapiro-Wilk normality test revealed a non-normal distribution of data, therefore all analyses were carried out in log scale (Ln) to normalise the data, and averages were expressed as Geometric Means (GM). A Pearson correlation coefficient was used to measure the degree of correlation among cytokines. Correlations were considered "low" if Rho < 0.33, "moderate" if Rho was between 0.33 and 0.66, or "high" if Rho > 0.66. Ratios between cytokines were obtained by dividing the values from each child, transforming that value to its natural log, then calculating the mean for the total number of children, and finally transforming that value to its exponential (GM). T-test and chi-square or fisher exact test were used to compare mean levels or prevalence (tertiles of distribution) respectively, between IPTi treatment groups. Unadjusted and adjusted linear regression models were also used to estimate the effect of IPTi-SP on the cytokine levels at each visit.

To identify what variables were independently associated with cytokine responses (intervention group [SP or placebo], age at visit, neighborhood of residence, current infection and the occurrence of previous clinical episodes [yes/no]), linear regression models with random intercept were estimated, assuming normal distribution of residuals. Selection of the variables for the final model was done through a stepwise procedure, where the criterion for including or excluding a variable was having a p-value ≤ 0.05 or ≥ 0.10 respectively from the Likelihood ratio test.

To evaluate the relationship between each of the individual cytokines at ages 5 and 12 months and the incidence of clinical malaria episodes from 5 to 12 or from 12 to 24 months, respectively, unadjusted and adjusted negative binomial regression models were estimated using a stepwise procedure. In the study of multiple episodes, the probabilities that the criteria to apply Poisson regressions are met are very low, and the phenomenon of over dispersion can occur, resulting in underestimation of CI. The negative binomial regression can estimate part of the variance that Poisson regression cannot identify, and it is exactly the same when there is no over dispersion. Cytokine responses were treated as continuous values, or categorized by tertiles (cut-offs in Additional file 1, Table S1). The incidence rate ratio (IRR) of children with cytokine levels in the highest tertile against those in the lowest tertile was estimated, as well as the IRR per 2-fold increase in the value of cytokines. Co-variates included were: sex, neighborhood of residence, treatment group and previous malaria episodes.

Two additional separate negative binomial regression models were estimated including (i) all the cytokine responses together and (ii) all the cytokine together plus antibody variables assessed in our previous studies (IgG against whole parasite and variant antigens on the surface of infected erythrocytes, IgG [total or isotypes] and IgM against the merozoite antigens AMA-1, EBA-175 and MSP-1₁₉) [58, 59] to identify what immunological variables were independently associated with the subsequent incidence of clinical malaria.

Data analysis was performed using Stata 11 for Windows (Stata Corporation, College Station, TX). Statistical significance was defined as a P of < 0.05. Crude P values reported in the manuscript were interpreted for internal coherence and consistency of results and biological plausibility. In addition, P values were adjusted statistically for multiple comparisons by the Bonferroni correction (P values not reported). Most of the associations were no longer statistically significant (P ≥ 0.05) once adjusted for multiple comparisons, except for a few associations which remained significant (indicated in the text and tables).

The administration of IPTi-SP did not significantly affect the proportion of CD3+ cells producing intracellular IFN-γ, IL-4 or IL-10 in peripheral whole blood samples stimulated ex-vivo with P. falciparum extract antigens, at any of the time points, neither by unadjusted analysis nor by adjusting for previous episodes of clinical malaria (Table 1).

Also, in general, no significant differences were found in plasma cytokine or chemokine responses between treatment groups. There were three isolated exceptions (IL-12 at 9 months, IL-7 at 12 months, G-CSF at 24 months) in which cases cytokine/chemokine concentrations were consistently higher in children who received placebo (Table 1), but associations were no longer statistically significant once adjusted for multiple comparisons. The complete raw data depicting levels of each cytokine and chemokine studied per age and treatment group is shown in Additional file 3, Figure S1.

Cytokine/chemokine levels varied according to the age of the child. All of them showed a decline from 5 to 9 months of age. Thereafter, some cytokines continued decreasing gradually throughout all measurements (plasma IL-2, TNF, IL-4, IL-10), others reached their highest levels at 24 months (plasma IL-12, IL-13, IL-17, MCP-1, G-CSF), and the remaining varied without any specific pattern (Table 2; Additional file 3, Figure S1).

In addition to age, plasma IL-10 levels were also associated with having had previous episodes of clinical malaria (proportional difference = 1.45 [1.14-1.86], P = 0.0028) and having current infection (proportional difference = 1.80 [1.41-2.29], P < 0.0001). IL-12 and IL-13 were also associated with having had previous malaria episodes (proportional difference = 1.35 [1.02-1.77], P = 0.0340; proportional difference = 1.32 [1.04-1.67], P = 0.0234). None of the other factors analyzed in the multivariate analyses (gender, neighborhood of residence, treatment) had any significant association with the cytokines assessed here.

We observed no correlation between the magnitude of IL-4, IL-10 or IFN-γ responses measured intracellularly upon antigen-stimulation and plasma concentrations (data not shown). A few correlations were found between plasma cytokines when analyzed against each other (Additional file 4, Figure S2). Concentrations of Th1 (IL-2, IL-12, IFN-γ) and Th2 (IL-4, IL-5, IL-13) cytokines correlated moderately within themselves, and both these two types showed a moderate/high correlation in between them (Th1 vs Th2). Pro-inflammatory (IL-1β, IL-6, TNF, IL-17) and anti-inflammatory/regulatory (IL-10, IL-7) cytokines correlated moderately within themselves, and both these showed a moderate/low correlation between them (pro- vs anti-inflammatory). Chemokines G-CSF and GM-CSF both correlated moderately/highly to Th1 and Th2 type cytokines, whereas MCP-1 correlated with most anti-inflammatory cytokines and MIP-1 β and IL-8 correlated with most pro-inflammatory cytokines.

We examined whether there was a transition from Th2 to Th1 [55] or from anti-inflammatory to pro-inflammatory cytokines with age in early infancy by analyzing the ratios. Overall, Th1 responses were higher than Th2 responses over the 2 years of life (Additional file 3, Figure S1), and the ratio of IFN-γ:IL-4 increased from 5 to 24 months (Table 3). When data was stratified by treatment, IFN-γ:IL-4 ratios were higher for placebo than for SP recipients at all ages (Table 3), particularly at age 24 months coinciding with the highest prevalence of P. falciparum infection (7.7% at 5 months, 13.5% at 9 months, 8% at 12 months and 26.6% at 24 months). Overall, the pro-inflammatory responses were higher than anti-inflammatory responses over the 2 years of life (Additional file 3, Figure S1), and the ratio of TNF:IL-10 decreased from 5 to 24 months (Table 3).

Elevated cytokine responses, both antigen-specific and non-specific, were not associated with a reduction in the subsequent incidence of clinical malaria in unadjusted or adjusted analyses (Table 4). In some cases there was an association between higher cytokine levels and higher malaria incidence and this was manifested at certain age intervals (5-12 months or 12-24 months). In particular, intracellular IFN-γ and IL-10 production by CD3+ cells was associated with higher incidence of malaria during the first year of age but not in the second year (Table 4). Adjusting for previous episodes did not affect these results. After Bonferroni correction for multiple comparisons, a non significant trend association was still observed (Table 4).

Associations between individual plasma cytokines and incidence of malaria were observed throughout the two years, but mostly in the second year. High concentrations of IL-10 and MCP-1 in plasma were associated with higher malaria incidence in both time intervals and by all analysis methods (Table 4). High IL-13 was also positively associated with incidence of malaria as analysed by all methods but only during the second year. A few disperse significant associations for plasma IFN-γ, IL-1β, IL-4, IL-5, IL-7, G-CSF and GM-CSF were also observed during the second year, when adjusted the analyses by the variables included in the multivariate model; all were in the same direction of higher concentrations correlating with increased malaria risk (Table 4). After correction for multiple comparisons, some statistically significant associations remained for MCP-1, IFN-γ, IL-10 and IL-5 (Table 4).

We did a multiple regression analysis in which we used a stepwise procedure to determine the variables that were independently associated with incidence of malaria. All cytokines were included in the list of variables taken into account for the stepwise procedure. Significant associations were only observed between higher levels of intracellular IFN-γ at 5 months and higher malaria incidence in the follow up interval 5-12 months (IRR 1.28, 95% CI 1.03-1.58, P = 0.0234). Similarly, higher concentrations of plasma IFN-γ (IRR 1.15, 95% CI 1.05-1.27, P = 0.0040) and MCP-1 (IRR 1.60, 95% CI 1.22-2.10, P = 0.0007) at 12 months were associated with higher malaria risk during the 12-24 month period. In contrast, in this analysis higher plasma concentrations of IL-17 at 12 months were associated with lower malaria incidence during the interval 12-24 months (IRR 0.80, 95% CI 0.69-0.93, P = 0.0035). Other co-variates that were significantly associated with malaria incidence, independently of the cytokine concentrations, were having had a previous episode in the second year of life (IRR 4.63, 95% CI 2.59-8.26, P < 0.0001), and the child's neighborhood of residence in both time intervals (P < 0.0001).

When antibody variables from prior studies were also included in the stepwise model together with all cytokines and co-variates, significant associations were only observed for higher intracellular IFN-γ at 5 months and higher malaria incidence during the follow up interval 5-12 months (IRR 1.26, 95% CI 1.01-1.57, P = 0.0381). At 12 months, higher plasma IFN-γ (IRR 1.26, 95% CI 1.11-1.44, P = 0.0004) and IL-5 (IRR 1.27, 95% CI 1.04-1.55, P = 0.0203) correlated with higher incidence of malaria up to 24 months, while higher IL-17 plasma concentrations were associated with lower malaria risk during the 12-24 month period (IRR 0.84, 95% CI 0.71-0.98, P = 0.0300). Likewise, the additional variables that independently showed significant associations with malaria risk were having had a previous episode in the second year of life (IRR 3.16, 95% CI 1.64-6.09, P < 0.0006), and the neighborhood in both time intervals (P < 0.0001).

Finally, no significant differences in cytokine or chemokine levels were found in relation to parasite density (data not shown). Levels of cytokines or chemokines could not be analysed in relation to incidence of severe malaria due to its low incidence (Additional file 2, Table S2).