High production of pro-inflammatory cytokines by maternal blood mononuclear cells is associated with reduced maternal malaria but increased cord blood infection

The study was conducted at the Centro de Investigação em Saúde de Manhiça (CISM) in southern Mozambique [36]. CISM runs demographic and morbidity surveillance systems at the Manhiça District Hospital (MDH) and nearby health posts where standardized information on paediatric outpatient visits and admissions is collected. Recruitment and follow up of participants were done at the Maragra Health Post (MHP) (September 2005–March 2009). Transmission of P. falciparum is perennial with marked seasonality (warm rainy season November–April, cool dry season rest of year) and of moderate intensity.

HIV-negative pregnant women resident in Manhiça were recruited during the third trimester of pregnancy at the antenatal clinic [35]. Exclusion criteria included birth weight < 2 kg, same gender twins, congenital malformations, birth asphyxia or apparent health problems. 349 eligible newborns were enrolled.

Samples were collected at delivery: 10 mL of maternal venous blood (EDTA vacutainer), 8 mL of cord blood (EDTA vacutainer), blood slides, bloodspots onto filter papers and placental tissue. When delivery occurred outside the maternity, only maternal blood samples were collected.

Children were followed up until age 24 months. Weekly active case detection was conducted from birth to age 10.5 months, and monthly home visits from 10.5 to 24 months. Children with fever (axillary temperature ≥ 37.5 °C) or history of fever in the preceding 24 h were taken to the MHP for examination, parasitaemia and haematocrit assessment. Passive case detection was carried out at the MHP and MDH to monitor attendances to outpatient/inpatient clinics.

Parasitaemia was quantified in blood slides at CISM following standard quality-controlled procedures [35]. Haematocrit was measured after centrifugation in heparinized microcapillary tubes with a haematocrit reader (Hawksley & Sons Ltd, Lancing, UK), and full blood counts performed using a KX-21N cell counter (Sysmex Corporation, Kobe, Japan). Placental biopsies collected from the maternal side were placed into 10% neutral buffered formalin and presence of parasites or pigment evaluated [37]. Real-time quantitative PCR (qPCR) for P. falciparum 18SrRNA (ABI PRISM 7900HT Fast Real-Time System) using TaqMan^® probe and FAM/TAMRA, in duplicates, in filter papers of maternal and cord blood [38].

Peripheral blood mononuclear cells (PBMC) and cord blood mononuclear cells (CBMC) were isolated using a Ficoll-Hypaque 1.077 gradient, resuspended in complete medium (RPMI 1640 culture medium with glutamine and antibiotics plus decomplemented 10% fetal calf serum), and stimulated as previously reported [39]. Briefly, 1.2 million fresh PBMC and CBMC were incubated with 20 μL of a 3D7 P. falciparum schizont extract corresponding to 2 million synchronized infected red blood cells (iRBC, lysed by freezing–thawing cycles, endotoxin and Mycoplasma tested), or with 20 μL of lysate from uninfected RBC (uRBC, control) in 24-well plates in a total volume of 590 μL. Supernatants were collected after 24, 48 or 72 h cultures and frozen at − 80 °C. The PBMC and CBMC pellets were collected in Trizol (InVitrogen) and frozen at − 80 °C. Supernatants were shipped in dry ice to ISGlobal where they were thawed and IL-12p70, IFN-γ, IL-2, IL-10, IL-8, IL-6, IL-4, IL-5, IL-1β, TNF, TNF-β quantified with the Bender Human T_H1/T_H2 11plex FlowCytomix Multiplex Kit (MedSystems, Aachen, Germany) as previously described [39]. Cytokine production in culture supernatants of unstimulated maternal and neonatal samples (uRBC) was not subtracted from the iRBC stimulated samples [40], but shown side by side as it is possibly biologically relevant [41‐43]. Limits of detection for cytokine/chemokine concentrations are as reported [39].

Messenger RNA (mRNA) and complementary DNA (cDNA) from 44 maternal PBMC samples stimulated for 24 h with parasite iRBC lysate and uRBC control were purified and cytokine transcript levels measured at QIMR [44]. RNA was extracted from PBMC in Trizol (Invitrogen, Carlsbad, CA, USA). Total RNA was reverse transcribed using Superscript™III R (Invitrogen). cDNA was quantified by reverse transcriptase qPCR (RT-qPCR) in triplicate using commercially available primers and probes (Taqman™, Applied Biosystems, CA, USA) specific for human IFN-γ (Hs00174143_m1), TNF (Hs00174128_m1), IL-2 (Hs00174114_m1), IL-4 (Hs00174122_m1), IL-6 (Hs00174131_m1), IL-10 (Hs00174086_m1), IL-13 (Hs00174379_m1) and ribosomal protein L13a (RPL13a) (Hs03043885_g1) as reference gene (Apte JI 2010, unpublished). Thermal cycler parameters (Corbett Rotor-Gene 3000, Mortlake, Australia) were 95 °C for 2 min and 60 °C for 30 s, 45 cycles at 95 °C for 5 s, and 60 °C for 30 s. Results were calculated by the number of molecules of each sample as determined from the standard curve for each gene, which was then standardized against the number of molecules of RPL13a (reference gene) for that sample.

Pregnant women were classified into primigravidae (PG) and those with at least one previous pregnancy (multigravidae, MG). Maternal age was categorized as ≤ 20, 21–24 and ≥ 25 years. Distance to the river was calculated with Hawth’s tools in ArcGIS software (ESRI) as the mean distance of each neighbourhood centroid to the Incomati river, and neighbourhoods classified as those adjacent to the river (mean distance ≤ 2.5 km) or those at a greater distance (> 2.5 km) [45]. Season during pregnancy was defined as dry if the gestation period included May–October, and rainy if it included ≥ 4 rainy months (November–April). Maternal malaria was defined as parasites in peripheral blood by microscopy and/or qPCR. Placental malaria was defined as parasites and/or pigment by histology. Placental inflammation was defined as presence of > 5 mononuclear cells and/or polymorphonuclear leukocytes in intervillous spaces, assessed in 10 high power fields at 400× [2]. Congenital malaria was defined as parasites on cord blood by qPCR. The primary case definition of clinical malaria in children was fever or history of fever plus P. falciparum parasites of any density [35].

Cytokine and chemokine concentrations (pg/mL) were logarithmically transformed. Continuous variables were compared by ANOVA and categorical variables by Fisher’s exact test. Spearman’s test and scatter plots assessed correlations among cytokines and chemokines. The Bonferroni correction was applied to p values to adjust for multiple comparisons in correlation analyses. Linear regression models were used to assess whether age, parity, neighbourhood, season (rainy vs dry), use of insecticide-treated net (ITN), or use of indoor residual spraying (IRS), affected the magnitude of maternal and fetal cytokine and chemokine responses and whether there was an interaction between parity and these variables. Analyses were done univariate and multivariable, adjusting for all the other variables. When a significant interaction with parity was found, an additional set of regression analyses was performed stratifying women by parity. Logistic (for infection and inflammation) or linear (for weight and haematocrit) regression models were used to evaluate the effect of doubling the levels of maternal and fetal cytokines and chemokines on pregnancy outcomes: P. falciparum infection (maternal, placental, cord), placental inflammation, maternal haematocrit and birth weight. Analyses were adjusted for age, parity, season, neighbourhood, ITN, IRS and, when applicable, P. falciparum infection. In women with peripheral parasitaemia, Spearman’s correlations between parasite density (microscopy and qPCR) and cytokine and chemokine levels, and between parasite density and placental inflammation, were performed. Adjusted negative binomial regression models evaluated the effect of doubling the levels of maternal and fetal cytokines and chemokines on the incidence of multiple malaria episodes in children up to 12 months of age. Significance was defined at p < 0.05. Crude p values are interpreted for internal coherence, consistency of results and biological plausibility. Analyses were performed using Stata/SE 10.1 (College Station, TX, USA).

To determine the optimal timepoint for cytokine and chemokine secretion in supernatants after stimulation with iRBC and uRBC lysates, a pilot study was conducted with 44 maternal PBMC and 44 matched CBMC samples (28 aparasitaemic and 16 parasitaemic mothers). Cytokine and chemokine concentrations were compared at 24, 48 and 72 h (n = 264 supernatants). In general, there was not a significant difference in cytokine and chemokine concentrations between timepoints. For some cytokines (IFN-γ, IL-6, IL-10, TNF) responses tended to be higher at 24 h and thus this timepoint was chosen for subsequent studies. Another pilot study was conducted to determine the optimal timepoint for cytokine mRNA production with 11 maternal PBMC samples and the 24 h incubation was also selected as the most appropriate.

Analyses reported here are based on cytokine and chemokine data obtained from 174 women (Table 1, see Additional file 1) in whom PBMC and/or CBMC of enough quantity and quality were obtained, cells could be stimulated with antigen ex vivo and supernatants processed. Table 2 presents the Spearman’s ρ coefficients and p values (Bonferroni adjusted) of the correlation analyses. Pro-inflammatory cytokines and chemokines IL-1β, IL-6, TNF and IL-8 were highly correlated (Fig. 1 PBMC supernatants, see Additional file 2 CBMC supernatants, Additional file 3 mRNA). A significant correlation was also found between pro-inflammatory cytokines and the regulatory cytokine IL-10. Moderate correlations were seen among T_H1 cytokines, T_H2 cytokines, and also between T_H2 and T_H1 cytokines (Table 2). Inverse correlations were observed for IL-8 and IFN-γ or IL-12. No substantial differences were found between peripheral and cord cytokines but peripheral responses tended to be slightly higher (see Additional file 4). Some cytokine and chemokine responses appeared not to be malaria-specific, as spontaneous release of pro-inflammatory cytokines was also found in control stimulations (see Additional file 5).

The correlations between secreted proteins and intracellular mRNA cytokines were also examined. For the positive IFN-γ, TNF, IL-10 and IL-6 responses, there was a high correlation between the levels of cytokines secreted in the supernatants and the mRNA levels contained within the PBMC (Fig. 2). However, a number of supernatant cytokine responses were below the detection limit of the FlowCytomix kit but were measurable as mRNA by RT-qPCR, and this rendered the correlations non-significant (Fig. 2). This could be because (i) the RT-qPCR method was more sensitive, and/or (ii) the mRNA transcribed in the cytoplasm was not translated into protein and secreted to the supernatant, and/or (iii) the cytokines/chemokines secreted were bound to receptors and internalized by the cells. For IL-2 and IL-4, concentrations were overall low. Subsequent analyses included only supernatant proteins due to the limited number of samples in which mRNA cytokine levels were quantified.

Concentrations were significantly affected by age, parity, seasonality and ITN use, but not by neighbourhood or IRS. After ex vivo stimulation with iRBC and uRBC, PBMC from older women secreted significantly higher levels of IL-1β and IL-6 than younger women and the differences remained significant after adjustment (Table 3). Upon iRBC stimulation, CBMC from MG secreted significantly higher levels of IL-8 than PG women (Table 3). A significant interaction between age and parity was observed for IL-5 production by PBMC (p = 0.026). After stratifying the analyses by parity, higher IL-5 concentrations in PG but not in MG women were significantly associated with older age (Table 3).

Depending on the season during pregnancy, the magnitude of IL-1β, IL-6, TNF and IL-8 responses at delivery varied significantly for both peripheral and cord blood cells following incubation with iRBC and uRBC. Lower levels of these pro-inflammatory cytokines and chemokines were significantly and consistently associated with the dry season (Table 3). In addition, cord IL-2 and IL-4 after stimulation with iRBC were also significantly lower in the dry season.

When examining the effect of malaria control tools on cytokine levels, CBMC from women sleeping under ITN produced higher levels of IL-10 than those not using ITN (Table 3). A weak interaction was noted between parity and ITN (p = 0.15); in the analyses stratified by parity the significant effect of ITN on IL-10 levels appeared to be attributable to MG women (Table 3).

Finally, maternal haematocrit was significantly associated with cord IL-8 in the adjusted analysis (ratio of mean IL-8 levels by haematocrit = − 2.46, 95% CI [− 4.28; − 0.63], p = 0.009 for iRBC, and − 2.83 [− 5.13; − 0.52], p = 0.018 for uRBC).

Lower production of IL-1β by PBMC following incubation with iRBC and uRBC was significantly associated with maternal peripheral infection (Table 4). Some interactions between peripheral infection and parity were noted for TNF (p = 0.054), IL-12 (p = 0.061), IL-2 (p = 0.003), IFN-γ (p = 0.025), IL-8 (p = 0.151), and IL-5 (p = 0.031). In the analysis stratified by parity, higher TNF and lower IL-12 secretion after iRBC and uRBC stimulations were significantly associated with maternal peripheral infection in PG women (Table 4). In contrast, lower IL-8 and higher IL-2 were associated with peripheral infection in MG women (Table 4). Such associations were not observed for CBMC stimulations (p > 0.05).

When taking into account parasite density in infected women, only IL-5 production by CBMC correlated inversely with parasitaemia by microscopy (Spearman’s ρ = − 0.67, p = 0.023). In contrast, when parasites were quantified by qPCR, a consistent inverse correlation emerged for PBMC T_H1 cytokines IFN-γ (ρ = − 0.34, p = 0.06), IL-2 (ρ = − 0.35, p = 0.070) and IL-12 (ρ = − 0.41, p = 0.025), and a direct correlation for PBMC TNF (ρ = 0.38, p = 0.042) and CBMC T_H2 cytokines IL-4 (ρ = 0.40, p = 0.032) and IL-5 (ρ = 0.36, p = 0.060), but no association was shown for IL-1β, IL-6, IL-10 or any other cytokine. Geometric mean parasite densities by qPCR but not by microscopy were significantly higher in women with placental inflammation (10.38 parasites/µL, standard deviation 4.62) than without (0.46 parasites/µL, standard deviation 1.63, p = 0.034).

Placental infection was significantly associated with lower production of IL-6 and TNF by PBMC (Table 4). Higher IL-5 production after iRBC stimulation was significantly associated with placental infection in MG (p = 0.023, Table 4). Placental inflammation was significantly associated with higher TNF secretion by PBMC (Table 4), only significant for PG in the stratified analysis.

Cord blood infection was significantly associated with higher concentrations of IL-1β secreted by PBMC (Table 4). It was not possible to perform this multivariable analysis stratified by parity due to the low number of congenital infections (n = 8: 2 PG, 6 MG).

Finally, cytokine or chemokine concentrations were not significantly associated with maternal haematocrit, birth weight or incidence of clinical malaria in the children (p > 0.05).