Polymorphisms in genes of interleukin 12 and its receptors and their association with protection against severe malarial anaemia in children in western Kenya

This study was part of Asembo Bay Cohort Project, an immuno-epidemiologic birth cohort project conducted from June 1992 through September 1996. Characteristics of this study have been detailed elsewhere [24]. Briefly, the study was carried out in a rural area of western Kenya where more than 95% of the residents belong to the Luo ethnic group, malaria transmission is holoendemic, and severe anaemia is a major cause of malaria-related mortality in children less than two years of age. Pregnant women were enrolled in their second or last trimester; their newborn babies were enrolled soon after birth and followed up to three to five years of life. Village monitors visited each mother-child pair every two weeks for clinical observation. At monthly intervals (and between intervals when children were sick), blood samples were obtained for measurement of parasitaemia and haemoglobin concentrations. Resulting data, along with time to first documented malaria infection, parasite density of first infection, and cumulative treatment with sulphadoxine/pyrimethamine were available as part of routine data collection. For the current genetic association study, longitudinal clinical information up to two-year follow-up from the time of enrollment was used since this age group accounted for most of the observed malarial anaemia [25].

Inclusion criteria for children in this genetic association study included having at least 1 monthly follow-up visit up to two years of age and availability of sample. Mean numbers of visits including both monthly follow-ups and sick visits during first two years of life were 19.8 (SD = 11.5) for the included children. Additionally, for genetic association analyses, it is important that all participants are genetically unrelated individuals. Therefore, only children from singleton births or one of the twins were included in the study. If more than one family member (siblings) was eligible based on the above criteria, only the first child in birth order was included in the study.

Informed written consent was obtained from parents/guardians of all study participants. This study was approved by the Institutional Review Board of the Kenya Medical Research Institute and Centers for Disease Control and Prevention.

A classification scheme [26], based on previous epidemiological analysis of the same cohort [25], was used in this study. SMA was defined as a haemoglobin concentration of less than 6 g/dL and the presence of P. falciparum greater than 10,000/μl. Three other subphenotypes used in this study were: severe anaemia (SA) (Hb < 6 g/dL), high-density parasitaemia (HDP) (> 10,000 P. falciparum parasites/μl), and severe anaemia with any level of parasitaemia (SAP) (Hb < 6 g/dL with any detectable P. falciparum parasitaemia). All data points collected monthly and at sick visits for two years in children were used and the clinical phenotypes are presented as episode crude incidence/1,000 person-months.

For tagging SNP (tagSNP) selection, LDSelect software at default settings (r² > 0.65) was used to select tagSNPs from all common variation within four IL12 related genes from a public database of SeattleSNPs [27]. LDSelect uses an efficient selection algorithm to select tagging SNPs based on linkage disequilibrium (LD) statistic r² and does not require direct haplotype inference [28]. At each round of selection, the binning algorithm identifies single SNP, which exceeds threshold r² with the maximum number of other SNPs, and sets this group of SNPs as a bin. Then each SNP within the bin is analysed to determine whether it exceeds the threshold r² with all other SNPs in the bin. All SNPs in a bin that meet this criterion are designated as tagSNPs. Only one tagSNP needs to be typed per bin. Binning criteria for tagSNP selection used in this study were minor allele frequency (MAF) cut-off of 10% and an r² threshold of 0.65. LD varies across different populations [29, 30]; therefore, only genotype data for the African-American samples present in the database was used, as this population is most similar to the samples in this study from western Kenya. Fifty five tagSNPs, 7 in IL12 A, 10 in IL12B, 22 in IL12RB1, and 16 in IL12RB2 were selected. Many more SNPs were needed for IL12RB1 gene than for other genes tested due to its larger size and the genetic diversity.

DNA was extracted from frozen blood pellets using a QiaAMP DNA purification kit (Qiagen). For each sample, 10 ng DNA was spotted into 384-well plates (ABgene), dried and frozen until just before use in genotyping assays. SNPs were genotyped using the MassARRAY^® iPLEX assay (Sequenom)[31]. Multiplex PCR primers (up to 29-plex) were designed in a region of approximately 100 base pairs around the SNP of interest and an extension primer was designed immediately adjacent to the SNP using MassArray Assay Design software (version 3.1). The single base primer extension step was performed using a standardized cycling programme. Primer extension products were dispensed on 384-well SpectroChip^®s that were automatically read by a Compact™ MALDI-TOF mass spectrometer (Sequenom). Data analysis was performed on MassArray Typer software (version 3.4).

Genotypes were individually examined via a detailed quality control process involving duplicate calling of genotypes, control samples, evaluation of missing calls, and Hardy--Weinberg Equilibrium (HWE) testing. The iPLEX genotype clusters were manually checked. Only SNPs with > 90% calling rate and MAF > 2% were including in the final analysis. Each SNP was analysed using univariate methods and then multivariate Poisson regression analysis to ascertain the association between genotypes and malaria-associated morbidity. Generalized estimating equations and an independent working correlation structure to adjust for correlation between multiple visits from the same individual were used. In the multivariate Poisson regression, SNPs were independently evaluated as covariates using one of four genetic models: dominant ((AA + Aa) vs aa), recessive (AA vs (Aa + aa)), additive (aa vs Aa vs AA, which treat the marker as a continuous variable), and heterozygous advantage (Aa vs (AA + aa)) with three levels based on the most common allele frequency. Here, "A" refers the common allele. The results of the Poisson regression were reported as rate ratios (or risk ratios, RR). Multivariate models were adjusted for confounders, sickle cell type and treatment with anti-malarial drugs since these two factors were significantly associated with all four clinical phenotypes in univariate analysis (Table 1). Data was analysed using SAS software package (version 9.1). Bonferroni adjusted critical alpha level of 0.007 for IL12A, 0.006 for IL12B, 0.005 for IL12RB1, and 0.004 for IL12RB2 were used. Haploview was used to obtain values for LD and perform haplotype block analysis of SNPs in the same gene [32, 33]. D prime values (D') were calculated as statistical values for pairwise LD analysis between SNPs.

Among 1,459 children enrolled in ABCP, 913 meeting inclusion criteria (62.6%) were selected for genotyping. In this study, incidence of at least one SMA episode in the same individual was 16.5% during the first two years of life, 44.5% for SA, 79.5% for HDP, and 40.0% for SAP. Only sickle cell type and history of anti-malarial use were significantly associated with all four clinical phenotypes of interest in the unadjusted analysis (Table 1). These two covariates were therefore included in the final model of multivariate Poisson regression analysis. Other covariates, such as bed net use, amount of rain in previous 120 days, parasitaemia in the mothers at the time of delivery, and low birth weight (< 2500 g) or pre-term birth (< 37 weeks gestation), were not significantly associated with all four clinical phenotypes although parasitaemia in mother was associated with SA only and low birth weight and pre-term birth were associated with HDP.

Fifty-five tagSNPs were selected for genotyping from genes encoding IL12 A, IL12 B, IL12RB1, and IL12RB2. Eight SNPs were not polymorphic in this population, and nine did not perform sufficiently in the multiplexed assay or pass quality control process (Table 2). Thirty-eight SNPs were included in the final analysis (Additional file 1).

This study identified an association with SMA for two SNPs, rs2243140 of IL12A and rs429774 of IL12RB1, both of which were associated with protection from SMA (Table 3). Specifically, for rs2243140 of IL12A, an increasing copy number of minor allele C (TT vs TC vs CC) was significantly associated with a reduced risk of SMA (P = 0.006). For rs429774 of IL12RB1, individuals possessing two copies of minor variant C appeared to be associated with a reduced risk of SMA (P = 5 × 10^-5). The risk for developing SMA over time based on risk ratios calculated from multivariate Poisson regression was reduced by 82% (95% CI: 31%-95%) if carrying two copies of minor allele (CC) of rs429774 compared with those carrying one or no copies of the C allele. No statistically significant interaction between these two SNPs (P = 0.38) was found. These two minor alleles are common in the population studied here, 11% and 24%, for rs2243140 and rs429774, respectively (Table 2).

Another SNP in IL12A, rs2243138, was associated with protection from SA (P = 0.004) and SAP (P = 0.004) (Table 4). The children homozygous for the rare allele (TT), of which the prevalence is 21.8% in this population, had significantly lower risk for developing SA (RR = 0.69, 95% CI: 0.53-0.89) and SAP (RR = 0.66, 95% CI: 0.49-0.89). A similar trend of association was seen with SMA, but was not statistically significant after adjusting for multiple comparisons (Additional file 1).

Another SNP in IL12RB1, rs383483, was associated with HDP (P = 0.003) (Table 4). Having an AG genotype, which is common (45.8%) in this population, was significantly associated with susceptibility to HDP at 1.21 (95% CI: 1.07-1.36) times greater risk compared with those with AA or TT genotypes. No association between SNPs from IL12B and IL12RB2 and malaria-related morbidity was identified.

Because the single SNP regression analysis demonstrated that multiple sites within IL12A and IL12RB1 genes are significantly associated with protection from severe malaria anaemia (Additional file 1), analysis of LD in these two genes (Figure 1) was performed respectively. The two hit SNPs of IL12A, rs2243140 and rs2243138, were in high LD (D' = 0.98), but SNPs of IL12RB1, rs383483 and rs429774, were not (D' = 0.47). This suggests that the associations shown by SNPs rs2243140 and rs2243138 of IL12A may be related whereas the rs383483 and rs429774 of IL12RB1 are independent.