TLR9 polymorphisms in African populations: no association with severe malaria, but evidence of cis-variants acting on gene expression

Patient samples were collected as part of ongoing epidemiological studies of severe malaria at the Royal Victoria Hospital, Banjul, The Gambia (885 severe malaria cases, 629 controls; 1,044 family trios) and the Queen Elizabeth Central Hospital, Blantyre, Malawi (712 severe malaria cases, 416 controls; 248 family trios). The set of nuclear family trios comprise a severe malaria affected child and its two (biological) parents, and were assessed as 'true trios' using the Nuclear software package[18]. All DNA samples were collected and genotyped following approval from the relevant research ethics committees and informed consent from participants.

All cases were children admitted to hospital with evidence of Plasmodium falciparum on blood film and clinical features of severe malaria [19]. Subjects were defined as having had 'cerebral malaria' if their Blantyre coma score was less than or equal to 3 on presentation or early during admission. A second phenotypic subset of 'severe malarial anaemia' was defined as those subjects having had a haemoglobin density of less than 5 g/dl or a haematocrit less than 15%. Participants with co-existing severe or chronic medical conditions (e.g. bacterial pneumonia, kwashiorkor) unrelated to a severe malarial infection were excluded. Controls were cord blood samples obtained from birth clinics in the same locality as the cases, and thought to approximate a random sample of the population thus, reflecting the true allele frequency. Cerebral malaria (CM) and severe malarial anaemia (SMA) were the most common symptoms defining severe malaria.

For the association studies (participants described below) genomic DNA samples underwent whole genome amplification through either Primer Extension Pre-amplification [20] or Multiple Displacement Amplification [21], before genotyping on a Sequenom MassArray genotyping platform. Only the four common SNPs (minor allele frequency (MAF) of at least 5% in the HapMap Yoruba population) in the TLR9 region were genotyped. The majority of polymorphisms in this region are monomorphic.

For the ASE assay 19 unrelated CEU and 20 YRI individuals were used, selected from the HapMap collection. Established lymphoblastoid cell lines (LCLs) from these individuals were obtained from the Centre d'Etude du Polymorphisme Humain (CEPH) collection (Coriell Institute for Medical Research). Cell lines were cultured has described previously [22].

For expression analysis, total RNA was extracted from cells lines from each individual using Tri reagent (Sigma). Poly A+ RNA was isolated from total RNA using the Dynabeads mRNA Purification Kit (Dynal). Synthesis of First-Strand cDNA was processed according to the StrataScript™ First-Strand Synthesis System (Stratagene). The Allelotype platform from Massarray (Sequenom) was utilized for accurate relative quantification of allele specific cDNA and gDNA transcripts [23]. Primers were designed using the dedicated software Spectrodesigner (Sequenom). Allele specific expression quantification was performed according to the protocol of Forton et al [15] and Campino et a l [24]. Briefly, for each cDNA and gDNA assays, nine technical replicates were performed. Biological replicates from independent cultures were assayed equally. For a given cDNA or gDNA assay the allelic transcript ratio was calculated on each of the nine technical replicates from the relative quantity of the two allele-specific transcripts. The mean allelic transcript ratio for the whole assay was then calculated, and normalized to the mean allelic transcript ratio for the genomic controls. An assay was accepted for further analysis if the standard error of the mean for the technical and biological replicates was less than 10%.

Haplotypes were downloaded from the HapMap Phase II database http://​www.​hapmap.​org. For data analysis, 770,394 markers were used; all monomorphic markers were excluded from the analysis. Locations of the SNP markers were based on those of the human reference sequence http://​genome.​ucsc.​edu of May 2004 (hg 17, build 35).

Genotypic deviations from Hardy-Weinberg equilibrium (HWE) were assessed using a chi-square statistical test. Case-control association analysis was undertaken by logistic regression and included the covariates: ethnic group, gender and the HbS polymorphism. Family-based association analysis was performed using the transmission disequilibrium test (TDT) [25], as well as a case-pseudo-control approach with conditional logistic regression [26]. The TDT test is robust to the effects of population stratification or sub-structure. All analysis was performed using the R statistical package.

For the association analysis, log-transformed ratios of allele expression values for individuals with the heterozygous genotype at the exonic SNP (rs352140) were used. To map putative cis-variants a statistical analysis approach was used, based on the linear regression method (similar to that developed by Teare et al and Campino et al)[24, 27]. In particular, it was assessed whether there is evidence of an increasing or decreasing (linear) relationship between the ratio of expressions for the phase-known heterozygous type 1 (e.g. AG), homozygous types (e.g. AA and GG), and heterozygous type 2 (e.g. GA) genotypes. The rationale for the approach is that allelic imbalances would be observed only in individuals heterozygous for the cis-acting polymorphism. All analysis was performed using the R statistical package.

Of the four common SNPs in the TLR9 region genotyped in the Malawian and Gambian cohorts, two were located in the promoter (-1486 T>C rs187084, -1237 C>T rs5743836), one was intronic (+1174 A>G rs352139) and the last was a synonymous SNP located on exon 2 (+2848 G>A rs352140). The observed MAFs observed in the cohorts are presented in Table 1. There is no evidence of distortion from Hardy-Weinberg equilibrium (HWE) in the controls or parents (P > 0.05). There is a high degree of concordance of the allele frequencies within country and between the two study designs. The intronic SNP (+1174-X) is in high linkage disequilibrium (LD) with the other SNPs, especially with the exonic SNP (+2848-X) with which it shows a near complete LD (D' = 0.95 in the Malawian cohort and 0.99 in the Gambian) (Figure 1).

The allele frequencies between the Gambian cases and controls and among the parents and children from the trios are very similar (odds ratios (OR) close to 1 in both studies) (see Table 2). In the Malawian case-control cohort we observed a weak association between allele 1174-A and severe malaria (OR = 1.21, 95% CI 1.01 – 1.45; p = 0.04). The allele 2848-T showed a marginal protective effect (OR = 0.82, 95% CI 0.68 – 1.00; p = 0.05). The + 1174 and +2848 SNPs are in near perfect LD (D' close to one) and only three possible haplotypes exist (1174-A/2848-C, 1174-G/2848-C, 1174-G/2848-T). The relative risk of disease (95% CI) for haplotypes 1174-G/2848-C, 1174-G/2848-T and 1174-A/2848-C are 0.95 (95% CI: 1.01–1.33), 0.86 (95% CI 0.80–1.11) and 1.15 (95% CI: 0.73–1.03) with p-values of 0.05, 0.45 and 0.10 respectively. However the effect of these polymorphisms was not reproducible in the trios, where allele frequencies between parents and children were comparable, and there was no evidence of transmission distortion and association (Table 2). In the case-control analyses, the allelic tests do not adjust for any confounding effects, such as population substructure, that can affect bias estimates in case-control study designs. In the genotype analysis of the case-controls (Table 3), adjusted for the HbS sickle cell polymorphism, and also ethnicity in the Gambian study, no strong signals of genotypic association was found for any of the four polymorphisms.

Candidate gene genetic studies are generally restricted to SNPs covering the gene area as potential regulatory SNPs localized away from a gene are difficult to identify. Therefore, as no association was found with SNPs in the TLR9 gene area, there was further exploration into whether other SNPs which can be far away from the gene could have an important role on the regulation of TLR9 expression.

Numerous studies have recently contributed to the identification of putative regulatory variants by correlating total gene expression or allele specific expression data with genetic variability[11, 16, 17, 28] Using a similar approach, in this report an allele specific expression (ASE) assay was performed on a panel of HapMap B-lymphoblastoid cell lines (LCL) for which genotyped data is available and facilitates the mapping of functional polymorphisms. These cell lines have been widely used to map cis/trans regulatory regions and recent publications have shown high consistence between cis-regulation results obtained using HapMap cell lines and primary tissue[29].

The strength of the ASE assay is the possibility to compare the relative expression of the two alleles in heterozygous individuals within the same cellular sample. Therefore, both alleles will be exposed to the same environmental, technical and genetic factors (e.g. trans-acting effect). If the expression of the alleles is not equal (allelic imbalance), it suggests that the expression of the gene is under cis- regulation.

Cell lines from individuals of northern and western Europe origin (CEU) and from Yoruba of Ibadan, Nigeria (YRI) were selected from the HapMap collection [13] and were genotyped for a exonic synonymous SNP present on the TLR9 gene. Since this SNP is located within a part of the gene that is transcribed, it is possible to discriminate between allelic transcripts in heterozygous individuals.

Twenty-one unrelated CEU and nineteen YRI individuals were found to be heterozygous for this SNP and were used for the ASE analysis. For these individuals, their genomic DNA was used to obtain the reference equi-molar ratio between the two alleles (allele1 and allele2), and this was then compared with their allelic RNA expression. Allelic expression imbalance (AEI) was determined if independent replicate assays in the RNA samples showed differential allele expression ratios that deviated from the genomic DNA. If there is no allelic imbalance then the ratio of the two alleles (allele1/allele2) will be equal to one. A ratio greater than one indicates an over-expression of allele1, while a ratio smaller than one indicates an over-expression of allele2.

Of the 21 CEU individuals, 17 (84%) showed no or low magnitude allelic variation (both alleles expressed nearly equally) (Figure 2). However, for three individuals, the allelic transcript A was clearly over-expressed (ratios between 1.3 and 1.87) and for one individual this allelic transcript was clearly under-expressed (average ratio of 0.5). As for the CEU population, almost all the YRI samples (80%) showed low magnitude allelic variation (Figure 2). However, three individuals showed over-expression (ratios between 1.2 and 1.67) of allele A, and one individual relatively under-expression of the same allele (average ratio of 0.7). Biological replicates from independent cultures yielded convergent results (Pearson correlation coefficient r = 0.84). Since the overall experimental variability was very low, being little influenced by variations in cell culture or other experimental techniques, these results strongly suggest that the allelic imbalances observed in the six individuals studied here represent a true biological phenomenon. As the allelic imbalance data observed was bidirectional, i.e. showing over-expression of different alleles among the cell lines, the cis-acting variant is likely to be in low LD with the transcript SNP [15, 27].

To map potential cis- acting variants that might be responsible for the allelic imbalances observed in the TLR9 gene, the allele ratio data is correlated with haplotypes (~400 kb surrounding each side of the TLR9 gene) of the HapMap individuals considered using a linear regression approach (described in materials and methods) [15, 24]. CEU and YRI sample sets were analysed independently. No significant associations were found between allelic imbalances and the 278 SNPs surrounding TLR9 gene including the promoter and intronic SNPs analysed (Figure 3). The fact that it was not possible to map any putative cis-variants might be because the allelic imbalances observed were caused by variants still not represented by HapMap data at the current level of SNP ascertainment or due to the reduced statistical power resulting from the small sample size used (21 CEU and 19 YRI).

Overall, while no functional polymorphism were mapped using the current HapMap genotyping data, this data shows that TLR9 expression is potentially modulated through cis-regulatory variants.