In collaboration with Dr. Fernando Polack (Vanderbilt University) and the Infant Foundation in Buenos Aires, Argentina we recruited a prospective cohort of children aged 0-9 months of age from 5 hospitals in Buenos Aires. The primary clinical outcome is severity of disease (O2 saturation <93%) and the secondary outcomes are RSV titer and Th2 polarization (blood lymphocytes are isolated and challenged with pro-inflammatory agonists including LPS). From 2003-2006, we recruited approximately 800 children to the study: 400 infected with RSV (240 with severe disease, 160 with mild disease) and 400 controls (uninfected). Our initial genetic analyses have focused on the role of functional SNPs in TLR4 (Asp299Gly and Thr399Ile), and NRF2 (-653A/G, -651G/A and -617C/A). These genes have been identified as candidate susceptibility genes for RSV infection in mice, and TLR4 has also been associated with RSV infection in children. In our initial analyses, we found that the -651G/A SNP was only found in severe RSV cases and conferred greater risk of disease relative to the wild type (OR: 1.9;CI: 1.4, 2.7). We also found that children heterozygous for the Asp299Gly/Thr399Ile mutations in TLR4 were at higher risk of severe RSV disease relative to children with wild-type TLR4 genotype (OR: 1.7;CI: 1.2, 2.6). However, when socioeconomic status (SES) was considered, we found children from low SES families and heterozygous for TLR4 Asp299Gly/Thr399Ile mutations had significantly less severe disease compared to wild type (4% severe disease vs 11% severe disease, p<0.05) compared with children from middle/high SES (9% severe disease vs 2% mild disease, p<0.05). These results are consistent with a role for TLR4 in RSV disease severity, but interaction of TLR4 genotype with environment must be considered when evaluating risk of these mutations. We have used human lymphoblastoid cell lines (LCL;Coriell Institute) to evaluate inter-individual variation in response to RSV, test the role of candidate susceptibility genes, and identify basal gene expression patterns that predict RSV responsivity. Databases on human sequence variation (dbSNP and HapMap) and functional annotation were queried for relevant single nucleotide polymorphisms (SNPs) and used to select LCLs with particular genotypes with SNPs of interest. These LCLs were infected with RSV and significant inter-individual variation in RSV infectivity across multiple LCLs indicated that genetic background is an important determinant of suscepti bility to infection. Moreover, enhanced (3-5 fold) viral load was found in LCLs with a non-synonymous coding SNP (rs469390) in MX1, which is involved in suppressing viral replication in infected cells. In a case control study of infants with mild or severe RSV disease, a positive association was found between individuals homozygous for the minor allele of rs469390 and development of severe disease. We also developed a model to predict RSV infection using data from individuals in the HapMap collection. Baseline mRNA expression of HapMap individuals from six different microarray data sets were initially assessed for correlation with viral load. Of transcripts that were positively correlated, the 5 highest and 5 lowest viral load responders were selected for the model, resulting in 61 genes that associated with viral load and were differentially expressed in at least 3 of 6 baseline data sets. Using heuristic modeling of the 61 genes, a 27 gene transcript model was the best predictor (RSV prediction, p<0.001;rsq 0.95) of viral load in individuals for which infection response had already been quantified. We found 75% accuracy in predicting RSV infectivity in test set LCLs with similar expression profiles. This novel cell model of RSV disease can thus be used translationally to identify functionally relevant candidate susceptibility genes. We also developed a mouse model of RSV disease by infecting 34 inbred mouse strains (JAX) with RSV-19 or vehicle control. Significant inter-strain variation in protein, cellular, mucus cell metaplasia (MCM), and RSV-N mRNA expression phenotypes following infection were found among the 34 inbred strains. The most severe RSV disease pathology was found in A/J and Balb/cJ mice, while C3H/HeJ mice were among the most resistant. Collaborating with Dr. Tim Wiltshire (UNC), we then used the SNPster algorithm for haplotype association mapping (HAM) of maximum RSV response phenotypes. Significant associations were found for MCM, RSV-N mRNA, and BAL monocytes, lymphocytes, and PMNs. The most significant association was found on chromosome 1 for BAL monocytes. Within this region are 6 genes that have SNPs that associate with BAL monocytes, including Marco. We identified a non-synonymous coding SNP (rs30741725) that causes a threonine to serine substitution in amino acid 476, and based on predicted protein structure change, this substitution severely affects receptor binding. Further, RSV-induced BAL monocytes and lymphocytes were significantly greater 5 d PI in strains that are homozygous for this SNP relative to strains that have the wild type allele, suggesting a functional role for the mutation in RSV disease. We also found that relative to Marco+/+ mice, significantly greater histopathological changes and numbers of monocytes and PMNs were found in Marco-/- mice 5 d PI providing additional evidence for a protective role for Marco in RSV disease. Collaborating with Dr. Fernando Polack, each subject from the RSV cohort (described above) was genotyped for a functional SNP in the MARCO promoter (rs3806496;chrom 2, 119.698574 Mb). Logistic regression was used for analysis where the response variable was O2 saturation. The allele frequency of MARCO (T allele) was 0.47 and in HWE. The likelihood ratio test was 16.8 with 1 df, p=4x10-5. The relative risk estimate for the T allele is 0.66 which means that it is protective or the C allele (RR=1.52) is causative. While essential to repeat these findings in another independent population, these novel studies support the hypothesis that MARCO is an important determinant of RSV disease severity, and demonstrates the translational value of the model to understand human RSV disease. We then used a combined genetics and genomics approach (genetical genomics) to understand the genetic basis of RSV disease susceptibility. To identify transcripts differentially expressed at baseline that predict response to RSV, baseline lung gene transcript expression for 29 inbred mouse strains (Novartis) was correlated to phenotype data from the RSV strain screen (described above) using linear and random forest regression. These analyses identified a battery of gene transcripts that were differentially expressed at baseline and significantly correlated with RSV disease severity including Ptgs2 (prostaglandin-endoperoxide synthase 2), Nqo1 (NAD(P)H dehydrogenase, quinone 1), and IL18bp (interleukin 18 binding protein). To determine whether genotype differences account for differential gene expression we utilized expression quantitative trait locus (eQTL) analysis using FastMap. At baseline, there were 125 genes with significant cis-eQTLs, where an association exists between gene expression and genotype at that particular genes locus. Expression of three of these genes significantly correlated with RSV disease phenotypes. Together, these approaches have identified genetic markers of susceptibility to RSV disease to predict severe responders and potentially provide more effective therapeutic targets.
Showing the most recent 10 out of 13 publications
