Environmental Correlation Analysis for Genes Associated with Protection against Malaria.

Margaret J Mackinnon,Sophie Uyoga,Gavin Band,Dominic P Kwiatkowski,Kirk A Rockett,Robert W Snow,Anna Rautanen,Thomas N Williams,Carolyne Ndila,Alex Macharia

doi:10.1093/molbev/msw004

Abstract

Genome-wide searches for loci involved in human resistance to malaria are currently being conducted on a large scale in Africa using case-control studies. Here, we explore the utility of an alternative approach—“environmental correlation analysis, ECA,” which tests for clines in allele frequencies across a gradient of an environmental selection pressure—to identify genes that have historically protected against death from malaria. We collected genotype data from 12,425 newborns on 57 candidate malaria resistance loci and 9,756 single nucleotide polymorphisms (SNPs) selected at random from across the genome, and examined their allele frequencies for geographic correlations with long-term malaria prevalence data based on 84,042 individuals living under different historical selection pressures from malaria in coastal Kenya. None of the 57 candidate SNPs showed significant (P < 0.05) correlations in allele frequency with local malaria transmission intensity after adjusting for population structure and multiple testing. In contrast, two of the random SNPs that had highly significant correlations (P < 0.01) were in genes previously linked to malaria resistance, namely, CDH13, encoding cadherin 13, and HS3ST3B1, encoding heparan sulfate 3-O-sulfotransferase 3B1. Both proteins play a role in glycoprotein-mediated cell-cell adhesion which has been widely implicated in cerebral malaria, the most life-threatening form of this disease. Other top genes, including CTNND2 which encodes δ-catenin, a molecular partner to cadherin, were significantly enriched in cadherin-mediated pathways affecting inflammation of the brain vascular endothelium. These results demonstrate the utility of ECA in the discovery of novel genes and pathways affecting infectious disease.

Highlights

Of the 57 single nucleotide polymorphism (SNP) loci representing 39 “candidate” genes selected for the first phase of the MalariaGEN Consortium large multipopulation case-control studies (MalariaGEN 2008), only five of these loci—the sickle cell-causing allele of the beta-hemoglobin gene (HBB), the “O” allele in the ABO gene that determines ABO blood group, both of which were well established as malaria protective prior to the advent of GWAS studies, G6PD, CD40LG, and ATP2B4—were confirmed (MalariaGEN 2014)
We applied environmental correlation analysis” (ECA) in a human population living under different malaria transmission intensities in a small geographic area in coastal Kenya in order to search for genes involved in resistance to malaria
We examined 57 candidate loci representing 39 genes that have been implicated in malaria pathogenesis: while 15 of these showed significant (P < 0.05) malaria-related clines in allele frequency, none of these reached significance when compared to a large set of randomly chosen loci and when population genetic structure in the background genome was controlled for in the analysis

Summary

Introduction

Many genes affecting susceptibility to malaria have been reported in the literature (Kwiatkowski 2005), their validation in large, multisite, genome-wide phenotype– genotype association studies (GWAS) has been disappointing, producing only weak signals (Jallow et al 2009; Timmann et al 2012; MalariaGEN 2014) or inconsistent results across different studies (Atkinson et al 2007; Cserti and Dzik 2007; Fry, Auburn, et al 2008; Clark et al 2009; Mangano et al 2009; Teo et al 2010; MalariaGEN 2014). One explanation for the discrepancies between results from large, multipopulation GWAS, and single-site studies may lie in the profound degree of genetic diversity seen over very small distances in African populations (Tishkoff and Williams 2002) which, due to undetected population structure, coupled with variation in disease transmission, can generate both false positive and false negative results (Marchini et al 2004) Another explanation is the lack of power of marker-based genome scans as a consequence of low levels of linkage disequilibrium in African genomes (Conrad et al 2006; Jallow et al 2009).

Objectives

Results

Conclusion