First genome-wide association study of non-severe malaria in two birth cohorts in Benin

Abstract

2 20 ABSTRACT 21 Recent research efforts to identify genes involved in the susceptibility to P. falciparum malaria have 22 focused on severe forms of malaria, with several genome-wide association studies (GWAS) and 23 multi-center analyses published this past decade. Here we present the first GWAS performed on 24 mild malaria susceptibility in young children, designed to identify genetic variants involved in 25 innate immunity or innate resistance mechanisms. Two cohorts of infants from southern Benin (525 26 and 250 individuals respectively), used as discovery and replication cohorts, were closely followed 27 from birth to 18-24 months of age, with an assessment of a space-and time-dependent risk of 28 exposure to vector bites. GWAS was performed on 15.5 million genotyped and imputed variants, 29 with the susceptibility of infants to mild malaria attacks and to malaria infections as a whole (both 30 symptomatic and asymptomatic). Infant susceptibility to malaria was assessed by considering all 31 malaria events occurring during the follow-up using a Cox-model for recurrent events. We found 32 strong statistical support for a role of PTPRT, a tyrosine phosphatase receptors involved in STAT3 33 pathway, with the protection against both mild malaria attacks and malaria infections (p=9.70x10 -8 34 and p=1.78x10 -7 respectively in the discovery cohort, and both p < 0.05 in the replication cohort).35 Furthermore, our study highlights several other genes, among them, UROC1, ACER3 and the 36 PLAG2 cluster, whose biological functions are relevant in malaria infection. Results show that 37 despite the difficulty of setting up such longitudinal field studies, GWAS on non-severe malaria can 38 successfully identify new candidate genes and inform physiological mechanisms underlying natural 39 protection against malaria. 40 41 AUTHOR SUMMARY 42 Malaria remains a major worldwide public health problem with circa 219 million cases and 435,000 43 deaths per year. As for many infectious diseases, the genetics of the host plays a role in the disease 44 course. So far, most studies on such genetic factors focused on severe malaria forms. Our study 45 aimed to find factors associated with simple forms (asymptomatic forms and uncomplicated clinical 3 46 forms). We used dense genome-wide data of two cohorts of infants closely followed during two 47 years in Benin, and we incorporated an environmental risk of exposure estimated at individual level 48 from entomological, climatic, and environmental data. This allowed us to unravel several genes that 49 appear to play a role in the susceptibility to malaria infection, thus demonstrating the feasibility of a 50 genome-wide approach on non-severe forms. Functional studies are needed to confirm the role of 51 the genetic factors highlighted here. Some of these genes represent interesting targets for future 52 prevention strategies and drug development. 4 53 INTRODUCTION 54 55 In spite of numerous prevention and control efforts in recent years, malaria remains a major 56 global public health ...

Mixed logistic regression in genome-wide association studies

“…We analyzed two data sets with a simulated binary phenotype. Simulations were based either on real genotype data from a recent GWAS [ 16 ], or on large cohort simulated with a coalescent model. Both data sets present cryptic relatedness and population stratification, with two strata (or two cohorts) with different disease prevalence.…”

“…We retrieved similar results in an analysis of a simulated phenotype with large differences of prevalence among strata, based on genotype data from two geographically close cohorts ( ca. 20 km apart) from South Benin [ 16 ], but with different self-reported ethnicities. Heterogeneity of prevalence may result from environmental factors (e.g.…”

“…We used genotype data from a GWAS on mild malaria susceptibility performed on two cohorts in South Benin [ 16 ]. The participants were ascertained in two sites distant of 20 km from each other, in three different health centers for the first cohort and two for the second one.…”

“…To illustrate the method, we applied it on the data from the GWAS in South Benin described in [ 16 ]. We used as binary phenotype the presence/absence of any malaria infection during the follow-up (there were 229 individuals with no infection and 546 with at least one infection).…”

“…In the sequel, we evaluate the capacity of logistic regression (LR), MLM, and MLR (using either the score test or the methods mentioned above) to properly take into account population structure associated with heterogeneous prevalence. While Chen et al were interested in geographically distant populations, spanning several countries in South America, part of our work focuses two geographically very close populations in West Africa, using real genotype data from a recent GWAS [ 16 ]. We also use data simulated with a coalescent model [ 17 ], reproducing the simulations presented in [ 14 ].…”