Genetics of human populations: evolutionary and epidemiological applications = Genética de las poblaciones humanas: aplicaciones evolutivas y epidemiológicas

Abstract

The present thesis deals with a deep analysis of the genetic structure of extant human worldwide populations, with a special focus on the Mediterranean.

Different genetic markers, 18 Alu insertions and 367 single nucleotide polymorphisms (SNPs), were analysed for the first time in general populations and Bedouins from Jordan. The same SNPs were also genotyped in a consistent set of samples from the Mediterranean Basin. Specifically, Alu markers were used to assess the genetic diversity within Jordan and to define the relationships across present-day Middle Eastern, North African, and European populations. While SNPs located in the top coronary artery disease (CAD) risk regions (1p13, 1q41, 9p21, and 10q11) were used to explore the genetic variation and the LD patterns across 19 populations from Europe, Middle East and North Africa, together with Asian and African samples from the 1000 Genomes Project. Furthermore, Tunisian and Moroccan case-control samples were used to test for the first time the applicability of European GWAS results in these four genomic regions to North African populations. Finally, simulated genomic data were used to evaluate the consistency of association results across European, Asian and Sub-Saharan African case-control samples.

Neutral and CAD risk markers pointed out a low but significant geographic structure in the Mediterranean area. As a particular case in the Mediterranean context, Bedouins and general population from Jordan showed statistical differences (p=0.038). The genetic heterogeneity detected is likely related to the fact that urban areas have been subject to several external influences but Bedouins have preserved their own genetic background because of their nomadic and isolated lifestyle.

The general patterns of genetic variation observed in the four CAD risk regions analysed are in agreement with the genetic structure generated by demographic processes. Despite this fact, the present work also identified potential signals of positive and balancing selection in some specific risk markers and haplotypes. Actually, potential signatures of positive selection in the 9p21 region and of balancing selection in 9p21 and 10q11 have been observed. The regions under positive or balancing selection are shared across continents but the specific markers under selection are, in some cases, continent-specific.

Results from the association study in North Africa revealed significant association signals in the four genomic regions analysed in Moroccan and Tunisian samples confirming the trans-ethnic importance of these genomic regions for the risk of CAD. However, in the meta-analyses performed with the North African case-control samples, only the 9p21 and 10q11 regions showed association signals and none of the North African risk SNPs was associated with CAD in Europe. These arguments suggest the existence of continent-specific CAD risk variants in these chromosomal regions.

Significant differences in the LD and in the haplotype patterns confirm the heterogeneity in the genomic structure between North Africans and Europeans. In addition, differences in the LD structure persist even across Europeans, North Africans, Asians, and Sub-Saharan Africans.

Finally, the simulation study highlighted an overall consistency in effect direction across populations of different ancestry, specially between Europeans and Asians. Because simulations assumed no actual differences in genetic effect size between populations and because the rate of between-population effect size differences generated by our study was considerably lower than the rate reported from empirical studies using real data, our study supports the existence of real differences in the genetic risk architecture of complex traits between populations. In general, this study confirmed that a substantial fraction of association signals are shared across people of different ancestries. However, some discrepancies in the effect size across populations highlighted the importance of accounting for population structure in association studies.

Esta tesis analiza la variación genética de muestras poblacionales europeas, africanas y asiáticas, con particular énfasis en la región mediterránea. Se usaron 18 marcadores Alu para evaluar la diversidad genética dentro de Jordania (en beduinos y en población general) y para definir la relación entre las actuales poblaciones del Mediterráneo. Por otra parte, 367 polimorfismos de un solo nucleótido (SNPs) localizados en cuatro regiones de riesgo para la enfermedad arterial coronaria (CAD) (1p13, 1q41, 9p21 y 10q11) se utilizaron para explorar la variación genética en 19 poblaciones de Europa, Oriente Medio y África del Norte, junto a las muestras asiáticas y africanas del Proyecto 1000 Genomas. Estas cuatro regiones de riesgo CAD fueron analizadas también en un conjunto de muestras de casos y controles de Marruecos y Túnez. Por último, se simularon muestras de casos y controles africanos, europeos, y asiáticos para evaluar el nivel de consistencia interpoblacional en los estudios de asociación.

Como caso particular en el Mediterráneo, los beduinos y la población general de Jordania mostraron diferencias significativas (p=0,038). Los beduinos se agruparon con los norteafricanos, mientras que los jordanos generales se agruparon con las poblaciones más orientales del Mediterráneo.

Los patrones generales de variación observados en las cuatro regiones de riesgo CAD mostraron una estructura genética plausiblemente modelada por procesos demográficos. A pesar de esto, el presente trabajo también detectó señales potenciales de selección natural positiva en la región 9p21 y de selección equilibradora en las regiones 9p21 y 10q11.

También fueron detectadas asociaciones significativas en las cuatro regiones genómicas analizadas, previamente identificadas en europeos, en Marruecos y Túnez. Sin embargo, sólo las regiones 9p21 y 10q11 mostraron asociación en el meta-análisis realizado utilizando las muestras de casos y controles norteafricanos.

El estudio de simulación destacó una alta replicabilidad de las señales de asociación entre los distintos continentes. Sin embargo, algunas discrepancias en el tamaño del efecto revelaron la posible existencia de diferencias interpoblacionales en la arquitectura del riesgo genético de caracteres complejos y destacaron la importancia de incluir la estructura poblacional en los estudios de asociación.