Evidence for Pathogen-Driven Selection Acting on HLA-DPB1 in Response to Plasmodium falciparum Malaria in West Africa

Abstract

African populations remain underrepresented in studies of human genetic diversity, despite a growing interest in understanding how they have adapted to the diverse environments they live in. In particular, understanding the genetic basis of immune adaptation to pathogens is of paramount importance in a continent such as Africa, where the burden of infectious diseases is a major public health challenge. In this study, we investigated the molecular variation of four Human Leukocyte Antigens (HLA) class II genes (DRB1, DQA1, DQB1 and DPB1), directly involved in the immune response to parasitic infections, in more than 1000 individuals from 23 populations across North, East, Central and West Africa. By analyzing the HLA molecular diversity of these populations in relation to various geographical, cultural and environmental factors, we identified divergent genetic profiles for several (semi-)nomadic populations of the Sahel belt as a signature of their unique demography. In addition, we observed significant genetic structuring supporting both substantial geographic and linguistic differentiations within West Africa. Furthermore, neutrality tests suggest balancing selection has been shaping the diversity of these four HLA class II genes, which is consistent with molecular comparisons between HLA genes and their orthologs in chimpanzees (Patr). However, the most striking observation comes from linear modeling, demonstrating that the prevalence of Plasmodium falciparum, the primary pathogen of malaria in Africa, significantly explains a large proportion of the nucleotide diversity observed at the DPB1 gene. DPB1*01:01, a highly frequent allele in Burkinabé populations, is identified as a potential protective allele against malaria, suggesting that strong pathogen-driven positive selection at this gene has shaped HLA variation in Africa. Additionally, two low-frequency DRB1 alleles, DRB1*08:06 and DRB1*11:02, also show significant associations with P. falciparum prevalence, supporting resistance to malaria is determined by multigenic and/or multiallelic combinations rather than single allele effects.

Keywords: Africa; HLA; human molecular diversity; malaria; pathogen‐driven selection; plasmodium falciparum.

FIGURE 1

Map of the populations sampled. Map of the populations sampled and investigated in this study, with their sampling location and linguistic families. Dashed lines indicate the different geographical regions considered here. BED: Senegal‐Bedik; MAN: Senegal‐Mandenka; SRR: Senegal‐Serer; SEF: Senegal‐Fulani; DOG: Mali‐Dogon; MAF: Mali‐Fulani; GUR: BurkinaFaso‐Gurmantche; GRS: BurkinaFaso‐Gurunsi; MOS: BurkinaFaso‐Mossi; BAG: Chad‐BaggaraArabs; DAN: Chad‐Dangaleat; DAZ: Chad‐Daza; MAB: Chad‐Maba; AMH: Ethiopia‐Amhara‐(Keketeya); ORO: Ethiopia‐Oromo; BEJ: Sudan‐BejaHadendoa; NUB: Sudan‐Nubians; RAS: Sudan‐RashaaydaArabs; SUD: Sudan‐SudaneseArabs; ALA: Algeria‐(Annaba); ALC: Algeria‐(Constantine); ALG: Algeria‐(Ghardaia); ALT: Algeria‐(Tamanrasset); AMI: Morocco‐Amazigh‐(Amizmiz); ASN: Morocco‐Amazigh‐(Asni); FIG: Morocco‐Amazigh‐(Figuig).

FIGURE 2

Non‐Metric Multidimensional Scaling analyses (nMDS). Non‐Metric Multidimensional Scaling analyses (nMDS) performed on Reynolds distances Θ _w computed for each locus. The nMDS for DQA1 and DPB1 (top right and bottom right, respectively) were computed on two axes, whereas the nMDS for DRB1 and DQB1 required a third axis to reduce the Stress value (left for DRB1 with top: Axes 1 and 2 and bottom: Axes 1 and 3; middle for DQB1 with top: Axes 1 and 2 and bottom: Axes 1 and 3, respectively). Dashed lines between two populations indicate non‐significant Θ _w 's. Gridlines are spaced 0.05 apart in all plots to highlight the scale differences between the loci. Colors correspond to geographic regions, that is, W‐AFR: West Africa; C‐AFR: Central Africa; E‐AFR: East Africa and N‐AFR: North Africa. Short population names correspond to BED: Senegal‐Bedik; MAN: Senegal‐Mandenka; SRR: Senegal‐Serer; SEF: Senegal‐Fulani; MAF: Mali‐Fulani; GUR: BurkinaFaso Gurmantche; GRS: BurkinaFaso‐Gurunsi; MOS: BurkinaFaso‐Mossi; BAG: Chad‐BaggaraArabs; DAN: Chad‐Dangaleat; DAZ: Chad‐Daza; MAB: Chad‐Maba; AMH: Ethiopia‐Amhara‐(Keketeya); ORO: Ethiopia‐Oromo; BEJ: Sudan‐BejaHadendoa; NUB: Sudan‐Nubians; RAS: Sudan‐RashaaydaArabs; SUD: Sudan‐SudaneseArabs; ALC: Algeria‐(Constantine); ALT: Algeria‐(Tamanrasset); AMI: Morocco‐Amazigh‐(Amizmiz); ASN: Morocco‐Amazigh‐(Asni); FIG: Morocco‐Amazigh‐(Figuig).

FIGURE 3

Allelic and molecular diversity indexes. Top: Violin plots of the allelic richness (ar) and expected heterozygosity (He). Bottom: Violin plots of the average nucleotide diversity per site (π _n ), average number of polymorphic sites (S _n ) and Tajima's D distributions. All statistics are computed for the four HLA loci and 23 populations under study (20 for DQB1). Arrows with “n.s.” indicate non‐significant differences between two distributions (Kruskal‐Wallis test, α = 0.05, fdr correction (Benjamini and Hochberg 1995), see Table 1 for adjusted Pvalues).

FIGURE 4

HLA sequences associated with Plasmodium falciparum prevalence. For the 5 HLA alleles identified as strongly correlated with Plasmodium falciparum prevalence (pfpr2000, Table S10.1), relationship between the observed frequency in each of the 23 population (20 for DQB1) and pfpr2000 at their sampling location. Three of these alleles, namely DPB1#66, DRB1#3144 and DRB1#3155, remain significantly associated with pfpr2000 after fdr correction for multiple testing. See Table S3 for the correspondence between sequence names and nominal HLA alleles.

FIGURE 5

Correspondence analysis. Correspondence analysis (axes 1 and 2) performed on the allele frequencies of 20 populations (BED, MAF and ASN were excluded due to the lack of data at DQB1). Axes 1 and 2 capture 21% and 9% of the total variance, respectively. The continuous color scale pfpr2000 corresponds to Plasmodium falciparum prevalence in year 2000 at the sampling locations of each population represented. For the sake of clarity, allele names are reported without the locus prefix but are colored according to the locus (pink: DRB1; violet: DQA1; blue: DQB1; green: DPB1). All alleles with frequencies above 20% in at least one population as well as all alleles displaying a significant association with pfpr2000 either before (underlined) or after (double underlined) fdr correction for multiple testing are represented. Allele DPB1#66 is shown in a larger format because of its very high frequencies in some populations. See Table S3 for the correspondence between sequence names and nominal HLA alleles.