The genomics of t'ef and finger millet domestication and spread

Abstract

The Northern Highlands of Ethiopia and Eritrea (NHE) were a centre for food production in Africa, hosting one of the earliest agriculture-based complex societies on the continent. The NHE's geographical connections with the Arabian Peninsula, and Nilotic cultures led to the cultivation of southwest Asian crops and African native domesticates in its territory. Additionally, the NHE were also the domestication centre for crops like t'ef (Eragrostis tef (Zucc.) Trotter) and finger millet (Eleusine coracana L. Gaertn L.), after well-adapted local wild plants. Considering the paucity of the archaeobotanical record in the region and food remains' preservation issues, in this study, we aim to investigate the domestication and spread of t'ef and finger millet using genomics and interpreting the results in the light of archaeological proxies. Our data confirmed Eragrostis pilosa and Eleusine coracana subsp. africana as the sole wild progenitors of t'ef and finger millet, respectively. T'ef was initially domesticated in the NHE before spreading into southern Ethiopia and eastwards into southern Arabia. Finger millet spread followed two routes: one leading eastwards through the Red Sea to India, and the other southwards, through Kenya and Uganda, reaching southern Africa.This article is part of the theme issue 'Unravelling domestication: multi-disciplinary perspectives on human and non-human relationships in the past, present and future'.

Keywords: DArTSeq; Ethiopia; archaeobotany; diversity; origins of agriculture.

Figure 1.

Population structure of 94 wild and domesticated Eragrostis accessions, based on 4455 DArTSeq SNPs. (A) First and second principal components of a PCA. Each point is an accession, the closer the points are the more genetically similar they are. Accessions are coloured according to the taxonomical classification reported in the germplasm bank’s passport data. (B) NJ phylogenetic tree based on pairwise genetic distances calculated in the R package adagenet. (C) Graphical representation of a Q-matrix of the Structure K = 2 model. Vertical lines correspond to accessions, and they are apportioned according to the proportional membership of each one of the two modelled clusters, or the proportion of alleles each accession gets from a hypothetical ancestral population.

Figure 2.

Population structure of 60 Ethiopian t’ef accessions, based on 5311 DArTSeq SNPs. (A) The first and second principal components of a PCA, with accessions coloured according to the cluster in the Structure K = 3 model from which they have they get most of their proportional membership. (B) Proportional memberships in the Structure K = 3 model with each accession represented in a geographical map as a pie chart in the location it was collected and with each slice indicating the proportional membership to each of the three groups.

Figure 3.

Population structure of 46 wild and domesticated Eleusine accessions, based on 5674 DArTSeq SNPs. (A) First and second principal components of a PCA. Points are coloured according to the taxonomical classification reported in the germplasm bank’s passport data and their country of origin. (B) NJ phylogenetic tree based on pairwise genetic distances calculated in the R package adagenet. (C) Graphical representation of a Q-matrix of the Structure K = 3 model.

Figure 4.

Population structure of 39 cultivated finger millet accessions, based on 2044 DArTSeq SNPs. (A) First and second principal components of a PCA, with accessions coloured according to their provenance. (B) Proportional memberships in the Structure K = 3 model with each accession represented in a geographical map as a pie chart in the location it was collected and with each slice indicating the proportional membership to each of the three groups.

Figure 5.

Proposed routes of dispersal for t’ef and finger millet from their core area of domestication based on the geographical distribution of present-day genetic diversity.