Genetic differentiation and geographical structure
The coefficient of population differentiation with no account the distances among haplotypes (Gst) and the coefficient of differentiation based on the pairwise difference between alleles that takes into account the distances among haplotypes (Nst) were found respectively, equal to 0.196 and 0.374; but the difference was not significant (P > 0.05). In Citrullus spp. Mujaju et al. (2011) found Gst = 0.56 and Nst = 0.49 for sweet watermelon and Gst = 0.71, Nst = 0.81 for cow watermelon. The fact that the differentiation parameter based on the pairwise difference between alleles is greater than the one calculated without permutation (i.e. Nst > Gst) indicates that the collection is characterized by clear geographic structure (Grivet, 2002; Dane et al., 2007; Guicking et al., 2011). Also the significant value of the total gene diversity across all four geographical regions (hT = 0.917, standard error = 0.0320) is indicating a strong structure in the population (Pons & Petit, 1996; Sun et al., 2019; Zhao et al., 2019).
Levi et al. (2017) observed that accessions of C. colocynthis were sub-divided into five groups in general agreement with their centres of diversification and origin. Our findings indicated that regional genetic differentiation statistics support Levi et al. (2017)’s conclusions, with sub-samples from different regions exhibiting genetic differentiation associated with their likely centers of diversification. Also, haplotypes of C. amarus were mostly grouped in Southern Africa, which is assumed to be the origin of that species (Dane & Liu, 2007; Chomicki & Renner, 2015).
Citrullus chloroplast sequences analysis with TCS 1.21 resulted in a network where haplotypes widely sampled throughout West Africa were placed at the root. While coalescence theory predicts that older alleles will prevail in a population due to a higher number of descending lineages and associated wider geographic distributions (Crandall & Templeton, 1993), such an observation may depend on sample sizes and evolutionary/domestication histories and also the lack of subs. cordophanus (from northeast Africa) in the germplasm studied. In this study, H1 is the most frequently sampled haplotype and has the most connections with other haplotypes; thus H1 may be considered the most ancient haplotype. This ancient haplotype was sampled most frequently in West Africa (i.e. Nigeria and Benin) and was highly shared by accessions of both C. lanatus and C. mucosospermus . These results support the findings of Chomicki and Renner (2015) and Renner et al. (2019) who used eleven gene regions to infer phylogeny among Citrullus species, and also a 3500-year-old leaf sample from the Egyptian tomb to infer close relationship between C. lanatus and C. mucosospermus . Our findings, based upon a large set of egusi melon and watermelon accessions from four continents, provide further evidence of that close relationship between these two species. However, they are indeed two different species, as previous crosses between them (e.g. Charleston Gray x PI 560006) resulted in high levels of sterility (Gusmini et al., 2004). The very limited haplotype diversity among the two species suggests an old split, with chlorotype fixation (Dane & Liu, 2007) and ancient types of C . mucosospermus originating from Western Africa (Renner et al., 2014). However, to the best of our knowledge, no wild populations have been confirmed in West Africa. Spontaneous plants may have been found earlier, but those individuals certainly escaped from cultivation. A region-wide collecting mission by the first author yielded no wild population of C. mucosospermus in West Africa (Achigan-Dako et al., 2015) though, the presence in West Africa of the ‘neri’ type [Fig. 9f in Achigan-Dako et al. (2015) and Fig. 1 in Minsart et al. (2011)], another cultivated egusi melon that exhibits smaller seeds with yellow soft coat, should be highlighted as a contributor to the genepool ofCitrullus is the region. While this neri type (C. lanatus ) is morphologically distinct from C. mucosospermus , it has been rarely studied.
Archaeological evidence indicates north-east Africa as a center of origin and domestication (Chomicki et al., 2020). Authors reported wild dessert watermelon in that region (Paris, 2015) or the genetic affinity with the C. lanatus var. cordophanus (a sweet white-fleshed cultivar) (Renner et al., 2019). However, within the genus Citrullus mucosospermus remains the closest relative species to C. lanatus . The presence of an ancient haplotype in West Africa on the one hand and the close relationship between C. lanatus and subsp. cordophanus of Darfur in north-eastern Africa as revealed by Renner et al. (2019) on the second hand, calls for further molecular and archaeological investigations to generate sufficient knowledge on newly published results, including those reported here. New molecular investigations should include more materials from Sudan and neighbouring countries where wild populations of watermelon have been found (Paris, 2015). Moreover, our data showed that one of the Egyptian accessions (PI 525083), indicated to beC. amarus and observed by Levi et al. (2013) to cluster with dessert watermelon, exhibits a unique haplotype (H32). That accession is several mutations away from C. colocynthis and closer to watermelon and egusi melon haplotype. Previous findings of Levi et al. (2017) showed that PI 525083 rather clustered with C. lanatus var. lanatus . In addition, the hypothesis that watermelon is from north-eastern Africa does not explain how an endemic species such asC. mucosospermus shares the same haplotype with dessert watermelon, while other accessions from the region (e.g. PI 525083) shows unique haplotype. If C. lanatus did indeed spread to the world from west or north-eastern Africa, how and when was it domesticated in those region as New Kingdom Egyptians were cultivating sweet red-fleshed watermelon more than 3500 years ago? From which species was C. mucosospermus domesticated? Through what mechanisms was C. lanatus spread to Asia, and when? More germplasm collections from all continents are necessary to fully understand the phylogeographical relationships among Citrullusspecies. In Africa the focus should be on both west and north-eastern regions to resolve the domestication history of modern cultivars.