Genetic diversity and population structure of Brachiaria brizantha (A.Rich.) Stapf accessions from Ethiopia

Brachiaria is a tropical, warm-season grass native to Africa. It is an extensively cultivated forage in the tropics with proven benefits on livestock productivity. Brachiaria is well-known for high biomass production, animal nutrition, carbon sequestration, biological nitrification inhibition, soil conservation, and adaptation to drought and low fertility soils. However, the use of Brachiaria grass for fodder production in Africa has been little explored largely due to lack of cultivars suitable to different production environments. The exploration and use of natural diversity is fundamental for an efficient Brachiaria breeding program. We analysed genetic diversity and population structure of 112 Ethiopian Brachiaria brizantha accessions using 23 microsatellite markers. A total of 459 alleles were detected with an average polymorphic information content of 0.75 suggesting high discriminating ability of these markers. The molecular variance analysis showed a high contribution (86%) of within-cluster differences to the total variation. Three allelic pools revealed by STRUCTURE analysis in 112 accessions were in agreement with the clustering patterns seen in neighbor-joining tree and principal coordinates analyses. A core collection of 39 B. brizantha accessions was constituted. This study concludes a high genetic diversity of Ethiopian B. brizantha accessions and their importance in Brachiaria breeding programs.

Twenty-three simple sequence repeat (SSR) primers initially developed for B. ruziziensis with proven transferability to other Brachiaria species were used (Silva et al. 2013; Supplementary Table 2). Primer optimisation, gel electrophoresis, multiplex PCR and capillary electrophoresis were performed as described previously (Kuwi et al. 2018). Allele calling and sizing were performed manually using GeneMapper 4.1 software (Applied Biosystems, Foster City, CA, USA). Due to the polyploid nature of B. brizantha, SSRs were scored as dominant markers (Jungmann et al. 2010;Vigna et al. 2011). Both allelic and binary data were used to assess genetic diversity.
The Bayesian model-based clustering algorithm was implemented in STRUCTURE 2.3.4 software to infer population structure (Pritchard et al. 2000). To estimate the posterior probabilities (qK) a 150 000 burn-in period was used, followed by 300 000 iterations using a model allowing for admixture and correlated allele frequencies with no a priori location and population information. A batch job with values of K ranging from 1 to 10 was set, with 20 independent runs for each K setting and other parameters at default values. The ∆K was calculated for each value of K using Structure Harvester (Evanno et al. 2005;Earl and vonHoldt 2011). Accessions were included in a cluster when a Q value for any cluster was ≥0.60. Bar plots were generated with average results of runs for the most probable K value using Distruct 1.1 (Rosenberg 2003).
Pairwise genetic dissimilarity matrices were calculated using Dice's similarity coefficient, and a dendrogram was constructed (for ecotypes and six improved cultivars) using the unweighted neighbour-joining (UNJ) method. The bootstrap calculation was performed based on 1 000 replications in DARwin 6.0 (Perrier and Jacquemoud-Collet 2006). Roger's genetic distance was used to construct the pairwise genetic distance between each pair of accessions. Principal coordinate analysis (PCoA) and Analysis of Molecular Variance (AMOVA) were performed using GenAlEx 6.5 (Peakall and Smouse 2012). A genetic core collection was assembled using the Min SD subset algorithm procedure in DARwin 6.0 (Perrier and Jacquemoud-Collet 2006).
Twenty-three SSR markers detected 459 alleles with an average of 4.9 alleles per locus (Supplementary Table 2). The polymorphic information content (PIC) ranged from 0.32 to 0.94 and Br0130, Br0149 and Br0235 were the three most polymorphic loci with the highest PIC values. The STRUCTURE analysis combined with Evanno ∆K statistics revealed 3 distinct gene pools (K = 3) and minor peaks at K = 5. The clusters 1, 2 and 3 were represented by 26, 39 and 47 accessions, respectively ( Figure 1). Five accessions from cluster 2 and four accessions from cluster 3 were an admixture.
The relationships among accessions were visualised in the UNJ tree ( Figure 2). To compare the grouping of accessions in the UNJ and STRUCTURE analyses, the branches of the tree were coloured according to STRUCTURE simulations for preset ∆K = 3 (clusters 1-3). Accessions (except a few) from cluster 1 grouped together in the UNJ tree. Similarly, most accessions from cluster 2 grouped together in the UNJ tree. Only 17 accessions from cluster 3 were grouped together in the UNJ tree, whereas the remainder were placed in different groups. All six improved cultivars grouped together in the UNJ tree. The genetic distance among most pairs of individuals clearly showed a high level of variation, which ranged from 0.6 to 0.90 (Supplementary Figure 1).
The PCoA well supported the UNJ tree and STUCTURE analysis. The first two coordinates explained 24% of the total variation and grouped 112 B. brizantha accessions in three major clusters (Figure 3). The accessions that formed cluster 1 in the STRUCTURE analysis (green) were also observed in a single cluster in the PCoA. Similarly, most cluster 2 accessions in the STRUCTURE analysis (red) * * * * * * * * * 14.32% Admix Cluster I Cluster II Cluster III formed a group in the PCoA. However, cluster 3 accessions (blue) were somewhat dispersed.
A hierarchical partitioning of genetic variation showed a significant (p < 0.001) variation among clusters. The analysis indicated that 86% of the total variation was contributed by within-cluster differences (Supplementary Table 3 Table 4).
The microsatellite markers used in this study were highly informative and well-differentiated Ethiopian B. brizantha accessions. The average PIC value for these markers (0.75) was higher than that in most previous studies (Jungmann et al. 2009a;Vigna et al. 2011;Silva et al. 2013;Ondabu et al. 2017;Kuwi et al. 2018). Similarly, the average number of alleles per locus (19.90) was higher than that in previous studies (7.70 to 16.96) (Jungmann et al. 2009b;Vigna et al. 2011;Pessoa-Filho et al. 2015;Kuwi et al. 2018). The high number of alleles in this study was attributed to multiple factors, including diverse geographical origins of accessions, and the use of markers with high PIC values and transferability to other Brachiaria species.
This study reports the high contribution of within-cluster differences (86%) to total variation in Ethiopian B. brizantha accessions. This result was in agreement with other studies on B. brizantha and B. ruziziensis (Vigna et al. 2011;Pessoa-Filho et al. 2015), but was higher than that in a study on B. humidocola (Jungmann et al. 2010). Such variable results are likely due to polyploidy, apomictic reproduction and occurrence of natural hybridisation in Brachiaria species (do Valle and Savidan 1996;Jungmann et al. 2010).
The STRUCTURE analysis grouped 112 B. brizantha accessions into three clusters suggesting scope for crossing individuals from different clusters to explore heterosis (Vigna et al. 2011). The clustering of the accessions was independent of their geographical origins as reported in other studies on B. brizantha and B. humidocola accessions (Jungmann et al. 2010;Vigna et al. 2011). Interestingly, some accessions from cluster 3 of the STRUCTURE bar plot (Figure 1) were dispersed in the UNJ and PCoA analyses which highlighted their genetic differences. Polyploidy, apomictic reproduction and high morphological variation in B. brizantha could be causes for such genetic dissimilarity (do Valle and Savidan 1996;Renvoize et al. 1996;Vigna et al. 2011).
The Ethiopian B. brizantha accessions analysed in this study showed higher genetic diversity than the six improved cultivars representing four species (B. brizantha, B. decumbens, B. humidicola and B. ruziziensis; Figure 2). Similar results were described for Kenyan ecotypes and Tanzania accessions (Ondabu et al. 2017;Kuwi et al. 2018). The presence of three allelic pools and high genetic diversity among Ethiopian B. brizantha accessions determines the importance of this germplasm in a Brachiaria breeding program. The core collection of 39 accessions constituted in this study will be a valuable resource for Brachiaria breeding and conservation programs.