Edinburgh Research Explorer Genetic Diversity and Population Structure of Brachiaria (syn. Urochloa) Ecotypes from Uganda

: Brachiaria (syn. Urochloa ) grass is an important tropical forage of African origin that supports millions of livestock and wildlife in the tropics. Overgrazing, conversion of grasslands for crop production and non-agricultural uses, and the introduction of improved forages have threatened the natural diversity of Brachiaria grass in Uganda. This study established a national collection of Brachiaria ecotypes in Uganda and analyzed them for genetic diversity and population structure using 24 simple sequence repeats (SSR) markers. These markers had a high discriminating ability with an average polymorphism information content (PIC) of 0.89 and detected 584 alleles in 99 ecotypes. Analysis of molecular variance revealed a high within populations variance (98%) indicating a high gene exchange or low genetic di ﬀ erentiation (PhiPT = 00.016) among the ecotype populations. The Bayesian model based clustering algorithm showed three allelic pools in Ugandan ecotypes. The principal component analysis (PCA) of ecotypes, and Neighbor-joining (NJ) tree of ecotypes and six commercial cultivars showed three main groups with variable membership coe ﬃ cients. About 95% of ecotype pairs had Rogers’ genetic distance above 0.75, suggesting most of them were distantly related. This study conﬁrms the high value of these ecotypes in Brachiaria grass conservation and improvement programs in Uganda and elsewhere. SSR Ecotypes rich


Introduction
The genus Brachiaria (Trin.) Griseb. (syn. Urochloa) belongs to the tribe Paniceae in the subfamily Panicoideae of the family Poaceae [1]. It consists of about 100 species distributed throughout the tropics especially in Africa [2]. Seven perennial species of African origins-Brachiaria arrecta (Hack. ex. Th. Dur and Schinz) Stent, Brachiaria brizantha (A. Rich.) Stapf., Brachiaria decumbens Stapf, Brachiaria dictyoneura ( Fig. and De Not.) Stapf, Brachiaria humidicola (Rendle) Schweick, Brachiaria mutica (Forssk.) Stapf and Brachiaria ruziziensis Germain and Evrard-have been used as fodder plants [3]. All Brachiaria species with known forage values occur naturally in eastern Africa which represents the center of diversity of the genus [3]. These agriculturally important Brachiaria species were introduced to other parts of the The origins and details of ecotype collection sites are presented in Table S1. Two-week old leaves from 2-month old regrowth were collected separately for each ecotype for total genomic DNA extraction. Leaves were put into Ziploc bags and transported in ice chests to the Biosciences eastern and central Africa-International Livestock Research Institute (BecA-IRLI) Hub laboratory in Nairobi, Kenya. Young leaf samples of six commercial cultivars: B. brizantha cv. MG4, B. brizantha cv. Piatã, B. decumbens cv. Basilisk, B. humidicola cv. Humidicola, B. humidicola cv. Llanero, and Brachiaria hybrid Mulato-II were also included in the study. All leaf samples were freeze-dried and stored at −80 °C prior to DNA extraction.

DNA Extraction
Total genomic DNA was extracted from freeze-dried and ground leaf samples of 99 ecotypes and six commercial cultivars using the Quick-DNA Plant/Seed Miniprep Kit (Zymo Research, Irvine, The origins and details of ecotype collection sites are presented in Table S1. Two-week old leaves from 2-month old regrowth were collected separately for each ecotype for total genomic DNA extraction. Leaves were put into Ziploc bags and transported in ice chests to the Biosciences eastern and central Africa-International Livestock Research Institute (BecA-IRLI) Hub laboratory in Nairobi, Kenya. Young leaf samples of six commercial cultivars: B. brizantha cv. MG4, B. brizantha cv. Piatã, B. decumbens cv. Basilisk, B. humidicola cv. Humidicola, B. humidicola cv. Llanero, and Brachiaria hybrid Mulato-II were also included in the study. All leaf samples were freeze-dried and stored at −80 • C prior to DNA extraction.

DNA Extraction
Total genomic DNA was extracted from freeze-dried and ground leaf samples of 99 ecotypes and six commercial cultivars using the Quick-DNA Plant/Seed Miniprep Kit (Zymo Research, Irvine, CA, USA) following the manufacturer's instructions. DNA concentration and purity were determined using a NanoDrop2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Samples were normalized to a concentration of 20 ng/µL, then run on a 1% agarose-0.5xTBE gel stained with 0.25× GelRed at 100 volts for 45 min and visualized under UV light (UVP Bio-Imaging Systems, Upland, CA, USA) to assess the integrity of DNA.

PCR Amplification and Capillary Electrophoresis
A total of 24 fluorescent-labeled SSR markers were used for genotyping (Table 1). These markers were initially developed for B. ruziziensis, had high polymorphism information content (PIC) values, and confirmed for transferability to other Brachiaria species [21]. The forward primers were labeled with the fluorescent dyes-6-FAM, PET, NED, and VIC. Markers used in this study consisted of dinucleotide and trinucleotide repeat motifs. Gradient PCR was carried out for each primer sets and the annealing temperature that gave clear bands was identified for each primer set. Multiplex panels were designed based on annealing temperature and dye label. Each of the eight selected panels consisted of 1-4 sets of primers.
PCR reaction was performed on a total volume of 10 µL using AccuPower ® PCR PreMix, negative dye (Bioneer, Daejeon, Korea), 40 ng template DNA, 0.09 µM each of forward and reverse primer, additional 0.5 µM MgCl 2, and 7.2 µL triple distilled water. The PCR reaction was performed in a GeneAmp PCR system 9700 thermocycler (Applied Biosystem, Foster City, CA, USA) using the following program: initial denaturation at 95 • C for 3 min, followed by 30 cycles of 30 s at 94 • C, 1 min at optimized annealing temperature for each primer pair, and 2 min at 72 • C for extension. This was followed by 20 min final extension at 72 • C and hold at 15 • C. The PCR products were run on a 2% agarose gel in 0.5x TBE buffer stained with 0.25x GelRed at 6.7 V/cm for 45 min and visualized under UV light. The size of the band was estimated using 1 kb plus ladder.

Allelic Scoring
Allele calling and sizing were performed manually using GeneMapper Software v4.1 (Applied Biosystems, Foster City, CA, USA). The SSR fragments were analyzed following a dominant scoring scheme as the information on ploidy levels for Ugandan ecotypes was not available [29][30][31]. Well defined clear peaks were binned according to expected product size and data was exported to Microsoft Excel for analysis. The allele calls were converted to binary data (0 = absence and 1 = presence of alleles) using ALS-Binary Software [32] for subsequent analyses. Both allelic and binary data were used in the genetic diversity analysis. The SSR genotyping data for the commercial cultivars was used in the construction of neighbor-joining tree only.

Population Genetic Analyses
The model-based clustering approach implemented in the software package STRUCTURE version 2.3.4 was used to analyze the population structure [33]. To estimate the posterior probabilities (qK), a 100,000 burn-in period was used, followed by 100,000 iterations using a model allowing for admixture and correlated allele frequencies with no prior location or population information. At least 10 independent runs of STRUCTURE were performed by setting K from 1-10 with 15 replicates for each K. The Delta K was calculated for each value of K using the web-based Structure Harvester [34,35]. A line was assigned to a given cluster when the proportion of its genome in the cluster (qK) was higher than a threshold value of 50%.
Matrices of Roger's genetic distance [36], gene diversity, PIC value, and allele frequency for each locus were calculated between each pair of lines using PowerMarker v3.2.5 [37]. The Pearson correlation coefficient matrix was computed to examine what type of linear relationships of rainfall, altitude, sample size, and the allele frequencies of regional ecotype populations had using XLSTAT software [38]. The genetic distance matrices for ecotypes and six commercial Brachiaria cultivars were used for constructing Neighbor-joining trees using PowerMarker and the resulting trees were visualized using MEGA version 5.0 [39]. Analysis of Molecular Variance (AMOVA) [40,41] was used to partition the variation among and within group (population) components in GenAlEx version 6.5 [42] which enabled the estimation of standardized PhiPT and the allelic patterns across different populations [43]. Significance levels for variance component estimates were computed using 9999 permutations. Principal component analysis (PCA) was performed to visualize similarities and variations among Brachiaria ecotypes from Uganda in DARwin software version 6.0.15 [44].

Microsatellite Diversity and Analysis of Molecular Variance
Twenty-four SSR markers were used for genotyping 99 Ugandan Brachiaria grass ecotypes (Table 1) and six commercial cultivars. These markers detected 584 alleles of different sizes ranging from 111 bp (Br0028) to 358 bp (Br0214) in 99 Ugandan ecotypes ( Table 1). The PIC values for these markers ranged between 0.60 (Br0118) and 0.96 (Br0028) with 0.89 average. The analysis of molecular variance (AMOVA) revealed a high contribution of within the population differences (98%) to the total genetic variations, whereas the rest (2%) was contributed by populations' differences. The genetic differentiation among the ecotype populations (PhiPT) was low (0.016) ( Table 2).

Allelic Diversity in the Regional Populations
The allelic diversity in five regional populations of Brachiaria ecotypes is presented in Figure 2. The populations varied for mean numbers of different alleles (Na) that ranged from 2.92 (western (WST)) to 4.63 (central (CTR)). Similarly, differences were observed among the populations for the mean number of private alleles (Np) ranging from 0.674 (south dryland (SDL)) to 1.542 (CTR), as well as for the mean number of effective alleles (Ne) which ranged from 2.39 (southwestern (SWT)) to 3.33 (northern (NTN)). We detected the highest mean genetic diversity in NTN population whereas the WST population had the least mean genetic diversity (0.74). The expected heterozygosity (He) of the populations ranged between 0.36 (WST) and 0.56 (NTN). 111 bp (Br0028) to 358 bp (Br0214) in 99 Ugandan ecotypes ( Table 1). The PIC values for these markers ranged between 0.60 (Br0118) and 0.96 (Br0028) with 0.89 average. The analysis of molecular variance (AMOVA) revealed a high contribution of within the population differences (98%) to the total genetic variations, whereas the rest (2%) was contributed by populations' differences. The genetic differentiation among the ecotype populations (PhiPT) was low (0.016) ( Table 2).

Allelic Diversity in the Regional Populations
The allelic diversity in five regional populations of Brachiaria ecotypes is presented in Figure 2. The populations varied for mean numbers of different alleles (Na) that ranged from 2.92 (western (WST)) to 4.63 (central (CTR)). Similarly, differences were observed among the populations for the mean number of private alleles (Np) ranging from 0.674 (south dryland (SDL)) to 1.542 (CTR), as well as for the mean number of effective alleles (Ne) which ranged from 2.39 (southwestern (SWT)) to 3.33 (northern (NTN)). We detected the highest mean genetic diversity in NTN population whereas the WST population had the least mean genetic diversity (0.74). The expected heterozygosity (He) of the populations ranged between 0.36 (WST) and 0.56 (NTN). The Pearson correlation coefficient analysis showed a strongly positive linear relationship between the number of ecotypes in the regional populations and the number of different alleles (r = 0.972; p = 0.006). Similar holds between the sample size and the number of private alleles (r = 0.920; p = 0.027). However, the linear relationships of altitude and rainfall with both Na and Np were negative and non-significant.

Similarity-Based Analysis
Neighbor-joining (NJ) dendrogram illustrates the genetic relationship among ecotypes as well as between ecotypes and commercial Brachiaria cultivars (Figure 3). The NJ tree constructed based on the genetic distances showed 99 ecotypes and six commercial cultivars in three major groups. The Pearson correlation coefficient analysis showed a strongly positive linear relationship between the number of ecotypes in the regional populations and the number of different alleles (r = 0.972; p = 0.006). Similar holds between the sample size and the number of private alleles (r = 0.920; p = 0.027). However, the linear relationships of altitude and rainfall with both Na and Np were negative and non-significant.

Similarity-Based Analysis
Neighbor-joining (NJ) dendrogram illustrates the genetic relationship among ecotypes as well as between ecotypes and commercial Brachiaria cultivars (Figure 3). The NJ tree constructed based on the genetic distances showed 99 ecotypes and six commercial cultivars in three major groups.  The pairwise Rogers genetic distance for 99 Brachiaria ecotypes showed a wide range of genetic differences (≤0.1500 to over ≥0.9001) among the ecotype pairs ( Figure 4). About 67.2% of the ecotype pairs had a genetic distance of over 0.9001, 27.6% pairs had a genetic distance between 0.7501 and 0.9000, while 5.2% pairs had a genetic distance of ≤0.7500. The pairwise Rogers genetic distance for 99 Brachiaria ecotypes showed a wide range of genetic differences (≤0.1500 to over ≥0.9001) among the ecotype pairs ( Figure 4). About 67.2% of the ecotype pairs had a genetic distance of over 0.9001, 27.6% pairs had a genetic distance between 0.7501 and 0.9000, while 5.2% pairs had a genetic distance of ≤0.7500.
(47 K), B. brizantha cv. MG4 (48 K), Brachiaria hybrid Mulato II (49 K), B. humidicola cv. Llanero (50 K), and B. decumbens cv. Basilisk (52 K)). Group 3 was further divided into five subgroups with the commercial cultivars present in two subgroups only. The pairwise Rogers genetic distance for 99 Brachiaria ecotypes showed a wide range of genetic differences (≤0.1500 to over ≥0.9001) among the ecotype pairs ( Figure 4). About 67.2% of the ecotype pairs had a genetic distance of over 0.9001, 27.6% pairs had a genetic distance between 0.7501 and 0.9000, while 5.2% pairs had a genetic distance of ≤0.7500.

Principal Component Analysis
The genetic relationships among ecotypes were visualized through principal component analysis. Principal component analysis based on allele frequencies generated from 24 SSR markers detected three-major groupings of 99 ecotypes ( Figure 5). The percentage variation explained by PC1 and PC2 were 40.6% and 18.2%, respectively.

Principal Component Analysis
The genetic relationships among ecotypes were visualized through principal component analysis. Principal component analysis based on allele frequencies generated from 24 SSR markers detected three-major groupings of 99 ecotypes ( Figure 5). The percentage variation explained by PC1 and PC2 were 40.6% and 18.2%, respectively.

Structure Analysis
The Bayesian model based clustering algorithm implemented in STRUCTURE software confirmed three distinct clusters (ΔK = 3) among 99 ecotypes ( Figure S1). These are indicated in different colors-Cluster I (red), Cluster II (green), and Cluster III (blue). These clusters consisted of

Structure Analysis
The Bayesian model based clustering algorithm implemented in STRUCTURE software confirmed three distinct clusters (∆K = 3) among 99 ecotypes ( Figure S1). These are indicated in different colors-Cluster I (red), Cluster II (green), and Cluster III (blue). These clusters consisted of pure lines and some admixture individuals with two or three gene pools (Figure 6a). For ∆K = 4, four allelic pools were identified with four different colors as red, green, blue, and yellow; while ∆K = 6, six allelic pools were identified with six colors as red, green, blue, yellow, purple, and pink. These clusters had pure lines as well as some admixture individuals (Figure 6b,c). As reported in previous study the clustering of ecotypes was independent of their geographical origin [29]. For ∆K = 3, most of the ecotypes from the North, showed the greater probability of ancestral membership (80.5%) for cluster I and II (Table 3). pure lines and some admixture individuals with two or three gene pools (Figure 6a). For ΔK = 4, four allelic pools were identified with four different colors as red, green, blue, and yellow; while ΔK = 6, six allelic pools were identified with six colors as red, green, blue, yellow, purple, and pink. These clusters had pure lines as well as some admixture individuals (Figure 6b,c). As reported in previous study the clustering of ecotypes was independent of their geographical origin [29]. For ΔK = 3, most of the ecotypes from the North, showed the greater probability of ancestral membership (80.5%) for cluster I and II (Table 3).

Discussion
The genus Brachiaria exhibits a great diversity between and within species for genetic composition, morphology, growth habits, adaptation, and agricultural utility. The understanding of the diversity in natural populations is important for genetic conservation as well as for the improvement of a plant species for desirable traits including in Brachiaria grass. Of the 100 documented Brachiaria species, 33 are represented in the various gene banks, and only seven perennial species of African origin have been explored for forage production [3]. For the past few years, the popularity of improved Brachiaria grass cultivars for pasture production has been increased among livestock farmers in Africa. However, all improved Brachiaria cultivars that are grown in Africa were developed for alien environments in Australia and South America, suggesting a lack of improved Brachiaria cultivars for African environments. This study reports the establishment of the first national collection of Brachiaria ecotypes in Uganda, their genetic diversity profiles, and population structure based on SSR markers to facilitate the Brachiaria improvement programs in Uganda. The SSR markers have multiple uses including cultivar identification, genetic diversity studies, and genome mapping [45]. For example, SSR markers have been used to assess genetic diversity in various plant species such as pearl millet, rice, sweet cassava, and Brachiaria grass [26,[29][30][31]45,46].
The average polymorphism information content of SSR markers used in this study was 0.89 conferring them as highly informative and capable to differentiate well among the Ugandan Brachiaria ecotypes ( Table 1). The mean PIC values (0.89) deduced for markers in this study were comparable to studies of Silva et al. for the top 30 most informative markers [21], Kuwi et al. [30], and Pessoa-Filho et al. [47] although it was higher than those reported in other studies [31,48,49]. Interestingly, the number of SSR alleles detected in this study (n = 584) was higher than those reported by Vigna et al. [49], Jungmann et al. [1], and Pessoa-Filho et al. [47], but was lower than in the study of Trivino et al. [50]. Differences in PIC values and SSR alleles among these studies could be attributed to several factors such as differences in number, genetic background, and genetic complexity of Brachiaria genotypes; variation in numbers and types of markers used in the analysis, and the difference among studies in allele scoring system and combinations thereof. A relatively higher number of alleles detected in this study may have been associated with geographical position of Uganda in the region where Africa's seven distinct biogeographic regions or phytochoria converge [51] and the region also represents the center of diversity of the genus Brachiaria [3]. Besides the robustness of markers in detecting a high number of alleles, there were some challenges in alleles scoring, especially in differentiating stutter and true peaks as reported by other authors [52,53].
The Brachiaria ecotypes analyzed in this study were collected from central, northern, south dryland, southwestern, and western regions that represent eight of eleven agroecological zones in Uganda [25]. The majority of these ecotypes (n = 95) were collected from sites with an altitude range of 1080-1521 m above sea level and an annual rainfall of between 1000 and 1500 mm. Despite differences among the collection sites for altitude and amount of precipitation, differences in the allelic patterns (e.g., number of effective alleles and private alleles) in the regional population were mainly influenced by sample size (data not shown) as reported in previous studies [30,54]. The detection of private alleles in all five regional populations suggests them as a valuable sources of genetic variation for breeding programs [55] targeting adaptation and other traits.
Analysis of molecular variance showed a high contribution of within-population difference to the total variation inferring high genetic diversity among the ecotypes. This result is substantiated by a low level of genetic variations among the populations, a high pair-wise Roger's genetic distance of most ecotype pairs, and a fair representation of ecotype from all regions in structure analysis clusters particularly in ∆K = 3. Such differences among the ecotypes is anticipated due to the apomictic mode of reproduction in favor of maternal genotype regardless of the level of heterozygosity [56] and the polyploidy nature of Brachiaria species that's often associated with meiotic anomalies leading to reduced pollen fertility [57]. Many Brachiaria species are known to have variable ploidy levels, for example, presence of tetra-, penta-and hexaploidy in B. brizantha population [58]. Polyploidy benefits plants from heterosis, gene redundancy, and loss of self-incompatibility and gain of asexual reproduction [59]. Our results were similar to studies of Pessoa-Filho et al. [47] and Vigna et al. [49] on Brachiaira ruziziensis and Brachiaria brizantha, respectively. The partitioning of molecular variations for the Ugandan ecotype population was similar to those reported in other studies [30,31,60].
The STRUCTURE analysis showed the presence of three distinct gene pools in Ugandan Brachiaria ecotypes. The three gene pools detected in this study agrees to previous studies in Tanzania, Ethiopia, and Brazil [30,31,49]. In agreement with STRUCTURE analysis, the NJ tree showed Ugandan ecotypes and six commercial cultivars in three distinct groups, but the membership coefficient to each group differed between two analyses. We observed a high degree of relatedness between ecotypes and the commercial cultivars. Ecotypes in groups 1, and subgroups of groups 2 and 3 that clustered exclusive of improved cultivars may require further analysis to know where they belong, they could possess unique traits of agricultural importance. Such grouping and sub-grouping of ecotypes and improved cultivars in NJ trees signifies a high level of genetic diversity in Ugandan ecotypes compared to six improved Brachiaria cultivars that belong to three species (B. brizantha, B. decumbens, and B. humidicola) and a hybrid of B. brizantha × B. decumbens × B. ruziziensis. These observations were anticipated as since many Brachiaria species occur naturally in eastern Africa and the region represents a center of diversity of the genus [3]. We also guess that this collection of 99 ecotypes may have representation of several Brachiaria species.
Most improved Brachiaria cultivars that are in use for pasture production were derived from the direct selection of naturally occurring genotypes from the East Africa [9]. Therefore, the evaluations of these Ugandan Brachiaria ecotypes for major agricultural traits, e.g., biomass yield, animal nutrition, livestock productivity, pests and disease resistance, and adaptation to drought at different agroclimatic zones is necessary to develop locally adapted and improved cultivars for commercial cultivation. The SSR markers revealed a high genetic diversity in Ugandan Brachiaria ecotypes and their high value in Brachiaria improvement programs. Crosses of distantly related ecotypes could be a good strategy to broaden the genetic base. High density genotyping and association mapping would help to shorten the time necessary for completing a breeding cycle and developing new varieties. The complexity of the Brachiaria genome, limited understanding of reproductive biology, and morphological agility within and between the species have limited the pace of Brachiaria breeding. Therefore, there is a need to enrich the current understanding of Brachiaria biology and promote integrated use of conventional and molecular breeding methods for better exploitation of genetic resources from this collection as well as those available elsewhere.

Conclusions
Through this study, we successfully established the first nation-wide collection of Brachiaria ecotypes in Uganda covering five regions representing eight different agroecological zones in the country. Ecotypes are maintained in the vegetative field gene bank at Tororo, Uganda in the guardianship of NaLIRRI, and these materials can be accessed by other researchers following the Ugandan government's guidelines for accessing genetic resources and benefit sharing. This study documented genetic diversity and population structure of these ecotypes using SSR markers. Ecotypes were rich in allelic diversity, genetically diverse and they had three distinct gene pools. High contribution of within ecotypes genetic difference to total diversity observed in these ecotypes was consistent with the reproductive mode, dispersal mechanism, and genetic attributes of the Brachiaria species. The genetic materials (ecotypes) and genetic information produced in this study will form a basis for Brachiaria grass conservation and improvement programs targeting agricultural and environmental applications in Uganda and beyond.