About Us | Help Videos | Contact Us | Subscriptions
 

Crop Science - Article

 

 

This article in CS


  1. unlockOPEN ACCESS
     
    Received: Jan 21, 2017
    Accepted: Apr 03, 2017
    Published: July 13, 2017


    * Corresponding author(s): mlworthi@uark.edu
 View
 Download
 Alerts
 Permissions
Request Permissions
 Share

doi:10.2135/cropsci2017.01.0045

Genetic Diversity and Population Structure of Brachiaria Species and Breeding Populations

  1. Narda Jimena Triviñoa,
  2. Juan Guillermo Pereza,
  3. Maria Eugenia Recioa,
  4. Masumi Ebinab,
  5. Naoki Yamanakac,
  6. Shin-ichi Tsurutabd,
  7. Manabu Ishitania and
  8. Margaret Worthington *ae
  1. a International Center for Tropical Agriculture (CIAT), Apartado Aereo 6713, Palmira, Colombia, 763537
    b National Agriculture and Food Research Organization (NARO), Institute of Livestock and Grassland Science, Nasushiobara, Tochigi 392-2793, Japan
    c Japan International Research Center for Agricultural Sciences (JIRCAS), 1-1 Ohwashi, Tsukuba, Ibaraki 305-8686, Japan
    d current address, Japan International Research Center for Agricultural Sciences (JIRCAS), Tropical Agriculture Research Front, 1091-1 Maezato-Kawarabaru, Ishigaki, Okinawa 907-0002, Japan
    e current address, Univ. of Arkansas, Dep. of Horticulture, 306 Plant Sciences Bldg., Fayetteville, AR 72701

Abstract

Several apomictic Brachiaria (Trin.) Griseb. (syn. Urochloa P. Beauv.) species are commercially important tropical forage grasses, but little is known about the interspecific diversity and population structure within this genus. Previously published genus-level Brachiaria phylogenies were conducted with few genotypes and contradicted well-established morphological evidence and proven interspecific fertility in the B. brizantha (Hochst. ex A. Rich.) Stapf., B. decumbens Stapf., and B. ruziziensis (R. Germ. & C.M. Evrard) agamic complex. In this study, we characterized the genetic diversity and population structure of 261 genotypes from 14 Brachiaria species and a Panicum maximum Jacq. outgroup using 39 simple sequence repeat primers with 701 polymorphic bands. The genotypes included in the panel included germplasm accessions, commercial cultivars, and sexually reproducing breeding populations. Results of STRUCTURE, neighbor joining, unweighted pair group method with arithmetic mean, and multiple correspondence analyses confirmed the relatedness of the important commercial species B. brizantha, B. decumbens, and B. ruziziensis. Brachiaria decumbens was most closely related to B. ruziziensis, and the diploid sexual and tetraploid apomict B. decumbens accessions formed into two related but distinct groups. The close relationship between B. humidicola (Rendle) Schweick and B. dictyoneura (Figari. and De Not) Stapf. and the unique genetic makeup of the lone sexually reproducing B. humidicola accession were also corroborated by these results. Our findings largely supported morphology-based taxonomic groupings in Brachiaria and indicated that genus-level phylogenies are made more robust by the inclusion of many polymorphic markers and multiple genotypes from each species.


Abbreviations

    AMOVA, analysis of molecular variance; CIAT, International Center for Tropical Agriculture; EMBRAPA, Brazilian Agricultural Research Corporation; ILRI, International Livestock Research Institute; ISSR, inter simple sequence repeat; ITS, internal transcribed spacer; JIRCAS, Japan International Research Center for Agricultural Sciences; MCA, multiple correspondence analysis; NJ, neighbor joining; PC, principal component; PCR, polymerase chain reaction; SSR, simple sequence repeat; RAPD, random amplified polymorphic DNA; UPGMA, unweighted pair group method with arithmetic mean

Grasses from the genus Brachiaria have played an important role in Latin American livestock production, especially in Brazil where there are extensive grazing lands and demand for productive, high-quality, and well-adapted forages for cattle nutrition (Jank et al., 2014). The commercial Brachiaria species [B. brizantha (Hochst. ex A. Rich.), B. decumbens Stapf., B. humidicola (Rendle) Schweick, and B. ruziziensis (R. Germ. & C.M. Evrard)] are native to East Africa and were introduced to tropical Latin America as natural germplasm accessions during the mid-20th century (Renvoize et al., 1996). Most important Brachiaria cultivars have been developed from direct selection of naturally occurring genotypes collected in East Africa (Jank et al., 2014). The International Center for Tropical Agriculture (CIAT) forage genebank maintains a collection 601 Brachiaria accessions, most of them collected in Africa between 1984 and 1985 in a mission cosponsored by CIAT and the International Livestock Research Institute (ILRI). Most of these accessions are held in duplicate in germplasm collections at ILRI and the Brazilian Agricultural Research Corporation (EMBRAPA) (Keller-Grein et al., 1996).

Brachiaria breeding began relatively recently when CIAT and EMBRAPA achieved compatibility between species with different ploidy levels in the late 1980s. The commercial species B. brizantha, B. decumbens, and B. ruziziensis form an agamic complex, meaning that some species are exclusively diploid and sexual reproducing, while other closely related species exist primarily as polyploid apomicts (Valle and Savidan, 1996). Researchers used colchicine doubling to create a sexual tetraploid B. ruziziensis genotype (Swenne et al., 1981) that could be used as a female parent in crosses with tetraploid apomictic accessions of B. decumbens and B. brizantha. The progeny of these original crosses were used to form fully sexual synthetic recurrent selection populations that formed the basis of breeding programs at CIAT and EMBRAPA (Miles et al., 2006). Recently, a third interspecific Brachiaria breeding program was established in Japan at the National Institute of Livestock and Grassland Science (NILGS) and the Japan International Research Center for Agricultural Sciences (JIRCAS) (Tsuruta et al., 2015) after tetraploid induction in B. ruziziensis (Ishigaki et al., 2009). Separate intraspecific B. humidicola breeding programs were also established at CIAT and EMBRAPA in the mid-2000s after the discovery of a naturally occurring sexual polyploid germplasm accession that could be crossed with other apomictic polyploid B. humidicola pollen donors (Jungmann et al., 2010).

Brachiaria is classified as a monophyletic complex within the genus Urochloa, both belonging to the Poaceae (Torres González and Morton, 2005). This BrachiariaUrochloa complex is a sister group of the genus Panicum L. (Giussani et al., 2001). The first taxonomic review of Brachiaria classified the genus into nine groups on the bases of inflorescence and panicle morphology (Renvoize et al., 1996). The commercial species were assigned to two principal taxonomic groups. The agamic complex species (B. ruziziensis, B. decumbens, and B. brizantha) were assigned to taxonomic Group 5, whereas B. humidicola and B. dictyoneura were located in Group 6 (Renvoize et al., 1996). Other methodologies such as flow cytometry (Penteado et al., 2000) and chromosome counting (Mendes-Bonato et al., 2006) have been also been used to characterize Brachiaria germplasm collections in support of breeding programs.

Molecular characterization of Brachiaria germplasm collections began recently with analysis of internal transcribed spacer (ITS) (Torres González and Morton, 2005), random amplified polymorphic DNA (RAPD) (Ambiel et al., 2008), simple sequence repeat (SSR) (Jungmann et al., 2010; Vigna et al., 2011; Ferreira et al., 2016), and inter simple sequence repeat (ISSR) (Nitthaisong et al., 2016) markers. Detailed intraspecific phylogenies have been conducted of the EMBRAPA B. brizantha (Vigna et al., 2011), B. decumbens (Ferreira et al., 2016), and B. humidicola (Jungmann et al., 2010) collections. The B. brizantha (Vigna et al., 2011) and B. humidicola (Jungmann et al., 2010) studies compared the genetic structure of the germplasm with the geographical distance between collection sites, ploidy levels, and reproductive modes of the accessions and found no significant correlations (Jungmann et al., 2010; Vigna et al., 2011). Possible explanations for these findings include the long distances traveled by seeds consumed by cattle or native savanna species and the low frequency of sexual reproduction and recombination expected for facultative apomicts.

In contrast with the detailed intraspecific phylogenies available for the most important commercial species, genus-level Brachiaria phylogenies have involved fewer genotypes and molecular markers and have often contradicted previously described taxonomic patterns (Renvoize et al., 1996) and proven interspecific fertility in the agamic complex (Valle and Savidan, 1996). Torres González and Morton (2005) conducted a phylogeny study of 22 Brachiaria genotypes, each from a different species, using ITS analysis and found no relationship between morphological traits described by Renvoize et al. (1996) and their molecular topology. Brachiaria decumbens was found to be most closely related to B. eruciformis (Sm.) Griseb., B. subulifolia (Mez) Clayton and Melinis repens (Willd) Zizka, whereas B. ruziziensis and B. brizantha species were placed together in a different clade with B. dura Stapf and B. comata (Hochst. ex A. Rich.) Sosef (Torres González and Morton, 2005). More recently, Nitthaisong et al. (2016) used 441 polymorphic ISSR bands to assess genetic structure among 28 genotypes from 11 different Brachiaria species. Brachiaria ruziziensis was placed in a monophyletic group with B. dexflexa (Shumach.) H. Scholz, B. ramosa (L.) T.Q. Nguyen, B. plantaginea (Link) Hitchc., and B. xantholeuca (Hack. ex Schinz) H. Scholz, which was even more distantly related to B. brizantha and B. decumbens than the Panicum maximum (syn. Megathyrsus maximus) outgroup.

In this study, we characterized the genetic diversity and population structure of 14 Brachiaria species from the CIAT tropical forages genebank and breeding programs. Our specific objectives were to investigate the relationship between morphology-based taxonomic assessments (Renvoize et al., 1996) and genetic variation within Brachiaria using a much larger number of genotypes and polymorphic markers than previous interspecific phylogenies and to examine the relationships between interspecific hybrid cultivars and breeding populations with germplasm accessions from the agamic complex species (B. brizantha, B. decumbens, and B. ruziziensis).


MATERIALS AND METHODS

Plant Materials

The CIAT Brachiaria collection is composed of 601 accessions and has a high representation of the commercial species B. decumbens (65), B. humidicola (66), B. ruziziensis (41), and especially B. brizantha (309). The remaining 120 accessions are made up of 10 noncommercial African Brachiaria species. Approximately one-third of the germplasm accessions for each of the 14 species in the CIAT Brachiaria collection were selected at random for inclusion in this study. The founding parents of the CIAT interspecific Brachiaria breeding program and B. humidicola breeding program, as well as several important cultivars including B. brizantha cv. Marandu (CIAT 6297), B. decumbens cv. Basilisk (CIAT 606), and B. humidicola cv. Tully (CIAT 679), Llanero (CIAT 6133), and Tupi (CIAT 26149), were also added to the selected group of accessions. A total of 226 accessions of the 14 African Brachiaria species were chosen from the tropical forage collection of the CIAT Genetic Resources Program (Supplemental Table S1). The selected genotypes included B. arrecta (Hack. ex T. Durand & Schinz) Stent (2), B. bovonei (Chiov.) Robyns (2), B. brizantha (112), B. decumbens (12), B. dictyoneura (2), B. dura (1), B. eruciformis (1), B. humidicola (58), B. jubata (Fig. & De Not.) Stapf (9), B. nigropedata (Munro ex Ficalho & Hiern) Stapf (8), B. platynota (K. Schum.) Robyns (2), B. ruziziensis (13), B. serrata (Thunb.) Stapf (2), and B. subulifolia (2). The Brachiaria germplasm accessions included in this study can be obtained from the CIAT Genetic Resources Program (http://isa.ciat.cgiar.org/).

Four sexually reproducing synthetic autotetraploid B. ruziziensis genotypes were also included in the phylogeny panel (Supplemental Table S1). One synthetic autotetraploid genotype, BRX 44-02, was the original source of sexuality in the CIAT hybrid Brachiaria breeding program. This plant was derived by doubling the chromosome number of a diploid sexual B. ruziziensis accession collected in Burundi (Swenne et al., 1981). The other three tetraploid sexual B. ruziziensis genotypes (MOK 8-4, MOK 8-6, and MOK 8-7) were developed by Ishigaki et al. (2009) and provided by Miyazaki University.

Four apomictic hybrid cultivars from the CIAT B. ruziziensis × B. decumbens × B. brizantha interspecific breeding program (Mulato [CIAT 36061], Mulato II [CIAT 36087], Cayman [BR02/1752], and Cobra [BR02/1794]) were also included in the study, along with three interspecific apomictic selections from the JIRCAS breeding program (JIRCAS 185, JIRCAS 203, and JIRCAS 226) (Supplemental Table S1). The JIRCAS hybrids were generated by crossing a sexual, tetraploid B. ruziziensis accession (Ishigaki et al., 2009) to Mulato (CIAT 36061), the first interspecific hybrid cultivar released from the CIAT breeding program (Thaikua et al., 2015). Ten sexually reproducing genotypes from the fourth cycle of recurrent selection for specific combining ability (SX14 cycle) of the CIAT B. ruziziensis × B. decumbens × B. brizantha interspecific breeding program and 10 sexually reproducing genotypes from the first cycle of recurrent selection (BhSX12 cycle) of the B. humidicola breeding program were also included in the study. More information on the origin and methodology of the CIAT B. ruziziensis × B. decumbens × B. brizantha recurrent selection program can be found in Miles et al. (2006) and Miles (2007). The CIAT B. humidicola recurrent selection population was generated by open pollination of sexual progeny derived from crosses between the natural polyploid sexual accession (CIAT 26146) and 18 apomictic B. humidicola accessions determined by flow cytometry to have 36 chromosomes (Penteado et al., 2000).

Lastly, three apomictic P. maximum cultivars [Mombaça (CIAT 6962), Paikaji, and Umak], as well as the P. maximum × P. infestum hybrid cv. Massai, were included as an outgroup. More information about the JIRCAS P. maximum cultivars (Paikaji and Umak) is available in Ebina et al. (2007). Identifying information including species, taxonomic group (Renvoize et al., 1996), geographic coordinates of the collection site, and reproductive mode is provided for the 261 genotypes included in the study in Supplemental Table S1. Reproductive mode was assessed with the primer pair p779/p780, a diagnostic marker for apomixis in Brachiaria, following Worthington et al. (2016).

DNA Extraction and Microsatellite Genotyping

Tissue was collected from young leaves of plants maintained in greenhouse conditions at CIAT headquarters in Palmira, Colombia (1001 m asl; 3.5833° N, 76.2500° W) and in a field planting in Popayan, Colombia (1760 m asl; 2.4542° N, 76.6092° W). Leaf tissue was collected in liquid nitrogen and stored at −20°C. DNA extraction was conducted with the modified mixed alkyltrimethylammonium bromide (MATAB)-chloroform protocol following Risterucci et al. (2000). Samples were resuspended overnight at room temperature and treated with RNase enzyme at 37°C. DNA quality was determined through visualization on an agarose gel and confirmed with ITS4 primers. Afterward, DNA was quantified with a NanoDrop 1000 spectrophotometer (Thermo Scientific) and diluted to 5 ng μL−1.

Sequences containing SSRs were first identified using MASTCOMMANDER software (Faircloth, 2008) applied to Roche 454 FLX+ sequence data from the interspecific Brachiaria hybrid Mulato II (CIAT 36087). Primers were developed from 902 SSR-containing sequences using Primer3Plus (Untergasser et al., 2007). These primers were then tested for amplification in three B. humidicola accessions (CIAT 679, CIAT 16888, and CIAT 26146). A total of 39 primers with good amplification and polymorphic across species were selected to evaluate the panel of 261 genotypes described above. The primers sequences and amplification conditions for the 39 SSR markers are provided in Supplemental Table S2.

Forward primers were designed with a 5′-TGTAAACGACGGCCAGTATGC M-13 reverse sequence tail for universal fluorescent labeling following Schuelke (2000). Polymerase chain reactions (PCRs) were performed on an ABI 9700 thermocycler (Applied Biosystems) in 10-μL final volume of reaction mixture consisting of buffer (100 mM Tris-HCl pH 8.3 at 37.0—C, 10 mM KCl, 6 mM (NH4)SO4, 0.1% Triton X-100, 0.001% bovine serum albumin [BSA], and 2.5 mM MgCl2), 10 ng genomic DNA, 0.25 μM M-13 labeled forward primer, 1.0 μM reverse primer, 1.0 μM M-13 labeled dye (FAM or HEX, Applied Biosystems), 250 μM deoxynucleotides, and 0.5 μL of Prime STAR HS DNA polymerase (TaKaRa Bio). After an initial denaturation step at 98.0°C for 1 min, PCR amplifications were performed with 30 cycles of 98.0°C for 20 s, 55.0°C for 20 s, and 72.0°C for 30 s, followed by eight cycles of 98.0°C for 20 s, 50.0°C for 20 s, and 72.0°C for 30 s, and a final elongation step at 72.0°C for 10 min. Polymerase chain reaction products were then visualized on an ABI 3130xl genetic analyzer and scored using GeneMapper Software 5 (Applied Biosystems, 2012). Because Brachiaria is a largely polyploid genus, each band was scored separately for presence or absence. Only bands with strong amplification peaks and clear separation from neighboring bands were used in this study. A binary matrix was obtained and used for phylogenetic and statistical analysis. Polymorphism information content and heterozygosity were calculated using CERVUS 3.0 (Marshall et al., 1998).

Statistical Analyses

Analysis of molecular variance (AMOVA) (Excoffier et al., 2005) was first performed to determine the diversity among and within the Brachiaria groups using GenAlEx 6.5 (Peakall and Smouse, 2012). A priori groups were assigned according to the nine morphological groups previously defined by Renvoize et al. (1996). Population structure was then examined with STRUCTURE 2.3.4 (Pritchard et al., 2000), which used a Bayesian approach to infer the optimal number of populations within the Brachiaria collection using only genotype scores as a priori information. A “no admixture” model was selected with five iterations for each K value (K = 1–15) and a 500,000 burn-in period with 800,000 Markov Chain Monte Carlo repetitions. The no admixture model was used to calculate the posterior probabilities of belonging to each K group for each individual. Each individual was then assigned to the group with the highest probability of membership. This setting was chosen because we wanted to consider discrete populations with no prior knowledge about the origin of the populations given the apomictic reproductive mode of most genotypes and the fact that our collection included several distantly related species that are not interfertile. We used an independent allele frequencies model, which assumes that allele frequencies are reasonably different across distinct populations. The optimal K value was determined following Evanno et al. (2005) using Structure Harvester (Earl and vonHoldt, 2012).

Both neighbor joining (NJ) and unweighted pair group method with arithmetic mean (UPGMA) approaches were used in this study as an additional means of validation. Many previously published interspecific Brachiaria phylogenies (Torres González and Morton, 2005; Nitthaisong et al., 2016) used just one or two approaches to analyze their data and arrived at conclusions that contradict well-established biological evidence. We explored population structure in our collection with a wide array of clustering methods to make sure our results were robust. Two pairwise dissimilarity matrices were created using Jaccard’s coefficient (Saitou & Nei, 1987) and Nei’s distance (Nei, 1978) to generate NJ and UPGMA trees. Both dendrograms were built using R software (version 3.3.1; R Development Core Team, 2016) with the “ape, phangorn, adegenet, and phytools” libraries for phylogenetic analysis. The bootstrap method was used to evaluate the parsimony and reliability of tree topology with 10,000 replicates. Finally, the collection was analyzed using multiple correspondence analysis (MCA) with graphical representation through R with the FactoMiner, ade4, and Pca3D libraries.


RESULTS

Microsatellite Diversity and Analysis of Molecular Variance

Thirty-nine SSR markers with 701 alleles were used to genotype the 257 Brachiaria accessions, cultivars, and breeding populations and four P. maximum cultivars included in this study (Supplemental Table S2). According to AMOVA analysis, 57% of total genetic variation was found within the Renvoize et al. (1996) taxonomic groups, whereas 43% of variation was distributed among taxonomic groups (Table 1). Because the SSR sequences were identified in the interspecific Brachiaria hybrid Mulato II (CIAT 36087) and selected on the basis of their amplification and polymorphism in three B. humidicola accessions (CIAT 679, CIAT 16888, and CIAT 26146), ascertainment bias could pose a problem in this study. However, we found no evidence of ascertainment bias in the data, as there was no relationship between the number of alleles present in each species and its taxonomic group. The average number of marker bands found per species ranged from 26 to 51 (Supplemental Table S3). The highest average number of marker bands was found in B. brizantha, which belongs to the same taxonomic group as Mulato II (Group 5). However, the distantly related species B. eruciformis (Group 7) and B. serrata (Group 2) had the second and third highest numbers of present marker bands (Supplemental Table S3).


View Full Table | Close Full ViewTable 1.

Analysis of molecular variance for populations assigned a priori according to Renvoize et al. (1996) taxonomic groups.

 
Source of variation Df Sum of squares Variance components Variance
%
Among taxonomic groups 5 5,423 37.6 43
Within taxonomic groups 256 12,946 50.6 57
Total 261 18,369 88.2 100

STRUCTURE Analysis

Bayesian STRUCTURE analysis revealed that the collection of 261 genotypes formed four distinct groups (K = 4). At K = 4, the species B. ruziziensis and B. decumbens, the interspecific hybrid cultivars and sexual breeding population, and one accession labeled as B. jubata (CIAT 26326) were located in the first group (C I), the B. brizantha accessions were placed in the second group (C II), the B. humidicola germplasm accessions and sexual B. humidicola breeding population were located in the third group (C III), and the fourth group (C IV) included the noncommercial species B. arrecta, B. bovonei, B. dictyoneura, B. jubata, B. platynota, B. subulifolia, B. nigropedata, B. serrata, B. eruciformis, and the P. maximum outgroup (Fig. 1). Although the B. brizantha, B. ruziziensis, and B. decumbens agamic complex was split into two groups (C I and C II), many genotypes in these groups displayed significant admixture. The apomictic polyploid B. decumbens accessions (CIAT 606, CIAT 6693, CIAT 16489, CIAT 16497, CIAT 16501, CIAT 26112, CIAT 26182, and CIAT 26183) and the apomictic interspecific hybrid cultivar Mulato II (CIAT 36087) showed admixture with the B. brizantha group (C II), whereas the diploid sexual B. decumbens accessions (CIAT 16495, CIAT 26186, CIAT 26301, and CIAT 26308) were conclusively placed in group C I (Fig. 1). Several B. brizantha accessions (CIAT 6297, CIAT 6780, CIAT 16827, CIAT 16829, CIAT 16830, CIAT 16835, and CIAT 26032) also had significant admixture with group C I. Within B. humidicola, three accessions (CIAT 675, CIAT 679, and CIAT 26146) showed a significant degree of admixture with cluster C IV in the STRUCTURE analysis (Fig. 1).

Fig. 1.
Fig. 1.

STRUCTURE assigned groupings of the 261 genotypes included in this study. Vertical bars indicate the estimated membership coefficients (Q) of each individual for each population cluster and are color coded according to the four optimal clusters: CI, yellow; CII, red; CIII, blue; CIV, green. The species designation of each genotype according to the CIAT Genetic Resources database is indicated by a color-coded dot below its numerical identification code. Renvoize et al. (1996) taxonomic group assignments for each species are provided in parentheses in the figure legend.

 

At K = 8, groups C I, C II, and C III remained intact, whereas the noncommercial species in C IV were split into five subgroups (data not shown). The first subgroup was composed of B. arrecta, B. dictyoneura, and B. dura. Another subgroup was contained B. bovonei, B. jubata, B. nigropedata, and B. subulifolia accessions. The third subgroup consisted of B. eruciformis and B. serrata accessions. The B. platynota and P. maximum accessions each formed their own subgroups. These subgroupings had little correspondence with the taxonomic groups developed by Renvoize et al. (1996).

Similarity-Based Analyses

The results of NJ and UPGMA analyses were generally in strong agreement (Fig. 2, Supplemental Fig. S1). The NJ tree was composed of two primary clusters, with the B. brizantha, B. decumbens, and B. ruziziensis agamic complex separated from the rest of the collection (Fig. 2). The agamic complex was further divided into three subgroups. The first subgroup was composed of B. ruziziensis, B. decumbens, and four interspecific hybrids. The four synthetic autotetraploid B. ruziziensis genotypes were clustered tightly with the four interspecific hybrid genotypes JIRCAS 185, JIRCAS 203, JIRCAS 226, and CIAT 36061 (Fig. 2). The diploid sexual B. decumbens accessions and CIAT 26326 (B. jubata) were clustered with the diploid B. ruziziensis accessions, whereas the seven tetraploid B. decumbens accessions formed an independent cluster (Fig. 2). The second subgroup was composed entirely of B. brizantha accessions, and the third subgroup included the SX14 sexual breeding population, three hybrid cultivars (BR02/1752, BR02/1794, and CIAT 36087), and seven B. brizantha accessions (CIAT 6297, CIAT 6780, CIAT 16827, CIAT 16829, CIAT 16830, CIAT 16835, and CIAT 26032) (Fig. 2). These clustering patterns were corroborated by UPGMA analysis, although the B. decumbens accessions formed a monophyletic branch in the UPGMA dendrogram (Supplemental Fig. S1).

Fig. 2.
Fig. 2.

Neighbor joining tree of the 261 genotypes included in this study according to Jaccard’s coefficient. Each genotype is color coded according to its species designation in the CIAT Genetic Resources database. Renvoize et al. (1996) taxonomic group assignments for each species are provided in parentheses in the figure legend.

 

Although the species in Renvoize et al. (1996) taxonomic Group 6 were split into two clusters by STRUCTURE at K = 4 (B. humidicola in C III; B. bovonei, B. dictyoneura, B. jubata, B. platynota, and B. subulifolia in C IV), all the species except B. platynota formed a monophyletic branch in the UPGMA tree (Supplemental Fig. S1). Brachiaria bovonei, B. jubata, and B. subulifolia all appeared very closely related, as they formed a monophyletic branch distinct from the other taxonomic Group 6 species in the NJ and UPGMA dendrograms (Fig. 2, Supplemental Fig. S1). Brachiaria dictyoneura was intermediate between B. humidicola and the B. bovonei, B. jubata, and B. subulifolia cluster. Within B. humidicola, three accessions (CIAT 675, CIAT 679, and CIAT 26146) were highly differentiated from the other 55 B. humidicola accessions in NJ and UPGMA analyses (Fig. 2, Supplemental Fig. S1). The 10 sexual genotypes from the first cycle of intraspecific recurrent selection in B. humidicola (BhSX12s) formed an intermediate group between CIAT 26146 and the rest of the B. humidicola collection (Fig. 2, Supplemental Fig. S1).

The two species from Renvoize et al. (1996) taxonomic Group 2 included in this study, B. nigropedata (n = 8) and B. serrata (n = 2), were placed in adjacent clusters along with the B. eruciformis (n = 1) and B. arrecta (n = 2) accessions between the P. maximum outgroup and the agamic complex species in the NJ trees (Fig. 2). However, the B. nigropedata accessions were tightly clustered together and very distinct from all other noncommercial species from STRUCTURE cluster C IV in the UPGMA analysis (Supplemental Fig. S1). The two B. platynota accessions (Renvoize et al., 1996; Group 6) were placed adjacent to the only B. dura accession (Renvoize et al., 1996; Group 5) in NJ and UPGMA analyses (Fig. 2, Supplemental Fig. S1).

Multiple Correspondence Analysis

The three primary principal components identified by MCA explained only 6.72, 4.74, and 4.64% of the population variance respectively (Fig. 3). The three agamic complex species were closely aligned on Principal Component 1 (PC1), with B. ruziziensis placed most closely to the origin and the interspecific hybrid genotypes, B. decumbens, and B. brizantha further to the left. The B. humidicola accessions and sexual breeding population were clustered together to on the far right of PC1, whereas the most of the noncommercial species were loosely clustered with the P. maximum outgroup near the origin. The two B. platynota accessions and the B. serrata and B. eruciformis accessions formed two separate clusters that were very distant from the rest of the collection along PC2 and PC3.

Fig. 3.
Fig. 3.

Multiple correspondence analysis of the 261 genotypes included in this study. Each genotype is color coded according to its species designation in the CIAT Genetic Resources database. Renvoize et al. (1996) morphological group assignments for each species are provided in parentheses in the figure legend. The four optimal STRUCTURE clusters are outlined and color coded (CI, yellow; CII, red; CIII, blue; CIV, green). Outlying B. platynota, B. serrata, and B. eruciformis genotypes are indicated with arrows. PC, principal component.

 


DISCUSSION

Taxonomic Group 5 and the Agamic Complex

The close genetic relationships revealed between the species comprising the agamic complex (B. brizantha, B. decumbens, and B. ruziziensis) in the NJ, UPGMA, and MCA analyses are contradicted by the results of Torres González and Morton (2005) and Nitthaisong et al. (2016). However, our results are corroborated by documented interfertility (Valle and Savidan, 1996), transferability of SSR markers (Ferreira et al., 2016), and morphological similarities across species described by Renvoize et al. (1996), who placed all three species in taxonomic Group 5. However, B. dura (n = 1), the only other taxonomic Group 5 species represented in this study, appeared distantly related to the agamic complex and was placed with the noncommercial and outgroup species in STRUCTURE, NJ, UPGMA, and MCA analyses.

Brachiaria ruziziensis and B. decumbens were clearly the two most closely related species of the agamic complex according to STRUCTURE, NJ, and UPGMA analyses. These results conflict with the findings of Ambiel et al. (2008), who used 107 polymorphic RAPD marker bands to create a UPGMA dendrogram and found that the B. ruziziensis accessions were all clustered together in one branch, whereas the B. decumbens and B. brizantha accessions and cultivars were intermixed in two other branches. However, Ferreira et al. (2016) also found that B. decumbens and B. ruziziensis were more closely related to one another than to B. brizantha in their NJ analysis of 34 genotypes from the EMBRAPA collection. The close genetic relationship between B. decumbens and B. ruziziensis observed in this study is further supported by morphological similarities between the two species, including decumbent habit, lanceolate leaf blades, winged rachis, and spikelets borne in two rows (Renvoize et al., 1996).

Ferreira et al. (2016) conducted STRUCTURE and NJ analyses on a collection of 24 B. decumbens accessions from the EMBRAPA Brachiaria collection and found no evidence for genetic differentiation between diploid sexual and polyploid apomictic genotypes. In contrast, we found that the diploid sexual and tetraploid apomict B. decumbens accessions formed two distinct subclusters in the NJ and UPGMA dendrograms and the STRUCTURE analysis, where only tetraploid apomict accessions showed moderate admixture with the B. brizantha group. The genetic distance observed between diploid and tetraploid B. decumbens accessions is not surprising given the reproductive barriers imposed by differences in ploidy and apomixis. Conclusive evidence of segmental allopolyploidy was recently documented in a tetraploid apomictic accession of B. decumbens (CIAT 606, cv. Basilisk) (Worthington et al., 2016). The admixture with B. brizantha observed in the tetraploid apomictic B. decumbens accessions in the STRUCTURE analysis suggests that one subgenome of B. decumbens may be closely related to B. ruziziensis, whereas the other is more similar to B. brizantha.

The population structure of interspecific hybrid cultivars and sexual breeding populations provides evidence for the effects of selection during the past 20 yr. The synthetic sexual autotetraploid B. ruziziensis accessions (BRX 44-02, MOK 8-4, MOK 8-6, and MOK 8-7) were closely related to the diploid B. ruziziensis accessions but formed a distinct subgroup in the UPGMA and NJ trees. The first interspecific cultivar released from the CIAT breeding program (CIAT 36061, cv. Mulato) and the three interspecific JIRCAS hybrids (JIRCAS 185, JIRCAS 203, and JIRCAS 226) were placed with the tetraploid B. ruziziensis subgroup in both dendrograms. Although the male parent of CIAT 36061 is unknown, its female parent was derived from a first generation cross between B. ruziziensis (BRX 44-02) and B. brizantha (CIAT 6297, cv. Marandu). The interspecific JIRCAS hybrids were selected from crosses between the synthetic autotetraploid B. ruziziensis genotypes (MOK 8-4, MOK 8-6, and MOK 8-7) and CIAT 36061 (Thaikua et al., 2015).

Unlike older interspecific hybrids, the 10 genotypes from the most recent cycle of CIAT’s fully sexual interspecific recurrent selection population (SX14s) and the three more recently released apomictic hybrid cultivars (CIAT 36087, cv. Mulato II; BR02/1752, cv. Cayman; and BR02/1794, cv. Cobra) were placed with a subgroup of B. brizantha in the NJ and UPGMA trees. The seven B. brizantha accessions (CIAT 6297, CIAT 6780, CIAT 16827, CIAT 16829, CIAT 16830, CIAT 16835, and CIAT 26032) that grouped with the newer interspecific hybrids formed a distinct clade apart from the rest of the B. brizantha accessions in UPGMA and NJ trees and showed 10 to 70% admixture with B. ruziziensis in the STRUCTURE analyses. Three of the seven B. brizantha accessions in this cluster (CIAT 6297, cv. Marandu; CIAT 16827; and CIAT 16829) were founding parents of CIAT’s synthetic, tetraploid, sexual interspecific recurrent selection population, respectively contributing 5.17, 10.34, and 8.62% of the parentage of the population (Miles et al., 2006). Interestingly, two other founding parents, BRX 44-02 (B. ruziziensis) and CIAT 606 (B. decumbens), contributed far more parentage to the first cycle of recurrent selection (50 and 17.24% respectively) than the B. brizantha founders (Miles et al., 2006). BRX 44-02 and CIAT 606 are both very susceptible to spittlebug (Hemiptera: Cercopidae), whereas CIAT 6297, CIAT 16827, and CIAT 16829 are the three most spittlebug-resistant genotypes of the 10 founding parents (Miles et al., 2006). Thus, the probable cause for the outsize contribution of these three B. brizantha parents to the most recent recurrent selection generations and interspecific hybrid cultivars is the intense selection pressure for spittlebug resistance that was applied during the first six cycles of intrapopulation population recurrent selection (Miles et al., 2006).

In addition to information about the genetic structure of natural and breeding populations, the results of this study have also provided clarification on the species identity and reproductive mode of several CIAT genebank accessions within the agamic complex. Although CIAT 606 (cv. Basilisk) is listed as B. decumbens in the CIAT genebank, taxonomists including Renvoize et al. (1996) have questioned this assignment and tentatively classified the accession as B. brizantha. Molecular evidence, including placement of CIAT 606 with commercial B. brizantha cultivars in the UPGMA tree by Ambiel et al. (2008), further called into question the placement of CIAT 606 in B. decumbens. However, this study conclusively places CIAT 606 with other tetraploid apomictic B. decumbens accessions in STRUCTURE, NJ, and UPGMA analyses, confirming its original species assignment (Oram, 1990). CIAT 26326 was the only accession from the agamic complex that was clearly misidentified in the CIAT genebank. Despite being classified as B. jubata in the CIAT genetic resources database, this accession was grouped with the diploid sexual B. decumbens accessions in all four analyses.

The results of this study also provide some clarification on the reproductive mode of two accessions, CIAT 26179 and CIAT 26186. CIAT 26179 was classified as an apomictic tetraploid accession of B. brizantha in the EMBRAPA genebank, and this assignment was subsequently confirmed by studies of reproductive mode (Vigna et al., 2011) and chromosome counts (Mendes-Bonato et al., 2002). However, CIAT 26179 is listed as B. ruziziensis in the CIAT genetic resources database and was classified as sexually reproducing by the diagnostic apomixis marker p779/p780 (Worthington et al., 2016). Similarly, the B. decumbens accession CIAT 26186 was classified as sexually reproducing by marker analysis with p779/p780 but was described as a tetraploid apomict by EMBRAPA (Penteado et al., 2000; Valle et al., 2008). In this study, CIAT 26179 was clearly placed with the diploid B. ruziziensis accessions, and CIAT 26186 clustered with the diploid B. decumbens accessions in both dendrograms and STRUCTURE analysis. Thus, it seems likely that these accessions labeled as CIAT 26179 and 26186 at CIAT and EMBRAPA are actually different genotypes and not rare recombinants in the apospory-specific genomic region, as have been reported by Conner et al. (2013) in Cenchrus.

Taxonomic Group 6 and B. humidicola

The UPGMA dendrogram, NJ tree, and MCA results presented in this study support the classification of B. humidicola, B. bovonei, B. dictyoneura, B. jubata, and B. subulifolia into taxonomic Group 6 (Renvoize et al., 1996). The species B. bovonei, B. jubata, and B. subulifolia all appeared very closely related, as they formed a monophyletic branch distinct from the other Group 6 species in the UPGMA dendrogram. The only taxonomic Group 6 species that did not appear closely related to the others was B. platynota (n = 2), which was placed adjacent to the lone B. dura accession in the UPGMA and NJ analyses and situated alone as an outgroup in the MCA.

Brachiaria dictyoneura and B. humidicola are morphologically similar species, differentiated primarily by growth habit; B. humidicola is more stoloniferous than B. dictyoneura, which has a more tufted habit (Renvoize et al., 1996). Both species have been shown to have a base chromosome number of six instead of the more common Brachiaria base chromosome numbers x = 7 and 9 (Risso-Pascotto et al., 2006; Boldrini et al., 2009). Furthermore, these species have at times been considered synonyms (Skerman and Riveros, 1990), and B. humidicola CIAT 6133 (cv. Llanero) was often described as B. dictyoneura in older literature (Bogdan, 1977). The results of this study support the conclusion that B. humidicola and B. dictyoneura are two closely related but distinct species. Despite its earlier misclassification as B. dictyoneura, CIAT 6133 (cv. Llanero) was clearly placed with other B. humidicola accessions in STRUCTURE, NJ, and UPGMA analyses.

Within B. humidicola, three accessions (CIAT 675, CIAT 679, and CIAT 26146) were highly differentiated from the rest of the B. humidicola entries in STRUCTURE, NJ, and UPGMA analyses. The lone polyploid sexual accession, CIAT 26146, was particularly distantly related to the other B. humidicola accessions, as was previously shown in the intraspecific phylogeny of B. humidicola conducted by Jungmann et al. (2010). The 10 sexual genotypes from the first cycle of intraspecific recurrent selection in B. humidicola (BhSX12s) were derived from open pollination between sexual progeny of crosses between CIAT 26146 and 18 apomictic B. humidicola accessions. Consequently, this breeding population formed a tightly clustered intermediate group between CIAT 26146 and the rest of the B. humidicola collection.

Noncommercial and Outgroup Species

The two species from Renvoize et al. (1996) taxonomic Group 2 included in this study, B. nigropedata (n = 8) and B. serrata (n = 2), did not appear closely related in any of the analyses. The B. nigropedata accessions were tightly clustered together and very distinct from all other species, whereas the B. serrata accessions formed an outgroup with the lone B. eruciformis accession (CIAT 16942). The extremely close genetic relationship between B. eruciformis and B. serrata found in this study is highly improbable, given the placement of these species in different taxonomic groups by Renvoize et al. (1996). Thus, it seems possible that CIAT 16942 has been misclassified as B. eruciformis in the CIAT genebank and is actually an accession of B. serrata. The placement of the two B. platynota accessions (Renvoize et al., 1996; Group 6) tightly adjacent to the only B. dura accession (Renvoize et al., 1996; Group 5) in NJ and UPGMA analyses is also suspect and suggests that one or more of those genotypes may be misclassified. The increased incidence of potentially misidentified genotypes observed in the outgroup species could be due to decreased familiarity with noncommercial species or to biased results caused by small sample size in this study.

Implications for Brachiaria Breeding

Current Brachiaria breeding programs at CIAT and EMBRAPA use recurrent selection for specific combining ability to improve the mean performance of a sexually reproducing population that is crossed to the same apomictic “tester” genotype each cycle (Miles, 2007; Barrios et al., 2013). The recent discovery of a broadly diagnostic molecular marker for apomixis in multiple Brachiaria species makes it logistically possible to determine the reproductive mode of many genotypes in large segregating populations (Worthington et al., 2016). This advance facilitates the simultaneous improvement of two complementary populations segregating for apomixis using reciprocal recurrent selection as outlined by Worthington and Miles (2015). The first step toward implementing this new procedure is the identification or establishment of complementary heterotic groups. Brachiaria brizantha and B. decumbens have been proposed as promising heterotic groups given their interfertility and complementary traits of interest (Worthington and Miles, 2015). Brachiaria brizantha is resistant to spittlebug damage, whereas B. decumbens has greater persistency and resistance to aluminum toxicity (Miles et al., 2006). The clear differentiation and genetic distance observed between B. brizantha and B. decumbens in this study supports their use as heterotic groups, although the performance of progeny produced from crosses of parents in the same and opposing groups should be evaluated to test heterotic patterns. The 10 genotypes from the most recent cycle of the SX14 recurrent selection population were grouped tightly with three founding parents of the interspecific breeding program (CIAT 6297, CIAT 16827, and CIAT 16829) and four other B. brizantha accessions in a cluster distinct from the rest of the B. brizantha collection. The differentiation between these subgroups suggests that there may also be intraspecific heterotic patterns to exploit within B. brizantha. The strong differentiation between the only sexually reproducing polyploid B. humidicola accession (CIAT 26146) and the BhSX12 breeding population from the rest of the B. humidicola collection suggests that it should be possible to create complementary heterotic groups segregating for reproductive mode in B. humidicola as well.


CONCLUSIONS

Large intraspecific phylogenies of commercial Brachiaria species have provided useful information about the genetic structure of the largely overlapping CIAT and EMBRAPA Brachiaria collections (Jungmann et al., 2010; Vigna et al., 2011). However, previously published genus-level phylogenies (Torres González and Morton, 2005; Nitthaisong et al., 2016) have largely contradicted well-established evidence of morphological patterns among and within taxonomic groups (Renvoize et al., 1996) and interfertility among species in the B. brizantha, B. decumbens, B. ruziziensis agamic complex (Valle and Savidan, 1996). The current study, which used 701 polymorphic SSR bands to explore population structure in a collection of 261 genotypes representing 14 Brachiaria species and a P. maximum outgroup through STRUCTURE, UPGMA, NJ, and MCA methodology, largely corroborated conventional taxonomic groupings based on similarity of panicle and inflorescence morphology, and interfertility patterns. Furthermore, the inclusion of hybrid cultivars and genotypes from the CIAT and JIRCAS breeding programs showed the effects of selection for spittlebug resistance on the genetic structure on recurrent selection populations. These findings demonstrate that interspecific phylogenies are made more useful and robust by the inclusion of large numbers of genotypes and corroboration of results with multiple clustering methods.

Conflict of Interest

The authors declare that there is no conflict of interest.

Supplemental Material Available

Supplemental material for this article is available online.

Acknowledgments

Thanks to the CIAT Genetic Resources team for providing germplasm accessions and to Lucia Chavez for her assistance with DNA extraction and shipments. Thanks to Miho In-nami, Etsuko Kaneda, and Yuri Ohnishi, who helped with PCR reactions and scoring marker bands. Thanks also to Dr. G. Ishigaki and Professor R. Akashi of Miyazaki University for providing tetraploid sexual B. ruziziensis accessions, and to John Miles for his careful review of this manuscript. Funding for this research was provided by the CGIAR Livestock and Fish coordinated research project.

 

References

Footnotes



Files:

Comments
Be the first to comment.



Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.