Introduction
Rice production in the upland ecology holds enormous potentials to meet the dual objectives of providing food and poverty escape in Africa (Gassner et al., Reference Gassner, Harris, Mausch, Terheggen, Lopes, Finlayson and Dobie2019). Agriculture is the traditional mainstay of the African economy with agricultural activities predominantly in the upland ecology, which covers a meaningful proportion of the arable land areas in the region (Allurri et al., Reference Alluri, Mahsatra and Lawson1979; Diagne et al., Reference Diagne, Amovin-Assagba, Futakuchi and Wopereis2013). Rice grain yield in the ecology is characteristically low (Oikeh et al., Reference Oikeh, Nwilene, Agunbiade, Oladimeji, Ajayi, Mande, Tsunematsu and Samejima2008; Diagne et al., Reference Diagne, Amovin-Assagba, Futakuchi and Wopereis2013) because the poor resource farmers cannot access sophisticated infrastructures, including irrigation technology, and chemicals to boost their productivity asides the myriad of biotic and abiotic constraints. Despite these limitations, rice production in Africa is significantly dependent on the upland ecology, which accounts for about 32% of the rice-growing areas in the region (Diagne et al., Reference Diagne, Amovin-Assagba, Futakuchi and Wopereis2013).
Consequently, rice improvement programmes are exploring various procedures to boost rice productivity in the upland ecology (Saito et al., Reference Saito, Asai, Zhao, Laborte and Grenier2018). Breeding works aimed at finding promising cultivars with superior characteristics over existing varieties are also intensified (Saito et al., Reference Saito, Sokei and Wopereis2012; Tollens et al., Reference Tollens, Demont, Sié, Diagne, Saito, Wopereis, Wopereis, Johnson, Ahmadi, Tollens and Jalloh2013; Saito et al., Reference Saito, Asai, Zhao, Laborte and Grenier2018). However, the reality of changing climate may further complicate plant breeding efforts (Wassmann et al., Reference Wassmann, Jagadish, Heuer, Ismail, Redona, Serraj, Singh, Howell, Pathak and Sumfleth2009; Ceccarelli et al., Reference Ceccarelli, Grando, Maatougui, Michael, Slash, Haghparast, Rahmanian, Taheri, Al-Yassin, Benbelkacem, Labdi, Mimoun and Nachit2010; Dreccer et al., Reference Dreccer, Bonnett, Lafarge, Christou, Savin, Costa-Pierce, Misztal and Whitelaw2013) such that potential cultivars need to demonstrate the ability to adapt to the changing environment.
Plant breeding potentials in improving crops for yield adaptation to environmental stresses rely on genetic variability within the germplasm (Lenaerts et al., Reference Lenaerts, Collard and Demont2019). Some works are available on genetic variation and diversity studies of rice germplasm, being a model crop (Garris et al., Reference Garris, Tai, Coburn, Kresovich and McCouch2005; Yamasaki and Ideta, Reference Yamasaki and Ideta2013; Ramadan et al., Reference Ramadan, Elmoghazy and Mowafi2015; Islam et al., Reference Islam, Khalequzzaman, Prince, Siddique, Rashid, Ahmed, Pittendrigh and Ali2018; Ndjiondjop et al., Reference Ndjiondjop, Semagn, Sow, Manneh, Gouda, Kpeki, Pegalepo, Wambugu, Sié and Warburton2018). However, rice genetic diversity keeps expanding with the development of new breeding lines and continuous germplasm collections. Thus, a need to continually elucidate genetic diversity among germplasms to aid rice breeding programmes, and population development, particularly for the upland ecology.
Crop diversity may be assayed at phenotypic and genotypic levels, exploring statistical procedures to partition phenotypic or genetic descriptors into genetic or environmental components (Cobb et al., Reference Cobb, Declerck, Greenberg, Clark and McCouch2013). The use of genetic markers, most especially single nucleotide polymorphism (SNP) markers have proven more effective in detecting patterns of crop diversities (Fischer et al., Reference Fischer, Rellstab, Leuzinger, Roumet, Gugerli, Shimizu, Holderegger and Widmer2017). In recent times, advanced Genotyping-by-Sequencing (GBS) platforms, such as Diversity Array Technology (DArT) (Sansaloni et al., Reference Sansaloni, Petroli, Jaccoud, Carling, Detering, Grattapaglia and Kilian2011), have enabled timely mining of high-density SNP information at a significantly reduced cost (Poland and Rife, Reference Poland and Rife2012). Consequently, diversity studies of several crop species have shown relatively better coverage and reduced missing data with DArT compared to other GBS platforms (Chen et al., Reference Chen, Zavala, Ortega, Petroli, Franco, Burgueño, Costich and Hearne2016; Nazzicari et al., Reference Nazzicari, Biscarini, Cozzi, Brummer and Annicchiarico2016; Yang et al., Reference Yang, Wei, Liu, Wu, Li, Zhang, Jian, Li, Tai, Zhang, Zhang, Jiang, Xia and Wan2016a).
The objectives of this study were to examine the genetic diversity and population structure of 176 upland rice germplasm using the DArT-SNP markers to gain a better understanding of their potential utilization in rice improvement programmes for the upland ecology.
Materials and methods
Plant materials
A population of 176 upland rice germplasm consisting of breeding lines, improved varieties and landraces was studied (online Supplementary Table S1). The breeding lines were developed, and preserved with other germplasm at Africa Rice Center (AfricaRice), Ibadan, Nigeria. Fresh leaves were collected from each of the germplasm, lyophilized and shipped in 96-plex to DArT Pty Ltd, Canberra, Australia, for genotyping by sequencing using diversity array technology (Sansaloni et al., Reference Sansaloni, Petroli, Jaccoud, Carling, Detering, Grattapaglia and Kilian2011; Ren et al., Reference Ren, Ray, Li, Xu, Zhang, Liu, Yao, Kilian and Yang2015).
DArT-based Genotyping-By-Sequencing
DArT-based GBS platform was used for the genotypic analysis and SNP calling. Procedures for DNA extraction, complexity reduction, cloning, library construction, cleaning and DNA quality assessment were previously described by Egea et al. (Reference Egea, Mérida-García, Kilian, Hernandez and Dorado2017). Amplification fragments of DNA sequence on the IlluminaHiseq 2500 (www.illumina.com), and the final SNPs called using the DArTsoft analytical pipeline (http://www.diversityarrays.com/darttechnology-package-dartSoft).
Data analysis
SNP filtering and genome characterization
The DArT GBS platform generated SNP characteristics, including missing rate, polymorphism information content (PIC) and frequency of heterozygosity for all SNPs. The filtering of the SNPs was done in two stages to exclude all SNPs unmapped to any of the 12 rice chromosomes and remove redundant markers. Markers with minor allele frequency below 5%, polymorphism information index (PIC) <10% and missing rate of more than 10% were considered redundant. The genome summary plugin in TASSELv.5.2.37 software (Bradbury et al., Reference Bradbury, Zhang, Kroon, Casstevens, Ramdoss and Buckler2007) generated the allele frequency and MAF of markers.
Clustering and structure analysis
Clustering and principal component (PC) analyses of the 176 upland rice germplasm were implemented in TASSELv.5.2.37 software (Bradbury et al., Reference Bradbury, Zhang, Kroon, Casstevens, Ramdoss and Buckler2007). Grouping of genotypes into clusters was obtained using the Unweighted Pair Group Method with Arithmetic Means (UPGMA) and displayed with the Archaeopteryx Tree plugin.
The structure of the population was further estimated using a model-based (Bayesian) analysis in STRUCTUREv2.3.4 (Pritchard et al., Reference Pritchard, Stephens and Donnelly2000). The k-values representing the fixed number of population subgroups (1–10) obtained in four independent analyses. The burn-in iterations and Markov Chain Monte Carlo (MCMC) replication were each programmed at 100,000. The k-value that is the best fit for the population was by Structure Harvester (Earl and Vonholdt, Reference Earl and vonHoldt2012). The CLUMPP files for each k-value with several curves, including that of the log probability of the data [LnP(D)] and derived statistics (ΔK) constructed based on the rate of change in [LnP(D)] between successive K-values were generated from Structure Harvester. The structure of the population was determined by k-value, corresponding to the maximum ΔK. Also, individual Q-values were obtained from the corresponding CLUMPP file generated, and genotypes with Q-value >0.5 belonged to the same group.
Analysis of molecular variance
The results of population structures were subjected to the analysis of molecular variance, which estimates and partitions the pattern of variation within and among the population using GenAlEx v6.5 (Peakall and Smouse, Reference Peakall and Smouse2012). Genetic differentiation between populations was determined using Phi-statistics (PhiPT) value. The number of migrants (Nm value) represents the gene flow index. The probability value for the significance of variance components was estimated based on 1000 standard permutations.
Analysis of allelic pattern
Analysis of allelic patterns across the estimated population structures was implemented in GenAlEx v6.5 (Peakall and Smouse, Reference Peakall and Smouse2012). Genetic diversity within and among the population was determined using the following indices: number of private alleles (Pa), number of different alleles (Na), number of effective alleles (Ne), Shannon's information index (I), diversity (h), unbiased diversity (uh) and percentage of polymorphic loci (%PPL) (Peakall and Smouse, Reference Peakall and Smouse2012).
Results
Genome characterization using DArT-based SNP markers
DArT-based GBS generated 23,000 SNPs for the 176 upland rice germplasm. Out of the 23,000 SNPs, the first stage of filtering excluded 2614 unmapped SNPs while retaining 20,386 mapped to the 12 rice chromosomes (Table 1). SNPs per chromosome ranged from 1090 SNPs on chromosome 9 to 2391 SNPs on chromosome 1, with an average of 1699 per chromosome.
Distribution of DArT-SNP markers on the 12 rice chromosomes (filtering at MAF>0.05, missing<0.1 and PIC>0.1)
The second stage of filtering returned 7063 SNPs (Table 1). The SNPs were reasonably distributed across the rice genome, ranging from 432 SNPs (6%) on chromosome 9 to 989 SNPs (14%) on chromosome 1. The PIC ranged from 0.11 to 0.5 and averaged at 0.25 (Fig. 1 and online Supplementary Table S2), while gene diversity (or heterozygosity) of the 7063 SNPs ranged from 0 to 0.57 and averaged at 0.03 (Fig. 2 and online Supplementary Table S2). Almost 98% of the markers had heterozygosity values below 0.1. The minimum MAF was 0.05 (online Supplementary Table S2). The frequencies of transition and transversion SNP types were 0.63 and 0.27, respectively (online Supplementary Table S3).
Polymorphic information content (PIC) of 7063 DArT-SNP markers for 176 upland rice germplasm.
Gene diversity (heterozygosity) of 7063 DArT-SNP markers for 176 upland rice germplasm.
Clustering and structure analysis
The clustering analysis using the UPGMA and the PC analysis identified three clusters for the 176 upland germplasm designated as P1, P2 and P3 (online Supplementary Figs. S1 and S2). The model-based (Bayesian) analysis, however, estimated two groups (Q1 and Q2) (Fig. 3) corresponding to the best-fit K-value of 2 (online Supplementary Fig. S3). There were 15 individuals in group Q1 and 161 individuals in Q2 at cluster value 0.11 and 0.89, respectively. The mean fixation index was 0.431 and 0.1618, while the expected heterozygosity was 0.3529 and 0.7203 for group Q1 and Q2, respectively (online Supplementary Table S4). The allele frequency divergence (net nucleotide distance) between Q1 and Q2 equals 0.4752. The individuals in Q2 sub-divides into two subgroups corresponding to P2 and P3 clusters obtained from the clustering and PC analyses (online Supplementary Table S1).
Population structure of 176 upland rice germplasm base on 7063 SNP markers (K = 2). The colours represent the estimated proportions of the two populations.sgg
Molecular variance and gene flow across the populations
Results from the analysis of molecular variance revealed that all the genetic variance was within the population (100%), and none was attributable to the population structure (0%) estimated from the model-based analysis (Table 2).
Summary analysis of molecular variance (AMOVA) for 176 upland rice germplasm based on 7063 DArT-SNP markers
Probability, P (rand ⩾ data), for PhiPT is based on standard permutation across the full data set.
The PhiPT value for testing the statistical significance of the estimated population structures was 0.001 (P = 0.197), and the number of migrants (Nm) was 622.65.
Allelic patterns across the populations
The average value for the numbers of different (Na), effective (Ne) and private alleles (Pa) across the populations were 1.94, 1.03 and 0.82, respectively (Table 3). Higher values of 2.67 and 1.54 were obtained in Q2 for Na and Pa, while Q1 had 1.22 and 0.10, respectively. The number of effective alleles was 1.03 for both Q1 and Q2.
Genetic diversity indices for the two estimated population structures of 176 upland rice germplasm based on 7063 DArT-SNP markers
Na, No. of different alleles; Ne, No. of effective alleles; I, Shannon's information index; Pa, No. of private alleles; h, diversity; uh, unbiased diversity; %PPL, percentage of polymorphic loci.
The population had mean values for Shannon's information index (I) diversity index (h), unbiased diversity index (uh) of 0.7, 0.02 and 0.03, respectively. The average percentage of polymorphic loci was 55.9%, with 20.9% in Q1 and 90.9% in Q2, respectively.
Discussion
The goal of analysing genetic structure and diversity within the population of individuals is to estimate the pattern of genetic variations, determining their further use in plant breeding programmes, including marker-traits analysis and selection of parental lines for linkage analysis. In recent years, molecular markers – particularly the GBS-SNP markers – have proven more efficient in estimating genome-wide genetic diversity (Fischer et al., Reference Fischer, Rellstab, Leuzinger, Roumet, Gugerli, Shimizu, Holderegger and Widmer2017). In the present study, 23,000 SNPs from the DArT-Seq GBS platform were analysed and filtered to retain 7063 polymorphic SNPs distributed across the 12 rice chromosomes. According to Anderson et al. (Reference Anderson, Pettersson, Clarke, Cardon, Morris and Zondervan2010), well-described reasons for marker exclusion include large amounts of missing data, lower variability and genotyping errors.
GBS platforms such as DArT generate large numbers of SNPs for use in genetic analyses and genotyping (Beissinger et al., Reference Beissinger, Hirsch, Sekhon, Foerster, Johnson, Muttoni, Vaillancourt, Buell, Kaeppler and de Leon2013). However, some authors have reported possibilities of redundant markers that may not be functional or useful for specific genetic analyses (Pootakham et al., Reference Pootakham, Shearman, Ruang-areerate, Sonthirod, Sangsrakru, Jomchai, Yoocha, Triwitayakorn, Tragoonrung and Tangphatsornruang2014; Yang et al., Reference Yang, Chen, Chen, Sun, Huang, Zhou, Huang, Wang, Liu, Wang, Chen and Guo2019). The usefulness of a marker for a particular purpose depends on the number of alleles it has and corresponding relative frequency (Shete et al., Reference Shete, Tiwari and Elston2000). All the 7063 SNP passed the quality control analyses used to remove redundant markers, and are considered suitable for estimating genetic diversity and structure of the studied 176 upland rice germplasm.
The results from the assessment of SNP marker distribution across the chromosome indicated that chromosome 9 had the lowest genetic diversity with the fewest number of polymorphic markers. Previous studies suggested that fewer polymorphic markers are an indication of lower genetic diversity (Chao et al., Reference Chao, Zhang, Akhunov, Sherman, Ma, Luo and Dubcovsky2009; Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018). The transition SNPs occurred twice as frequently as transversion SNPs in our study, which indicates that mutation does not occur randomly within the markers. The transition SNPs were more frequent than transversions in previous studies on maize (Morton et al., Reference Morton, Bi, McMullen and Gaut2006), Camellia sinensis (Yang et al., Reference Yang, Ren, Ray, Xu, Li, Zhang, Liu, Yao and Kilian2016b), Camelina sativa (Luo et al., Reference Luo, Brock, Dyer, Kutchan, Schachtman, Augustin, Ge, Fahlgren and Abdel-Haleem2019), Brassica napus (Huang et al., Reference Huang, Deng, Guan, Li, Lu, Wang, Fu, Mason, Liu and Hua2013), Brassica rapa (Park et al., Reference Park, Yu, Mun and Lee2010) and may be due to synonymous mutations in protein-coding (Guo et al., Reference Guo, McDowell, Nodzenski, Scholtens, Allen, Lowe and Reddy2017).
The values of PIC and heterozygosity index of markers are indications of the diversity level of the population and molecular markers. The range of PIC values for the studied rice germplasm revealed a moderate level of genetic diversity, which is in line with the use of PIC values for genetic diversity by Botstein et al. (Reference Botstein, White, Skolnick and Davis1980). Results from the present research are also similar to the findings of Luo et al. (Reference Luo, Brock, Dyer, Kutchan, Schachtman, Augustin, Ge, Fahlgren and Abdel-Haleem2019) and Eltaher et al. (Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018), who studied genetic diversity in C. sativa and Nebraska wheat, respectively. Eltaher et al. (Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018) argued that moderate PIC values might be expected due to low mutation rates (Coates et al., Reference Coates, Sumerford, Miller, Kim, Sappington, Siegfried, Siegfried and Lewis2009), as indicated by the ratio of transition to transversion SNPs in this study. It may also be attributed to the bi-allelic nature of SNPs, restricting PIC values to 0.5 when the two alleles have identical frequencies (Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018). More so, PIC values of SNPs are usually low compared to that of SSR (Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018). The heterozygosity values of below 0.1 found in almost 98% of the markers further reiterate their suitability in estimating the genetic structure and diversity of the studied rice germplasm. Zhu et al. (Reference Zhu, Niu, Shi and Mou2019) estimated genetic diversity in olive using SNP markers with heterozygosity values of <0.1. According to Ndjiondjop et al. (Reference Ndjiondjop, Semagn, Gouda, Kpeki, Dro Tia, Sow, Goungoulou, Sie, Perrier, Ghesquiere and Warburton2017), rice plants are self-pollinators, which may account for the low heterozygosity in this study.
Some previous studies have reported clear genetic structures for many rice germplasm (Garris et al., Reference Garris, Tai, Coburn, Kresovich and McCouch2005; Zhao et al., Reference Zhao, Tung, Eizenga, Wright, Ali, Price, Norton, Islam, Reynolds, Mezey, McClung, Bustamante and McCouch2011; Ndjiondjop et al., Reference Ndjiondjop, Semagn, Sow, Manneh, Gouda, Kpeki, Pegalepo, Wambugu, Sié and Warburton2018). From our results, the cluster and PC analyses revealed three stratifications (P1, P2 and P3), while the model-based analysis structured the population into two distinct groups of 15 (Q1) and 161 (Q2) individuals. The model-based (Bayesian) analysis structured clusters P2 and P3 of the PC analysis into a group, suggesting the two clusters are probably not differentiated.
Furthermore, the stratifications estimated from this study, including the model-based analysis, appear random and do not reflect any specific pedigree, which may be an indication that the genotypes are closely related. The analysis of molecular variance revealed that all the genetic variance was within the population, and none attributed to population structures estimated from either the clustering analysis or the model-based analysis. Previous studies on crops, such as C. sativa (Luo et al., Reference Luo, Brock, Dyer, Kutchan, Schachtman, Augustin, Ge, Fahlgren and Abdel-Haleem2019), wheat (Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018), Bangladesh rice (Islam et al., Reference Islam, Khalequzzaman, Prince, Siddique, Rashid, Ahmed, Pittendrigh and Ali2018) and Japanese rice (Yamasaki and Ideta, Reference Yamasaki and Ideta2013) attributed larger proportion of genetic variation to the individual differences than the population structure.
The estimates of PhiPT from AMOVA indicated that there were no significant differences between the groups or clusters obtained from the population structure analysis (PhiPT = 0.001, P = 0.197). Wright (Reference Wright1978), who studied two allelic systems and used FST for population differentiation, suggested that FST (analogue of PhiPT) value of 0.25 indicates good differentiation between population groups, the range 0.15–0.25 indicates moderate differentiation while differentiation is negligible if the FST value is 0.05 or less.
Lack of significant differentiation among the estimated population structure may be due to high genetic exchange and gene flow, as indicated by the high value of the number of migrants (Nm) (Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018; Luo et al., Reference Luo, Brock, Dyer, Kutchan, Schachtman, Augustin, Ge, Fahlgren and Abdel-Haleem2019). Nuijten and Richards (Reference Nuijten, Richards, Wopereis, Johnson, Ahmadi, Tollens and Jalloh2013) discussed the mechanisms of gene flow in rice farmer's field, implying that various cultural, as well as agro-ecological factors influence gene flow in rice. The studied germplasm is a mix of Oryza sativa (indica and japonica subspecies) and Oryza glaberrima, which are either released varieties or breeding lines selected for better yielding and adaptation to a specific ecology, and this might have influenced the low genetic variation observed.
A specific pattern of population stratifications (structures) may be a result of selection, genetic drift (Buckler and Thornsberry, Reference Buckler and Thornsberry2002; Breseghello and Sorrells, Reference Breseghello and Sorrells2006) and gene exchange (Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018; Luo et al., Reference Luo, Brock, Dyer, Kutchan, Schachtman, Augustin, Ge, Fahlgren and Abdel-Haleem2019), which influences their use in plant breeding programmes (Eltaher et al., Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018). Estimating the patterns of the genetic structure is crucial for rice improvement strategies involving marker-trait association studies, such as genome-wide association scanning for quantitative trait loci of interest and genomic selections. The information obtained from such estimation will build confidence in the outcome of the potential association that may be detected.
Several reports estimated the level of diversity among population groups using indices such as number of private alleles (Pa), number of different alleles (Na), number of effective alleles (Ne), Shannon's information index (I), diversity (h), unbiased diversity (uh) and percentage of polymorphic loci (%PPL) (Peakall and Smouse, Reference Peakall and Smouse2012). Eltaher et al. (Reference Eltaher, Sallam, Belamkar, Emara, Nower, Salem, Poland and Baenziger2018) and Luo et al. (Reference Luo, Brock, Dyer, Kutchan, Schachtman, Augustin, Ge, Fahlgren and Abdel-Haleem2019) both suggested that higher value for these indices is an indication of a higher level of genetic diversity. In our study, comparing genetic diversity indices such as Ne, h and uh also revealed the population structures are not exactly different. However, group Q2 appeared more diverse with higher values of Pa, Na and I. The level of diversity may be attributed to the high gene flow index (Arora et al., Reference Arora, Kundu, Dilbaghi, Sharma and Tiwari2014) but also represents valuable resources for future rice improvement programmes.
In summary, the DArT GBS platform generated high-density markers with reasonable distribution across the whole rice genome. The DArT markers identified two population structures for the 176 upland rice germplasm, which are not significantly differentiated. Nonetheless, substantial diversity existed within the population. These materials may therefore be useful in breeding programmes targeted at recombining favourable alleles within adapted gene pools (McCouch et al., Reference McCouch, Wing, Semon, Venuprasad, Atlin, Sorrells, Jannink, Wopereis, Johnson, Ahmadi, Tollens and Jalloh2013), such as improvement of defensive traits, grain quality characteristics or the improvement of yield per se (Lamkey and Lee, Reference Lamkey and Lee2006; McCouch et al., Reference McCouch, Wing, Semon, Venuprasad, Atlin, Sorrells, Jannink, Wopereis, Johnson, Ahmadi, Tollens and Jalloh2013). More so, introgression of novel alleles from divergent sources (Tanskley and McCouch, Reference Tanksley and McCouch1997; McCouch et al., Reference McCouch, Sweeney, Li, Jiang, Thomson, Septiningsih, Edwards, Moncada, Xiao and Garris2007; Venuprasad et al., Reference Venuprasad, Dalid, Del Valle, Zhao, Espiritu, Sta Cruz, Amante, Kumar and Atlin2009, Reference Venuprasad, Impa, VowdaGowda, Atlin and Serraj2011) and advanced genomic techniques, such as CRISPR/Cas9 technology (Cong et al., Reference Cong, Ran, Cox, Lin, Barretto, Habib, Hsu, Wu, Jiang, Marraffini and Zhang2013; Hsu et al., Reference Hsu, Lander and Zhang2014), may be explored to develop a more diverse population. Our study provided useful insight into the genetic structure and diversity of upland rice germplasm that would guide in planning rice improvement programme in the upland ecology.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1479262120000362
Acknowledgement
We thank the World Bank's Africa Center of Excellence in Agriculture and Sustainable Environment (CEADESE), Federal University of Agriculture, Abeokuta, Nigeria, for providing funds (Grant ACE023) for the genotyping through the Ph.D. Scholarships awarded to Kehinde A. Adeboye and Olayinka E. Oyedeji. We are also grateful to the management of Africa Rice Center, Ibadan, Nigeria, for supplying the plant materials.
