Modelling the genetic architecture of flowering time control in barley through nested association mapping

Modelling the genetic architecture of flowering time control in barley through nested association... Background: Barley, globally the fourth most important cereal, provides food and beverages for humans and feed for animal husbandry. Maximizing grain yield under varying climate conditions largely depends on the optimal timing of flowering. Therefore, regulation of flowering time is of extraordinary importance to meet future food and feed demands. We developed the first barley nested association mapping (NAM) population, HEB-25, by crossing 25 wild barleys with one elite barley cultivar, and used it to dissect the genetic architecture of flowering time. Results: Upon cultivation of 1,420 lines in multi-field trials and applying a genome-wide association study, eight major quantitative trait loci (QTL) were identified as main determinants to control flowering time in barley. These QTL accounted for 64% of the cross-validated proportion of explained genotypic variance (p ). The strongest single QTL effect corresponded to the known photoperiod response gene Ppd-H1. After sequencing the causative part of Ppd-H1, we differentiated twelve haplotypes in HEB-25, whereof the strongest exotic haplotype accelerated flowering time by 11 days compared to the elite barley haplotype. Applying a whole genome prediction model including main effects and epistatic interactions allowed predicting flowering time with an unmatched accuracy of 77% of cross-validated p . Conclusions: The elaborated causal models represent a fundamental step to explain flowering time in barley. In addition, our study confirms that the exotic biodiversity present in HEB-25 is a valuable toolbox to dissect the genetic architecture of important agronomic traits and to replenish the elite barley breeding pool with favorable, trait-improving exotic alleles. Keywords: Barley, Wild barley, Nested association mapping (NAM), Flowering time, Genome-wide association study (GWAS), Quantitative trait locus (QTL), Genomic prediction, Epistasis, Haplotypes Background one promising approach to replenish the elite breeding Barley is among the oldest crop species human civilization pool with new favorable alleles [6-13]. The enriched diver- was built on. Approximately 10,500 years ago, barley was sity may be pivotal to boost the rate of genetic improve- domesticated in the Fertile Crescent [1,2], presumably ment and to cope with the anticipated enhanced effects of followed by additional independent domestication events biotic and abiotic stresses due to climate change. in East Asia [3,4]. Domestication and subsequent genetic In this regard, time of flowering is expected to play a selection led to gene erosion in most crop species’ gene major role in future crop improvement. It is a key trait pools [5,6], threatening future breeding advances. Utilizing for the successful completion of a plant’s life cycle and, the untapped biodiversity, present in wild progenitors is therefore, it has a strong impact on grain yield [14]. Flowering time indicates the transition from vegetative * Correspondence: klaus.pillen@landw.uni-halle.de to reproductive stage, which is mainly influenced by en- Equal contributors vironmental cues like day length (photoperiod) and pro- Institute of Agricultural and Nutritional Sciences, Martin Luther University longed exposure to cold temperatures (vernalization). In Halle Wittenberg, Betty-Heimann-Str. 3, 06120 Halle, Germany Full list of author information is available at the end of the article © 2015 Maurer et al.; licensee BioMed Central. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Maurer et al. BMC Genomics (2015) 16:290 Page 2 of 12 barley, the day length determining light signal is trans- Nested association mapping (NAM) emerged as a multi- mitted from a circadian clock oscillator, with Ppd-H1, parental mapping design to investigate genomic regions a PSEUDO-RESPONSE REGULATOR 7 (PRR7) gene, in with unprecedented genetic resolution and allelic variation its center [15]. Under long day condition, Ppd-H1, through by combining the advantages of linkage analysis and associ- mediation of CONSTANS (CO), promotes the expression of ation mapping [26]. Hence, it facilitates the elucidation of a the floral inducer Vrn-H3, a homolog of the Arabidopsis trait’s genetic architecture via genome-wide association thaliana FLOWERING LOCUS T (FT) gene [16]. On the study (GWAS). Until now, the NAM design was applied to other hand, Vrn-H2, azinc-finger CONSTANS, CO-like and the allogamous species maize and sorghum [26,27]. NAM TOC1 (CCT)-domain protein (ZCCT1)acts asarepressor populations for autogamous species like barley or wheat of Vrn-H3 [17]. Vrn-H2,inturn, isrepressed by Vrn-H1, an have not been developed, yet. In maize, the genetic dissec- APETALA1 family MADS-box transcription factor [18], tion of various agronomic traits, including flowering time, which is up-regulated during vernalization. Thus, after has been investigated [28-34]. However, it was not possible vernalization, the repression of Vrn-H3 is abolished to completely dissect the genetic architecture of flowering and flowering is induced. Based on its vernalization time in maize due to its complex inheritance and the multi- requirement, winter barley and spring barley can be dis- tude of involved small effect QTL. We developed the first tinguished. Spring barley lacks the vernalization require- NAM population in the autogamous species barley, termed ment due to a deletion of Vrn-H2 [19]. ‘Halle Exotic Barley 25’ (HEB-25). The population results Besides photoperiod and vernalization, there are also from initial crosses between the spring barley elite cultivar genetic mechanisms acting independently of environmen- Barke (Hordeum vulgare ssp. vulgare, Hv) and 25 highly tal cues, so-called earliness per se [20]. Although several divergent exotic barley accessions, contributing an ideal key regulatory cereal genes of flowering time were charac- instrument to studybiodiversity. Theexoticdonorscom- terized and finally cloned during the last two decades, still prise 24 wild barley accessions of H. vulgare ssp. sponta- little is known about the genetic architecture underlying neum (Hsp), the progenitor of domesticated barley, and one flowering time regulation in temperate cereals, as com- Tibetian H. vulgare ssp. agriocrithon (Hag) accession. Barke pared to the model species A. thaliana [14,21-23]. So far, was selected since it was also used as a parent of a barley a holistic explanation of flowering time in a segregating high-resolution mapping population [35] and as a genetic germplasm population and the accurate prediction of a stock for mutation screening [36]. The exotic donors were plant’s time of flowering, based on the combined action selected from Badr et al. [37] to represent a substantial part and interaction of major genes, is still not fully achieved in of the genetic diversity that is present across the Fertile cereal species. Furthermore, it is reported that wild barley Crescent, where barley domestication occurred. To gener- accessions possess a rich reservoir of beneficial alleles ate the nested population, F plants were backcrossed to controlling flowering time [7,24,25]. Barke and, subsequently, selfed three times (Figure 1). In Figure 1 Development of the nested association mapping population HEB-25. HEB-25 is made of 25 families with 1,420 NAM lines in BC S . Per 1 3 NAM line, one chromosome pair is illustrated as a double bar. Black and colored bars represent chromosome segments originating from Barke and the exotic donor accessions, respectively. At each SNP locus, HEB-25 is expected to segregate into 71.875% homozygous Barke, 6.25% heterozygous and 21.875% homozygous donor genotypes. Maurer et al. BMC Genomics (2015) 16:290 Page 3 of 12 total, HEB-25 consists of 1,420 BC S lines, divided into 25 to collect flowering time data. The HEB lines flowered 1 3 HEB families of up to 75 lines per family (Additional file 1). on average 68.1 days after sowing with a range from 51.0 to In the present study we investigated the genetic archi- 98.9 days and a standard deviation of 6.5 days (Additional tecture of flowering time in barley. For this purpose, the files 4 and 5). The broad variation in flowering time, cover- NAM population HEB-25 was grown from 2011 to 2013 ing almost 50 days among the 1,420 HEB lines, and a high in multi-field trials to gather data on flowering time. By heritability of 91.6%, as well as the genetic properties of the combining these data with high-density SNP marker infor- NAM population provided an excellent starting point to mation via genome-wide association studies and genomic study the genetic architecture of flowering time through prediction models, we could show that flowering time in GWAS. barley mainly depends on a low number of large-effect QTL and epistatic interactions. Genome-wide association study For GWAS, we initially applied the multiple linear re- gression Model-B with step-wise selection of cofactors, Results and discussion as outlined in Liu et al. [38]. Model-B was found most Characterization of HEB-25 suitable to study traits across multiple related families The inheritance of parental segments across the genomes [39], where a family effect and additional SNPs, selected of the 1,420 HEB lines was characterized through genotyp- as cofactors, are included in the model. GWAS revealed ing 5,709 informative, genic single nucleotide polymorph- eight highly significant major QTL regions controlling ism (SNP) markers [35]. Marker saturation was high with flowering time with P <1.0 E-10 (Table 1, BON-HOLM an average genetic distance of 0.17 cM and a maximum Figure 2, Additional files 6 and 7). Testing the combined gap of 11.1 cM between adjacent markers. Linkage disequi- explanatory power of the single peak markers of the eight librium (LD) among the 26 parents decayed rapidly major QTL revealed a cross-validated explained propor- (Additional file 2) enabling a high mapping resolution tion of genotypic variance (p , [40]) of 64% (Figure 3). To [26]. The SNP data revealed a low degree of genetic check if genetic relatedness, as reported elsewhere [41], similarity between Barke and the donors, ranging from affects this parameter in HEB-25, we also estimated p for 40 to 54% (Additional file 1). Parents and the HEB-25 different sets of eight randomly chosen SNPs, excluding population could be clearly separated in a principal regions with significant QTL. However, since the cross- component analysis (PCA) (Additional file 3). Also, validated explained p remained low with an average of HEB families could be ordered in the PCA based on 8%, we conclude that genetic relatedness between individ- their geographical origin. These findings point to the ual lines does not play a major role in HEB-25. This em- high genetic diversity that is present among HEB-25 phasizes the power and precision of QTL detection in and its parents. HEB-25, which may be a combined effect of the low ex- Diversity in HEB-25 was also visible phenotypically. tent of LD and the particular mating design, resulting in During the seasons 2011 through 2013, HEB-25 was cul- an elevated rate of chromosomal recombination. Thus, tivated at the Halle University Experimental Field Station flowering time of barley can be reliably predicted based on Table 1 List of eight major QTL controlling flowering time in HEB-25 a b b c d e f g h i QTL Chr Pos Range Peak marker No. Seg. Fam. P p CV Freq. Effect CG BON-HOLM G QFt.HEB25-1b 1H 128.3 128.0-128.3 SCRI_RS_150786 25 2.41E-18 0.01 68 −1.4 HvELF3 [46,47] QFt.HEB25-2b 2H 23.0 16.8-23.8 BK_16 24 3.39E-130 0.36 100 −9.5 Ppd-H1 [15] QFt.HEB25-2c 2H 57.4 56.4-58.1 BOPA2_12_30265 25 2.25E-42 0.05 84 −3.0 HvCEN [35] QFt.HEB25-3c 3H 108.4 107.8-109.2 BOPA1_ABC07496_ pHv1343_02 23 2.62E-62 0.04 83 −3.1 denso [45] QFt.HEB25-4a 4H 3.5 3.5 BOPA2_12_31458 24 5.08E-15 0.05 82 3.2 QFt.HEB25-4e 4H 113.4 113.4-114.3 SCRI_RS_216897 24 4.58E-17 0.02 100 2.2 Vrn-H2 [17] QFt.HEB25-5d 5H 125.5 125.5-125.8 BOPA1_4795_782 24 2.31E-33 0.06 60 3.8 Vrn-H1 [18] QFt.HEB25-7a 7H 34.3 25.9-34.3 BOPA2_12_30895 23 6.04E-69 0.07 100 4.1 Vrn-H3 [16] Barley chromosome on which the QTL was determined. Genetic position of the peak marker and range of the QTL in cM, based on Comadran et al. [35]. Marker of the QTL with the highest significance (peak marker). Number of families, in which peak marker is segregating. Significance of the peak marker, expressed as P . BON-HOLM Cross-validated proportion of explained genotypic variance of peak marker. Frequency of significant detections of the peak marker in 100 five-fold cross-validation runs. Difference between the wild genotype and the cultivated genotype in days until flowering. Early flowering effects of exotic alleles are indicated in red. Candidate gene, potentially explaining the QTL effect with reference. Maurer et al. BMC Genomics (2015) 16:290 Page 4 of 12 Figure 2 Genetic architecture of flowering time in HEB-25. Barley chromosomes are indicated as colored bars on the inner circle, centromeres are highlighted as transparent boxes. a) Grey connector lines represent the genetic position of SNPs on the chromosomes. b) Frequency of QTL detection in 100 cross-validation runs via GWAS (0 to 100, grid line spacing: 25); markers with > 50 detections are colored in red. c) Additive effect of the SNP obtained from the BayesCπ genomic prediction model. d) Links in the center of the circle represent significant (P < 0.05) di-genic BON-HOLM interactions between SNP markers via GWAS. Clusters of significant SNP interactions are indicated by different colors. Position of candidate genes, potentially explaining major effects and epistatic effects, correspond to Table 1 and are indicated in blue outside the circle. eight major QTL. This finding is in contrast to flowering acid (GA) in flowering time regulation [44] through detec- time regulation in the allogamous species maize and tion of denso [45] and HvELF3 [46,47] as two further sorghum, where only small effect minor QTL were de- major QTL. Both genes are shown to be involved in GA tected [28,42]. The most significant association in HEB-25 biosynthesis [45,48]. So far, only the QTL on 4HS could (P = 3.4 E-130) was observed on the short arm not been referenced. This QTL, thus, remains a subject BON-HOLM of chromosome 2H and explained a p of 36%. This SNP for further genetic investigations. is directly located within Ppd-H1,the majordeterminant The eight major QTL were located with high genetic of photoperiod response in barley under long day condi- precision, with four QTL restricted to confidence inter- tion [15]. Seven further genomic regions of extraordinary vals of less than 0.9 cM (Table 1). In cases where gene- high significance were detected on chromosome arms specific SNPs were available (i.e. Ppd-H1 and Vrn-H3), 1HL, 2HS, 3HL, 4HS, 4HL, 5HL, and 7HS. All except one exactly those SNPs revealed the highest significance QTL (4HS) could be assigned to known flowering time within the respective QTL window (Additional file 6). genes (Table 1, Figure 2 and Additional file 7). Besides The exotic alleles at Ppd-H1 and Vrn-H3 revealed the Ppd-H1, also the vernalization genes Vrn-H1 and Vrn-H2, strongest effects, accelerating flowering time by 9.5 days as well as the floral inducer Vrn-H3 and its putative para- and delaying flowering time by 4.1 days, respectively. log HvCEN [43] exhibited highly significant effects. In The drastic effects of single QTL outline the high poten- addition, we could confirm the importance of gibberellic tial of introducing wild barley alleles from HEB-25 in Maurer et al. BMC Genomics (2015) 16:290 Page 5 of 12 slight increase of the cross-validated explained p from 36% to 38%. This finding implies that modelling haplotype-specific effects for a substantial portion of the barley gene space may result in an improved prediction of flowering time in HEB-25. However, a genome-wide re-sequencing of HEB-25 lines will be required to identify and distinguish those haplotypes. Applying genomic prediction models To check whether we could further elucidate the genetic architecture of barley flowering time we applied genomic prediction models that considered all markers simultan- eously. Genomic prediction evolved in animal breeding as a tool to predict a phenotype based on modelling a large set of SNP data [50]. It is used for selection of improved Figure 3 Cross-validated proportion of explained genotypic variance (p ) of different applied models. The box-whisker plots depict the genotypes based on estimated genomic breeding values. variation of explained genotypic variance after 100 cross-validations. Applying RR-BLUP [51] and BayesCπ [52] models, we The tested QTL models are (i) thesingleSNP locus Ppd-H1 (Mean could further increase the cross-validated explained p to p =0.36), (ii) GWAS with peak markers, representing the eight 71% and 74%, respectively (Figure 3). These findings are in major QTL indicated in Table 1 (Mean p = 0.64), (iii) the whole agreement with comparisons of multiple linear regression genome ridge regression best linear unbiased prediction (RR-BLUP, mean p = 0.71), (iv) the BayesCπ prediction (Mean p = 0.74), and and genomic prediction of traits in bi-parental plant popu- G G (v) RR-BLUP including epistasis (Mean p = 0.77). G lations [53]. However, our p values substantially exceed the prediction accuracies of genomic prediction models reported in comparable studies [54-56], underlining the order to adapt flowering time to environmental require- tremendous predictive power of HEB-25. Interestingly, ments and to enhance biodiversity in the elite barley compared to GWAS, only a few additional loci had non- breeding pool. zero effects in the BayesCπ model, indicating that flower- ing time is indeed mainly controlled by the eight major loci detected via GWAS. Ppd-H1 haplotype study We assume that important reasons for the slightly As we used bi-allelic SNP markers, additive effects were higher explained p of genomic prediction compared to estimated across the NAM population. Theoretically, GWAS are that minor QTL effects and marginally exist- there may be up to 26 different alleles present at each ing genetic relatedness [55,57] among HEB lines may be locus in HEB-25. Thus, distinct alleles that show con- better modeled in the first case. Furthermore, modeling trasting effects between families potentially escaped de- all makers simultaneously enables a better prediction of tection in our SNP-based GWAS. Contrasting effects are flowering time due to the estimation of family-specific illustrated in Figure 4 and, in detail, in Additional file 6. QTL effects. This is indicated by the occurrence of op- For instance, SNPs at position 46.2 cM on chromosome posing additive effects between HEB families alongside 3H, which are tightly linked to HvGI [49], revealed tightly linked SNPs (Figure 2 and Additional file 6). opposing effects across HEB families. We tested the po- tential to integrate SNP haplotypes in the GWAS model A model including epistasis to maximize the cross-validated for Ppd-H1, which exhibited the largest p . After re- explained p G G sequencing the last two exons and three introns of Ppd- A final increase of the cross-validated explained p to an H1, twelve haplotypes could be distinguished (Additional extraordinary high level of 77% was achieved by including file 1). All Hsp donor haplotypes at Ppd-H1 showed a sig- di-genic epistatic interactions between significant main ef- nificantly reduced flowering time (Additional file 8 and fect SNPs in the RR-BLUP model. This finding indicates Figure 5), where a maximum reduction of flowering time that epistasis explains a portion of the ‘missing heritability’ was associated with H-6 (−11.1 days compared to elite [58] of flowering time regulation in barley, whereas in barley haplotype H-2). Only the Hag haplotype H-45 did maize it does not [28]. The term ‘missing heritability’ is not differ from H-2. This finding confirms the presence of highly debated in quantitative genetics and refers to the haplotype-specific effects present in HEB-25. Conse- observation that the explained genotypic variance of com- quently, we expect the existence of further haplotype ef- bined marker effects is usually lower than the heritability fects for other candidate genes controlling flowering of the trait. Epistatic interactions between candidate genes time. The haplotype-based Ppd-H1 model resulted in a may point to functional relationships and genetic Maurer et al. BMC Genomics (2015) 16:290 Page 6 of 12 Figure 4 Visualization of family-wise SNP effects. Barley chromosomes are indicated as inner circle of colored bars, centromeres are highlighted as transparent boxes. Grey connector lines represent the genetic position of SNPs on chromosomes. Each track displays one HEB family (F01 – F25, from inside to outside). The heatmap indicates the difference in days between the donor and Barke genotype. Blue and red colors specify early and late flowering, respectively, caused by the donor genotype. White color indicates no SNP effect or SNPs monomorphic in the respective family. Candidate genes (Table 1) are indicated outside the circle. Black frames highlight their family-specific effects as indicated in Additional file 6. networks [59]. Our findings indicate that the flowering spring type barley cultivar that lacks the vernalization re- time genes HvGI, Vrn-H2, Vrn-H1 and HvCO1 [60] on quirement due to a deletion of Vrn-H2 [19]. Hence, our chromosomes 3H, 4H, 5H and 7H, respectively, are prob- findings may indicate that the epistatic interaction found ably major players of di-genic epistatic interactions in between the two regions on chromosomes 4H and 5H is HEB-25. All four genes potentially interact with each based on the presence (exotic allele) or absence (Barke other as well as with further genes on additional chromo- allele) of Vrn-H2,the target of Vrn-H1. In general, the somes (Figure 2 and Additional file 9). These observations epistatic interactions detected in HEB-25 may provide are in agreement with independent studies in barley and hints for the presence of so far unknown functional A. thaliana where these interacting genes were placed in a networks of genes, which assist in fine-tuning flowering day length and temperature depending signaling network time in barley. Studies with knock out lines of these genes that controls flowering time [14,21-23]. It is, thus, likely may be used to validate the observed interaction effects. that the observed interaction between the chromosomal regions in HEB-25 may be a function of the mentioned Conclusions flowering time genes. As an example we refer to the po- The first barley NAM population HEB-25 provides great tential interaction found between Vrn-H1 and Vrn-H2. opportunities for future research and breeding. The genetic Epistatic interactions between these loci were already constitution of HEB-25 allows to carry out detailed studies reported [17,61,62] and support the model that Vrn-H2 is on the genetic architecture of important agronomic traits, a long-day suppressor of flowering, that is itself sup- as exemplified by flowering time. The present study sub- pressed by Vrn-H1 following vernalization [63]. Barke is a stantiated that flowering time in barley is primarily Maurer et al. BMC Genomics (2015) 16:290 Page 7 of 12 originating from Tibet, China, was classified as Hordeum vulgare ssp. agriocrithon (Åberg). F plants of the initial crosses were backcrossed with Barke as the female parent. Twenty BC plants per cross were subsequently selfed three times, using the single seed descent (SSD) technique to generate the next generations. The resulting BC S gener- 1 3 ation consists of 1,420 individual lines, classified in 25 HEB families with 22 to 75 individual lines per family (Additional file 1). Subsequently, each HEB line was bulk propagated until BC S to achieve sufficient seed numbers for field 1 3:6 testing. No artificial selection was carried out during the de- velopment of HEB-25. Collecting single nucleotide polymorphism (SNP) data SNP genotype data were collected at TraitGenetics, Figure 5 Box-whisker plots of flowering time BLUEs for Ppd-H1 Gatersleben, Germany, for all 1,420 individual BC S lines 1 3 haplotypes. Green box-whisker-plots display the distribution of and their corresponding parents with the barley Infinium flowering time BLUEs of all HEB lines carrying the respective haplotype. Horizontal lines and diamonds indicate median and mean, respectively, iSelect 9k chip consisting of 7,864 SNPs [35]. At each locus, for each haplotype. The extension of vertical lines indicates minimum three genotypes were differentiated, with an expected and maximum observations, excluding outliers, which are indicated as BC S segregation ratio of 0.71875 : 0.0625 : 0.21875 1 3 circles. The red dotted horizontal line indicates the BLUE of cultivar for homozygous recipient (i.e. Barke), heterozygous and Barke (68.2 days). H-2 represents the haplotype of the Barke genotype homozygous donor genotypes, respectively. In total, present in HEB lines. All haplotypes except H-45 differ significantly (P < 0.05) from H-2, as indicated by red asterisks. Further information to 1,027 monomorphic SNPs and 1,128 SNPs with high haplotypes is given in Additional files 1 and 8. failure rates (i.e. no call in >10% of HEB lines) were ex- cluded from the dataset, resulting in 5,709 informative SNPs for further analyses. determined by large-effect QTL and epistatic interactions. This finding is in contrast to flowering time regulation in Extraction of genomic DNA the allogamous species maize and sorghum, where only DNA was extracted from leaf tissue of 1,420 single small effect minor QTL were detected [28,42], indicating founder HEB plants in generation BC S . The subse- 1 3 that the mating system may control the genetic architec- quent seed propagation of HEB lines was based on these ture of adaptive traits [28]. founder HEB plants. For Barke and the wild barley ac- In future, the NAM population HEB-25 will be utilized cessions leaf material from three to four plants was used in two directions: On the one hand, HEB-25 may support to create bulked samples. The plants were cultivated in a elucidating the genetic architecture of quantitatively inher- glasshouse and 50 to 100 mg of leaf material was harvested ited agronomic traits, ultimately resulting in cloning of yet for each sample. DNA was extracted according to the man- unknown causal genes. On the other hand, HEB-25 will be ufacturer’s protocol, using the BioSprint 96 DNA Plant Kit exploited by breeders to enhance biodiversity of the elite and a BioSprint work station (Qiagen, Hilden, Germany), barley gene pool. This may occur through introgression of and finally dissolved in distilled water at approximately favorable wild alleles with the aim to sustainably increase 50 ng/μl. yield and stress tolerances against disadvantageous climate conditions like drought, heat and salt. SNP mapping The chromosomal positions of 3,391 out of 5,709 SNPs Methods were taken from Comadran et al. [35]. The remaining Development of the NAM population SNPs were fitted next to the mapped SNPs applying chi- The development of the NAM population ‘Halle Exotic square tests of independence. Each non-mapped SNP Barley 25’ (HEB-25) was initiated in 2007 conducting was compared to each mapped SNP based on genotype crosses between the spring barley cultivar Barke (Hordeum segregation across all HEB lines. If two SNPs segregated vulgare ssp. vulgare)and 25 highly divergentexoticwild completely independent from each other, i.e. in case of barley accessions. The latter were used as pollen donors. no linkage disequilibrium (LD), one expects to find all Twenty-four accessions, originating from Afghanistan, Iran, possible genotype combinations according to the product Iraq, Israel, Lebanon, Turkey, and Syria (Hordeum vulgare of their single locus genotype frequencies. However, in case ssp. spontaneum), were selected to maximize the genetic di- of tight linkage, there should be a significant deviation from versity in HEB-25. One further accession, HID380, the expected genotype combination frequencies due to Maurer et al. BMC Genomics (2015) 16:290 Page 8 of 12 reduced recombination between these markers. Conse- between SNP pairs against their genetic distance. A quently, a high chi-square statistic and a low P-value likely second-degree smoothed loess curve [69] was fitted in indicate a tight linkage. Therefore, we assigned the position SAS with Proc Loess. The population-specific baseline r of the SNP with the lowest P-value (minimum: P < 0.001) was defined as the 95% percentile of the distribution of r to the non-mapped SNP under investigation. If there were for unlinked markers [70]. LD decay was defined as the more than one SNP with the same P-value, the position of distance, at which this baseline crossed the loess curve. the unmapped SNP was defined as the average of the mini- mum and the maximum position of the respective markers. Ppd-H1 haplotype definition In this way, all except six of the non-mapped SNPs were For sequencing of the Ppd-H1 locus on chromosome placed into the Comadran map. 2HS we used the following primers: PP05 (forward) 5′-GTGCAAAGCATAATATCAGTGTCC-3′ and PP04 SNP calling (reverse) 5′-GGCCAAAGACACAAGAATCAG-3′. These The differentiation of the HEB genotypes was based on an primers amplify the last two exons and three introns of identity-by-state approach. Based on parental genotype in- Ppd-H1 covering the CCT domain that contains SNP22, formation, the exotic allele could be specified in each segre- the causal SNP of Ppd-H1 [15]. Identical sequences were gating family. Thus, HEB lines that showed a homozygous grouped into haplotypes. A detailed description of the exotic genotype were assigned a value of 2 and HEB lines sequencing is given in Jakob et al. [71]. that showed a homozygous Barke genotype were assigned a value of 0. Consequently, heterozygous HEB lines were assigned a value of 1. If a SNP was monomorphic in one HEB-25 field trials HEB family but polymorphic in a second family, lines of Between 2011 and 2013, three field trials were conducted thefirst HEB familywereassignedagenotype value of0 to at the ‘Kühnfeld Experimental Station’ of the University of keep a full genotype data set, which is a pre-requisite for Halle to gather phenotype data on flowering time. In the subsequent multiple regression analysis. For the same 2011, the field trial was conducted with selfed progenies reason, missing genotypes were estimated applying the of BC S lines (so-called BC S ). Sowing occurred in sin- 1 3 1 3:4 mean imputation (MNI) approach [64]. For this, each miss- gle to five row plots with a length of 1.50 m and a distance ing SNP value was replaced with the mean of the non- of 0.20 m between rows. The number of rows per HEB missing values of that SNP in the respective HEB family. line and the position inside the field trial depended on the Quantitative SNP genotypes were subsequently used for number of available BC S seeds. Lines with seed num- 1 3:4 multiple regression analysis. bers lower than ten were planted in plots with a length of 0.50 m. In 2012 and 2013, the field trials were conducted Evaluation of genetic diversity with the selfed progenies in BC S and BC S , respect- 1 3:5 1 3:6 SAS 9.4 Software (SAS Institute Inc., Cary, NC, USA) was ively. Two replications per HEB line, arranged in two used to evaluate genetic diversity among parents and pro- randomized complete blocks, were cultivated in 2012 and genies of the HEB-25 population. Genetic similarities (GS) 2013. The plots consisted of two rows (30 seeds each) with between HEB lines and their parents and among HEB a length of 1.50 m and a distance of 0.20 m between rows. lines were calculated with Proc Distance, based on a sim- All field trials were sown in spring between March and ple matching comparison between the three possible April with fertilization and pest management following genotype states across all informative SNPs. In addition, local practice. we performed principal component analysis (PCA) using R [65]. First we applied PCA for the 26 parents (the culti- var Barke and 25 wild donors) based on the SNP matrix. Phenotypic data The first two PCs explained 51.9 and 4.8% of the variation. The occurrence of flowering time was recorded as days Then, all progenies of HEB-25 were projected to the space after sowing, when the first awns were visible (BBCH49 spanned by the two PCs (Additional file 3) as outlined in [72]) for 50% of all plants of a plot. We performed a one- detail elsewhere [66]. step phenotypic data analysis with SAS, using a linear mixed model with effects for genotype (i.e. 1,420 HEB Linkage disequilibrium (LD) lines), environment (i.e. 3 years) and interaction of geno- LD was calculated as r [67] between all mapped SNPs type and environment. To estimate variance components, with the software package TASSEL [68]. For this pur- all effects were assumed to be random. Broad-sense herit- pose, heterozygous genotypes and SNPs with a minor ability (h ) was estimated on an entry-mean basis. Best lin- allele frequency < 0.05 were excluded. LD was calcu- ear unbiased estimates (BLUEs) of flowering time were lated across the 26 parents of HEB-25. LD decay across calculated for each genotype assuming fixed genotype intra-chromosomal SNPs was displayed by plotting r effects. Maurer et al. BMC Genomics (2015) 16:290 Page 9 of 12 Genome-wide association study (GWAS) corresponding design matrix and e is the residual term. For GWAS, we applied Model-B as outlined in detail by 2 In the model we assumed that g e N 0; σ , e e N Liu et al. [38]. This model was found most suitable to 2 2 2 2 0; σ , where σ ¼ σ =m for SNP markers and σ ¼ e g G e carry out GWAS with multiple families [39]. It is based on 2 2 2 σ =l . Here σ and σ are the genotypic and residual multiple regression considering an SNP effect and a family R G R variance components obtained in the mixed model in effect in addition to cofactors, which control both popula- the phenotypic data analysis. The penalty parameter is tion structure and genetic background [39]. Cofactor se- 2 2 lection was carried out by applying Proc Glmselect in SAS λ ¼ σ =l = σ =m . The estimation of marker effects is R G and minimizing the Schwarz Bayesian Criterion [73]. then given by the mixed model equations [76]. The genome-wide scan for presence of marker-trait as- The basic model of BayesC π is the same as RR-BLUP. sociations was implemented in the statistical software R However, all parameters are treated as random variables [65], excluding cofactors linked closer than 1 cM to the in a Bayesian framework. First, we defined the prior dis- SNP under investigation. The Bonferroni–Holm pro- 2 2 tributions as g e N 0; σ ; e e N 0; σ . The prior of g e cedure [74] was used to adjust marker-trait associations μ is a constant. The prior distribution of σ is assumed for multiple testing. Significant marker main effects were g to be zero with probability π and a scaled inverse chi- accepted with P < 0.05. Additive effects for each BON-HOLM squared distribution with probability (1-π). The prob- SNP were estimated based on regression across but also ability π is a random variable whose prior distribution is within families. Significant marker trait associations were uniform on the interval [0,1]. The prior distribution of grouped to a singled QTL if the significant SNPs were linked by less than 5 cM and revealed additive effects σ is also scaled inverse chi-squared. A Gibbs sampler of the same direction, i.e. both exotic alleles increased or algorithm was then implemented to infer all the parame- decreased flowering time. In addition, a two-dimensional ters in the model. It was run for 10,000 cycles and the epistasis scan was carried out to identify pairwise marker first 1,000 cycles were discarded as burn in. The samples interactions. For this, the GWAS Model-B was extended of g from all later cycles were averaged to obtain esti- to cover a second main SNP effect and the interaction mates of the marker effects. effect between the two SNPs. Cross-validation for additive models Haplotype-based association mapping for Ppd-H1 The accuracy of the prediction of flowering time by A haplotype-based association mapping test was imple- GWAS and the two genomic prediction approaches mented in HEB-25 to test for effects of haplotypes at Ppd- were evaluated using five-fold cross-validations [77]. In H1. We used the same GWAS procedure with cofactor each run of cross-validation, the estimation set in- selection as mentioned before. However, bi-allelic SNPs cluded 80% of HEB lines, randomly selected per HEB covering the region of Ppd-H1 were replaced by a qualita- family, while the remaining 20% of HEB lines were tive variable containing the defined Ppd-H1 haplotype. assigned to build the test set. For GWAS, we per- BLUEs were determined for each haplotype. Subsequently, formed an association mapping scan within the estima- pairwise comparisons between all haplotype BLUEs were tion set and recorded the detected significant markers. performed using the Tukey-Kramer [75] multiple compari- To determine the cross-validated proportion of ex- son test. plained genotypic variance (p ), we estimated the ef- fects of the significant peak markers within the Genomic prediction estimation set and predicted the genotypic value of the Based on BLUEs of the 1,420 HEB genotypes, two ap- lines in the test set [40]. We then calculated the cross- proaches for genomic prediction were applied considering validated p as the squared Pearson product–moment additive effects: ridge regression best linear unbiased pre- correlation between predicted and observed genotypic diction (RR-BLUP [51]) and BayesCπ [52]. All statistical values in the test set standardized with the heritability. procedures for genomic prediction approaches were exe- The mean p in 100 cross-validation runs (20 times five- cuted using R. We briefly describe the two models in the fold cross-validations) was taken as the final record. In following. addition, the number of significances for each SNP was Let n be the number of genotypes, m be the number cumulated across all runs and is referred to as QTL de- of markers and l be the number of environments. The tection rate. RR-BLUP model has the form y =1 μ + Xg + e, where y For genomic prediction we estimated the effects for all is the vector of BLUEs of flowering time scores for all markers using the estimation set and predicted the HEB genotypes across environments, 1 denotes the vec- genotypic value of the lines in the test set. The cross- tor of 1’s, μ is the overall mean, g is the vector of marker validated p was calculated as in GWAS. effects (for SNP markers, allele effects), X is the G Maurer et al. BMC Genomics (2015) 16:290 Page 10 of 12 Exploiting additive and additive times additive epistatic Additional file 4: Distribution of flowering time. Figure showing the effects in genomic prediction frequency distribution of flowering time BLUEs across three field trials and illustrating contrasting phenotypes in the field. We extended the RR-BLUP based on main effects Additional file 5: Phenotype and genotype data for HEB-25. Table to model also epistasis for markers with significant listing the complete phenotype and genotype data of HEB-25 underlying main effects in the GWAS. The model is y ¼ 1 μþ this study as well as marker information. X X X g þ X ⋅X f þ e,where y is the vector of i j l Additional file 6: Estimates of single marker GWAS and genomic jl i¼1 j<l prediction effects across HEB-25 and within individual HEB families. BLUEs of flowering time for all HEB genotypes, 1 denotes Table listing the results of GWAS and genomic prediction across HEB-25 the vector of 1’s, μ is the overall mean, g is the main and within individual HEB families. additive effect of the i-th marker, X is the vector of marker Additional file 7: GWAS Manhattan plot for flowering time. Figure displaying the GWAS results through plotting the significance and effects indices, f is the epistatic effects of the j- and the l-th jl of markers in a Manhattan plot. marker, X X is the point-wise product of the two vectors j l Additional file 8: Ppd-H1 haplotype comparison. Two tables X and X,and e is the vector of residual terms. Note that j l contrasting the different Ppd-H1 haplotype effects by comparison of their in the third term of the right hand side of the formula, the BLUEs. sum is not taken over all pairs of markers but only pairs of Additional file 9: Significant epistatic interactions via GWAS. Table listing all significant (P < 0.05) epistatic interactions between BON-HOLM markers exhibiting a significant additive effect in the SNPs that were obtained via GWAS. GWAS study performed previously. Hence, in different cross-validation runs, different pairs of markers were con- Abbreviations sidered in the model. The model assumptions are similar BC : Progeny of F after back-crossing with Barke; BC S : Progeny of BC 1 1 1 3 1 to the usual RR-BLUP, except treating additive and epistatic after three rounds of selfing; BC S :Progeny of BC S individual after x-3 1 3:x 1 3 rounds of bulk propagation; BLUE: Best linear unbiased estimate; effects separately. We assumed g e N 0; σ , f e N i g jl cM: Centimorgan; F : First generation after initial crosses; GA: Gibberellic acid; GS: Genetic similarity; GWAS: Genome-wide association study; 2 2 2 2 2 0; σ ,where σ ¼ p σ =m, σ ¼ðÞ 1−p σ =p.Here P f g G G f G G Hag: Hordeum agriocrithon; HEB: Halle exotic barley; HID: Hordeum identity; Hsp: Hordeum spontaneum; Hv: Hordeum vulgare;LD: Linkage is the cross-validated proportion of explained genotypic disequilibrium; MNI: Mean imputation; NAM: Nested association mapping; variance for genomic prediction, obtained previously by PC: Principal component; PCA: Principal component analysis; p : Proportion of explained genotypic variance; QTL: Quantitative trait the RR-BLUP, only considering additive effects, m is the G locus/loci; RR-BLUP: Ridge regression best linear unbiased prediction; number of markers, p is the number of pairs of markers SNP: Single nucleotide polymorphism; SSD: Single seed descent. having significant additive effect. Therefore the penalty parameter λ is different for additive and epistatic effects. Competing interests The authors declare that they have no competing interests. Using the above extended model, for each cross-validation run we estimated the additive effects of all markers and Authors’ contributions epistatic effects of all pairs of markers exhibiting sig- AM planned and conducted the field trials in 2012 and 2013, analyzed the nificant additive effects in GWAS using the estimation genotype and phenotype data, carried out the GWAS, created the figures, and drafted the manuscript. VD was involved in the development of the HEB-25 set. Then we predicted genotypic values of the lines in population, gathered the genotype data, and planned and conducted the field the test set and calculated the p in thesamewayas trial in 2011. YJ performed the genomic prediction and cross-validation outlined above. approaches. FS was involved in the development of the HEB-25 population. RS performed the re-sequencing of Ppd-H1. ES supervised the field trials. BK supervised the Ppd-H1 re-sequencing. JCR supervised the genomic prediction Availability of supporting data and cross-validation approaches. KP acquired the funding, supervised the development of the HEB-25 population and all analyses, and drafted the Raw data, including data on SNPs, Ppd-H1 haplotypes manuscript. All authors read and approved the final manuscript. and GWAS, and all other supporting data are provided as additional files. Acknowledgements This work was supported by Deutsche Forschungsgemeinschaft (DFG), Bonn (projects Pi339/7-1 and Pi 339/7-2), and the Interdisciplinary Centre for Crop Additional files Plant Research (IZN), Halle. We are grateful to Roswitha Ende, Jana Müglitz, Diana Rarisch, Helga Sängerlaub, Brigitte Schröder, Bernd Kollmorgen and Additional file 1: Genetic constitution of HEB-25: classification of various student assistants for excellent technical help and to TraitGenetics families and donors. Tabular overview of the genetic constitution of GmbH, Gatersleben, Germany, for genotyping HEB-25 with the Infinium HEB-25, classifying the 25 families and donors and containing the Ppd-H1 iSelect 9k SNP chip. haplotypes. Additional file 2: LD decay of intra-chromosomal markers among Author details HEB-25 parents. Figure showing the LD decay of intra-chromosomal Institute of Agricultural and Nutritional Sciences, Martin Luther University markers among HEB-25 parents by plotting r against the genetic marker Halle Wittenberg, Betty-Heimann-Str. 3, 06120 Halle, Germany. distance. Interdisciplinary Center for Crop Plant Research (IZN), Betty-Heimann-Str. 3, 06120 Halle, Germany. Leibniz-Institute of Plant Genetics and Crop Plant Additional file 3: Principal component analysis for HEB-25 and its Research (IPK), Corrensstr. 3, 06466 Stadt Seeland, OT Gatersleben, Germany. parents. Figure showing the relatedness of HEB-25 lines by plotting of Current address: University of Dundee at the James Hutton Institute, the first two principal components of a principal component Invergowrie, Dundee DD2 5DA, UK. Current address: Bayer CropScience NV, analysis for HEB-25 and its parents. Technologiepark 38, 9052 Ghent, Belgium. Maurer et al. BMC Genomics (2015) 16:290 Page 11 of 12 Received: 12 October 2014 Accepted: 9 March 2015 25. Nevo E, Fu Y-B, Pavlicek T, Khalifa S, Tavasi M, Beiles A. Evolution of wild cereals during 28 years of global warming in Israel. Proc Natl Acad Sci U S A. 2012;109:3412–5. 26. Yu J, Holland JB, McMullen MD, Buckler ES. Genetic design and statistical power of nested association mapping in maize. Genetics. 2008;178:539–51. References 27. Jordan D, Mace E, Cruickshank A, Hunt C, Henzell R. Exploring and 1. Sakuma S, Salomon B, Komatsuda T. The domestication syndrome genes exploiting genetic variation from unadapted sorghum germplasm in a responsible for the major changes in plant form in the Triticeae crops. breeding program. Crop Sci. 2011;51:1444–57. Plant Cell Physiol. 2011;52:738–49. 28. Buckler ES, Holland JB, Bradbury PJ, Acharya CB, Brown PJ, Browne C, et al. 2. Zohary D, Hopf M, Weiss E. Domestication of Plants in the Old World: The The Genetic Architecture of Maize Flowering Time. Science. 2009;325:714–8. origin and spread of domesticated plants in Southwest Asia, Europe, and 29. Kump KL, Bradbury PJ, Wisser RJ, Buckler ES, Belcher AR, Oropeza-Rosas MA, the Mediterranean Basin. 4th ed. Oxford: Oxford University Press; 2012. et al. Genome-wide association study of quantitative resistance to southern 3. Morrell PL, Clegg MT. Genetic evidence for a second domestication of leaf blight in the maize nested association mapping population. Nat Genet. barley (Hordeum vulgare) east of the Fertile Crescent. Proc Natl Acad Sci U 2011;43:163–8. S A. 2007;104:3289–94. 30. Poland JA, Bradbury PJ, Buckler ES, Nelson RJ. Genome-wide nested association 4. Dai F, Nevo E, Wu D, Comadran J, Zhou M, Qiu L, et al. Tibet is one of the mapping of quantitative resistance to northern leaf blight in maize. Proc Natl centers of domestication of cultivated barley. Proc Natl Acad Sci U S A. Acad Sci U S A. 2011;108:6893–8. 2012;109:16969–73. 31. Tian F, Bradbury PJ, Brown PJ, Hung H, Sun Q, Flint-Garcia S, et al. Genome-wide 5. Tanksley SD, McCouch SR. Seed Banks and Molecular Maps: Unlocking association study of leaf architecture in the maize nested association mapping Genetic Potential from the Wild. Science. 1997;277:1063–6. population. Nat Genet. 2011;43:159–62. 6. Zamir D. Improving plant breeding with exotic genetic libraries. Nat Rev 32. Cook JP, McMullen MD, Holland JB, Tian F, Bradbury P, Ross-Ibarra J, et al. Genet. 2001;2:983–9. Genetic Architecture of Maize Kernel Composition in the Nested Association 7. Ellis RP, Forster BP, Robinson D, Handley LL, Gordon DC, Russell JR, et al. Mapping and Inbred Association Panels. Plant Physiol. 2012;158:824–34. Wild barley: a source of genes for crop improvement in the 21st century? 33. Peiffer JA, Romay MC, Gore MA, Flint-Garcia SA, Zhang Z, Millard MJ, et al. J Exp Bot. 2000;51:9–17. The genetic architecture of maize height. Genetics. 2014;196:1337–56. 8. Pillen K, Zacharias A, Leon J. Advanced backcross QTL analysis in barley 34. Wallace J, Larsson S, Buckler E. Entering the second century of maize (Hordeum vulgare L.). Theor Appl Genet. 2003;107:340–52. quantitative genetics. Heredity (Edinb). 2014;112:30–8. 9. Wang G, Schmalenbach I, von Korff M, Leon J, Kilian B, Rode J, et al. 35. Comadran J, Kilian B, Russell J, Ramsay L, Stein N, Ganal M, et al. Natural Association of barley photoperiod and vernalization genes with QTLs for variation in a homolog of Antirrhinum CENTRORADIALIS contributed to flowering time and agronomic traits in a BC(2)DH population and a set of spring growth habit and environmental adaptation in cultivated barley. wild barley introgression lines. Theor Appl Genet. 2010;120:1559–74. Nat Genet. 2012;44:1388–92. 10. Schmalenbach I, March TJ, Bringezu T, Waugh R, Pillen K. High-Resolution 36. Gottwald S, Bauer P, Komatsuda T, Lundqvist U, Stein N. TILLING in the Genotyping of Wild Barley Introgression Lines and Fine-Mapping of the two-rowed barley cultivar ‘Barke’ reveals preferred sites of functional Threshability Locus thresh-1 Using the Illumina GoldenGate Assay. G3 diversity in the gene HvHox1. BMC Res Notes. 2009;2:258. (Bethesda). 2011;1:187–96. 11. Ma X, Li C, Wang A, Duan R, Jiao G, Nevo E, et al. Genetic diversity of wild 37. Badr A, Muller K, Schafer-Pregl R, El Rabey H, Effgen S, Ibrahim HH, et al. barley (Hordeum vulgare ssp. spontaneum) and its utilization for barley On the origin and domestication history of barley (Hordeum vulgare). improvement. Sci Cold Arid Reg. 2012;4:453–61. Mol Biol Evol. 2000;17:499–510. 38. Liu W, Gowda M, Steinhoff J, Maurer HP, Würschum T, Longin CFH, et al. 12. McCouch S, Baute GJ, Bradeen J, Bramel P, Bretting PK, Buckler E, et al. Association mapping in an elite maize breeding population. Theor Appl Agriculture: Feeding the future. Nature. 2013;499:23–4. Genet. 2011;123:847–58. 13. Schnaithmann F, Kopahnke D, Pillen K. A first step toward the development 39. Würschum T, Liu W, Gowda M, Maurer H, Fischer S, Schechert A, et al. of a barley NAM population and its utilization to detect QTLs conferring leaf Comparison of biometrical models for joint linkage association mapping. rust seedling resistance. Theor Appl Genet. 2014;127:1513–25. Heredity (Edinb). 2012;108:332–40. 14. Cockram J, Jones H, Leigh FJ, O’Sullivan D, Powell W, Laurie DA, et al. 40. Utz HF, Melchinger AE, Schön CC. Bias and sampling error of the estimated Control of flowering time in temperate cereals: genes, domestication, and proportion of genotypic variance explained by quantitative trait loci sustainable productivity. J Exp Bot. 2007;58:1231–44. determined from experimental data in maize using cross validation and 15. Turner A, Beales J, Faure S, Dunford RP, Laurie DA. The pseudo-response validation with independent samples. Genetics. 2000;154:1839–49. regulator Ppd-H1 provides adaptation to photoperiod in barley. Science. 2005;310:1031–4. 41. Gowda M, Zhao Y, Würschum T, Longin CF, Miedaner T, Ebmeyer E, et al. 16. Yan L, Fu D, Li C, Blechl A, Tranquilli G, Bonafede M, et al. The wheat and Relatedness severely impacts accuracy of marker-assisted selection for barley vernalization gene VRN3 is an orthologue of FT. Proc Natl Acad Sci U disease resistance in hybrid wheat. Heredity (Edinb). 2014;112:552–61. S A. 2006;103:19581–6. 42. Mace E, Hunt C, Jordan D. Supermodels: sorghum and maize provide 17. Yan L, Loukoianov A, Blechl A, Tranquilli G, Ramakrishna W, SanMiguel P, mutual insight into the genetics of flowering time. Theor Appl Genet. et al. The wheat VRN2 gene is a flowering repressor down-regulated by 2013;126:1377–95. vernalization. Science. 2004;303:1640–4. 43. Loscos J, Igartua E, Contreras-Moreira B, Gracia MP, Casas AM. HvFT1 polymorphism and effect – survey of barley germplasm and expression analysis. Front Plant Sci. 18. Yan L, Loukoianov A, Tranquilli G, Helguera M, Fahima T, Dubcovsky J. 2014;5:251. Positional cloning of the wheat vernalization gene VRN1. Proc Natl Acad Sci U S A. 2003;100:6263–8. 44. Mutasa-Göttgens E, Hedden P. Gibberellin as a factor in floral regulatory 19. von Zitzewitz J, Szűcs P, Dubcovsky J, Yan L, Francia E, Pecchioni N, et al. networks. J Exp Bot. 2009;60:1979–89. Molecular and structural characterization of barley vernalization genes. 45. Jia Q, Zhang J, Westcott S, Zhang X-Q, Bellgard M, Lance R, et al. GA-20 oxidase Plant Mol Biol. 2005;59:449–67. as a candidate for the semidwarf gene sdw1/denso in barley. Funct Integr 20. Laurie D, Pratchett N, Snape J, Bezant J. RFLP mapping of five major genes Genomics. 2009;9:255–62. and eight quantitative trait loci controlling flowering time in a winter × 46. Faure S, Turner AS, Gruszka D, Christodoulou V, Davis SJ, von Korff M, et al. spring barley (Hordeum vulgare L.) cross. Genome. 1995;38:575–85. Mutation at the circadian clock gene EARLY MATURITY 8 adapts domesticated 21. Distelfeld A, Li C, Dubcovsky J. Regulation of flowering in temperate cereals. barley (Hordeum vulgare) to short growing seasons. Proc Natl Acad Sci U S A. Curr Opin Plant Biol. 2009;12:178–84. 2012;109:8328–33. 22. Jung C, Mueller AE. Flowering time control and applications in plant 47. Zakhrabekova S, Gough SP, Braumann I, Muller AH, Lundqvist J, Ahmann K, et al. breeding. Trends Plant Sci. 2009;14:563–73. Induced mutations in circadian clock regulator Mat-a facilitated short-season 23. Milec Z, Valárik M, Bartoš J, Šafář J. Can a late bloomer become an early adaptation and range extension in cultivated barley. Proc Natl Acad Sci U S A. bird? Tools for flowering time adjustment. Biotechnol Adv. 2014;32:200–14. 2012;109:4326–31. 24. Cockram J, Hones H, O’Sullivan DM. Genetic variation at flowering time loci 48. Boden SA, Weiss D, Ross JJ, Davies NW, Trevaskis B, Chandler PM, et al. in wild and cultivated barley. Plant Genet Resour. 2011;9:264–7. EARLY FLOWERING3 Regulates Flowering in Spring Barley by Mediating Maurer et al. BMC Genomics (2015) 16:290 Page 12 of 12 Gibberellin Production and FLOWERING LOCUS T Expression. Plant Cell. 75. Kramer CY. Extension of multiple range tests to group means with unequal 2014;26:1557–69. numbers of replications. Biometrics. 1956;12:307–10. 49. Dunford RP, Griffiths S, Christodoulou V, Laurie DA. Characterisation of a 76. Henderson CR. Applications of Linear Models in Animal Breeding. Guelph: barley (Hordeum vulgare L.) homologue of the Arabidopsis flowering time University of Guelph; 1984. regulator GIGANTEA. Theor Appl Genet. 2005;110:925–31. 77. Hjorth JU. Computer intensive statistical methods: Validation, model 50. Meuwissen THE, Hayes BJ, Goddard ME. Prediction of Total Genetic Value selection, and bootstrap. London: Chapman & Hall/CRC; 1993. Using Genome-Wide Dense Marker Maps. Genetics. 2001;157:1819–29. 51. Whittaker JC, Thompson R, Denham MC. Marker-assisted selection using ridge regression. Genet Res. 2000;75:249–52. 52. Habier D, Fernando RL, Kizilkaya K, Garrick DJ. Extension of the Bayesian alphabet for genomic selection. BMC Bioinf. 2011;12:186. 53. Lorenzana RE, Bernardo R. Accuracy of genotypic value predictions for marker-based selection in biparental plant populations. Theor Appl Genet. 2009;120:151–61. 54. Guo ZG, Tucker DM, Lu JW, Kishore V, Gay G. Evaluation of genome-wide selection efficiency in maize nested association mapping populations. Theor Appl Genet. 2012;124:261–75. 55. Guo Z, Tucker DM, Basten CJ, Gandhi H, Ersoz E, Guo B, et al. The impact of population structure on genomic prediction in stratified populations. Theor Appl Genet. 2014;127:749–62. 56. Zhao Y, Mette MF, Gowda M, Longin CFH, Reif JC. Bridging the gap between marker-assisted and genomic selection of heading time and plant height in hybrid wheat. Heredity (Edinb). 2014;112:638–45. 57. Habier D, Fernando R, Dekkers J. The impact of genetic relationship information on genome-assisted breeding values. Genetics. 2007;177:2389–97. 58. Mackay TF. Epistasis and quantitative traits: using model organisms to study gene-gene interactions. Nat Rev Genet. 2013;15:22–33. 59. von Korff M, Léon J, Pillen K. Detection of epistatic interactions between exotic alleles introgressed from wild barley (H. vulgare ssp. spontaneum). Theor Appl Genet. 2010;121:1455–64. 60. Griffiths S, Dunford RP, Coupland G, Laurie DA. The evolution of CONSTANS-like gene families in barley, rice, and Arabidopsis. Plant Physiol. 2003;131:1855–67. 61. Tranquilli G, Dubcovsky J. Epistatic interaction between vernalization genes Vrn-Am1 and Vrn-Am2 in diploid wheat. J Hered. 2000;91:304–6. 62. Szűcs P, Skinner JS, Karsai I, Cuesta-Marcos A, Haggard KG, Corey AE, et al. Validation of the VRN-H2/VRN-H1 epistatic model in barley reveals that intron length variation in VRN-H1 may account for a continuum of vernalization sensitivity. Mol Genet Genomics. 2007;277:249–61. 63. Chen A, Dubcovsky J. Wheat TILLING mutants show that the vernalization gene VRN1 down-regulates the flowering repressor VRN2 in leaves but is not essential for flowering. PLoS Genet. 2012;8:e1003134. 64. Rutkoski JE, Poland J, Jannink J-L, Sorrells ME. Imputation of Unordered Markers and the Impact on Genomic Selection Accuracy. G3 (Bethesda). 2013;3:427–39. 65. R Development Core Team (2010) R: A language and environment for statistical computing, R Foundation for Statistical Computing. http://www.R-project.org 66. Xia X, Reif J, Melchinger A, Frisch M, Hoisington D, Beck D, et al. Genetic diversity among CIMMYT maize inbred lines investigated with SSR markers. Crop Sci. 2005;45:2573–82. 67. Weir BS. Genetic Data Analysis II. Sunderland, MA: Sinauer Associates, Inc.; 1996. 68. Bradbury PJ, Zhang Z, Kroon DE, Casstevens TM, Ramdoss Y, Buckler ES. TASSEL: software for association mapping of complex traits in diverse samples. Bioinformatics. 2007;23:2633–5. 69. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc. 1979;74:829–36. 70. Breseghello F, Sorrells ME. Association mapping of kernel size and milling quality in wheat (Triticum aestivum L.) cultivars. Genetics. 2006;172:1165–77. 71. Jakob SS, Rödder D, Engler JO, Shaaf S, Özkan H, Blattner FR, et al. Submit your next manuscript to BioMed Central Evolutionary History of Wild Barley (Hordeum vulgare subsp. spontaneum) and take full advantage of: Analyzed Using Multilocus Sequence Data and Paleodistribution Modeling. Genome Biol Evol. 2014;6:685–702. • Convenient online submission 72. Lancashire PD, Bleiholder H, Boom TVD, Langelüddeke P, Stauss R, Weber E, • Thorough peer review et al. A uniform decimal code for growth stages of crops and weeds. Ann Appl Biol. 1991;119:561–601. • No space constraints or color figure charges 73. Schwarz G. Estimating the dimension of a model. Ann Stat. 1978;6:461–4. • Immediate publication on acceptance 74. Holm S. A simple sequentially rejective multiple test procedure. Scand J • Inclusion in PubMed, CAS, Scopus and Google Scholar Stat. 1979;6:65–70. • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Genomics Springer Journals

Modelling the genetic architecture of flowering time control in barley through nested association mapping

Free
12 pages

Loading next page...
 
/lp/springer_journal/modelling-the-genetic-architecture-of-flowering-time-control-in-barley-KzBqWUOaMm
Publisher
Springer Journals
Copyright
Copyright © 2015 by Maurer et al.; licensee BioMed Central.
Subject
Life Sciences; Life Sciences, general; Microarrays; Proteomics; Animal Genetics and Genomics; Microbial Genetics and Genomics; Plant Genetics and Genomics
eISSN
1471-2164
D.O.I.
10.1186/s12864-015-1459-7
Publisher site
See Article on Publisher Site

Abstract

Background: Barley, globally the fourth most important cereal, provides food and beverages for humans and feed for animal husbandry. Maximizing grain yield under varying climate conditions largely depends on the optimal timing of flowering. Therefore, regulation of flowering time is of extraordinary importance to meet future food and feed demands. We developed the first barley nested association mapping (NAM) population, HEB-25, by crossing 25 wild barleys with one elite barley cultivar, and used it to dissect the genetic architecture of flowering time. Results: Upon cultivation of 1,420 lines in multi-field trials and applying a genome-wide association study, eight major quantitative trait loci (QTL) were identified as main determinants to control flowering time in barley. These QTL accounted for 64% of the cross-validated proportion of explained genotypic variance (p ). The strongest single QTL effect corresponded to the known photoperiod response gene Ppd-H1. After sequencing the causative part of Ppd-H1, we differentiated twelve haplotypes in HEB-25, whereof the strongest exotic haplotype accelerated flowering time by 11 days compared to the elite barley haplotype. Applying a whole genome prediction model including main effects and epistatic interactions allowed predicting flowering time with an unmatched accuracy of 77% of cross-validated p . Conclusions: The elaborated causal models represent a fundamental step to explain flowering time in barley. In addition, our study confirms that the exotic biodiversity present in HEB-25 is a valuable toolbox to dissect the genetic architecture of important agronomic traits and to replenish the elite barley breeding pool with favorable, trait-improving exotic alleles. Keywords: Barley, Wild barley, Nested association mapping (NAM), Flowering time, Genome-wide association study (GWAS), Quantitative trait locus (QTL), Genomic prediction, Epistasis, Haplotypes Background one promising approach to replenish the elite breeding Barley is among the oldest crop species human civilization pool with new favorable alleles [6-13]. The enriched diver- was built on. Approximately 10,500 years ago, barley was sity may be pivotal to boost the rate of genetic improve- domesticated in the Fertile Crescent [1,2], presumably ment and to cope with the anticipated enhanced effects of followed by additional independent domestication events biotic and abiotic stresses due to climate change. in East Asia [3,4]. Domestication and subsequent genetic In this regard, time of flowering is expected to play a selection led to gene erosion in most crop species’ gene major role in future crop improvement. It is a key trait pools [5,6], threatening future breeding advances. Utilizing for the successful completion of a plant’s life cycle and, the untapped biodiversity, present in wild progenitors is therefore, it has a strong impact on grain yield [14]. Flowering time indicates the transition from vegetative * Correspondence: klaus.pillen@landw.uni-halle.de to reproductive stage, which is mainly influenced by en- Equal contributors vironmental cues like day length (photoperiod) and pro- Institute of Agricultural and Nutritional Sciences, Martin Luther University longed exposure to cold temperatures (vernalization). In Halle Wittenberg, Betty-Heimann-Str. 3, 06120 Halle, Germany Full list of author information is available at the end of the article © 2015 Maurer et al.; licensee BioMed Central. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Maurer et al. BMC Genomics (2015) 16:290 Page 2 of 12 barley, the day length determining light signal is trans- Nested association mapping (NAM) emerged as a multi- mitted from a circadian clock oscillator, with Ppd-H1, parental mapping design to investigate genomic regions a PSEUDO-RESPONSE REGULATOR 7 (PRR7) gene, in with unprecedented genetic resolution and allelic variation its center [15]. Under long day condition, Ppd-H1, through by combining the advantages of linkage analysis and associ- mediation of CONSTANS (CO), promotes the expression of ation mapping [26]. Hence, it facilitates the elucidation of a the floral inducer Vrn-H3, a homolog of the Arabidopsis trait’s genetic architecture via genome-wide association thaliana FLOWERING LOCUS T (FT) gene [16]. On the study (GWAS). Until now, the NAM design was applied to other hand, Vrn-H2, azinc-finger CONSTANS, CO-like and the allogamous species maize and sorghum [26,27]. NAM TOC1 (CCT)-domain protein (ZCCT1)acts asarepressor populations for autogamous species like barley or wheat of Vrn-H3 [17]. Vrn-H2,inturn, isrepressed by Vrn-H1, an have not been developed, yet. In maize, the genetic dissec- APETALA1 family MADS-box transcription factor [18], tion of various agronomic traits, including flowering time, which is up-regulated during vernalization. Thus, after has been investigated [28-34]. However, it was not possible vernalization, the repression of Vrn-H3 is abolished to completely dissect the genetic architecture of flowering and flowering is induced. Based on its vernalization time in maize due to its complex inheritance and the multi- requirement, winter barley and spring barley can be dis- tude of involved small effect QTL. We developed the first tinguished. Spring barley lacks the vernalization require- NAM population in the autogamous species barley, termed ment due to a deletion of Vrn-H2 [19]. ‘Halle Exotic Barley 25’ (HEB-25). The population results Besides photoperiod and vernalization, there are also from initial crosses between the spring barley elite cultivar genetic mechanisms acting independently of environmen- Barke (Hordeum vulgare ssp. vulgare, Hv) and 25 highly tal cues, so-called earliness per se [20]. Although several divergent exotic barley accessions, contributing an ideal key regulatory cereal genes of flowering time were charac- instrument to studybiodiversity. Theexoticdonorscom- terized and finally cloned during the last two decades, still prise 24 wild barley accessions of H. vulgare ssp. sponta- little is known about the genetic architecture underlying neum (Hsp), the progenitor of domesticated barley, and one flowering time regulation in temperate cereals, as com- Tibetian H. vulgare ssp. agriocrithon (Hag) accession. Barke pared to the model species A. thaliana [14,21-23]. So far, was selected since it was also used as a parent of a barley a holistic explanation of flowering time in a segregating high-resolution mapping population [35] and as a genetic germplasm population and the accurate prediction of a stock for mutation screening [36]. The exotic donors were plant’s time of flowering, based on the combined action selected from Badr et al. [37] to represent a substantial part and interaction of major genes, is still not fully achieved in of the genetic diversity that is present across the Fertile cereal species. Furthermore, it is reported that wild barley Crescent, where barley domestication occurred. To gener- accessions possess a rich reservoir of beneficial alleles ate the nested population, F plants were backcrossed to controlling flowering time [7,24,25]. Barke and, subsequently, selfed three times (Figure 1). In Figure 1 Development of the nested association mapping population HEB-25. HEB-25 is made of 25 families with 1,420 NAM lines in BC S . Per 1 3 NAM line, one chromosome pair is illustrated as a double bar. Black and colored bars represent chromosome segments originating from Barke and the exotic donor accessions, respectively. At each SNP locus, HEB-25 is expected to segregate into 71.875% homozygous Barke, 6.25% heterozygous and 21.875% homozygous donor genotypes. Maurer et al. BMC Genomics (2015) 16:290 Page 3 of 12 total, HEB-25 consists of 1,420 BC S lines, divided into 25 to collect flowering time data. The HEB lines flowered 1 3 HEB families of up to 75 lines per family (Additional file 1). on average 68.1 days after sowing with a range from 51.0 to In the present study we investigated the genetic archi- 98.9 days and a standard deviation of 6.5 days (Additional tecture of flowering time in barley. For this purpose, the files 4 and 5). The broad variation in flowering time, cover- NAM population HEB-25 was grown from 2011 to 2013 ing almost 50 days among the 1,420 HEB lines, and a high in multi-field trials to gather data on flowering time. By heritability of 91.6%, as well as the genetic properties of the combining these data with high-density SNP marker infor- NAM population provided an excellent starting point to mation via genome-wide association studies and genomic study the genetic architecture of flowering time through prediction models, we could show that flowering time in GWAS. barley mainly depends on a low number of large-effect QTL and epistatic interactions. Genome-wide association study For GWAS, we initially applied the multiple linear re- gression Model-B with step-wise selection of cofactors, Results and discussion as outlined in Liu et al. [38]. Model-B was found most Characterization of HEB-25 suitable to study traits across multiple related families The inheritance of parental segments across the genomes [39], where a family effect and additional SNPs, selected of the 1,420 HEB lines was characterized through genotyp- as cofactors, are included in the model. GWAS revealed ing 5,709 informative, genic single nucleotide polymorph- eight highly significant major QTL regions controlling ism (SNP) markers [35]. Marker saturation was high with flowering time with P <1.0 E-10 (Table 1, BON-HOLM an average genetic distance of 0.17 cM and a maximum Figure 2, Additional files 6 and 7). Testing the combined gap of 11.1 cM between adjacent markers. Linkage disequi- explanatory power of the single peak markers of the eight librium (LD) among the 26 parents decayed rapidly major QTL revealed a cross-validated explained propor- (Additional file 2) enabling a high mapping resolution tion of genotypic variance (p , [40]) of 64% (Figure 3). To [26]. The SNP data revealed a low degree of genetic check if genetic relatedness, as reported elsewhere [41], similarity between Barke and the donors, ranging from affects this parameter in HEB-25, we also estimated p for 40 to 54% (Additional file 1). Parents and the HEB-25 different sets of eight randomly chosen SNPs, excluding population could be clearly separated in a principal regions with significant QTL. However, since the cross- component analysis (PCA) (Additional file 3). Also, validated explained p remained low with an average of HEB families could be ordered in the PCA based on 8%, we conclude that genetic relatedness between individ- their geographical origin. These findings point to the ual lines does not play a major role in HEB-25. This em- high genetic diversity that is present among HEB-25 phasizes the power and precision of QTL detection in and its parents. HEB-25, which may be a combined effect of the low ex- Diversity in HEB-25 was also visible phenotypically. tent of LD and the particular mating design, resulting in During the seasons 2011 through 2013, HEB-25 was cul- an elevated rate of chromosomal recombination. Thus, tivated at the Halle University Experimental Field Station flowering time of barley can be reliably predicted based on Table 1 List of eight major QTL controlling flowering time in HEB-25 a b b c d e f g h i QTL Chr Pos Range Peak marker No. Seg. Fam. P p CV Freq. Effect CG BON-HOLM G QFt.HEB25-1b 1H 128.3 128.0-128.3 SCRI_RS_150786 25 2.41E-18 0.01 68 −1.4 HvELF3 [46,47] QFt.HEB25-2b 2H 23.0 16.8-23.8 BK_16 24 3.39E-130 0.36 100 −9.5 Ppd-H1 [15] QFt.HEB25-2c 2H 57.4 56.4-58.1 BOPA2_12_30265 25 2.25E-42 0.05 84 −3.0 HvCEN [35] QFt.HEB25-3c 3H 108.4 107.8-109.2 BOPA1_ABC07496_ pHv1343_02 23 2.62E-62 0.04 83 −3.1 denso [45] QFt.HEB25-4a 4H 3.5 3.5 BOPA2_12_31458 24 5.08E-15 0.05 82 3.2 QFt.HEB25-4e 4H 113.4 113.4-114.3 SCRI_RS_216897 24 4.58E-17 0.02 100 2.2 Vrn-H2 [17] QFt.HEB25-5d 5H 125.5 125.5-125.8 BOPA1_4795_782 24 2.31E-33 0.06 60 3.8 Vrn-H1 [18] QFt.HEB25-7a 7H 34.3 25.9-34.3 BOPA2_12_30895 23 6.04E-69 0.07 100 4.1 Vrn-H3 [16] Barley chromosome on which the QTL was determined. Genetic position of the peak marker and range of the QTL in cM, based on Comadran et al. [35]. Marker of the QTL with the highest significance (peak marker). Number of families, in which peak marker is segregating. Significance of the peak marker, expressed as P . BON-HOLM Cross-validated proportion of explained genotypic variance of peak marker. Frequency of significant detections of the peak marker in 100 five-fold cross-validation runs. Difference between the wild genotype and the cultivated genotype in days until flowering. Early flowering effects of exotic alleles are indicated in red. Candidate gene, potentially explaining the QTL effect with reference. Maurer et al. BMC Genomics (2015) 16:290 Page 4 of 12 Figure 2 Genetic architecture of flowering time in HEB-25. Barley chromosomes are indicated as colored bars on the inner circle, centromeres are highlighted as transparent boxes. a) Grey connector lines represent the genetic position of SNPs on the chromosomes. b) Frequency of QTL detection in 100 cross-validation runs via GWAS (0 to 100, grid line spacing: 25); markers with > 50 detections are colored in red. c) Additive effect of the SNP obtained from the BayesCπ genomic prediction model. d) Links in the center of the circle represent significant (P < 0.05) di-genic BON-HOLM interactions between SNP markers via GWAS. Clusters of significant SNP interactions are indicated by different colors. Position of candidate genes, potentially explaining major effects and epistatic effects, correspond to Table 1 and are indicated in blue outside the circle. eight major QTL. This finding is in contrast to flowering acid (GA) in flowering time regulation [44] through detec- time regulation in the allogamous species maize and tion of denso [45] and HvELF3 [46,47] as two further sorghum, where only small effect minor QTL were de- major QTL. Both genes are shown to be involved in GA tected [28,42]. The most significant association in HEB-25 biosynthesis [45,48]. So far, only the QTL on 4HS could (P = 3.4 E-130) was observed on the short arm not been referenced. This QTL, thus, remains a subject BON-HOLM of chromosome 2H and explained a p of 36%. This SNP for further genetic investigations. is directly located within Ppd-H1,the majordeterminant The eight major QTL were located with high genetic of photoperiod response in barley under long day condi- precision, with four QTL restricted to confidence inter- tion [15]. Seven further genomic regions of extraordinary vals of less than 0.9 cM (Table 1). In cases where gene- high significance were detected on chromosome arms specific SNPs were available (i.e. Ppd-H1 and Vrn-H3), 1HL, 2HS, 3HL, 4HS, 4HL, 5HL, and 7HS. All except one exactly those SNPs revealed the highest significance QTL (4HS) could be assigned to known flowering time within the respective QTL window (Additional file 6). genes (Table 1, Figure 2 and Additional file 7). Besides The exotic alleles at Ppd-H1 and Vrn-H3 revealed the Ppd-H1, also the vernalization genes Vrn-H1 and Vrn-H2, strongest effects, accelerating flowering time by 9.5 days as well as the floral inducer Vrn-H3 and its putative para- and delaying flowering time by 4.1 days, respectively. log HvCEN [43] exhibited highly significant effects. In The drastic effects of single QTL outline the high poten- addition, we could confirm the importance of gibberellic tial of introducing wild barley alleles from HEB-25 in Maurer et al. BMC Genomics (2015) 16:290 Page 5 of 12 slight increase of the cross-validated explained p from 36% to 38%. This finding implies that modelling haplotype-specific effects for a substantial portion of the barley gene space may result in an improved prediction of flowering time in HEB-25. However, a genome-wide re-sequencing of HEB-25 lines will be required to identify and distinguish those haplotypes. Applying genomic prediction models To check whether we could further elucidate the genetic architecture of barley flowering time we applied genomic prediction models that considered all markers simultan- eously. Genomic prediction evolved in animal breeding as a tool to predict a phenotype based on modelling a large set of SNP data [50]. It is used for selection of improved Figure 3 Cross-validated proportion of explained genotypic variance (p ) of different applied models. The box-whisker plots depict the genotypes based on estimated genomic breeding values. variation of explained genotypic variance after 100 cross-validations. Applying RR-BLUP [51] and BayesCπ [52] models, we The tested QTL models are (i) thesingleSNP locus Ppd-H1 (Mean could further increase the cross-validated explained p to p =0.36), (ii) GWAS with peak markers, representing the eight 71% and 74%, respectively (Figure 3). These findings are in major QTL indicated in Table 1 (Mean p = 0.64), (iii) the whole agreement with comparisons of multiple linear regression genome ridge regression best linear unbiased prediction (RR-BLUP, mean p = 0.71), (iv) the BayesCπ prediction (Mean p = 0.74), and and genomic prediction of traits in bi-parental plant popu- G G (v) RR-BLUP including epistasis (Mean p = 0.77). G lations [53]. However, our p values substantially exceed the prediction accuracies of genomic prediction models reported in comparable studies [54-56], underlining the order to adapt flowering time to environmental require- tremendous predictive power of HEB-25. Interestingly, ments and to enhance biodiversity in the elite barley compared to GWAS, only a few additional loci had non- breeding pool. zero effects in the BayesCπ model, indicating that flower- ing time is indeed mainly controlled by the eight major loci detected via GWAS. Ppd-H1 haplotype study We assume that important reasons for the slightly As we used bi-allelic SNP markers, additive effects were higher explained p of genomic prediction compared to estimated across the NAM population. Theoretically, GWAS are that minor QTL effects and marginally exist- there may be up to 26 different alleles present at each ing genetic relatedness [55,57] among HEB lines may be locus in HEB-25. Thus, distinct alleles that show con- better modeled in the first case. Furthermore, modeling trasting effects between families potentially escaped de- all makers simultaneously enables a better prediction of tection in our SNP-based GWAS. Contrasting effects are flowering time due to the estimation of family-specific illustrated in Figure 4 and, in detail, in Additional file 6. QTL effects. This is indicated by the occurrence of op- For instance, SNPs at position 46.2 cM on chromosome posing additive effects between HEB families alongside 3H, which are tightly linked to HvGI [49], revealed tightly linked SNPs (Figure 2 and Additional file 6). opposing effects across HEB families. We tested the po- tential to integrate SNP haplotypes in the GWAS model A model including epistasis to maximize the cross-validated for Ppd-H1, which exhibited the largest p . After re- explained p G G sequencing the last two exons and three introns of Ppd- A final increase of the cross-validated explained p to an H1, twelve haplotypes could be distinguished (Additional extraordinary high level of 77% was achieved by including file 1). All Hsp donor haplotypes at Ppd-H1 showed a sig- di-genic epistatic interactions between significant main ef- nificantly reduced flowering time (Additional file 8 and fect SNPs in the RR-BLUP model. This finding indicates Figure 5), where a maximum reduction of flowering time that epistasis explains a portion of the ‘missing heritability’ was associated with H-6 (−11.1 days compared to elite [58] of flowering time regulation in barley, whereas in barley haplotype H-2). Only the Hag haplotype H-45 did maize it does not [28]. The term ‘missing heritability’ is not differ from H-2. This finding confirms the presence of highly debated in quantitative genetics and refers to the haplotype-specific effects present in HEB-25. Conse- observation that the explained genotypic variance of com- quently, we expect the existence of further haplotype ef- bined marker effects is usually lower than the heritability fects for other candidate genes controlling flowering of the trait. Epistatic interactions between candidate genes time. The haplotype-based Ppd-H1 model resulted in a may point to functional relationships and genetic Maurer et al. BMC Genomics (2015) 16:290 Page 6 of 12 Figure 4 Visualization of family-wise SNP effects. Barley chromosomes are indicated as inner circle of colored bars, centromeres are highlighted as transparent boxes. Grey connector lines represent the genetic position of SNPs on chromosomes. Each track displays one HEB family (F01 – F25, from inside to outside). The heatmap indicates the difference in days between the donor and Barke genotype. Blue and red colors specify early and late flowering, respectively, caused by the donor genotype. White color indicates no SNP effect or SNPs monomorphic in the respective family. Candidate genes (Table 1) are indicated outside the circle. Black frames highlight their family-specific effects as indicated in Additional file 6. networks [59]. Our findings indicate that the flowering spring type barley cultivar that lacks the vernalization re- time genes HvGI, Vrn-H2, Vrn-H1 and HvCO1 [60] on quirement due to a deletion of Vrn-H2 [19]. Hence, our chromosomes 3H, 4H, 5H and 7H, respectively, are prob- findings may indicate that the epistatic interaction found ably major players of di-genic epistatic interactions in between the two regions on chromosomes 4H and 5H is HEB-25. All four genes potentially interact with each based on the presence (exotic allele) or absence (Barke other as well as with further genes on additional chromo- allele) of Vrn-H2,the target of Vrn-H1. In general, the somes (Figure 2 and Additional file 9). These observations epistatic interactions detected in HEB-25 may provide are in agreement with independent studies in barley and hints for the presence of so far unknown functional A. thaliana where these interacting genes were placed in a networks of genes, which assist in fine-tuning flowering day length and temperature depending signaling network time in barley. Studies with knock out lines of these genes that controls flowering time [14,21-23]. It is, thus, likely may be used to validate the observed interaction effects. that the observed interaction between the chromosomal regions in HEB-25 may be a function of the mentioned Conclusions flowering time genes. As an example we refer to the po- The first barley NAM population HEB-25 provides great tential interaction found between Vrn-H1 and Vrn-H2. opportunities for future research and breeding. The genetic Epistatic interactions between these loci were already constitution of HEB-25 allows to carry out detailed studies reported [17,61,62] and support the model that Vrn-H2 is on the genetic architecture of important agronomic traits, a long-day suppressor of flowering, that is itself sup- as exemplified by flowering time. The present study sub- pressed by Vrn-H1 following vernalization [63]. Barke is a stantiated that flowering time in barley is primarily Maurer et al. BMC Genomics (2015) 16:290 Page 7 of 12 originating from Tibet, China, was classified as Hordeum vulgare ssp. agriocrithon (Åberg). F plants of the initial crosses were backcrossed with Barke as the female parent. Twenty BC plants per cross were subsequently selfed three times, using the single seed descent (SSD) technique to generate the next generations. The resulting BC S gener- 1 3 ation consists of 1,420 individual lines, classified in 25 HEB families with 22 to 75 individual lines per family (Additional file 1). Subsequently, each HEB line was bulk propagated until BC S to achieve sufficient seed numbers for field 1 3:6 testing. No artificial selection was carried out during the de- velopment of HEB-25. Collecting single nucleotide polymorphism (SNP) data SNP genotype data were collected at TraitGenetics, Figure 5 Box-whisker plots of flowering time BLUEs for Ppd-H1 Gatersleben, Germany, for all 1,420 individual BC S lines 1 3 haplotypes. Green box-whisker-plots display the distribution of and their corresponding parents with the barley Infinium flowering time BLUEs of all HEB lines carrying the respective haplotype. Horizontal lines and diamonds indicate median and mean, respectively, iSelect 9k chip consisting of 7,864 SNPs [35]. At each locus, for each haplotype. The extension of vertical lines indicates minimum three genotypes were differentiated, with an expected and maximum observations, excluding outliers, which are indicated as BC S segregation ratio of 0.71875 : 0.0625 : 0.21875 1 3 circles. The red dotted horizontal line indicates the BLUE of cultivar for homozygous recipient (i.e. Barke), heterozygous and Barke (68.2 days). H-2 represents the haplotype of the Barke genotype homozygous donor genotypes, respectively. In total, present in HEB lines. All haplotypes except H-45 differ significantly (P < 0.05) from H-2, as indicated by red asterisks. Further information to 1,027 monomorphic SNPs and 1,128 SNPs with high haplotypes is given in Additional files 1 and 8. failure rates (i.e. no call in >10% of HEB lines) were ex- cluded from the dataset, resulting in 5,709 informative SNPs for further analyses. determined by large-effect QTL and epistatic interactions. This finding is in contrast to flowering time regulation in Extraction of genomic DNA the allogamous species maize and sorghum, where only DNA was extracted from leaf tissue of 1,420 single small effect minor QTL were detected [28,42], indicating founder HEB plants in generation BC S . The subse- 1 3 that the mating system may control the genetic architec- quent seed propagation of HEB lines was based on these ture of adaptive traits [28]. founder HEB plants. For Barke and the wild barley ac- In future, the NAM population HEB-25 will be utilized cessions leaf material from three to four plants was used in two directions: On the one hand, HEB-25 may support to create bulked samples. The plants were cultivated in a elucidating the genetic architecture of quantitatively inher- glasshouse and 50 to 100 mg of leaf material was harvested ited agronomic traits, ultimately resulting in cloning of yet for each sample. DNA was extracted according to the man- unknown causal genes. On the other hand, HEB-25 will be ufacturer’s protocol, using the BioSprint 96 DNA Plant Kit exploited by breeders to enhance biodiversity of the elite and a BioSprint work station (Qiagen, Hilden, Germany), barley gene pool. This may occur through introgression of and finally dissolved in distilled water at approximately favorable wild alleles with the aim to sustainably increase 50 ng/μl. yield and stress tolerances against disadvantageous climate conditions like drought, heat and salt. SNP mapping The chromosomal positions of 3,391 out of 5,709 SNPs Methods were taken from Comadran et al. [35]. The remaining Development of the NAM population SNPs were fitted next to the mapped SNPs applying chi- The development of the NAM population ‘Halle Exotic square tests of independence. Each non-mapped SNP Barley 25’ (HEB-25) was initiated in 2007 conducting was compared to each mapped SNP based on genotype crosses between the spring barley cultivar Barke (Hordeum segregation across all HEB lines. If two SNPs segregated vulgare ssp. vulgare)and 25 highly divergentexoticwild completely independent from each other, i.e. in case of barley accessions. The latter were used as pollen donors. no linkage disequilibrium (LD), one expects to find all Twenty-four accessions, originating from Afghanistan, Iran, possible genotype combinations according to the product Iraq, Israel, Lebanon, Turkey, and Syria (Hordeum vulgare of their single locus genotype frequencies. However, in case ssp. spontaneum), were selected to maximize the genetic di- of tight linkage, there should be a significant deviation from versity in HEB-25. One further accession, HID380, the expected genotype combination frequencies due to Maurer et al. BMC Genomics (2015) 16:290 Page 8 of 12 reduced recombination between these markers. Conse- between SNP pairs against their genetic distance. A quently, a high chi-square statistic and a low P-value likely second-degree smoothed loess curve [69] was fitted in indicate a tight linkage. Therefore, we assigned the position SAS with Proc Loess. The population-specific baseline r of the SNP with the lowest P-value (minimum: P < 0.001) was defined as the 95% percentile of the distribution of r to the non-mapped SNP under investigation. If there were for unlinked markers [70]. LD decay was defined as the more than one SNP with the same P-value, the position of distance, at which this baseline crossed the loess curve. the unmapped SNP was defined as the average of the mini- mum and the maximum position of the respective markers. Ppd-H1 haplotype definition In this way, all except six of the non-mapped SNPs were For sequencing of the Ppd-H1 locus on chromosome placed into the Comadran map. 2HS we used the following primers: PP05 (forward) 5′-GTGCAAAGCATAATATCAGTGTCC-3′ and PP04 SNP calling (reverse) 5′-GGCCAAAGACACAAGAATCAG-3′. These The differentiation of the HEB genotypes was based on an primers amplify the last two exons and three introns of identity-by-state approach. Based on parental genotype in- Ppd-H1 covering the CCT domain that contains SNP22, formation, the exotic allele could be specified in each segre- the causal SNP of Ppd-H1 [15]. Identical sequences were gating family. Thus, HEB lines that showed a homozygous grouped into haplotypes. A detailed description of the exotic genotype were assigned a value of 2 and HEB lines sequencing is given in Jakob et al. [71]. that showed a homozygous Barke genotype were assigned a value of 0. Consequently, heterozygous HEB lines were assigned a value of 1. If a SNP was monomorphic in one HEB-25 field trials HEB family but polymorphic in a second family, lines of Between 2011 and 2013, three field trials were conducted thefirst HEB familywereassignedagenotype value of0 to at the ‘Kühnfeld Experimental Station’ of the University of keep a full genotype data set, which is a pre-requisite for Halle to gather phenotype data on flowering time. In the subsequent multiple regression analysis. For the same 2011, the field trial was conducted with selfed progenies reason, missing genotypes were estimated applying the of BC S lines (so-called BC S ). Sowing occurred in sin- 1 3 1 3:4 mean imputation (MNI) approach [64]. For this, each miss- gle to five row plots with a length of 1.50 m and a distance ing SNP value was replaced with the mean of the non- of 0.20 m between rows. The number of rows per HEB missing values of that SNP in the respective HEB family. line and the position inside the field trial depended on the Quantitative SNP genotypes were subsequently used for number of available BC S seeds. Lines with seed num- 1 3:4 multiple regression analysis. bers lower than ten were planted in plots with a length of 0.50 m. In 2012 and 2013, the field trials were conducted Evaluation of genetic diversity with the selfed progenies in BC S and BC S , respect- 1 3:5 1 3:6 SAS 9.4 Software (SAS Institute Inc., Cary, NC, USA) was ively. Two replications per HEB line, arranged in two used to evaluate genetic diversity among parents and pro- randomized complete blocks, were cultivated in 2012 and genies of the HEB-25 population. Genetic similarities (GS) 2013. The plots consisted of two rows (30 seeds each) with between HEB lines and their parents and among HEB a length of 1.50 m and a distance of 0.20 m between rows. lines were calculated with Proc Distance, based on a sim- All field trials were sown in spring between March and ple matching comparison between the three possible April with fertilization and pest management following genotype states across all informative SNPs. In addition, local practice. we performed principal component analysis (PCA) using R [65]. First we applied PCA for the 26 parents (the culti- var Barke and 25 wild donors) based on the SNP matrix. Phenotypic data The first two PCs explained 51.9 and 4.8% of the variation. The occurrence of flowering time was recorded as days Then, all progenies of HEB-25 were projected to the space after sowing, when the first awns were visible (BBCH49 spanned by the two PCs (Additional file 3) as outlined in [72]) for 50% of all plants of a plot. We performed a one- detail elsewhere [66]. step phenotypic data analysis with SAS, using a linear mixed model with effects for genotype (i.e. 1,420 HEB Linkage disequilibrium (LD) lines), environment (i.e. 3 years) and interaction of geno- LD was calculated as r [67] between all mapped SNPs type and environment. To estimate variance components, with the software package TASSEL [68]. For this pur- all effects were assumed to be random. Broad-sense herit- pose, heterozygous genotypes and SNPs with a minor ability (h ) was estimated on an entry-mean basis. Best lin- allele frequency < 0.05 were excluded. LD was calcu- ear unbiased estimates (BLUEs) of flowering time were lated across the 26 parents of HEB-25. LD decay across calculated for each genotype assuming fixed genotype intra-chromosomal SNPs was displayed by plotting r effects. Maurer et al. BMC Genomics (2015) 16:290 Page 9 of 12 Genome-wide association study (GWAS) corresponding design matrix and e is the residual term. For GWAS, we applied Model-B as outlined in detail by 2 In the model we assumed that g e N 0; σ , e e N Liu et al. [38]. This model was found most suitable to 2 2 2 2 0; σ , where σ ¼ σ =m for SNP markers and σ ¼ e g G e carry out GWAS with multiple families [39]. It is based on 2 2 2 σ =l . Here σ and σ are the genotypic and residual multiple regression considering an SNP effect and a family R G R variance components obtained in the mixed model in effect in addition to cofactors, which control both popula- the phenotypic data analysis. The penalty parameter is tion structure and genetic background [39]. Cofactor se- 2 2 lection was carried out by applying Proc Glmselect in SAS λ ¼ σ =l = σ =m . The estimation of marker effects is R G and minimizing the Schwarz Bayesian Criterion [73]. then given by the mixed model equations [76]. The genome-wide scan for presence of marker-trait as- The basic model of BayesC π is the same as RR-BLUP. sociations was implemented in the statistical software R However, all parameters are treated as random variables [65], excluding cofactors linked closer than 1 cM to the in a Bayesian framework. First, we defined the prior dis- SNP under investigation. The Bonferroni–Holm pro- 2 2 tributions as g e N 0; σ ; e e N 0; σ . The prior of g e cedure [74] was used to adjust marker-trait associations μ is a constant. The prior distribution of σ is assumed for multiple testing. Significant marker main effects were g to be zero with probability π and a scaled inverse chi- accepted with P < 0.05. Additive effects for each BON-HOLM squared distribution with probability (1-π). The prob- SNP were estimated based on regression across but also ability π is a random variable whose prior distribution is within families. Significant marker trait associations were uniform on the interval [0,1]. The prior distribution of grouped to a singled QTL if the significant SNPs were linked by less than 5 cM and revealed additive effects σ is also scaled inverse chi-squared. A Gibbs sampler of the same direction, i.e. both exotic alleles increased or algorithm was then implemented to infer all the parame- decreased flowering time. In addition, a two-dimensional ters in the model. It was run for 10,000 cycles and the epistasis scan was carried out to identify pairwise marker first 1,000 cycles were discarded as burn in. The samples interactions. For this, the GWAS Model-B was extended of g from all later cycles were averaged to obtain esti- to cover a second main SNP effect and the interaction mates of the marker effects. effect between the two SNPs. Cross-validation for additive models Haplotype-based association mapping for Ppd-H1 The accuracy of the prediction of flowering time by A haplotype-based association mapping test was imple- GWAS and the two genomic prediction approaches mented in HEB-25 to test for effects of haplotypes at Ppd- were evaluated using five-fold cross-validations [77]. In H1. We used the same GWAS procedure with cofactor each run of cross-validation, the estimation set in- selection as mentioned before. However, bi-allelic SNPs cluded 80% of HEB lines, randomly selected per HEB covering the region of Ppd-H1 were replaced by a qualita- family, while the remaining 20% of HEB lines were tive variable containing the defined Ppd-H1 haplotype. assigned to build the test set. For GWAS, we per- BLUEs were determined for each haplotype. Subsequently, formed an association mapping scan within the estima- pairwise comparisons between all haplotype BLUEs were tion set and recorded the detected significant markers. performed using the Tukey-Kramer [75] multiple compari- To determine the cross-validated proportion of ex- son test. plained genotypic variance (p ), we estimated the ef- fects of the significant peak markers within the Genomic prediction estimation set and predicted the genotypic value of the Based on BLUEs of the 1,420 HEB genotypes, two ap- lines in the test set [40]. We then calculated the cross- proaches for genomic prediction were applied considering validated p as the squared Pearson product–moment additive effects: ridge regression best linear unbiased pre- correlation between predicted and observed genotypic diction (RR-BLUP [51]) and BayesCπ [52]. All statistical values in the test set standardized with the heritability. procedures for genomic prediction approaches were exe- The mean p in 100 cross-validation runs (20 times five- cuted using R. We briefly describe the two models in the fold cross-validations) was taken as the final record. In following. addition, the number of significances for each SNP was Let n be the number of genotypes, m be the number cumulated across all runs and is referred to as QTL de- of markers and l be the number of environments. The tection rate. RR-BLUP model has the form y =1 μ + Xg + e, where y For genomic prediction we estimated the effects for all is the vector of BLUEs of flowering time scores for all markers using the estimation set and predicted the HEB genotypes across environments, 1 denotes the vec- genotypic value of the lines in the test set. The cross- tor of 1’s, μ is the overall mean, g is the vector of marker validated p was calculated as in GWAS. effects (for SNP markers, allele effects), X is the G Maurer et al. BMC Genomics (2015) 16:290 Page 10 of 12 Exploiting additive and additive times additive epistatic Additional file 4: Distribution of flowering time. Figure showing the effects in genomic prediction frequency distribution of flowering time BLUEs across three field trials and illustrating contrasting phenotypes in the field. We extended the RR-BLUP based on main effects Additional file 5: Phenotype and genotype data for HEB-25. Table to model also epistasis for markers with significant listing the complete phenotype and genotype data of HEB-25 underlying main effects in the GWAS. The model is y ¼ 1 μþ this study as well as marker information. X X X g þ X ⋅X f þ e,where y is the vector of i j l Additional file 6: Estimates of single marker GWAS and genomic jl i¼1 j<l prediction effects across HEB-25 and within individual HEB families. BLUEs of flowering time for all HEB genotypes, 1 denotes Table listing the results of GWAS and genomic prediction across HEB-25 the vector of 1’s, μ is the overall mean, g is the main and within individual HEB families. additive effect of the i-th marker, X is the vector of marker Additional file 7: GWAS Manhattan plot for flowering time. Figure displaying the GWAS results through plotting the significance and effects indices, f is the epistatic effects of the j- and the l-th jl of markers in a Manhattan plot. marker, X X is the point-wise product of the two vectors j l Additional file 8: Ppd-H1 haplotype comparison. Two tables X and X,and e is the vector of residual terms. Note that j l contrasting the different Ppd-H1 haplotype effects by comparison of their in the third term of the right hand side of the formula, the BLUEs. sum is not taken over all pairs of markers but only pairs of Additional file 9: Significant epistatic interactions via GWAS. Table listing all significant (P < 0.05) epistatic interactions between BON-HOLM markers exhibiting a significant additive effect in the SNPs that were obtained via GWAS. GWAS study performed previously. Hence, in different cross-validation runs, different pairs of markers were con- Abbreviations sidered in the model. The model assumptions are similar BC : Progeny of F after back-crossing with Barke; BC S : Progeny of BC 1 1 1 3 1 to the usual RR-BLUP, except treating additive and epistatic after three rounds of selfing; BC S :Progeny of BC S individual after x-3 1 3:x 1 3 rounds of bulk propagation; BLUE: Best linear unbiased estimate; effects separately. We assumed g e N 0; σ , f e N i g jl cM: Centimorgan; F : First generation after initial crosses; GA: Gibberellic acid; GS: Genetic similarity; GWAS: Genome-wide association study; 2 2 2 2 2 0; σ ,where σ ¼ p σ =m, σ ¼ðÞ 1−p σ =p.Here P f g G G f G G Hag: Hordeum agriocrithon; HEB: Halle exotic barley; HID: Hordeum identity; Hsp: Hordeum spontaneum; Hv: Hordeum vulgare;LD: Linkage is the cross-validated proportion of explained genotypic disequilibrium; MNI: Mean imputation; NAM: Nested association mapping; variance for genomic prediction, obtained previously by PC: Principal component; PCA: Principal component analysis; p : Proportion of explained genotypic variance; QTL: Quantitative trait the RR-BLUP, only considering additive effects, m is the G locus/loci; RR-BLUP: Ridge regression best linear unbiased prediction; number of markers, p is the number of pairs of markers SNP: Single nucleotide polymorphism; SSD: Single seed descent. having significant additive effect. Therefore the penalty parameter λ is different for additive and epistatic effects. Competing interests The authors declare that they have no competing interests. Using the above extended model, for each cross-validation run we estimated the additive effects of all markers and Authors’ contributions epistatic effects of all pairs of markers exhibiting sig- AM planned and conducted the field trials in 2012 and 2013, analyzed the nificant additive effects in GWAS using the estimation genotype and phenotype data, carried out the GWAS, created the figures, and drafted the manuscript. VD was involved in the development of the HEB-25 set. Then we predicted genotypic values of the lines in population, gathered the genotype data, and planned and conducted the field the test set and calculated the p in thesamewayas trial in 2011. YJ performed the genomic prediction and cross-validation outlined above. approaches. FS was involved in the development of the HEB-25 population. RS performed the re-sequencing of Ppd-H1. ES supervised the field trials. BK supervised the Ppd-H1 re-sequencing. JCR supervised the genomic prediction Availability of supporting data and cross-validation approaches. KP acquired the funding, supervised the development of the HEB-25 population and all analyses, and drafted the Raw data, including data on SNPs, Ppd-H1 haplotypes manuscript. All authors read and approved the final manuscript. and GWAS, and all other supporting data are provided as additional files. Acknowledgements This work was supported by Deutsche Forschungsgemeinschaft (DFG), Bonn (projects Pi339/7-1 and Pi 339/7-2), and the Interdisciplinary Centre for Crop Additional files Plant Research (IZN), Halle. We are grateful to Roswitha Ende, Jana Müglitz, Diana Rarisch, Helga Sängerlaub, Brigitte Schröder, Bernd Kollmorgen and Additional file 1: Genetic constitution of HEB-25: classification of various student assistants for excellent technical help and to TraitGenetics families and donors. Tabular overview of the genetic constitution of GmbH, Gatersleben, Germany, for genotyping HEB-25 with the Infinium HEB-25, classifying the 25 families and donors and containing the Ppd-H1 iSelect 9k SNP chip. haplotypes. Additional file 2: LD decay of intra-chromosomal markers among Author details HEB-25 parents. Figure showing the LD decay of intra-chromosomal Institute of Agricultural and Nutritional Sciences, Martin Luther University markers among HEB-25 parents by plotting r against the genetic marker Halle Wittenberg, Betty-Heimann-Str. 3, 06120 Halle, Germany. distance. Interdisciplinary Center for Crop Plant Research (IZN), Betty-Heimann-Str. 3, 06120 Halle, Germany. Leibniz-Institute of Plant Genetics and Crop Plant Additional file 3: Principal component analysis for HEB-25 and its Research (IPK), Corrensstr. 3, 06466 Stadt Seeland, OT Gatersleben, Germany. parents. Figure showing the relatedness of HEB-25 lines by plotting of Current address: University of Dundee at the James Hutton Institute, the first two principal components of a principal component Invergowrie, Dundee DD2 5DA, UK. Current address: Bayer CropScience NV, analysis for HEB-25 and its parents. Technologiepark 38, 9052 Ghent, Belgium. Maurer et al. BMC Genomics (2015) 16:290 Page 11 of 12 Received: 12 October 2014 Accepted: 9 March 2015 25. Nevo E, Fu Y-B, Pavlicek T, Khalifa S, Tavasi M, Beiles A. Evolution of wild cereals during 28 years of global warming in Israel. Proc Natl Acad Sci U S A. 2012;109:3412–5. 26. Yu J, Holland JB, McMullen MD, Buckler ES. Genetic design and statistical power of nested association mapping in maize. Genetics. 2008;178:539–51. References 27. Jordan D, Mace E, Cruickshank A, Hunt C, Henzell R. Exploring and 1. Sakuma S, Salomon B, Komatsuda T. The domestication syndrome genes exploiting genetic variation from unadapted sorghum germplasm in a responsible for the major changes in plant form in the Triticeae crops. breeding program. Crop Sci. 2011;51:1444–57. Plant Cell Physiol. 2011;52:738–49. 28. Buckler ES, Holland JB, Bradbury PJ, Acharya CB, Brown PJ, Browne C, et al. 2. Zohary D, Hopf M, Weiss E. Domestication of Plants in the Old World: The The Genetic Architecture of Maize Flowering Time. Science. 2009;325:714–8. origin and spread of domesticated plants in Southwest Asia, Europe, and 29. Kump KL, Bradbury PJ, Wisser RJ, Buckler ES, Belcher AR, Oropeza-Rosas MA, the Mediterranean Basin. 4th ed. Oxford: Oxford University Press; 2012. et al. Genome-wide association study of quantitative resistance to southern 3. Morrell PL, Clegg MT. Genetic evidence for a second domestication of leaf blight in the maize nested association mapping population. Nat Genet. barley (Hordeum vulgare) east of the Fertile Crescent. Proc Natl Acad Sci U 2011;43:163–8. S A. 2007;104:3289–94. 30. Poland JA, Bradbury PJ, Buckler ES, Nelson RJ. Genome-wide nested association 4. Dai F, Nevo E, Wu D, Comadran J, Zhou M, Qiu L, et al. Tibet is one of the mapping of quantitative resistance to northern leaf blight in maize. Proc Natl centers of domestication of cultivated barley. Proc Natl Acad Sci U S A. Acad Sci U S A. 2011;108:6893–8. 2012;109:16969–73. 31. Tian F, Bradbury PJ, Brown PJ, Hung H, Sun Q, Flint-Garcia S, et al. Genome-wide 5. Tanksley SD, McCouch SR. Seed Banks and Molecular Maps: Unlocking association study of leaf architecture in the maize nested association mapping Genetic Potential from the Wild. Science. 1997;277:1063–6. population. Nat Genet. 2011;43:159–62. 6. Zamir D. Improving plant breeding with exotic genetic libraries. Nat Rev 32. Cook JP, McMullen MD, Holland JB, Tian F, Bradbury P, Ross-Ibarra J, et al. Genet. 2001;2:983–9. Genetic Architecture of Maize Kernel Composition in the Nested Association 7. Ellis RP, Forster BP, Robinson D, Handley LL, Gordon DC, Russell JR, et al. Mapping and Inbred Association Panels. Plant Physiol. 2012;158:824–34. Wild barley: a source of genes for crop improvement in the 21st century? 33. Peiffer JA, Romay MC, Gore MA, Flint-Garcia SA, Zhang Z, Millard MJ, et al. J Exp Bot. 2000;51:9–17. The genetic architecture of maize height. Genetics. 2014;196:1337–56. 8. Pillen K, Zacharias A, Leon J. Advanced backcross QTL analysis in barley 34. Wallace J, Larsson S, Buckler E. Entering the second century of maize (Hordeum vulgare L.). Theor Appl Genet. 2003;107:340–52. quantitative genetics. Heredity (Edinb). 2014;112:30–8. 9. Wang G, Schmalenbach I, von Korff M, Leon J, Kilian B, Rode J, et al. 35. Comadran J, Kilian B, Russell J, Ramsay L, Stein N, Ganal M, et al. Natural Association of barley photoperiod and vernalization genes with QTLs for variation in a homolog of Antirrhinum CENTRORADIALIS contributed to flowering time and agronomic traits in a BC(2)DH population and a set of spring growth habit and environmental adaptation in cultivated barley. wild barley introgression lines. Theor Appl Genet. 2010;120:1559–74. Nat Genet. 2012;44:1388–92. 10. Schmalenbach I, March TJ, Bringezu T, Waugh R, Pillen K. High-Resolution 36. Gottwald S, Bauer P, Komatsuda T, Lundqvist U, Stein N. TILLING in the Genotyping of Wild Barley Introgression Lines and Fine-Mapping of the two-rowed barley cultivar ‘Barke’ reveals preferred sites of functional Threshability Locus thresh-1 Using the Illumina GoldenGate Assay. G3 diversity in the gene HvHox1. BMC Res Notes. 2009;2:258. (Bethesda). 2011;1:187–96. 11. Ma X, Li C, Wang A, Duan R, Jiao G, Nevo E, et al. Genetic diversity of wild 37. Badr A, Muller K, Schafer-Pregl R, El Rabey H, Effgen S, Ibrahim HH, et al. barley (Hordeum vulgare ssp. spontaneum) and its utilization for barley On the origin and domestication history of barley (Hordeum vulgare). improvement. Sci Cold Arid Reg. 2012;4:453–61. Mol Biol Evol. 2000;17:499–510. 38. Liu W, Gowda M, Steinhoff J, Maurer HP, Würschum T, Longin CFH, et al. 12. McCouch S, Baute GJ, Bradeen J, Bramel P, Bretting PK, Buckler E, et al. Association mapping in an elite maize breeding population. Theor Appl Agriculture: Feeding the future. Nature. 2013;499:23–4. Genet. 2011;123:847–58. 13. Schnaithmann F, Kopahnke D, Pillen K. A first step toward the development 39. Würschum T, Liu W, Gowda M, Maurer H, Fischer S, Schechert A, et al. of a barley NAM population and its utilization to detect QTLs conferring leaf Comparison of biometrical models for joint linkage association mapping. rust seedling resistance. Theor Appl Genet. 2014;127:1513–25. Heredity (Edinb). 2012;108:332–40. 14. Cockram J, Jones H, Leigh FJ, O’Sullivan D, Powell W, Laurie DA, et al. 40. Utz HF, Melchinger AE, Schön CC. Bias and sampling error of the estimated Control of flowering time in temperate cereals: genes, domestication, and proportion of genotypic variance explained by quantitative trait loci sustainable productivity. J Exp Bot. 2007;58:1231–44. determined from experimental data in maize using cross validation and 15. Turner A, Beales J, Faure S, Dunford RP, Laurie DA. The pseudo-response validation with independent samples. Genetics. 2000;154:1839–49. regulator Ppd-H1 provides adaptation to photoperiod in barley. Science. 2005;310:1031–4. 41. Gowda M, Zhao Y, Würschum T, Longin CF, Miedaner T, Ebmeyer E, et al. 16. Yan L, Fu D, Li C, Blechl A, Tranquilli G, Bonafede M, et al. The wheat and Relatedness severely impacts accuracy of marker-assisted selection for barley vernalization gene VRN3 is an orthologue of FT. Proc Natl Acad Sci U disease resistance in hybrid wheat. Heredity (Edinb). 2014;112:552–61. S A. 2006;103:19581–6. 42. Mace E, Hunt C, Jordan D. Supermodels: sorghum and maize provide 17. Yan L, Loukoianov A, Blechl A, Tranquilli G, Ramakrishna W, SanMiguel P, mutual insight into the genetics of flowering time. Theor Appl Genet. et al. The wheat VRN2 gene is a flowering repressor down-regulated by 2013;126:1377–95. vernalization. Science. 2004;303:1640–4. 43. Loscos J, Igartua E, Contreras-Moreira B, Gracia MP, Casas AM. HvFT1 polymorphism and effect – survey of barley germplasm and expression analysis. Front Plant Sci. 18. Yan L, Loukoianov A, Tranquilli G, Helguera M, Fahima T, Dubcovsky J. 2014;5:251. Positional cloning of the wheat vernalization gene VRN1. Proc Natl Acad Sci U S A. 2003;100:6263–8. 44. Mutasa-Göttgens E, Hedden P. Gibberellin as a factor in floral regulatory 19. von Zitzewitz J, Szűcs P, Dubcovsky J, Yan L, Francia E, Pecchioni N, et al. networks. J Exp Bot. 2009;60:1979–89. Molecular and structural characterization of barley vernalization genes. 45. Jia Q, Zhang J, Westcott S, Zhang X-Q, Bellgard M, Lance R, et al. GA-20 oxidase Plant Mol Biol. 2005;59:449–67. as a candidate for the semidwarf gene sdw1/denso in barley. Funct Integr 20. Laurie D, Pratchett N, Snape J, Bezant J. RFLP mapping of five major genes Genomics. 2009;9:255–62. and eight quantitative trait loci controlling flowering time in a winter × 46. Faure S, Turner AS, Gruszka D, Christodoulou V, Davis SJ, von Korff M, et al. spring barley (Hordeum vulgare L.) cross. Genome. 1995;38:575–85. Mutation at the circadian clock gene EARLY MATURITY 8 adapts domesticated 21. Distelfeld A, Li C, Dubcovsky J. Regulation of flowering in temperate cereals. barley (Hordeum vulgare) to short growing seasons. Proc Natl Acad Sci U S A. Curr Opin Plant Biol. 2009;12:178–84. 2012;109:8328–33. 22. Jung C, Mueller AE. Flowering time control and applications in plant 47. Zakhrabekova S, Gough SP, Braumann I, Muller AH, Lundqvist J, Ahmann K, et al. breeding. Trends Plant Sci. 2009;14:563–73. Induced mutations in circadian clock regulator Mat-a facilitated short-season 23. Milec Z, Valárik M, Bartoš J, Šafář J. Can a late bloomer become an early adaptation and range extension in cultivated barley. Proc Natl Acad Sci U S A. bird? Tools for flowering time adjustment. Biotechnol Adv. 2014;32:200–14. 2012;109:4326–31. 24. Cockram J, Hones H, O’Sullivan DM. Genetic variation at flowering time loci 48. Boden SA, Weiss D, Ross JJ, Davies NW, Trevaskis B, Chandler PM, et al. in wild and cultivated barley. Plant Genet Resour. 2011;9:264–7. EARLY FLOWERING3 Regulates Flowering in Spring Barley by Mediating Maurer et al. BMC Genomics (2015) 16:290 Page 12 of 12 Gibberellin Production and FLOWERING LOCUS T Expression. Plant Cell. 75. Kramer CY. Extension of multiple range tests to group means with unequal 2014;26:1557–69. numbers of replications. Biometrics. 1956;12:307–10. 49. Dunford RP, Griffiths S, Christodoulou V, Laurie DA. Characterisation of a 76. Henderson CR. Applications of Linear Models in Animal Breeding. Guelph: barley (Hordeum vulgare L.) homologue of the Arabidopsis flowering time University of Guelph; 1984. regulator GIGANTEA. Theor Appl Genet. 2005;110:925–31. 77. Hjorth JU. Computer intensive statistical methods: Validation, model 50. Meuwissen THE, Hayes BJ, Goddard ME. Prediction of Total Genetic Value selection, and bootstrap. London: Chapman & Hall/CRC; 1993. Using Genome-Wide Dense Marker Maps. Genetics. 2001;157:1819–29. 51. Whittaker JC, Thompson R, Denham MC. Marker-assisted selection using ridge regression. Genet Res. 2000;75:249–52. 52. Habier D, Fernando RL, Kizilkaya K, Garrick DJ. Extension of the Bayesian alphabet for genomic selection. BMC Bioinf. 2011;12:186. 53. Lorenzana RE, Bernardo R. Accuracy of genotypic value predictions for marker-based selection in biparental plant populations. Theor Appl Genet. 2009;120:151–61. 54. Guo ZG, Tucker DM, Lu JW, Kishore V, Gay G. Evaluation of genome-wide selection efficiency in maize nested association mapping populations. Theor Appl Genet. 2012;124:261–75. 55. Guo Z, Tucker DM, Basten CJ, Gandhi H, Ersoz E, Guo B, et al. The impact of population structure on genomic prediction in stratified populations. Theor Appl Genet. 2014;127:749–62. 56. Zhao Y, Mette MF, Gowda M, Longin CFH, Reif JC. Bridging the gap between marker-assisted and genomic selection of heading time and plant height in hybrid wheat. Heredity (Edinb). 2014;112:638–45. 57. Habier D, Fernando R, Dekkers J. The impact of genetic relationship information on genome-assisted breeding values. Genetics. 2007;177:2389–97. 58. Mackay TF. Epistasis and quantitative traits: using model organisms to study gene-gene interactions. Nat Rev Genet. 2013;15:22–33. 59. von Korff M, Léon J, Pillen K. Detection of epistatic interactions between exotic alleles introgressed from wild barley (H. vulgare ssp. spontaneum). Theor Appl Genet. 2010;121:1455–64. 60. Griffiths S, Dunford RP, Coupland G, Laurie DA. The evolution of CONSTANS-like gene families in barley, rice, and Arabidopsis. Plant Physiol. 2003;131:1855–67. 61. Tranquilli G, Dubcovsky J. Epistatic interaction between vernalization genes Vrn-Am1 and Vrn-Am2 in diploid wheat. J Hered. 2000;91:304–6. 62. Szűcs P, Skinner JS, Karsai I, Cuesta-Marcos A, Haggard KG, Corey AE, et al. Validation of the VRN-H2/VRN-H1 epistatic model in barley reveals that intron length variation in VRN-H1 may account for a continuum of vernalization sensitivity. Mol Genet Genomics. 2007;277:249–61. 63. Chen A, Dubcovsky J. Wheat TILLING mutants show that the vernalization gene VRN1 down-regulates the flowering repressor VRN2 in leaves but is not essential for flowering. PLoS Genet. 2012;8:e1003134. 64. Rutkoski JE, Poland J, Jannink J-L, Sorrells ME. Imputation of Unordered Markers and the Impact on Genomic Selection Accuracy. G3 (Bethesda). 2013;3:427–39. 65. R Development Core Team (2010) R: A language and environment for statistical computing, R Foundation for Statistical Computing. http://www.R-project.org 66. Xia X, Reif J, Melchinger A, Frisch M, Hoisington D, Beck D, et al. Genetic diversity among CIMMYT maize inbred lines investigated with SSR markers. Crop Sci. 2005;45:2573–82. 67. Weir BS. Genetic Data Analysis II. Sunderland, MA: Sinauer Associates, Inc.; 1996. 68. Bradbury PJ, Zhang Z, Kroon DE, Casstevens TM, Ramdoss Y, Buckler ES. TASSEL: software for association mapping of complex traits in diverse samples. Bioinformatics. 2007;23:2633–5. 69. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc. 1979;74:829–36. 70. Breseghello F, Sorrells ME. Association mapping of kernel size and milling quality in wheat (Triticum aestivum L.) cultivars. Genetics. 2006;172:1165–77. 71. Jakob SS, Rödder D, Engler JO, Shaaf S, Özkan H, Blattner FR, et al. Submit your next manuscript to BioMed Central Evolutionary History of Wild Barley (Hordeum vulgare subsp. spontaneum) and take full advantage of: Analyzed Using Multilocus Sequence Data and Paleodistribution Modeling. Genome Biol Evol. 2014;6:685–702. • Convenient online submission 72. Lancashire PD, Bleiholder H, Boom TVD, Langelüddeke P, Stauss R, Weber E, • Thorough peer review et al. A uniform decimal code for growth stages of crops and weeds. Ann Appl Biol. 1991;119:561–601. • No space constraints or color figure charges 73. Schwarz G. Estimating the dimension of a model. Ann Stat. 1978;6:461–4. • Immediate publication on acceptance 74. Holm S. A simple sequentially rejective multiple test procedure. Scand J • Inclusion in PubMed, CAS, Scopus and Google Scholar Stat. 1979;6:65–70. • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit

Journal

BMC GenomicsSpringer Journals

Published: Apr 12, 2015

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off