Brassinosteroid and Gibberellin Control of Plant Height in Maize (Zea Mays

Iowa State University Capstones, Theses and Graduate Theses and Dissertations Dissertations

2016 Brassinosteroid and gibberellin control of plant height in maize (Zea mays. L) Songlin Hu Iowa State University

Follow this and additional works at: https://lib.dr.iastate.edu/etd Part of the Agricultural Science Commons, Agriculture Commons, and the Agronomy and Crop Sciences Commons

Recommended Citation Hu, Songlin, "Brassinosteroid and gibberellin control of plant height in maize (Zea mays. L)" (2016). Graduate Theses and Dissertations. 15936. https://lib.dr.iastate.edu/etd/15936

This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected].

Brassinosteroid and gibberellin control of plant height in maize (Zea mays. L)

Songlin Hu

A dissertation submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

Major: Plant Breeding

Program of Study Committee: Thomas Lübberstedt, Major Professor Maria G. Salas Fernandez Yanhai Yin Michael Blanco Peng Liu

Iowa State University

Ames, Iowa

2016

TABLE OF CONTENTS

ACKNOWLEDGMENTS ...... iv

ABSTRACT………………………………...... v

CHAPTER 1. GENERAL INTRODUCTION ...... 1

Brassinosteroids ...... 3 Gibberellins ...... 6 Quantitative Genetics of PHT in Maize ...... 7 References ...... 10

CHAPTER 2. GETTING THE ‘MOST’ OUT OF CROP IMPROVEMENT ...... 19

Abstract ...... 19 Introduction ...... 20 Development of MOST treatments ...... 25 Concluding Remarks ...... 31 References ...... 32 Figures and Tables ...... 40

CHAPTER 3. GENETIC ARCHITECTURE OF PLANT HEIGHT IN MAIZE PHENOTYPE-SELECTED INTROGRESSION FAMILIES ...... 46 Abstract ...... 46 Introduction ...... 47 Materials and Methods ...... 51 Results ...... 55 Discussion ...... 58 References ...... 64 Figures and Tables ...... 71

CHAPTER 4. GIBBERELLINS PROMOTE BRASSINOSTEROIDS AND BOTH INCREASE HETEROSIS FOR PLANT HEIGHT IN MAIZE ...... 92 Abstract ...... 92 Introduction ...... 94 Results ...... 97 Discussion ...... 102

iii

Conclusions ...... 109 Methods ...... 110 References ...... 118 Figures and Tables ...... 126

CHAPTER 5. BRASSINOSTEROID AND GIBBERELLIN CONTROL OF SEEDLING TRAITS IN MAIZE ...... 166 Summary ...... 166 Introduction ...... 166 Results ...... 170 Discussion ...... 176 Experimental Procedures ...... 183 References ...... 189 Figures and Tables ...... 197

CHAPTER 6. GENERAL CONCLUSIONS ...... 218

ACKNOWLEDGMENTS

I would like to take this chance to thank those who helped me throughout my graduate career for support in all facets. First, I want to thank my advisor Thomas

Lübberstedt and the entire lab members for their efforts in helping me with advice and support. Next, I want to thank my POS committee members, who have always been willing to give helpful advice and a lot of support. I would also like to thank my collaborators Dr.

Alex Lipka and Dr. Candice Gardener on my projects. Finally, I would like to thank my parents for their love and support, also my friends, colleagues, the department faculty and staff for making my time at Iowa State University a wonderful experience.

ABSTRACT

Brassinosteroids (BRs) and gibberellins (GAs) are two major plant hormones involved in many plant developmental processes. However, the knowledge of BR and GA control of agronomic traits in maize is limited compared to model species Arabidopsis and rice, especially BR. This PhD project focused on BR and GA control of plant height from genetics and plant biology perspectives, as these two plant hormones have shown great impact for shaping plant height, and a short/tall stature of maize height is beneficial for grain/biomass production, respectively. We introgressed multiple exotic accessions into two maize heterotic groups Stiff Stalk and Non Stiff Stalk, with phenotype selection for taller plant height and undistinguishable flowering time as temperate elite lines. Finally, we created two backcross libraries varied with plant height and the level of heterozygosity. Moreover, another two doubled-haploid libraries were established without any phenotype selection as a comparison. We conducted genome-wide association studies to investigate plant height, BR and GA candidate genes, and explored co-localizations. In addition, we compared BR and

GA control of plant height in heterozygotes (backcross families) and inbred lines (doubled haploid lines) by correlating seeding BR/GA inhibitor response with field plant height. We found that discovered plant height associated genomic regions were overlapped with BR and

GA candidate genes. Moreover, we found that seedling stage BR and GA inhibitor response was able to predict plant height for heterozygotes, but not inbred lines. Path analyses showed that higher level of heterozygosity increased GA level, and GA promoted BR (by crosstalk), finally BR and GA will increase heterosis in maize plant height.

CHAPTER 1.

GENERAL INTRODUCTION

Modern agriculture remarkably increased food production, yet one in nine people remain undernourished and unable to consume sufficient protein and energy [1]. Although plant breeding activities have been very successful in the past few decades (e.g. Green Revolution) as a linear increase at an average rate of 32 million metric tons per year has been achieved for yield

[2], this is not sufficient, as the global population likely exceeding 9 billion by 2050. An average annual increase of 44 million metric tons per year is needed [1, 3] to address the food crisis.

Moreover, altered consumption habits, climate change and production of raw materials for biofuels and other products may require further increase in crop productivity and stability.

Manipulation of plant architecture is an effective way to improve crop yield, also for increasing pest and disease resistance. Primary growth, branching and reiteration allow the plant to develop its 3D structure to escape or survive to attacks [4]. The principles and controls of plant architecture underlie every successful breeding program. For example, Over the past 70 years, genetic improvements as well as farming practices have produced a >6-fold increase in corn yields. These have not resulted from increases in individual plant yields, but rather from increased tolerance to higher cropping densities, enabled in large part by changes in shoot architecture such as altered leaf angle (more upright) [5, 6]. Plant hormones determine traits such as plant height (PHT), biomass, organ size, yield, branching, stress tolerance, all of which affect plant performance. The knowledge of metabolism, transport and signaling of hormone pathways is important for controlling plant architecture traits and for crop improvement.

Among plant architecture traits, PHT is a key architectural trait intimately related to both grain and biomass yield [7, 8]. The “green revolution” was due to development and widespread adoption of high-yielding semi-dwarf varieties of rice and wheat, in combination with the application of increased amounts of nitrogen fertilizer [9]. Application of nitrogen fertilizer promotes grain development to a greater extent than vegetative growth in such dwarf varieties, which are now used extensively worldwide. Additionally, the use of dwarf crop plants allows reduced use of pesticides and fertilizers, better water use efficiency, and higher planting densities.

Thus, increased use of semi-dwarf crop varieties may help meet future agricultural production demands and environmental challenges. However, for dual purpose or biomass crop, a tall stature is more desirable [10]. For example, increased stover biomass adds value (grain is harvested for food or feed, stover for bioenergy conversion) for dual-purpose maize. In 2009, more than 10 billion gallons of ethanol were produced in the US - mostly as corn-starch ethanol

(www.ethanolrfa.org) and production of 36 billion gallons is targeted by the US congress for

2022 (Renewable Fuels Standard, EPA). However, seed-based biofuel production is limited by competing requirements for feed and food corn. Cellulosic biomass provides an alternative basis for biofuels, but the shift to cellulosic biomass requires the deployment of new interrelated plant traits such as greater vegetative biomass, altered stature (taller plants) and biomass composition

[11, 12]. To overcome lodging associated with taller PHT, breeders can either produce maize varieties with increased lignin level (highly lignified cell walls are more suitable for thermochemical conversion for biomass production) or a stronger root system to stabilize plants.

Thus, the maize plant ideotype for biomass production is different from the grain production ideotype.

Considerable information is available on the hormonal control of PHT and single gene loci are known with profound effects on this trait. Many of the genes involved in gibberellin (GA) or brassinosteroid (BR) synthesis or signaling were originally discovered by virtue of their dwarf mutant phenotypes. In maize, several maize dwarf genes have been well characterized, but are not intentionally used in breeding programs due to their adverse impact on grain yield, such as dwarf3 [13], dwarf8 and dwarf9 and nana plant1 [14-16], and brd1 [17]. All these genes cause defects in either the BR or the GA pathway, stressing the importance of these two plant hormones in the control of maize PHT. It was reported that BR and GA have the most direct effects on PHT without major negative pleiotropic effects as compared to other plant hormones

[10]. Understanding the profound role of hormones in controlling plant architecture has led to the development of synthetic growth retardants, which are routinely used to reduce stem elongation in a variety of crops and ornamentals. All the commonly used growth retardants are inhibitors of gibberellin biosynthesis and limit stem or shoot growth by reducing elongation. However, most of these compounds are detrimental to human and animal health and are only used for non-food plant species. Uncovering the molecular mechanisms controlling plant growth will help optimize crop ideotypes, and allow increased yields with reduced dependence on synthetic growth regulators.

Brassinosteroids

BRs are a type of hormone that is found throughout the plant kingdom. They are steroid hormones similar to those found in animals [18]. They promote cell growth, even if at low concentrations, by regulating cell divisions and elongation [19, 20]. As mentioned, mutants

4 defective in BR biosynthesis or signaling display dwarf phenotypes while increased BR can increase plant size, biomass and seed yield [18, 21]. Mutations that disrupt enzymatic steps in

BR biosynthesis can decrease PHT due to impaired cell elongation [18]. In maize, reduced expression of ZmDWF1 resulted in dwarf plants [22], whereas overexpression of the maize

DWF4 increased BR biosynthesis and activity in rice [23]. Recently, nana plant1 (na1) and brd1 were reported as BR-deficient mutants in maize with strong reduction in internode elongation as well as novel phenotypes, including sex determination and leaf auricle development [16, 17].

Na1 is a homolog of DET2, which encodes a steroid 5α-reductase, while brd1 encodes a BR-C-6 oxidase. These studies indicate that BR synthesis in maize is similar to Arabidopsis, but that BRs have novel functions in maize architecture. Remarkably, some BR functions appear opposite in maize vs. Arabidopsis or other plants. Down-regulated BR signaling in maize, such as in the recently isolated url1 mutant– the maize homolog of the BR regulatory transcription factor

OsDLT1 - causes an increased leaf angle in stark opposition to the strong epinasty shown in

Arabidopsis leaves (The “classical” BR mutant phenotype in Arabidopsis includes a significant increase in leaf number and shoot branches, whereas maize BR mutants show a strong suppression of tiller branching, which is epistatic even to high-tillering mutants such as teosinte branched1 (tb1). This was corroborated with suppression of tillering in the close maize relative,

Sorghum bicolor, by pharmacological inhibition of BR biosynthesis.

BR signal transduction [24-26] initiates with binding to BRI1, a plasma membrane- localized leucine rich repeat (LRR) receptor kinase. BRI1 has been studied in several crops including the dwarf 61 (d61) mutant in rice [27, 28] and the “green revolution” semi-dwarf

“uzu” barley [29]. In the absence of BR, the negative regulator BKI1 binds and inhibits BRI1

5 function [30, 31]. Transcription factors BES1/BZR1 are phosphorylated by BIN2 and are inhibited by several mechanisms [32-36]. Binding of BR to BRI1 leads to release of BKI1, which interacts with 14-3-3 and thus promotes nuclear accumulation of BES1 [37]. Association of BRI1 with co-receptor BAK1 activates BRI1 kinase activity [38-46]. Activated BRI1 signals through BSK1 and CDG1 kinases as well as BSU1 phosphatase to inhibit BIN2 kinase activity

[47-49]. The inhibition of BIN2 and action of PP2A phosphatase allow accumulation of

BES1/BZR1 in the nucleus, which regulates gene expression in combination with other transcription regulators [50-53]. Other transcription factors are also regulated by BIN2 in response to BR [54, 55]. Multiple mechanisms function to attenuate BR signaling. Accumulated

BES1 and BZR1 repress the expression of BR biosynthesis genes and activate BR degradation enzymes as a feedback inhibition mechanism [56]. Moreover, LCMT, a Leucine C-terminal

Methyltransferase, activates PP2A, which dephosphorylates BRI1, turning off the BR signaling

[57, 58]. BRI1 auto-phosphorylation can also deactivate itself [59]. BES1 and BZR1 bind to E- box or BRRE promoter elements to regulate target gene expression [52, 53]. ChIP-chip experiments identified BES1 and BZR1 target genes, which are enriched for functions in cellular growth and other signaling pathways such as auxin, abscisic acid (ABA), and stress responses, confirming that BRs interact with many other signaling systems to regulate various biological responses [60, 61]. OsBZR1, a rice homolog of BES1/BZR1, similarly mediate BR responses

[62]. Several additional transcription factors are involved in BR responses in rice and/or

Arabidopsis [63-66]. Some of them, such as DLT [64] and LIC [67] appear unique in cereals, suggesting that not all the BR signaling components are conserved between Arabidopsis and cereals.

Gibberellins

GAs are cyclic diterpene compounds that promote stem elongation, and mutants in GA synthesis or signaling show dwarf phenotypes. The semi-dwarf habits of several green revolution crops result from mutations in GA related genes [68, 69]. For example, the Green Revolution semi-dwarf IR8 rice was a GA deficient mutant with a mutation in the GA20ox-2 gene. In maize, five recessive dwarf mutants are known to block enzymatic steps in early GA biosynthesis [13,

70-72]. In rice, GA homeostasis is maintained through a balance of anabolic (GA20ox and

GA3ox) and catabolic (GA2ox and EUI1) activities, regulated through feedback transcriptional control [73-76]. The semi-dwarf ‘green revolution’ phenotype in rice results from a mutation in the semidwarf1 (sd1) gene, which encodes a GA20oxidase (OsGA20ox2) [69], and dwarf18 (d18) dwarves result from mutations in a GA3ox gene (OsGA3ox2) [74].An increase in the amount of

EUI causes a reduction in GA levels, and consequently leads to dwarfism [77]. All the known metabolic genes except EUI have been identified in maize and most are encoded by multigene families with differential expression patterns [78].

For GA signaling, key components of the pathway have been identified in plants such as

Arabidopsis, rice, maize, and others. It was shown in rice that (GIBBERELLIN INSENSITIVE

DWARF1) GID1 is the GA receptor [79]. Different GAs have different affinities to GID1 and the one with the highest affinity is GA4, the major bioactive GA in rice [80]. Both positive and negative regulators of GA signaling have demonstrated effects on PHT. In rice, gid1-1 mutants have dark-green leaf blades with a dwarf stature resembling GA related mutants, and show greatly reduced sensitivity to applied GA [79]. GA signaling induces gene expression by targeting the DELLA proteins for degradation. In rice, GA binding by the GA receptor (GID1)

7 induces a GA-dependent interaction between GID1 and the DELLA protein SLENDER RICE1

(SLR1) [79-81]. This interaction leads to ubiquitination and proteolysis of SLR1, mediated by

GID2 [82]. Thus, GID1 and GID2 are positive regulators of GA signaling and SLR1 is a negative regulator. Loss of function mutations in GID1 or GID2 result in dwarf phenotypes because of reduced GA signaling, whereas GID1 overexpression causes GA hypersensitivity and internode elongation [79]. In the absence of GA, DELLA proteins accumulate in the nucleus and negatively regulate GA responsive gene expression [83]. In the presence of GA, DELLA proteins are degraded by 26S proteasome-mediated proteolysis, which in turn de-represses GA responsive gene transcription. DELLA proteins are negative regulators that repress GA-induced transcription, and GA signaling targets DELLA proteins for degradation, thereby de-repressing

GA responsive genes [84]. DELLA mutations that disrupt the interactions with GID1 or GID2 proteins cause dominant GA-insensitive dwarf phenotypes in all species. For example, in maize, dominant mutations in D8 or D9 DELLA genes cause GA insensitive dwarf phenotypes [68, 85], and in wheat, Rht1 similarly cause dominant, GA-insensitive dwarf phenotypes [68, 86].

Quantitative Genetics of PHT in Maize

PHT is one of the most heritable and easily measured traits in maize. Given a pedigree or estimates of the genomic identity-by-state among related plants, height is also accurately predictable [87]. But, mapping alleles explaining natural variation in maize height remains a formidable challenge as PHT is a complicated quantitative trait that is controlled by a large number of genes. PHT shows polygenic inheritance, with over 100 maize QTL reported in at least 40 genomic locations [88]. As many as 30 QTL have been detected in a single mapping experiment [89]. PHT loci have been cloned and resolved by molecular tagging of large effect

8 alleles often induced by mutagenesis [8, 90]. Over 40 maize genes at which mutations have large effects on plant height have been identified [91]. These are involved in hormone synthesis, transport, and signaling [92]. The large effects of these loci suggest they may not commonly segregate in natural populations due to rapid loss or fixation. However, admixture of exotic germplasm and locally adapted subpopulations may also allow persistent segregation of large- effect alleles due to evolutionary capacitance [93].

For individual QTL, all modes of inheritance (additive, partial, complete, or over- dominance) have been found with a tendency toward dominance for alleles that increase PHT

[94-96]. PHT shows substantial hybrid vigor and is an excellent model trait to unravel the molecular basis of heterosis [97]. Maize breeders maintain inbred heterotic groups with established general combining abilities but select individuals based upon specific hybrid performance. This practice may allow segregation of large recessive effects. QTL identified from individual mapping studies usually have large confidence intervals, obscuring co-localization of

QTL and underlying genes. 11 candidate genes were identified in meta-analyses attempting to identify the most likely position of QTL ("real QTL") consistently found across PHT studies [88,

98]. One of the likely candidate loci was na1 which functions in BR biosynthesis. However, not all known maize PHT genes mapped to QTL and for most PHT QTL, no candidate genes have yet been identified. This could be due to a lack of genetic variation for PHT genes in the QTL mapping populations or technical limitations of these studies.

The overall goal of this manuscript was to 1) introgress multiple exotic tropical and subtropical accessions into two maize heterotic groups Stiff Stalk and Non stalk to improve PHT

9 performance, 2) identify putative causative loci throughout the genome associated with PHT, BR and GA, and 3) investigate the BR and GA control of PHT from a genetics and physiological perspective. The first objective (Chapter 2) in this project was to propose a “Molecular

Strengthening” (MOST) concept to improve trait expression (such as the control of flowering time for tropical germplasm adaptation in temperate areas) from a non-genetics perspective.

Unraveling the function of genes affecting agronomic traits is accelerating due to progress in

DNA sequencing and other high-throughput genomic approaches. Characterized genes can be exploited by plant breeders by using either marker assisted selection (MAS) or transgenic procedures. As a third ‘outlet’, MOST can be used as an alternative option for exploiting detailed molecular understanding of trait expression. Informed by -omics approaches, molecules can be produced or extracted to target biological molecules or structures to manipulate crop performance, which is comparable to the pharmaceutical treatment of human diseases. The second objective (Chapter 3) was to propose a “phenotype-selected introgression library” method for utilizing exotic germplasm, which can be used for improving complex traits such as plant height. We backcrossed multiple tropical and subtropical exotic accessions with PHB47 and

PHZ51 to broaden the genetic variation of the two maize heterotic groups. At the same time, we did phenotype selection to improve PHT and select backcross families with undistinguishable flowering time as their recurrent parents, with the goal to produce pre-breeding materials with potential for biofuel maize production to be grown at temperate area. We measured PHT performance and flowering time across multiple years and locations to evaluate efficiency of this

“phenotype-selected introgression library” method. The third objective (Chapter 4) was to investigate the BR and GA control of PHT. To do this we used the two phenotype-selected introgression libraries from Chapter three to conduct a genome-wide association study to explore

10 which genomic regions are controlling PHT, BR, and GA, and test the hypothesis that there are co-localizations among these genomic regions. Moreover, we used plant hormone inhibitors to treat different maize backcross families (with different levels of heterozygosity), and correlated hormone inhibitor response with the level of heterozygosity and PHT, and tested the hypothesis that seedling stage hormone inhibitor response can be used to predict heterosis in maize PHT.

The fourth objective (Chapter 5) was to investigate the BR and GA control of seedling traits, as a high quality seedling establishment is required for stable field performance such as PHT and yield. As BRs and GAs have shown strong correlation with seedling vigor and biotic/abiotic stress tolerance, we did genome-wide association analysis across 207 diverse doubled haploid lines to investigate the genomic regions associated with BR and GA pathway that controlling seedling architecture traits. At the same time, we compared these regions to known QTL regions regulating stress tolerance such as cold tolerance. Moreover, we tested the hypothesis that seedling stage hormone level can be used to predict PHT, yield and flowering time for an inbred genetic background.

Organization of the thesis

This thesis contains two published research articles (Chapters 2 and 3) and one manuscripts under review (Chapter 4), and one manuscript in preparation (Chapter 5). The conclusions of all studies are summarized in a final chapter (Chapter 6). As each chapter contains its own introduction, the general introduction was kept brief. Literature for each individual experiment and procedure is introduced and discussed within the respective chapters.

REFERENCES

1. Godfray HCJ, Beddington JR, Crute IR, Haddad L, Lawrence D, Muir JF et al. Food security: the challenge of feeding 9 billion people. Science. 2010;327:812-8.

2. Alston JM, Beddow JM, Pardey PG. Agricultural research, productivity, and food prices in the long run. Science. 2009;325:1209-10.

3. Tester M, Langridge P. Breeding technologies to increase crop production in a changing world. Science. 2010;327:818-22. doi:10.1126/science.1183700.

4. Costes E, Lauri PE, Simon S, Andrieu B. Plant architecture, its diversity and manipulation in agronomic conditions, in relation with pest and pathogen attacks. Eur J Plant Pathol. 2013;135:455-70. doi:10.1007/s10658-012-0158-3.

5. Duvick D. Genetic progress in yield of United States maize (Zea mays L.). Maydica. 2005;50:193.

6. Duvick D, Smith J, Cooper M. Long term selection in a commercial hybrid maize breeding program. Janick I Plant Breeding Reviews Part. 2010;2:109-52.

7. Lubberstedt T, Melchinger AE, Schon CC, Utz HF, Klein D. QTL mapping in testcrosses of European flint lines of maize .1. Comparison of different testers for forage yield traits. Crop Sci. 1997;37:921-31.

8. Fernandez MGS, Becraft PW, Yin Y, Lübberstedt T. From dwarves to giants? Plant height manipulation for biomass yield. Trends Plant Sci. 2009;14:454-61.

9. Sakamoto T, Matsuoka M. Generating high-yielding varieties by genetic manipulation of plant architecture. Curr Opin Biotechnol. 2004;15:144-7. doi:10.1016/j.copbio.2004.02.003.

10. Salas Fernandez MG, Becraft PW, Yin Y, Lübberstedt T. From dwarves to giants? Plant height manipulation for biomass yield. Trends Plant Sci. 2009;14:454-61.

11. Ragauskas AJ, Williams CK, Davison BH, Britovsek G, Cairney J, Eckert CA et al. The path forward for biofuels and biomaterials. Science. 2006;311:484-9.

12. Somerville C. Biofuels. Curr Biol. 2007;17:R115-R9.

13. Winkler RG, Helentjaris T. The maize Dwarf3 gene encodes a cytochrome P450-mediated early step in Gibberellin biosynthesis. Plant Cell. 1995;7:1307-17. doi:10.1105/tpc.7.8.1307.

14. Lawit SJ, Wych HM, Xu D, Kundu S, Tomes DT. Maize DELLA proteins dwarf plant8 and dwarf plant9 as modulators of plant development. Plant and cell physiology. 2010;51:1854-68.

15. Thornsberry JM, Goodman MM, Doebley J, Kresovich S, Nielsen D, Buckler ESt. Dwarf8 polymorphisms associate with variation in flowering time. Nat Genet. 2001;28:286-9. doi:10.1038/90135.

16. Hartwig T, Chuck GS, Fujioka S, Klempien A, Weizbauer R, Potluri DPV et al. Brassinosteroid control of sex determination in maize. Proceedings of the National Academy of Sciences. 2011;108:19814-9.

17. Makarevitch I, Thompson A, Muehlbauer GJ, Springer NM. Brd1 gene in maize encodes a brassinosteroid C-6 oxidase. Plos One. 2012;7:e30798. doi:10.1371/journal.pone.0030798.

18. Fujioka S, Yokota T. Biosynthesis and metabolism of brassinosteroids. Annu Rev Plant Biol. 2003;54:137-64.

19. Clouse SD. Molecular genetic studies confirm the role of brassinosteroids in plant growth and development. Plant J. 1996;10:1-8. doi:DOI 10.1046/j.1365-313X.1996.10010001.x.

20. Mandava NB. Plant growth-promoting brassinosteroids. Annu Rev Plant Physiol Plant Mol Biol. 1988;39:23-52.

21. Bishop GJ. Brassinosteroid mutants of crops. J Plant Growth Regul. 2003;22:325-35.

22. Tao Y, Zheng J, Xu Z, Zhang X, Zhang K, Wang G. Functional analysis of ZmDWF1, a maize homolog of the Arabidopsis brassinosteroids biosynthetic DWF1/DIM gene. Plant Sci. 2004;167:743-51.

23. Wu C-y, Trieu A, Radhakrishnan P, Kwok SF, Harris S, Zhang K et al. Brassinosteroids regulate grain filling in rice. The Plant Cell. 2008;20:2130-45.

24. Clouse SD. Brassinosteroid Signal Transduction: From Receptor Kinase Activation to Transcriptional Networks Regulating Plant Development. Plant Cell. 2011;23:1219-30. doi:DOI 10.1105/tpc.111.084475.

25. Ye H. Recent Advances in the Regulation of Brassinosteroid Signaling and Biosynthesis PathwaysF. Journal of integrative plant biology. 2011;53:455.

26. Kim T-W, Wang Z-Y. Brassinosteroid Signal Transduction from Receptor Kinases to Transcription Factors. In: Merchant S, Briggs WR, Ort D, editors. Annual Review of Plant Biology, Vol 61. Annual Review of Plant Biology, 2010. p. 681-704.

27. Nakamura A, Fujioka S, Sunohara H, Kamiya N, Hong Z, Inukai Y et al. The role of OsBRI1 and its homologous genes, OsBRL1 and OsBRL3, in rice. Plant Physiol. 2006;140:580-90.

28. Yamamuro C, Ihara Y, Wu X, Noguchi T, Fujioka S, Takatsuto S et al. Loss of function of a rice brassinosteroid insensitive1 homolog prevents internode elongation and bending of the lamina joint. The Plant Cell. 2000;12:1591-605.

29. Chono M, Honda I, Zeniya H, Yoneyama K, Saisho D, Takeda K et al. A semidwarf phenotype of barley uzu results from a nucleotide substitution in the gene encoding a putative brassinosteroid receptor. Plant Physiol. 2003;133:1209-19.

30. Jaillais Y, Chory J. Unraveling the paradoxes of plant hormone signaling integration. Nat Struct Mol Biol. 2010;17:642-5.

31. Wang X, Chory J. Brassinosteroids regulate dissociation of BKI1, a negative regulator of BRI1 signaling, from the plasma membrane. Science. 2006;313:1118-22.

32. Choe S, Schmitz RJ, Fujioka S, Takatsuto S, Lee M-O, Yoshida S et al. Arabidopsis Brassinosteroid-Insensitive dwarf12Mutants Are Semidominant and Defective in a Glycogen Synthase Kinase 3β-Like Kinase. Plant Physiol. 2002;130:1506-15.

33. He J-X, Gendron JM, Yang Y, Li J, Wang Z-Y. The GSK3-like kinase BIN2 phosphorylates and destabilizes BZR1, a positive regulator of the brassinosteroid signaling pathway in Arabidopsis. Proceedings of the National Academy of Sciences. 2002;99:10185-90.

34. Li J, Nam KH. Regulation of brassinosteroid signaling by a GSK3/SHAGGY-like kinase. Science. 2002;295:1299-301.

35. Wang Z-Y, Nakano T, Gendron J, He J, Chen M, Vafeados D et al. Nuclear-localized BZR1 mediates brassinosteroid-induced growth and feedback suppression of brassinosteroid biosynthesis. Dev Cell. 2002;2:505-13.

36. Yin Y, Wang Z-Y, Mora-Garcia S, Li J, Yoshida S, Asami T et al. BES1 accumulates in the nucleus in response to brassinosteroids to regulate gene expression and promote stem elongation. Cell. 2002;109:181-91.

37. Wang X, Zeng J, Shi M, Zhao S, Bai W, Cao W et al. Targeted blockage of signal transducer and activator of transcription 5 signaling pathway with decoy oligodeoxynucleotides suppresses leukemic K562 cell growth. DNA Cell Biol. 2011;30:71-8. doi:10.1089/dna.2010.1112.

38. Gou X, Yin H, He K, Du J, Yi J, Xu S et al. Genetic evidence for an indispensable role of somatic embryogenesis receptor kinases in brassinosteroid signaling. PLoS Genet. 2012;8:e1002452. doi:10.1371/journal.pgen.1002452

PGENETICS-D-11-01756 [pii].

39. Oh MH, Wang X, Kota U, Goshe MB, Clouse SD, Huber SC. Tyrosine phosphorylation of the BRI1 receptor kinase emerges as a component of brassinosteroid signaling in Arabidopsis. Proc Natl Acad Sci U S A. 2009;106:658-63.

40. Oh MH, Wu X, Clouse SD, Huber SC. Functional importance of BAK1 tyrosine phosphorylation in vivo. Plant Signal Behav. 2011;6. doi:14337 [pii].

41. Wang X, Kota U, He K, Blackburn K, Li J, Goshe MB et al. Sequential transphosphorylation of the BRI1/BAK1 receptor kinase complex impacts early events in brassinosteroid signaling. Dev Cell. 2008;15:220-35.

42. Wang X, Li X, Meisenhelder J, Hunter T, Yoshida S, Asami T et al. Autoregulation and homodimerization are involved in the activation of the plant steroid receptor BRI1. Dev Cell. 2005;8:855-65. doi:S1534-5807(05)00170-X [pii]

10.1016/j.devcel.2005.05.001.

43. Wang X, Goshe MB, Soderblom EJ, Phinney BS, Kuchar JA, Li J et al. Identification and functional analysis of in vivo phosphorylation sites of the Arabidopsis BRASSINOSTEROID- INSENSITIVE1 receptor kinase. Plant Cell. 2005;17:1685-703.

44. Russinova E, Borst JW, Kwaaitaal M, Cano-Delgado A, Yin Y, Chory J et al. Heterodimerization and endocytosis of Arabidopsis brassinosteroid receptors BRI1 and AtSERK3 (BAK1). Plant Cell. 2004;16:3216-29. doi:tpc.104.025387 [pii]

10.1105/tpc.104.025387.

45. Nam KH, Li J. BRI1/BAK1, a receptor kinase pair mediating brassinosteroid signaling. Cell. 2002;110:203-12.

46. Li J, Wen J, Lease KA, Doke JT, Tax FE, Walker JC. BAK1, an Arabidopsis LRR receptor- like protein kinase, interacts with BRI1 and modulates brassinosteroid signaling. Cell. 2002;110:213-22.

47. Kim T-W, Guan S, Burlingame AL, Wang Z-Y. The CDG1 kinase mediates brassinosteroid signal transduction from BRI1 receptor kinase to BSU1 phosphatase and GSK3-like kinase BIN2. Mol Cell. 2011;43:561-71.

48. Kim T-W, Guan S, Sun Y, Deng Z, Tang W, Shang J-X et al. Brassinosteroid signal transduction from cell-surface receptor kinases to nuclear transcription factors. Nat Cell Biol. 2009;11:1254-60.

49. Tang W, Kim T-W, Oses-Prieto JA, Sun Y, Deng Z, Zhu S et al. BSKs mediate signal transduction from the receptor kinase BRI1 in Arabidopsis. Science. 2008;321:557-60.

50. Tang W, Yuan M, Wang R, Yang Y, Wang C, Oses-Prieto JA et al. PP2A activates brassinosteroid-responsive gene expression and plant growth by dephosphorylating BZR1. Nat Cell Biol. 2011;13:124-31.

51. Ryu H, Kim K, Cho H, Hwang I. Predominant actions of cytosolic BSU1 and nuclear BIN2 regulate subcellular localization of BES1 in brassinosteroid signaling. Molecules and cells. 2010;29:291-6.

52. Yin Y, Vafeados D, Tao Y, Yoshida S, Asami T, Chory J. A new class of transcription factors mediates brassinosteroid-regulated gene expression in Arabidopsis. Cell. 2005;120:249- 59.

53. He J-X, Gendron JM, Sun Y, Gampala SS, Gendron N, Sun CQ et al. BZR1 is a transcriptional repressor with dual roles in brassinosteroid homeostasis and growth responses. Science. 2005;307:1634-8.

54. Tong H, Liu L, Jin Y, Du L, Yin Y, Qian Q et al. DWARF AND LOW-TILLERING acts as a direct downstream target of a GSK3/SHAGGY-like kinase to mediate brassinosteroid responses in rice. The Plant Cell. 2012;24:2562-77.

55. Ye H, Li L, Guo H, Yin Y. MYBL2 is a substrate of GSK3-like kinase BIN2 and acts as a corepressor of BES1 in brassinosteroid signaling pathway in Arabidopsis. Proceedings of the National Academy of Sciences. 2012;109:20142-7.

56. Ye H, Li L, Yin Y. Recent Advances in the Regulation of Brassinosteroid Signaling and Biosynthesis PathwaysF. Journal of integrative plant biology. 2011;53:455-68.

57. Wu G, Wang X, Li X, Kamiya Y, Otegui MS, Chory J. Methylation of a phosphatase specifies dephosphorylation and degradation of activated brassinosteroid receptors. Science signaling. 2011;4:ra29.

58. Di Rubbo S, Irani NG, Russinova E. PP2A phosphatases: the “on-off” regulatory switches of brassinosteroid signaling. Science signaling. 2011;4:pe25-pe.

59. Oh M-H, Wang X, Clouse SD, Huber SC. Deactivation of the Arabidopsis BRASSINOSTEROID INSENSITIVE 1 (BRI1) receptor kinase by autophosphorylation within the glycine-rich loop. Proceedings of the National Academy of Sciences. 2012;109:327-32.

60. Sun Y, Fan X-Y, Cao D-M, Tang W, He K, Zhu J-Y et al. Integration of brassinosteroid signal transduction with the transcription network for plant growth regulation in Arabidopsis. Dev Cell. 2010;19:765-77.

61. Yu X, Li L, Zola J, Aluru M, Ye H, Foudree A et al. A brassinosteroid transcriptional network revealed by genome‐wide identification of BESI target genes in Arabidopsis thaliana. The Plant Journal. 2011;65:634-46.

62. Bai M-Y, Zhang L-Y, Gampala SS, Zhu S-W, Song W-Y, Chong K et al. Functions of OsBZR1 and 14-3-3 proteins in brassinosteroid signaling in rice. Proceedings of the National Academy of Sciences. 2007;104:13839-44.

63. Duan K, Li L, Hu P, Xu SP, Xu ZH, Xue HW. A brassinolide‐suppressed rice MADS‐box transcription factor, OsMDP1, has a negative regulatory role in BR signaling. The Plant Journal. 2006;47:519-31.

64. Tong H, Jin Y, Liu W, Li F, Fang J, Yin Y et al. DWARF AND LOW‐TILLERING, a new member of the GRAS family, plays positive roles in brassinosteroid signaling in rice. The Plant Journal. 2009;58:803-16.

65. Wang H, Zhu Y, Fujioka S, Asami T, Li J, Li J. Regulation of Arabidopsis brassinosteroid signaling by atypical basic helix-loop-helix proteins. The Plant Cell. 2009;21:3781-91.

66. Luo X-M, Lin W-H, Zhu S, Zhu J-Y, Sun Y, Fan X-Y et al. Integration of light-and brassinosteroid-signaling pathways by a GATA transcription factor in Arabidopsis. Dev Cell. 2010;19:872-83.

67. Zhang C, Xu Y, Guo S, Zhu J, Huan Q, Liu H et al. Dynamics of brassinosteroid response modulated by negative regulator LIC in rice. PLoS Genet. 2012;8:e1002686.

68. Peng J, Richards DE, Hartley NM, Murphy GP, Devos KM, Flintham JE et al. ‘Green revolution’genes encode mutant gibberellin response modulators. Nature. 1999;400:256-61.

69. Sasaki A, Ashikari M, Ueguchi-Tanaka M, Itoh H, Nishimura A, Swapan D et al. Green revolution: a mutant gibberellin-synthesis gene in rice. Nature. 2002;416:701-2.

70. Yamaguchi S. Gibberellin metabolism and its regulation. Annu Rev Plant Biol. 2008;59:225- 51.

71. Bensen RJ, Johal GS, Crane VC, Tossberg JT, Schnable PS, Meeley RB et al. Cloning and characterization of the maize An1 gene. Plant Cell. 1995;7:75-84. doi:10.1105/tpc.7.1.75.

72. Fujioka S, Yamane H, Spray CR, Gaskin P, Macmillan J, Phinney BO et al. Qualitative and Quantitative Analyses of Gibberellins in Vegetative Shoots of Normal, dwarf-1, dwarf-2, dwarf- 3, and dwarf-5 Seedlings of Zea mays L. Plant Physiol. 1988;88:1367-72.

73. Yamaguchi S. Gibberellin metabolism and its regulation. Annu Rev Plant Biol. 2008;59:225- 51.

74. Itoh H, Ueguchi-Tanaka M, Sentoku N, Kitano H, Matsuoka M, Kobayashi M. Cloning and functional analysis of two gibberellin 3β-hydroxylase genes that are differently expressed during the growth of rice. Proceedings of the National Academy of Sciences. 2001;98:8909-14.

75. Lo S-F, Yang S-Y, Chen K-T, Hsing Y-I, Zeevaart JA, Chen L-J et al. A novel class of gibberellin 2-oxidases control semidwarfism, tillering, and root development in rice. The Plant Cell. 2008;20:2603-18.

76. Zhu Y, Nomura T, Xu Y, Zhang Y, Peng Y, Mao B et al. ELONGATED UPPERMOST INTERNODE encodes a cytochrome P450 monooxygenase that epoxidizes gibberellins in a novel deactivation reaction in rice. The Plant Cell. 2006;18:442-56.

77. Zhang Y, Zhu Y, Peng Y, Yan D, Li Q, Wang J et al. Gibberellin homeostasis and plant height control by EUI and a role for gibberellin in root gravity responses in rice. Cell Res. 2008;18:412-21.

78. Song J, Guo B, Song F, Peng H, Yao Y, Zhang Y et al. Genome-wide identification of gibberellins metabolic enzyme genes and expression profiling analysis during seed germination in maize. Gene. 2011;482:34-42.

79. Ueguchi-Tanaka M, Ashikari M, Nakajima M, Itoh H, Katoh E, Kobayashi M et al. GIBBERELLIN INSENSITIVE DWARF1 encodes a soluble receptor for gibberellin. Nature. 2005;437:693-8.

80. Ueguchi-Tanaka M, Nakajima M, Katoh E, Ohmiya H, Asano K, Saji S et al. Molecular interactions of a soluble gibberellin receptor, GID1, with a rice DELLA protein, SLR1, and gibberellin. The Plant Cell. 2007;19:2140-55.

81. Ikeda A, Ueguchi-Tanaka M, Sonoda Y, Kitano H, Koshioka M, Futsuhara Y et al. slender rice, a constitutive gibberellin response mutant, is caused by a null mutation of the SLR1 gene, an ortholog of the height-regulating gene GAI/RGA/RHT/D8. The Plant Cell. 2001;13:999-1010.

82. Sasaki A, Itoh H, Gomi K, Ueguchi-Tanaka M, Ishiyama K, Kobayashi M et al. Accumulation of phosphorylated repressor for gibberellin signaling in an F-box mutant. Science. 2003;299:1896-8.

83. de Lucas M, Davière J-M, Rodríguez-Falcón M, Pontin M, Iglesias-Pedraz JM, Lorrain S et al. A molecular framework for light and gibberellin control of cell elongation. Nature. 2008;451:480-4.

84. Zentella R, Zhang Z-L, Park M, Thomas SG, Endo A, Murase K et al. Global analysis of DELLA direct targets in early gibberellin signaling in Arabidopsis. The Plant Cell. 2007;19:3037-57.

85. Lawit SJ, Wych HM, Xu DP, Kundu S, Tomes DT. Maize DELLA Proteins dwarf plant8 and dwarf plant9 as Modulators of Plant Development. Plant and Cell Physiology. 2010;51:1854-68. doi:Doi 10.1093/Pcp/Pcq153.

86. Winkler RG, Freeling M. Physiological genetics of the dominant gibberellin-nonresponsive maize dwarfs, Dwarf8 and Dwarf9. Planta. 1994;193:341-8.

87. Peiffer JA, Romay MC, Gore MA, Flint-Garcia SA, Zhang Z, Millard MJ et al. The genetic architecture of maize height. Genetics. 2014;196:1337-56. doi:10.1534/genetics.113.159152.

88. Wang Y, Yao J, Zhang Z, Zheng Y. The comparative analysis based on maize integrated QTL map and meta-analysis of plant height QTLs. Chin Sci Bull. 2006;51:2219-30.

89. Schön CC, Utz HF, Groh S, Truberg B, Openshaw S, Melchinger AE. Quantitative trait locus mapping based on resampling in a vast maize testcross experiment and its relevance to quantitative genetics for complex traits. Genetics. 2004;167:485-98.

90. Andorf CM, Lawrence CJ, Harper LC, Schaeffer ML, Campbell DA, Sen TZ. The Locus Lookup tool at MaizeGDB: identification of genomic regions in maize by integrating sequence information with physical and genetic maps. Bioinformatics. 2010;26:434-6.

91. Multani DS, Briggs SP, Chamberlin MA, Blakeslee JJ, Murphy AS, Johal GS. Loss of an MDR transporter in compact stalks of maize br2 and sorghum dw3 mutants. Science. 2003;302:81-4. doi:10.1126/science.1086072.

92. Wang Y, Li J. Molecular basis of plant architecture. Annu Rev Plant Biol. 2008;59:253-79.

93. Masel J. Evolutionary capacitance may be favored by natural selection. Genetics. 2005;170:1359-71. doi:DOI 10.1534/genetics.105.040493.

94. Beavis WD, Grant D, Albertsen M, Fincher R. Quantitative Trait Loci for Plant Height in 4 Maize Populations and Their Associations with Qualitative Genetic-Loci. Theor Appl Genet. 1991;83:141-5.

95. Lin Y-R, Schertz KF, Paterson AH. Comparative analysis of QTLs affecting plant height and maturity across the Poaceae, in reference to an interspecific sorghum population. Genetics. 1995;141:391-411.

96. Pereira M, Lee M. Identification of genomic regions affecting plant height in sorghum and maize. Theor Appl Genet. 1995;90:380-8.

97. Użarowska A, Keller B, Piepho H-P, Schwarz G, Ingvardsen C, Wenzel G et al. Comparative expression profiling in meristems of inbred-hybrid triplets of maize based on morphological investigations of heterosis for plant height. Plant Mol Biol. 2007;63:21-34.

98. Chardon F, Virlon B, Moreau L, Falque M, Joets J, Decousset L et al. Genetic architecture of flowering time in maize as inferred from quantitative trait loci meta-analysis and synteny conservation with the rice genome. Genetics. 2004;168:2169-85. doi:DOI 10.1534/genetics.104.032375.

CHAPTER 2.

GETTING THE ‘MOST’ OUT OF CROP IMPROVEMENT

Songlin Hu1 and Thomas Lubberstedt1

Paper published in Trends in Plant Science Journal. Abstract, structure and references are all formatted according to journal standards.

SH and TL contributed equally to this work and finished the manuscript together.

Abstract

Unraveling the function of genes affecting agronomic traits is accelerating due to progress in DNA sequencing and other high-throughput genomic approaches. Characterized genes can be exploited by plant breeders by using either marker-aided selection (MAS) or transgenic procedures. Here, we propose a third ‘outlet’, ‘molecular strengthening’ (MOST), as alternative option for exploiting detailed molecular understanding of trait expression, which is comparable to the pharmaceutical treatment of human diseases. MOST treatments can be used to enhance yield stability. Alternatively, they can be used to control traits temporally, such as flowering time to facilitate crosses for plant breeders. We also discuss the essence for developing

MOST treatments, their prospects, and limitations.

Understanding of The Molecular Basis of Traits Is Accumulating

The amount of sequences deposited in GenBank (ftp://ftp. ncbi.nih.gov/genbank/gbrel.txt) increased from 11 billion to 165 billion base pairs from 2000 to 2014. Due to progress in high- throughput sequencing approaches (next- and third-generation sequencing [1,2]), resequencing of well characterized species has become affordable [3,4], as well as de novo sequencing of species without reference genomes [5,6]. Even for polyploid species, major breakthroughs have been reported [7]. Resulting DNA sequence information can be converted into effective molecular marker applications [8,9] to generate genetic and physical maps, thus facilitating mapping of agronomic traits [10,11]. Alternatively, sequencing can be used directly for genome- wide association studies (GWAS; see Glossary) [12,13] or candidate gene-based association studies [14,15] to identify genes and intragenic polymorphisms associated with target traits. As a result, the number of discoveries of novel genes has increased exponentially [16,17]. In addition, functional [18] and comparative genomic studies [7,19] benefit substantially from progress in sequencing technology, contributing to the accelerated isolation and characterization of genes with impacts on agronomic traits [20,21]. Moreover, due to rapid progress in transcriptome analysis [22,23], chemical genomics [24,25], proteomics [26,27], metabolomics [28,29], and emerging hormone studies [30,31], understanding of gene and protein function as well as gene networks and metabolic pathways has increased dramatically over the past decade. Traditionally, detailed molecular genetic information of trait expression is exploited by either MAS or transgenic procedures. However, the recent implementation of genomic prediction and selection

[32–34] has caused a paradigm shift in plant breeding that questions the need for understanding the underlying molecular genetics (i.e., gene function or pathways) of agronomic traits. It was shown initially in animals and later in plant breeding, that a ‘black box approach’ based on low-

21 cost and dense coverage markers across the genome [35,36] can efficiently enhance breeding programs. This raises two questions in relation to crop improvement: (i) How relevant is the understanding of gene and pathway information for agronomic trait improvement? (ii) In which ways can the accumulating -omics information be exploited, besides providing low cost markers?

Comparable to medical genomics, we propose MOST treatments as alternative and potentially superior routes for exploiting molecular information in plants. Our objectives here are to: (i) systematically introduce the concept of MOST; (ii) compare the advantages and disadvantages of the three avenues for exploiting understanding of the molecular basis of agronomic trait expression (MAS, MOST, and transgenic methods); and (iii) use examples to illustrate the significance of MOST treatments.

Three Major Avenues to Use Molecular Genomic Information for Crop Improvement

Knowledge of genes affecting agronomic traits is successfully exploited in transgenic and

MAS approaches for crop improvement. However, in human genetics, pharmaceutical genomics is the primary route for exploiting gene information [37–40], turning key molecules into drugs to provide health benefits [41,42]. This pharmaceutical approach is less explored in crop improvement, but has the potential to become a third important outlet to fully utilize the accumulating genomic information (Figure 1).

MAS

Once a gene has been identified, linked, gene-derived, or functional markers [43] can be used as diagnostic tools for breeders to select valuable parents or progenies. Currently, major

MAS applications include marker assisted backcrossing, marker-assisted introgression, gene pyramiding, marker-assisted recurrent selection, and F2 enrichment [44,45]. Successes based on

MAS have been reported, such as food quality improvement in tomato (Solanum lycopersicum)

[46], yield improvement in maize (Zea mays) under water-limited conditions [47], gene pyramiding in rice (Oryza sativa) [48], and the development of disease-resistant soybeans

(Glycine max (L.) Merr) [49]. However, successes recorded so far have mainly been for genes with major effects. By contrast, most agronomic traits are quantitatively inherited, with minor contributions from many genes. For those traits, genomic selection seems to be more promising

[32], because it does not require prior characterization of genes or quantitative trait loci (QTL) affecting traits of interest.

Transgenics

Since the release of the first genetically modified (GM) crop variety in 1992 (‘FlavrSavr’ tomato), a massive expansion of GM crops has occurred. In 2013, 173 million hectares were planted with commercial GM crops (http://isaaa.org/ resources/publications/pocketk/16/default.asp), especially for maize, soybean, and cotton

(Gossypium spp.) [50]. Most of the transgenes that were introduced to these crops are related to pest, disease, or herbicide resistance. The Bt trait [a transgenic trait conferring resistance to lepidopteran insects found in Bacillus thuringiensis (Bt)] in cotton reduces costs relating to pesticides and labor, increases yield, and, thus, leads to economic and health benefits for farmers

[51]. The main advantage of transgenic approaches is that they can utilize genes from other species, introducing novel beneficial traits. However, legal approval for a new transgenic product

23 costs approximately US$35 million for the regulatory process (http://www.isaaa.org/ gmapprovaldatabase/default.asp). As a consequence, only a few transgenic events promising sufficient return on investment are used for crop improvement. This limits the number of genes and crop species for which the use of transgenic varieties is economically feasible. Moreover, public concern about the safety and environmental impact of GM crops prohibits their use in various countries. Novel genome-editing methods [52] might provide a fresh perspective with acceptance of transgene products due to a more targeted manipulation of genomes.

MOST

We will use the term ‘MOST’ for the non-genetic manipulation of crop improvement, based on understanding of molecular mechanisms. Informed by -omics approaches, molecules can be produced or extracted to target biological molecules or structures to manipulate crop performance, similar to medical treatment of humans. MOST treatments can be used to enhance yield stability under different abiotic or biotic stress conditions, to reduce fertilizer input, and to modulate plant developmental or morphological traits, such as flowering time. The idea of plant treatment is not novel: from the 1940s to the late 1970s, application of ethylene, gibberellins, auxins, cytokinins, and their antagonists and protagonists was used extensively by plant biologists following the discovery that 2,4- dichlorophenoxyacetic acid (2-4-D) induces plant growth [53]. Application of chlorocholine chloride (CCC) to avoid wheat lodging was prevalent during the early 1960s, although CCC was initially discovered to reduce growth. Only several years later was its relation to gibberellic acid (GA) synthesis inhibition uncovered [54]. In contrast to these traditional studies, MOST is a sequence-based ‘reverse crop improvement’

24 strategy, which depends on an understanding of gene function or molecular pathways, while earlier work on phytohormones precedes this knowledge; in addition, MOST is not focused on plant hormones. Instead, it includes a broad range of effector molecules, such as DNA, RNA, proteins, metabolites, and plant extracts. In addition, nanoparticle approaches provide novel opportunities to deliver proteins or other molecules into plant cells.

Comparison of MAS, transgenic procedures, and MOST treatments

MAS benefits from new genotyping technologies, and the price per marker data point has dropped substantially over the past decade. MAS is routine in commercial breeding programs for simple and oligogenic inherited traits. However, the identification of minor-effect QTL and the pyramiding of those into elite germplasm did not appear to be a feasible strategy, which is why such QTLs were replaced by genomic selection (GS). The major advantage of transgenic approaches compared with MAS and MOST procedures is that it is possible to create novel trait expressions that are unavailable in the original plant species. However, costs and public concerns are often prohibitive to using this option (Table 1). The essential difference between MOST and

MAS and transgenic procedures is that the latter two approaches act at the genetic level to improve genetic potential of breeding materials over multiple generations, whereas MOST does not affect genetic potential, but rather aims at maximizing trait expression for any given genotype (e.g., to counteract environmental conditions). Alternatively, MOST might override genetic determination to manipulate traits temporarily (e.g., flowering control for controlled crosses by plant breeders). In contrast to MAS and transgenic methods, which need multiple generations for the introgression of favorable alleles or transgenic events into improved genetic

25 backgrounds, once a new MOST treatment is established, it may become broadly applicable across genotypes and species without further delay. However, MOST has some potential drawbacks, such as efficient ways of application, adverse effects, and potential interactions with environmental conditions. Given the history of the negative impact on human and environmental conditions, such as the application of 2-4-D, dioxin, and dichlorodiphenyltrichloroethane (DDT), novel MOST treatments need to be carefully examined before field applications can go ahead.

Development of MOST Treatments

The development of MOST treatments starts with a thorough understanding of the underlying molecular components and mechanisms (i.e., genes, RNAs, proteins, and pathways) affecting the expression of a particular trait. This molecular information can then be translated into potential effectors, that is, molecules that activate, modulate, or inactivate the action of specific molecular targets (Figure 2). After a perfect ‘hit’ between effectors and molecular targets is identified, extensive field trials are needed before MOST treatments can be developed that are applicable under practical conditions.

Molecular targets

Plant molecules as potential molecular targets include DNA, RNA, proteins, and lipids.

DNA targets include regulatory sequences, sites underlying epigenetic regulation, and binding sites for transcription factors. Coding RNA (mRNA) and noncoding RNA (tRNA and rRNA) can be targeted to control protein synthesis. Other RNA classes are siRNA and miRNA, which affect

26 the abiotic stress response [55,56]. Protein targets can have a functional or structural role.

Functional proteins include enzymes, transcription factors, transport proteins, and receptors.

Structural proteins, such as cell wall proteins, are crucial for support and cellular motility functions [57–59]. Lipids, such as phospholipids, are key components of cell membranes and important players in plant metabolism [59,60]. Steroids as functional lipids may act as hormones or hormone precursors. Moreover, metabolites, including hormones, can be molecular targets, as well as part of DNA–protein, RNA–protein, protein–protein, and higher-order complexes. For developing MOST strategies, the preferred molecular target is specifically affected by a corresponding effector. If the molecular target is involved in multiple biological processes, implications for other traits need to be considered. For medical research, next generation sequencing (NGS) data have been successfully used in the early stages of target identification

[61]. RNAseq data from whole human exomes can help to identify potential new targets for disease pathology for rare Mendelian disorders [62].

Effectors

DNA, RNA, protein, hormones, plant extracts, and chemically synthesized molecules can function as effectors to affect molecular targets and metabolic pathways to manipulate trait expression. The transient introduction of DNA (e.g., by Agro-infiltration [63]) may stimulate ectopic expression of pivotal proteins or metabolites. In contrast to transgenic methods, DNA as an effector is transiently expressed in particular target cells without genome integration. RNAi is a useful tool for gene knockdown and the fact that siRNA and miRNA are both naturally used by cells for gene expression regulation means that RNA effectors will be nontoxic and highly

27 effective. Designed double stranded RNA (dsRNA) can be used as effectors to reduce the expression of unfavorable gene sets in a targeted way. Alternatively, RNA silencing can be used as a plant immune system against virus infection [64]. For the purpose of medical treatments, the use of protein-based drugs has dramatically increased since the introduction of insulin. In plants, proteins or recombinant proteins can be applied to execute essential functions, or to interact with potential targets. However, it is harder to deliver proteins into plant cells than into animal cells because of the plant cell wall. Nanoparticle approaches for delivering proteins directly into plant cells offer great insights into the delivery process for plants [65]. Transcription factors or enzymes can either modulate gene expressions or affect important biochemical reactions. Plant hormones are involved in key developmental processes and mediate responses to both biotic and abiotic stresses throughout the whole life cycle of a plant [66]. Hormones not only orchestrate intrinsic genetic programs, but also integrate environmental cues to adapt to environments. By controlling hormone abundances, crop performance can be optimized [67]. Plant extracts have shown to be efficient in controlling enzymes related to disease resistance [68]. Instead of using natural biological molecules, synthesized molecules can be designed and applied. For example, some chemicals could act as hormone inhibitors to control hormone synthesis and signaling pathways.

Systematic discovery of target–effector hits

For medical research, improper target selection may lead to a high rate of failure for drug development [69]. Thus, the first critical step to develop a new MOST treatment is proper target identification and validation. Accumulating sequence-based -omics information provides

28 massive opportunities for target selection. Bioinformatic tools combined with genetic and biological models can be used to predict and prioritize molecular targets [70]. For target validation, emerging genome-editing techniques are promising. Based on zinc-finger nucleases

(ZFNs), transcription activator-like endonucleases (TALENs), and clustered regularly interspaced short palindromic repeat [CRISPR)/ CRISPR-associated protein 9 (Cas9)] systems, it is possible to introduce or revert mutations at specific sites in a genome and in key regulatory proteins [71]. Subtle target site modification can validate predicted phenotypes. After that, chemical genomics or chemogenomics can test interactions between molecular targets and small molecules to identify hits [72]. For example, a transmembrane protein BIL4 related to the control of cell elongation was found to be affected by the brassinosteroid synthesis inhibitor brassinazole

[73]. A chemically synthesized molecule Gravicin has been shown to interact with PGP19, a member of abscisic acid (ABA) transporters in Arabidopsis thaliana [74]. Bioassays to screen diverse chemical, bioactive molecule collections for their effect on particular cells or plants in vitro provide a strong basis [75] for future MOST developments. Alternatively, in silico tools can be used for identifying hits, and these have been used extensively for drug discoveries [76]. For example, TargetHunter has been developed for predicting bio-targets of chemical compounds

[77]. This web-based target prediction tool implements an in silico target prediction algorism, fishing targets and similar counterparts by exploring the largest available chemogenomic databases, reaching a prediction accuracy of 91.1%. NGS technologies have been successfully applied in high-throughput compound screening by using a method called ‘encoded library technology’ [78]. This method searches for compounds that are covalently linked to short oligonucleotide labels. By sequencing oligonucleotide tags, chemical compounds can be identified in chemical libraries by affinity selection. This hit discovery process can produce

29 results specific for different organs or developmental stages, thus offering clues about when and where to apply MOST.

Application of MOST treatments

MOST treatments have to be economic and easily applicable, such as by seed coating or foliar spraying. MOST application might be optional (e.g., no drought treatment in case of sufficient water supply). In addition, the effector should be stable across environments, easily absorbed by plants, and degradable. Extensive field examinations are a prerequisite to establish robust MOST treatments. It is important that potentially adverse influences of effectors on unrelated traits are examined. As for other agricultural applications (such as use of pesticides),

MOST treatments need to meet country-specific regulations, such as the Environmental

Protection Agency in the USA (http:// www.epa.gov/) with regard to potential ecological and human health risks, including adverse effects on soil, water systems, and wildlife species.

Effectors should be biodegradable so that residues will not enter the food chain. Potential hazard identification, dose response, exposure assessment, and risk characterization of MOST treatments need to be examined for either field handlers or final product consumers. The registration of a new pesticide costs approximately US$500 000 (http://www2. epa.gov/pesticide-registration), and is likely to be similar for any new MOST treatment.

Compared with transgenic approaches, for which each new event needs to be approved and registered, once a novel MOST treatmentis approved, it can be applied broadly and extensively to different species.

Examples

MOST treatments can be used to induce disease resistance (Box 1). Alternatively, MOST treatments can be used to provide tolerance to abiotic stress factors, to control crop morphology and development, or to reduce fertilizer input. Drought stress exerts a dramatic impact on crop production. Traditionally, farmers use irrigation (if available) or, in some cases, polyethylene cover to protect soil. However, these methods have inherent problems, such as adverse environmental influences and a high price [79,80]. Alternatively, exploiting biological mechanisms for plant drought tolerance are promising [81]. ABA acts as a key regulator of yield increase under drought conditions [82]. Intervention in ABA signaling by small molecules could help plants to reduce the impact of drought [83]. However, direct application of ABA is affected by chemical instability and rapid catabolism by plant enzymes. A recently discovered ABA- mimicking ligand is promising as a potential effector. It acts as a potent activator of multiple members of ABA receptors, and proved to be efficient in protecting crops from drought by external application [84]. Alternatively, molecular targeting of receptor-like kinases (RLKs) was found to increase drought tolerance [80]. RLKs and related peptide ligands are evolutionarily highly conserved, which facilitates their transfer from model to crop species [85]. For control of plant architecture, phytohormones can be targeted. For example, by reducing brassinosteroid levels, rice leaves become more erect, which increases yield and biomass production [86]. This can be achieved by the application of the brassinosteroid synthesis inhibitor propiconazole, which is easy to apply and available at low cost [87]. In maize, the loss of ability to produce brassinosteroids results in female maize plants [88]. This may be useful for maize hybrid seed production, because producers must remove tassels from female parents to avoid self-pollination.

Usually, phytohormones have pleiotropic effects. Thus, the impact of MOST treatments on other

31 traits needs to be carefully examined. For example, low levels of brassinosteroids might lead to oxidative damage induced by water stress [89]. However, this can be complemented by external application of ABA [89]. Combinations of effectors addressing different hormone pathways to execute unique or complementary functions will be required to address the problem of correlated traits. Finally, MOST treatments may be used for decreasing fertilizer inputs. For example, external application of secret small peptides increased both lateral root elongation and nitrogen use-efficiency as a result of the upregulation of nitrate transporter genes [90].

Concluding Remarks

Here, we have highlighted MOST as a third important route for utilizing detailed molecular genomic information generated from current large-scale sequencing and -omics projects. Although classic methods such as MAS and transgenics are successful, they have inherent limitations and could be substituted or complemented by MOST treatments. Some current limitations of the existing methods are that, in European countries, transgenics are not accepted and MAS cannot be utilized if there is no natural variation for a desired trait. However, if there is prior knowledge underlying a desirable trait in other species, MOST treatments may be used to trigger specific metabolic pathways, which may lead to a target phenotype. Nonetheless, the applicability is depending on the acceptance by plant breeders and/or consumers of the specific treatment method. For the purpose of complementing existing methods, MOST treatments can be used for maximizing trait expression after an elite cultivar is produced by

MAS or transgenic approaches. For example, MOST can increase biotic or abiotic resistance for yield stability. Moreover, MOST treatments can be used to induce the expression of a transgene

32 at a defined time point. Promoters that will be induced by a known effector can be used to construct transgenic events [91]. Finally, detailed understanding of molecular processes in plants can be exploited not only for crop improvement (MOST), but also for identification of molecules of high economic value for other applications [92].

REFERENCES

1 Shendure, J. and Ji, H. (2008) Next-generation DNA sequencing. Nat. Biotechnol. 26, 1135– 1145

2 Chin, C-S. et al. (2013) Nonhybrid, finished microbial genome assemblies from long-read SMRT sequencing data. Nat. Methods 10, 563–569

3 Xu, X. et al. (2012) Resequencing 50 accessions of cultivated and wild rice yields markers for identifying agronomically important genes. Nat. Biotechnol. 30, 105–111

4 Huang, X. et al. (2012) A map of rice genome variation reveals the origin of cultivated rice. Nature 490, 497–501

5 Varshney, R.K. et al. (2009) Next-generation sequencing technologies and their implications for crop genetics and breeding. Trends Biotechnol. 27, 522–530

6 Rounsley, S. et al. (2009) De novo next generation sequencing of plant genomes. Rice 2, 35– 43

7 Mayer, K.F. et al. (2011) Unlocking the barley genome by chromosomal and comparative genomics. Plant Cell 23, 1249–1263

8 Barbazuk, W.B. et al. (2007) SNP discovery via 454 transcriptome sequencing. Plant J. 51, 910–918

9 Ossowski, S. et al. (2008) Sequencing of natural strains of Arabidopsis thaliana with short reads. Genome Res. 18, 2024–2033

10 Paux, E. et al. (2012) Sequence-based marker development in wheat: advances and applications to breeding. Biotechnol. Adv. 30, 1071–1088

11 Xia, Z. et al. (2012) Positional cloning and characterization reveal the molecular basis for soybean maturity locus E1 that regulates photoperiodic flowering. Proc. Natl. Acad. Sci. U.S.A. 109, E2155–E2164

12 Ersoz, E.S. et al. (2007) Applications of linkage disequilibrium and association mapping in crop plants. In Genomics-assisted Crop Improvement (Varshney, R.K. and Tuberosa, R., eds), pp. 97–119, Springer

13 Nordborg, M. and Weigel, D. (2008) Next-generation genetics in plants. Nature 456, 720–723 14 Rafalski, J.A. (2010) Association genetics in crop improvement. Curr. Opin. Plant Biol. 13, 174–180

15 Thudi, M. et al. (2014) Genetic dissection of drought and heat tolerance in chickpea through genome-wide and candidate genebased association mapping approaches. PLoS ONE 9, e96758

16 Feuillet, C. et al. (2011) Crop genome sequencing: lessons and rationales. Trends Plant Sci. 16, 77–88

17 Bolger, M.E. et al. (2014) Plant genome sequencing: applications for crop improvement. Curr. Opin. Biotechnol. 26, 31–37

18 Morozova, O. and Marra, M.A. (2008) Applications of next-generation sequencing technologies in functional genomics. Genomics 92, 255–264

19 Monaco, M.K. et al. (2014) Gramene 2013: comparative plant genomics resources. Nucleic Acids Res. 42, D1193–D1199

20 Al-Dous, E.K. et al. (2011) De novo genome sequencing and comparative genomics of date palm (Phoenix dactylifera). Nat. Biotechnol. 29, 521–527

21 Morrell, P.L. et al. (2011) Crop genomics: advances and applications. Nat. Rev. Genet. 13, 85–96

22 Hiremath, P.J. et al. (2011) Large-scale transcriptome analysis in chickpea (Cicer arietinum L.), an orphan legume crop of the semi-arid tropics of Asia and Africa. Plant Biotechnol. J. 9, 922–931

23 Mudalkar, S. et al. (2014) De novo transcriptome analysis of an imminent biofuel crop, Camelina sativa L. using Illumina GAIIX sequencing platform and identification of SSR markers. Plant Mol. Biol. 84, 159–171

24 Toth, R. and van der Hoorn, R.A. (2010) Emerging principles in plant chemical genetics. Trends Plant Sci. 15, 81–88

25 Khan, M. et al. (2013) Brassinosteroid-regulated GSK3/Shaggy-like kinases phosphorylate mitogen-activated protein (MAP) kinases, which control stomata development in Arabidopsis thaliana. J. Biol. Chem. 288, 7519–7527

26 Cramer, R. et al. (2013) Plant proteomics in crop improvement. Proteomics 13, 1771

27 Kim, S.T. et al. (2014) Rice proteomics: a model system for crop improvement and food security. Proteomics 14, 593–610

28 Kusano, M. and Saito, K. (2012) Role of metabolomics in crop improvement. J. Plant Biochem. Biotechnol. 21, S24–S31

29 Amiour, N. et al. (2012) The use of metabolomics integrated with transcriptomic and proteomic studies for identifying key steps involved in the control of nitrogen metabolism in crops such as maize. J. Exp. Bot. 63, 5017–5033

30 Kudo, T. et al. (2013) UniVIO: a multiple omics database with hormonome and transcriptome data from rice. Plant Cell Physiol. 54, e9

31 Seo, M. et al. (2011) Profiling of hormones and related metabolites in seed dormancy and germination studies. In Seed Dormancy (Kermode, A.R., ed.), pp. 99–111, Springer

32 Jannink, J-L. et al. (2010) Genomic selection in plant breeding: from theory to practice. Brief. Funct. Genomics 9, 166–177

33 Heffner, E.L. et al. (2010) Plant breeding with genomic selection: gain per unit time and cost. Crop. Sci. 50, 1681–1690

34 Heslot, N. et al. (2012) Genomic selection in plant breeding: a comparison of models. Crop Sci. 52, 146–160

35 Rubin, C-J. et al. (2010) Whole-genome resequencing reveals loci under selection during chicken domestication. Nature 464, 587–591

36 Goddard, M. (2009) Genomic selection: prediction of accuracy and maximization of long term response. Genetica 136, 245–257

37 Sander, C. (2000) Genomic medicine and the future of health care. Science 287, 1977–1978

38 Cockett, M. et al. (2000) Applied genomics: integration of the technology within pharmaceutical research and development. Curr. Opin. Biotechnol. 11, 602–609

39 Fullerton, S.M. et al. (2012) Return of individual research results from genome-wide association studies: experience of the Electronic Medical Records and Genomics (eMERGE) Network. Genet. Med. 14, 424–431

40 Biesecker, L.G. (2010) Exome sequencing makes medical genomics a reality. Nat. Genet. 42, 13

41 Bredel, M. and Jacoby, E. (2004) Chemogenomics: an emerging strategy for rapid target and drug discovery. Nat. Rev. Genet. 5, 262–275

42 Koul, A. et al. (2011) The challenge of new drug discovery for tuberculosis. Nature 469, 483– 490

43 Andersen, J.R. and Lubberstedt, T. (2003) Functional markers in plants. Trends Plant Sci. 8, 554–560

44 Collard, B.C. and Mackill, D.J. (2008) Marker-assisted selection: an approach for precision plant breeding in the twenty-first century. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 363, 557– 572

45 Xu, Y. and Crouch, J.H. (2008) Marker-assisted selection in plant breeding: from publications to practice. Crop Sci. 48, 391–407

46 Lecomte, L. et al. (2004) Marker-assisted introgression of five QTLs controlling fruit quality traits into three tomato lines revealed interactions between QTLs and genetic backgrounds. Theor. Appl. Genet. 109, 658–668

47 Ribaut, J-M. and Ragot, M. (2007) Marker-assisted selection to improve drought adaptation in maize: the backcross approach, perspectives, limitations, and alternatives. J. Exp. Bot. 58, 351–360

48 Ashikari, M. et al. (2005) Cytokinin oxidase regulates rice grain production. Science 309, 741–745

49 Concibido, V.C. et al. (2004) A decade of QTL mapping for cystnematode resistance in soybean. Crop Sci. 44, 1121–1131

50 Dunwell, J.M. (2000) Transgenic approaches to crop improvement. J. Exp. Bot. 51, 487–496

51 Pray, C.E. et al. (2002) Five years of Bt cotton in China: the benefits continue. Plant J. 31, 423–430

52 Perez-Pinera, P. et al. (2012) Advances in targeted genome editing. Curr. Opin. Chem. Biol. 16, 268–277

53 Guern, J. (1987) Regulation from within: the hormone dilemma. Ann. Bot. 60, 75–102

54 Lang, A. (1970) Gibberellins: structure and metabolism. Annu. Rev. Plant Physiol. 21, 537– 570

55 Khraiwesh, B. et al. (2012) Role of miRNAs and siRNAs in biotic and abiotic stress responses of plants. Biochim. Biophys. Acta 1819, 137–148

56 Seki, M. et al. (2007) Regulatory metabolic networks in drought stress responses. Curr. Opin. Plant Biol. 10, 296–302

57 Witke, W. (2004) The role of profilin complexes in cell motility and other cellular processes. Trends Cell Biol. 14, 461–469

58 Cassab, G.I. (1998) Plant cell wall proteins. Annu. Rev. Plant Biol. 49, 281–309

59 Showalter, A.M. (1993) Structure and function of plant cell wall proteins. Plant Cell 5, 9

60 Simon, E. (1974) Phospholipids and plant membrane permeability. New Phytol. 73, 377–420

61 Woollard, P.M. et al. (2011) The application of next-generation sequencing technologies to drug discovery and development. Drug Discov. Today 16, 512–519

62 Ng, S.B. et al. (2010) Exome sequencing identifies the cause of a Mendelian disorder. Nat. Genet. 42, 30–35

63 Santos-Rosa, M. et al. (2008) Development of a transient expression system in grapevine via agro-infiltration. Plant Cell Rep. 27, 1053–1063

64 Voinnet, O. (2001) RNA silencing as a plant immune system against viruses. Trends Genet. 17, 449–459

65 Martin-Ortigosa, S. et al. (2012) Gold functionalized mesoporous silica nanoparticle mediated protein and DNA codelivery to plant cells via the biolistic method. Adv. Funct. Mater. 22, 3576– 3582

66 Santner, A. and Estelle, M. (2009) Recent advances and emerging trends in plant hormone signalling. Nature 459, 1071–1078

67 Wilkinson, S. et al. (2012) Plant hormone interactions: innovative targets for crop breeding and management. J. Exp. Bot. 63, 3499–3509

68 Nisha, S. et al. (2012) Effect of plant compounds on induced activities of defense-related enzymes and pathogenesis related protein in bacterial blight disease susceptible rice plant. Physiol. Mol. Plant Pathol. 80, 1–9

69 Butcher, S.P. (2003) Target discovery and validation in the postgenomic era. Neurochem. Res. 28, 367–371

70 Yang, Y.L. et al. (2009) Target discovery from data mining approaches. Drug Discov. Today 14, 147–154

71 Gaj, T. et al. (2013) ZFN, TALEN, and CRISPR/Cas-based methods for genome engineering. Trends Biotechnol. 31, 397–405

72 Wuster, A. and Madan Babu, M. (2008) Chemogenomics and biotechnology. Trends Biotechnol. 26, 252–258

73 Yamagami, A. et al. (2009) Chemical genetics reveal the novel transmembrane protein BIL4, which mediates plant cell elongation in brassinosteroid signaling. Biosci. Biotechnol. Biochem. 73, 415–421

74 Rea, P.A. (2007) Plant ATP-Binding cassette transporters. Annu. Rev. Plant Biol. 58, 347– 375

75 Hicks, G.R. and Raikhel, N.V. (2014) Plant chemical biology: are we meeting the promise? Front. Plant Sci. 5, 455

76 Ekins, S. et al. (2007) In silico pharmacology for drug discovery: applications to targets and beyond. Br. J. Pharmacol. 152, 21–37

77 Wang, L. et al. (2013) TargetHunter: an in silico target identification tool for predicting therapeutic potential of small organic molecules based on chemogenomic database. AAPS J. 15, 395–406

78 Clark, M.A. et al. (2009) Design, synthesis and selection of DNA encoded small-molecule libraries. Nat. Chem. Biol. 5, 647–654

79 Athar, H. and Ashraf, M. (2009) Strategies for crop improvement against salinity and drought stress: An overview. In Salinity and Water Stress (Ashraf, M. et al., eds), pp. 1–16, Springer

80 Marshall, A. et al. (2012) Tackling drought stress: RECEPTOR-LIKE KINASES present new approaches. Plant Cell 24, 2262–2278

81 Peleg, Z. and Blumwald, E. (2011) Hormone balance and abiotic stress tolerance in crop plants. Curr. Opin. Plant Biol. 14, 290–295

82 Hirayama, T. and Shinozaki, K. (2010) Research on plant abiotic stress responses in the post- genome era: past, present and future. Plant J. 61, 1041–1052

83 Wilkinson, S. and Hartung, W. (2009) Food production: reducing water consumption by manipulating long-distance chemical signalling in plants. J. Exp. Bot. 60, 1885–1891

84 Cao, M.J. et al. (2013) An ABA-mimicking ligand that reduces water loss and promotes drought resistance in plants. Cell Res. 23, 1043–1054

85 Shiu, S-H. et al. (2004) Comparative analysis of the receptor-like kinase family in Arabidopsis and rice. Plant Cell 16, 1220–1234

86 Sakamoto, T. et al. (2005) Erect leaves caused by brassinosteroid deficiency increase biomass production and grain yield in rice. Nat. Biotechnol. 24, 105–109

87 Hartwig, T. et al. (2012) Propiconazole Is a specific and accessible brassinosteroid (BR) biosynthesis inhibitor for Arabidopsis and maize. PLoS ONE 7, e36625

88 Hartwig, T. et al. (2011) Brassinosteroid control of sex determination in maize. Proc. Natl. Acad. Sci. U.S.A. 108, 19814–19819

89 Zhang, A. et al. (2011) Nitric oxide mediates brassinosteroid-induced ABA biosynthesis involved in oxidative stress tolerance in maize leaves. Plant Cell Physiol. 52, 181–192

90 Tabata, R. et al. (2014) Perception of root-derived peptides by shoot LRR-RKs mediates systemic N-demand signaling. Science 346, 343–346

91 Akram, N.A. and Ashraf, M. (2013) chemical inducers. Plant Growth Regul. 32, 663–679

92 Pogue, G.P. et al. (2010) Production of pharmaceutical-grade recombinant aprotinin and a monoclonal antibody product using plant-based transient expression systems. Plant Biotechnol. J. 8, 6–8

93 Nino-Liu, D.O. et al. (2006) Xanthomonas oryzae pathovars: model pathogens of a model crop. Mol. Plant Pathol. 7, 303–324

94 Singh, S. et al. (2001) Pyramiding three bacterial blight resistance genes (xa5, xa13 and Xa21) using marker-assisted selection into indica rice cultivar PR106. Theor. Appl. Genet. 102, 1011– 1015

95 Huang, B. et al. (2012) Introgression of bacterial blight resistance genes Xa7, Xa21, Xa22 and Xa23 into hybrid rice restorer lines by molecular marker-assisted selection. Euphytica 187, 449– 459

96 Luo, Y.C. et al. (2012) Marker-assisted breeding of Xa4, Xa21 and Xa27 in the restorer lines of hybrid rice for broad-spectrum and enhanced disease resistance to bacterial blight. Mol. Breed. 30, 1601–1610

97 Tu, J. et al. (1998) Transgenic rice variety ‘IR72’ with Xa21 is resistant to bacterial blight. Theor. Appl. Genet. 97, 31–36

98 WANG, W. et al. (2011) Resistance of antimicrobial peptide gene transgenic rice to bacterial blight. Rice Sci. 18, 10–16

99 Sharma, A. et al. (2000) Transgenic expression of cecropin B, an antibacterial peptide from Bombyx mori, confers enhanced resistance to bacterial leaf blight in rice. FEBS Lett. 485, 208

100 Taniguchi, S. et al. (2014) Jasmonate induction of the monoterpene linalool confers resistance to rice bacterial blight and its biosynthesis is regulated by JAZ protein in rice. Plant Cell Environ. 37, 451–461

Figure 1. Three ways to utilize detailed molecular genetic information for crop improvement.

Phenotypes generated from the field, high-throughput sequencing techniques, and various -omics approaches generate vast amounts of valuable data in the information pool for crop improvement.

Traditionally, this detailed molecular genomic information is exploited by marker-aided selection (MAS) (i) and transgenic approaches (ii) for producing better new varieties. MAS can be used for both parental and progeny selection, whereas transgenic methods provide opportunities for introducing novel beneficial traits. Molecular strengthening (MOST) treatments

(iii) as a third outlet can facilitate the breeding process (i.e., flowering time regulation or

Figure 1 continued controlled crosses). MOST treatments can also be applied directly in the field for crop management, for reduced fertilizer input, or improved yield stability due to increased biotic and/or abiotic resistance. Abbreviations: GWAS, genome-wide association studies; QTL, quantitative trait loci.

Figure 2. Possible interactions between effectors and molecular targets. The triangles represent six classes of effector. Green represents DNA as effectors and/or targets, red represents RNA, orange represents proteins, purple represents hormones, dark red represents metabolites, silver represents lipids, yellow represents synthesized chemical molecules, and green represents plant extracts. Possible interactions are shown as triangles and targets combined.

Table 1. Comparison of MAS, transgenics and MOST

Trait MAS Transgenics MOST Effect on environment No effect May have effect May have effect

Transferability across species Low Mediana High

Multitrait integration Median Low Highb

Novel trait induction No High Medianc

Cost for producers Median Expensive Median

Cost for farmersd Low Median Low

Public concern No High Median

a Plant transformation protocol may not be available for species of interest. b Combination of molecules targeting multiple traits is feasible. c Metabolic pathways may not exist in target species. d MOST treatments are applied only when it is needed. Compared with farmers who encounter severe yield losses without MOST application, farmers using MOST will consider it cost efficient.

Box 1. Comparison of MAS, transgenics and MOST methods

Bacterial blight is one of the most important bacterial diseases of rice and is caused by

Xanthomonas oryzae pv. oryzae (Xoo). It can reduce rice yields by up to 50% [93]; thus, worldwide efforts have been devoted to solve this problem. Successful examples of increasing the disease resistance by MAS, transgenic, and MOST methods are available. A combination of all three methods might be superior to each of those individually.

MAS

MAS is mainly used to pyramid different resistance genes into elite materials. Different combinations of these resistance genes provide high levels of resistance, such as introgression of xa5, xa13, and Xa21 into elite line cv. PR106 [94]; Xa7, Xa21, Xa22, and Xa23 into the elite rice line Huahui 1035 [95; and Xa4, Xa21, and Xa27 into Mianhui 725 [96]. These MAS pyramiding strategies were efficient for developing varieties resistant to rice bacterial blight. However, the introgression process is time consuming, and has to be repeated for novel cultivars. Once new pathogen isolates evolve that can overcome specific resistance genes, further R genes need to be introgressed.

Transgenics

Similar to MAS, transgenic approaches aim to introduce resistant genes into elite rice varieties.

Besides introducing rice resistance genes, for example Xa21 into IR72 [97], genes from other species can be introduced in this way, such as antimicrobial peptide genes Np3 and Np5 from

Chinese shrimp [98] and antibacterial peptide cecropin B from Bombyx mori [99]. Compared with MAS, transgenic approaches are more expensive and strictly regulated. Thus, MAS is preferable for rice genes. However, for genes from non-crossable species, transgenic approaches are the only option.

MOST

MOST treatments do not change the genetic information of an organism. However, in contrast to traditional fungicides, MOST targets the genome, transcriptome, or proteome of the host plant instead of the invading pathogens, although the mode of action may be similar. Exogenous application of jasmonic acid (JA) can induce resistance to Xoo [100]. Expression of linalool synthase is increased and leads to the upregulation of defense-related genes. Alternatively, some plant extracts can induce defense-related enzymes, such as protoporphyrinogen oxidase (PPO), prolyl oligopeptidase (PO) and β-1,3-glucanase [68, which can suppress the disease by up to

70% under field conditions. However, environmental, social, and economical consequences need to be considered before large-scale application.

CHAPTER 3.

GENETIC ARCHITECTURE OF PLANT HEIGHT IN MAIZE

PHENOTYPE-SELECTED INTROGRESSION FAMILIES

Adel H. Abdel-Ghani*, Songlin Hu*, Yongsheng Chen, Everton A. Brenner, Bharath Kumar,

Micheal Blanco, Thomas Lubberstedt

Paper published in Plant Breeding Journal. Abstract, structure, and references are formatted according to the journal standards.

AAG and SH contributed equally to this work. AAG, YC, EAB and BK established the experimental population (PIFs) and collected data. SH and AAG analyzed the data and wrote the manuscript. MB and TL designed the experiment and involved in the revision of the manuscript.

All authors read and approved the manuscript.

Abstract

This study aimed at developing, characterizing and evaluating two maize phenotypic- selected introgression libraries for a collection of dominant plant height (PHT)-increasing alleles by introgressing donor chromosome segments (DCS) from Germplasm Enhancement of Maize

(GEM) accessions into elite inbred lines: PHB47 and PHZ51. Different backcross generations

(BC1-BC4) were developed and the tallest 23 phenotype-selected introgression families (PIFs) from each introgression library (PHB47 or PHZ51) were selected for single nucleotide polymorphism genotyping to localize DCS underlying PHT. The result shows that most PIFs carrying DCS were significantly (α= 0.01) taller than the respective recurrent parent. In addition, they contained larger donor genome proportions than expected in the absence of selection or random mating across all BC generations. The DCS were distributed over the whole genome, indicating a complex genetic nature underlying PHT. We conclude that our PIFs are enriched for favorable PHT-increasing alleles. These two libraries offer opportunities for future PHT gene isolation and allele characterization and for breeding purposes, such as novel cultivars for biofuel production.

Introduction

Plant height (PHT) is an important agronomic trait that is associated with grain yield.

PHT exhibits a strong correlation with relative yield (grain production; r2 = 0.67) and moderate, but highly significant, correlations with ear height and the number of leaves per plant (r2 = 0.88)

(Sibov et al. 2003, Aastveit and Aastveit 1993, Lima et al. 2006), illustrating the importance of

PHT as fitness trait in the agricultural context. Currently, the focus is on high grain yield due to increasing plant population densities of moderate-sized, lodging-tolerant cultivars. However, an increased demand for lignocellulosic biomass for biofuels has led to a paradigm shift in plant architecture (Fernandez et al. 2009, Schwietzke et al. 2009, Hu and Lubberstedt 2015). As tall plants can produce more vegetative tissue, there is a positive and strong correlation between PHT

48 and biomass in maize (r 2 > 0.73) (Lubberstedt et al. 1997). Moreover, for dual-purpose crops for which grain is still of high value, increased stover biomass is an added value. In the Midwestern and Central Plains regions, 100 million metric tons of maize stover can be produced for biofuel production (Graham et al. 2007). Moreover, selection for taller plant stature increases the yield capacity under water-limiting conditions, as PHT shows a positive and high correlation with yield under drought (Edmeades et al. 1999, He et al. 2010, Sayed 2011). Although tall plants are more susceptible to lodging, selection for traits that enhance the plant stability (Fernandez et al.

2009) can avoid the lodging problem.

However, continued selection of moderate PHT within elite germplasm may cause the reduction in genetic diversity (Hawkes 1977, Goodman 1999). Introgression library may act as a potential method to add extra genetic diversity – as introgression of unadapted exotic germplasm into elite breeding materials can reduce genetic vulnerability and broaden the variability available for selection (Tanksley and Nelson 1996). Despite poor agronomic performance, exotic germplasm has repeatedly shown to contain favorable genes (Frey et al. 1984, Elgin and Miller

1989, DeVicente and Tanksley 1993), which provides novel opportunities to maximize the genetic diversity and produce new cultivars.

Previously, introgression libraries were proposed as powerful tool to disrupt elite genetic background as little as possible by introducing a limited number of exotic segments by backcrossing (Eshed et al. 1992). Introgression involves the repeated backcrossing with the elite line as recurrent parent (RP), thereby minimizing the negative side effects attributable to genetic

49 interactions between exotic and elite genomes. Because Eshed and Zamir (1994) constructed the first introgression line library in tomato carrying single wild tomato (Lycopersicon pennellii) chromosomal segments in the genetic background of cultivated tomato (Lycopersicon esculentum), introgression libraries were successfully used in other crop species. Beneficial effects of alleles from exotic germplasm have been illustrated for PHT (Miedaner et al. 2011), various agronomic traits (Huang et al. 2003, Pillen et al. 2003, Septiningsih et al. 2003, Tian et al.

2006, Tan et al. 2007, Von Korff et al. 2008), quality-related traits (Matus et al. 2003, Pillen et al.

2003, Kunert et al. 2007) and biotic and abiotic stress tolerance (Von Korff et al. 2005, Siangliw et al. 2007, Gu et al. 2012).

Introgression libraries ideally consist of a set of homozygous, near-isogenic lines, each carrying a single marker-defined donor chromosome segment (DCS) in the genetic background of an elite line (Zamir 2001), which facilitates the characterization and utilization of exotic segments by plant breeders (Eshed et al. 1996). These DCS are introduced by marker-assisted backcrossing (MABC), and all DCS in the introgression library jointly represent the total donor genome. So far, introgression libraries were developed for maize (Ribaut and Ragot 2007,

Szalma et al. 2007) and other economically important crop species including tomato (Eshed et al.

1992, Eshed and Zamir 1994), Brassica napus (Howell et al. 1996, Ramsay et al. 1996), rice (Lin et al. 1998), barley (Matus et al. 2003, Von Korff et al. 2004, 2008), soybean (Concibido et al.

2003), lettuce (Jeuken and Lindhout 2004), melon (Eduardo et al. 2005), wheat (Liu et al. 2006) and rye (Miedaner et al. 2011).

However, traditional establishment of introgression libraries using MABC is time- consuming, laborious (Zamir and Eshed 1998) and requires extensive genotyping process, although significant improvements have been made such as how to more efficiently use the high- throughput marker selection system and different crossing schemes to shorten backcross (BC) generations (Herzog et al. 2014). Moreover, for traditional introgression libraries, only two parents will be involved to construct introgression libraries, which may limit their genetic diversity. In this manuscript, we address the idea of phenotypic-selected introgression library to accumulate PHT-increasing genetic variation. Phenotype-selected introgression families (PIFs) still need several generations of backcrossing, but without marker selection or genotyping, this method is more cost-effective. Moreover, it maximizes the genetic diversity for the target trait by using more donor accessions. In summary, it has three major differences compared to traditional introgression line libraries: (i) PIFs for PHT do not necessarily cover all regions of the genome, but accumulate DCS affecting PHT within an elite/adapted genetic background; (ii) DCS originate from multiple donors; and (iii) only phenotypic selection was employed. PIF libraries enable broader and more targeted (trait-specific) use of exotic germplasm. Multiple exotic donors have the advantage over a single donor to provide a greater variety of chromosomal segments in any genomic location (Holland 2004, Zeng et al. 2007, Ladizinsky 2012). In each generation, only tall plants were selected and backcrossed, which leads to an enrichment of chromosome segments and genes with impact on increasing the PHT in isogenic background.

The objectives of our study were to (i) develop two maize PIF libraries using the

Germplasm Enhancement of Maize (GEM) accessions as a source of DCS into the genetic

51 background of PHB47 and PHZ51 elite maize lines, (ii) investigate the per se performance of introgression families resulting from BC individuals selected for PHT to study the potential utility of establishing trait-based introgression libraries using PHT as well-characterized quantitative genetic model character, and (iii) identify, localize and characterize DCS of individual PIFs to investigate the genetic architecture underlying maize PHT for future maize

PHT improvement programs using markers.

Materials and Methods

Two libraries of introgression families were developed. Exotic donor parents as source materials were obtained from the GEM project (Salhuana and Pollak 2006), which aims at developing adapted germplasm from unadapted sources from the Latin American Maize Project

(LAMP) (Salhuana and Pollak 2006) that can be utilized in the major corn-growing areas of the

United States. These exotic maize accessions were crossed with expired PVP (Plant Variety

Protection) lines PHZ51 and PHB47 as RPs, resulting in two libraries based on either PHB47 or

PHZ51 genetic background as PIFB47 or PIFZ51. PHB47 belongs to the Iowa stiff stalk (BSSS) gene pool, while PHZ51 is belonging to the non-stiff-stalk gene pool. As illustrated in (Fig. 1), for each donor accession, bulked pollen from donor individuals was used to cross with RP

(PHB47 or PHZ51) to produce F1 seed. Five random F1 individuals were backcrossed with the

RP plants to produce five BC1F1 families (BC1F1 ears were kept separate as BC1 families).

Among the five BC1F1 families, only the families flowering about the same time (within 5 days) as the RP were advanced. On average, one to three BC1F1 families were selected for each donor accession. The two tallest plants in each selected BC1F1 family were picked and backcrossed to

the RP. Seed from the tallest plant was used to produce a BC1:2 family (Fig. 1). BC1:2 seed of the other selected plant was kept as a backup. This procedure was repeated to produce advanced

BC2:3 or BC3:4 families with the goal to produce one advanced BC family per accession × background combination. In total, 58 and 62 exotic accessions were used to establish the PIFB47 and PIFZ51 libraries, respectively, including 17 accessions commonly used for both PIF libraries

(Table 1). For PIFZ51, there are 2 BC1 families, 79 BC2 families, 42 BC3 families and 3 BC4 families. For PIFB47, we developed 4 BC1 families, 39 BC2 families, 53 BC3 families and 4 BC4 families.

Field evaluation

In the 2010 and 2011 growing seasons, all BC families (126 from PIFZ51, 100 from

PIFB47 – some donor accessions produce more than one PIF) were planted at two locations:

Agronomy farm, Iowa State University, Ames, IA (Ames), and Plant Introduction Station, IA

(Ames), in 2-m plots. Twenty-five plants were planted per plot with 0.75 m between plots. BC families (226 in total) and the two parental inbred lines (PHB47 and PHZ51), included as checks, were planted in unreplicated field trials in 2010 and in a randomized complete block (RCBD) design in two replicated field trials in 2011 at two locations. Days to tasseling was recorded, when at least 50% of the plants per plot showed emerged tassels. Two weeks after tasseling, the plant with a median PHT within each plot was selected and measured; PHT was recorded from the ground to the top of the tassel. The average of PHT across locations and replications was calculated to represent the PHT performance for each family.

DNA extraction and genotyping

Leaf samples of the selected PIFs were collected for DNA extraction after tasseling. The

DNA of the BC families was extracted from the leaf materials according to the protocol described by Saghai-Maroof et al. (1984). The tallest 23 families from each introgression library

(PIFB47 and PIFZ51) were selected (=subset of ‘selected introgression libraries’), named as

PIFB47S and PIFZ51S for genotyping with 207 mapped single nucleotide polymorphism markers (SNP) [based on ISU SNP map (Liu et al. 2010)]. The markers are distributed throughout the genome using the SequenomMassARRAY System at the Genomic Technologies

Facility (http://www. plantgenomics.iastate.edu/) at ISU. Estimation of recurrent and donor parent genome percentages for each BC family was carried out by considering alleles differing from those of the RP as donor introgressions, while marker alleles identical to the RP alleles were classified as RP alleles. Fifty BC1 doubled haploid (DH) lines developed from the GEM project were used for comparison as unselected library in this study (Brenner et al. 2012). Briefly,

GEM-DH lines were generated by introgressing the same set of exotic maize races into the genetic background of PHZ51 and PHB47, respectively, to produce GEM-DH-B47 and GEM-

DH-Z51 lines. The difference between GEM-DH lines and PIFs is that instead of repeated backcrossing with phenotypic selection, DH lines were generated from BC1 individuals without any phenotypic selection. The 50 GEM-DH lines were genotyped with 199 SNP markers with a subset of 139 SNP markers overlapping with the 207 used for genotyping the PIFs.

Statistical analyses

Phenotypic data were subjected to analyses of variance based on a RCBD. Entry means were used in the combined analyses across locations. Only data collected in 2011 were used in the combined analyses. Location was considered as a fixed factor, whereas genotypes were

2 considered as a random factor. The variance components for genotypes (σ g), genotype ×

2 2 environment variance (σ g×e) and error variance (σ e) were estimated according to Hallauer and

Miranda (1988). Heritability h2 was calculated on an entry mean basis. After excluding monomorphic markers and markers polymorphic for only one of 23 families in either PIFB47S or PIFZ51S, 71 and 84 markers were used for further analyses in PIFB47S and PIFZ51S.

Graphical genotypes were obtained using the software package Graphical GenoTypes (GGT)

(Van Berloo 2008). Genotyped BC families were characterized for the percentage of donor parent compared to RP alleles. Based on genotype representation with GGT, the number of DCS as well as the percentage of donor genome was determined. For the calculation of segment length and genome ratios, the half-intervals flanking a marker locus were considered to be of the same genotype as implemented by the GGT software. For a few missing marker data points, because the flanking markers are RP genotype, BC families with missing data were assumed to have the genotype of the RP for this particular marker. We define hot spots in this manuscript as meta- analyses results of previous published PHT QTL regions from Wang et al. (2006).

Results

Average PHT was 220.7 and 237.3 cm for the RPs PHZ51 and PHB47, respectively (Fig.

2). As expected, the means of most families from PIFB47 and PIFZ51 exceeded their respective

RPs. For PIFB47 and PIFZ51, PHT ranged from 215.5 to 296.0 cm and from 231.1 to 285.3 cm,

55 respectively; 55% of PIFB47 families and 98% of PIFZ51 families were significantly taller than their corresponding RP (α = 0.01). In contrast, only 10% and 9% of GEM-DH-B47 and GEM-

DH-Z51 lines were significantly taller than their corresponding RP – most GEM-DH lines showed no difference in PHT, supporting the efficiency of phenotypic selection and indicating that the PHT-increasing donor alleles have been successfully accumulated in PIFB47 and

PIFZ51 introgression libraries. DTT (Days to Tasseling) means were close to RPs in both libraries. All BC families reached tasseling within five days for PIFB47 and within seven days for PIFZ51, except for one family (PHZ51 < 2 > /(BOFO DGO123/ PHZ51)/PHZ51#005)#001 >

01/Z51:1050)-B) with 10 days delayed tasseling compared to PHZ51. In contrast, most unselected GEM-DH lines show later flowering time, supporting the efficiency of phenotypic selection in preparing our PIF introgression libraries. Pearson’s phenotypic correlations between

PHT and DTT were low, although significant (r = 0.19**). When correlations were calculated separately for each library, low significant positive correlations were obtained within the PHB47 library (r = 0.21**). No significant correlation was found within the PHZ51 library between PHT and DTT. Variance components for genotype generally exceeded those for genotype 9 environment (GxE) interactions (Table 2). Nevertheless, GxE interaction components were significant (P = 0.01). Estimates for h2 were high: h2 for PHT and DTT were 0.79 and 0.74, respectively.

Genetic characterization of backcross families

Generally, BC families displayed a higher percentage of donor genome than the expected

25%, 12.5%, 6.25% and 3.125% for BC1, BC2, BC3 and BC4 generations (Table 3 and Table 4).

For example, there are on average 21.56% and 11.8% for BC2 and BC3 in the PIFZ51S (Table 3)

and 16.24% and 11.47% for BC2 and BC3 in the PIFB47S (Table 4). The percentage of donor introgressions ranged from 8.4% to 64.6% and from 6.3% to 32.5% for PIFZ51S and PIFB47S, respectively. The number of donor segments ranged from 4 to 19 in PIFZ51S and 4 to 18 in

PIFB47S.

Advanced BC generations in both selected introgression libraries (PIFB47S and PIFZ51S) showed an obvious decrease in the number of DCS and the proportion of donor genome per PIF

(Table 3 and Table 4). In PIFB47S, the percentage of donor parent decreased from 24.5% (10 segments) in BC1, 16.3% (seven segments) in BC2, to 11.5% (six segments) in BC3 families. In

PIFZ51S, the proportion of donor parent decreased from 21.6% (nine segments) in BC2 to 11.8%

(six segments) in BC3 families. However, one BC4 family had a 56% donor genome proportion

(19 segments), which may be due to random drift or a large linkage block inherited from its original donor. The observed donor genome proportion was greater for PIFs compared to the unselected GEM-DH lines (Table 3 and Table 4). On average, unselected GEM-DH lines had a reduced percentage of donor introgressions (21.3% and 16.5% in GEM-DH-Z51 and GEMDH-

B47 lines, respectively) compared with the expected value of 25.0% in BC1.

Comparison between the two selected libraries (PIFB47S and PIFZ51S)

Donor chromosome segments are distributed over the whole genome for the two selected libraries (Figs 3 and 4). The maximum frequency of donor introgressions was found on

Chromosome 4 in both libraries (21.4% and 33.2% in PIFB47S and PIFZ51S, respectively).

Chromosomes 1 (21.2%) and 6 (17.3%) for PIFB47S and Chromosomes 2 (21.8%) and 3 (24.0%) for PIFZ51S also displayed high percentages of donor introgressions. The lowest frequency of donor allele introgressions was observed for Chromosomes 2 (9.3%), 3 (10.4%) and 7 (10.0%) in

PIFB47S and for Chromosomes 6 (8.8%) and 7 (12.4%) in PIFZ51S. In PIFZ51S, seven genomic regions displayed the highest donor genome percentages, which suggests that in these regions the RP alleles tend to be replaced by donor segments during the phenotypic selection process. The highest genome contribution (45%) was detected within the intervals ranging from

16.0 to 29.7 cM on Chromosome 4, followed by nine regions with 35% donor genome contribution distributed over five chromosomes. The second highest donor genome contributions were located on Chromosomes 1, 2, 3 and 4. Regions with low frequency of RP allele substitution by donor segments were clustered in 10 regions on seven chromosomes:

Chromosome 1 (0 and 201.4– 230.1 cM), Chromosome 2 (99.3 cM), Chromosome 4 (17.7–29.0 cM), Chromosome 6 (12.9 cM), Chromosome 7 (11.5–34.3 cM), Chromosome 9 (30.3 and 81.2 cM) and Chromosome 10 (83.6 and 146.7 cM). For PIFB47S, 10 regions with a high donor replacement frequency were distributed over six chromosomes. The highest donor genome percentages were detected on Chromosome 6 at 12.1 cM (65%) and 47.7 cM (45%), followed by the telomeric regions of Chromosomes 1, 4, 6 and 10 with 35% donor allele contributions.

Sixteen regions with a low replacement frequency were identified for PIFB47S distributed over the whole genome with 15% of donor genome.

In several cases, introgressions coming from different donor parents were shared by selected BC families in both libraries. The most frequent donor regions detected in both selected libraries were at 93.5 and 117.1 cM on Chromosome 1, 21.4 cM on Chromosome 2, 201.9 cM on

Chromosome 3, 16.0 and 101.0 cM on Chromosome 4, 12.1 cM on Chromosome 6, 41.3 and

77.5 cM on Chromosome 8 and 72.4 cM on Chromosome 9.

Discussion

Our study aimed to establish two phenotypic-selected introgression libraries to accumulate dominant maize PHT-increasing alleles from various donors into the same genetic background. Tall plants were selected that flowered about the same time as the RP, to reduce the confounding effect of flowering time and PHT. Various unrelated PHT QTL from different donor parents may facilitate the discovery of novel PHT-related genes, and they may be useful for breeding varieties aiming at maximizing biomass yield (Fernandez et al. 2009).

The two introgression libraries used in this study consist of BC families carrying DCS, which were introgressed into a common elite genetic background of either PHB47 or PHZ51.

The selected and genotyped tallest 46 BC families (23 in each RP background) with significantly

(α = 0.01) taller PHT than their RPs contained 4–19 DCS with donor genome percentages ranging from 6.3% to 64.6%. Compared to high-throughput genotyping technologies that produce thousands of markers for each individual, the number of markers used in this study was limited. However, because most of the individuals are only BC2-BC3, the number of SNP markers used in this study is sufficient to capture large DCS (Figs 3 and 4) – in PIFZ51S, the average donor genome proportion is 21.3%, and there are on average 9.6 donor segments, so the average length is 21.3%/9.6 = 2.22% of the genome. We used 84 markers across the genome and

59 the average length between each two markers is 100%/84 = 1.19% < 1.83 – the number of markers are able to capture the large donor segments. Distribution of DCS across the whole genome in these selected PIFs is consistent with quantitative inheritance of PHT. In the present study, most PIFs were generated after two to three generations of backcrossing. Without selection, the proportion of donor genome should decrease by 50% after each generation of backcrossing. A lower-than-expected donor genome proportion was observed (BC1DH = 16 <

25%) for the unselected GEMDH lines. This can be explained by monomorphic markers between the RP and donors, which leads to an underestimated donor genome proportion. In contrast, PIFs had a higher level (on average 8%) of donor genome proportion (BC2 = 19.1%, BC3 = 11.6%, etc.) than expected without selection. Because some markers are monomorphic between RP and donor parent and the number of SNP markers is relatively small in this study, the true donor genome proportion in PIFs is expected to be even higher. This indicates that the selection process was successful and that donor alleles related to PHT were accumulated in our PIF libraries. Some alleles with small effects related to PHT may be lost during the selection process. However, PIFs are still able to capture more DCS relating to tallness compared with traditional introgression methods. Substantial hitchhiking and linkage drag around the target locus can be expected (Stam and Zeven 1981), resulting on average in 36-cM-long donor regions in BC3 for a chromosome of

200 cM. Our phenotypic selection process likely selected multiple loci related to PHT in each

PIF, explaining the much higher donor genome proportion found in our study. A rough calculation could be that if only one locus affecting PHT was selected, then additional donor genome proportion is expected to be 2% [36 cM/1800 cM – ISU SNP map (Liu et al. 2010)]. If two independent loci are selected, then it should be 4%, etc. Based on the BC1DH information, when the expected donor genome proportion is 25%, on average there is 9% (25–16%) lower

60 donor genome proportion likely due to monomorphic markers, which means that when the observed donor proportion is 16%, there will be 9% underestimated. So on average, when the observed extra donor genome proportion is 8%, there will be 8%×9%/16% = 4.5%. The extra donor genome proportion for the PIFs should then be (8% + 8%× (9%/ 16%) = 12.5%), which means that there were around seven (12.5%/2%) loci selected for each PIF on average. For the

PIF with the largest donor genome proportion, the number of loci would be 41 (one BC2F1 –

(64.6–12.5) ×(25/16)/2).

To differentiate genome regions with the effect on PHT from regions carrying DCS and present due to random drift, we determined hot spots for PHT in the genome. A hot spot in our study was defined as genomic region, where the RP allele is replaced by DCS in more than seven

PIFs across both PIFB47S and PIFZ51S. A DCS is expected to occur at 50% in BC1F1, 25% in

BC2F1 and 12.5% in BC3F1 due to the random drift. Because there were 1 BC1, 33 BC2, 10 BC3 and 2 BC4 families within our PIF library, we calculated an approximate weighted probability for a DCS to occur by random drift among our selected 46 PIFs as P = 0.25× (1/46) + 0.125×(33/46)

+ 0.0625× (10/46) + 0.03125×(2/46) = 0.1. Based on a binomial distribution mass function

(Frisch et al. 1999) of F (7, 46, 0.1) = 0.92, the probability of more than seven individuals carrying a donor allele SNP by random drift rather than by selection pressure is smaller than 0.08.

The more often a DCS substitutes a particular region, the more likely it becomes that this substitution is not only by chance but indeed due to selection. Our hypothesis is that hot spots are more likely to carry PHT-related QTL. This is similar to introgression mapping process proposed by Thurber et al. (2013). Therein, for all BC4-derived lines, the theoretical expectation of finding

61 a donor segment without selection was 3% and a cut-off of 4% was used to define the regions retained due to selection. As a result, 28 hot spots met our criterion, spread across the whole genome (Table 5). The number of hot spots in our study was comparable with the number of

QTL identified in previous studies of maize using large mapping populations. In a study based on a population size of N = 976 in maize, a total of 30 QTL were detected for PHT (Schon et al.

2004). Melchinger et al. (1998) detected a total of 31 QTL for PHT among 344 maize F3 families derived from a biparental cross. Wang et al. (2006) found 40 ‘real’ PHT QTL using meta analyses by collecting all previous PHT QTL results distributed over the whole maize genome.

We found that 12 hot spots in our study overlapped with respective ‘real’ QTL, which means that these 12 hot spots overlapped with repeatedly reported QTL in previous mapping studies for

PHT QTL. For example, the hot spot on Chromosome 4 with the maximum frequency of donor introgressions overlapped with previously reported PHT QTL (Wang et al. 2006). This number of overlapping with previous published PHT hot spots is higher than expected by chance: the probability of finding 12 hot spots by chance based on 28 regions detected in our study is below

0.1. Around 31% of the total markers overlap with the ‘real’ QTL. Based on a binomial distribution mass function (Frisch et al. 1999) of F (11, 28, 0.31) ~0.9, the probability of finding

12 or more overlap regions with the ‘real’ QTL just by chance is around 0.1. These 12 overlap regions may be the main clusters of PHT-related genes segregating within elite maize materials.

For the 126 PIFs from the PIFZ51 and 100 PIFs from the PIFB47, we only genotyped 23

PIFs from each library and only the regions without an effect on flowering time were selected.

This may be the reason that we did not capture more ‘real’ QTL. We also detected 16 hot spots

62 different from the 40 ‘real’ QTL. As most of the previous PHT QTL mapping studies were based on elite germplasm, it is no surprise to find novel PHT related regions in our study, because we utilized offspring of exotic × elite crosses. Across the 28 regions, the region around 12.1 cM on

Chromosome 6 and 101 cM on Chromosome 4 [based on the ISU SNP map (Liu et al. 2010)] had the highest DCS substitution rates: 18 of 46 PIFs carried DCS in these two regions.

Our approach of utilizing a small population size while tapping into extensive genetic diversity by using multiple donor accessions is different from bi-parental QTL mapping and introgression mapping populations. Small population sizes influence the power to detect QTL and the accuracy of estimated QTL effects (Melchinger et al. 1998). In a study performed by

Beavis (1994) using 400 maize F3 families derived from the B73 × Mo17 cross to determine the effect of population size on the number of QTL detected, only four QTL were detected in the combined mapping population of size N = 400, further reduced to 1–3 in each of the subsets of N

= 100 families. In another study performed by Schon et al. (2004), a total of 30 QTL for PHT were detected, but with a very large mapping population of N = 976 maize F2:5 families derived from a bi-parental cross. When multiple sets of smaller population sizes N = 488, 244 and 122

F2:5 from the same bi-parental cross were used by sampling without replacement, the number of

QTL detected decreased to 17.6, 12.0 and 9.1, respectively. Therefore, our PIF library method provides a promising opportunity to efficiently use the genetic potential of exotic germplasm and targeting genomic regions underlying quantitative traits based on a small number of selected families.

The method to use introgression libraries to discover PHT QTL was successful in previous studies (Pillen et al. 2003, Septiningsih et al. 2003, Liu et al. 2006, Von Korff et al.

2008) and for other agronomically important traits (Huang et al. 2003, Pillen et al. 2003,

Septiningsih et al. 2003, Tian et al. 2006, Tan et al. 2007, Von Korff et al. 2008). Even for introgression lines with extra donor segments, new statistical methods have been developed for

QTL detection (Falke et al. 2014). A pre-selection for target genes underlying a quantitative trait or phenotype in previous BC generations was shown to be helpful for accelerating the recovery of the elite RP genome (Frisch 2005, Falke et al. 2008). Our PIF library method is comparable to other traditional introgression libraries, with the difference being that only phenotypic selection was executed, multiple donor parents were used and we only focused on one trait. Compared with the traditional introgression libraries, the use of PIFs has advantages: (i) low cost phenotypic selection, (ii) multiple donors and (iii) focus on only one trait to accumulate respective genomic regions. However, this approach of using PIFs may only be useful for highly heritable traits and a limited use for only one trait – PIFs may not be suitable for complex traits that are more difficult to phenotype with lower heritability. However, the fast developing novel high-throughput phenotyping techniques may provide more opportunities for PIF method.

Without high-throughput marker-assisted selection, there may be linkage drag problems associated with PIF method; however, PIF is still a very efficient way to maximize the genetic variation and to establish pre-breeding materials.

Flowering time and PHT are closely correlated. We intentionally and successfully enriched for DCS affecting PHT, but not flowering time as documented by much lower correlations between both traits (r2 = 0.19) compared to other studies, for example r2 = 0.34 in the NAM (Nested Association Mapping) population and r2 = 0.78 in NCRPIS panel (Peiffer et al.

2014). In this way, the PHT-related genomic regions will not be confounded with flowering time.

This is valuable to develop silage and dual-purpose maize varieties to increase the biomass yield via PHT without affecting the maturity.

REFERENCES

Aastveit, A., and K. Aastveit, 1993: Effects of genotype-environment interactions on genetic correlations. Theor. Appl. Genet. 86, 1007—1013.

Beavis, W. D., 1994: The power and deceit of QTL experiments: lessons from comparative QTL studies Proceedings of the forty-ninth annual corn and sorghum industry research conference, 250—266. Chicago, IL.

Brenner, E. A., M. Blanco, C. Gardner, and T. Lubberstedt, 2012: Genotypic and phenotypic characterization of isogenic doubled haploid exotic introgression lines in maize. Mol. Breed. 30, 1001—1016.

Concibido, V., B. La Vallee, P. Mclaird, N. Pineda, J. Meyer, L. Hummel, J. Yang, K. Wu, and X. Delannay, 2003: Introgression of a quantitative trait locus for yield from Glycine soja into commercial soybean cultivars. Theor. Appl. Genet. 106, 575—582.

DeVicente, M. C., and S. D. Tanksley, 1993: QTL analyses of transgressive segregation in an interspecific tomato cross. Genetics 134, 585—596.

Edmeades, G., S. Chapman, and H. Lafitte, 1999: Selection improves drought tolerance in tropical maize populations: I. Gains in biomass, grain yield, and harvest index. Crop Sci. 39, 1306—1315.

Eduardo, I., P. Arus, and A. J. Monforte, 2005: Development of a genomic library of near isogenic lines (NILs) in melon (Cucumis melo L.) from the exotic accession PI161375. Theor. Appl. Genet. 112, 139—148.

Elgin, J. H. Jr, and P. A. Miller, 1989: Enhancement of Plant Germplasm Biotic Diversity and Germplasm Preservation, Global Imperatives, 311—322. Springer, Netherlands.

Eshed, Y., and D. Zamir, 1994: A genomic library of Lycopersicon pennellii in L. esculentum: a tool for fine mapping of genes. Euphytica 79, 175—179.

Eshed, Y., M. Abu-Abied, Y. Saranga, and D. Zamir, 1992: Lycopersicon esculentum lines containing small overlapping introgressions from L. pennellii. Theor. Appl. Genet. 83, 1027— 1034.

Eshed, Y., G. Gera, and D. Zamir, 1996: A genome-wide search for wild-species alleles that increase horticultural yield of processing tomatoes.

Theor. Appl. Genet. 93, 877—886. Falke, K., Z. Susic, B. Hackauf, V. Korzun, J. Schondelmaier, P. Wilde, P. Wehling, H. Wortmann, R. Mank, and J. R. Van Der Voort, 2008: Establishment of introgression libraries in hybrid rye (Secale cereale L.) from an Iranian primitive accession as a new tool for rye breeding and genomics. Theor. Appl. Genet. 117, 641—652.

Falke, K., G. S. Mahone, E. Bauer, G. Haseneyer, T. Miedaner, F. Breuer, and M. Frisch, 2014: Genome-wide prediction methods for detecting genetic effects of donor chromosome segments in introgression populations. BMC Genomics. 15, 782.

Fernandez, M. G. S., P. W. Becraft, Y. Yin, and T. L€ubberstedt, 2009: From dwarves to giants? Plant height manipulation for biomass yield. Trends Plant Sci. 14, 454—461.

Frey, K., T. Cox, D. Rodgers, and P. Bramel-Cox, 1984: Increasing cereal yields with genes from wild and weedy species Genetics: new frontiers: proceedings of the XV International Congress of Genetics/editors, VL Chopra, c1984. Oxford & IBH Publishing Co., New Delhi.

Frisch, M., 2004: Breeding Strategies: Optimum Design of Marker Assisted Backcross Programs Molecular Marker Systems in Plant Breeding and Crop Improvement, 319—334. Springer, Verlag, Berlin.

Frisch, M., M. Bohn, and A. A. Melchinger, 1999: Minimum sample size and optimal positioning of flanking markers in marker-assisted backcrossing for transfer of a target gene. Crop Sci. 39, 967—975.

Goodman, M. M., 1999: Broadening the Genetic Diversity in Maize Breeding by Use of Exotic Germplasm. The Genetics and Exploitation of Heterosis in Crops, 139—148. ASA-CSSA-SSSA, Madison, WI.

Graham, R. L., R. Nelson, J. Sheehan, R. Perlack, and L. L. Wright, 2007: Current and potential US corn stover supplies. Agron. J. 99, 1—11.

Gu, J., X. Yin, P. C. Struik, T. J. Stomph, and H. Wang, 2012: Using chromosome introgression lines to map quantitative trait loci for photosynthesis parameters in rice (Oryza sativa L.) leaves under drought and well-watered field conditions. J. Exp. Bot. 63, 455—469.

Hawkes, J., 1977: The importance of wild germplasm in plant breeding. Euphytica 26, 615—621.

Hallauer, A. R., and J. B. Miranda, 1988: Quantitative Genetics in Maize Breeding, 2nd edn. By A. R. Hallauer & J. B. Miranda Filho. xii+468 pages. Ames, Iowa State University Press.

He, Y., T. Zheng, X. Hao, L. Wang, Y. Gao, Z. Hua, H. Zhai, J. Xu, Z. Xu, and L. Zhu, 2010: Yield performances of japonica introgression lines selected for drought tolerance in a BC breeding programme. Plant Breed. 129, 167—175.

Herzog, E., K. C. Falke, T. Presterl, D. Scheuermann, M. Ouzunova, and M. Frisch, 2014: Selection strategies for the development of maize introgression populations. PLoS ONE 9, e92429.

Holland, J. B., 2004: Breeding: incorporation of exotic germplasm. Encycl. Plant Crop Sci., Marcel Dekker, New York. 222—224.

Howell, P., D. Lydiate, and D. Marshall, 1996: Towards developing intervarietal substitution lines in Brassica napus using marker-assisted selection. Genome 39, 348—358.

Hu, S., and T. Lubberstedt, 2015: Getting the ‘MOST’ out of crop improvement. Trends Plant Sci. 20, 372—379. doi:10.1016/ j.tplants.2015.03.002.

Huang, X., H. Coster, M. Ganal, and M. Roder, 2003: Advanced backcross QTL analyses for the identification of quantitative trait loci alleles from wild relatives of wheat (Triticum aestivum L.). Theor. Appl. Genet. 106, 1379—1389.

Jeuken, M., and P. Lindhout, 2004: The development of lettuce backcross inbred lines (BILs) for exploitation of the Lactuca saligna (wild lettuce) germplasm. Theor. Appl. Genet. 109, 394— 401.

Koester, R. P., P. H. Sisco, and C. W. Stuber, 1993: Identification of quantitative trait loci controlling days to flowering and plant height in two near isogenic lines of maize. Crop Sci. 33, 1209—1216.

Kunert, A., A. A. Naz, O. Dedeck, K. Pillen, and J. Leon, 2007: ABQTL analyses in winter wheat: I. Synthetic hexaploid wheat (T. turgidum ssp. dicoccoides × T. tauschii) as a source of favourable alleles for milling and baking quality traits. Theor. Appl. Genet. 115, 683—695.

Ladizinsky, G., 2012: Plant Evolution Under Domestication. Springer Science & Business Media, Netherlands.

Lima, M.d. L. A., C. L. Jr de Souza, D. A. V. Bento, A. P. de Souza, and L. A. Carlini-Garcia, 2006: Mapping QTL for grain yield and plant traits in a tropical maize population. Mol. Breed. 17, 227—239.

Lin, S., T. Sasaki, and M. Yano, 1998: Mapping quantitative trait loci controlling seed dormancy and heading date in rice, Oryza sativa L., using backcross inbred lines. Theor. Appl. Genet. 96, 997—1003.

Liu, S., R. Zhou, Y. Dong, P. Li, and J. Jia, 2006: Development, utilization of introgression lines using a synthetic wheat as donor. Theor. Appl. Genet. 112, 1360—1373.

Liu, S., H. D. Chen, I. Makarevitch, R. Shirmer, S. J. Emrich, C. R. Dietrich, W. B. Barbazuk, N. M. Springer, and P. S. Schnable, 2010: High-throughput genetic mapping of mutants via quantitative single nucleotide polymorphism typing. Genetics 184, 19—26.

Lubberstedt, T., A. E. Melchinger, C. C. Schon, H. F. Utz, and D. Klein, 1997: QTL mapping in testcrosses of European flint lines of maize: I. Comparison of different testers for forage yield traits. Crop Sci. 37, 921—931.

Matus, I., A. Corey, T. Filichkin, P. Hayes, M. Vales, J. Kling, O. RieraLizarazu, K. Sato, W. Powell, and R. Waugh, 2003: Development and characterization of recombinant chromosome substitution lines (RCSLs) using Hordeum vulgare subsp. spontaneum as a source of donor alleles in a Hordeum vulgare subsp. vulgare background. Genome 46, 1010—1023.

Melchinger, A. E., H. F. Utz, and C. C. Sch€on, 1998: Quantitative trait locus (QTL) mapping using different testers and independent population samples in maize reveals low power of QTL detection and large bias in estimates of QTL effects. Genetics 149, 383—403.

Miedaner, T., B. Muller, H. P. Piepho, and K. Falke, 2011: Genetic architecture of plant height in winter rye introgression libraries. Plant Breed. 130, 209—216.

Peiffer, J. A., M. C. Romay, M. A. Gore, S. A. Flint-Garcia, Z. Zhang, M. J. Millard, C. A. Gardner, M. D. McMullen, J. B. Holland, and P. J. Bradbury, 2014: The genetic architecture of maize height. Genetics 196, 1337—1356.

Pillen, K., A. Zacharias, and J. Leon, 2003: Advanced backcross QTL analyses in barley (Hordeum vulgare L.). Theor. Appl. Genet. 107, 340—352.

Ramsay, L., D. Jennings, M. Kearsey, D. Marshall, E. Bohuon, A. Arthur, and D. Lydiate, 1996: The construction of a substitution library of recombinant backcross lines in Brassica oleracea for the precision mapping of quantitative trait loci. Genome 39, 558—567.

Ribaut, J.-M., and M. Ragot, 2007: Marker-assisted selection to improve drought adaptation in maize: the backcross approach, perspectives, limitations, and alternatives. J. Exp. Bot. 58, 351— 360.

Saghai-Maroof, M. A., K. M. Soliman, R. A. Jorgensen, and R. Allard, 1984: Ribosomal DNA spacer-length polymorphisms in barley: Mendelian inheritance, chromosomal location, and population dynamics.

Proc. Natl Acad. Sci. USA 81, 8014—8018. Salhuana, W., and L. Pollak, 2006: Latin American maize project (LAMP) and germplasm enhancement of maize (GEM) project: generating useful breeding germplasm. Maydica 51, 339.

Sayed, M. A. E.-A. A., 2011: QTL analyses for drought tolerance related to root and shoot traits in barley (Hordeum vulgare L.), Bonn, Univ., Diss., 2011.

Schon, C. C., H. F. Utz, S. Groh, B. Truberg, S. Openshaw, and A. E. Melchinger, 2004: Quantitative trait locus mapping based on resampling in a vast maize testcross experiment and its relevance to quantitative genetics for complex traits. Genetics 167, 485—498.

Schwietzke, S., Y. Kim, E. Ximenes, N. Mosier, and M. Ladisch, 2009: Ethanol Production from Maize Molecular Genetic Approaches to Maize Improvement, 347—364. Springer, Verlag Berlin Heidelberg.

Septiningsih, E., J. Prasetiyono, E. Lubis, T. Tai, T. Tjubaryat, S. Moeljopawiro, and S. McCouch, 2003: Identification of quantitative trait loci for yield and yield components in an advanced backcross population derived from the Oryza sativa variety IR64 and the wild relative O. rufipogon. Theor. Appl. Genet. 107, 1419—1432.

Siangliw, J. L., B. Jongdee, G. Pantuwan, and T. Toojinda, 2007: Developing KDML105 backcross introgression lines using marker-assisted selection for QTLs associated with drought tolerance in rice. Sci. Asia 33, 207—214.

Sibov, S. T., J. de Souza, C. Lopes, A. A. Garcia, A. R. Silva, A. Garcia, C. A. Mangolin, L. L. Benchimol, and A. P. de Souza, 2003: Molecular mapping in tropical maize (Zea mays L.) using microsatellite markers. 2. Quantitative trait loci (QTL) for grain yield, plant height, ear height and grain moisture. Heredities 139, 107—115.

Stam, P., and A. Zeven, 1981: The theoretical proportion of the donor genome in near-isogenic lines of self-fertilizers bred by backcrossing. Euphytica 30, 227—238.

Szalma, S., B. Hostert, J. LeDeaux, C. Stuber, and J. Holland, 2007: QTL mapping with near- isogenic lines in maize. Theor. Appl. Genet.114, 1211—1228.

Tan, L., F. Liu, W. Xue, G. Wang, S. Ye, Z. Zhu, Y. Fu, X. Wang, and C. Sun, 2007: Development of Oryza rufipogon and O. sativa introgression lines and assessment for yield- related quantitative trait loci. J. Integr. Plant Biol. 49, 871—884.

Tanksley, S., and J. Nelson, 1996: Advanced backcross QTL analyses: a method for the simultaneous discovery and transfer of valuable QTLs from unadapted germplasm into elite breeding lines. Theor. Appl. Genet. 92, 191—203.

Thurber, C. S., J. M. Ma, R. H. Higgins, and P. J. Brown, 2013: Retrospective genomic analyses of sorghum adaptation to temperate-zone grain production. Genome Biol. 14, R68.

Tian, F., D. J. Li, Q. Fu, Z. F. Zhu, Y. C. Fu, X. K. Wang, and C. Q. Sun, 2006: Construction of introgression lines carrying wild rice (Oryza rufipogon Griff.) segments in cultivated rice (Oryza

70 sativa L.) background and characterization of introgressed segments associated with yield-related traits. Theor. Appl. Genet. 112, 570—580.

Van Berloo, R., 2008: GGT 2.0: versatile software for visualization and analyses of genetic data. J. Hered. 99, 232—236.

Von Korff, M., H. Wang, J. Leon, and K. Pillen, 2004: Development of candidate introgression lines using an exotic barley accession (Hordeum vulgare ssp. spontaneum) as donor. Theor. Appl. Genet. 109, 1736—1745.

Von Korff, M., H. Wang, J. Leon, and K. Pillen, 2005: AB-QTL analyses in spring barley. I. Detection of resistance genes against powdery mildew, leaf rust and scald introgressed from wild barley. Theor. Appl. Genet. 111, 583—590.

Von Korff, M., H. Wang, J. Leon, and K. Pillen, 2008: AB-QTL analyses in spring barley: III. Identification of exotic alleles for the improvement of malting quality in spring barley (H. vulgare ssp. spontaneum). Mol. Breed. 21, 81—93.

Wang, Y., J. Yao, Z. Zhang, and Y. Zheng, 2006: The comparative analyses based on maize integrated QTL map and meta-analyses of plant height QTLs. Chin. Sci. Bull. 51, 2219—2230.

Zamir, D., 2001: Improving plant breeding with exotic genetic libraries. Nat. Rev. Genet. 2, 983—989.

Zamir, D., and Y. Eshed, 1998: Tomato Genetics and Breeding Using Nearly Isogenic Introgression Lines Derived from Wild Species. Molecular Dissection of Complex Traits, 207— 217. CRC, Boca Raton, FL, USA.

Zeng, L., W. R. Meredith, D. L. Boykin, and E. Taliercio, 2007: Evaluation of an exotic germplasm population derived from multiple crosses among Gossypium tetraploid species. J. Cotton Sci 11, 118—127.

Figure 1. Breeding scheme to generate one PIF/the donor accession is a segregating exotic source population. Bulked pollen from the donor accession was used to pollinate the recurrent parent (RP). As the F1 is segregating, random F1 plants were used to backcross (BC) with RP.

BC1F1 families were harvested separately, and for selected BC1F1 families (similar flowering time with RP), the two tallest plants (O) were selected and backcrossed with RP, one to be planted next season and one as a backup. This process was repeated in following BC generations.

The most advanced PIF per donor × background combination in terms of BC generations was used for further phenotypic and genotypic characterization in this study.

Table 1. Exotic accessions used for establishing PIFs

No. PIFB47 PIFZ51

Amarillo ABANERO 1 Huancabamba – CUN342 PIU 17B

ALTIPLANO 2 ANCASH 291 BOV903

Amarillo ANDELA 3 Huancabamba – PIU ECU699 ICA 38B B

ANCASHINO Araguito – VEN 4 ANC102 678

ANDAQUI Arequipeno – 5 CAQ307 FT ARQ 1 B

ANDELA ECU Argentino – 6 699 BOV 920

ARAGUITO Arizona – LIB 7 VEN678 16 B

Table 1 continued

Arequipeno – AvatiPichingaIh 8 ARQ 1 B) u – BR 2830

AvatiMorotiGuapi Blanco Ayabaca 9 – PAG 139 B – PIU119 B

Blanco Ayabaca – Blanco Blandito 10 PIU119 B – ECU 523 B

Blanco Blandito – 11 BOFO DGO123 ECU 523 B

BR105:N99v99 12 BR106:S99n99n v

CABUYA Calchaqui white 13 SAN316 FTC flint – ARG 2420

CAMELIA CANGUIL GR. 14 CHI411 ECU447

CANDELA Canilla – 15 ECU531 VEN693 B

Table 1 continued

Canilla – VEN Capiorosado – 16 693 B ARG 460

CatetoNortista – Chandelle – 17 GIN I B VEN 409

Chillo – ECU 458 Chillo – ECU 18 B 411

Chirimito – VEN Chirimito – 19 703 B VEN 703 B

Comiteco – GUA Clavito – 20 515 B ECU884 B

CON NORT Clavo – NAR 21 ZAC161 329

CON PUNT Comiteco – 22 CUZ13 GUA 515 B

CONICO CON PN 23 [PUE116]{CIMYT CUZ13

Table 1 continued

Costeño – Coroicoblanco 24 Antioquia 394 B – BOV 406

COSTENO Cravoriogrande 25 [VEN775]{ICA} nse – RGS VII B

Cravoriograndens Cristal – SP X 26 e – RGS VII B B

Cubanodentado 27 Cuzco gigante – BOV 585

DULCILLO DE Culli – ARG 28 NO SON57 471

CUZCO 29 DZ B GUA131 CUZ217

EARLY 30 CARIBBEAN Cuzco gigante MAR9

Table 1 continued

ELOTES DE T CAL. 31 OCCIDENT NAY29 GUA159

GORDO DULCILLO 32 [CHH131]{CIMYT} DEL NO SON57

ELOTES IMA 66 33 OCCIDENT Ames8554 DGO236

ELOTES KARAPAMPA 34 OCCIDENT BOV978 NAY29

MISHCA Huachano – 35 ECU321 LIM 43 B

MONTANA Huarmaca – 36 NAR625 PIU 72

MORADO Huevito – VEN 37 BOV567 396 B

Table 1 continued

38 MOROC APC67 Jora – ANC 1 B

NANO # KCELO 39 BOV1032 BOV948

MISHC 40 OLITA DGO159 ECU321

OLOTON Morchon – 41 [GUA383]{CIMYT} ECU454B

ONAVENO Morocho – APC 42 SON24 77

MOROCHO 43 PARO CUZ76 APC67

PATILLO 44 MOROTI PEI GRANDE BOV649

Negrita de 45 PERLA ANC23 tierrafria – GUA 522

NINUELO 46 PISAN BOV344 BOV1088

Table 1 continued

POJ CHICO 47 Oke – ARG 539 BOV800

PUE116 OLLO CUN424 48 {CIMYT} ICA

Puya – MAG 355 Patillo – ECU 49 B 417

Rabo De Zorro – Piricinco – SM 50 ANC 325 B 8 B

Racimo de Uva – PISAN 51 ECU 517 B BOV344

Rabo De Zorro 52 SABN CUN367 – ANC 325 B

SAN MARCENO Racimo de Uva 53 # [GUA724]{INIA} – ECU 517 B

Semi Rienda – CAJ 54 dentadopaulista – 80 B PAG I B

Table 1 continued

SARCO 55 TEPE CHS225 ANC184

Tepecintle – GUA SG HUANCAV 56 65 B JUN164

Sin Vandeño – GRO 57 Clasificación – 96 B CAU454

YUCATAN TuxpeñoNorteñ 58 TOL389 ICA o – CHH287

UCHIM 59 ECU681

Uchuquilla – 60 BOV318

Vandeño – 61 GRO 96 B

YUNGUENO 62 BOV362

Figure 2. Ranges of plant height of 126 (PIFZ51) and 100 (PIFB47) PIFs and the 23 selected phenotype-selected introgression families (PIFB47S and PIFZ51S)/the box defined between the first quartile (Q1) and the third quartile (Q3) represent 50% of the backcross family values. The line in the box indicates the median value. The two horizontal rows represent the means of the two recurrent parents, the upper one represents for PHB47, and the bottom one represents for

PHZ51.

Table 2. Estimates of variance components, broad-sense heritability h2 for phenotype-selected introgression families (PIFs).

Trait Variance component

Location Genotype G × E Error h2

PHT 450.26** 154.54** 31.81** 50.94** 0.8

DTT 10.31** 0.74** 0.25** 0.31** 0.7

PHT, plant height; DTT, days to tasseling; G × E, genotype by environment interaction. ‘*’ and ‘**’ show the significances at 0.05 and 0.01 of probability level, respectively.

Table 3. Genetic characterization of PIFZ51S: Proportion of donor parent genome and its expected values and the observed number of donor chromosomal segments (DCS).

Expected Observed PIF Donor BC donor donor No.DCS

Chandelle – 1 BC 12.5 22.5 10 VEN 409 2

Capiorosado 2 BC 12.5 10.1 7 – ARG 460 2

Negrita de 3 tierrafria – BC2 12.5 17.3 11 GUA 522

Blanco 4 Blandito – ECU BC2 12.5 35.3 15 523 B

Rabo De 5 Zorro – ANC BC2 12.5 12.7 8 325 B

Huevito – 6 BC 12.5 21.2 10 VEN 396 B 2

Piricinco – 7 BC 12.5 23.4 10 SM 8 B 2

Culli – ARG 8 BC 12.5 13.4 9 471 2

Table 3 continued

Huachano – 9 BC 12.5 31.8 12 LIM 43 B 2

BOFO 10 BC 3.125 56 19 DGO123 4

Racimo de 11 Uva – ECU 517 BC2 12.5 64.6 10 B

Chillo – ECU 12 BC 12.5 25.4 11 458 2

BOFO 13 BC 3.125 10.9 7 DGO123 4

BOFO 14 BC 6.25 14.3 7 DGO123 3

Sin 15 Clasificación – BC2 12.5 18.1 9 CAU454

Jora – ANC 16 BC 12.5 14.3 8 1 B 2

Patillo – 17 BC 12.5 15.5 10 ECU 417 2

Comiteco – 18 BC 12.5 20.4 11 GUA 515 B 2

Table 3 continued

BOFO 19 BC 6.25 12.4 7 DGO123 3

Arizona – 20 BC 6.25 8.7 4 LIB 16 B 3

Chillo – ECU 21 BC 12.5 19.6 10 411 2

Cuzco 22 BC 12.5 8.4 4 gigante 2

Canilla – 23 BC 12.5 14.1 11 VEN693 B 2

Avg. PIFZ51S (23 10.8 21.3 9.6 ILs) Avg. GEM- BC DH-Z51 lines 1 25 16.5 8.95 -DH (20 ILs)

BC, backcross generation; DH, doubled haploid; GEM, Germplasm Enhancement of Maize; PIF, phenotype-selected introgression families.

Table 4. Genetic characterization of PIFB47S: Proportion of donor parent genome and its expected values and the observed number of donor chromosomal segments (DCS).

Expected Observed PIF Donor BC donor donor No.DCS

Blanco Ayabaca – 1 BC 12.5 10 4 PIU119 B 2

ELOTES OCCIDENT 2 BC 6.25 32.5 18 NAY29 3

Racimo de Uva – ECU 3 BC 12.5 9 9 517 B 2

4 Vandeño – GRO 96 B BC2 12.5 20.1 13

Amarillo 5 Huancabamba – PIU 38B BC2 12.5 18.2 12 B

PHB47/Chillo – ECU 6 BC 12.5 11.6 8 458 2

Comiteco – GUA 515 7 BC 12.5 10.5 5 B 2

Rabo De Zorro – ANC 8 BC 12.5 18.7 10 325 B 2

Table 4 continued

CatetoNortista – GIN I 9 BC 12.5 29.8 11 B 2

B47NANO # 10 BC 6.25 5.4 5 BOV1032/PHB47 3

Semi dentadopaulista – 11 BC 12.5 19.5 9 PAG I B 2

12 CON PUNT CUZ13 BC3 6.25 6.7 4

GORDO 13 [CHH131]{CIMYT})-B- BC2 12.5 8.6 5 B-SIB-023-001-001)#001

14 CAMELIA CHI411 BC3 6.25 10.3 8

Costeño – Antioquia 15 BC 12.5 22.2 11 394 B 2

B47/COSTENO 16 [VEN775]{ICA})-B-B- BC3 6.25 12.2 8 SIB-018-001-001)#001

Blanco Ayabaca – 17 BC 12.5 15.6 9 PIU119 B 2

18 Cuzco gigante BC1 25 24.5 12

Table 4 continued

19 Cuzco gigante BC2 12.5 11.5 5

Blanco Ayabaca – 20 BC 6.25 6.9 4 PIU119 B 3

21 Canilla – VEN 693 B BC2 12.5 20.4 7

22 TEPE CHS225 BC3 6.25 6.3 4

Blanco Blandito – 23 BC 12.5 18 11 ECU 523 B 2

Mixe Avg. PIFB47S (23 ILs) 11.1 15.2 8.3 d

Avg. GEM-DH-B47 BC - 1 25 15.5 6.8 lines (20 ILs) DH

BC, backcross generation; DH, doubled haploid; GEM, Germplasm Enhancement of Maize; PIF, phenotype-selected introgression families.

Figure 3: Genotypic characterization of the top 23 phenotype-selected introgression families within PIFB47S/the marker position (vertical bars) is presented above the figure, black represents heterozygous state of the donors, and grey represents for recurrent parents.

Figure 4: Genotypic characterization of the top 23 phenotype-selected introgression families within PIFZ51S/the marker position (vertical bars) is presented above the figure, black represents heterozygous state of the donors, and grey represents for recurrent parents.

Table 5. Positions and frequencies of selected hot spots.

Position Donor Chr. (cM) frequency 1 0 19.55 1 93.5 21.75 1 107.2 15.2 1 117.1 23.9 1 125.3 17.39 1 174.5 17.39 1 221 15.2 2 1.1 19.57 2 21.4 23.9 3 174.5 17.39 3 201.9 26.05 4 16 28.25 4 29.7 21.74 4 94.1 15.2 4 101 34.75 5 16 19.55 5 70.8 15.2 5 71.9 17.35 6 12.1 41.3 6 47.7 21.74 6 79.7 17.39 8 41.3 19.55 8 59 15.2 8 62.2 15.2 8 77.5 17.4 8 80.8 17.35 9 72.4 19.55 10 140.1 19.57

CHAPTER 4.

GIBBERELLINS PROMOTE BRASSINOSTEROIDS AND BOTH

INCREASE HETEROSIS FOR PLANT HEIGHT IN MAIZE

Songlin Hu1, Cuiling Wang2, Darlene L. Sanchez1, Alexander E. Lipka3, Peng Liu4, Yanhai Yin5,

Michael Blanco6, Thomas Lübberstedt1*

Paper submitted to PLoS Genetics Journal. Abstract, structure, and references are formatted according to the journal standards.

SH and TL conceived the study, designed the experiments, discussed the results and finalized the manuscript; CW identified the candidate genes in BR and GA pathway; DLS collected genotype and phenotype data for doubled haploids as a comparison; AEL, PL helped with statistical analysis. YY established protocol for measuring BR/GA inhibitor response. TL, AEL, PL, YY,

MB edited the manuscript; all authors read and approved the final manuscript.

Abstract

Brassinosteroids (BR) and Gibberellins (GA) are two classes of plant hormones which have shown great agronomic potential to promote plant height (PHT). A dwarf or tall stature of crops can be achieved through an increased or decreased BR and GA level or signaling, which is beneficial for grain or biomass production, respectively. In this study, we established two maize

92 backcross libraries with substantial variation for the level of heterozygosity and PHT, with two doubled haploid libraries as comparison. We conducted genome wide association studies to explore the genetic control of PHT by BR and GA. In addition, we used BR and GA inhibitors to compare the relationship between PHT, BR, and GA in inbred lines and heterozygotes from a physiological and biological perspective. A total of 73 genomic loci were discovered to be associated with PHT, with 7 co-localized with GA, and 2 very close to BR candidate genes. Field

PHT was significantly correlated with seedling stage BR and GA inhibitor responses. However, this observation was only valid for maize heterozygotes, but not for inbred lines. Our path analysis results demonstrated that higher level of heterozygosity led to elevated GA activities.

Increased GA promoted BR level, and both increase heterosis in maize PHT. We concluded that seedling stage hormone inhibitor response is promising for predicting heterosis for field PHT.

Author Summary

Heterosis is the phenomenon in which hybrids exhibit superior phenotypes compared to their parents. Although plant breeders have utilized heterosis for the development of superior yielding varieties in many important crop species, the molecular basis of heterosis has not been fully understood. Here, we provided new insights to show that two major plant hormones brassinosteroids (BR) and gibberellins (GA) was involved in the regulation of heterosis in maize plant height. We found that seedling stage BR and GA inhibitor response can be correlated with field plant height for heterozygotes, but not for inbred lines. We employed path analysis, which revealed that higher level of heterozygosity lead to higher GA and BR (due to crosstalk with GA) levels, and these two plant hormones promote heterosis for plant height. Compared to previous

93 reports to use genes, metabolites, or transcripts, we were able to use plant hormones to predict heterosis for plant height in maize. This knowledge enriches the knowledge of plant hormone control of plant performance, and will benefit future maize hybrid breeding projects.

Introduction

Increasing demand for biomass production led to a paradigm shift regarding plant height

(PHT) from dwarfs to giants [1]. Tall maize varieties are more desirable, if maize is used as dual purpose or biomass crop. For dual-purpose maize, increased stover biomass adds value: grain is harvested for food or feed, stover for bioenergy conversion. While PHT increasing alleles contribute to increasing biofuel production, they also increase the risk of lodging. To overcome this negative side effect, breeders can either produce maize varieties with increased lignin level or a stronger root system to stabilize plants [1]. Increased lignin levels are not desirable for biochemical, but favorable for thermochemical conversion of biomass [2].

PHT is an important agronomic trait in modern maize and more generally cereal breeding programs, and has been manipulated during maize domestication as it shows significant correlations with different agronomic traits such as grain yield [3, 4]. For grain maize production, breeders prefer a short stature, as high yielding maize varieties need to be lodging-tolerant under high nitrogen levels and high density planting conditions. Breeders use semi-dwarf genes to moderately decrease PHT in cereals, such as the green revolution genes sd-1 in rice [5] and Rht in wheat [6]. In maize, several maize dwarf genes have been well characterized, but are not intentionally used in breeding programs due to their adverse impact on grain yield, such as dwarf3 [7], dwarf8 and dwarf9 [8, 9], nana plant1 [10], and brd1 [11]. All these genes cause

94 defects in either the brassinosteroid (BR) or the gibberellin (GA) pathway, stressing the importance of these two plant hormones in the control of PHT. It was reported that BR and GA have the most direct effects on PHT without major negative pleiotropic effects as compared to other plant hormones [1].

Plant hormone inhibitors are powerful tools for elucidating plant hormone functions during plant development. For example, BR inhibitor Propiconazole (Pcz) has successfully been employed for studying BR control of sex determination and PHT in maize [12]. GA inhibitor

Uniconazole (Ucz) was used to explore root growth and nitrogen transfer in soybean [13].

Instead of using hormone pathway mutants, plant hormone inhibitors can phenocopy hormone- deficient mutants in crops, with the advantage that deficiency levels of hormones can be controlled. Moreover, plant hormone inhibitors can help with identification and characterization of hormone deficient mutants without prior knowledge of the mutant phenotype [12]. In maize,

Pcz and Ucz are two popular plant hormone inhibitors of BR and GA, respectively, due to their easy accessibility and low costs [12, 14]. Application of Pcz and Ucz reduce mesocotyl elongation, and genotypes with elevated BR or GA level are more tolerant to Pcz or Ucz, resulting in alleviated reduction of mesocotyl length [12]. Therefore, Pcz and Ucz can be employed to explore the relationship between morphological traits, BR and GA activities in maize.

In this study, two sets of backcross (BC) libraries derived by introgression of a diverse set of tropical maize accessions into inbred lines, representing two major maize heterotic groups (Iowa

Stiff Stalk and Non Stiff Stalk), were evaluated for PHT. Moreover, we applied BR and GA

95 inhibitors Pcz and Ucz at seedling stage to compare BR and GA inhibitor responses between tall and short maize BC families within each library, and between the tallest and shortest individuals within BC families. In addition, two sets of doubled haploid (DH) libraries derived from the same parents producing the two BC libraries were tested for PHT and BR/GA activities as a comparison. Our objectives were (1) to evaluate the PHT performance of the two BC libraries and the two DH libraries; (2) to conduct a genome-wide association study (GWAS) to investigate the genomic regions associated with BR, GA, and PHT in these BC families; and (3) to apply

Pcz and Ucz treatments to these two BC libraries and two DH libraries to investigate the relationship between BR, GA, and PHT in both inbred lines (the two DH libraries) and heterozygotes (the two BC libraries).

Results

All statements about statistical inferences are based on a type I error rate (or family-wise error rate) of 5% unless P-values are specifically given in the result section. As described in the section of Methods, two libraries of phenotypic-selected introgression families (PIFs): PIFB47

(PHB47 as recurrent parent) and PIFZ51 (PHZ51 as recurrent parent), and two doubled haploid

(DH) libraries (DHB47 and DHZ51, without any phenotype selection) were compared with their recurrent parent. Most PIFs were significantly taller than their recurrent parent (Fig 1) and showed consistent PHT performance (heritability: h2=0.81) across three years (S1 Fig). The percentages of PIFs significantly taller than their recurrent parent in each year were: 91.6%

(2013), 92.5% (2014), 92.0% (2015) for PIFB47; and 85.7% (2013), 85.5% (2014), and 90.0%

(2015) for PIFZ51. Even for the shortest individual within each PIF (SPHT), 91.2% and 59.4% of these individuals were taller than recurrent parent within PIFB47 or PIFZ51, respectively. In

96 contrast, the proportion of taller BGEM lines (DH lines from DHB47 and DHZ51: See methods section) compared with their recurrent parent was only 23% for DHZ51 and 18% for DHB47.

PIFB47 was on average 23 cm taller than PIFZ51, and PHB47 was on average 8 cm taller than

PHZ51. For the BGEM lines, DHB47 was on average 2 cm taller than DHZ51. None of the PIFs were significantly taller than the F1 hybrid of PHB47×PHZ51.

All PIFs flowered within three days compared to the respective recurrent parent for both days to tasseling (DDT) and days to silking (DTS). The average anthesis - silking interval (ASI) was 0.3 and 0.6 days for PIFB47 and PIFZ51, respectively, and thus similar to PHB47 (0.2 days) and PHZ51 (0.4 days). Differences of DDT and DTS between BGEM lines and recurrent parent ranged from 0-13 days in DHB47 and 0-15 days in DHZ51. The F1 hybrid of PHB47×PHZ51 flowered four or six days earlier than PHB47 and PHZ51, respectively. Selection for PHT affected other agronomic traits including ear height (EHT), number of nodes (NNode), and tassel length (TL) but not leaf angle (LA). EHT, NNode and TL increased along with an increased PHT

(Table 1) and LA is not associated with other traits. Summary statistics for other traits are listed in S1 Table.

Genotypic characterization

We defined donor genome proportion (DGP) of PIFs as 50% of heterozygosity (percentage of heterozygous markers) in this study, as for each heterozygous locus, only one allele is from the donor parent. On average, PIFs had a higher DGP than expected by chance (Table 2). For example, the expected DGP without selection is on average 0.78% for BC6 individuals, but the observed DGP for BC6 PIFs was 17.2% and 6.2% in PIFB47 and PIFZ51, respectively. On

97 average, PIFB47 contained 12.9% higher DGP compared to PIFZ51. In contrast, BGEM lines showed a slightly reduced average DGP of 22.3% in DHB47 and 23.8% in DHZ51, compared with the expected 25% of BC1-DH. DGP and PHT are significantly correlated for both PIFB47

(r=0.72) and PIFZ51 (r=0.70). However, there was no significant linear relationship between

DGP and PHT for BGEM lines.

Hormone inhibitor responses of PIFs and BGEM lines

We applied BR and GA inhibitors to PIFB47, PIFZ51, DHB47 and DHZ51 to calculate the

BR and GA inhibitor responses. The BR and GA inhibitor Pcz and Ucz greatly reduced mesocotyl length compared to water treatment (Fig 2). Heritabilities of BR inhibitor response

(MP/MW; see methods section) and GA inhibitor response (MU/MW) were 0.74 and 0.80, respectively. Ucz showed stronger effects than Pcz for reducing the mesocotyl length for both

PIFs and BGEM lines (Fig 3). PHZ51 showed stronger BR and GA inhibitor response compared to PHB47. MP/MW was 0.19 and 0.35 for PHB47 and PHZ51, and MU/MW was 0.07 and 0.20 for PHB47 and PHZ51, respectively. 69% (PIFB47) and 45% (PIFZ51) of PIFs had increased

BR inhibitor response (e.g., PIF58 in S2 Fig) compared to their respective recurrent parent. GA inbititor response was 68% (PIFB47) and 38% (PIFZ51) increased for PIFs (e.g., PIF59 in S2

Fig), compared to their respective recurrent parent. In contrast, only 6% to 13% of BGEM lines showed significant increased BR/GA inhibitor response compared to their recurrent parent, and the BR and GA inhibitor responses of the recurrent parents were always close to the median of

DHB47 and DHZ51 (without phenotype selection) (Fig 3). The F1 hybrid of PHB47 × PHZ51 showed significantly higher hormone inhibitor response for both BR and GA than most of the

PIFs. Only one PIF from PIFZ51 showed higher BR inhibitor response than the hybrid of PHB47

× PHZ51 (Fig 3).

BR and GA control of PHT

Within both PIFB47 and PIFZ51: BR and GA inhibitor responses, and PHT were significantly correlated (Table 3). PIFs with stronger BR or GA inhibitor response were taller in the field. In contrast, no significant correlation was found between BR inhibitor response and

PHT or between GA inhibitor response and PHT for BGEM lines. When assessed between the tallest individual and the shortest individual within PIFs, seed harvested from the tallest individual showed stronger hormone inhibitor response (either BR or GA or both) compared with the seed harvested from the shortest individual (S2 Table). For example, seed from the tallest individual was more tolerant to GA inhibitor compared to the seed from the shortest one of

PIF113 (P=0.0003) (Fig 4). These PIFs contained a high level of heterozygosity (%) with on average 40.4% (PIFB47) and 17.8% (PIFZ51).

Path analysis was used to connect the level of heterozygosity, BR inhibitor response, GA inhibitor response, and PHT separately within both PIFB47 and PIFZ51. Starting from the proposed model (see methods section), new links were added and old links were deleted based on the correlation structure. Path analysis for PIFB47 and PIFZ51 arrived at the same final best model, for which the direct correlation between heterozygosity and BR was deleted, and new links between BR and GA, and between heterozygosity and PHT was added (Fig 5) compared to the original proposed model. This was the only model fitting all criteria (Summary statistics in

S3 Table) and it was associated with the lowest Akaike information criterion (AIC) value.

According to the final model a high level of heterozygosity has a direct positive effect on PHT and two indirect paths to impact PHT: (i) a higher level of heterozygosity increases the GA level

(synthesis or signaling), and increased GA promotes PHT; (ii) increased GA causes an increased

BR level/signaling, which then promotes PHT.

Genome-wide association analysis

Three sub-populations were obtained for joint analysis of PIFB47 and PIFZ51 (S3 Fig), with two subpopulations containing PHB47 and PHZ51, respectively (with limited donor introgression), and a mixed subpopulation containing a large (~30 %) donor genome proportion.

With three GWAS models MLM, GLM+Q, FarmCPU for balancing false positives and false negatives (See Methods Section), we found that none of the SNP markers was found to be significantly associated with BR/GA inhibitor responses, PHT and other agronomic traits with a

MLM (Q+K) model. With the FarmCPU method (Q+K model, but with controlled confounding effect between covariates (Q, K) and testing markers), we found one SNP SYN38535 on

Chromosome 5 (282,014,45 bp; P-value 1.39 × 10-15) significantly associated with PHT (Fig 6) and one SNP PZE-108005623 on Chromosome 8 (567,899,5 bp; P-value 5.45 × 10-8) significantly associated with GA inhibitor response (Fig 6). QQ-plot showed that population structure and kinship controlled false positives effectively (Fig 6). With a GLM+Q model, 73

SNPs were significantly associated with PHT (S4 Table) including SYN38535 from FarmCPU, and 22 of them overlapping with the 40 published PHT quantitative trait loci (QTL) hot spots

[15]. Of these 73 SNPs, 41 were located within genes and these gene functions were summarized in S5 Table according to the function of their orthologues in rice and Arabidopsis. Among these

100 genes, the four genes GRMZM2G348452, GRMZM2G082191, GRMZM2G100423, and

GRMZM2G025171 are promising for further investigation, as the orthologues of these four genes

- cytokinin oxidase, receptor-like protein kinase, cytochrome P450 and ATP synthase in

Arabidopsis and rice are functioning in PHT control [16-19].

For GA and BR inhibitor response, we found five, and two significant SNPs with GLM+Q method, respectively (S4 Table), with one overlapping SNP marker SYN19928 on Chromosome

8 (128,133,446 bp) associated with both BR (P-value 3.62×10-5) and GA (P-value 1.66×10-5) inhibitor response. This SNP marker was found to be the closest linked marker in our dataset to a

BR signaling pathway candidate gene ZmBSU1 with a 200kb physical distance. For the other four SNPs associated with GA inhibitor response, three were close to GA candidate genes: PZE-

108005623 (overlapped with FarmCPU method) and PZA00058.6 (Chromosome 8; 602,357,7 bp) are clustered and 2 Mb away from GA signaling candidate gene ZmRBX1A. Marker PZE-

108070783 (Chromosome 8; 123,786,918 bp) is 3Mb away from GA signaling candidate gene

ZmCUL1. These three candidate genes ZmBSU1, ZmRBX1A, and ZmCUL1 are expressed at seedling stage with the same absolute expression value of 3475.53 [20]. For the other traits, four markers were associated with EHT, one with NNode, and one with LA (S4 Table). None of these markers were close to any BR or GA candidate genes.

Co-localization of BR and GA candidate genes with PHT associated SNP markers

We systematically used orthologue information from Arabidopsis and rice to find all GA and

BR candidate genes in maize. Compared with a different method, which used all previously reported genes encoding GA metabolism enzymes in maize and other species as BLAST queries

101 to find GA candidate genes, all the GA candidate genes found in this study matched with their results [21]. With 1Mb bin size, seven PHT associated SNP markers co-localized with GA candidate genes (Table 4). Except for marker PZA02388.01 and PZE-108073190, which are the third and second closest SNP markers to the candidate gene, all the other SNP markers are the closest SNP markers included in our marker dataset to the candidate gene. The probability of finding seven PHT associated SNP markers co-localized with GA candidate genes by chance with a 1Mb bin size within our marker dataset is P=0.041 based on a non-parametric resampling method. We did not find BR candidate genes co-localized with PHT associated SNP markers with 1Mb bin size. The two closest PHT associated SNP markers PZE-108100090 and PZE-

109116158 were 1.13 and 1.25 Mb away from two BR candidate genes, ZmBRI1-1 and

ZmBSK1-1, respectively.

Discussion

Phenotypic selection for PHT and flowering time was successful in this study. Most PIFs

(~90%) were significantly taller than their respective recurrent parent (Fig 1), while flowering undistinguishable from their recurrent parents. Selection for PHT resulted in high levels of heterozygosity (Table 2). The percentage of heterozygous markers for PIFB47 and PIFZ51 after four to six generations of backcrossing was on average 34% and 9%, respectively. Percentage of heterozygosity was correlated with PHT in both PIFB47 and PIFZ51 (r~0.7). In contrast, only about 20% of BGEM lines were significantly taller than their recurrent parent, and there was no significant correlation between donor genome proportion (DGP) and PHT. Without selection, the observed DGP is around 23%, close to the expected 25% for BC1-DH lines.

102

Stronger selection was observed in PIFB47. PIFB47 was on average 23 cm taller than

PIFZ51, and this difference was larger than the difference between PHB47 and PHZ51 (8cm).

Moreover, PIFB47 contained on average 25% more heterozygous markers than PIFZ51. In comparison, DHB47 was on average 2cm taller than DHZ51, and both of them had an observed

DGP around 23%. In addition, we compared a subset of PIFs from PIFB47 and PIFZ51, which were derived from the same set of donor parents backcrossed with both PHB47 and PHZ51. On average, these PIFs were 30 cm taller for PIFB47 compared to PIFZ51. The stiff stalk group

(including PHB47) is more distantly related to tropical germplasm than the non-stiff stalk group

(including PHZ51) [22]. Thus, the chance of heterotic effects increases for crosses between

PHB47 and tropical germplasm. Previous studies for U.S. maize germplasm showed that panmictic midparent heterosis (PMPH) linearly increased with increasing genetic distance [23], and PMPH also increased with increasing genetic distances among tropical and U.S. germplasm, unless affected by maladaptation problems [24]. An increased divergence between two genotypes of a heterotic pattern increases the probability to select for complementary favorable alleles at different loci [25], which is consistent with our findings of accumulating more exotic regions at heterozygous state in PHB47 background.

BR and GA inhibitor responses are correlated with PHT for PIFs

BR and GA are two classes of plant hormones that are regarded as major pathways controlling PHT [26]. The effect of BR and GA on PHT is manifested primarily through enhanced internode elongation resulting from both increased cell elongation and cell division [1,

27]. Previous studies focused on two aspects to connect BR, GA, and PHT in maize from a

103 genetic perspective: 1) identification of genes with profound effect on PHT leading to dwarf mutants, caused by defects in BR or GA synthesis or signaling pathway genes [1, 6, 10, 11, 26];

2) identification of PHT QTL, co-localized with BR and GA candidate genes [1, 4, 28, 29]. In this study, we found seven co-localized GA candidate genes with PHT associated SNP markers using a 1Mb bin size, and this observed frequency of co-localization is significantly different from random co-segregation. In addition, two BR candidate genes were found to be 1.13 and

1.25 Mb away from two PHT associated SNPs. With a diverse panel of 7000 accessions, Yang et al [30] successfully discovered co-localizations with 1Mb bin size. Considering the limited number of backcross generations for generating PIFs, 1Mb is a reasonable cutoff to serve as a linkage block in this study. In maize, there are around 32,000 genes predicted, with a 2700 Mb genome size of maize [31]. On average, there are around 32000/2700=12 genes within a 1Mb region. Our assumption is, that BR and GA candidate genes are promising candidates for PHT control. In sorghum, two recent studies have used the Sequenom (SQNM) MassARRAY iPLEX platform [32] to develop molecular markers from GA and BR candidate genes [33, 34] for association studies, this method is promising to be applied in maize populations with low linkage disequilibrium but do not add more information for populations with extended linkage blocks such as in PIFs.

Previous studies used Pcz and Ucz as BR and GA inhibitors to phenocopy the effect of BR and GA biosynthetic gene mutations [12]. With Pcz and Ucz treatment, mesocotyl length of maize seedlings is significantly reduced, and genotypes with higher BR and GA biosynthesis or signaling show increased tolerance to corresponding inhibitors. In this study, we applied Pcz and

104

Ucz to compare the BR and GA inhibitor responses among PIFs and BGEM lines. We found that the pattern of hormone inhibitor responses was very similar to the pattern of PHT performance

(Figs 1 and 3), as PIFs are on average more tolerant to both BR and GA inhibitors compared to their respective recurrent parents and BGEM lines. In addition, taller PIFs are associated with stronger BR and GA inhibitor response within both PIFB47 and PIFZ51 (Table 3). Even within the same PIF, there were significant differences for hormone inhibitor response between offspring from the tallest and the shortest plants. In this study, the only selection applied was for

PHT, not for hormone inhibitor response. This indicates that 1) “stronger” hormone pathway genes were directly selected for by selecting for increased PHT, or 2) there was a common pathway contributing to both taller PHT and increased hormone inhibitor response for PIFs.

Crosstalk between BR and GA

We found significant correlations between BR and GA inhibitor responses for both the PIFs and BGEM lines (Table 3). PIFs and BGEM lines that are more tolerant to one of these two hormone inhibitors also tend to show tolerance to the other hormone inhibitor. Pathway analysis suggested that there was a direct crosstalk between the BR and GA pathways, supported by recent studies in Arabidopsis and rice [35-40]. In addition, we found the same SNP, co-localized with a BR candidate gene ZmBSU1, to be associated with both BR and GA. This indicates that

BSU1 may have a function in the interaction between BR and GA. A previous study showed that

BZR1, a transcription factor activated by BR signaling and a DELLA protein, which inhibits the

GA signaling pathway, are the key genes mediating crosstalk between BR and GA signaling

105 pathways [17, 38]. Since BSU1 is activating BZR1 activity [41], it explains detection of BSU1 as a modulator.

Prediction of heterosis in PHT by early monitoring of BR and GA levels

Heterosis and inbreeding depression are considered two sides of the same coin [42] and both of them are defined and quantified in relation to a reference population. Any effect of inbreeding in a population will increase heterosis by the same amount [43]. Heterosis is clearly related to heterozygosity, but it has long been debated how heterozygosity results in heterosis [44].

Inbreeding depression is caused by increased homozygosity of individuals [45], as increased levels of homozygosity accumulate detrimental recessive mutations and reduce heterozygote advantages. Positive correlations between trait expression and level of heterozygosity are recognized as suggestive evidence for heterosis and inbreeding depression, and inbreeding coefficients estimated using homozygous SNPs was found to correlate well with pedigree inbreeding coefficient to infer inbreeding depression [46]. In this study, we found significant positive correlations between PHT and the level of marker heterozygosity (r~0.7). Path analysis indicates that the level of heterozygosity is directly and positively correlated with GA levels

(r~0.5), and seedling stage BR/GA inhibitor responses were positively correlated with field PHT for PIFs (r~0.5), but not BGEM lines (r~0.1). We were not able to directly correlate BR/GA with heterosis in PHT, as we only have recurrent parent and backcross progeny information (exotic donor parents are segregating accessions without available seed source and characterization).

However, we were able to calculate the level of homozygosity, and correlate the level of homozygosity with PHT. Increased inbreeding significantly reduced PHT, with correlation

106

(between the level of homozygosity and PHT) around 0.7. As the level of homozygosity is closely correlated with inbreeding depression [46], and inbreeding depression is the flip side of heterosis, we conclude that BR and GA promote heterosis for PHT (α=0.05) in this study.

Previous studies have shown that GA levels are correlated with seedling heterosis: 1) hybrids were associated with higher GA levels than parental inbred lines in maize [47, 48], sorghum [49], poplar [50] and rice [51], and when heterosis was not displayed due to unfavorable environmental conditions, the hybrid contained equal levels of endogenous GA-like substances [52] as the inbred parents; 2) maize inbred lines were more responsive to external application of GAs compared to hybrid progeny, suggesting a deficiency of endogenous GAs in inbred lines [48]; 3) GA biosynthesis and positive signaling components were up-regulated in hybrids, whereas genes deactivating bioactive GAs, and negative GA signaling components were down-regulated, which together increase seedling heterosis [51]. In addition to seedling heterosis,

GA content was also found to be correlated with heterosis in PHT: it was reported that increased elongation of the uppermost internode contributed most to heterosis for PHT in wheat hybrids

[27]. Examination of GA levels and activities in the uppermost internode tissue of wheat hybrids revealed, that 1) genes promoting GA biosynthesis were up-regulated and GA deactivating genes were down-regulated, which resulted in higher level of GA in hybrids; 2) upregulation of GA receptors GID1 (for GA INSENSITIVE DWARF1) and positive regulator GAMYB, and down- regulation of negative component GAI resulted in enhanced GA sensitivity; 3) GA promoted the expression of expansion genes such as gibberellins induced proteins (GIPs) and endoxyloglucan transferase (XET), which promoted cell division and cell elongation, finally contributed to increased internode elongation and heterosis in PHT. After four days of germinating rice seed,

107

the content of GA4 started to be significantly higher in hybrids compared to their inbred parents

[51]. At both seedling and adult stages, GA levels were increased in hybrids. This explains why the seedling stage GA level was correlated with heterosis in PHT for PIFs. There are only few correlation studies between BR and heterosis available, although BRs function in both cell division and elongation in model species rice and Arabidopsis [35, 53-55]. Despite BRs’ broad effects on the physiological and developmental processes of plants, they were not widely recognized as plant hormones until the mid-1990s [56, 57]. We observed that the BR level was increased for PIFs with higher levels of heterozygosity, and path analysis indicated that this was due to an indirect effect from the crosstalk with the GA pathway. Nonetheless, BR was shown to directly increase PHT. BR and GA coordinately promote plant growth and development by jointly regulating the expression of specific groups of genes [58]. Heterosis for PHT is also affected by other factors. For example, we did not measure the auxin response due to the complexity of the auxin biosynthesis pathway [59], but polar auxin transport has been shown to affect PHT [26].

Both DNA markers, transcripts, and metabolites have been evaluated for prediction of heterosis with various approaches [60-62]. Molecular markers associated with genomic regions contributing to heterosis can be identified with linkage or association mapping methods, and used in a linear regression approach to predict heterosis [63-66]. The advantage of transcriptome based prediction of heterosis over DNA markers is that transcript abundancies are resulted from the integration of the whole genome information, including DNA methylation level and histone modification status, thus the prediction accuracy is higher [67]. For metabolism based prediction of heterosis, metabolic markers were used for the prediction of biomass heterosis in Arabidopsis

108

[61], as metabolite levels are the result of more genes than those represented by genetic markers.

Here, we used plant hormone inhibitors to assess the BR and GA level in different maize genotypes, and this information is significantly correlated with heterosis in PHT. With both seedling stage BR and GA inhibitor responses incorporated into a linear model, the prediction accuracies (r) are 0.62 in PIFB47 and 0.63 in PIFZ51. Our results indicate the possibility to use seedling stage hormone inhibitor response to predict heterosis for PHT in maize breeding projects, especially for biomass maize production. However, it needs to be noted that our prediction is based on the heterozygotes (PIFs) per se, instead of using their parental information.

In other words, previous studies used molecular data from inbred parents to predict heterosis, whereas our prediction was from the same genotypes. To further investigate the usage of hormone based prediction, we measured 201 hybrids for PHT across four replications, which were derived from BGEM lines (DHB47) × PHZ51, BGEM lines (DHZ51) × PHB47, and

PHB47 × PHZ51 (Sanchez et al., 2016 in preparation). We calculated the Mid-parent heterosis as we measured PHT for F1, BGEM lines, PHB47 and PHZ51. We used parental BR and GA inhibitor response to predict Mid-parent heterosis, with data available in S6 Table. Neither BR nor GA inhibitor responses of the inbred parents were associated with heterosis in PHT, with correlations less than 0.1. Thus, our hormone based prediction method can only be used to predict adult performance based on seedling information of the same genotype, instead of using inbred parental information to predict heterosis in hybrids.

Conclusions

In this study, we used phenotypic selection to improve PHT and broadened the genetic variation of two maize heterotic groups (Stiff Stalk and Non Stiff Stalk) by adapting multiple

109 tropical and subtropical accessions into these two genetic pools. We found that phenotypic selection of PHT was genotype dependent, and stronger selection was observed when crosses were made between Stiff Stalk and tropical germplasm. For heterozygotes, GA activities were elevated with an increased level of heterozygosity. Increased GA promotes BR and they together lead to increased heterosis in maize PHT. If this can be generalized for other populations of hybrids (in addition to the crosses between temperate lines and tropical accessions), early stage monitoring of plant hormones is promising for predicting PHT for maize heterozygotes without growing all seed in the field.

Methods

Plant materials

Two libraries of phenotypic-selected introgression families (PIFs): PIFB47 (PHB47 as recurrent parent) and PIFZ51 (PHZ51 as recurrent parent) were used in this study. PHB47 and

PHZ51 are two elite expired PVP (Plant Variety Protection) inbred lines which belong to Iowa

Stiff Stalk and Non Stiff Stalk heterotic groups, respectively. Donor parents were tropical or subtropical accessions from the Germplasm Enhancement of Maize (GEM) project [68], of which 42 different accessions were used for PIFB47 and 46 for PIFZ51, with 5 accessions being used for both PIFB47 and PIFZ51 (S7 Table). The process of constructing PIF libraries was described in a previous study [3]. Briefly, in each backcross (BC) generation, phenotypic selection for PHT

(selection for tallness) and flowering time (synchrony with recurrent parent) was carried out to accumulate PHT increasing alleles from donor parents into elite maize background (PHB47 and

PHZ51) and to minimize confounding effects between flowering time and PHT. As a result, 75

110 and 71 PIFs were produced for PIFB47 and PIFZ51, respectively (S7 Table). In addition, two doubled haploid (DH) libraries (DHB47 and DHZ51) were used as unselected groups for comparison, including 103 and 66 BGEM (DH lines from GEM project) lines [69], respectively.

The method used to create the BGEM lines was described in a previous study [70]. Briefly, donor parents from GEM project were used for BGEM lines as for PIFs, and same recurrent parents (PHB47 and PHZ51) were used for backcrossing. Different from PIF development,

BGEM lines were produced (induced and doubled) from BC1 individuals, and there was no phenotypic selection during the BC process. The hybrid of PHB47×PHZ51 was included in field experiments for comparison.

Field experiments

PIFs together with their recurrent parents were evaluated for PHT across three years (2013,

2014, 2015), with one location in 2013 (Plant Introduction Station: PSI, Ames, IA), three locations in 2014 (Agronomy farm (AG), Ames, IA, PSI, and Neely-Kinyon Memorial Research and Demonstration Farm (NK) – Greenfield, IA) and one location in 2015 (AG) with a randomized complete block design (RCBD) within each location (two blocks per location). All locations were under standard farm management practices (tillage, nutrient management, etc.), with 15 plants per plot and 0.75m between plots. PIFs were compared with their recurrent parents and hybrid of PHB47×PHZ51 for PHT and flowering time. Two weeks after tasseling, a representative plant with median PHT (MPHT) was selected per plot, and PHT was recorded from ground to tassel tip. In 2014, the shortest (SPHT) and the tallest individual (TPHT) within each PIF were also measured. In 2014 and 2015, flowering time were recorded for each plot, when at least 50% of the plants shed pollen or showed silks.

111

DHB47, DHZ51 together with their recurrent parent were characterized for PHT and flowering time in 2014 and 2015 with a RCBD design with two replications at AG. In each plot, one representative individual grown in the middle of the plot was used to measure PHT, and flowering time was measured in the same way as for PIFs.

In 2014 at PSI, leaf samples of the two tallest plants from each PIF together with six PHB47 and six PHZ51 individuals (304 individuals in total) were collected after tasseling for genotyping.

PHT, ear height (EHT), node number (NNode), leaf angle of the first leaf below the flag leaf (LA) and tassel length (TL) were measured for each genotyped individual. EHT was measured from the ground to the primary ear node. NNode was scored as the number of nodes between the top brace root node and the flag leaf excluding the variation in brace root nodes and subterranean nodes. LA was measured as the angle between horizontal position and the midrib of the first leaf below the flag leaf. TL was measured from the non-branching node present below the lowermost primary branch to the tip of central spike. In 2015, the tallest and shortest individual from each

PIF was selected to produce backcross seed for hormone inhibitor response comparison.

Hormone inhibitor test

All PIFs and BGEM lines were evaluated for responses to BR and GA inhibitors in three independent experiments (BR, GA, and water). All entries were treated with 80µM BR inhibitor

Pcz, 80µM GA inhibitor Ucz and water (mock treatment), with three replications per experiment in a completely randomized design (CRD). Each treatment was applied to 12 kernels from each

PIF, and 6 kernels from each BGEM line for each replication within each experiment. Seed was soaked for 24 hours, then transferred to paper rolls [71] containing the same soaking solution

112

(i.e., water, Pcz, or Ucz). Rolls were placed in buckets with a cover in a growth chamber without light at 25C. After eight days, seedlings were removed from the growth chamber and mesocotyl length was measured from the root-shoot transition zone to the first node for each seedling [12] as an indicator for hormone inhibitor response. The average mesocotyl length for the 12 (or 6) seedlings under Pcz, Ucz, and water treatment was recorded as MP (mesocotyl length with Pcz treatment), MU (mesocotyl length with Ucz treatment), and MW (mesocotyl length with water treatment), respectively. BR inhibitor response was defined as MP/MW, and GA inhibitor response as MU/MW [12, 72], as shown in S4 Fig. Seed harvested from the tallest and the shortest individual within each PIF in 2015 was tested using the same protocol described above in three independent experiments.

DNA extraction and genotyping

DNA was extracted using Qiagen BioSprint96 at the Genomic Technology Facility at Iowa

State University. Samples were genotyped with the MaizeSNP3K chip (3072 SNPs across the whole genome), which is a subset of the Illumina MaizeSNP50 BeadChip [73]. Genotyping was executed using the Illumina GoldenGate SNP genotyping platform [74] at the National Maize

Improvement Center of China (China Agricultural University). DNA from all BGEM lines was extracted using 1-10 kernels per line, and genotyped with 8523 (chip-based) SNPs across the genome, in which 7739 polymorphic markers were used in the analysis, with 920 markers overlapping with the MaizeSNP3K subset across the genome. Both DNA extraction and SNP chip genotyping of the BGEM lines were done at KWS SAAT SE (company in Germany).

113

Marker quality control

The twelve genotyped recurrent parent individuals (six PHB47 and six PHZ51) were analyzed for marker quality control: SNP markers missing across all RP individuals or showing heterozygous state across at least three (out of six) recurrent parent individuals were excluded, as residual heterozygosity is random and unlikely to frequently occur at the same genomic position within a fixed inbred line. Subsequently, all PIF genotypes were jointly analyzed. SNP markers with >5% missing data were discarded. For missing SNP markers, the LD k- nearest neighbor algorithm (LD KNNi imputation) was used for imputation using TASSEL5.2.15

[75].

Phenotype and genotype statistical analysis

Phenotypic data was analyzed using SAS 9.3 software (SAS Institute). For each evaluated trait, variance component estimates were obtained from a mixed linear model fitted across all environments in SAS PROC MIXDED. Variance components ( ) were estimated where corresponds to genotypic variance, genotype by environmental interaction variance, and error variance respectively. Entry mean-based heritability was calculated from variance components estimates as where r is the number of replications within each location, and n is the number of locations [76]. Comparisons between PIFs and recurrent parents or between DH lines and recurrent parents were done using Fisher’s Least

Significant Difference (LSD) with a significance level of 0.05. Pearson’s correlation coefficients were used to assess the relationship between different traits with the SAS PROC CORR function.

Linear regression analysis was performed using the SAS PROC GLM function. Graphs were obtained using ggplot2 in R [77].

114

Genome-wide association study

We conducted GWAS for all measured field traits and BR/GA inhibitor responses of the PIFs

(292 genotypes) with 2930 polymorphic genome wide SNPs (minor allele frequency>5%).

Population structure was estimated using Structure 2.3.4 software [78]. The parameter settings included a burn-in length of 50,000 followed by 50,000 iterations of setting K (clusters - number of subpopulations) from 1-10, with 5 replications for each K [79]. The most probable K value was picked by plotting each K (x-axis) with its estimated Ln probability of data (y-axis). The K value was selected when the estimated Ln probability of data reached a plateau [78].

To balance false positives and false negatives, we used three models for GWAS analysis: 1) a Generalized Linear Model (GLM) + Q, with the covariates Q from STRUCTURE included in the model as fixed effects to account for population structure; 2) a mixed linear model (MLM)

[80] with population structure and kinship as covariates; and 3) FarmCPU (Fixed and random model Circulating Probability Unification) with kinship and population structure as covariates, but with additional algorithms solving the confounding problems between testing markers and covariates [81]. GLM was conducted with the software program TASSEL 5.0 [82]. MLM was used as implemented in GAPIT (Genome Association and Prediction Integrated Tool-R package)

[83]. FarmCPU was applied with R package FarmCPU [81]. Statistical program simpleM implemented in R was used to account for multiple testing [84]. The threshold level was based on an effective number of independent tests m, and m was used in a similar way as the

Bonferroni correction method [79]. In this study, for a family-wise error rate of 0.05 the threshold for significant trait-marker associations was set as 4.44×10-5 (multiple testing threshold level).

115

Comparison with published PHT QTL regions

SNP markers associated with PHT were compared with previously published PHT quantitative trait loci (QTL). The dataset for comparison is based on a maize PHT QTL meta- analysis aimed at identifying the most likely position of PHT QTL that are consistently found across studies. This dataset summarized published PHT QTL into 40 hot spots across the maize genome based on the maize B73 RefGen_v2 physical map [15]. PHT associated SNP markers from this study were compared with these hot spots to study co-localizations.

BR and GA candidate genes

The information about the genes involved in BR and GA biosynthesis and signaling pathway was collected from Arabidopsis and rice (model species). The protein sequence of these genes was obtained from the National Center for Biotechnology Information (NCBI) databases, and was used to identify orthologous genes in maize through BLASTP in Gramene [85]. Based on

BLASTP score, % Identity, and E-value, hits were ranked. From the most likely hit, a reverse

BLAST search was conducted: maize genes identified using the approach described above were blasted back to model species to identify orthologous genes, the goal was to confirm that the gene with the highest score was the original BR/GA pathway gene from model species. If the gene identified was not the original gene (used to find maize candidate genes), it was discarded.

Finally, each BR/GA gene from the model species was aligned with at least one candidate gene in maize, with a few genes aligned with 2-4 maize candidate genes (which are likely due to gene duplications in maize). Finally, we compared the GA candidate genes found in this study with

GA candidate genes published with another method [21], which used all previously reported genes encoding GA metabolism enzymes in maize and other species as BLAST queries.

116

Co-localization of BR and GA candidate genes and PHT associated SNPs

We used a bin size of 1Mb [30] around each PHT associated SNP to capture BR/GA candidate genes. Moreover, we used a re-sampling based non-parametric method to test the hypothesis, that the observed number of BR or GA candidate genes co-localized with PHT associated SNP markers using a 1Mb bin size is not different from randomly occurrence. First, we defined the number of PHT associated SNP markers capturing BR/GA candidate genes using a 1Mb bin size as “observed value”. Second, the same number of PHT associated SNP markers from GWAS results was randomly sampled (sampling with replacement) from the whole set of

SNP markers. In each random sample, all sampled SNP markers used a bin size of 1Mb to capture BR/GA candidate genes. The number of SNP markers capturing a BR or GA candidate gene was recorded as test statistic. This procedure was repeated for 10,000 times, and the number of test statistics exceeding the observed value was divided by the total number of simulations to compute a P-value [30, 86].

Path analysis between BR, GA, PHT, and heterozygosity

For each PIF, we recorded the BR inhibitor response, GA inhibitor response, PHT performance and the level of heterozygosity (the percentage of heterozygous markers). We used path analysis implemented in IBM SPSS Amos 20 [87, 88] to investigate the direct or indirect correlations between these four variables by fitting proposed path models according to the covariance structure of the underlying data [88]. The correlations between each pair of these four variables are used to calculate total, direct, and indirect effects between the observed variables.

We started with a proposed model with the hypothesis that the level of heterozygosity indirectly

117 impacts PHT through the regulation of BR and GA instead of a direct correlation. In other words, there are no other factors connecting heterozygosity and PHT except for BR and GA. Based on the data structure, path analysis suggests to add new links or to delete old links between variables.

The final best fit was defined by a stringent criterion with chi square P-value>0.05, root mean square error of approximation (RMSEA)<0.05, comparative fit index (CFI)>0.95, goodness of fit index (GFI)>0.8, adjusted goodness of fit index (AGFI)>0.9, normed fit index (NFI)>0.9, relative fit index (RFI) >0.9, incremental fit index (IFI) >0.9. If more than one model fit all criteria, then selection was based on a minimum Akaike’s information criterion (AIC) value [88,

89].

REFERENCES

1. Fernandez MGS, Becraft PW, Yin Y, Lübberstedt T. From dwarves to giants? Plant height manipulation for biomass yield. Trends Plant Sci. 2009;14:454-61.

2. Mendu V, Harman-Ware AE, Crocker M, Jae J, Stork J, Morton S et al. Identification and thermochemical analysis of high-lignin feedstocks for biofuel and biochemical production. Biotechnol Biofuels. 2011;4:1.

3. Abdel‐Ghani AH, Hu S, Chen Y, Brenner EA, Kumar B, Blanco M et al. Genetic architecture of plant height in maize phenotype‐selected introgression families. Plant Breed. 2016;135:429-38.

4. Teng F, Zhai L, Liu R, Bai W, Wang L, Huo D et al. ZmGA3ox2, a candidate gene for a major QTL, qPH3. 1, for plant height in maize. Plant J. 2013;73:405-16.

5. Sasaki A, Ashikari M, Ueguchi-Tanaka M, Itoh H, Nishimura A, Swapan D et al. Green revolution: a mutant gibberellin-synthesis gene in rice. Nature. 2002;416:701-2.

6. Peng J, Richards DE, Hartley NM, Murphy GP, Devos KM, Flintham JE et al. ‘Green revolution’genes encode mutant gibberellin response modulators. Nature. 1999;400:256-61.

7. Winkler RG, Helentjaris T. The maize Dwarf3 gene encodes a cytochrome P450-mediated early step in Gibberellin biosynthesis. Plant Cell. 1995;7:1307-17.

118

8. Lawit SJ, Wych HM, Xu D, Kundu S, Tomes DT. Maize DELLA proteins dwarf plant8 and dwarf plant9 as modulators of plant development. Plant Cell Physiol. 2010;51:1854-68.

9. Thornsberry JM, Goodman MM, Doebley J, Kresovich S, Nielsen D, Buckler ES. Dwarf8 polymorphisms associate with variation in flowering time. Nat Genet. 2001;28:286-9.

10. Hartwig T, Chuck GS, Fujioka S, Klempien A, Weizbauer R, Potluri DPV et al. Brassinosteroid control of sex determination in maize. Proc. Natl. Acad. Sci. U.S.A. 2011;108:19814-9.

11. Makarevitch I, Thompson A, Muehlbauer GJ, Springer NM. Brd1 gene in maize encodes a brassinosteroid C-6 oxidase. PloS one. 2012;7:e30798.

12. Hartwig T, Corvalan C, Best NB, Budka JS, Zhu J-Y, Choe S et al. Propiconazole is a specific and accessible brassinosteroid (BR) biosynthesis inhibitor for Arabidopsis and maize. PLoS One. 2012;7:e36625.

13. Yan W, Yanhong Y, Wenyu Y, Taiwen Y, Weiguo L, Xiaochun W. Responses of root growth and nitrogen transfer metabolism to uniconazole, a growth retardant, during the seedling stage of soybean under relay strip intercropping system. Commun Soil Sci Plant Anal. 2013;44:3267-80.

14. Rademacher W. Growth retardants: effects on gibberellin biosynthesis and other metabolic pathways. Annu Rev Plant Biol. 2000;51:501-31.

15. Wang Y, Yao J, Zhang Z, Zheng Y. The comparative analysis based on maize integrated QTL map and meta-analysis of plant height QTLs. Chin Sci Bull. 2006;51:2219-30.

16. Gao S, Fang J, Xu F, Wang W, Sun X, Chu J et al. CYTOKININ OXIDASE/DEHYDROGENASE4 integrates cytokinin and auxin signaling to control rice crown root formation. Plant Physiol. 2014;165:1035-46.

17. Li J, Wen J, Lease KA, Doke JT, Tax FE, Walker JC. BAK1, an Arabidopsis LRR receptor- like protein kinase, interacts with BRI1 and modulates brassinosteroid signaling. Cell. 2002;110:213-22.

18. Zhu Y, Nomura T, Xu Y, Zhang Y, Peng Y, Mao B et al. ELONGATED UPPERMOST INTERNODE encodes a cytochrome P450 monooxygenase that epoxidizes gibberellins in a novel deactivation reaction in rice. Plant Cell. 2006;18:442-56.

19. Komorisono M, Ueguchi-Tanaka M, Aichi I, Hasegawa Y, Ashikari M, Kitano H et al. Analysis of the rice mutant dwarf and gladius leaf 1. Aberrant katanin-mediated microtubule organization causes up-regulation of gibberellin biosynthetic genes independently of gibberellin signaling. Plant Physiol. 2005;138:1982-93.

20. Sekhon RS, Lin H, Childs KL, Hansey CN, Buell CR, de Leon N et al. Genome‐wide atlas of transcription during maize development. Plant J. 2011;66:553-63.

119

21. Song J, Guo B, Song F, Peng H, Yao Y, Zhang Y et al. Genome-wide identification of gibberellins metabolic enzyme genes and expression profiling analysis during seed germination in maize. Gene. 2011;482:34-42.

22. Liu K, Goodman M, Muse S, Smith JS, Buckler E, Doebley J. Genetic structure and diversity among maize inbred lines as inferred from DNA microsatellites. Genetics. 2003;165:2117-28.

23. Moll RH, Salhuana W, Robinson H. Heterosis and genetic diversity in variety crosses of maize. Crop Sci. 1962;2:197-8.

24. Moll R, Lonnquist J, Fortuno JV, Johnson E. The relationship of heterosis and genetic divergence in maize. Genetics. 1965;52:139.

25. Reif J, Hallauer A, Melchinger A. Heterosis and heterotic patterns in maize. Maydica. 2005;50:215.

26. Wang Y, Li J. Molecular basis of plant architecture. Annu Rev Plant Biol. 2008;59:253-79.

27. Zhang Y, Ni Z, Yao Y, Nie X, Sun Q. Gibberellins and heterosis of plant height in wheat (Triticum aestivum L.). BMC Genet. 2007;8:1.

28. Peiffer JA, Romay MC, Gore MA, Flint-Garcia SA, Zhang Z, Millard MJ et al. The genetic architecture of maize height. Genetics. 2014;196:1337-56.

29. Weng J, Xie C, Hao Z, Wang J, Liu C, Li M et al. Genome-wide association study identifies candidate genes that affect plant height in Chinese elite maize (Zea mays L.) inbred lines. PLoS One. 2011;6:e29229.

30. Yang J, Jiang H, Yeh CT, Yu J, Jeddeloh JA, Nettleton D et al. Extreme‐phenotype genome‐wide association study (XP‐GWAS): a method for identifying trait‐associated variants by sequencing pools of individuals selected from a diversity panel. Plant J. 2015;84:587-96.

31. Schnable PS, Ware D, Fulton RS, Stein JC, Wei F, Pasternak S et al. The B73 maize genome: complexity, diversity, and dynamics. Science. 2009;326:1112-5.

32. Gabriel S, Ziaugra L, Tabbaa D. SNP genotyping using the Sequenom MassARRAY iPLEX platform. Curr Protoc Hum Genet. 2009:2.12. 1-2.. 6.

33. Zhao J, Mantilla Perez MB, Hu J, Salas Fernandez MG. Genome-Wide Association Study for Nine Plant Architecture Traits in Sorghum. Plant Genome. 2016.

34. Perez MBM, Zhao J, Yin Y, Hu J, Fernandez MGS. Association mapping of brassinosteroid candidate genes and plant architecture in a diverse panel of Sorghum bicolor. Theor Appl Genet. 2014;127:2645-62.

35. Tong H, Xiao Y, Liu D, Gao S, Liu L, Yin Y et al. Brassinosteroid regulates cell elongation by modulating gibberellin metabolism in rice. Plant Cell. 2014;26:4376-93.

120

36. Hofmann NR. Taking hormone crosstalk to a new level: brassinosteroids regulate gibberellin biosynthesis. Plant Cell. 2015;27:2081-.

37. Unterholzner SJ, Rozhon W, Papacek M, Ciomas J, Lange T, Kugler KG et al. Brassinosteroids are master regulators of gibberellin biosynthesis in Arabidopsis. Plant Cell. 2015;27:2261-72.

38. Bai M-Y, Shang J-X, Oh E, Fan M, Bai Y, Zentella R et al. Brassinosteroid, gibberellin and phytochrome impinge on a common transcription module in Arabidopsis. Nat Cell Biol. 2012;14:810-7.

39. Gallego-Bartolomé J, Minguet EG, Grau-Enguix F, Abbas M, Locascio A, Thomas SG et al. Molecular mechanism for the interaction between gibberellin and brassinosteroid signaling pathways in Arabidopsis. Proc. Natl. Acad. Sci. U.S.A. 2012;109:13446-51.

40. Li Q-F, Wang C, Jiang L, Li S, Sun SS, He J-X. An interaction between BZR1 and DELLAs mediates direct signaling crosstalk between brassinosteroids and gibberellins in Arabidopsis. Sci Signal. 2012;5:72.

41. Ye H, Li L, Yin Y. Recent Advances in the Regulation of Brassinosteroid Signaling and Biosynthesis Pathways. J Integr Plant Biol. 2011;53:455-68.

42. Mather K, Jinks JL. Biometrical genetics: the study of continuous variation. Springer; 2013.

43. Miranda Filho J, Brasil S, editors. Inbreeding and heterosis. Book of abstracts The Genetics and Exploitation of Heterosis in Crops An International Symposium México; 1997.

44. Li Z-K, Luo L, Mei H, Wang D, Shu Q, Tabien R et al. Overdominant epistatic loci are the primary genetic basis of inbreeding depression and heterosis in rice. I. Biomass and grain yield. Genetics. 2001;158:1737-53.

45. Charlesworth D, Willis JH. The genetics of inbreeding depression. Nat Rev Genet. 2009;10:783-96.

46. Kim E-S, Sonstegard TS, Van Tassell CP, Wiggans G, Rothschild MF. The relationship between runs of homozygosity and inbreeding in Jersey cattle under selection. PloS one. 2015;10:e0129967.

47. Rood S, Buzzell R, Major D, Pharis R. Gibberellins and heterosis in maize: quantitative relationships. Crop Sci. 1990;30:281-6.

48. Rood SB, Buzzell RI, Mander LN, Pearce D, Pharis RP. Gibberellins: a phytohormonal basis for heterosis in maize. Science. 1988;241:1216-8.

49. Rood SB, Witbeck TJ, Major DJ, Miller FR. Gibberellins and heterosis in sorghum. Crop Sci. 1992;32:713-8.

121

50. Bate NJ, Rood SB, Blake TJ. Gibberellins and heterosis in poplar. Can J Bot. 1988;66:1148- 52.

51. Ma Q, Hedden P, Zhang Q. Heterosis in rice seedlings: its relationship to gibberellin content and expression of gibberellin metabolism and signaling genes. Plant Physiol. 2011;156:1905-20.

52. Rood S, Major D, Jones M, Pharis R. Low temperature eliminates heterosis for growth and gibberellin content in maize. Crop Sci. 1985;25:1063-8.

53. Szekeres M, Németh K, Koncz-Kálmán Z, Mathur J, Kauschmann A, Altmann T et al. Brassinosteroids rescue the deficiency of CYP90, a cytochrome P450, controlling cell elongation and de-etiolation in Arabidopsis. Cell. 1996;85:171-82.

54. Hu Y, Bao F, Li J. Promotive effect of brassinosteroids on cell division involves a distinct CycD3‐induction pathway in Arabidopsis. Plant J. 2000;24:693-701.

55. Zhiponova MK, Vanhoutte I, Boudolf V, Betti C, Dhondt S, Coppens F et al. Brassinosteroid production and signaling differentially control cell division and expansion in the leaf. New Phytol. 2013;197:490-502.

56. Clouse SD. Molecular genetic studies confirm the role of brassinosteroids in plant growth and development. Plant J. 1996;10:1-8.

57. Divi UK, Krishna P. Brassinosteroid: a biotechnological target for enhancing crop yield and stress tolerance. N Biotechnol. 2009;26:131-6.

58. Yang G-X, Jan A, Shen S-H, Yazaki J, Ishikawa M, Shimatani Z et al. Microarray analysis of brassinosteroids-and gibberellin-regulated gene expression in rice seedlings. Mol Genet Genomics. 2004;271:468-78.

59. Zhao Y. Auxin biosynthesis and its role in plant development. Annu Rev Plant Biol. 2010;61:49.

60. Maenhout S, De Baets B, Haesaert G. Prediction of maize single-cross hybrid performance: support vector machine regression versus best linear prediction. Theor Appl Genet. 2010;120:415-27.

61. Andorf S, Selbig J, Altmann T, Poos K, Witucka-Wall H, Repsilber D. Enriched partial correlations in genome-wide gene expression profiles of hybrids (A. thaliana): a systems biological approach towards the molecular basis of heterosis. Theor Appl Genet. 2010;120:249- 59.

62. Steinfath M, Gärtner T, Lisec J, Meyer RC, Altmann T, Willmitzer L et al. Prediction of hybrid biomass in Arabidopsis thaliana by selected parental SNP and metabolic markers. Theor Appl Genet. 2010;120:239-47.

122

63. Tang J, Yan J, Ma X, Teng W, Wu W, Dai J et al. Dissection of the genetic basis of heterosis in an elite maize hybrid by QTL mapping in an immortalized F2 population. Theor Appl Genet. 2010;120:333-40.

64. Meyer RC, Kusterer B, Lisec J, Steinfath M, Becher M, Scharr H et al. QTL analysis of early stage heterosis for biomass in Arabidopsis. Theor Appl Genet. 2010;120:227-37.

65. Basunanda P, Radoev M, Ecke W, Friedt W, Becker H, Snowdon R. Comparative mapping of quantitative trait loci involved in heterosis for seedling and yield traits in oilseed rape (Brassica napus L.). Theor Appl Genet. 2010;120:271-81.

66. Vuylsteke M, Kuiper M, Stam P. Chromosomal regions involved in hybrid performance and heterosis: their AFLP®-based identification and practical use in prediction models. Heredity. 2000;85:208-18.

67. Scholten S, Thiemann A. Transcriptome-Based Prediction of Heterosis and Hybrid Performance. Diagnostics in Plant Breeding. Springer; 2013. p. 265-279.

68. Salhuana W, Pollak L. Latin American maize project (LAMP) and germplasm enhancement of maize (GEM) project: Generating useful breeding germplasm. Maydica. 2006;51:339.

69. BGEM Line Information Database. http://www.public.iastate.edu/~usda- gem/Official_Documents/GEM_Documents.htm. Accessed 12 August 2016.

70. Brenner EA, Blanco M, Gardner C, Lübberstedt T. Genotypic and phenotypic characterization of isogenic doubled haploid exotic introgression lines in maize. Mol Breed. 2012;30:1001-16.

71. Kumar B, Abdel‐Ghani AH, Reyes‐Matamoros J, Hochholdinger F, Lübberstedt T. Genotypic variation for root architecture traits in seedlings of maize (Zea mays L.) inbred lines. Plant breed. 2012;131:465-78.

72. Huang J, Zhang J, Li W, Hu W, Duan L, Feng Y et al. Genome‐wide association analysis of ten chilling tolerance indices at the germination and seedling stages in maize. J Integr Plant Biol. 2013;55:735-44.

73. Ganal MW, Durstewitz G, Polley A, Bérard A, Buckler ES, Charcosset A et al. A large maize (Zea mays L.) SNP genotyping array: development and germplasm genotyping, and genetic mapping to compare with the B73 reference genome. PloS one. 2011;6:e28334.

74. Fan JB, Gunderson KL, Bibikova M, Yeakley JM, Chen J, Garcia EW et al. [3] illumina universal bead arrays. Methods Enzymol. 2006;410:57-73.

75. Money D, Gardner K, Migicovsky Z, Schwaninger H, Zhong G-Y, Myles S. LinkImpute: Fast and accurate genotype imputation for nonmodel organisms. G3: Genes| Genomes| Genetics. 2015;5:2383-90.

123

76. Holland JB, Nyquist WE, Cervantes-Martínez CT. Estimating and interpreting heritability for plant breeding: an update. Plant Breed Rev. 2003;22:9-112.

77. Wickham H. ggplot2: elegant graphics for data analysis. Springer Science & Business Media; 2009.

78. Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics. 2000;155:945-59.

79. Pace J, Lee N, Naik HS, Ganapathysubramanian B, Lübberstedt T. Analysis of maize (Zea mays L.) seedling roots with the high-throughput image analysis tool ARIA (Automatic Root Image Analysis). PLoS One. 2014;9:e108255.

80. Yu J, Pressoir G, Briggs WH, Bi IV, Yamasaki M, Doebley JF et al. A unified mixed-model method for association mapping that accounts for multiple levels of relatedness. Nat Genet. 2006;38:203-8.

81. Liu X, Huang M, Fan B, Buckler ES, Zhang Z. Iterative usage of fixed and random effect models for powerful and efficient genome-wide association studies. PLoS Genet. 2016;12:e1005767.

82. 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.

83. Lipka AE, Tian F, Wang Q, Peiffer J, Li M, Bradbury PJ et al. GAPIT: genome association and prediction integrated tool. Bioinformatics. 2012;28:2397-9.

84. Johnson RC, Nelson GW, Troyer JL, Lautenberger JA, Kessing BD, Winkler CA et al. Accounting for multiple comparisons in a genome-wide association study (GWAS). BMC Genomics. 2010;11:1.

85. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997;25:3389-402.

86. Johnson C, Drgon T, Walther D, Uhl GR. Genomic regions identified by overlapping clusters PloS one. 2011;6:e19210.

87. Arbuckle JL. Amos AJ. 7.0 User's Guide. Chicago, IL: SPSS. Inc; 2006.

88. Jansen C, de Leon N, Lauter N, Hirsch C, Ruff L, Lübberstedt T. Genetic and morphometric analysis of cob architecture and biomass-related traits in the intermated B73× Mo17 recombinant inbred lines of maize. BioEnergy Research. 2013;6:903-16.

89. Thompson B. The Ten Commandments of Good Structural Equation Modeling Behavior: A User-Friendly, Introductory Primer on SEM. 1998.

124

Fig 1. Box plots of PHT performance of the BGEM lines and PIFs. The box in the middle between the first (Q1) and third quartile (Q3) represents 50% values. The lines within in each box indicate the median value for each group. The three bold horizontal lines represent the F1 hybrid of PHB47 × PHZ51, PHB47 and PHZ51, respectively.

125

Fig 2. Hormone inhibitor responses of PHB47, PHZ51 and their F1 hybrid of PHB47 ×

PHZ51. From left to right, PHB47, PHZ51, F1 hybrid of PHB47 × PHZ51 were treated with gibberellin inhibitor Ucz (green), brassinosteroid inhibitor Pcz (orange), and Water (blue), respectively.

126

Fig 3. Hormone inhibitor responses of BGEM lines and PIFs. X-axis: Groups from left to right represent different combinations of library and hormone inhibitor treatments. For example,

DHB47-BR represents the DHB47 library treated with BR inhibitor. Y-axis reflects the hormone inhibitor response which is calculated as (mesocotyl length with hormone inhibitor treatment)/(mesocotyl length with water treatment). The box in the middle between the first (Q1) and third quartile (Q3) represents 50% values. The lines within in each box indicate the median value for each group. Solid and dashed lines indicate BR and GA treatment, respectively. The six horizontal lines from top to bottom indicate F1 (hybrid of PHB47 × PHZ51) BR inhibitor

127

Fig 3. continued

response, F1 GA inhibitor response, PHZ51 BR inhibitor response, PHZ51 GA inhibitor response,

PHB47 BR inhibitor response, PHB47 GA inhibitor response, respectively.

128

Fig 4. Hormone inhibitor response of seed from tallest and shortest individuals of PIF113.

Green, orange and blue represents for gibberellin inhibitor Ucz, brassinosteroid inhibitor Pcz, and water treatment, respectively.

129

Fig 5. The final best model from path analysis for PIFB47 and PIFZ51. Heterozygosity represents for the proportion of heterozygous markers for PIFs. BR and GA represents for brassinosteroid and gibberellin inhibitor response, respectively. PHT represents for plant height.

The numbers represent for P-values of direct relationships between each two variables (without confounding effects).

130

Fig 6. Manhattan plot and QQ-plot of the FarmCPU results for plant height (PHT) and gibberellin inhibitor response (GA). In the left panel, X-axis represents for the ten chromosomes in maize and Y-axis represents for negative log10-transformed P-values. In the right panel, X and Y axis represents for the expected and observed negative log10-transformed P- values for FarmCPU model, respectively.

131

Table 1. Phenotypic correlations between different agronomic traits in PIFB47 and PIFZ51.

PIFZ51 PHT EHT TL NNode LA PHT - 0.70** 0.36** 0.31** -0.05 EHT - 0.26* 0.28** 0.19 TL - -0.01 0.04 NNode - -0.17 LA - PIFB47 PHT EHT TL NNode LA PHT - 0.73** 0.38** 0.31** 0.1 EHT - 0.28** 0.25* 0.15 TL - 0.05 0.09 NNode - 0.04 LA -

‘*’ and ‘**’ show the significances at α=0.01 and α=0.001, respectively. PHT, EHT, TL, NNode and LA represents for plant height, ear height, tassle length, number of nodes, and leaf angle, respectively.

132

Table 2. Genetic characterization of PIFB47 and PIFZ51.

Generation Expected DGP Observed DGP Observed DGP (%) (%) of PIFB47 (%) of PIFZ51

BC3 6.25 17.3 4.1

BC4 3.13 16.7 2.7 BC5 1.56 17.1 4.9 BC6 0.78 17.2 6.2 DGP represents for donor genome proportion.

133

Table 3. Correlations between BR/GA inhibitor response, PHT, and heterozygosity for PIFs and BGEM lines.

PIFB47 PHT BR GA Heterozygosity PHT - 0.46*** 0.59*** 0.72*** BR - 0.52*** 0.34** GA - 0.53*** Heterozygosity - PIFZ51 PHT BR GA Heterozygosity PHT - 0.51*** 0.58*** 0.70*** BR - 0.54*** 0.32** GA - 0.41** Heterozygosity - DHB47 PHT BR GA - PHT - -0.12 -0.07 - BR - 0.68*** - GA - - DHZ51 PHT BR GA - PHT - -0.08 -0.04 - BR - 0.49*** - GA - - ‘*’, ‘**’, ‘***’ show the significances at α=0.01, α=0.001, α=0.0001, respectively; BR and GA represents for brassinosteroid and gibberellin inhibitor response, respectively; PHT represents for plant height.

134

Table 4. Co-localization of gibberellin (GA) candidate genes and PHT loci.

Distance (Mb) to Candidate Candidate gene Marker Pathway candidate gene ID gene

PZE- GA 0.47 ZmCPS3 GRMZM2G044481 101087901 biosynthesis

PZE- ZmDELLAs- GA 0.39 GRMZM2G001426 103056753 1 signaling

PZE- GA 0.43 ZmGA2ox7 GRMZM2G078798 106070979 biosynthesis

GA PZA02388.1 0.77 ZmPIF3-2 GRMZM2G387528 signaling

PZE- GA 0.75 ZmCUL1-1 GRMZM2G166089 108073190 signaling

GA SYN13209 0.87 ZmGA20ox4 GRMZM2G049418 biosynthesis

PZE- GA 0.36 ZmKAO1 GRMZM2G089803 109092637 biosynthesis

135

Figure S1: PHT (plant height) performance of PIB47 and PIFZ51 across the three backcross generations.

A. MPHT (median plant height) performance of PIFB47 across three years (three generations), the red line at the very bottom represents for recurrent parent (PHB47), and the very top line represents for the hybrid of PHB47×PHZ51.

136

Figure S1 continued

B. MPHT (median plant height) performance of PIFZ51 across three years (three generations), the red line at the very bottom represents for recurrent parent (PHZ51), and the very top line represents for the hybrid ofPHB47×PHZ51.

137

Figure S2A: A representative of brassinosteroid (BR) inhibitor tolerant PIF (PIF58) compared with its recurrent parent (RP).

Figure S2B: A representative of gibberellin inhibitor tolerant PIF (PIF59) compared with its recurrent parent (RP).

Blue, orange and green represents for water, BR inhibitor and GA inhibitor treatment, respectively.

138

Figure S3: Population structure estimates for PIFs. The three different colors illustrate the proportion of each subpopulation.

139

Figure S4: Brassinosteroid (BR) and gibberellin (GA) inhibitor response calculation. Pcz and Ucz represents for BR and GA inhibitor, respectively. MU, MP and MW represents for mesocotyl length with Ucz, Pcz and water treatment. BR inhibitor response was calculated as

MP/MW, and GA inhibitor response was calculated as MU/MW.

140

Table S1: Summary statistics for other agronomic traits (besides PHT) in PIFB47 and PIFZ51.

Recurrent Library Trait Average STDEV parent PIFB47 EHT(cm) 113.4 12.35 86 PIFB47 TL(cm) 35.42 5.25 28 PIFB47 NNode 13.83 0.61 13 PIFB47 LA(°) 50 5.96 50 PIFZ51 EHT(cm) 90.73 11.77 85 PIFZ51 TL(cm) 37.41 3.9 34 PIFZ51 NNode 14.34 0.87 13 PIFZ51 LA(°) 40.81 9.07 47

PHT, EHT, TL, NNode and LA represents for plant height, ear height, tassle length, number of nodes, leaf angle respectively; STDEV stands for standard deviation.

141

Table S2: Comparison between the seed from the tallest/shortest individual within one PIF for hormone inhibitor response.

P-value of the P-value of the Donor Shortest Tallest difference for difference for genome individual individual BR inhibitor GA inhibitor proportion response response (DGP %) PIF17S PIF17T 0.9466 0.0313 19.34 PIF34S PIF34T 0.0173 0.6992 20.09 PIF58S PIF58T 0.9581 0.0047 20.87 PIF59S PIF59T <.0001 0.0654 20.5 PIF90S PIF90T 0.0009 0.4546 7.43 PIF99S PIF99T 0.0205 <.0001 5.53 PIF113S PIF113T 0.9975 0.0003 11.88 PIF134S PIF134T 0.0449 0.1328 10.86

BR and GA represents for brassinosteroid and gibberellin, respectively.

142

Table S3: Parameters of the best model from path analysis for PIFB47 and PIFZ51.

Model fit PIFB47 PIFZ51 P-value for Chi square 0.479 0.272 test RMSEA 0.0001 0.05 CFI 1 1 GFI 0.996 0.991 AFGI 0.964 0.913 NFI 0.995 0.988 RFI 0.972 0.912 IFI 1.005 0.998 AIC 18.502 19.207 RMSEA, CFI, GFI, AFGI, NFI, RFI, IFI, and AIC represents for root mean square approximation, comparative fit index, goodness of fit index, adjusted goodness of fit index, normed fit index, relative fit index, incremental fit index, Akaike's information criterion, respectively.

143

Table S4: Markers associated with PHT, EHT, NNode, and LA from association study.

Method Trait SNP Chromosome Position P.value 1.39E- FarmCPU PHT SYN38535 5 28201445 15 PZE- 5.45E- FarmCPU GA 8 5678995 108005623 08 9.08E- GLM+Q PHT SYN20493 3 186435833 09 PZE- 1.11E- GLM+Q PHT 9 74588641 109044053 08 PZE- 2.27E- GLM+Q PHT 9 74588367 109044064 08 5.35E- GLM+Q PHT SYN32654 8 160366514 08 PZE- 5.69E- GLM+Q PHT 8 130213045 108074750 08 PZE- 7.41E- GLM+Q PHT 7 34172216 107028505 08 PZE- 1.15E- GLM+Q PHT 9 152920575 109116158 07 PZE- 2.39E- GLM+Q PHT 4 215385635 104132368 07 PZE- 2.60E- GLM+Q PHT 8 139343576 108082592 07 3.67E- GLM+Q PHT SYN23237 3 184601452 07 ZM013369- 6.29E- GLM+Q PHT 9 74171446 0605 07 PZE- 9.88E- GLM+Q PHT 8 79934432 108047536 07 1.07E- GLM+Q PHT SYN31220 3 180977949 06 1.35E- GLM+Q PHT SYN29112 4 231412099 06 PZE- 1.80E- GLM+Q EHT 6 155654988 106105143 06 1.84E- GLM+Q PHT SYN19565 2 30589468 06 PZE- 2.15E- GLM+Q PHT 6 145180751 106087627 06 PZE- 2.73E- GLM+Q PHT 10 112204539 110058308 06 PZE- 3.07E- GLM+Q GA 8 5678995 108005623 06

144

Table S4 continued PZE- 3.54E- GLM+Q PHT 6 114178476 106062991 06 PZE- 3.82E- GLM+Q PHT 7 35162894 107029002 06 PZE- 4.74E- GLM+Q PHT 8 158762013 108103023 06 PZE- 4.83E- GLM+Q PHT 9 86397696 109049656 06 PZE- 5.25E- GLM+Q PHT 8 79424912 108047365 06 PZE- 5.82E- GLM+Q PHT 6 114185516 106063009 06 PZE- 5.83E- GLM+Q PHT 3 122077074 103073770 06 PZE- 6.34E- GLM+Q PHT 6 125452957 106070979 06 PZE- 6.46E- GLM+Q PHT 8 156213656 108100090 06 PZE- 6.56E- GLM+Q EHT 1 205219399 101161892 06 6.95E- GLM+Q PHT SYN19755 8 138390481 06 PZE- 7.69E- GLM+Q PHT 3 125427407 103075614 06 PZE- 8.06E- GLM+Q PHT 2 179538585 102129469 06 PZE- 8.10E- GLM+Q PHT 8 127625290 108073190 06 9.80E- GLM+Q PHT PZA02388.1 8 169137 06 PZE- 9.91E- GLM+Q PHT 9 139661789 109092637 06 PZE- 1.04E- GLM+Q PHT 6 114181162 106063007 05 PZE- 1.06E- GLM+Q PHT 8 68313123 108041894 05 PZE- 1.07E- GLM+Q PHT 3 110461222 103060293 05 PZE- 1.15E- GLM+Q PHT 4 180053538 104103747 05 PZE- 1.16E- GLM+Q PHT 2 59631929 102077234 05 PZE- 1.45E- GLM+Q LA 7 136261616 107081317 05

145

Table S4 continued PZE- 1.46E- GLM+Q PHT 7 149854371 107094423 05 1.56E- GLM+Q GA PZA00058.6 8 6023577 05 1.59E- GLM+Q PHT SYN26803 9 28506261 05 1.66E- GLM+Q GA SYN19928 8 128133446 05 PZE- 1.70E- GLM+Q PHT 3 69558888 103056753 05 PZE- 1.75E- GLM+Q EHT 4 215385635 104132368 05 1.78E- GLM+Q PHT PZA02479.1 4 212747880 05 PZE- 1.80E- GLM+Q PHT 8 103638340 108058109 05 PZE- 1.84E- GLM+Q PHT 9 94798138 109054725 05 PZE- 1.96E- GLM+Q GA 8 123786918 108070783 05 1.96E- GLM+Q PHT SYN10393 1 259824332 05 1.97E- GLM+Q PHT SYN13209 8 164063270 05 1.97E- GLM+Q PHT SYN35767 1 69881711 05 1.98E- GLM+Q NNode SYN10369 2 5379183 05 PZE- 2.02E- GLM+Q PHT 3 205260873 103151399 05 PZE- 2.15E- GLM+Q PHT 8 85863945 108049639 05 PZE- 2.23E- GLM+Q PHT 4 196930705 104119909 05 2.42E- GLM+Q EHT SYN4646 6 147717038 05 PZE- 2.49E- GLM+Q PHT 7 171926341 107130789 05 2.53E- GLM+Q PHT SYN35471 4 199278965 05 PZE- 2.56E- GLM+Q PHT 1 207712012 101164570 05 PZE- 2.74E- GLM+Q PHT 9 13755379 109013469 05

146

Table S4 continued PZE- 2.74E- GLM+Q PHT 10 1018195 110000136 05 2.75E- GLM+Q PHT SYN11836 8 131337588 05 PZE- 2.79E- GLM+Q PHT 1 26496101 101039301 05 PZE- 2.81E- GLM+Q PHT 9 154964992 109121160 05 2.81E- GLM+Q PHT SYN35141 6 115294816 05 2.81E- GLM+Q PHT SYN38535 5 28201445 05 PZE- 2.83E- GLM+Q PHT 2 49913767 102071022 05 PZE- 2.88E- GLM+Q PHT 8 87037136 108050016 05 PZE- 2.89E- GLM+Q PHT 9 110930390 109066773 05 PZE- 2.90E- GLM+Q PHT 1 79236710 101087901 05 PZE- 2.94E- GLM+Q PHT 8 80503266 108048010 05 PZE- 3.13E- GLM+Q PHT 4 181429757 104105165 05 PZE- 3.36E- GLM+Q BR 8 26074383 108026715 05 PZE- 3.40E- GLM+Q PHT 4 202279305 104124953 05 PZE- 3.42E- GLM+Q PHT 2 43372951 102065424 05 PZE- 3.53E- GLM+Q PHT 9 12983891 109012555 05 3.60E- GLM+Q PHT SYN9992 7 175637786 05 3.62E- GLM+Q BR SYN19928 8 128133446 05 PZE- 3.79E- GLM+Q GA 8 136129416 108080005 05 PZE- 3.80E- GLM+Q PHT 7 35080464 107028964 05 3.90E- GLM+Q PHT SYN28264 8 115029108 05 PZE- 4.08E- GLM+Q PHT 8 93929253 108052803 05

147

Table S4 continued PZE- 4.11E- GLM+Q PHT 3 209798825 103157755 05

PHT, ETH, NNode, LA represents for plant height, ear height, number of nodes, leaf angle, respectively; GLM+Q represents for Generalized linear model+ population structure.

148

Table S5: PHT associated SNP marker annotations and gene functions.

SNP Chr Position Gene annotation PZE- 1 26496101 NA 101039301 SYN35767 1 69881711 DUF581 domain containing protein expressed PZE- 1 79236710 Thymidine kinase 101087901 PZE- 1 207712012 Cytokinin oxidase 101164570 SYN10393 1 259824332 Plasma-membrane choline transporter SYN19565 2 30589468 Receptor-like protein kinase PZE- 2 43372951 Transporter 102065424 PZE- 2 49913767 PHD-finger domain protein 102071022 PZE- 2 59631929 NA 102077234 PZE- 2 179538585 NA 102129469 PZE- 3 69558888 NA 103056753 PZE- 3 110461222 NA 103060293 PZE- 3 122077074 NA 103073770 PZE- 3 125427407 NA 103075614 SYN31220 3 180977949 NA SYN23237 3 184601452 NA SYN20493 3 186435833 GTPase-activating protein PZE- 3 205260873 NA 103151399 PZE- 3 209798825 RING/U-box superfamily protein 103157755 PZE- 4 180053538 NA 104103747 PZE- 4 181429757 ATP synthase 104105165 PZE- 4 196930705 Tetratricopeptide repeat (TPR)-like superfamily protein 104119909 SYN35471 4 199278965 Squamosa promoter binding protein PZE- 4 202279305 Cytochrome P450 104124953

149

Table S5 continued PZA02479.1 4 212747880 Phox-associated domain PZE- 4 215385635 NA 104132368 SYN29112 4 231412099 RNA-binding SYN38535 5 28201445 NA PZE- GATA type zinc finger transcription factor - binding 6 114178476 106062991 proteins PZE- 6 114181162 NA 106063007 PZE- 6 114185516 NA 106063009 SYN35141 6 115294816 (ADL3, CF1, DL3, DRP2B) dynamin-like 3 PZE- 6 125452957 NA 106070979 PZE- 6 145180751 NA 106087627 PZE- 7 34172216 Lipid binding protein 107028505 PZE- 7 35080464 Methyltransferase 107028964 PZE- 7 35162894 NA 107029002 PZE- 7 149854371 NA 107094423 PZE- 7 171926341 kinase 107130789 SYN9992 7 175637786 NA PZA02388.1 8 169137 Squamosa promoter binding PZE- 8 68313123 tRNA-binding 108041894 PZE- 8 79424912 NA 108047365 PZE- 8 79934432 NA 108047536 PZE- 8 80503266 NA 108048010 PZE- 8 85863945 (ATCHX19, CHX19) cation/H+ exchanger 108049639 PZE- 8 87037136 (UXS4) UDP-xylose synthase 108050016 PZE- 8 93929253 Acyltransferase 108052803 PZE- 8 103638340 NA

150

SYN28264 8 115029108 Protein kinase PZE- 8 127625290 NA 108073190 PZE- 8 130213045 NA 108074750 SYN11836 8 131337588 MYB family transcription factor SYN19755 8 138390481 NA PZE- 8 139343576 ZOS1-10 - C2H2 zinc finger protein 108082592 PZE- 8 156213656 RING/FYVE/PHD zinc finger protein 108100090 PZE- 8 158762013 Phytochromobilin: ferredoxin oxidoreductase 108103023 SYN32654 8 160366514 RNA helicase SYN13209 8 164063270 ARM repeat superfamily protein PZE- 9 12983891 NA 109012555 PZE- 9 13755379 NA 109013469 SYN26803 9 28506261 NA ZM013369- 9 74171446 NA 0605 PZE- 9 74588367 NA 109044064 PZE- 9 74588641 NA 109044053 PZE- 9 86397696 NA 109049656 PZE- 9 94798138 NA 109054725 PZE- 9 110930390 NA 109066773 PZE- 9 139661789 NA 109092637 PZE- 9 152920575 NA 109116158 PZE- 9 154964992 NA 109121160 PZE- 10 1018195 NA 110000136 PZE- 10 112204539 Nudix hydrolase 110058308

151

Table S6: Raw data for the prediction of Mid-parent heterosis for plant height.

Hybrid P1BR P1GA P2BR P2GA MPHT BGEM-0001- 0.5 0.2 0.19 0.07 110.005 N/PHB47 BGEM-0003- 0.57 0.25 0.19 0.07 95.005 N/PHB47 BGEM-0005- 0.33 0.11 0.19 0.07 40.625 N/PHB47 BGEM-0006- 0.29 0.12 0.35 0.2 94.585 S/PHZ51 BGEM-0008- 0.31 0.14 0.35 0.2 90.215 S/PHZ51 BGEM-0010- 0.28 0.07 0.35 0.2 98.965 S/PHZ51 BGEM-0011- 0.37 0.15 0.35 0.2 87.085 S/PHZ51 BGEM-0012- 0.36 0.16 0.35 0.2 88.965 S/PHZ51 BGEM-0013- 0.28 0.07 0.35 0.2 107.085 S/PHZ51 BGEM-0014- 0.34 0.24 0.35 0.2 119.585 S/PHZ51 BGEM-0018- 0.18 0.07 0.35 0.2 95.835 S/PHZ51 BGEM-0019- 0.17 0.06 0.35 0.2 107.085 S/PHZ51 BGEM-0020- 0.28 0.06 0.35 0.2 94.585 S/PHZ51 BGEM-0022- 0.33 0.07 0.35 0.2 93.335 S/PHZ51 BGEM-0023- 0.22 0.07 0.35 0.2 108.335 S/PHZ51 BGEM-0024- 0.34 0.13 0.35 0.2 100.835 S/PHZ51 BGEM-0025- 0.34 0.13 0.35 0.2 110.215 S/PHZ51 BGEM-0026- 0.2 0.1 0.35 0.2 107.715 S/PHZ51 BGEM-0027- 0.26 0.07 0.35 0.2 90.835 S/PHZ51 BGEM-0028- 0.25 0.07 0.35 0.2 105.835 S/PHZ51 BGEM-0029- 0.32 0.09 0.35 0.2 110.215 S/PHZ51

152

Table S6 continued BGEM-0030- 0.28 0.12 0.35 0.2 108.335 S/PHZ51 BGEM-0031- 0.23 0.11 0.35 0.2 97.085 S/PHZ51 BGEM-0032- 0.5 0.15 0.35 0.2 109.585 S/PHZ51 BGEM-0033- 0.54 0.17 0.35 0.2 102.085 S/PHZ51 BGEM-0034- 0.26 0.1 0.35 0.2 129.585 S/PHZ51 BGEM-0035- 0.35 0.12 0.35 0.2 111.465 S/PHZ51 BGEM-0036- 0.49 0.12 0.35 0.2 120.835 S/PHZ51 BGEM-0037- 0.38 0.14 0.35 0.2 99.585 S/PHZ51 BGEM-0039- 0.4 0.23 0.19 0.07 118.125 N/PHB47 BGEM-0040- 0.56 0.21 0.19 0.07 113.755 N/PHB47 BGEM-0041- 0.21 0.06 0.35 0.2 98.465 S/PHZ51 BGEM-0042- 0.2 0.05 0.35 0.2 113.335 S/PHZ51 BGEM-0044- 0.31 0.08 0.35 0.2 115.215 S/PHZ51 BGEM-0045- 0.34 0.17 0.19 0.07 88.755 N/PHB47 BGEM-0046- 0.25 0.08 0.19 0.07 95.005 N/PHB47 BGEM-0047- 0.21 0.06 0.35 0.2 94.585 S/PHZ51 BGEM-0048- 0.4 0.2 0.35 0.2 106.465 S/PHZ51 BGEM-0049- 0.26 0.11 0.35 0.2 78.085 S/PHZ51 BGEM-0050- 0.21 0.07 0.35 0.2 97.085 S/PHZ51 BGEM-0051- 0.34 0.1 0.35 0.2 107.085 S/PHZ51 BGEM-0052- 0.4 0.17 0.35 0.2 76.465 S/PHZ51 BGEM-0054- 0.15 0.06 0.35 0.2 103.335 S/PHZ51

153

Table S6 continued BGEM-0055- 0.16 0.04 0.35 0.2 107.085 S/PHZ51 BGEM-0056- 0.25 0.08 0.35 0.2 106.465 S/PHZ51 BGEM-0057- 0.2 0.05 0.35 0.2 112.085 S/PHZ51 BGEM-0059- 0.28 0.04 0.35 0.2 105.215 S/PHZ51 BGEM-0063- 0.56 0.22 0.19 0.07 95.625 N/PHB47 BGEM-0066- 0.72 0.5 0.19 0.07 96.255 N/PHB47 BGEM-0067- 0.44 0.09 0.35 0.2 91.465 S/PHZ51 BGEM-0068- 0.39 0.1 0.35 0.2 105.835 S/PHZ51 BGEM-0070- 0.21 0.11 0.35 0.2 93.965 S/PHZ51 BGEM-0071- 0.4 0.09 0.35 0.2 122.085 S/PHZ51 BGEM-0072- 0.39 0.1 0.35 0.2 102.085 S/PHZ51 BGEM-0073- 0.23 0.09 0.35 0.2 123.965 S/PHZ51 BGEM-0074- 0.19 0.08 0.35 0.2 115.215 S/PHZ51 BGEM-0077- 0.44 0.13 0.35 0.2 99.585 S/PHZ51 BGEM-0078- 0.57 0.2 0.35 0.2 64.585 S/PHZ51 BGEM-0079- 0.12 0.05 0.35 0.2 50.215 S/PHZ51 BGEM-0080- 0.3 0.07 0.35 0.2 112.715 S/PHZ51 BGEM-0082- 0.19 0.07 0.35 0.2 90.215 S/PHZ51 BGEM-0083- 0.26 0.06 0.35 0.2 94.585 S/PHZ51 BGEM-0085- 0.55 0.29 0.19 0.07 94.375 N/PHB47 BGEM-0087- 0.44 0.16 0.19 0.07 95.005 N/PHB47 BGEM-0088- 0.67 0.28 0.19 0.07 103.755 N/PHB47

154

Table S6 continued BGEM-0089- 0.49 0.3 0.19 0.07 93.125 N/PHB47 BGEM-0090- 0.32 0.27 0.19 0.07 95.005 N/PHB47 BGEM-0094- 0.2 0.08 0.35 0.2 55.215 S/PHZ51 BGEM-0095- 0.51 0.12 0.35 0.2 97.715 S/PHZ51 BGEM-0097- 0.21 0.06 0.35 0.2 107.085 S/PHZ51 BGEM-0099- 0.41 0.14 0.35 0.2 104.585 S/PHZ51 BGEM-0100- 0.39 0.07 0.35 0.2 101.465 S/PHZ51 BGEM-0102- 0.4 0.13 0.19 0.07 100.005 N/PHB47 BGEM-0107- 0.24 0.06 0.19 0.07 68.125 N/PHB47 BGEM-0108- 0.37 0.19 0.35 0.2 107.715 S/PHZ51 BGEM-0110- 0.35 0.19 0.19 0.07 102.505 N/PHB47 BGEM-0111- 0.47 0.28 0.35 0.2 106.465 S/PHZ51 BGEM-0112- 0.34 0.2 0.35 0.2 75.835 S/PHZ51 BGEM-0113- 0.45 0.26 0.35 0.2 108.965 S/PHZ51 BGEM-0114- 0.47 0.23 0.35 0.2 112.085 S/PHZ51 BGEM-0115- 0.55 0.26 0.35 0.2 102.085 S/PHZ51 BGEM-0116- 0.37 0.22 0.35 0.2 90.835 S/PHZ51 BGEM-0117- 0.41 0.24 0.35 0.2 103.335 S/PHZ51 BGEM-0118- 0.42 0.19 0.35 0.2 113.335 S/PHZ51 BGEM-0120- 0.4 0.19 0.19 0.07 123.125 N/PHB47 BGEM-0121- 0.36 0.11 0.19 0.07 110.625 N/PHB47 BGEM-0122- 0.34 0.29 0.19 0.07 103.125 N/PHB47

155

Table S6 continued BGEM-0123- 0.44 0.24 0.19 0.07 116.255 N/PHB47 BGEM-0124- 0.44 0.23 0.19 0.07 105.625 N/PHB47 BGEM-0125- 0.5 0.18 0.19 0.07 83.125 N/PHB47 BGEM-0126- 0.39 0.19 0.19 0.07 94.375 N/PHB47 BGEM-0127- 0.39 0.25 0.19 0.07 100.625 N/PHB47 BGEM-0129- 0.21 0.07 0.19 0.07 68.755 N/PHB47 BGEM-0130- 0.31 0.22 0.19 0.07 97.505 N/PHB47 BGEM-0131- 0.4 0.09 0.19 0.07 118.125 N/PHB47 BGEM-0133- 0.25 0.07 0.35 0.2 108.965 S/PHZ51 BGEM-0134- 0.13 0.07 0.35 0.2 95.215 S/PHZ51 BGEM-0136- 0.43 0.1 0.35 0.2 105.835 S/PHZ51 BGEM-0137- 0.18 0.09 0.35 0.2 103.965 S/PHZ51 BGEM-0138- 0.26 0.14 0.35 0.2 96.465 S/PHZ51 BGEM-0140- 0.46 0.26 0.19 0.07 95.625 N/PHB47 BGEM-0141- 0.4 0.23 0.19 0.07 115.625 N/PHB47 BGEM-0143- 0.33 0.3 0.19 0.07 96.255 N/PHB47 BGEM-0144- 0.48 0.13 0.19 0.07 107.505 N/PHB47 BGEM-0145- 0.27 0.12 0.19 0.07 109.375 N/PHB47 BGEM-0146- 0.31 0.17 0.19 0.07 100.625 N/PHB47 BGEM-0149- 0.41 0.13 0.35 0.2 92.715 S/PHZ51 BGEM-0150- 0.22 0.05 0.35 0.2 113.335 S/PHZ51 BGEM-0151- 0.33 0.16 0.19 0.07 73.755 N/PHB47

156

Table S6 continued BGEM-0152- 0.37 0.17 0.19 0.07 56.875 N/PHB47 BGEM-0153- 0.24 0.08 0.35 0.2 116.465 S/PHZ51 BGEM-0154- 0.33 0.14 0.35 0.2 105.215 S/PHZ51 BGEM-0155- 0.2 0.05 0.35 0.2 114.585 S/PHZ51 BGEM-0156- 0.35 0.14 0.35 0.2 130.835 S/PHZ51 BGEM-0157- 0.45 0.15 0.19 0.07 96.255 N/PHB47 BGEM-0158- 0.43 0.11 0.35 0.2 75.215 S/PHZ51 BGEM-0159- 0.2 0.07 0.35 0.2 112.085 S/PHZ51 BGEM-0160- 0.23 0.09 0.35 0.2 105.215 S/PHZ51 BGEM-0161- 0.28 0.17 0.35 0.2 113.965 S/PHZ51 BGEM-0162- 0.2 0.09 0.35 0.2 104.585 S/PHZ51 BGEM-0164- 0.25 0.08 0.35 0.2 95.215 S/PHZ51 BGEM-0165- 0.44 0.17 0.35 0.2 63.335 S/PHZ51 BGEM-0166- 0.38 0.18 0.35 0.2 77.085 S/PHZ51 BGEM-0167- 0.39 0.15 0.35 0.2 108.965 S/PHZ51 BGEM-0169- 0.26 0.17 0.35 0.2 111.465 S/PHZ51 BGEM-0170- 0.33 0.13 0.35 0.2 90.215 S/PHZ51 BGEM-0172- 0.21 0.07 0.35 0.2 113.965 S/PHZ51 BGEM-0173- 0.4 0.07 0.35 0.2 111.465 S/PHZ51 BGEM-0174- 0.39 0.21 0.19 0.07 110.625 N/PHB47 BGEM-0175- 0.31 0.19 0.35 0.2 63.335 S/PHZ51 BGEM-0178- 0.31 0.07 0.35 0.2 87.085 S/PHZ51

157

Table S6 continued BGEM-0179- 0.28 0.05 0.35 0.2 107.085 S/PHZ51 BGEM-0182- 0.43 0.21 0.19 0.07 104.375 N/PHB47 BGEM-0184- 0.32 0.16 0.19 0.07 95.625 N/PHB47 BGEM-0186- 0.43 0.11 0.35 0.2 108.965 S/PHZ51 BGEM-0187- 0.28 0.12 0.35 0.2 133.335 S/PHZ51 BGEM-0188- 0.35 0.04 0.35 0.2 115.215 S/PHZ51 BGEM-0190- 0.44 0.31 0.19 0.07 100.625 N/PHB47 BGEM-0191- 0.65 0.56 0.19 0.07 105.625 N/PHB47 BGEM-0192- 0.24 0.21 0.19 0.07 96.255 N/PHB47 BGEM-0193- 0.28 0.18 0.19 0.07 101.255 N/PHB47 BGEM-0196- 0.43 0.26 0.19 0.07 95.005 N/PHB47 BGEM-0197- 0.36 0.12 0.35 0.2 92.085 S/PHZ51 BGEM-0198- 0.19 0.05 0.35 0.2 107.715 S/PHZ51 BGEM-0199- 0.32 0.07 0.35 0.2 92.085 S/PHZ51 BGEM-0200- 0.23 0.07 0.35 0.2 122.715 S/PHZ51 BGEM-0201- 0.37 0.15 0.19 0.07 109.375 N/PHB47 BGEM-0202- 0.42 0.29 0.19 0.07 113.125 N/PHB47 BGEM-0203- 0.36 0.19 0.19 0.07 95.625 N/PHB47 BGEM-0204- 0.42 0.22 0.19 0.07 85.625 N/PHB47 BGEM-0205- 0.45 0.25 0.19 0.07 73.125 N/PHB47 BGEM-0206- 0.3 0.23 0.19 0.07 95.625 N/PHB47 BGEM-0207- 0.18 0.29 0.19 0.07 109.375 N/PHB47

158

Table S6 continued BGEM-0209- 0.67 0.34 0.19 0.07 90.625 N/PHB47 BGEM-0210- 0.54 0.25 0.19 0.07 85.005 N/PHB47 BGEM-0211- 0.52 0.44 0.19 0.07 119.375 N/PHB47 BGEM-0212- 0.54 0.31 0.19 0.07 86.875 N/PHB47 BGEM-0213- 0.24 0.09 0.35 0.2 101.465 S/PHZ51 BGEM-0215- 0.29 0.02 0.19 0.07 76.875 N/PHB47 BGEM-0216- 0.39 0.16 0.19 0.07 100.005 N/PHB47 BGEM-0218- 0.45 0.16 0.35 0.2 89.585 S/PHZ51 BGEM-0221- 0.26 0.09 0.35 0.2 90.835 S/PHZ51 BGEM-0222- 0.16 0.1 0.35 0.2 107.715 S/PHZ51 BGEM-0223- 0.54 0.13 0.19 0.07 101.875 N/PHB47 BGEM-0224- 0.81 0.25 0.19 0.07 138.125 N/PHB47 BGEM-0225- 0.41 0.15 0.19 0.07 88.125 N/PHB47 BGEM-0226- 0.3 0.09 0.35 0.2 90.835 S/PHZ51 BGEM-0227- 0.36 0.08 0.19 0.07 88.755 N/PHB47 BGEM-0228- 0.42 0.23 0.19 0.07 107.505 N/PHB47 BGEM-0233- 0.24 0.07 0.35 0.2 92.085 S/PHZ51 BGEM-0235- 0.32 0.14 0.19 0.07 92.505 N/PHB47 BGEM-0236- 0.24 0.11 0.35 0.2 108.335 S/PHZ51 BGEM-0237- 0.39 0.16 0.19 0.07 108.755 N/PHB47 BGEM-0239- 0.36 0.14 0.19 0.07 102.505 N/PHB47 BGEM-0240- 0.33 0.21 0.19 0.07 108.125 N/PHB47

159

Table S6 continued BGEM-0242- 0.34 0.27 0.19 0.07 106.875 N/PHB47 BGEM-0243- 0.34 0.07 0.35 0.2 82.085 S/PHZ51 BGEM-0244- 0.24 0.07 0.35 0.2 110.215 S/PHZ51 BGEM-0247- 0.47 0.14 0.19 0.07 126.875 N/PHB47 BGEM-0248- 0.21 0.06 0.19 0.07 72.505 N/PHB47 BGEM-0250- 0.42 0.11 0.35 0.2 132.085 S/PHZ51 BGEM-0252- 0.25 0.08 0.35 0.2 110.215 S/PHZ51 BGEM-0253- 0.6 0.15 0.19 0.07 104.375 N/PHB47 BGEM-0254- 0.49 0.15 0.35 0.2 111.465 S/PHZ51 BGEM-0255- 0.29 0.11 0.35 0.2 97.715 S/PHZ51 BGEM-0256- 0.46 0.14 0.19 0.07 101.875 N/PHB47 BGEM-0257- 0.37 0.13 0.35 0.2 117.715 S/PHZ51 BGEM-0258- 0.27 0.12 0.35 0.2 120.835 S/PHZ51 BGEM-0259- 0.37 0.25 0.19 0.07 87.505 N/PHB47 BGEM-0260- 0.33 0.13 0.19 0.07 48.125 N/PHB47 BGEM-0261- 0.31 0.1 0.35 0.2 98.335 S/PHZ51 BGEM-0262- 0.39 0.27 0.35 0.2 60.215 S/PHZ51 BGEM-0263- 0.39 0.24 0.35 0.2 57.715 S/PHZ51 BGEM-0264- 0.26 0.14 0.35 0.2 63.335 S/PHZ51 BGEM-0265- 0.38 0.04 0.35 0.2 87.715 S/PHZ51 BGEM-0266- 0.39 0.25 0.35 0.2 62.085 S/PHZ51 BGEM-0268- 0.31 0.15 0.35 0.2 94.585 S/PHZ51

160

Table S6 continued BGEM-0269- 0.63 0.32 0.35 0.2 72.085 S/PHZ51 BGEM-0270- 0.45 0.3 0.35 0.2 105.835 S/PHZ51 BGEM-0272- 0.39 0.1 0.35 0.2 75.835 S/PHZ51 PHB47/PHZ51 0.19 0.07 0.35 0.2 95.51

P1BR, P2BR, P1GA, P2GA, represents for parent 1 or 2 with BR or GA inhibitor response, respectively. MPHT represents for mid-parent heterosis for plant height.

161

Table S7: Donor parent information and backcross generations for PIFs. Library Year Donor PIFB47 2013 2014 2015 Blanco Ayabaca - PIU119 B PIF1 BC4 BC5 BC6 OLOTON [GUA383]{CIMYT} PIF2 BC4 BC5 BC6 Rabo De Zorro - ANC 325 B PIF3 BC3 BC4 BC5 SABN CUN367 PIF4 BC3 BC4 BC5 MONTANA NAR625 PIF5 BC3 BC4 BC5 CABUYA SAN316 FTC PIF6 BC3 BC4 BC5 OLOTON [GUA383]{CIMYT} PIF7 BC4 BC5 BC6 COSTENO [VEN775]{ICA} PIF8 BC4 BC5 BC6 Avati Moroti Guapi - PAG 139 B PIF9 BC3 BC4 BC5 MOROCHO APUC77 PIF10 BC4 BC5 BC6 Amarillo Huancabamba - PIU 38B B PIF11 BC3 BC4 BC5 YUCATAN TOL389 ICA PIF12 BC5 BC6 BC7 COSTENO [VEN775]{ICA} PIF13 BC2 BC3 BC4 OLOTON [GUA383]{CIMYT} PIF14 BC4 BC5 BC6 ARAGUITO VEN678 PIF15 BC4 BC5 BC6 MOROC APC67 PIF16 BC4 BC5 BC6 CONICO [PUE116]{CIMYT} PIF17 BC3 BC4 BC5 CONICO [PUE116]{CIMYT} PIF18 BC4 BC5 BC6 DZ B GUA131 PIF19 BC5 BC6 BC7 Canilla - VEN 693 B PIF20 BC4 BC5 BC6 Rabo De Zorro - ANC 325 B PIF21 BC3 BC4 BC5 COSTENO [VEN775]{ICA} PIF22 BC2 BC3 BC4 ANDAQUI CAQ307 FT PIF23 BC3 BC4 BC5 MOROC APC67 PIF24 BC3 BC4 BC5 OLOTON [GUA383]{CIMYT} PIF25 BC4 BC5 BC6 Blanco Blandito - ECU 523 B PIF26 BC3 BC4 BC5 Chillo - ECU 458 B PIF27 BC3 BC4 BC5 MORADO BOV567 PIF28 BC3 BC4 BC5 ANDAQUI CAQ307 FT PIF29 BC4 BC5 BC6 ANCASHINO ANC102 PIF30 BC4 BC5 BC6 CABUYA SAN316 FTC PIF31 BC5 BC6 BC7 IMA 66 Ames8554 PIF32 BC5 BC6 BC7 PERLA ANC23 PIF33 BC4 BC5 BC6 CABUYA SAN316 FTC PIF34 BC5 BC6 BC7 Cuzco gigante PIF35 BC4 BC5 BC6 NANO # BOV1032 PIF36 BC4 BC5 BC6 YUCATAN TOL389 ICA PIF37 BC4 BC5 BC6 PARO CUZ76 PIF38 BC4 BC5 BC6 SAN MARCENO # [GUA724]{INIA} PIF39 BC5 BC6 BC7 Chirimito - VEN 703 B PIF40 BC2 BC3 BC4

162

Table S7 continued IMA 66 Ames8554 PIF41 BC3 BC4 BC5 YUCATAN TOL389 ICA PIF42 BC4 BC5 BC6 Blanco Ayabaca - PIU119 B PIF43 BC3 BC4 BC5 ALTIPLANO BOV903 PIF44 BC5 BC6 BC7 OLITA DGO159 PIF45 BC4 BC5 BC6 OLOTON [GUA383]{CIMYT} PIF46 BC3 BC4 BC5 TEPE CHS225 PIF47 BC5 BC6 BC7 POJ CHICO BOV800 PIF48 BC5 BC6 BC7 ALTIPLANO BOV903 PIF49 BC4 BC5 BC6 SAN MARCENO # [GUA724]{INIA} PIF50 BC5 BC6 BC7 Blanco Ayabaca - PIU119 B PIF51 BC3 BC4 BC5 Arequipeno - ARQ 1 B PIF52 BC3 BC4 BC5 ARAGUITO VEN678 PIF53 BC3 BC4 BC5 Chillo - ECU 458 B PIF54 BC4 BC5 BC6 CON NORT ZAC161 PIF55 BC4 BC5 BC6 PISAN BOV344 PIF56 BC5 BC6 BC7 ONAVENO SON24 PIF57 BC3 BC4 BC5 EARLY CARIBBEAN MAR9 PIF58 BC3 BC4 BC5 PISAN BOV344 PIF59 BC4 BC5 BC6 Chirimito - VEN 703 B PIF60 BC2 BC3 BC4 GORDO [CHH131]{CIMYT} PIF61 BC4 BC5 BC6 COSTENO [VEN775]{ICA} PIF62 BC4 BC5 BC6 TEPE CHS225 PIF63 BC4 BC5 BC6 CAMELIA CHI411 PIF64 BC3 BC4 BC5 Costeño - Antioquia 394 B PIF65 BC2 BC3 BC4 Canilla - VEN 693 B PIF66 BC4 BC5 BC6 Blanco Blandito - ECU 523 B PIF67 BC3 BC4 BC5 Cuzco gigante PIF68 BC2 BC3 BC4 ALTIPLANO BOV903 PIF69 BC3 BC4 BC5 OLOTON [GUA383]{CIMYT} PIF70 BC3 BC4 BC5 MISHCA ECU321 PIF71 BC4 BC5 BC6 COSTENO [VEN775]{ICA} PIF72 BC4 BC5 BC6 SABN CUN367 PIF73 BC4 BC5 BC6 SAN MARCENO # [GUA724]{INIA} PIF74 BC4 BC5 BC6 ELOTES OCCIDENT NAY29 PIF75 BC3 BC4 BC5 Donor PIFZ51 2013 2014 2015 Rienda - CAJ 80 B PIF76 BC4 BC5 BC6 NINUELO BOV1088 PIF77 BC4 BC5 BC6 Canilla - VEN693 B PIF78 BC4 BC5 BC6 DE T CAL. GUA159 PIF79 BC2 BC3 BC4 Chillo - ECU 458 PIF80 BC4 BC5 BC6 Canilla - VEN693 B PIF81 BC4 BC5 BC6 Chillo - ECU 411 PIF82 BC2 BC3 BC4

163

Table S7 continued Arizona - LIB 16 B PIF83 BC3 BC4 BC5 MISHC ECU321 PIF84 BC4 BC5 BC6 Arequipeno - ARQ 1 B PIF85 BC3 BC4 BC5 Cristal - SP X B PIF86 BC3 BC4 BC5 Argentino - BOV 920 PIF87 BC2 BC3 BC4 Chandelle - VEN 409 PIF88 BC2 BC3 BC4 BR105:N99v99v PIF89 BC2 BC3 BC4 Rienda - CAJ 80 B PIF90 BC4 BC5 BC6 Chirimito - VEN 703 B PIF91 BC4 BC5 BC6 SARCO ANC184 PIF92 BC4 BC5 BC6 CANGUIL GR. ECU447 PIF93 BC3 BC4 BC5 Uchuquilla - BOV318 PIF94 BC3 BC4 BC5 Rabo De Zorro - ANC 325 B PIF95 BC3 BC4 BC5 Morocho - APC 77 PIF96 BC3 BC4 BC5 CUZCO CUZ363 PIF97 BC4 BC5 BC6 Huevito - VEN 396 B PIF98 BC4 BC5 BC6 Avati Pichinga Ihu - BR 2830 PIF99 BC4 BC5 BC6 Negrita de tierra fria - GUA 522 PIF100 BC2 BC3 BC4 Oke - ARG 539 PIF101 BC3 BC4 BC5 Vandeño - GRO 96 B PIF102 BC3 BC4 BC5 Huarmaca - PIU 72 PIF103 BC3 BC4 BC5 MOROCHO APC67 PIF104 BC3 BC4 BC5 CON PN CUZ13 PIF105 BC5 BC6 BC7 DULCILLO DEL NO SON57 PIF106 BC3 BC4 BC5 Araguito - VEN 678 PIF107 BC5 BC6 BC7 ANDELA ECU699 ICA PIF108 BC5 BC6 BC7 YUNGUENO BOV362 PIF109 BC3 BC4 BC5 Cristal - SP X B PIF110 BC3 BC4 BC5 Comiteco - GUA 515 B PIF111 BC3 BC4 BC5 Blanco Ayabaca - PIU119 B PIF112 BC2 BC3 BC4 Araguito - VEN 678 PIF113 BC3 BC4 BC5 Clavo - NAR 329 PIF114 BC3 BC4 BC5 ANDELA ECU699 ICA PIF115 BC5 BC6 BC7 Araguito - VEN 678 PIF116 BC5 BC6 BC7 CON PN CUZ13 PIF117 BC5 BC6 BC7 Rabo De Zorro - ANC 325 B PIF118 BC4 BC5 BC6 Patillo - ECU 417 PIF119 BC4 BC5 BC6 Comiteco - GUA 515 B PIF120 BC4 BC5 BC6 SG HUANCAV JUN164 PIF121 BC3 BC4 BC5 CON PN CUZ13 PIF122 BC5 BC6 BC7 Avati Pichinga Ihu - BR 2830 PIF123 BC4 BC5 BC6 ANDELA ECU699 ICA PIF124 BC5 BC6 BC7 Patillo - ECU 417 PIF125 BC4 BC5 BC6

164

Table S7 continued Araguito - VEN 678 PIF126 BC5 BC6 BC7 Capio rosado - ARG 460 PIF127 BC3 BC4 BC5 MOROTI PEI PIF128 BC4 BC5 BC6 Comiteco - GUA 515 B PIF129 BC4 BC5 BC6 CON PN CUZ13 PIF130 BC5 BC6 BC7 Calchaqui white flint - ARG 2420 PIF131 BC3 BC4 BC5 ELOTES OCCIDENT NAY29 PIF132 BC3 BC4 BC5 Chillo - ECU 458 PIF133 BC5 BC6 BC7 Rienda - CAJ 80 B PIF134 BC4 BC5 BC6 Rabo De Zorro - ANC 325 B PIF135 BC4 BC5 BC6 DULCILLO DEL NO SON57 PIF136 BC3 BC4 BC5 Blanco Blandito - ECU 523 B PIF137 BC3 BC4 BC5 Culli - ARG 471 PIF138 BC2 BC3 BC4 Jora - ANC 1 B PIF139 BC3 BC4 BC5 BOFO DGO123 PIF140 BC5 BC6 BC7 Araguito - VEN 678 PIF141 BC3 BC4 BC5 Chillo - ECU 458 PIF142 BC2 BC3 BC4 NINUELO BOV1088 PIF143 BC4 BC5 BC6 Sin Clasificación - CAU454 PIF144 BC3 BC4 BC5 Chirimito - VEN 703 B PIF145 BC4 BC5 BC6 Chirimito - VEN 703 B PIF146 BC4 BC5 BC6

165

CHAPTER 5.

BRASSINOSTEROID AND GIBBERELLIN CONTROL OF SEEDLING

TRAITS IN MAIZE

Songlin Hu1, Darlene Sanchez1, Cuiling Wang2, Alexander E. Lipka3, Yanhai Yin4,

Candice A.C. Gardner1,5, Thomas Lübberstedt1

Paper in preparation to be submitted to the Plant Journal. Abstract, structure, and references are formatted according to the journal standards.

SH and TL conceived the study, designed the experiments, discussed the results and finalized the manuscript; DLS collected field data for BGEM lines and helped with genotype analyses; CW helped with plant hormone related experiments; AEL helped with statistical analysis. YY established protocol for measuring BR/GA inhibitor response. TL, AEL, YY, CG edited the manuscript; all authors read and approved the final manuscript.

Summary

Brassinosteroids (BRs) and gibberellins (GAs) are two major plant hormones regulating various plant developmental processes. In maize, BRs and GAs were shown to regulate field traits such as plant height and sex determination. In this study, we established two doubled haploid (DH) libraries with a total of 207 DH lines. We applied BR and GA inhibitors to all these

DH lines at seedling stage and measured BR and GA inhibitor response for four seedling traits.

Moreover, we evaluated plant height, flowering time, and yield for each DH line. We conducted

166 genome-wide association studies (GWAS) with 62,049 genome wide SNPs to explore the genetic control of seedling traits by BR and GA. In addition, we correlate seedling stage hormone inhibitor response with field traits. Large variation for BR and GA inhibitor response and field traits was observed across these DH lines. Seedling stage BR and GA inhibitor response was able to predict yield and flowering time. With three GWAS models of a general linear model, mixed linear model, and fixed and random model Circulating Probability Unification (FarmCPU), multiple SNPs were discovered to be significantly associated with BR and GA inhibitor response with some localized within gene models. SNPs from gene model GRMZM2G013391 were associated with GA inhibitor response across all three GWAS models. This gene is expressed in roots and shoots and was shown to regulate GA signaling. These results show that BRs and GAs have a great impact for controlling seedling growth. Gene models from GWAS results could be targets for seeding traits improvement.

Significance Statement

BR and GA related candidate genes were identified in this study for future maize improvement projects. In addition, seedling BR and GA inhibitor response was able to predict field traits for inbred lines.

Introduction

Brassinosteroids (BRs) and gibberellins(GAs) are two groups of plant hormones that control various plant developmental processes. Aberrations occurring in BR/GA biosynthesis or signaling can greatly alter plant stature and change plant responses to environmental and

167 developmental cues. BRs are steroid hormones similar to those found in animals (Fujioka and

Yokota, 2003). They regulate important traits such as germination, cell elongation, root development, fertility, flowering time, and plant architecture traits such as leaf angle and plant height (Clouse, 1996; Taiz and Zeiger, 2010; Vriet et al., 2012). In addition, BRs also play positive roles in resistance to both biotic and abiotic stresses such as drought, salt, heat, cold, oxidative stress and pathogen attacks (Krishna, 2003). BR biosynthesis and signaling pathways have been well established in the model species Arabidopsis thaliana (Clouse, 2011; Zhao and Li,

2012; Hao et al., 2013; Jiang et al., 2013; Zhu et al., 2013), and in rice (Oryza sativa) (Bishop and Koncz, 2002; Tong and Chu, 2012). These BR pathway components have shown a great potential to serve as targets for genetic engineering for crop improvement. For example, seed- specific over-expression of AtDWF4 in Arabidopsis can enhance cold tolerance, which was attributed to the up-regulation of the cold-responsive gene COR15A (Divi and Krishna, 2009).

Loss-of-function of OsGSK improved rice tolerance to cold, heat, salt, and drought stress (Koh et al., 2007). In contrast, role and pathway information of BRs in maize is still limited (Hartwig et al., 2011; Makarevitch et al., 2012).

GAs are a large group of cyclic diterpene compounds (Sakamoto et al., 2004). They are involved in seed germination and vegetative growth including elongation of stems, roots, and the expansion of leaves, development of flowers, fruit set, and the control of fertilization (Dellaporta and Calderon-Urrea, 1994; Sasaki et al., 2002; Oikawa et al., 2004; Fleet and Sun, 2005; Lange and Lange, 2006; Zhang et al., 2008). GA biosynthesis and signaling pathways have been extensively studied in Arabidopsis, rice, maize and other crops (Fleet and Sun, 2005; Ueguchi-

Tanaka et al., 2005; Ueguchi-Tanaka et al., 2007; de Lucas et al., 2008; Tong and Chu, 2012).

168

Mutations in GA pathway genes played a vital role in the green revolution such as the semidwarf1 (sd1) gene (GA biosynthesis pathway), which encodes a GA20oxidase (OsGA20ox2)

(Sasaki et al., 2002) and reduced height-1 (RHT-1) (GA signaling pathway) of wheat (Triticum aestivumL.) (Peng et al., 1999). With the knowledge of BR and GA control of traits and stress tolerance, crop improvement can greatly benefit in three ways 1) genetic engineering of BR and

GA biosynthesis and/or signaling pathways can be employed to improve stress tolerance, and perhaps biomass and yield of agricultural crops (Choudhary et al., 2012); 2) Molecular strengthening (Hu and Lübberstedt, 2015) treatments can be developed to optimize trait expression temporally; 3) functional markers (Andersen and Lübberstedt, 2003) can be developed for BR/GA pathway genes to breed better cultivars.

Establishment of seedlings is a key factor for uniform, high-yielding field stands and, consequently, stable yields (Amram et al., 2015). Several traits were used to measure seedling vigor in maize, including seed germination rate, seedling shoot length, root length, seedling dry weight, root dry weight etc., and these traits have been extensively used for correlation studies with adult plant traits in different maize genotypes (Asghar and Khan, 2005; Cervantes-Ortiz et al., 2007; Molatudi and Mariga, 2009; Abdel-Ghani et al., 2013). Early maize seedling vigor greatly enhances crop establishment, especially in stressful environments such as under low temperatures in central Europe and northern Mediterranean areas (Hund et al., 2004). In low precipitation areas, where the soil moisture in the upper layer is quite limited and seeds are exposed to frequent dehydration events, deep sowing could ensure adequate seed-zone moisture before germination and thereby enhance seedling establishment (Richards et al., 2010; Zhang et al., 2012). However, when sown deep, an elongated and vigorous mesocotyl is needed to push

169 the shoots to the surface. It was reported that deep-seeding tolerance was significantly associated with mesocotyl elongation (Troyer, 1997; Zhao et al., 2009). BRs and GAs control maize seedling architecture traits such as mesocotyl length, seeding and root length (Hartwig et al.,

2012). In addition, they are associated with seedling stress tolerance such as chilling tolerance and salinity stress (Anuradha and Rao, 2001; Li et al., 2013; Strigens et al., 2013). In this study, we used the BR and GA inhibitors Propiconazole (Pcz) and Uniconazole (Ucz) to explore the BR and GA levels in different maize genotypes due to their easy accessibility, high stability and low costs (Rademacher, 2000; Hartwig et al., 2012), and investigated the control of seedling traits by

BR and GA from physiological and genetic perspectives. Our main objectives were to i) study phenotypic variation of BR and GA inhibitor responses of four seedling traits within a diverse maize association panel, ii) use Genome-wide association studies (GWAS) to identify SNPs throughout the genome associated with BR and GA inhibitor responses, iii) investigate the correlation between seedling stage hormone level with field traits of plant height, flowering time and yield.

Results

Designations of measured BR and GA inhibitor responses of traits was summarized in

Table 1. All traits have shown considerable variation and diversity for BR and GA inhibitor responses within both DHB47 and DHZ51. The standard deviation for BRR and GAR varied the most with values of 0.14 and 0.11 within DHZ51, and 0.13 and 0.1 within DHB47. All trait BR and GA inhibitor responses maximum, minimum, mean, median and standard deviations are listed in Table 2. Specifically, averages of BRM, BRS, BRR, GAM, GAS, and GAR ranged from

0.2 to 0.8 across the entire panel. For these traits, the BR and GA inhibitor responses were below

170

1.0, indicating that the mesocotyl, shoot, and primary root lengths were inhibited by hormone inhibitor application. Moreover, GA inhibitor Ucz showed a stronger effect compared to BR inhibitor, as the values for GAM, GAS, and GAR were 0.21, 0.38, 0.71 in DHZ51, lower than those for BRM, BRS, and BRR, which were 0.42, 0.57, 0.84, respectively. Similar results were observed in DHB47 (Table 2). The means for BRM and GAM were lowest with 0.3-0.4 for BRM, and 0.1-0.2 for GAM, compared to other traits with BRR, BRS, BRW ranged from 0.5~1 and

GAR, GAS, GAW ranged from 0.3-1. This indicated that the mesocotyl length was more sensitive in response to BR and GA inhibitors compared to shoot length, primary root length and seedling dry weight. We noticed that the median of trait values in DHB47 and DHZ51 was close to their recurrent parent (PHB47 and PHZ51) performance (Table 2). For example, the median of

BRS in DHB47 was 0.49, close to BRS (0.47) for PHB47. The median of BRS in DHZ51 was

0.55, close to BRS (0.54) for PHZ51. This indicated that around half of the BGEM lines were more tolerant to BR and GA inhibitors compared to their recurrent parents. For heritability, BRM and GAM showed the highest values (H2~0.8) compared to seedling length (H2~0.55), root length (H2~0.3), and dry weight (H2~0.1). BRM, BRS, BRR, GAM, GAS, GAR were significantly and positively correlated (P=0.001) with each other (Figure 1), suggesting that higher levels of BR and GA increase elongation of mesocotyl length, shoot length and primary root length. However, BRW and GAW were not significantly correlated with any other traits, although significantly correlated with each other (P=0.001).

Field performance of BGEM lines

Large variation was also observed for field traits (Table 2). Standard deviation of

Growing Degree Units (GDUs; see methods section) for anthesis and silking was similar with

171

28.8 °C and 40.7 °C in DHZ51, and 32.7°C and 32.9 °C in DHB47. For ASI (GDUs), standard deviation in DHB47 was 15.1°C, smaller than 22.8°C in DHZ51. For PHT (cm) and yield (tons per hectare), the standard deviation within DHB47 was 20 cm, 0.9 tons per hectare, almost the same as in DHZ51 which was 21.5 cm and 0.9 tons per hectare, respectively. When BGEM lines were compared to the recurrent parents, the median of field trait values in DHB47 and DHZ51 was close to PHB47 and PHZ51 (Table 2). For example, the median value of PHT in DHB47 and

DHZ51 were 225 and 222.2 cm, respectively, and for PHB47 and PHZ51 230.8 and 225 cm, respectively. However, there was one exception for ASI (GDUs) in DHB47, where the median value is 8.8°C, while PHB47 was 0.2°C. Moreover, for yield, PHB47 and PHZ51 performed better than most of the BGEM lines (PHB47 was with the maximum value for yield). For heritability, PHT showed the highest value of H2~0.9, followed by flowering time (anthesis and silking) with H2~0.7, yield with H2~0.6, and ASI with H2~0.4 in DHB47 and H2~0.5 in DHZ51.

The highest correlation was observed between GDUs to anthesis and silking, with r=0.85. Both traits were significantly and negatively correlated with yield (α=0.001). ASI was negatively correlated with yield and PHT, and there was a positive correlation between yield and PHT

(α=0.001).

Correlation between hormone inhibitor responses and field traits

There was no significant correlation (α=0.01) between field traits PHT, GDUs for anthesis, and seedling BR and GA inhibitor responses. However, field traits yield, GDUs to silking, and ASI were significantly (α=0.01) correlated with seedling BR and GA inhibitor responses. Specifically, yield was significantly (α=0.001) and negatively correlated with BRM,

BRS, BRR, GAM, GAS, and GAR. The highest correlation was between yield and BRM with a

172 negative correlation of 0.3. GDUs for silking was positively and significantly (α=0.01) correlated with BRM, BRS, BRR, GAM, GAS, and GAR. The highest correlation was between GDUs for silking and BRM with a positive correlation of 0.26. ASI was significantly (α=0.001) and positively correlated with BRM, BRS, BRR, GAM, and GAS. The highest correlation was observed between BRM, GAM, and ASI, with a positive correlation of 0.33. In summary, seedling stage hormone inhibitor responses were predictive for field traits yield, ASI and GGD for silking for BGEM lines, and BRM, GAM are the best predictors (with highest correlation).

Population structure

Consistent with the two recurrent parent subsets, we found two sub-populations, used for joint analysis of DHB47 and DHZ51. One subpopulation comprised 60% BGEM lines used for

GWAS including PHB47, which include mostly BGEM lines from DHB47 group, and nine

BGEM lines BGEM-0005-N, BGEM-0085-N, BGEM-0107N, BGEM-0121N, BGEM-0129N,

BGEM-0215N, BGEM-0227N, BGEM-0248N, BGEM-0260N from DHZ51 group. Another subpopulation comprised 40% BGEM lines including PHZ51, including mostly BGEM lines from DHZ51 group, with nines BGEM-0007-S, BGEM-0052S, BGEM-0078-S, BGEM-0094-S,

BGEM-0165-S, BGEM-0166-S, BGEM-0175-S, BGEM-0220-S, BGEM-0266-S from the

DHB47 group. These BGEM lines are with on average around 50% donor genome proportion, larger than the average of donor genome proportion for the whole BGEM collection which was

18.3%.

173

Genome-wide association analyses

Three SNP markers S8_174376713 (P=1.57×10-7), S8_174338368 (P=1.24×10-6), and

S8_174376891 (P=1.75×10-6) were significantly associated with GAM (H2~0.85) using MLM and are located on Chromosome 8: S8_174376713 and S8_174376891 were from the same gene

GRMZM2G013391 and S8_174338368 is located in an intergenic region 40kb upstream of

GRMZM2G013391 (Figure 2A). Linkage disequilibrium (LD) between S8_174338368 and

GRMZM2G013391 was r2=0.85. Tissue specific expression of GRMZM2G013391 has been determined using high density NimbleGen Microarrays based on B73. It was shown that

GRMZM2G013391 was expressed with absolute expression level of 422 in shoots and 1899.04 in primary roots at seedling stage (Sekhon et al., 2011). It means that this gene is involved in metabolic processes for seedling growth. The homologs of GRMZM2G013391 in model species

Arabidopsis and rice are zinc-finger type transcription factors, designated as WRKY, and it was shown that a rice WRKY gene encodes a transcriptional repressor of the gibberellin signaling pathway (Zhang et al., 2004). We searched for GA candidate genes near GRMZM2G013391 to test the hypothesis that the signal was caused by strong LD between GA candidate genes and

GRMZM2G013391. The closest GA candidate gene ZmGA2ox1 is 500kb away from

GRMZM2G013391. However, the r2value between a SNP derived from this GA candidate gene and the two SNPs (significantly associated with GAM) from GRMZM2G013391 was below 0.1

(Figure 2B).

With the FarmCPU model, 19 significant SNPs were found using the same threshold of

P=2.55×10-6 for GAM, GAS, BRM, and BRS. For root traits GAR and BRR (H2 ranged from

0.2-0.4), no significant markers were detected using the FarmCPU model. Eleven candidate

174 genes were predicted based on the 19 SNPs and their surrounding regions of 2kb up- and downstream of markers (Table 3). Among them, four genes GRMZM2G024657,

GRMZM2G091919, GRMZM2G177050, GRMZM2G114911 were associated with BR inhibitor response (BRM and BRS), and seven genes GRMZM2G013391, GRMZM2G022258,

GRMZM2G001169, GRMZM2G148229, GRMZM5G891656, GRMZM2G412085,

GRMZM2G132663 were associated with GA inhibitor responses (GAM and GAS). All 11 genes are expressed at seedling shoots based on B73 (Sekhon et al., 2011), with gene

GRMZM2G022258 associated with the highest expression value of 9118.35. Its orthologue in

Arabidopsis is exportin protein that mediates the nuclear export of proteins, rRNA, snRNA, and some mRNA (Andorf et al., 2010). Orthologues for the other genes in Arabidopsis and rice are

WRKY transcription factor, cysteine proteinases, kinase, and binding proteins, among others

(Table S1). None of these genes were co-localized with any BR or GA candidate genes. We noticed that the most significant marker associated with GAM is exactly the same detected both by FarmCPU (SNP S8_174376713 in Figure 2) and MLM, with P=3.22×10-11 for FarmCPU.

Considering that the use of a MLM and FarmCPU could generate false negative results as both kinship and population structure are included to control the false positive results, we identified the most significant associations with GLM+PCA method, using the same threshold as

MLM and FarmCPU of P=2.55×10-6, resulting in 134 significant markers for all traits (Table S2).

More significant markers were detected, and some markers overlapped with MLM and

FarmCPU results. In particular, the top three markers from GLM+PCA are exactly the same as the three significant markers with MLM method (for GAM), with P-values 3.23×10-13, 3.45×10-

13, 4.32×10-12 for S8_174376891, S8_174376713, and S8_174338368, respectively. In total, 24

175 markers were significantly associated with GAM and GAS on Chromosome 8, clustered within a

1Mb region up- and downstream of gene GRMZM2G013391. As a result, gene

GRMZM2G013391 is significantly associated with GAM across all three methods of MLM,

FarmCPU, and GLM and always associated with the lowest P-values.

Except for gene GRMZM2G013391, there were four genes consistent across FarmCPU and GLM+PCA methods with two associated with BR inhibitor response and another two with

GA inhibitor response. Specifically, GRMZM2G024657 and GRMZM2G114911 were associated with BRM and BRS, respectively. GRMZM2G024657 codes for a cysteine proteinase, and it was found in common bean that cysteine proteinase was regulated by BR for germination and seedling elongation (Yamauchi, 2007). GRMZM2G001169 and GRMZM5G891656 were associated with GAM, however, they did not overlap with predicted GA candidate genes.

GRMZM2G001169 is a gene with unknown function, and GRMZM5G891656 is an oxidoreductase family protein.

Discussion

Plant hormones are small molecules that regulate many aspects regarding to plant growth and developments, as well as responses to changing environmental conditions. By modifying the production, distribution or signal transduction of plant hormones, plants are able to regulate and coordinate both growth and/or stress tolerance to promote survival or escape from environmental stress (Colebrook et al., 2014). In this study, we focused on BR and GA control of seedling traits, as BR and GA promote seedling vigor and growth (Fuchs and Lieberman, 1968; Evans and

Poethig, 1995; Clouse, 2001; Özdemir et al., 2004; Ashraf et al., 2010) and they are positively

176 correlated to seedling stress tolerance of drought, salinity, cold/heat, and heavy metals in different plant species such as maize, sorghum, Arabidopsis and rice (Anuradha and Rao, 2001;

Vardhini and Rao, 2003; Kagale et al., 2007; Colebrook et al., 2014). For example, exogenous application of BRs increased the cold tolerance of maize seedlings (Singh et al., 2012), and increased endogenous levels of GAs were effective in protecting maize seedlings from drought stress (Wang et al., 2008). As a high quality seedling establishment can lead to uniform field stands and stable yields, we investigated the BR and GA activities at seedling stage across 207

BGEM lines. On average, an expected 25% (with observed 18.3%) tropical germplasm was introgressed into PHB47 and PHZ51, representing the two maize heterotic groups Stiff Stalk and

Non Stiff Stalk in the corn belt area of United States. We measured the hormone inhibitor responses of these 207 BGEM lines, and an extensive amount of phenotypic variation was found for all traits measured (Table 2). For example, BGEM-0191-N was 28 times more tolerant to GA inhibitor compared to BGEM-0215N with regard to GAM, and BGEM-0269-S was 6 times more tolerant to BR inhibitor compared to BGEM-0079-S. Moreover, without selection, the donor segments have random effects for hormone inhibitor response – around half BGEM lines are more tolerant/not tolerant to BR and GA inhibitors for each trait measured compared to recurrent parents. In conclusion, there is substantial variation for BR and GA control of seedling traits.

For both DHB47 and DHZ51, shoot length, mesocotyl length, and primary root length were reduced with the application of BR and GA inhibitors compared to water treatment. BRM,

BRS, GAM and GAS was below 1.0 for all the 207 BGEM lines except for one line BGEM-

0123-N with BRS equal to 1.1. Primary root length was less sensitive to BR/GA inhibitor compared with shoot and mesocotyl length, as on average BRR and GAR was larger than BRM,

177

BRS, GAM and GAS. We noticed that on average, BRW and GAW was close to 1 (Table 2). The dry weight of most BGEM lines was not affected with the application of BR and GA inhibitors, although the length of both shoot and primary root length was significantly reduced. It may be because BR and GA regulate cell length instead of cell number and weight at seedling stage, and we observed that the seedlings were “fatter” and “shorter” with hormone inhibitors compared to water treatment. However, BRW and GAW were with very low heritabilities for both DHB47 and DHZ51, thus the effect of BR and GA inhibitor on seedling dry weight was not reliable.

Heritability values ranged from 0.01 to 0.87. BRM and GAM showed the highest heritability (>0.8) across both DHB47 and DHZ51, followed by BRS and GAS (0.5-0.7). By keeping all conditions equal, mesocotyl was the most stable and repeatable trait for reflecting BR and GA inhibitor response across different maize genotypes. BRR and GAR had a heritability close to 0.35, this corresponds to previous studies with similar ranges of heritability for root traits both under controlled environmental and field conditions (Krill et al., 2010; Cai et al., 2012;

Pace et al., 2015). For BRW and GAW, the heritability was very low and less than 0.1. This may because biomass traits were not measured as accurately as length traits, or seedling dry weight was very sensitive to environmental conditions.

BR and GA inhibitor response prediction for field traits

Seedling traits are easy to measure in controlled environments and have been used for comparison with adult plant traits. For example, since root traits are difficult and laborious to measure at the adult stage, measurements of seedling root architectural traits were compared with field root traits and some root traits have been shown to be positively correlated with yield

(Abdel-Ghani et al., 2013; Pace et al., 2015). Instead of using seedling traits per se, we

178 calculated the correlation between seedling stage BR/GA inhibitor response and field traits.

Because BRs and GAs regulate both seedling growth and field traits such as plant height, sex determination, and yield in maize (Hartwig et al., 2011; Vriet et al., 2012; Best et al., 2016), our hypothesis is that genotypes with elevated early BR and/or GA activities are more tolerant to BR and GA inhibitors, and will grow taller and with higher yield. If so, we can use seedling stage BR and GA inhibitor response to predict field traits for inbred lines. We found that BRM, BRS, BRR,

GAM, GAS were positively (α=0.01) correlated with GDUs to silking and ASI, and were negatively correlated with yield. Reduced yield may result from a longer ASI, as grain yield and its component, ears per plant, showed a dependence on ASI with a negative correlation

(Edmeades et al., 2000), the asynchrony of male and female flowering can cause reduced pollination rate. Although there was a positive correlation between PHT and yield, seedling stage

BR and GA inhibitor response did not predict field PHT for BGEM lines. The reason may be that, the BR and GA genes functioning at seedling stage and the developmental process to form PHT are different sets of genes in an inbred background. With a heterozygous background, our results showed that seedling stage BR and GA inhibitor response showed a very strong and positive correlation with plant height in maize (Hu et al., 2016 submitted).

Genome-wide association analyses

Genome-wide association studies identified candidate genes or putative functional markers affecting field traits, seedling traits, stress tolerance, or nutritional quality in maize

(Buckler et al., 2009; Weng et al., 2011; Yan et al., 2011; Huang et al., 2013; Lipka et al., 2013;

Pace et al., 2015). In this study, we used three methods (GLM+PCA, MLM, FarmCPU) to conduct genome-wide association studies for BR and GA inhibitor response, and found in total 3,

179

19, and 134 significant SNPs with MLM, FarmCPU, and GLM+PCA methods, respectively. As noted in other studies, MLM can overfit a model and create type II errors (Xue et al., 2013) and

GLM method can create type I errors. We made an effort to reduce type I errors by using GLM by fitting a PCA matrix to control for population structure, and by applying correction for multiple testing. In addition, we applied a recently developed method FarmCPU, which included both population structure and kinship matrix in the GWAS model like for the MLM, but used additional algorithms to address confounding problems between testing markers and covariates

(Liu et al., 2016). As a result, it is less stringent compared to the MLM method, but more stringent than the GLM+PCA model, and markers not detected by MLM, but significant using

FarmCPU, are representing the causal variants masked by kinship or between subpopulations.

This explains the discrepancy of the number of significant associations detected by these three methods. However, significant markers from MLM are consistent with GLM+PCA and

FarmCPU methods. Using all three methods in conjunction is preferable to balance the false positives and false negatives for gene trait associations.

Across all three association models, SNPs from GRMZM2G013391 were consistently associated with GAM and were always associated with the lowest P-values. Although there is a

GA candidate gene close to (within 0.5 Mb) GRMZM2G013391, the LD between these two genes is below 0.1. Moreover, GRMZM2G013391 is expressed throughout seedling development with absolute expression level of 422 in B73 shoots (Sekhon et al., 2011). It needs to be noted that expression is only based on B73, and variation in transcriptome profiles between multiple inbred lines has been reported (Hansey et al., 2012). The gene model of GRMZM2G013391 codes for WRKY transcription factors predicted from MaizeGDB (Monaco et al., 2013).

180

Plant WRKY gene family represents an ancient and complex class of zinc-finger transcription factors (TFs) that are involved in the regulation of various physiological processes and many plant pathways such as development and senescence, and in plant response to many biotic and abiotic stresses (Wang et al., 2014; Zhang et al., 2015). In rice, it was shown that a WRKY gene encodes a transcriptional repressor of the GA signaling pathway in aleurone cells: OsWRKY71 blocks GA signaling by functionally interfering with GA-inducible transcriptional activator

OsGAMYB and exogenous GA treatment decreases the steady-state mRNA level of OsWRKY71 and destabilizes the GFP:OsWRKY71 fusion protein (Zhang et al., 2004).

Moreover, synergistic interaction of ABA-inducible WRKY genes was regulating GAMYB- mediated GA signaling in aleurone cells, thereby establishing a novel mechanism for ABA and

GA signaling cross-talk (Xie et al., 2006). If the function of GRMZM2G013391 is similar as

WRKY transcription factors in rice to regulate GA signaling, this candidate gene could be a vital player in regulating GA levels in maize seedlings. In maize, WRKY transcription factors were found to be associated with chilling tolerance of maize seedlings, with a GWAS analysis of chilling tolerance indices (ratio of measurements taken under chilling stress and control conditions) (Huang et al., 2013). Since WRKY genes are associated with both GA activities and chilling tolerance, it is possible that WRKY genes are regulating chilling tolerance through increasing endogenous GA levels. It was shown that endogenous GA levels were positively associated with chilling stress tolerance: GA3 treatment increased chilling tolerance with decreased electrolyte leakage and malondialdehyde content, increased proline content, and improved antioxidant enzyme activities. Treatment with GA inhibitors exacerbated chilling injury (Ding et al., 2015). This may explain that in our study, we detected WRKY genes with GA

181 inhibitor response of seedlings, and WRKY genes were also associated with seedling chilling tolerance in previous studies.

Except for GRMZM2G013391, ten genes were found to be significantly associated with

BRM, BRS, GAM, and GAS with FarmCPU method, and they are all expressed throughout seedling development at different expression levels in B73 (Table 3). Based on the known functions of the homologs of these candidate genes in Arabidopsis and rice, these candidate genes are related to proteinases, protein kinase, transferase, binding proteins, and SNARE

(Soluble NSF Attachment Protein Receptor) proteins (Table S1). Some genes with enzyme activity functions were found to be regulated by BR and GA in different species. For example,

GRMZM2G024657 is predicted to code for cysteine proteinases and associated with BRM in this study. It was shown that BRs and GAs are involved in the expression of cysteine proteinases in cotyledons of common beans for germination. BRs and GAs regulate the synthesis of cysteine proteinases to degrade storage proteins in cotyledons of legume plants, to provide enough nutrition for the germination and growing process (Yamauchi, 2007). In addition,

GRMZM2G091919 is predicted to code for glycosyl transferase and also associated with BRM.

Glycosyl transferase is one of the most important modification reactions towards plant secondary metabolites, and plays a key role in maintaining cell homeostasis (Wang and Hou, 2009). It was shown that glycosyl transferase regulates BR activities. When UDP-glycosyltransferase 73C5

(UGT73C5) was ectopically overexpressed in Arabidopsis, the transgenic plants displayed characteristic BR-deficient dwarf phenotypes, and this dwarfism was reverted to wild-type morphology by exogenous application of epi-BL (Poppenberger et al., 2005). The genes associated with BR and GA inhibitor responses in this study may represent BR/GA pathway

182 components, or they regulate BR/GA activities, or they are regulated by BR/GA. They may play key roles for the BR and GA control of seedling traits such as mesocotyl length, shoot length and primary root length. Several loci with small effects were likely missed in our study due to limited population size (Yan et al., 2011). However, the genes identified by GWAS in this study are in various cases consistent with findings of earlier studies of large effect loci, which could be used as direct targets of marker assisted selection (Huang et al., 2013).

Experimental Procedures

A set of 207 BGEM (BC1 derived doubled haploid (DH) lines – with expected25% donor and 75% recurrent parent genome composition) were used in this study, which were obtained from the USDA-ARS North Central Regional Plant Introduction Station (NCRPIS) in Ames,

Iowa. Donor parents are landraces included in the Germplasm Enhancement of Maize (GEM) project. Recurrent parents are two expired PVP (Plant Variety Protection) lines PHB47 and

PHZ51, which are from the two maize heterotic groups Stiff Stalk and Non Stiff Stalk in corn belt area of United States, respectively. The methods used to create the BGEM lines were described in a previous study (Brenner et al., 2012). Briefly, various landrace accessions were used as donor parents, and PHB47/PHZ51served as recurrent parents for one generation of backcrossing. BC1 individuals were subsequently used to produce DH lines. There was no intentional selection during BGEM line development. BGEM lines are intended as a resource for biological research and potential discovery of unique alleles and traits. In this study, 76 BGEM lines are in PHZ51 and 131 BGEM lines in PHB47 background (DHZ51 and DHB47 subset, respectively). The pedigree, DH code, and donor accession information of the 207 BGEM lines is listed in Table S3.

183

Phenotyping

A paper roll cultivation method was employed as described in a previous study (Kumar et al., 2012) to germinate seed. All BGEM lines were treated with three independent treatments of

80µM BR inhibitor propiconazole (Pcz), 80µM GA inhibitor uniconazole (Ucz) and water as mock treatment. Each treatment was applied in three independent experiments completed June 8,

2015, June 19, 2015, and June 29, 2015 using a completely randomized design (CRD). For each treatment, 18 seeds from each BGEM line were used with 6 seeds assigned to each experiment.

Seed was germinated in three steps: 1) seed was soaked 24 hours for absorption of sufficient treatment solution; 2) seed was transferred into paper rolls containing the respective soaking solution (water, Pcz, or Ucz); 3) paper rolls were placed in covered buckets in a growth chamber without light at 25C. After eight days, seedlings were removed from the growth chamber and different traits were measured. Each paper roll with six seedlings was considered as an experimental unit.

Four seedling traits - mesocotyl length, shoot length, primary root length, and total dry weight were manually evaluated for the three treatments (BR inhibitor, GA inhibitor, and water).

Mesocotyl length (cm) was measured with a ruler from the root-shoot transition zone to the first node of the seedling. Shoot length (cm) was measured with a ruler from the root-shoot transition zone to the tip of the seedling. Primary root length (cm) was measured with a ruler from the root- shoot transition zone to the tip of the primary root. After these three measurements, seedlings were dried for 48 hrs at 55C (Pace et al., 2015), to determine the seedling dry weight (g).

Hormone inhibitor response was calculated as the ratios of the measurements taken under BR (or

GA) inhibitor and mock treatment (Huang et al., 2013). Except for seedling traits, field traits

184 plant height (PHT), Growing Degree Units (GDUs) to anthesis, GDUs to silking, anthesis silking interval (ASI) and yield (tons per hectare) were measured under regular field conditions

(nutrition, irrigation etc.,). PHT was evaluated at Agronomy Farm (Boone, IA) across two years

2014-2015 using a randomized complete block design with two replications (50 individuals per plot – with two rows). Two weeks after tasseling, a representative plant in the center of each plot was selected and PHT was measured from the ground to the top of the tassel. Yield and flowering time (anthesis, silking, ASI) were evaluated across three environments (with one replication within each environment) at Agronomy Farm2014, Agronomy Farm 2015, and the ISU research farm (2015, Nashua IA) following a completely randomized design. In each replication, 50 individuals from each BGEM line were grown per plot. At maturity, ears were machine- harvested and yield was measured on a plot basis at 15.5% moisture content and subsequently converted to tons per hectare (details were described in Sanchez et al., 2016 in preparation).

Days to anthesis and silking were counted from date of sowing to the day, when 50% of the plants in a plot had tassels with anthers exerted from the glumes and ears with emerged silks, respectively. These were converted to GDUs: GDUs = (Tmax +Tmin)/2 –10, where Tmax is maximum daily temperature and is set equal to 30 °C when temperatures exceeded 30 °C. Tmin is the minimum daily temperature and is set equal to 10 °C when temperatures fall below 10 °C.

ASI was calculated as the difference in GDUs between anthesis and silking.

Phenotypic data analysis

From each experimental unit (each roll): four seedlings out of six were sampled to eliminate possible not or poorly germinating seedlings, and means were taken per roll. The additive model for analysis of variance of seedling traits, flowering time and yield was:

185

th yij=µ+Ei+Gj+eij (Pace et al., 2015), where yij represents the observation from the ij experimental unit, µ is the overall mean, Ei is the experiment, and Gj is the genotype, and eij is the error. The interaction between the genotype and experiment is confounded within the error eij. Best linear unbiased prediction (BLUP) was calculated by fitting genotype and experiment as random effects in SAS 9.2 (SAS Institute, 2008). The additive model for analysis of variance of PHT was:

th yijk=µ+Ei+Rj(Ei)+Gk+Gk*Ei+eijk, where yijk represents the observation from the ijk experimental unit, µ is overall mean, Ei is the environment, Rj (Ei) is the replication nested within each environment, Gk is the genotype, and Gk*Ei is the interaction between genotype and environment, ejik is the error. Best linear unbiased prediction (BLUP) was calculated by fitting genotype, environment and replication as random effects in SAS 9.2 (SAS Institute, 2008). Heritability for all traits was calculated based on a “plot basis" (Holland et al., 2003). For each evaluated trait, variance component estimates were obtained from a mixed linear model fitted across all environments in SAS PROC MIXDED. Variance components ( ) were estimated according where corresponds to genotypic variance, genotype by environmental interaction variance, and error variance, respectively. Entry mean-based heritability was calculated from variance component estimates as where r is the number of replications within each location, and n is the number of locations (Holland et al., 2003).

Phenotypic correlations were calculated based on Pearson’s correlation method, and graphs were obtained using ggplot2 in R (Wickham, 2009).

186

Marker data

The diversity panel was genotyped with a genotyping-by-sequencing (GBS) protocol

(Elshire et al., 2011) by Cornell University from Buckler lab. Removal of monomorphic and low-quality SNPs, as well as those with minor allele frequencies below 2.5% and missing data rate >25%, generated a data set with 247,775 SNPs. For SNPs located at the same position (with genetic distance equal to 0 cM) in the genetic map, only one randomly selected SNP per map position was retained, which reduced the final number of SNPs to 62,049. For missing SNP markers, the LD k-nearest neighbor algorithm (LD KNNi imputation) was used for imputation using TASSEL 5.0 (Bradbury et al., 2007). We observed many short recurrent parent segments interspersed with exotic donor genotypes, leading to more than 1,000 recombinations across the genome per line. Since we only conducted one generation of backcrossing, the recombination frequency should be much lower. Thus, we corrected for monomorphic markers within large donor segments based on Bayes theorem (Sanchez et al., in preparation), with underlying assumption that the short recurrent parent segments are monomorphic markers instead of due to double recombinations. These short recurrent parent segments were corrected or kept as original genotype based on P-values from the Bayes theorem. After correction the donor genome composition was closer to the expected 25% compared to original marker data, and the average recombination rate substantially reduced. The corrected dataset was used for genome-wide association studies.

Population structure, linkage disequilibrium, and genome-wide association analyses

Population structure was estimated using the genome-wide 62,049 SNPs with principle component analysis (PCA). PCA was calculated with R package GAPIT (Genome Association

187 and Prediction Integrated Tool-R package) (Lipka et al., 2012). The most probable number of subpopulations was picked by plotting the number of PCAs (X-axis) against the variance explained by PCA numbers (Y-axis). The best number of subpopulations was selected, when the decrease of variance reached a plateau (no more variance can be explained by adding more PCs)

(Lipka et al., 2012). The software program TASSEL 5.0 (Bradbury et al., 2007) was used to calculate linkage disequilibrium (LD) among SNP markers.

BLUPs of trait values for BR/GA inhibitor response of mesocotyl length, shoot length and primary root length were used for GWAS with 62,049 SNPs. GWAS analysis for field traits is summarized by Sanchez et al. (in preparation). Seedling weight was not included due to low heritability (<0.3). To balance false positives and false negatives, three association analysis methods were applied in this study:1) General Linear Model (GLM)+PCA, with covariate PCA from GAPIT was included as fixed effects to account for population structure, 2) mixed linear model (MLM) (Yu et al., 2006) with population structure (PCA) and kinship included as covariates, and 3) FarmCPU (Fixed and random model Circulating Probability Unification) with kinship and population structure (PCA) as covariates, but with additional algorithms solving the confounding problems between testing markers and covariates (Liu et al., 2016). The software program TASSEL 5.0 (Bradbury et al., 2007) was used to conduct GWAS with GLM+PCA method. GAPIT (Genome Association and Prediction Integrated Tool-R package) (Lipka et al.,

2012) was applied to conduct MLM. The R package FarmCPU was used to conduct GWAS with

FarmCPU model. The statistical program simpleM implemented in R was used to account for multiple testing (Johnson et al., 2010). The threshold level was based on the effective number of independent tests m (Meff_G) and m was used in a similar way as the Bonferroni correction

188

(Pace et al., 2015). To obtain Meff_G for SNP data, a correlation matrix for all markers needs to be constructed and corresponding eigenvalues for each SNP locus calculated. A composite LD

(CLD) correlation is calculated directly from SNP genotypes (Lipka et al., 2012). Once this SNP matrix is created, the effective number of independent tests is calculated. In this study, with

α=0.05 as family-wise error rate, the threshold for significant trait-marker associations was set as

2.55×10-6 (multiple testing threshold level).

REFERENCES

Abdel-Ghani, A.H., Kumar, B., Reyes-Matamoros, J., Gonzalez-Portilla, P.J., Jansen, C., San Martin, J.P., Lee, M. and Lübberstedt, T. (2013) Genotypic variation and relationships between seedling and adult plant traits in maize (Zea mays L.) inbred lines grown under contrasting nitrogen levels. Euphytica. 189, 123-133.

Amram, A., Fadida-Myers, A., Golan, G., Nashef, K., Ben-David, R. and Peleg, Z. (2015) Effect of GA-sensitivity on wheat early vigor and yield components under deep sowing. Frontiers in plant science. 6.

Andersen, J.R. and Lübberstedt, T. (2003) Functional markers in plants. Trends Plant Sci. 8, 554-560.

Andorf, C.M., Lawrence, C.J., Harper, L.C., Schaeffer, M.L., Campbell, D.A. and Sen, T.Z. (2010) The Locus Lookup tool at MaizeGDB: identification of genomic regions in maize by integrating sequence information with physical and genetic maps. Bioinformatics. 26, 434-436.

Anuradha, S. and Rao, S.S.R. (2001) Effect of brassinosteroids on salinity stress induced inhibition of seed germination and seedling growth of rice (Oryza sativa L.). Plant Growth Regulation. 33, 151-153.

Asghar, M.J. and Khan, I.A. (2005) Variability for seedling vigour in two maize (Zea mays L.) populations I: means, variance components and heritabilities. Pak. J. Biol. Sci. 8, 839-843.

Ashraf, M., Akram, N., Arteca, R. and Foolad, M. (2010) The physiological, biochemical and molecular roles of brassinosteroids and salicylic acid in plant processes and salt tolerance. Crit. Rev. Plant Sci. 29, 162-190.

Best, N.B., Hartwig, T., Budka, J., Fujioka, S., Johal, G.G.S., Schulz, B. and Dilkes, B.P. (2016) nana plant2 encodes a maize ortholog of the Arabidopsis brassinosteroid biosynthesis

189 protein Dwarf1, identifying developmental interactions between brassinosteroids and gibberellins. Plant Physiol., pp. 00399.02016.

Bishop, G.J. and Koncz, C. (2002) Brassinosteroids and plant steroid hormone signaling. The Plant Cell. 14, S97-S110.

Bradbury, P.J., Zhang, Z., Kroon, D.E., Casstevens, T.M., Ramdoss, Y. and Buckler, E.S. (2007) TASSEL: software for association mapping of complex traits in diverse samples. Bioinformatics. 23, 2633-2635.

Brenner, E.A., Blanco, M., Gardner, C. and Lubberstedt, T. (2012) Genotypic and phenotypic characterization of isogenic doubled haploid exotic introgression lines in maize. Mol. Breed. 30, 1001-1016.

Buckler, E.S., Holland, J.B., Bradbury, P.J., Acharya, C.B., Brown, P.J., Browne, C., Ersoz, E., Flint-Garcia, S., Garcia, A., Glaubitz, J.C., Goodman, M.M., Harjes, C., Guill, K., Kroon, D.E., Larsson, S., Lepak, N.K., Li, H.H., Mitchell, S.E., Pressoir, G., Peiffer, J.A., Rosas, M.O., Rocheford, T.R., Romay, M.C., Romero, S., Salvo, S., Villeda, H.S., da Silva, H.S., Sun, Q., Tian, F., Upadyayula, N., Ware, D., Yates, H., Yu, J.M., Zhang, Z.W., Kresovich, S. and McMullen, M.D. (2009) The Genetic Architecture of Maize Flowering Time. Science. 325, 714-718.

Cai, H., Chen, F., Mi, G., Zhang, F., Maurer, H.P., Liu, W., Reif, J.C. and Yuan, L. (2012) Mapping QTLs for root system architecture of maize (Zea mays L.) in the field at different developmental stages. Theor. Appl. Genet. 125, 1313-1324.

Cervantes-Ortiz, F., Carballo-Carballo, A., Bergvinson, D., Crossa, J., Mendoza-Elos, M. and Moreno-Martinez, E. (2007) Seedling vigor inheritance and its relationship to adult plant traits in inbred tropical maize lines.

Choudhary, S.P., Yu, J.-Q., Yamaguchi-Shinozaki, K., Shinozaki, K. and Tran, L.-S.P. (2012) Benefits of brassinosteroid crosstalk. Trends Plant Sci.

Clouse, S.D. (1996) Molecular genetic studies confirm the role of brassinosteroids in plant growth and development. Plant J. 10, 1-8.

Clouse, S.D. (2001) Integration of light and brassinosteroid signals in etiolated seedling growth. Trends Plant Sci. 6, 443-445.

Clouse, S.D. (2011) Brassinosteroid Signal Transduction: From Receptor Kinase Activation to Transcriptional Networks Regulating Plant Development. Plant Cell. 23, 1219-1230.

Colebrook, E.H., Thomas, S.G., Phillips, A.L. and Hedden, P. (2014) The role of gibberellin signalling in plant responses to abiotic stress. J. Exp. Biol. 217, 67-75.

190 de Lucas, M., Davière, J.-M., Rodríguez-Falcón, M., Pontin, M., Iglesias-Pedraz, J.M., Lorrain, S., Fankhauser, C., Blázquez, M.A., Titarenko, E. and Prat, S. (2008) A molecular framework for light and gibberellin control of cell elongation. Nature. 451, 480-484.

Dellaporta, S.L. and Calderon-Urrea, A. (1994) The sex determination process in maize. Science. 266, 1501.

Ding, Y., Sheng, J., Li, S., Nie, Y., Zhao, J., Zhu, Z., Wang, Z. and Tang, X. (2015) The role of gibberellins in the mitigation of chilling injury in cherry tomato (Solanum lycopersicum L.) fruit. Postharvest Biol. Technol. 101, 88-95.

Divi, U.K. and Krishna, P. (2009) Brassinosteroid: a biotechnological target for enhancing crop yield and stress tolerance. New Biotechnology. 26, 131-136.

Edmeades, G., Bolanos, J., Elings, A., Ribaut, J.-M., Bänziger, M. and Westgate, M. (2000) The role and regulation of the anthesis-silking interval in maize. Physiology and modeling kernel set in maize, 43-73.

Elshire, R.J., Glaubitz, J.C., Sun, Q., Poland, J.A., Kawamoto, K., Buckler, E.S. and Mitchell, S.E. (2011) A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PloS one. 6, e19379.

Evans, M.M. and Poethig, R.S. (1995) Gibberellins promote vegetative phase change and reproductive maturity in maize. Plant Physiol. 108, 475-487.

Fleet, C.M. and Sun, T.-p. (2005) A DELLAcate balance: the role of gibberellin in plant morphogenesis. Curr. Opin. Plant Biol. 8, 77-85.

Fuchs, Y. and Lieberman, M. (1968) Effects of kinetin, IAA, and gibberellin on ethylene production, and their interactions in growth of seedlings. Plant Physiol. 43, 2029-2036.

Fujioka, S. and Yokota, T. (2003) Biosynthesis and metabolism of brassinosteroids. Annu. Rev. Plant Biol. 54, 137-164.

Hansey, C.N., Vaillancourt, B., Sekhon, R.S., De Leon, N., Kaeppler, S.M. and Buell, C.R. (2012) Maize (Zea mays L.) genome diversity as revealed by RNA-sequencing. PloS one. 7, e33071.

Hao, J., Yin, Y. and Fei, S.-z. (2013) Brassinosteroid signaling network: implications on yield and stress tolerance. Plant Cell Rep. 32, 1017-1030.

Hartwig, T., Chuck, G.S., Fujioka, S., Klempien, A., Weizbauer, R., Potluri, D.P.V., Choe, S., Johal, G.S. and Schulz, B. (2011) Brassinosteroid control of sex determination in maize. Proceedings of the National Academy of Sciences. 108, 19814-19819.

191

Hartwig, T., Corvalan, C., Best, N.B., Budka, J.S., Zhu, J.Y., Choe, S. and Schulz, B. (2012) Propiconazole Is a Specific and Accessible Brassinosteroid (BR) Biosynthesis Inhibitor for Arabidopsis and Maize. Plos One. 7.

Holland, J.B., Nyquist, W.E. and Cervantes-Martínez, C.T. (2003) Estimating and interpreting heritability for plant breeding: an update. Plant breeding reviews. 22, 9-112.

Hu, S. and Lübberstedt, T. (2015) Getting the ‘MOST’out of crop improvement. Trends Plant Sci.

Huang, J., Zhang, J., Li, W., Hu, W., Duan, L., Feng, Y., Qiu, F. and Yue, B. (2013) Genome‐wide association analysis of ten chilling tolerance indices at the germination and seedling stages in maize. Journal of integrative plant biology. 55, 735-744.

Hund, A., Fracheboud, Y., Soldati, A., Frascaroli, E., Salvi, S. and Stamp, P. (2004) QTL controlling root and shoot traits of maize seedlings under cold stress. Theor. Appl. Genet. 109, 618-629.

Jiang, J., Zhang, C. and Wang, X. (2013) Ligand perception, activation, and early signaling of plant steroid receptor brassinosteroid insensitive 1. Journal of integrative plant biology. 55, 1198-1211.

Johnson, R.C., Nelson, G.W., Troyer, J.L., Lautenberger, J.A., Kessing, B.D., Winkler, C.A. and O'Brien, S.J. (2010) Accounting for multiple comparisons in a genome-wide association study (GWAS). BMC Genomics. 11, 1.

Kagale, S., Divi, U.K., Krochko, J.E., Keller, W.A. and Krishna, P. (2007) Brassinosteroid confers tolerance in Arabidopsis thaliana and Brassica napus to a range of abiotic stresses. Planta. 225, 353-364.

Koh, S., Lee, S.-C., Kim, M.-K., Koh, J.H., Lee, S., An, G., Choe, S. and Kim, S.-R. (2007) T- DNA tagged knockout mutation of rice OsGSK1, an orthologue of Arabidopsis BIN2, with enhanced tolerance to various abiotic stresses. Plant Mol. Biol. 65, 453-466.

Krill, A.M., Kirst, M., Kochian, L.V., Buckler, E.S. and Hoekenga, O.A. (2010) Association and linkage analysis of aluminum tolerance genes in maize. PloS one. 5, e9958.

Krishna, P. (2003) Brassinosteroid-mediated stress responses. J. Plant Growth Regul. 22, 289- 297.

Kumar, B., Abdel‐Ghani, A.H., Reyes‐Matamoros, J., Hochholdinger, F. and Lübberstedt, T. (2012) Genotypic variation for root architecture traits in seedlings of maize (Zea mays L.) inbred lines. Plant breeding. 131, 465-478.

Lange, M.P. and Lange, T. (2006) Gibberellin biosynthesis and the regulation of plant development. Plant Biol. 8, 281-290.

192

Li, X., Jiang, H., Liu, F., Cai, J., Dai, T., Cao, W. and Jiang, D. (2013) Induction of chilling tolerance in wheat during germination by pre-soaking seed with nitric oxide and gibberellin. Plant Growth Regulation. 71, 31-40.

Lipka, A.E., Gore, M.A., Magallanes-Lundback, M., Mesberg, A., Lin, H., Tiede, T., Chen, C., Buell, C.R., Buckler, E.S. and Rocheford, T. (2013) Genome-wide association study and pathway-level analysis of tocochromanol levels in maize grain. G3: Genes| Genomes| Genetics. 3, 1287-1299.

Lipka, A.E., Tian, F., Wang, Q., Peiffer, J., Li, M., Bradbury, P.J., Gore, M.A., Buckler, E.S. and Zhang, Z. (2012) GAPIT: genome association and prediction integrated tool. Bioinformatics. 28, 2397-2399.

Liu, X., Huang, M., Fan, B., Buckler, E.S. and Zhang, Z. (2016) Iterative usage of fixed and random effect models for powerful and efficient genome-wide association studies. PLoS Genet. 12, e1005767.

Makarevitch, I., Thompson, A., Muehlbauer, G.J. and Springer, N.M. (2012) Brd1 gene in maize encodes a brassinosteroid C-6 oxidase. PloS one. 7, e30798.

Molatudi, R. and Mariga, I. (2009) The effect of maize seed size and depth of planting on seedling emergence and seedling vigour. Journal of Applied Sciences Research. 5, 2234-2237.

Monaco, M.K., Sen, T.Z., Dharmawardhana, P.D., Ren, L., Schaeffer, M., Naithani, S., Amarasinghe, V., Thomason, J., Harper, L. and Gardiner, J. (2013) Maize metabolic network construction and transcriptome analysis. The Plant Genome. 6.

Oikawa, T., Koshioka, M., Kojima, K., Yoshida, H. and Kawata, M. (2004) A role of OsGA20ox1, encoding an isoform of gibberellin 20-oxidase, for regulation of plant stature in rice. Plant Mol. Biol. 55, 687-700.

Özdemir, F., Bor, M., Demiral, T. and Türkan, İ. (2004) Effects of 24-epibrassinolide on seed germination, seedling growth, lipid peroxidation, proline content and antioxidative system of rice (Oryza sativa L.) under salinity stress. Plant Growth Regulation. 42, 203-211.

Pace, J., Gardner, C., Romay, C., Ganapathysubramanian, B. and Lübberstedt, T. (2015) Genome-wide association analysis of seedling root development in maize (Zea mays L.). BMC Genomics. 16, 1.

Peng, J., Richards, D.E., Hartley, N.M., Murphy, G.P., Devos, K.M., Flintham, J.E., Beales, J., Fish, L.J., Worland, A.J. and Pelica, F. (1999) ‘Green revolution’genes encode mutant gibberellin response modulators. Nature. 400, 256-261.

Poppenberger, B., Fujioka, S., Soeno, K., George, G.L., Vaistij, F.E., Hiranuma, S., Seto, H., Takatsuto, S., Adam, G. and Yoshida, S. (2005) The UGT73C5 of Arabidopsis thaliana glucosylates brassinosteroids. Proceedings of the National Academy of Sciences of the United States of America. 102, 15253-15258.

193

Rademacher, W. (2000) Growth retardants: effects on gibberellin biosynthesis and other metabolic pathways. Annu. Rev. Plant Biol. 51, 501-531.

Richards, R.A., Rebetzke, G.J., Watt, M., Condon, A.T., Spielmeyer, W. and Dolferus, R. (2010) Breeding for improved water productivity in temperate cereals: phenotyping, quantitative trait loci, markers and the selection environment. Funct. Plant Biol. 37, 85-97.

Sakamoto, T., Miura, K., Itoh, H., Tatsumi, T., Ueguchi-Tanaka, M., Ishiyama, K., Kobayashi, M., Agrawal, G.K., Takeda, S. and Abe, K. (2004) An overview of gibberellin metabolism enzyme genes and their related mutants in rice. Plant Physiol. 134, 1642-1653.

Sasaki, A., Ashikari, M., Ueguchi-Tanaka, M., Itoh, H., Nishimura, A., Swapan, D., Ishiyama, K., Saito, T., Kobayashi, M. and Khush, G.S. (2002) Green revolution: a mutant gibberellin-synthesis gene in rice. Nature. 416, 701-702.

Sekhon, R.S., Lin, H., Childs, K.L., Hansey, C.N., Buell, C.R., de Leon, N. and Kaeppler, S.M. (2011) Genome‐wide atlas of transcription during maize development. The Plant Journal. 66, 553-563.

Singh, I., Kumar, U., Singh, S., Gupta, C., Singh, M. and Kushwaha, S. (2012) Physiological and biochemical effect of 24-epibrassinoslide on cold tolerance in maize seedlings. Physiol. Mol. Biol. Plants. 18, 229-236.

Strigens, A., Freitag, N.M., Gilbert, X., Grieder, C., Riedelsheimer, C., Schrag, T.A., Messmer, R. and Melchinger, A.E. (2013) Association mapping for chilling tolerance in elite flint and dent maize inbred lines evaluated in growth chamber and field experiments. Plant, Cell Environ. 36, 1871-1887.

Taiz, L. and Zeiger, E. (2010) Plant physiology 5th Ed. Sunderland: Sinauer Assoc.

Tong, H. and Chu, C. (2012) Brassinosteroid signaling and application in rice. Journal of Genetics and Genomics. 39, 3-9.

Troyer, A.F. (1997) The location of genes governing long first internode of corn. Genetics. 145, 1149-1154.

Ueguchi-Tanaka, M., Ashikari, M., Nakajima, M., Itoh, H., Katoh, E., Kobayashi, M., Chow, T.-y., Yue-ie, C.H., Kitano, H. and Yamaguchi, I. (2005) GIBBERELLIN INSENSITIVE DWARF1 encodes a soluble receptor for gibberellin. Nature. 437, 693-698.

Ueguchi-Tanaka, M., Nakajima, M., Katoh, E., Ohmiya, H., Asano, K., Saji, S., Hongyu, X., Ashikari, M., Kitano, H. and Yamaguchi, I. (2007) Molecular interactions of a soluble gibberellin receptor, GID1, with a rice DELLA protein, SLR1, and gibberellin. The Plant Cell. 19, 2140-2155.

194

Vardhini, B.V. and Rao, S.S.R. (2003) Amelioration of osmotic stress by brassinosteroids on seed germination and seedling growth of three varieties of sorghum. Plant Growth Regulation. 41, 25-31.

Vriet, C., Russinova, E. and Reuzeau, C. (2012) Boosting Crop Yields with Plant Steroids. Plant Cell. 24, 842-857.

Wang, C., Yang, A., Yin, H. and Zhang, J. (2008) Influence of water stress on endogenous hormone contents and cell damage of maize seedlings. Journal of integrative plant biology. 50, 427-434.

Wang, J. and Hou, B. (2009) Glycosyltransferases: key players involved in the modification of plant secondary metabolites. Frontiers of Biology in China. 4, 39-46.

Wang, L., Zhu, W., Fang, L., Sun, X., Su, L., Liang, Z., Wang, N., Londo, J.P., Li, S. and Xin, H. (2014) Genome-wide identification of WRKY family genes and their response to cold stress in Vitis vinifera. BMC Plant Biol. 14, 1.

Weng, J., Xie, C., Hao, Z., Wang, J., Liu, C., Li, M., Zhang, D., Bai, L., Zhang, S. and Li, X. (2011) Genome-wide association study identifies candidate genes that affect plant height in Chinese elite maize (Zea mays L.) inbred lines. Plos One. 6, e29229.

Wickham, H. (2009) ggplot2: elegant graphics for data analysis: Springer Science & Business Media.

Xie, Z., Zhang, Z.L., Zou, X., Yang, G., Komatsu, S. and Shen, Q.J. (2006) Interactions of two abscisic‐acid induced WRKY genes in repressing gibberellin signaling in aleurone cells. The Plant Journal. 46, 231-242.

Xue, Y., Warburton, M.L., Sawkins, M., Zhang, X., Setter, T., Xu, Y., Grudloyma, P., Gethi, J., Ribaut, J.-M. and Li, W. (2013) Genome-wide association analysis for nine agronomic traits in maize under well-watered and water-stressed conditions. Theor. Appl. Genet. 126, 2587-2596.

Yamauchi, D. (2007) Hormonal regulation of the expression of cysteine proteinase genes in germinated cotyledons of common bean seeds. Plant Biotechnol. 24, 231-235.

Yan, J., Warburton, M. and Crouch, J. (2011) Association mapping for enhancing maize (L.) genetic improvement. Crop Sci. 51, 433-449.

Yu, J., Pressoir, G., Briggs, W.H., Bi, I.V., Yamasaki, M., Doebley, J.F., McMullen, M.D., Gaut, B.S., Nielsen, D.M. and Holland, J.B. (2006) A unified mixed-model method for association mapping that accounts for multiple levels of relatedness. Nat. Genet. 38, 203-208.

Zhang, H., Ma, P., Zhao, Z., Zhao, G., Tian, B., Wang, J. and Wang, G. (2012) Mapping QTL controlling maize deep-seeding tolerance-related traits and confirmation of a major QTL for mesocotyl length. Theor. Appl. Genet. 124, 223-232.

195

Zhang, L., Gu, L., Ringler, P., Smith, S., Rushton, P.J. and Shen, Q.J. (2015) Three WRKY transcription factors additively repress abscisic acid and gibberellin signaling in aleurone cells. Plant Sci. 236, 214-222.

Zhang, Y., Zhu, Y., Peng, Y., Yan, D., Li, Q., Wang, J., Wang, L. and He, Z. (2008) Gibberellin homeostasis and plant height control by EUI and a role for gibberellin in root gravity responses in rice. Cell Res. 18, 412-421.

Zhang, Z.-L., Xie, Z., Zou, X., Casaretto, J., Ho, T.-h.D. and Shen, Q.J. (2004) A rice WRKY gene encodes a transcriptional repressor of the gibberellin signaling pathway in aleurone cells. Plant Physiol. 134, 1500-1513.

Zhao, B. and Li, J. (2012) Regulation of Brassinosteroid Biosynthesis and InactivationF. Journal of integrative plant biology. 54, 746-759.

Zhao, G., Ma, P., Wang, J. and Wang, G. (2009) Identification of deep-seeding tolerance in different maize inbred lines and their physiological response to deep-seeding condition. J. Maize Sci. 17, 9-13.

Zhu, J.-Y., Sae-Seaw, J. and Wang, Z.-Y. (2013) Brassinosteroid signalling. Development. 140, 1615-1620.

196

Figure 1. Pearson correlations between all 13 traits collected. * and ** represent significance levels at α = 0.01 and 0.001, respectively. BRM, BRS, BRR, BRW, GAM, GAS, GAR, and

GAW: see Table 1 for explanation; ASI, anthesis silking interval; PHT, plant height.

197

Figure 2. GWAS for GAM in maize.

(A) GWAS results from the mixed linear model (MLM) of GAM. GAM represents for GA control of mescocotyl length and calculated as ratio of mesocotyl length under GA inhibitor treatment to the length under water treatment). X-axis represents the ten chromosomes and y-axis represents for the -log10P-values. The horizontal line represents for the threshold from SimpleM.

(B) Scatter plot of association results from MLM analysis of GAM and LD estimates (r2) across

Chromosome 8 where SNPs were found to be significantly associated with GAM. -log10P-values

198

Figure 2 continued

(left, y-axis) are from GWAS results and r2 values (right, y-axis) represent the LD between all

SNPs from Chromosome 8 to the top marker which locates within GRMZM2G01339. The grey vertical lines are -log10 P-values for SNPs. Triangles are the r2 values of each SNP relative to the peak SNP (indicated in red) at 174,376,713 bp. The black horizontal dashed line indicated the threshold from SimpleM.

199

Table 1. Trait designations and descriptions.

Trait name Trait description

BRM Ratio of mesocotyl length with BR inhibitor treatment to water treatment

GAM Ratio of mesocotyl length with GA inhibitor treatment to water treatment

BRS Ratio of shoot length with BR inhibitor treatment to water treatment

GAS Ratio of shoot length with GA inhibitor treatment to water treatment

BRR Ratio of primary root length with BR inhibitor treatment to water treatment

Ratio of primary root length with GA inhibitor treatment to water GAR treatment

Ratio of seedling dry weight with BR inhibitor treatment to water BRW treatment

Ratio of seedling dry weight with GA inhibitor treatment to water GAW treatment

200

Table 2. Summary statistics for both hormone inhibitor response and field traits.

Library: DHZ51 Trait Min Max Mean Median PHZ51 SD H2 BRM 0.18 0.81 0.42 0.4 0.46 0.12 0.71 BRS 0.39 0.88 0.57 0.55 0.54 0.11 0.52 BRW 0.87 1.5 1.12 1.11 1.11 0.12 0.01 BRR 0.54 1.18 0.84 0.82 0.94 0.14 0.19 GAM 0.02 0.56 0.21 0.21 0.28 0.09 0.87 GAS 0.2 0.79 0.38 0.38 0.4 0.09 0.68 GAR 0.48 1.07 0.71 0.71 0.73 0.11 0.38 GAW 0.75 1.39 1.04 1.04 0.96 0.11 0.05 Anthesis (GDUs) 723.9 881.8 786.8 784.1 766.3 28.8 0.65 Silking (GDUs) 745.2 927.8 813.2 811.1 783.7 40.7 0.70 ASI (GDUs) -11.8 104.7 26.2 19.7 17 22.8 0.53 PHT(cm) 177.2 268.3 222.6 222.2 225 20 0.85 Yield (tons/hectare) 0.9 4.6 2.6 2.5 3.9 0.9 0.59 Library: DHB47 Trait Min Max Mean Median PHB47 SD H2 BRM 0.12 0.63 0.31 0.31 0.28 0.1 0.8 BRS 0.31 0.73 0.5 0.49 0.47 0.09 0.5 BRR 0.47 1.12 0.75 0.75 0.58 0.13 0.39 BRW 0.86 1.64 1.13 1.12 1.11 0.13 0.03 GAM 0.04 0.32 0.12 0.1 0.09 0.06 0.85 GAS 0.15 0.53 0.29 0.28 0.27 0.07 0.68 GAR 0.38 1.06 0.64 0.62 0.57 0.1 0.39 GAW 0.79 1.3 1.03 1.02 1 0.1 0.01 Anthesis (GGD) 664.6 864.8 778.9 780.1 746.1 32.7 0.73 Silking (GDUs) 691.1 867 787.4 784.7 746.1 32.9 0.65 ASI (GDUs) -24.7 45.8 9.2 8.8 0.2 15.1 0.39 PHT(cm) 173.2 279 225.4 225 230.8 21.5 0.88 Yield 1.1 5.3 3.2 3.1 5.3 0.9 0.62

GDUs, Growing Degree Units (°C); SD, standard deviation; H2, heritability; ASI, anthesis silking interval; PHT, plant height. For BRM, BRS, BRR, BRW, GAM, GAS, GAR, and GAW: see Table 1 for explanation.

201

Table 3. Significant SNPs from FarmCPU.

Absolute P- expression Trait SNP Chr Position Gene value value in seedling

3.29E- BRM S5_3538677 5 3538677 GRMZM2G024657 69.88 08 3.03E- BRM S7_143003800 7 143003800 GRMZM2G091919 102.82 07 2.50E- BRM S7_172879110 7 172879110 GRMZM2G177050 163.72 06 1.06E- BRS S4_154986604 4 154986604 GRMZM2G114911 2173.76 06 1.41E- BRS S3_184357485 3 184357485 Intergenic 06 2.44E- BRS S9_76294986 9 76294986 Intergenic 06 3.22E- GAM S8_174376713 8 174376713 GRMZM2G013391 422 11 2.63E- GAM S5_749312 5 749312 GRMZM2G022258 9118.35 09 5.05E- GAM S4_2511802 4 2511802 GRMZM2G001169 46.65 09 4.37E- GAM S7_170538823 7 170538823 GRMZM2G148229 8265.48 08 7.41E- GAM S8_4409026 8 4409026 Intergenic 08 1.25E- GAM S9_152757586 9 152757586 GRMZM5G891656 256.55 06 6.44E- GAS S8_174377648 8 174377648 GRMZM2G013391 422 12 1.85E- GAS S3_2813553 3 2813553 GRMZM2G412085 423.37 11 1.93E- GAS S8_119054591 8 119054591 GRMZM2G132663 1704.02 08 2.14E- GAS S4_2511802 4 2511802 GRMZM2G001169 46.65 08 1.34E- GAS S3_3123403 3 3123403 Intergenic 07 2.96E- GAS S1_292671200 1 292671200 Intergenic 07 4.33E- GAS S10_140978407 10 140978407 Intergenic 07 Chr represents for Chromosome, BRM, BRS, GAM, GAS: see Table 1 for explanation.

202

Table S1. Orthologues of genes from FarmCPU results in Arabidopsis and rice. Trait Gene Arabidopsis BRM GRMZM2G024657 Cysteine proteinases superfamily protein BRM GRMZM2G091919 Protein kinase superfamily protein BRM GRMZM2G177050 SOS3-interacting protein 4 BRS GRMZM2G114911 ORMDL family protein GAM GRMZM2G013391 WRKY DNA-binding protein 65 GAM GRMZM2G022258 (ATCRM1, ATXPO1, HIT2, XPO1, XPO1A) exportin 1A GAM GRMZM2G001169 Protein of unknown function (DUF1645) GAM GRMZM2G148229 SNARE associated Golgi protein family GAM GRMZM5G891656 FAD/NAD(P)-binding oxidoreductase family protein GAS GRMZM2G013391 WRKY DNA-binding protein 65 GAS GRMZM2G412085 Unknown GAS GRMZM2G132663 Unknown GAS GRMZM2G001169 Protein of unknown function (DUF1645)

203

Table S2. SNPs from GLM+PCA results. Trait Marker Chr Position (bp) P-value GAM S8_174376891 8 174376891 3.23E-13 GAM S8_174376713 8 174376713 3.45E-13 GAM S8_174338368 8 174338368 4.32E-12 GAM S8_174338241 8 174338241 1.83E-11 GAM S8_174377648 8 174377648 3.35E-11 GAM S6_115077944 6 115077944 1.62E-10 GAM S4_2511802 4 2511802 2.14E-10 GAM S8_173872193 8 173872193 8.35E-10 GAM S6_142405130 6 142405130 5.20E-09 GAS S3_197247986 3 197247986 6.11E-09 GAM S3_210838424 3 210838424 6.48E-09 GAS S3_210838424 3 210838424 7.07E-09 GAM S3_197247986 3 197247986 7.17E-09 GAM S8_173872578 8 173872578 8.47E-09 GAS S6_142405130 6 142405130 1.05E-08 GAM S4_29298 4 29298 1.28E-08 GAM S1_135726530 1 135726530 1.33E-08 GAM S8_171750406 8 171750406 1.41E-08 GAM S4_30007 4 30007 1.50E-08 GAM S8_173873070 8 173873070 1.55E-08 GAM S8_173888280 8 173888280 1.58E-08 GAM S8_171751561 8 171751561 1.69E-08 GAS S3_178727369 3 178727369 1.91E-08 GAM S8_171138492 8 171138492 2.08E-08 GAM S1_30215825 1 30215825 2.36E-08 GAS S1_156436847 1 156436847 3.04E-08 GAS S1_156531194 1 156531194 3.04E-08 GAS S3_176366179 3 176366179 3.24E-08 GAM S8_173872457 8 173872457 3.31E-08 GAR S6_87590837 6 87590837 3.77E-08 GAM S8_174398658 8 174398658 4.54E-08 BRS S4_154986604 4 154986604 5.02E-08 GAM S5_206554916 5 206554916 5.48E-08 GAM S5_206555336 5 206555336 5.48E-08 GAM S8_171137903 8 171137903 6.76E-08 GAM S8_171169875 8 171169875 6.85E-08 GAR S6_66346291 6 66346291 7.90E-08 GAM S3_178727369 3 178727369 8.11E-08

204

Table S2 continued GAM S3_176366179 3 176366179 8.24E-08 GAM S8_171679655 8 171679655 9.04E-08 GAM S8_171076169 8 171076169 9.64E-08 GAM S5_158289246 5 158289246 9.71E-08 GAS S5_57247757 5 57247757 1.10E-07 GAM S6_61674688 6 61674688 1.12E-07 GAM S8_174400650 8 174400650 1.16E-07 GAM S6_115641175 6 115641175 1.21E-07 GAM S8_171881360 8 171881360 1.58E-07 GAM S4_2512830 4 2512830 1.69E-07 GAM S8_174398862 8 174398862 1.71E-07 BRS S4_155009009 4 155009009 1.80E-07 GAM S4_30681 4 30681 1.80E-07 GAM S8_171750740 8 171750740 2.00E-07 GAM S8_174329369 8 174329369 2.01E-07 GAM S2_199389407 2 199389407 2.24E-07 GAM S5_158407271 5 158407271 2.33E-07 GAM S1_5328404 1 5328404 3.11E-07 GAM S1_156436847 1 156436847 3.61E-07 GAM S1_156531194 1 156531194 3.61E-07 GAM S9_152757586 9 152757586 3.94E-07 GAS S6_115077944 6 115077944 4.29E-07 GAM S8_171649742 8 171649742 4.33E-07 GAM S8_171745653 8 171745653 4.69E-07 GAM S8_174312656 8 174312656 4.82E-07 GAM S8_172680709 8 172680709 4.82E-07 GAM S4_2487687 4 2487687 5.06E-07 GAM S8_171745131 8 171745131 5.21E-07 GAS S6_61674688 6 61674688 5.21E-07 GAM S9_143892516 9 143892516 5.58E-07 GAM S2_236476080 2 236476080 5.75E-07 GAM S4_2513379 4 2513379 5.94E-07 GAS S3_172198353 3 172198353 6.06E-07 GAM S4_32475 4 32475 6.32E-07 GAM S2_199206870 2 199206870 6.46E-07 GAM S1_77891902 1 77891902 6.48E-07 GAM S4_2514646 4 2514646 6.53E-07 GAM S3_3202458 3 3202458 6.60E-07 GAM S4_2487570 4 2487570 7.76E-07

205

GAM S8_174316462 8 174316462 7.91E-07 GAM S5_6543680 5 6543680 8.88E-07 GAS S8_174338368 8 174338368 9.09E-07 GAM S8_171037601 8 171037601 9.62E-07 GAM S2_184698003 2 184698003 9.86E-07 GAM S8_171170018 8 171170018 9.91E-07 GAR S6_80949310 6 80949310 9.92E-07 GAM S2_199462332 2 199462332 1.00E-06 GAM S4_2513671 4 2513671 1.04E-06 GAR S6_103702345 6 103702345 1.06E-06 GAM S3_3472787 3 3472787 1.07E-06 GAS S8_174377648 8 174377648 1.12E-06 GAM S1_8266562 1 8266562 1.14E-06 GAM S1_211893357 1 211893357 1.19E-06 GAM S8_174398727 8 174398727 1.21E-06 BRS S9_82352749 9 82352749 1.22E-06 GAM S1_6163742 1 6163742 1.22E-06 GAM S3_3698698 3 3698698 1.24E-06 GAM S2_209921466 2 209921466 1.28E-06 GAM S3_81926 3 81926 1.28E-06 GAM S1_211894567 1 211894567 1.30E-06 BRS S9_76294986 9 76294986 1.31E-06 BRS S5_3538677 5 3538677 1.32E-06 GAM S4_31326 4 31326 1.41E-06 GAM S1_195720491 1 195720491 1.46E-06 GAM S8_171795838 8 171795838 1.47E-06 GAM S3_221218307 3 221218307 1.47E-06 GAR S6_92390733 6 92390733 1.54E-06 BRS S9_58847187 9 58847187 1.55E-06 GAM S1_208097436 1 208097436 1.62E-06 GAM S8_171647664 8 171647664 1.62E-06 GAM S8_171072149 8 171072149 1.66E-06 GAM S2_184646154 2 184646154 1.74E-06 GAS S1_161614796 1 161614796 1.74E-06 GAM S8_160922631 8 160922631 1.79E-06 GAM S8_171751162 8 171751162 1.88E-06 GAM S3_221218730 3 221218730 1.93E-06 GAM S8_160925172 8 160925172 1.93E-06 GAR S6_4896509 6 4896509 1.94E-06 GAM S8_160978625 8 160978625 1.95E-06

206

GAM S8_160927952 8 160927952 1.95E-06 GAM S8_172126644 8 172126644 1.96E-06 GAM S1_127046561 1 127046561 1.99E-06 GAS S8_174376891 8 174376891 2.01E-06 GAM S4_2066556 4 2066556 2.02E-06 GAM S8_171759875 8 171759875 2.02E-06 GAM S1_206268074 1 206268074 2.09E-06 GAS S5_28022363 5 28022363 2.17E-06 GAM S1_12596372 1 12596372 2.21E-06 GAM S5_111307116 5 111307116 2.22E-06 GAM S8_173389037 8 173389037 2.32E-06 GAS S8_171169875 8 171169875 2.34E-06 GAM S8_174329425 8 174329425 2.42E-06 GAM S8_173851481 8 173851481 2.48E-06 GAM S3_6677098 3 6677098 2.52E-06 GAM S6_114931605 6 114931605 2.55E-06 GAM S8_174329308 8 174329308 2.55E-06

GAM, GAS, GAR, BRS: see Table 1 for explanation.

207

Table S3. Detailed information about the 207 BGEM lines

Donor DH code Pedigree Accession number BGEM-0001- (ALTIPLANO BOV903/PHZ51)/PHZ51 #002-(2n)-003 PI 485364 N BGEM-0003- (ALTIPLANO BOV903/PHZ51)/PHZ51 #005-(2n)-001 PI 485364 N BGEM-0004- (ALTIPLANO BOV903/PHZ51)/PHZ51 #005-(2n)-003 PI 485364 N BGEM-0005- (ALTIPLANO BOV903/PHZ51)/PHZ51 #005-(2n)-005 PI 485364 N BGEM-0006- (ALTIPLANO BOV903/PHB47)/PHB47 #001-(2n)-002 PI 485364 S BGEM-0007- (ALTIPLANO BOV903/PHB47)/PHB47 #001-(2n)-004 PI 485364 S BGEM-0008- (ALTIPLANO BOV903/PHB47)/PHB47 #001-(2n)-005 PI 485364 S BGEM-0010- (ALTIPLANO BOV903/PHB47)/PHB47 #003-(2n)-002 PI 485364 S BGEM-0011- (ALTIPLANO BOV903/PHB47)/PHB47 #004-(2n)-002 PI 485364 S BGEM-0012- (ALTIPLANO BOV903/PHB47)/PHB47 #004-(2n)-003 PI 485364 S BGEM-0013- (ALTIPLANO BOV903/PHB47)/PHB47 #005-(2n)-001 PI 485364 S BGEM-0014- (ALTIPLANO BOV903/PHB47)/PHB47 #005-(2n)-003 PI 485364 S BGEM-0018- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #003-(2n)-002 PI 444284 S BGEM-0019- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #003-(2n)-004 PI 444285 S BGEM-0020- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #003-(2n)-005 PI 444286 S BGEM-0022- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #003-(2n)-007 PI 444287 S BGEM-0023- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #005-(2n)-001 PI 444288 S BGEM-0024- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #006-(2n)-001 PI 444289 S BGEM-0025- (ANDAQUI CAQ307 FT/PHB47)/PHB47 #007-(2n)-002 PI 444290 S BGEM-0026- (ARAGUITO VEN678/PHB47)/PHB47 #003-(2n)-002 NSL 283561 S

208

Table S3 continued BGEM-0027- (ARAGUITO VEN678/PHB47)/PHB47 #003-(2n)-003 NSL 283561 S BGEM-0028- (ARAGUITO VEN678/PHB47)/PHB47 #003-(2n)-004 NSL 283561 S BGEM-0029- (ARAGUITO VEN678/PHB47)/PHB47 #005-(2n)-003 NSL 283561 S BGEM-0030- (ARAGUITO VEN678/PHB47)/PHB47 #005-(2n)-004 NSL 283561 S BGEM-0031- (ARAGUITO VEN678/PHB47)/PHB47 #005-(2n)-005 NSL 283561 S BGEM-0032- (ARAGUITO VEN678/PHB47)/PHB47 #006-(2n)-002 NSL 283561 S BGEM-0033- (ARAGUITO VEN678/PHB47)/PHB47 #006-(2n)-004 NSL 283561 S BGEM-0034- (ARAGUITO VEN678/PHB47)/PHB47 #006-(2n)-006 NSL 283561 S BGEM-0035- (ARAGUITO VEN678/PHB47)/PHB47 #006-(2n)-007 NSL 283561 S BGEM-0036- (ARAGUITO VEN678/PHB47)/PHB47 #006-(2n)-008 NSL 283561 S BGEM-0037- (ARAGUITO VEN678/PHB47)/PHB47 #006-(2n)-009 NSL 283561 S BGEM-0039- ((Arizona - LIB 16/PHZ51 B)/PHZ51)-(2n)-002 PI 485359 N BGEM-0040- ((Arizona - LIB 16/PHZ51 B)/PHZ51)-(2n)-003 PI 485360 N BGEM-0041- ((Arizona - LIB 16/PHB47 B)/PHB47)-(2n)-001 PI 485361 S BGEM-0042- ((Avati Moroti Guapi - PAG 139/PHB47 B)/PHB47)-(2n)- PI 485458 S 003 BGEM-0044- ((Blanco Blandito - ECU 523/PHB47 B)/PHB47)-(2n)-001 PI 488113 S BGEM-0045- (BOFO DGO123/PHZ51)/PHZ51 #002-(2n)-001 Ames 28481 N BGEM-0046- (BOFO DGO123/PHZ51)/PHZ51 #002-(2n)-002 Ames 28481 N BGEM-0047- (BOFO DGO123/PHB47)/PHB47 #001-(2n)-001 Ames 28481 S BGEM-0048- (BOFO DGO123/PHB47)/PHB47 #002-(2n)-003 Ames 28481 S BGEM-0049- (BOFO DGO123/PHB47)/PHB47 #003-(2n)-003 Ames 28481 S BGEM-0050- (BOFO DGO123/PHB47)/PHB47 #004-(2n)-001 Ames 28481 S

209

Table S3 continued BGEM-0051- (BOFO DGO123/PHB47)/PHB47 #004-(2n)-002 Ames 28481 S BGEM-0052- (BOFO DGO123/PHB47)/PHB47 #004-(2n)-003 Ames 28481 S BGEM-0054- (CAMELIA CHI411/PHB47)/PHB47 #002-(2n)-001 NSL 42755 S BGEM-0055- (CAMELIA CHI411/PHB47)/PHB47 #005-(2n)-002 NSL 42755 S BGEM-0056- (CAMELIA CHI411/PHB47)/PHB47 #005-(2n)-003 NSL 42755 S BGEM-0057- (CAMELIA CHI411/PHB47)/PHB47 #005-(2n)-006 NSL 42755 S BGEM-0059- (CANDELA ECU531/PHB47)/PHB47 #003-(2n)-001 NSL 287040 S BGEM-0063- (CHANDELLE CUB68 CI/PHZ51)/PHZ51 #004-(2n)-001 Ames 28574 N BGEM-0066- (PHZ51/CON PN CUZ13)/PHZ51 #007-(2n)-001 Ames 28653 N BGEM-0067- (PHB47/CON PUNT CUZ13)/PHB47 #002-(2n)-001 Ames 28653 S BGEM-0068- (PHB47/CON PUNT CUZ13)/PHB47 #005-(2n)-001 Ames 28653 S BGEM-0070- (CON NORT ZAC161/PHB47)/PHB47 #003-(2n)-001 PI 515577 S BGEM-0071- (CON NORT ZAC161/PHB47)/PHB47 #004-(2n)-001 PI 515577 S BGEM-0072- (CON NORT ZAC161/PHB47)/PHB47 #004-(2n)-004 PI 515577 S BGEM-0073- (CON NORT ZAC161/PHB47)/PHB47 #005-(2n)-001 PI 515577 S BGEM-0074- (CON NORT ZAC161/PHB47)/PHB47 #005-(2n)-003 PI 515577 S BGEM-0077- (CRISTALINO AMAR AR21004/PHB47)/PHB47 #001- PI 516163 S (2n)-001 BGEM-0078- (CRISTALINO AMAR AR21004/PHB47)/PHB47 #001- PI 516163 S (2n)-002 BGEM-0079- (CRISTALINO AMAR AR21004/PHB47)/PHB47 #005- PI 516163 S (2n)-003 BGEM-0080- (CRISTALINO AMAR AR21004/PHB47)/PHB47 #005- PI 516163 S (2n)-004 BGEM-0082- (CRISTALINO AMAR AR21004/PHB47)/PHB47 #007- PI 516163 S (2n)-002 BGEM-0083- ((Cubano dentado - BOV 585/PHB47 B)/PHB47)-(2n)-001 PI 485383 S

210

Table S3 continued BGEM-0085- (CUZCO CUZ217/PHZ51)/PHZ51 #002-(2n)-001 PI 485274 N BGEM-0087- (DULCILLO DEL NO SON57/PHZ51)/PHZ51 #001-(2n)- PI 490973 N 001 BGEM-0088- (DULCILLO DEL NO SON57/PHZ51)/PHZ51 #002-(2n)- PI 490973 N 002 BGEM-0089- (DULCILLO DEL NO SON57/PHZ51)/PHZ51 #002-(2n)- PI 490973 N 003 BGEM-0090- (DULCILLO DEL NO SON57/PHZ51)/PHZ51 #002-(2n)- PI 490973 N 004 BGEM-0091- (DULCILLO DEL NO SON57/PHZ51)/PHZ51 #003-(2n)- PI 490973 N 002 BGEM-0094- (DULCILLO DE NO SON57/PHB47)/PHB47 #005-(2n)- PI 490973 S 003 BGEM-0095- (DULCILLO DE NO SON57/PHB47)/PHB47 #006-(2n)- PI 490973 S 001 BGEM-0097- (EARLY CARIBBEAN MAR9/PHB47)/PHB47 #001-(2n)- Ames 28579 S 002 BGEM-0099- (EARLY CARIBBEAN MAR9/PHB47)/PHB47 #006-(2n)- Ames 28579 S 002 BGEM-0100- (EARLY CARIBBEAN MAR9/PHB47)/PHB47 #007-(2n)- Ames 28579 S 001 BGEM-0102- (ELOTES OCCIDENT NAY29/PHZ51)/PHZ51 #002-(2n)- PI 484828 N 001 BGEM-0107- (ELOTES OCCIDENT NAY29/PHZ51)/PHZ51 #004-(2n)- PI 484828 N 004 BGEM-0108- (ELOTES OCCIDENT NAY29/PHB47)/PHB47 #002-(2n)- PI 484828 S 001 BGEM-0110- (ELOTES OCCIDENT DGO236/PHZ51)/PHZ51 #007-(2n)- PI 484828 N 001 BGEM-0111- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-001 BGEM-0112- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-002 BGEM-0113- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-003 BGEM-0114- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-004 BGEM-0115- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-005 BGEM-0116- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-006 BGEM-0117- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-007

211

Table S3 continued BGEM-0118- (B47/GORDO [CHH131]{CIMYT})-B-B-SIB-011-001-001- PI 484406 S (2n)-008 BGEM-0120- ((Jora - ANC 1/PHZ51 B)/PHZ51)-(2n)-001 PI 571477 N BGEM-0121- ((Jora - ANC 1/PHZ51 B)/PHZ51)-(2n)-002 PI 571477 N BGEM-0122- ((Jora - ANC 1/PHZ51 B)/PHZ51)-(2n)-003 PI 571477 N BGEM-0123- (KARAPAMPA BOV978/PHZ51)/PHZ51 #001-(2n)-001 NSL 286824 N BGEM-0124- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-001 NSL 286824 N BGEM-0125- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-002 NSL 286824 N BGEM-0126- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-003 NSL 286824 N BGEM-0127- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-004 NSL 286824 N BGEM-0129- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-006 NSL 286824 N BGEM-0130- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-007 NSL 286824 N BGEM-0131- (KARAPAMPA BOV978/PHZ51)/PHZ51 #003-(2n)-008 NSL 286824 N BGEM-0133- (KARAPAMPA BOV978/PHB47)/PHB47 #003-(2n)-001 NSL 286824 S BGEM-0134- (KARAPAMPA BOV978/PHB47)/PHB47 #004-(2n)-001 NSL 286824 S BGEM-0136- (KARAPAMPA BOV978/PHB47)/PHB47 #005-(2n)-002 NSL 286824 S BGEM-0137- (KARAPAMPA BOV978/PHB47)/PHB47 #005-(2n)-003 NSL 286824 S BGEM-0138- (KARAPAMPA BOV978/PHB47)/PHB47 #005-(2n)-005 NSL 286824 S BGEM-0140- (PHZ51/KCELO BOV948)/PHZ51 #001-(2n)-002 Ames 28740 N BGEM-0141- (PHZ51/KCELO BOV948)/PHZ51 #001-(2n)-005 Ames 28740 N BGEM-0143- (PHZ51/KCELO BOV948)/PHZ51 #003-(2n)-001 Ames 28740 N BGEM-0144- (PHZ51/KCELO BOV948)/PHZ51 #003-(2n)-002 Ames 28740 N BGEM-0145- (PHZ51/KCELO BOV948)/PHZ51 #004-(2n)-001 Ames 28740 N

212

Table S3 continued BGEM-0146- (PHZ51/KCELO BOV948)/PHZ51 #004-(2n)-002 Ames 28740 N BGEM-0147- (PHB47/KCELO BOV948)/PHB47 #001-(2n)-001 Ames 28740 S BGEM-0149- (PHB47/KCELO BOV948)/PHB47 #001-(2n)-003 Ames 28740 S BGEM-0150- (PHB47/KCELO BOV948)/PHB47 #003-(2n)-002 Ames 28740 S BGEM-0151- (PHZ51/MISHC ECU321)/PHZ51 #004-(2n)-002 PI 488016 N BGEM-0152- (PHZ51/MISHC ECU321)/PHZ51 #004-(2n)-003 PI 488016 N BGEM-0153- (PHB47/MISHCA ECU321)/PHB47 #002-(2n)-001 PI 488016 S BGEM-0154- (PHB47/MISHCA ECU321)/PHB47 #003-(2n)-001 PI 488016 S BGEM-0155- (PHB47/MISHCA ECU321)/PHB47 #004-(2n)-001 PI 488016 S BGEM-0156- (PHB47/MISHCA ECU321)/PHB47 #004-(2n)-002 PI 488016 S BGEM-0157- CUBA164:N(PHZ51)(PHZ51)-(2n)-002 PI 489361 N BGEM-0158- (MONTANA NAR625/PHB47)/PHB47 #003-(2n)-002 PI 445252 S BGEM-0159- (MONTANA NAR625/PHB47)/PHB47 #003-(2n)-003 PI 445252 S BGEM-0160- (MONTANA NAR625/PHB47)/PHB47 #006-(2n)-001 PI 445252 S BGEM-0161- (MONTANA NAR625/PHB47)/PHB47 #006-(2n)-003 PI 445252 S BGEM-0162- (MORADO BOV567/PHB47)/PHB47 #002-(2n)-001 PI 485373 S BGEM-0163- (MORADO BOV567/PHB47)/PHB47 #003-(2n)-001 PI 485373 S BGEM-0164- (MORADO BOV567/PHB47)/PHB47 #003-(2n)-002 PI 485373 S BGEM-0165- (MORADO BOV567/PHB47)/PHB47 #003-(2n)-003 PI 485373 S BGEM-0166- (MORADO BOV567/PHB47)/PHB47 #005-(2n)-002 PI 485373 S BGEM-0167- (MORADO BOV567/PHB47)/PHB47 #005-(2n)-003 PI 485373 S BGEM-0169- (MORADO BOV567/PHB47)/PHB47 #006-(2n)-002 PI 485373 S

213

Table S3 continued BGEM-0170- (MORADO BOV567/PHB47)/PHB47 #006-(2n)-003 PI 485373 S BGEM-0172- (MORADO CANTENO LIM55/PHB47)/PHB47 #002-(2n)- PI 515026 S 002 BGEM-0173- (MORADO CANTENO LIM55/PHB47)/PHB47 #002-(2n)- PI 515026 S 004 BGEM-0174- (PHZ51/(PHZ51/MOROCHO APC67) #001-(2n))-001 PI 571413 N BGEM-0175- (MOROCHO APUC77/PHB47)/PHB47 #003-(2n)-001 PI 503511 S BGEM-0178- (B47/(B47/OLOTON [GUA383]{CIMYT}))-SIB-005-001- Ames 28539 S 001-(2n)-001 BGEM-0179- (B47/(B47/OLOTON [GUA383]{CIMYT}))-SIB-005-001- Ames 28539 S 001-(2n)-003 BGEM-0180- (B47/(B47/OLOTON [GUA383]{CIMYT}))-SIB-007-001- Ames 28539 S 001-(2n)-001 BGEM-0182- (ONAVENO SON24/PHZ51)/PHZ51 #004-(2n)-001 PI 484880 N BGEM-0184- (ONAVENO SON24/PHZ51)/PHZ51 #005-(2n)-002 PI 484880 N BGEM-0186- (ONAVENO SON24/PHB47)/PHB47 #002-(2n)-001 PI 484880 S BGEM-0187- (ONAVENO SON24/PHB47)/PHB47 #003-(2n)-001 PI 484880 S BGEM-0188- (ONAVENO SON24/PHB47)/PHB47 #004-(2n)-001 PI 484880 S BGEM-0190- ((Patillo - ECU 417/PHZ51)/PHZ51)-(2n)-002 PI 488039 N BGEM-0191- (PATILLO BOV493/PHZ51)/PHZ51 #001-(2n)-001 Ames 28736 N BGEM-0192- (PATILLO BOV493/PHZ51)/PHZ51 #003-(2n)-001 Ames 28736 N BGEM-0193- (PATILLO BOV493/PHZ51)/PHZ51 #003-(2n)-002 Ames 28736 N BGEM-0196- (PATILLO GRANDE BOV649/PHZ51)/PHZ51 #006-(2n)- Ames 28748 N 003 BGEM-0197- (PATILLO GRANDE BOV649/PHB47)/PHB47 #002-(2n)- Ames 28748 S 001 BGEM-0198- (PATILLO GRANDE BOV649/PHB47)/PHB47 #003-(2n)- Ames 28748 S 001 BGEM-0199- (PATILLO GRANDE BOV649/PHB47)/PHB47 #006-(2n)- Ames 28748 S 001 BGEM-0200- (PHB47/PERLA ANC23)/PHB47 #002-(2n)-001 PI 571479 S

214

Table S3 continued BGEM-0201- (PIRA TOL405/PHZ51)/PHZ51 #001-(2n)-001 PI 445528 N BGEM-0202- (PIRA TOL405/PHZ51)/PHZ51 #002-(2n)-001 PI 445528 N BGEM-0203- (PHZ51/PISAN BOV344)/PHZ51 #001-(2n)-002 NSL 286568 N BGEM-0204- (PHZ51/PISAN BOV344)/PHZ51 #001-(2n)-003 NSL 286568 N BGEM-0205- (PHZ51/PISAN BOV344)/PHZ51 #001-(2n)-004 NSL 286568 N BGEM-0206- (PHZ51/PISAN BOV344)/PHZ51 #001-(2n)-005 NSL 286568 N BGEM-0207- (PHZ51/PISAN BOV344)/PHZ51 #001-(2n)-006 NSL 286568 N BGEM-0209- (PHZ51/PISAN BOV344)/PHZ51 #002-(2n)-002 NSL 286568 N BGEM-0210- (PHZ51/PISAN BOV344)/PHZ51 #003-(2n)-001 NSL 286568 N BGEM-0211- (PHZ51/PISAN BOV344)/PHZ51 #003-(2n)-002 NSL 286568 N BGEM-0212- (PHZ51/PISAN BOV344)/PHZ51 #005-(2n)-001 NSL 286568 N BGEM-0213- (PHB47/PISAN BOV344)/PHB47 #003-(2n)-001 NSL 286568 S BGEM-0215- (POJ CHICO BOV800/PHZ51)/PHZ51 #001-(2n)-001 NSL 286760 N BGEM-0216- (POJ CHICO BOV800/PHZ51)/PHZ51 #006-(2n)-001 NSL 286760 N BGEM-0218- (POJ CHICO BOV800/PHB47)/PHB47 #002-(2n)-001 NSL 286760 S BGEM-0220- (POJ CHICO BOV800/PHB47)/PHB47 #006-(2n)-004 NSL 286760 S BGEM-0221- (POJ CHICO BOV800/PHB47)/PHB47 #006-(2n)-005 NSL 286760 S BGEM-0222- (POJ CHICO BOV800/PHB47)/PHB47 #006-(2n)-006 NSL 286760 S BGEM-0223- (SG HUANCAV JUN164/PHZ51)/PHZ51 #005-(2n)-001 PI 503711 N BGEM-0224- (SG HUANCAV JUN164/PHZ51)/PHZ51 #005-(2n)-002 PI 503711 N BGEM-0225- (SG HUANCAV JUN164/PHZ51)/PHZ51 #005-(2n)-003 PI 503711 N BGEM-0226- ((Semi dentado paulista - PAG I/PHB47 B)/PHB47)-(2n)- PI 449576 S 002

215

Table S3 continued BGEM-0227- BR105:N(PHZ51)(PHZ51)-(2n)-001 Ames 26251 N BGEM-0228- BR105:N(PHZ51)(PHZ51)-(2n)-003 Ames 26251 N BGEM-0233- ((Tehua - CHS29/PHB47 B)/PHB47)-(2n)-003 Ames 29075 S BGEM-0235- ((Tuxpeño - GUA 456/PHZ51 B)/PHZ51)-(2n)-002 Ames 28567 N BGEM-0236- ((Vandeño - GRO 96/PHB47 B)/PHB47)-(2n)-001 Ames 28466 S BGEM-0237- (YUCATAN TOL389 ICA/PHZ51)/PHZ51 #003-(2n)-001 PI 445514 N BGEM-0239- (YUCATAN TOL389 ICA/PHZ51)/PHZ51 #005-(2n)-001 PI 445514 N BGEM-0240- (YUCATAN TOL389 ICA/PHZ51)/PHZ51 #005-(2n)-002 PI 445514 N BGEM-0242- (YUCATAN TOL389 ICA/PHZ51)/PHZ51 #005-(2n)-005 PI 445514 N BGEM-0243- (YUCATAN TOL389 ICA/PHB47)/PHB47 #002-(2n)-001 PI 445514 S BGEM-0244- (YUCATAN TOL389 ICA/PHB47)/PHB47 #003-(2n)-001 PI 445514 S BGEM-0247- (YUNGUENO BOV362/PHZ51)/PHZ51 #005-(2n)-002 NSL 286578 N BGEM-0248- (YUNGUENO BOV362/PHZ51)/PHZ51 #006-(2n)-002 NSL 286578 N BGEM-0250- (YUNGUENO BOV362/PHB47)/PHB47 #001-(2n)-001 NSL 286578 S BGEM-0252- (YUNQUILANO F AND ECU710/PHB47)/PHB47 #007- PI 485436 S (2n)-001 BGEM-0253- ((Avati Moroti Guapi - PAG 139/PHZ51 B)/PHZ51)-(2n)- PI 485458 N 001 BGEM-0254- ((Chandelle - VEN 409/PHB47 B)/PHB47)-(2n)-001 PI 488529 S BGEM-0255- ((Chirimito - VEN 703/PHB47 B)/PHB47)-(2n)-001 PI 504168 S BGEM-0256- ((Chirimito - VEN 703/PHZ51 B)/PHZ51)-(2n)-001 PI 504168 N BGEM-0257- ((Culli - ARG 471/PHB47 B)/PHB47)-(2n)-001 Ames 28799 S BGEM-0258- ((Culli - ARG 471/PHB47 B)/PHB47)-(2n)-002 Ames 28799 S BGEM-0259- ((PHZ51/Jala - JAL 44)/PHZ51)-(2n)-001 PI 484704 N

216

Table S3 continued BGEM-0260- ((Puya - MAG 355/PHZ51 B)/PHZ51)-(2n)-001 PI 444923 N BGEM-0261- ((Tepecintle - GUA 65/PHB47 B)/PHB47)-(2n)-001 Ames 29169 S BGEM-0262- ((Tuxpeño Norteño - CHH287/PHB47 B)/PHB47)-(2n)-001 PI 629142 S BGEM-0263- ((Tuxpeño Norteño - CHH287/PHB47 B)/PHB47)-(2n)-002 PI 629142 S BGEM-0264- ((Tuxpeño Norteño - CHH287/PHB47 B)/PHB47)-(2n)-005 PI 629142 S BGEM-0265- ((Capio rosado - ARG 460/PHB47 B)/PHB47)-(2n)-001 Ames 28794 S BGEM-0266- ((Cateto Nortista - GIN I/PHB47 B)/PHB47)-(2n)-003 PI 449478 S BGEM-0268- ((Capio rosado - ARG 460/PHB47 B)/PHB47)-(2n)-002 Ames 28794 S BGEM-0269- ((Cateto Nortista - GIN I/PHB47 B)/PHB47)-(2n)-002 PI 449478 S BGEM-0270- ((Chandelle - VEN 409/PHB47 B)/PHB47)-(2n)-002 PI 488529 S BGEM-0272- ((PHB47/Tuxpeño - VER 143)/PHB47)-(2n)-002 Ames 28567 S

217

CHAPTER 6. GENERAL CONCLUSIONS

BRs and GAs have shown great potential to be used as plant growth regulators in different crop species, and their biosynthesis and signaling pathways have been well established in model species rice and Arabidopsis. However, their mechanism in maize is relatively limited especially BR. Despite BRs’ broad effects on the physiological and developmental processes of plants, they were not widely recognized as plant hormones until the mid-1990s. In this study, we focused on BR and GA control of maize PHT from genetics and plant physiology perspectives, as PHT is significantly associated with maize grain yield and biomass production. To do this, we established two backcross libraries (phenotype-selected introgression libraries, representing heterozygotes) and two doubled haploid (DH) libraries (representing inbred background). We genotyped and phenotyped these populations and conducted genome-wide association studies to find genome-wide SNPs associated with PHT, BR, and GA and explore co-localizations. At the same time, we applied BR and GA inhibitors to these populations, to compare the BR and GA inhibitor response of heterozygotes and inbred lines, and to connect seedling stage hormone inhibitor response to field PHT.

In chapter two, we proposed a “MOST” concept to improve crop performance from a non-genetics perspective. We highlighted MOST as a third important route for utilizing detailed molecular genomic information generated from current large-scale sequencing and -omics projects, in additional to MAS and transgenics methods. In European countries, transgenics methods are not accepted and MAS cannot be utilized if there is no natural variation for a desired

218 trait. If there is prior knowledge underlying a desirable trait in other species, MOST treatments may be used to trigger specific metabolic pathways, which may lead to a target phenotype.

Moreover, MOST treatments can be used for maximizing trait expression after an elite cultivar is produced by MAS or transgenic approaches. With detailed BR and GA pathway established in maize, this knowledge can be used to develop effectors to address key proponents in BR or GA pathway that regulate trait expression to improve crop performance.

In chapter three, we developed, characterized and evaluated two maize two maize phenotypic-selected introgression libraries by introgressing donor chromosome segments (DCS) from Germplasm Enhancement of Maize (GEM) accessions into elite inbred lines: PHB47 and

PHZ51. We found that our phenotype selection was efficient, as most phenotype-selected introgression families (PIFs) carrying DCS were significantly (α= 0.01) taller than the respective recurrent parent. In addition, they contained larger donor genome proportions than expected in the absence of selection or random mating across all BC generations. The DCS were distributed over the whole genome, indicating a complex genetic nature underlying PHT. We found that our

PIFs were enriched for favorable PHT-increasing alleles, with normal flowering time at temperate area, and we found 12 hot spots overlapped with previous identified PHT QTL regions.

These two libraries offer opportunities for future PHT gene isolation and allele characterization and for breeding purposes, such as novel cultivars for biofuel production. Compared with the traditional introgression libraries, the use of PIFs has advantages: (i) low cost phenotypic selection, (ii) multiple donors and (iii) focus on only one trait to accumulate respective genomic regions. However, this approach of using PIFs may only be useful for highly heritable traits and

219 a limited use for only one trait – PIFs may not be suitable for complex traits that are more difficult to phenotype with lower heritability.

In chapter four, we genotyped the two phenotype-selected introgression libraries with

3072 genome wide SNPs. In addition, we genotyped (with 8,000 SNPs) and phenotyped two doubled haploid (DH) libraries as comparison. We found that on average, PIFs contained larger donor genome composition compared to DH lines because of phenotype selection, and PIFB47 has a stronger phenotype selection due to a further relatedness between Stiff Stalk (includes

PHB47) and tropical germplasm. We conducted genome wide association studies and found 73 genomic loci associated with PHT, with 7 co-localized with GA candidate genes and two very closed to BR candidate genes, and this number of co-localization is significantly different from random occurrence. We concluded that BR and GA had a great impact for PHT increase in maize. From a plant physiology perspective, we applied BR and GA inhibitors to all PIFs and

DH lines, and found that field PHT was significantly correlated with seedling stage BR and GA inhibitor responses. However, this observation was only valid for PIFs, but not for inbred lines.

Our path analysis results demonstrated that higher level of heterozygosity led to elevated GA activities. Increased GA promoted BR level, and both increase heterosis in maize PHT. To our knowledge, this is the first study in maize to connect BR and GA with heterosis for PHT.

In chapter five, we genotyped the two DH libraries (207 DH lines) with 62,049 GBS markers, and measured BR and GA inhibitor responses for mesocotyl length, seedling length, and primary root length. In addition, we measured plant height, yield and flowering time for

220 these 207 DH lines. We conducted genome-wide association studies with three models MLM,

GLM+Q and FarmCPU, and found 3, 19, 134 SNPs significantly associated with different seedling traits, with gene GRMZM2G013391 associated with GA inhibitor response for mesocotyl length across the three models. This gene was found to regulate GA signaling in rice.

With these three models, we were able to balance the false positives and false negatives for

GWAS analysis. We noticed that seedling stage hormone inhibitor response can be used to predict field performance yield and flowering time. Genotypes with increased BR and GA inhibitor response were with lower yield, which may due to longer anthesis-silking interval.

The studies described herein were a first step to understand the BR and GA control of

PHT in maize. The populations used in this PhD research were from the crosses between tropical germplasm and temperate maize lines, thus, other types of populations are needed to further verify our conclusions. We only measured PHT, yield, flowering time and seedling traits for the

DH lines, it will be interesting to know if seedling BR/GA inhibitor response can be correlated with other important agronomic traits such as leaf angle, and biotic/abiotic stress tolerance in maize. In addition, since other plant hormones such as auxin was shown to regulate PHT, it will be beneficial to incorporate other plant hormones, and investigate how they are controlling different traits in maize.