Plant Sciences Institute

Department of Agronomy

Raymond F. Baker Center for Plant Breeding

 

Crop Genetic Improvement and Adaptation UsingGene Discovery, Phenotypic Prediction, and Systems Engineering

 


Justification and Impact. 

Thegoals of this umbrella project include demonstrations of 1) how to rapidly deploy adaptive agronomic traits and 2) how to transform plant breeding from an art into an engineering discipline. This will be accomplished by meeting the following specific objectives: a) discovery of alleles associated with adaptive agronomic traits, b) development of predictive models for adaptive agronomic traits, c) development of designs that assure optimal genetic improvement strategies and d) implementation of accelerated genetic improvement strategies for adaptive agronomic traits
Genetic improvement exhibited by many crops during the last 70 years has been due primarily to increased capacity to evaluate larger samples of segregating generations through mechanization of planting, cultivation and harvest, logistics of seed transfer between northern and southern hemispheres and computational infrastructure. In the 1940’s a breeding program might evaluate hundreds of lines to identify a few superior cultivars over a period of ten to fifteen years. Today commercial maize and soybean breeding programs evaluate millions of progeny annually to identify superior cultivars in half as much time.
Plant breeder’s evaluations of phenotypes have been based on the assumption that performance in breeder’s field trials will translate intoequivalent future performance in farmer’s fields. Such an assumption is reasonable as long as the two types of environments are similar. Twenty-five years ago climate modelers stated that the Midwest had experienced very stable environments for at least a century, but we should expect variable environmental conditions in the near future. Their predictions were conservative relative to the actual fluctuations experienced during the last 10 years. Since 2005 there have been floods and droughts as well as excessive heat, unusual cold, and extended and shortened growing seasons, often within a single year. As a consequence, apparent genetic improvements in breeder’s trials have been associated with little change inaverage on-farm production of corn and soybeans in Iowa for the years 2004 - 2013. 
The premise of this umbrella project is that genetic improvements in the future will require identification and rapid deployment of alleles enabling crops topredictably adapt to a broad range of environments. Adaptation refers to the inheritance of plasticity in growth and development, also known as reaction norms. Such plasticity provides resilience to unstable environmental conditions.Thus, adaptive agronomic traits refer to characteristics that enable crops to successfully produce food, feed or fuel under unstable environmental conditions and changing management practices.

 

Previous Work and Present Outlook


a. Discover alleles associated with adaptive agronomic traits.
Methods to identify alleles associated with agronomic traitsbased on an assumption of additive genetic architectureshave been routinely applied for at least 20 years. An Adaptive Agronomic Trait (AAT) represents any trait that assures the farmer will produce harvestable yield on a per acre basis. If a crop cultivar does not produce a large quantity of the product for which it is grown, then there is no financial incentive to grow the cultivar.
The application of ‘omics’ technologies has revealed that measured phenotypes are products of genetic regulation of networks consisting of many genes that respond to both developmental and environmental signals. This contrasts with results from applications of statistical genetics that reveal the primary contributions to genetic variability are additive.
Cheverud and Routman (1995)demonstrated that the inconsistencies could be due to statistical genetic approaches that attribute most non-additive genetic interactions to additive genetic variability, i.e., the statistical genetic approaches are inherently biased. In 2014 Mackay reviewed the large literature on genetic architecture studies and found that every carefully studied quantitative trait has demonstrated ephemeral additive genetic variability based on actual underlying epistatic genetic architectures. To date, statistical genetic models that have overcome this inherent bias have not been developed.
Even though agronomic traits have long been recognized as products of genetic regulation and environmental influences throughout phases of growth and development, the current practice is to evaluate agronomic traits at a single time, usually after the individuals reach physiological maturity. Under relatively stable environmental conditions it makes sense for plant breeders to evaluate agronomic traits at a single time point or developmental stage. If the environmental influences are consistent from one growing season to the next, then it is reasonable to expect cultivars with the most desirable phenotypes in one season will be cultivars with the most desirable phenotypes in the next season. In the future, environmental stresses will affect growth and development of genotypic lines inconsistently from year to year and across regions within years. The result will be inconsistent phenotypic performance due to cross-over genotype by environment (GxE) interactions. An interesting and testable hypothesisis that phenotypic evaluations conducted throughout phases of growth and development will enable identification ofgenetic alleles responsible for GxE. A related testable hypothesis is that epistasis is an artifact of trait evaluations at a single time point. If either hypothesis is true, it should be possible to decompose these interactions into tractable (possibly additive) components through analysis of data from traits that are continuously evaluated throughout growth and development.
Phenotypic progression of growth and development of a genotype is known as its reaction norm.There exist reaction norms that transcend the entire life cycle as well as within specific phases such as pollination, seed development, germination, emergence, seedling vigor, etc. Growth and development of many traits, including height, flowering time, leaf angle, stem circumference, tillers, panicle length, panicle exertion, and number of stem internodes are all recognized to be components of yield for biomass and grain crops. We also know variable responses of these morphological traits are mediated by concentrations of hormones, e.g., auxins, brassinosteroids, carotenoids, gibberellins, that are in turn genetically regulated through cell signaling pathways.Variability in reaction norms among genotypes suggests potential for genotypic resilience through flexible responses, i.e., adaptation to variable environmental stressors. For example, adaptive plasticity will be needed for consistent germination and seedling vigor under both cool wet conditions and hot, dry conditions. In order to resolve genetic factors responsible for differential growth and development we will have to phenotype thousands of genotypic lines throughout their life cycle under field conditions.
High throughput phenotypingis also known as ‘phenomics’. The development of phenomic technologies requires interdisciplinary approaches and is receiving significant attention from funding agencies and commercial breeding organizations. The challenges facing the development of field-based phenomics include capturing high-resolution multi-spectra images under non-ideal field conditions,as well as image storage, processing and data reduction. Preliminary development of robots to capture phenotypes in field plots throughout the growing season has received national recognition.Similar drone-based technologies will likely become sufficiently well developed for routine acquisition of field based phenomic data during the period of this umbrella project.
Finding association between segregating alleles and estimatedfunctions for reaction norms will identify candidate genes associated with transitions between stages of growth and development, thus providing genetic insights into cell signaling of developmental physiology as a response to environment(Yin and Struik, 2008) The principles of this approach have been successfully applied to flowering phenology of barley(Yin, Struik, et al., 2005), maize leaf growth as a function of heat units(Reymond, Muller, et al., 2003) and to explore the non-linear genetic regulation of biomass accumulation and water content in developing maize kernels (Meade, unpublished dissertation).
With the general prediction of highly variable environments, we also anticipate some specific environmental changes that will require identification of alleles needed forAATsin the upper Midwest. Examples include heat shock tolerance, photo-protection, nitrogen use efficiency (NUE), and water use efficiency. Currently, very little is known about the underlying genetic architectures of these AATs, although preliminary linkage and genome wide association studies have been conducted on NUE in maize.
In addition to emergence of new abiotic stressors on growth and development, diseases and pests that have historically been confined to tropical and semi-tropical environments will challenge crop productivity in high lattitudes. Unlike identification of alleles for tolerance to abiotic stress, there has been considerable experience in identifying resistance alleles for many diseases and pests. Resistance alleles have been identified, mapped and validated through routine application of forward and reverse genetic approaches as well as routine bioinformatic approaches in crops with sequenced genomes. Most disease and pest resistance traits consist of single gene or simple oligogenic architectures. While resistance alleles have been identified for many existing diseases and pests, the diseases and pests have the ability to rapidly evolve and pose emerging threats that are capable of overwhelming all known resistance alleles. A specific example that soybean farmers in the Midwest are concerned about is the Soybean Cyst Nematode (SCN). Soybean growers know that there is only one known source of resistance that transcends the most prevalent SCN races. Soybean producers have asked the North Central Soybean Regional Program (NCSRP) to focus its funding on finding novel resistance alleles for a broad spectrum of races of SCN.
b. Develop predictive models for adaptive agronomic traits.
Biology is experiencing a Kuhnian Scientific Revolution. The most striking feature of the revolution is a change in how hypotheses are generated. Historically, biological hypotheses have been generated through observation of natural and experimental phenomena. Biologists would make discoveries such as described in objective a using this approach; Ernst Rutherford referred to this approach as akin to collecting postage stamps. In the future, hypotheses will be generated from predictions based on mathematical models of complex systems.For plant breeders the paradigm shift will result in predictions (hypotheses) from mathematical models of complex biological and breeding processes. For example, simulation modeling predicted that additive genetic variance (a component of heritability) is an emergent, and ephemeral, property of non-additive, i.e., epistatic genetic architectures(Cheverud and Routman, 1995)
In the future, models of complex AATs and plant breeding processes will consist of multiple linear and non-linear relationships. Thus, thousands to millions of simulated outcomes will be computationally explored to generate hypotheses and help experimentalists make decisions about the most appropriate experimental and breeding designs to employ. Currently the state of predictive modeling for complex systems include stochastic rather than deterministic models. While the emergence of phenomic technologies will provide insights to the genetic regulation of physiological processes, deterministic models for complex systems are not anticipatedto provide more accurate predictions than stochastic models for at least 20 years.
Genomic Prediction.About 15 years ago, genomic technologies enabled quantitative and population geneticists to develop genome-wide prediction (GP) methods. Initial assessments of GP based on simulations of plant breeding populations were encouraging(Yu J, Pressoir G, et al., 2006). Some even suggested that after a GP model was developed, there would be no further need for expensive field trials (Heffner, Sorrells, et al., 2009).GP methods now consist of about 20 statistical procedures that produce predictive models based on associations betweengenomic marker and phenotypic data. Based on the objective criterion of accuracy and use of simulated additive and non-additive traits,(Howard, Carriquiry, et al., 2014) evaluated 14 GP methods and found that they are all equivalent and reasonably good for traits withadditive genetic architectures. However, if the underlying genetic architecture of the traits consistsonly of epistasis, the predictions using parametric methods are inaccurate, even for highly heritable traits.In contrast, non-parametric methods provide reasonably accurate predictions. Thus, providing a diagnostic set of evaluation analyses.
Based on experimental data, it is now known that GP models need to be ‘retrained’ each cycle of breeding with both genotypic and phenotypic evaluations (Sebastian et al, 2010). Even within breeding cycles, GP has not lived up to the hyperbole. In a soon to be published manuscript, Lian et. al. (personal communication) evaluated 969 segregating maize hybrid populations using RRBLUP, a parametric GP method, and found that prediction accuracies in almost 1/3 of the populations were close to zero or negative. Was this poor outcome due to application of a parametric GP method that ignores epistasis and GxE?
Despite the assessments of more limited application,GP represents a significant contribution to the transformation of biology from a descriptive to a predictive science. Current applications of GP methods are enabling plant breeders to evaluate many more lines than they can afford to grow in field plots. Also, for some traitsthe time required to evaluate segregating lines canbe reduced by 20% before using the lines as parents to create a new generation of segregating populations. GP methods need to be developed for application to reaction norms of AATs, consisting of epistatic genetic architectures that respond differentially to environmental signals. Further there is a need for GP methods that not only accurately predict reaction norms within generations, but also accurately predict which lines are going to be the best parents for the next generation and subsequent generations; many high performing lines do not produce high performing progeny.
Integrated knowledge. In the long term, systems biology and comparative genomics coupled with field-based phenomics is going to produce knowledge of genetic regulatory and metabolic networks as well as associations with crop physiology and production systems. Such knowledge will produce the need for an integrated modeling approach for performance prediction. Fundamentally, the models will be hierarchical and consist of component complex models. The naïve approach will be to integrate inputs and outputs among the models for predictions based on simulations using massive parallel computing.
Applications of information theory, as well as experimental evidence about cell to cell communication have revealed that the amount of information that is shared among members within and between hierarchical levels of organization do not need to be comprehensive.Agent Based Models (ABM)may provide a means for developing very accurate predictive models for AAT systems. ABM provides a means of exploring how various entities from molecules to communities interact with each other while assessing and predicting performance of the whole system. ABM utilizes game theory, evolutionary dynamics, complexity theory and emergent systems to simulate the simultaneous interactions of multiple models in an attempt to predict the appearance of complex phenomena.

c. Utilize genetic discoveries and predictions to DESIGN optimal accelerated genetic improvement strategies.
Genetic improvement by plant breeding is a simple iterative process: create useful genetic variability by crossing distinct genotypes, identify and select the most desirable variant progeny and use these to create the next generation. For millennia, plant breeders have been implementing this process, constrained by reproductive biology of the species. During the last 70 years, cycles through the iterative process have accelerated and selection intensities have increased through integration of agricultural, information and bio-technologies. As a consequence, current plant breeding systems are diverse and consist of idiosyncratic activities based on historical artifacts unique to each breeding organization. However, it is not known if existing breeding programs are sustainable or optimal (Ted Crosbie, Monsanto; Ray Riley, Syngenta; Thomas Connelley, DuPont; personal communications), i.e., whether the probabilities of meeting objectives are maximized while minimizing costs and time.
As previously noted, genetic improvements in the future will require rapid deployment of predictable AATs enabling crops to produce food, fuel and fiber in variableenvironments.With emergence of climate change and 9 billion people by 2050, genetic improvement no longer can afford the luxury of ad hocdecisions or trial and error approaches. In the near future it will be essential to implement the tools of systems engineering to assure that plant breeding systems are optimized and accelerate the breeding cycles through development of doubled haploid technologies for all crops and in vitro nurseries.
Systems Engineering. Despite the idiosyncrasies associated with breeding different crops, plant breeding systems are fundamentally decision systems;they can be designed to be optimal using the same engineering principles that are used to optimize manufacturing, industrial and financial systems.. It is possible to optimize multi-objective projects by maximizingprobability of success while minimizing costs and time(Xu, Wang, et al., 2011).Implementing engineering design principles requires the plant breeder to clearly define his/her objectives, biological and cost constraints, then translate these into mathematical relationships known as an objective function. Optimal solutions to most classes of objective functions can be found with existing algorithms that have been implemented in software packages such as CPLEX. The challenge is for plant breeders to work with systems engineers to learn how to clearly define objectives and translate these into mathematical functions.
Doubled Haploids. In the last ten years an older discovery, di-haploids, has been implementedfor maize.Doubled haploids (DH) have enabled maize breeders to skip many generations needed to produce ‘true breeding’ lines for use in hybrid combinations. The current DH challenge in maize is to optimize the process, i.e., minimize the cost while maximizing the production of DH lines. For several important crop species such as sorghum, soybean and sunflower there are no DH system currently available. To accelerate the breeding programs for these crops it will be desirable to develop DH methods.
In vitro nurseries.Despite improvements from mechanization, logistics, informatics, systems engineering and doubled haploids, the length of plant life cycles (seed to seed) remains the primary constraint on cultivar development and genetic improvement. In vitro nurseries are a recentconcept for accelerating genetic improvement(De La Fuente, Frei, et al., 2013).Briefly, the concept is to maintain selected gametes or zygotes in cell cultures. New genotypic zygotes are formed by in vitro fusion of gametes. Each cycle of gamete and genotypic formation can be repeated multiple times before mature plants are regenerated. Such an in vitro system has the potential to immediately induce gametes for new crosses, or for genome doubling to produce homozygous cell lines. Selection within anin vitro nursery would be accomplished by genotyping gametes and application of GP or gene stacking to enabling targeted mating and resulting in desirableDH lines with significant reductions in time and resources.
d. Implement accelerated genetic improvement strategies for adaptive agronomic traits.
This objective is focused on translation of the experimental discoveries and theoretical developments represented in objectives a, b and c. As noted by (Bernardo, 2008), most experimental discoveries of useful alleles, by academic plant breeders, reside in journals on shelves of libraries. The same critique of GP and optimization of plant breeding programs also applies. Commercial corn and soybean breeders would not agree with this critique, but for the vast majority of crops the critique is justified. The primary reason for failure to translate is lack of resources (funding) for translational research by funding agencies that support academic research.


Procedures


Herein we describe infrastructure and approaches that will be needed to successfully demonstrate “principles of rapid identification and deployment of adaptive agronomic traits and transformation of plant breeding into an engineering discipline”. If successful, the greatest impact will be acollaborative faculty capable ofestablishing infrastructure that will enable them and future faculties to rapidly adapt to changing funding environments. We recognize that most of our research projects are supported by extramural funding, thus our approaches reflect this pragmatism. For example, we have not pre-defined which Adaptive Agronomic Traits (AATs)to pursue. Rather we will be responsive to community prioritized AATs as expressed through funding organizations at NSF, USDA-NIFA, various private foundations, various commodity boards, and commercial seed organizations. Some of our currently funded AATs include grain yield in soybean, biomass yield in sorghum, nitrogen use efficiency in maize, and pollen production, viability and vigor in sunflower.
Infrastructure. Because genotyping and phenotyping technologies are rapidly changing, approaches and procedures to meet all four specific objectiveswill need to utilize core facilities and service labs whenever possible. For example, there exist high throughput genotyping facilities at ISU and at external non-profit and commercial organizations. These facilities provide low cost, quality assured data and bioinformatic services. On the other hand, preparation of seed and fields, as well as the planting, cultivation and harvest are conducted by individual (silo) plant breeding programs. This is largely a historical artifact from a time when Hatch-Act funding enabled individual plant breeding programs to purchase and maintain field equipment and hire technical staff.We will work with the Iowa Crop Improvement Association (ICIA), the RF Baker Center for Plant Breeding and our colleagues at USDA-ARS (Ames) to develop a core ‘field-plot’facility that is capable of meeting the fluctuating (due to the nature of extramural funding) demands of projects. The advantages of core genotyping facilities including quality assurance, cost effectiveness and ability to rapidly respond to changing priorities of funding agencies are directly translatable to a core field plot facility and will enable all of our PI’s to be more competitive and successful. The greatest challenge in development of such a facility will be development of a sustainable funding model. We will address this through an organic approach. Initially, the core field plot facility will consist of an annual financial commitment on the part of RF Baker Center for Plant Breeding (and perhaps other on-campus unitsincluding ICIA and ARS) as well as a commitment by all of our faculty to include financing for the facility in the budgets of their grant proposals. After establishing commitments on the part of our members and close collaborators, we will evaluate the potential for extramural commitments.
Although the ability to assess reaction norms with phenomic capabilities represents an essential element of this proposal, we are not proposing to establish aphenomics core facility.We recognize that image capture can be accomplished with off-the-shelf technologies, however a phenomics core facility also requires development of quality assured software pipelines for image storage, processing and data-reduction. A seminar series supported by the Plant Sciences Institute in 2014, identified a number of engineers and computer scientists at ISUwith the expertise to develop these tools and services. We have also learned from our colleagues at USDA-ARS that such services are being offered by commercial engineering organizations.Rather than committing our resources to develop field-based phenomic technologies, we will collaborate with on-campus engineers and engineers at commercial organizations to identify providers of high quality data that can be used to meet our research objectives.
Data generated by various ‘omics’ technologies will overwhelm the capabilities of our individual research projects. We are not currently prepared to address either hardware (data servers) or software (data transfer and management) issues.For example, one of ourGWAS projects has recently collected data that exceeds the capacity of desktop and small clusters to accommodate simple merging and sub-setting of phenomic and genomic data sets. Ten years ago, the NSF began to develop infrastructurefor a number scientific disciplines, including plant biology (IPlant). IPlant is primarily focused on development of data storage and transfer of large data sets, but has also incorporated data management and analysis tools into their infrastructure. Also,about ten years ago, the Generation Challenge Program of the Consultative Group of International Agricultural Research (CGIAR) began funding development of a data management software system, known as the Integrated Breeding Platform (IBP). With funding from the Gates Foundation, members of our team have begun to collaborate with the CGIAR to learn and evaluate the data management components of the IBP. During the first year of this project we will assess capabilitiesfor the IBP and IPlant and identify other options that may be available.
A hidden cost associated with emerging public information resources is the potential learning curves that may have to be repeated with every new project and graduate student. There are a couple of approaches to address this hidden cost. First we could ask technical staff supported by the field core facility to learn the data management tools and provide quality assured data to faculty and students. A second approach is to ask students to develop short “how to” videos for use by future students and faculty as they learn the systems. We will evaluate which of these two approaches will be most cost effective during the first year of the project.
Computational requirements for simulation modeling, data analyses and interpretation of experimental data generated by this project will exceed the capacity of typical small clusters of processors found in single investigatorcomputing environments. An alternative is to apply for ‘unused cycles’ of computing resources at high performance computing (HPC) facilities at ISU, at IPlant and throughout the US. Two of our members (Beavis and Yu) and their students have learned the Linux OS and command line interfaces to do this. Our experience (as well as that of colleagues in Animal Breeding) is that most faculty and students will be uncomfortable with the infrequent need to relearn the OS and interface. Further, we have learned that managers of HPC resources are responsive to requests for massive parallel computing, but not to intermediate sized computing challenges that occur at infrequent intervals. Our projects during the period of this project will be characterized by such requests. To meet our anticipated needs, four of our faculty have pooled financial resources with the Baker Lab and crop-modeling faculty of the Agronomy Department to build an intermediate computational resource. The computing resource will consist of 64 Thread Intel Xeon CPUs, 512GB of RAM, 2TB SSD array+4TB Research Data, Dell Perc H730 RAID controller, Redundant 1100 Watt power supplies, 4 – 1GB Nics + 2 – 10GB Nics, Windows 8.1 Virtualized on Windows Server 2012 R2.The resource will be managed by the Baker Lab.
Approaches.Procedures for each of the specific objectives will depend on the reproductive biology, life-span, public information resources, accessible ‘omics’ technologies and budgets for each AAT in each plant species. The specific procedural details will be described in grant proposals to funding agencies. Rather than conjecture about unknown and unpredictable successful grant proposals, herein we describe general approaches that transcend the specifics associated with allcombinations of AAT and plant species.
a) Approaches for discovering alleles associated with adaptive agronomic traits, as represented by reaction norms, are well established and described in the section on Previous Work. The only issues that remain includeidentification of appropriate and fundable AATs, identification of appropriate phenomic technologies and establishment of infrastructure as described previously in this section on Procedures. Outcomes will consist of genomic regions associated with AAT’s will be identified and reported through peer reviewed literature.
b) Approaches for development of predictive models for adaptive agronomic traits, will be based primarily on simulation modeling. As noted in the section on Previous Work, GP methods based on additive genetic architectures are incapable of providing accurate predictions for unrelated and new generations of breeding lines(Sebastian, Streit, et al., 2010). We hypothesize that semi-parametric GP methods for reaction norms will provide more accurate models. This hypothesis will be investigated using data from simulations as well as empirical data generated by multi-institute collaborations that are available through public web resources. Outcomes will consist of publications in which novel GP methods are proposed and evaluated with respect to objective criteria of accuracy, precision and power.
c-i) Approaches fordevelopment of designs that assure optimal genetic improvement strategies, through applications of systems engineering. This approach represents an adaptation of established methods from Operations Research (OR) to plant breeding systems. These principles include a) defining the project objectives, b) translating project objectives into mathematical objective functions, c) developing algorithms to solve the objective functions and d) implementing the algorithms in computational solvers to obtain optimal solutions. While there exist libraries and software to address c) and d), a) and b) represent difficult principles because there are no generic approaches to translating specific project objectives into mathematical functions. (Xu, Wang, et al., 2011)and (Canzar and El-Kebir, 2011)demonstrated how to define and translate cultivar development objectives into mathematical objective functions. We will extend these preliminary simple examples to additional genetic improvement projects based on GP for AATs. Specific outcomes will include joint publications involving plant breeders and systems engineers, development and delivery of a summer short course on use of engineering principles in design of plant breeding projects.
c-ii) Approaches fordevelopment of designs that assure optimal genetic improvement strategies, through Doubled Haploids are well understood and have been implemented in maize, barley, and canola. We will investigate genetic sources of haploid induction for soybean, sorghum and sunflower. We will address the challenge to minimize the cost while maximizing the production of DH lines through response surface analyses and OR principles. Specific outcome will be expansion of the ISU DH facility to provide DH capabilities to crops other than corn.
c-iii) Approaches fordevelopment of designs that assure optimal genetic improvement strategies, through In vitro nurseries is currently a concept. As such, there are a large number of fundamental biological discoveries that will be needed before implementation. Design of experiments to discover the underlying mechanisms needed for an in vitro system will be based on Response Surface Designs. If successful, we will utilize OR approaches to integrate the nurseries into genetic improvement programs. Tangible outcomes will be fundamental discoveries on regulation of phase transitions in cellular growth and development.
d) Approaches forimplementing accelerated genetic improvement strategies for adaptive agronomic traits will depend on funding. While ISU does not have extensive resources for translational research, the Baker Center for Plant Breeding as well as individual plant breeding programs have some longer term funding for translational research in Corn, Soybean and Sorghum. Results from specific objectives a),b) and c) will establish a foundation for approaching funding agencies. For example, two of our members have participated in a successful collaborative translational proposal to accelerate genetic improvements in soybean through application of GP methods.A tangible outcome will be extramural funding of translational research into applied breeding projects.