Published in Crop Sci. 44:1907-1913 (2004).
© 2004 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
SYMPOSIUM ON GENOMICS AND PLANT BREEDING: THE EXPERIENCE OF THE INITIATIVE FOR FUTURE AGRICULTURAL AND FOOD SYSTEMS
Genomics, Genetics, and Plant Breeding
A Private Sector Perspective
Mark Cooper*,
Oscar S. Smith,
Geoff Graham,
Lane Arthur,
Lizhi Feng and
Dean W. Podlich
Pioneer Hi-Bred International Inc., 7250 N.W. 62nd Avenue, P.O. Box 552, Johnston, IA 50131, USA
* Corresponding author (mark.cooper{at}pioneer.com)
 |
Legacy of Phenotype-Based Pedigree Selection
|
|---|
PEDIGREE BREEDING STRATEGIES have been the basis for genetic improvement of corn (Zea mays L.) in Pioneer Hi-Bred from the foundation of the company in the 1920s through to the 1990s. Over this period, grain yield has undergone genetic improvement at a rate of around 75 kg ha–1 yr–1 (Duvick et al., 2004a, 2004b; Fig. 1a)
. It is widely understood that realized progress for grain yield in the U.S. corn belt has been an outcome of combining improved genetics with appropriate crop management strategies (e.g., plant populations). Systematic evaluations of the outcomes of this long-term corn breeding effort have shown that the performance phenotypes and genotypic composition of the elite germplasm pools of the breeding program can be changed by selecting directly on the trait phenotypes we seek to improve (e.g., Fig. 1; Duvick et al., 2004a, 2004b). Side-by-side phenotypic evaluations of a sequence of successful Pioneer corn hybrids, representing each decade from the 1930s to present, provides a description of the phenotypic changes for a number of the key traits that the breeders have directly or indirectly changed (Fig. 1a–c). Genetic fingerprints of the inbred parents of these hybrids provide a description of the genotypic changes that have occurred in association with the sustained breeding effort (Fig. 1d). Important phases can be identified over this period of breeding. Initially double-cross hybrids (1920s-1960s) were developed. From the 1960s there was a relatively rapid transition to the use of single-cross hybrids, the foundation of which was the organization of the corn germplasm into heterotic groups, represented in this example by the Stiff Stalk Synthetic (SS) and Non Stiff Stalk Synthetic (NSS) Groups (Fig. 1d).
View larger version (50K):
[in this window]
[in a new window]
|
Fig. 1. Best linear unbiased predictors (BLUPs ± SE) for phenotypic performance of three traits, (a) grain yield (hybrids grown at three densities, 30, 54, and 79 thousand plants/ha, and yield per hybrid is for the density giving the highest average yield), (b) percentage of plants not root-lodged, (c) anthesis to silking interval, measured in experiments conducted from 1990 to 2002, for a sequence of commercially successful Pioneer corn hybrids taken from an ERA study (unpublished data), and (d) a plot of the inbred scores on the first two principal components from analysis of SSR molecular marker profiles of the parents of the hybrids. (OPV = open pollinated variety, DC = double cross, TC = three parent cross, TC-MSC = three parent modified single cross, SC = single cross, SC-T = single cross transgenic; SS = Stiff Stalk Synthetic inbred line, NSS = non-Stiff Stalk Synthetic inbred line). The large boundaries distinguish three main groups of lines; Old = the old inbred lines used before the formation of the heterotic groups and the other two groups represent SS and NSS inbred lines. The arrows indicate the direction of the progression of inbred improvement in the SS and NSS heterotic groups.
|
|
|
Early Outcomes from Molecular Breeding
|
|---|
Advances in molecular genetics have reached a stage where breeding schemes can now be augmented with the use of a number of molecular technologies. Commercial breeding programs have and will continue to evaluate and invest in research that considers the prospects to either change or refine the in situ gene-to-phenotype system. From the 1990s, transgenic methods have been applied to key traits. Commercial transgenic hybrids have been developed for traits where there is a simple gene-to-phenotype relationship; e.g., Bt for insect resistance and multiple herbicide resistance genes. Some of the successful Pioneer Bt hybrids are included in the progression shown in Fig. 1. At this early stage of technology development it is tempting to conclude that transgenic solutions will only be applicable for trait targets under simple genetic control. However, this view is not accurate. Conventional genetic improvement of resistance to insect pests was viewed as a traditional complex trait problem before the widespread use of transgene sources of resistance to insect pests. Molecular marker based breeding strategies are being considered for improvement of simply inherited traits and quantitative traits that show complex inheritance in elite germplasm pools. Marker-assisted selection has been applied in breeder crosses when marker-trait associations are sufficient, the requisite marker polymorphism(s) can be identified and a high throughput system of assay is available. In most cases for complex traits the candidate regions identified by molecular markers have not been resolved to the level of candidate genes.
|
Organization of Genomics Efforts
|
|---|
Only over the last decade has the scientific community developed and had access to the range of molecular tools that provide the technological foundation that will be necessary to understand (i) the genetic architecture of the trait combinations we seek to manipulate, (ii) the nature of the genetic changes that were brought about by phenotypic selection, (iii) the power that can be attained in a breeding strategy (molecular and conventional) to achieve directed genetic changes that manipulate the trait phenotypes we seek to improve, and (iv) the limits that will ultimately be faced in using genetic technologies to make robust changes to plant phenotypes that improve the sustainability of agricultural systems.
Much of the genomic technological advancements used in plants were developed to meet the needs of the human genome effort. In most cases the application of these DNA-, RNA-, and protein-based technologies to study plant genomes has been straightforward. To take advantage of the opportunities that these genomic technologies provide to plant breeding, plant genomics efforts over the last decade have been heavily focused on plant specific gene discovery, gene function knowledge creation, and organization of the heterogeneous data sources that have emerged across the scientific community.
|
Creating a Molecular Breeding Focus
|
|---|
Today the concept of commercially successful molecular breeding is multifaceted and should be viewed as such. At its current stage of development as a proven breeding methodology, the term "molecular breeding" is a collective descriptor of the heterogeneous efforts, challenges, and opportunities being investigated to enhance the short-term and long-term success of the systematic procedures used to improve trait phenotypes by directed manipulation of the genotype at the DNA sequence level. At this time, molecular breeding is not an identifier of a single general breeding approach in the same way that "pedigree breeding" is such an identifier. Thus, many different breeding approaches are considered under the title of molecular breeding. Two major components are in use today: (i) direct movement of genes between individuals by a range of transgenic approaches and (ii) development of associations between interindividual DNA sequence variation and trait phenotypic variation in combination with the design of DNA based prognostics that can be used in high throughput systems as a component of a breeding program. The feasibility and the range of successful outcomes from both approaches are being enhanced for a range of traits by greater fundamental knowledge of plant genome organization and the functional properties of genes.
|
Improving a Breeding Strategy
|
|---|
The concept of evaluating alternatives and building on the strengths of an incumbent strategy is not new to plant breeding. The overriding motivation for considering molecular breeding strategies in place of conventional phenotypic–pedigree-based breeding strategies is that molecular-based selection does or with appropriate development will provide advantages over phenotype-based selection. Often many hidden assumptions are made in the theoretical discussions of the advantages that can be realized from molecular breeding strategies. One assumption that is often difficult to consider fully is the complexity of the genetics that the current strategy faces. Overly simplified genetic models can often give an associated overly optimistic assessment of the benefits, or in some cases lack of benefit, to be expected from an alternative strategy. Ultimately, validation by measuring realized benefits in situ are necessary. Because of the complex stochastic nature of the genotype–environment systems that breeding programs operate within, it has been resource intensive and difficult to demonstrate the advantage of one breeding strategy over another. A major difference between academic and commercial evaluations of molecular breeding strategies is the greater need by the commercial programs to make as many of the hidden assumptions that underlie the potential advantages and disadvantages as visible as possible for direct consideration. These advantages may come in the form of (i) reduced costs for achieving a given level of phenotypic improvement, (ii) improvements in the accuracy and precision with which we can make phenotypic changes, (iii) step–change improvements in phenotypes that were not previously accessible with comparable research investments into conventional breeding methods, and (iv) the identification of industry game-changing technologies for complex genotype–environment systems.
By emphasizing the need for a demonstrable advantage at the level of the commercial viability of breeding program outcomes, the criteria for success are set at a much higher level than would be the case if all that was required was a demonstration that genotype-based improvement of the phenotype, via manipulation of DNA sequence, was feasible. This is much the same process that was used by previous Pioneer breeders in judging the merits of alternatives to and refinements of the conventional pedigree-breeding program (Duvick et al., 2004a). Ultimately, for commercial breeding programs, the success of any alternative breeding strategy is based on the value that can be gained by all stakeholders from the improved phenotypes and the costs of attaining and maintaining these improved phenotypes.
Therefore, the challenge is to outperform the current breeding strategy for a wide range of situations. The range of approaches must work for the important traits, which will inevitably differ in genetic complexity. It is difficult to conduct comprehensive empirical evaluations of alternative breeding strategies for a large number of scenarios. An alternative approach is to use computer simulation (Cooper et al., 2002). Figure 2
provides a stochastic computer simulation comparison of a conventional phenotypic selection (PS) and marker-assisted selection (MAS) strategy for a series of putative quantitative trait models. In this figure the difference between the resulting performance of two groups of genotypes (Normalized difference in response; MAS-PS) selected after five cycles of breeding, using either MAS or phenotypic selection, is plotted against a measure of the genetic complexity of a trait [complexity here is quantified as an autocorrelation value estimated from sequences of genotypic values from random neighborhood walks in genetic space; see Cooper and Podlich (2002) for additional details]. The emphasis in this theoretical example is not on the details of the two breeding strategies, but on the average difference and variability in expected difference in response to selection between the two strategies for simple and complex genetic situations and the impact of both trait heritability and the level of knowledge of the genetic architecture of the trait that is available to the breeder to facilitate the implementation of the MAS strategy. Even though on average there is an advantage observed for MAS, this varies for different genetic architectures and also for different replicates for the same scenario.
The more effective the current breeding strategy the more difficult will be the challenge to outperform the incumbent strategy and demonstrate the advantages. As with most difficult challenges, the paths to improvement are many and any commitment to molecular breeding strategy development will be an iterative process. The commercial molecular breeding strategies we see today represent first or second cycle iterations of some of the potential paths to implementing molecular breeding strategies. These may be more accurately referred to as molecular enhanced breeding strategies, which apply molecular technologies around what are still predominantly large pedigree breeding strategies.
|
Foundations for Molecular Breeding: Integration of Genomics and Genetics
|
|---|
It is expected that the foundation for studying natural genetic variation at the DNA sequence level for traits in elite breeding germplasm pools and the design of successful molecular breeding strategies will involve integration of structural and functional genomics technologies within the framework of classical trait mapping methods. Through the latter half of the 1980s and the 1990s the development of a range of molecular marker technologies allowed mapping of traits to broad candidate regions [quantitative trait loci (QTL)] on genetic maps. In some cases continued investigation of these regions by fine mapping and ultimately direct sequencing of the resolved regions enabled identification and cloning of candidate genes and in a few cases verification of the causal gene and some knowledge of functional allelic variation. The availability of physical maps for a number of the important crop plants and the complete genome sequence for the model plant Arabidopsis thaliana (L.) Heynh. (The Arabidopsis Genome Initiative, 2000) and the crop plant rice (Oryza sativa; Goff et al., 2002; Yu et al., 2002) has enabled alignment of genetic maps with physical sequence and thus opportunity to target sequence data from genomic regions for gene discovery and gene function analysis. Further advances in technologies for the study of gene expression and protein interactions have opened up glimpses of some of the gene networks that are involved in the relationships between interindividual DNA sequence variation and trait phenotypic variation in elite breeding populations. The apparent complexity of the gene networks underlying the observed gene-to-phenotype relationships for plant development, specific pathways and traits has stimulated interest in the use of a range of advanced mathematical methods, within a systems biology framework, to develop and test gene-to-phenotype models for traits that operate across scales of biological organization. Validation of these models will be a critical step in this process and will be the ultimate assessment of the potential of such modeling approaches to create new gene-to-phenotype knowledge for traits that can be used to improve the traits and design robust molecular breeding strategies.
|
Foundations for Molecular Breeding: Some Current Considerations
|
|---|
Much of the outcomes from plant genomics efforts to date have involved large-scale data generation of a narrow sample of genotypic variation and its systematic organization, combined with descriptive efforts to annotate the features observed in the organized data. This process has revealed much about some important features of the structural organization of plant genomes within a broader evolutionary framework. Comparative approaches have been used to initiate candidate gene searches from model to target species. Moving from this broad comparative genomics view of gene discovery to a breeding strategy view that is focused on understanding the detailed organization of extant allelic variation for multiple traits within a selected elite germplasm pool, presents significant challenges. Nevertheless, for some traits this approach can be effective. The outcomes from experimental investigations structured around these genomic resources are being examined today and represent a logical step for testing comparative genetics hypotheses and refinement of the current genomic database annotations.
|
Foundations for Molecular Breeding: Power of Selection
|
|---|
When we consider the power that selection on phenotype has demonstrated in bringing about directed changes for traits, it is important to consider the robustness of this approach across a wide range of genetic situations where the breeder knew little about the detail of the genetic architecture of the traits. An important consideration in the design of any selection process, be it molecular or phenotypic, is that you get what you select for. Direct selection on the phenotype of the end-product traits of commercial significance is a relatively robust, albeit in some cases slow, approach when genetic variation exists for the target traits. Replacing or augmenting this system with a knowledge-based approach that targets selection at the level of DNA sequence variation will also rapidly bring about genetic changes. The rigor of the associations we develop between population level sequence variation and phenotypic variation will determine the robustness of this molecular breeding approach. We will be continually forced to refine our knowledge of trait genetics and gene-to-phenotype associations.
|
What Does a Genomics View Give Us Access to That We Did Not Previously Have?
|
|---|
Without appropriate investment into crop genomics research, we will always lack detailed knowledge in two areas critical for successful breeding: (i) the structural organization and functional properties of genetic variation for traits and (ii) the influences that plant breeding strategies have on the genetic variation that is widely used in agriculture. Without detailed knowledge in both of these domains, it is difficult to answer many of the important questions asked of breeding programs: (i) how sustainable is a breeding effort in the long term, (ii) how robust are the products of a given breeding program, and (iii) how important is it to and what are the appropriate procedures to maintain and utilize genetic resources? Consider two opposing views on the role of genomics in plant breeding.
The optimistic view is that much of the detail necessary for the design of molecular breeding strategies will flow from a comprehensive genomics view of our genetic resources. Broad information on the extent of colinearity and rearrangement of genomes among species is a first step. Revealing the detail of interindividual sequence colinearity within a species and within elite breeding germplasm pools is a necessary next step. Some surprises and testable hypotheses will emerge (e.g., Fu and Dooner, 2002). With a detailed knowledge of genome organization and gene function and appropriately designed experiments, it will become increasingly feasible to resolve some long-standing debates and competing genetic models of quantitative trait variation; e.g., the genetic basis of epistasis, pleiotropy, genotype x environment interactions, inbreeding depression, and heterosis. Further, a genomics view will enable the study of chromosomal recombination at the physical level. Importantly, detailed knowledge of sequence variation among individuals within a pedigree breeding structure will provide a powerful resource for understanding the distribution and genetic control of recombination and the historical importance of specific recombination events in the breeding process. There has been much debate in the scientific literature on the importance of minimizing and maximizing recombination for specific purposes. Ultimately, we would seek to be able to understand the genetic control of recombination, identify specific regions of the genome where it is important to restrict recombination and those regions where we need to create new recombinants. Collectively, knowledge generation of gene-to-phenotype relationships for traits would be greatly accelerated by access to a comprehensive view of genome organization and interindividual genomic variation. Testing the hypotheses of the organization of genes within gene networks requires experimental procedures for studying the structural and functional properties of genes, the functional nature of allelic variation and measurement of their coordinated regulation within a plant growth and development framework. Access to this genomic resource will enable the study of some of the key properties of gene networks and their involvement and roles in determination of phenotypes.
In contrast, the pessimistic view is that the biology for many of the traits targeted by a breeding program is so complex and interconnected that the context dependent knowledge generation that is required to achieve improvements in predictability of the system will be so great that it will be difficult or impossible to achieve sufficient knowledge to design a molecular breeding strategy that will consistently improve on large-scale targeted phenotypic selection (e.g., Fig. 2). This view is not void of theoretical consideration and is founded in considerations derived from complexity theory (Cooper and Podlich, 2002). However, to date there is no comprehensive experimental evidence to test such arguments. Nevertheless, experience suggests that it has been difficult to predict and that we understand little about the functional basis of many of the genetic improvements that have been achieved for quantitative traits by plant breeding programs. Within the human genetics scientific community, similar debates can be found on what are appropriate strategies for the study of and healthcare solutions for complex diseases (Sing et al., 2003; Botstein and Risch, 2003).
Genomics gives us some new perspectives on genetic variation. With access to data and information at the sequence level, our views of what contributes to the natural genetic variation that resides within the germplasm pools developed by breeding programs are changing. The traditional view of genetic variation as a function of loci with fixed effects acting in a predominantly additive manner is challenged by many of the properties of genes that are observed using genomics technologies. An important component of experimental evidence indicates that gene regulation is an important source of genetic variation. What appears to be linear when examined at the phenotypic level is not necessarily linear at the level of the gene network (Peccoud et al., 2004). This creates a complex situation where many of the effects of genes can be highly context dependent. Therefore, the genetic background and environments within which the genes are studied will influence the estimates of the effects of the genes. Plant breeders have always been exposed to this phenomenon but have never had the tools to investigate its genetic basis. The predominant models used to derive the statistical estimates of genetic effects do not yet take into account the nonlinear features of many of the context-dependent properties of genes. Thus, any changes in the effects of the genes as the genetic or environmental contexts change are not accommodated in our genotype–phenotype statistical association models. The important implication of this observation is that selection at the level of the phenotype can operate and utilize all types of genotype–phenotype associations, extending from simple to highly complex genetics, and progress from selection can still be observed. However, strategies based on manipulation of the genotype at the molecular level will only be able to utilize the currently available experimental information from statistically determined genotype-phenotype associations. Reminiscent of much of the debate around the effects of epistasis on response to selection, this forces the plant breeding community to ask specific questions of the implications of molecular breeding strategies for both short-term and long-term genetic improvement of complex traits.
The reality of many of the plant breeding situations we encounter in practice is likely a mixture of and somewhere between the extremes of the knowledge-driven optimistic view and the unpredictable complexity view. A major challenge for experimental genomics is to design experiments that help to resolve components of these issues and demonstrate and define uses of genomics to enhance breeding outcomes. Here, we identify two fundamentally different, but equally critical issues, that need to be considered in experimentally demonstrating the areas where genomics will affect multitrait improvement: (i) enhancing the rate of progress of a population of individuals toward a target genotype that has already been identified and defined, e.g., by accelerating a backcross process for simple traits or focusing a pedigree effort for more complex traits and (ii) the process of predicting, defining, and creating the new gene combinations that will provide performance enhancements that have not yet been discovered. Both are important features of achieving genetic progress in both the short-term and long-term. The former is easier than the latter.
|
Designing Molecular Breeding Strategies: Three Themes
|
|---|
Today we can consider coordinated development of three themes and associated research paths arising from research over the last 10 to 15 yr.- There is a growing knowledge base of the genetic architecture for some traits and how genetic variation is organized within unimproved and elite germplasm pools. Further work in this area will require the integration of genomics technologies with the study of genetic variation to conduct focused gene-to-phenotype studies. This will require fundamental questioning and in some cases refinement of our models of genetic variation. Development of our bioinformatics and computational modeling tools will be necessary.
- High throughput genetic profiling of individuals for key regions of the genome is now feasible for elite and unimproved germplasm pools. High performance management, manipulation, analysis and interpretation of molecular and phenotypic data will continue to be areas of research priority.
- Determining the power of molecular and conventional breeding strategies to achieve directed phenotypic changes for simple to complex traits.
This last area is the least developed of the three themes we have identified here. To date, we have a lot of practical experience with conventional breeding strategies and are now gaining some practical experience with molecular enhanced breeding strategies. High performance computing and simulation is being used to complement theoretical and experimental investigations. There is a clear need for further research into the appropriate statistical and biological modeling procedures for determining and testing gene-to-phenotype associations for complex traits. Demands for advances in this area will grow as we populate and explore data rich gene-genotype-phenotype knowledge bases.
|
Conclusions
|
|---|
Current structural and functional genomics methodologies provide the foundation for studying the genetic architecture and variation for traits. Quantitative integration of interindividual molecular and phenotypic variation is a challenging step that is an area of intense research in the study of gene-to-phenotype relationships. Applying genomic methods in parallel across many genotypes is considered an important step in enabling the study of genetic variation in elite germplasm and the design of molecular enhanced breeding strategies. The design of commercially viable molecular plant breeding strategies is an experiment in progress. Genomics has and will continue to make contributions to the knowledge base of our target crops. We will continue to improve our ability to identify the factors that have contributed to past successes in breeding and use this to identify potential new paths to improvement. As with the history of the development of conventional breeding strategies during the 20th century, it is expected that the design, evaluation, and commercial use of molecular breeding strategies will unfold in the 21st century from a rich mixture of both independent and collaborative contributions from the public and private sector research communities.
Received for publication July 23, 2003.
|
REFERENCES
|
|---|
- Botstein, D., and N. Risch. 2003. Discovering genotypes underlying human phenotypes: Past successes for mendelian disease, future approaches for complex disease. Nature Genet. Supp. 33:228–237.
- Cooper, M., and D.W. Podlich. 2002. The E(NK) model: Extending the NK model to incorporate gene-by-environment interactions and epistasis for diploid genomes. Complexity 7:31–47.
- Cooper, M., D.W. Podlich, K.P. Micallef, O.S. Smith, N.M. Jensen, S.C. Chapman et al. 2002. Complexity, quantitative traits and plant breeding: A role for simulation modeling in the genetic improvement of crops. p. 143–166. In M.S. Kang (ed.) Quantitative genetics, genomics and plant breeding, CABI Publishing, Wallingford, UK.
- Duvick, D.N., J.S.C. Smith, and M. Cooper. 2004a. Long-term selection in a commercial hybrid maize breeding program. Plant Breed. Rev. 24:109–151.
- Duvick, D.N., J.S.C. Smith, and M. Cooper. 2004b. Changes in performance, parentage, and genetic diversity of successful corn hybrids, 1930–2000. p. 65–97. In C.W. Smith et al. (ed.) Corn: Origin, history, technology, and production. John Wiley & Sons Inc., NJ.
- Fu, H., and H.K. Dooner. 2002. Intraspecific violation of genetic colinearity and its implications in maize. Proc. Natl. Acad. Sci. USA 99:9573–9578.[Abstract/Free Full Text]
- Goff, S.A. et al. 2002. A draft sequence of the rice genome (Oryza sativa L. spp. japonica). Science 296:92–100.[Abstract/Free Full Text]
- Peccoud, J., K. Vander Velden, D.W. Podlich, C.R. Winkler, W.L. Arthur, and M. Cooper. 2004. The selective values of alleles in a molecular network model are context dependent. Genetics 166:1715–1725.[Abstract/Free Full Text]
- Sing, C.F., J.H. Stengård, and S.L.R. Kardia. 2003. Genes, environment and cardiovascular disease. Arterioscler Thromb. Vasc. Biol. 23:1190–1196.[Abstract/Free Full Text]
- The Arabidopsis Genome Initiative. 2000. Analysis of the genome sequence of the flowering plant Arabidopsis thaliana. Nature 408:796–815.[Medline]
- Yu, J. et al. 2002. A draft sequence of the rice genome (Oryza sativa L. spp. indica). Science 296:79–92.[Abstract/Free Full Text]
Related articles in Crop Science:
- THIS ISSUE IN CROP SCIENCE
Crop Science 2004 44: 1889-1892.
[Full Text]
This article has been cited by other articles:
|
|
|
|
 
S. P. Moose and R. H. Mumm
Molecular Plant Breeding as the Foundation for 21st Century Crop Improvement
Plant Physiology,
July 1, 2008;
147(3):
969 - 977.
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. B. Egli
Comparison of Corn and Soybean Yields in the United States: Historical Trends and Future Prospects
Agron. J.,
May 7, 2008;
100(Supplement_3):
S-79 - S-88.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
Legacy of Phenotype-Based...
Early Outcomes from Molecular...
1 represents the more simple additive genetic models and the autocorrelation