Performance Story: Filling Gaps in Wheat Cultivar Development with Translational Research
Genomic technologies have the potential to improve breeding efficiency in wheat. The level of our understanding of the wheat genome at the DNA level through technological advances is unprecedented. Over the course of the past several years, millions of DNA variants that can be converted to markers and deployed in our wheat breeding programs have become available. Multiple genome sequences of wheat, including two Canadian cultivars, have also emerged, providing an opportunity to develop genomic strategies that are tailored to Canadian wheat breeding programs. However, a significant challenge exists to properly deploy these markers and genome sequence information to improve wheat cultivar development in Canada. The goal of this research was to develop strategies to translate these upstream genomic technologies into breeding applications with the aim to support continued improvement of wheat cultivars adapted to western Canadian conditions.
Just a few years ago, it was evident that there is a translational gap between wheat genomic resources and the application of this information into our breeding programs. This existed for three main reasons. First, it was cost prohibitive to screen large populations of breeding lines with all available DNA markers. Development of an informative set of DNA markers that can be deployed effectively was needed. Secondly, validation of these DNA markers was required such that only those DNA markers that are effective at improving genetic gain in our wheat breeding programs would be utilized. Indeed, some DNA markers are more useful than others, and validation is essential if we are to make most efficient utilization of genomic technologies. Thirdly, marker validation studies are limiting because large amounts of phenotypic information must be collected simultaneously with the genomic information to allow identification of those DNA markers most effective at predicting the phenotype in the target environment. This project was established to mitigate these challenges, and to bridge the gap between genomic resources and wheat cultivar development programs.
The first challenge we tackled was the development of a DNA testing platform to support breeding. Advances in sequencing and genotyping methods have enabled cost-effective production of high throughput DNA markers that detect differences in DNA sequence among wheat breeding lines. Many laboratories have developed a number of high-throughput tests and testing platforms in well over 25 crop species. While these platforms work well for genetic research, they are largely cost prohibitive for applied breeding programs. To that end, we established a unique “breeders’ chip” that could be used for both durum and bread wheat to support breeding programs. We first established a diverse collection of our germplasm, consisting of 85% CWRS material, with the remaining being CPS, CNHR, and GP wheat classes. These panels were genotyped with 90,000 DNA markers with the aim to select a smaller subset of useful markers for which to design the breeder chip. Based on this data set, we determined that approximately 5000 markers would likely be sufficient for whole genome profiling of our breeding material. To reduce the number of markers we first selected the most informative DNA markers – that is those DNA markers that adequately sampled the genetic diversity available to our breeding programs. Next, we applied filtering algorithms to select markers that adequately covered the wheat genome. We then evaluated the potential of this reduced marker set to identify markers associated with signatures of selection in the AAFC- Swift Current Research and Development Centre and Crop Development Centre durum wheat programs. We found that the distribution pattern of markers under selection was different between the 21 wheat chromosomes, and we were able to identify several genomic regions that were clearly the target of breeder selection. It was clear from our analysis that the DNA markers were not inherited independently, but rather as “DNA blocks” (aka haplotypes) and identified and validated those DNA blocks. The results of this study support that the 5000 markers selected for the breeder chip are effective to identify those haplotypes that are associated with signatures of selection. All these haplotypes will be helpful in the identification of genes related to many traits of interest and to further facilitate marker-assisted selection in breeding.
The second challenge we tackled was to begin the process of validating our breeder chip to select for important traits in our breeding programs. To accomplish this, carefully designed experiments were established to test the association of DNA markers with agronomic disease resistance, and end-use quality traits that are important to wheat producers. In the context of this project, we focused specifically on resistance to fusarium head blight (FHB) and related mycotoxin contamination [such as dexoynivalenol (DON) produced by Fusarium graminearum], straw strength, and gluten strength. These traits are controlled by many genes, and the growing environment has a significant impact on the final phenotype. With such quantitative traits, phenotyping requires growing populations of wheat lines in multiple years and locations in the target environment to establish an average response. In this research we expanded our testing capacity for FHB and DON (particularly in hexaploid wheat) and worked to establish additional testing sites on the Regina plains to support additional phenotyping experiments. A testing site near Rouleau, SK was identified and leased. This site included a building to store machinery and sufficient land for up to 4,000 plots per year managed using a 2-year rotation. Unfortunately, the environmental conditions were challenging over the next four years. In 2015, dry conditions after seeding led to poor establishment and grain yield data was not useful. In 2016, timely seeding helped with excellent establishment of plots, however, a severe midseason hailstorm caused complete site loss. In 2017, we struggled with a lack of moisture after seeding again which led to uneven establishment. Grain yield was not collected due to the uneven plant stand; however, we were able to take agronomic and disease notes on the wheat lines. The 2019 growing season was the best trialing year despite a severe drought. Yield plots were successfully established. Despite these challenges, we collected phenotypic data as follows: grain yield (2014, 2018, 2019), straw strength, plant height, foliar disease, Fusarium head blight, shattering, quality testing on samples harvested (2014, 2018, 2019).
For our final objective, we set out to utilize the phenotypic data collected for our early generation breeding populations, doubled haploid populations, parents, and select genetic populations coupled with genotypic information to begin the process of evaluating their potential in a larger scale genomic assisted breeding platform. Briefly, genome-wide selection uses an array of DNA markers that spans all 21 wheat chromosomes which can be used to train computer models to predict the phenotype of lines within the breeding program before extensive expensive field testing. For our work, we focused on traits that would most benefit from genomic translational experiments and focused on phenotypic information consisting of the following traits: yield, protein concentration, FHB index (%) and FHB severity (%) and gluten strength. Our selection of traits was an iterative process, where traits were chosen based on a) importance to Saskatchewan producers, b) availability of appropriate genetic and breeding populations that would allow us to “hit the ground running” and c) our ability to utilize and expand existing infrastructure and resources for the necessary phenotyping experiments. Our analysis revealed several regions in the genome associated with each trait and identified DNA markers that could be used to enrich our breeding populations for desirable alleles responsible for these traits.
Next, we further applied our genotypic/phenotypic data to develop predictive models for genomic selection. For this work, we focused on phenotypic information consisting of the following traits: yield, protein concentration, FHB index (%) and FHB severity (%) and gluten strength, but emphasis here was on our analysis of end-use quality traits. In CWRS wheat, there are many end-use quality traits that define end-use quality. We applied multiple training populations to develop predictive models for grain protein concentration (%), gluten strength, and dough extensibility. The predictive models were effective at predicting those lines with “poor quality”, that is low protein concentration and poor gluten strength and dough extensibility. This confirmed to us that these predictive values would be effective in the context of breeding because it would allow the breeder to cull poor quality lines well before lines are evaluated in field trials. It is worth mentioning that breeders do not select for individual traits but rather perform multi-trait selection to develop cultivars with superior performance. This is particularly wise for end-use quality given that some quality traits are negatively correlated with one another. We developed multi-trait selection models for end-use quality, and these models performed well at identifying those lines with the best combination of quality traits.
This project set out to translate upstream genomic research by filling in gaps with an inexpensive genotyping platform that would make molecular breeding more cost effective and by generating the required phenotypic data to utilize the genomic data more effectively. This work represented an unprecedented collaborative effort between the leading wheat breeding institutions in western Canada. To accomplish our goals, we carefully selected the appropriate breeding and genetic populations from our breeding programs and combined our expertise in genomic assisted breeding and in field-based phenotyping. Outcomes of these two objectives were merged to accomplish the third objective of piloting genomic selection. Through this study phenotypic information on relevant Canadian germplasm in existing environments and in the Regina plain environment, which is a major gap for selection, was generated. Validation of DNA markers for multiple agronomic, disease and quality traits were obtained with DNA markers identified that are effective for improving genetic gain in our wheat breeding programs on the Canadian prairies. The trait-marker relationships discovered in this project moves the translation of upstream technology towards efficient utilization of genomic technologies in breeding. The phenotypic data collected in this study in combination with the application of a less expensive wheat breeder chip marker array technology was synergistic in translating genomic selection into a viable wheat breeding option. With the combination of phenotypic and genotypic data generated in this project, genomic selection models were identified that optimize accuracy when applied to relevant western Canadian germplasm.
In summary, this research has been instrumental for us to establish a genomic selection pipeline in a portion of the wheat breeding programs. It has allowed us to compile a large training set data base to experiment with and to answer fundamental questions about genomic selection going forward so that we can adjust and optimize this new tool in our breeding programs. Indeed, the wheat breeder genotyping platform developed here was instrumental to begin implementing genomic selection and since the completion of this project, our breeding programs further refined our genotyping strategies. Lastly, the work accomplished has provided us with a genomics selection template that we are now using for other portions of our breeding effort, and we have expanded our experience to other wheat classes including Canada Western Amber Durum (CWAD). The application of these new breeding tools will assist with meeting our improved yield and yield stability targets in new cultivars for farmers.