By Tim Martinson and Michelle Podolec.
Public and private breeders of commodity crops such as maize and soybeans have long used sophisticated bioinformatics, statistical tools, and ‘big data’ to guide their breeding programs. These tools increasingly leverage detailed DNA sequence information to predict the performance of each crop’s breeding lines – and ultimately help breeders release improved varieties more quickly.
But these sophisticated breeding tools have not been available to breeders of smaller-scale specialty crops. With their limited resources, obtaining even rudimentary genetic maps has been time-consuming and expensive. So specialty crop breeding has relied heavily upon field observations (phenotyping), and the breeder’s individual knowledge and experience in making crosses and selecting (or discarding) progeny. Specialty crop breeding has been as much of an ‘art’ as a ‘science’.
Dramatic drops in the cost of DNA sequencing and genome assembly have, for the first time, made genetic data abundant and financially accessible to specialty crop breeders. In grapes, for example, obtaining high quality whole-genome sequences of the ~500 million DNA base pairs on 19 chromosomes has dropped in cost from millions of dollars in the early 2000s to about $3000 per accession today.
Now specialty crop breeders have access to a torrent of new DNA sequence data. But they still lack the infrastructure and access to bioinformatics that commodity crop breeders have enjoyed for decades.
USDA’s Breeding Insight project
Breeding Insight was established in 2018 by the USDA Agricultural Research Service to provide these resources to specialty crop breeding programs. According to project director Moira Sheehan, its purpose is to bring sophisticated data analysis and bioinformatics to a number of specialty crops.
“We build software that help our breeders manage their program, use new tools and technologies, build new capacities in their program, and to streamline their processes, getting rid of waste. We help our breeders decide where to prioritize their resources to meet their goals,” said Sheehan. “That’s Breeding Insight, in a nutshell.”
Initially, the program worked with six plant and animal crop species as the focus of USDA-ARS breeding programs. New crops added to the project this year include cranberry, pecan, lettuce, cucumber, honeybee, oat, and strawberry.
“We just started our third year and are currently focusing on six species. Two of them are fish species (North American Atlantic salmon and rainbow trout) and four of them are plants (forage alfalfa, sweet potato, blueberry, and table grapes).”
Breeding Insight includes biologists, programmers and communications specialists.
“We have a scientific team – people like me with PhDs in genomics or biology. We work directly with the breeders. But we also have an IT team, a software team, and these are computer programmers.” said Sheehan. “We make a really tight connection between the software group and the science group. Sitting in between those two groups is our communication and training lead, talking to both sides: ‘The breeder needs to do this. How do I create a training module for this?’”
Collaboration with VitisGen2
Grapes were selected for inclusion in the Breeding Insight program in part because of a longstanding USDA-ARS table grape breeding program, based in Parlier, California – but also because of the nationwide VitisGen (2012-2016) and VitisGen2 (2017-2022) projects, funded through the USDA’s Specialty Crops Research Initiative. VitisGen2 project co-leader Lance Cadle-Davidson, a USDA-ARS geneticist based at Cornell AgriTech, cited the project’s existing emphasis on DNA mapping and marker-assisted selection as a benefit to Breeding Insight:
“One of the reasons why we were chosen is that among the specialty crops, VitisGen2 has set a solid foundation for genotyping and DNA marker analysis, along with advanced phenotyping methods”, said Cadle-Davidson. “We were an example of a crop that’s pretty far along, and already had our own ideas about how we could contribute to and benefit from Breeding Insight’s efforts.”
Sheehan echoed that sentiment. “The table grape group in Breeding Insight is one of our fastest moving groups. They are farther ahead than any other group in Breeding Insight – they already had a large set of DNA markers, made available by VitisGen2 to others so that ‘anyone who wants to use them can’. Most of our other species didn’t have available markers, so we have to develop them from scratch.”
How VitisGen2 and Breeding Insight work together
Breeding Insight works closely with Vitisgen2 scientists Cadle-Davidson and Rachael Naegele – both USDA-ARS scientists who serve as domain experts.
“We work closely with Moira’s group to help guide how they invest their time”, said Cadle-Davidson. “As a grape group, whenever we meet with Breeding Insight there’s always university breeders, too, such as Bruce Reisch [Cornell University] and Matt Clark [University of Minnesota], plus Craig Ledbetter, who is the USDA breeder at Parlier.”
Other key collaborators from VitisGen2 on the bioinformatics side have been Cheng Zou, a postdoc bioinformaticist, Avi Karn, postdoc geneticist, who has now moved on to private industry, and Qi Sun, co-director of the Cornell Bioinformatics Facility.
Benefits to Breeding Insight
When asked how the collaboration with VitisGen2 benefitted Breeding Insight, Sheehan cited the strong support from industry groups such as the National Grape Research Alliance (NGRA). The group includes representatives of the table, raisin, wine, and juice grape sectors of the industry – who championed the project.
“Most [specialty crops] don’t have a very good centralized, organized coalition that can go to bat for them, nor can they decide from a pool of available research what is going to be the most beneficial to them.” Said Sheehan. “[for many crops], getting communication back up the food supply chain to breeders is really difficult. VitisGen2 does that.”
Sheehan also cited VitisGen2’s pioneering work in high-throughput phenotyping, especially to quantify variability of vine response to powdery mildew in the laboratory.
“Lance Cadle-Davidson and the VitisGen group have really been at the forefront of developing a machine [that captures images of powdery mildew] and uses computer image analysis to precisely phenotype powdery mildew resistance of each accession. Then they can now link these phenotypes back to specific parts of the genome that control those phenotypes. And that’s a powerful tool for breeders.”
Genomic characterization of USDA Grape Germplasm Collections
The USDA-ARS curates two extensive Vitis germplasm collections at Cornell AgriTech in Geneva, NY (cool climate accessions) and University of California at Davis (cold-sensitive accessions). Breeding Insight and VitisGen2 are jointly funding an effort to map the genomes of the 7,000 accessions housed at the two locations.
Using the so-called ‘rhAmpSeq’ platform developed by VitisGen2, each accession will have 2,000 DNA markers that cover 99% of the 19 grapevine chromosomes – allowing detailed comparisons of genetic traits across all the accessions.
“In the end, we will have the most detailed genotypic dataset for grapevine ever created”, said Sheehan.
Cadle-Davidson cited the benefits of characterizing the germplasm collections. “One of the reasons to do this is to discover what genetic diversity in Vitis exists out there. If we can capture all that genetic diversity, it will actually speed up our analysis of breeding lines and help us identify the genetic basis of desirable traits.”
“It’s akin to Google going out and scraping words from all the websites so that they can run their search algorithms faster.”
Bioinformatics and Breeding Programs
The value of genetic information derived through bioinformatics to breeders is both in the selection of which parents to cross, but also in the all-important decisions about ‘what to keep and what to discard’. By establishing associations between DNA markers and observed traits (phenotyping), breeders can use the DNA marker information for ‘marker-assisted selection’. The VitisGen2 project, for example, has been able to identify and ‘stack’ (combine in one plant) 4-6 unique powdery mildew resistance markers in breeding lines through DNA testing of grapevine seedlings.
Figure 1. This Illumina plate reader is where DNA sequence data is generated. Using the rhAmpSeq marker panel developed by VitisGen, one reaction can identify 2,000 genetic markers collected from up to 400 individual samples. “DNA bar codes” allow researchers to identify which markers come from which samples
To identify DNA markers within the genome, computer algorithms are needed to sort through the voluminous DNA sequence information.
“Cheng Zou from VitisGen and Dongyan Zhao of Breeding Insight both developed marker-assisted selection pipelines” said Cadle-Davidson. “We wanted an objective way to process the data and provide statistical and quantitative predictions, where before we were just looking at DNA marker data subjectively.”
The Future
For specialty crop breeders, knowledge gained from a career in observing a plant’s traits (phenotypes) has been far more important than knowing about the specific genetics of the parents used in breeding new crop varieties. “Many breeders may know a little bit about the pedigree, but what they’re doing is making selections based on know-how or art”, observed Sheehan.
This relies on generating high numbers of potential breeding lines. “It’s like the lotto. You plant out 30,000 lines and advance 3,000, based on your gut feeling or the plant’s appearance”
Marker-assisted selection (MAS) supplements breeders’ knowledge with genetic information “to inform decisions earlier and with more precision.” Said Sheehan. “Table grape breeders are using marker-assisted selection for powdery mildew resistance and other traits.”
This allows breeders to increase efficiency, and stock the breeding pipeline with better specialty crop selections.
Figure 2. The Blackbird phenotyping robot developed by the VitisGen project images leaf disks inoculated with powdery mildew spores, allowing researchers to quickly and accurately characterize powdery mildew resistance in grape breeding lines. Automating this processes allows grape breeders to quickly, consistently and efficiently determine the best seedlings out of thousands to keep, saving breeding programs time and money in the development of new introductions.
Looking to the future, ‘genomic selection’, or whole genome analysis and metrics, will be used quantitatively to predict what can happen within a specialty crop breeding program. While marker-assisted selection relies on a few key regions of the genome to predict traits, genomic selection will assign a value to the whole genome of an individual candidate for breeding. “Out of a population of several individuals, you will be able to make decisions based on a numerical value which will be the ‘expected breeding value’ of an individual.” This is analogous to information collected over decades by animal breeders such as the Dairy Herd Improvement (DHI) association, which uses production information to identify elite bloodlines and guide artificial insemination programs.
Moving from the ‘art and personal knowledge’ style of breeding to a more quantitative approach involving high-throughput phenotyping and detailed genetic information will pay dividends for specialty crop breeders and growers. Cadle-Davidson cited the power of using quantitative databases to capture breeders’ knowledge:
“There is tremendous value in capturing the art of breeding in a database for modern analytical tools to predict why the artist was successful – what traits were selected, and as importantly, what failed.”
“How do I talk to a breeder who’s about to retire, and convince them to use these [informatics] tools?” asked Sheehan. “They need to be able to hand off to the next person to continue the development and understand the context of the 35 years of work they have done.”
“If I can convince them to put something in a data management system and secure their data, it then becomes accessible to other people.”
The collaboration between Breeding Insight and the VitisGen programs has already yielded rich results for grape breeding programs. To date, the two VitisGen projects have discovered over 70 DNA markers identified in grape genes for fruit quality and disease resistance traits – results the VitisGen projects are leveraging into improved processes and data sets with the support of Breeding Insight. Together, these programs are providing the grape industry with exciting, cutting edge bioinformatics tools and processes to enhance their breeding programs, adding valuable quantitative information to the ancient art of the breeder and turbocharging breeding programs to produce improved varieties better and more efficiently.
Acknowledgements
Funding for VitisGen2 is provided by Specialty Crop Research Initiative Competitive Grant, Award No. 2017- 51181-26829, of the USDA National Institute of Food and Agriculture.
Breeding Insight is funded by the USDA Agricultural Research Service through a grant to Cornell University.
Resources
Breeding Insight Program website. https://breedinginsight.org/
USDA CRIS Project Portal. 2017 VitisGen2: Application of Next Generation Technologies to Accelerate Grapevine Cultivar Development. Project Proposal, 2017.
Martinson, T. E. and B. Reisch. 2020. The Core Grape Genome and Cheap DNA Sequencing: A New Roadmap for Grape Breeders. Appellation Cornell #42, August 2020
Martinson, T., Q. Sun, C. Zou, and L. Cadle-Davidson. 2019. Grape Breeders Search for Reliable DNA Markers: Why the Pinot noir PN40024 Reference Genome is Not Enough. Wine Business Monthly, December 2019
Tim Martinson is senior extension associate at Cornell AgriTech in Geneva, NY. Sam Filler is Executive Director of the New York Wine and Grape Foundation. Michelle Podolec is extension support specialist with the statewide viticulture extension program, based at Cornell AgriTech in Geneva, NY.