Plant breeding advances may spark annual corn and soybean yield increases
Gil Hodges was in a slump.
The Brooklyn Dodgers first baseman couldn’t buy a hit as his hitless streak in the 1952 World Series continued the following spring. Spurred on by Brooklyn fans, Hodges finally resumed hitting and never slumped in four more World Series appearances.
His hitting slump, though, was nothing compared to the 70-year yield stagnation that U.S corn farmers endured from 1866 to 1936. Average yields tallied just 26 bushels per acre (bpa) during this time.
“It is amazing that there was no appreciable change in productivity over that 70-year time period,” says Bob Nielsen, a Purdue University Extension agronomist who analyzed corn yields from 1866 to the present.
Plant breeders changed this, as they developed double-cross corn hybrids in the mid-1930s to replace the open-pollinated varieties farmers had planted for decades. Seemingly overnight, average yield increases zoomed from zero to 0.8 bpa every year from 1937 to 1955, says Nielsen.
Average annual corn yield increases stepped up again in 1955. Continued genetic improvements, teamed with increased adoption of mechanization, nitrogen fertilizer, pesticides, and overall improved soil and crop management practices, more than doubled increases to the average 1.9 bpa annual corn yield increase that continues today.
Soybeans, too, have kept an average upward yield clip, increasing an average 0.44 bpa since 1960, according to Agricultural Economic Insights.
Since then, however, average annual yield increases for both crops have not changed. Once again, though, potential exists for plant breeders to boost annual yield increases with technologies that include molecular markers, artificial intelligence, and gene editing, and other tools.
“So much technology has come about in the last 10 years that will enable us to blow the lid off genetic yield,” says Jesse Gilsinger, BASF North American soybean breeding lead.
Plant breeders historically considered just phenotypes, which are sets of visible characteristics of a crop plant. Breeders would cross plants based on visual attributes in hope of developing an inbred that would lead to a better hybrid or variety.
They knew little about what caused phenotypes.
“Back then, breeding was more of an art than a science,” says Wendy Srnic, Corteva Agriscience vice president of seeds research and development.
That changed when molecular markers surfaced in the late 1990s. A molecular marker is a specific DNA fragment in a known location associated with a particular gene or trait that impacts a plant characteristic.
Initially, they worked well for less complex traits, such as soybean cyst nematode resistance in soybean varieties.
“Before then, we had to do pathology screens to identify which potential varieties had the desired genes,” says Tim Pruski, a BASF soybean breeder. “They were laborious, time-consuming, and expensive. With molecular markers, the industry was able to deploy SCN resistance genes more efficiently and effectively.”
Breeders today also use molecular markers to develop more complex traits, such as yield or drought tolerance.
“We found out that you probably needed 15 to 20 really big signpost markers, and then 150 more that were smaller, but still influenced an outcome,” says Srnic.
Breeders now use data analytics to sort molecular markers and develop a mathematical formula to predict performance, she adds.
“When I started my career, we planted everything in the field because we couldn’t predict which lines would perform the best,” says Srnic. “Now, we can preload potential winners right from the start.”
“Tagging many genes with DNA markers allows breeding teams to select for these traits more precisely, much faster, and at greater scale than with field phenotyping alone,” says Trevor Hohls, head of global seed development at Syngenta Seeds. “This allows greater speed and precision throughout the breeding process, and means that we can ensure that the entire commercialization process is streamlined toward selecting the best seed products for farmers.”
Ever wish your corn had more kernels around the ear? These days, maybe it can.
Pairwise is teaming with Bayer to use gene editing to create more kernel rows on ears without impacting kernel size.
Genetic lines always have existed that could spur kernel numbers, explains Tom Adams, CEO of North Carolina-based Pairwise.
“The problem is they were in weird places, and not in elite breeding lines,” he says. “If you wanted to get it into elite breeding lines, you’d have to cross it with something else and then backcross it for years to get back to all the other elite characteristics. Gene editing allows you to start with the elite line and just alter the single trait.”
Pairwise uses CRISPR-Cas as its gene-editing tool. After cuts are made to the DNA, the plant’s repair system activates the gene that boosts kernel numbers.
Gene editing can make other changes to plants. Inari’s gene-editing strategy aims at boosting yield while improving nitrogen and water use efficiency through plant architecture changes. These are complex changes that require multiplexing, or multiple changes or edits to a plant’s genome, says Ponsi Trivisvavet, Inari CEO. Products developed by Inari are slated to appear in varieties sold by seed companies in a couple of years, she adds.
Gene editing is less expensive than transgenic trait technology in which a foreign gene is inserted into a plant. Trait development can cost up to $250 million, while the cost of sequencing a gene under gene editing can cost $200, she adds.
“Because gene editing creates variation that is already present in nature, the products can have a shorter and less complicated regulatory path than GMOs [genetically modified organisms] that involve inserting genes from other organisms into a crop genome,” adds Adams.
“When people talk about gene editing in the popular press, it sounds like we can just go in and write anything into the genome that we want,” says Adams. “It’s more complicated than that, as editing tools are complex to develop.”
Still, gene editing holds much promise, Adams says.
“We have other traits associated with yield that are coming down the pipeline,” he adds.
UAVs Boost Data Accuracy
Wading through thigh-high soybeans and head-high corn to evaluate products is a tough job for plant breeders and their technicians. Fortunately, unmanned aerial vehicles (UAVs) have made this job much easier.
“UAVs can do in 20 minutes what it would take a person a couple hours to do,” says Ben Stewart-Brown, a Bayer soybean product design scientist.
UAVs can also more quickly detect plots gone awry, such as one with half of its plants missing.
“This can prevent bad data from going into a data set,” says Stewart-Brown. “Better decisions about what lines we want to advance in our pipeline can result with better data.”
Dials work well for tuning in a radio program. In a sense, they’re akin to the technology that fuels Sound Agriculture’s On-Demand Breeding platform.
Through a field of science called epigenetics, this platform increases or reduces expression of existing genes in a hybrid or variety, similar to a dial used to fine-tune a radio station signal.
For example, a corn plant facing hot and dry conditions can naturally turn on or off certain genes to help it adapt, says Travis Bayer, co-founder and chief technology officer of Sound Agriculture.
“The whole goal [of On-Demand Breeding] is to make breeding much faster, and to develop traits on demand,” he says. “In traditional breeding, gene expression levels change when breeders make crosses and select for new traits. We can do that it in a more targeted and rapid way, such as changing the expression of genes that control drought tolerance or ear development.”
Sound Agriculture scientists can tweak gene expression in a matter of weeks, compared to the years taken by conventional breeding, he adds. If changes need to be made, scientists can quickly modify and retest them.
Unlike gene editing, no cuts are made to DNA. Nor are foreign genes inserted into DNA as occurs in transgenic technology. As a result, the regulatory process is less intensive, Bayer says.
Sound Agriculture is aiming On-Demand Breeding at corn, soybeans, and wheat in addition to specialty crops. The company is partnering with several seed firms to develop traits for existing varieties and hybrids, says Bayer.
Ever wonder why many National Basketball Association (NBA) games feature layups, dunking, or gunning from the three-point line?
Well, NBA teams have studied data that shows little difference in the success rate between midrange two-point shots and those from the three-point line.
“They’ve figured out that midrange shooting doesn’t make a lot of sense,” says Ben Stewart-Brown, a Bayer soybean product design scientist. “Some of the same things have happened in plant breeding, where we’ve brought in data scientists to gain insights into data we’ve collected.”
Data enables plant breeders to better and more quickly decipher the merits of genetic combinations. Compiling and organizing data from which to develop better corn hybrids and soybean varieties is a challenge, though.
The amount of genetic interactions that can occur during plant breeding is immense. For example, 1.2 million possible combinations can result among interactions between 13 soybean genes, says Ponsi Trivisvavet, Inari CEO.
“It would take forever to sort these out using Excel and paper,” she says.
Enter artificial intelligence, a technology that includes several subsets, such as machine learning, says Rania Khalaf, Inari chief information and data officer.
“Machine learning enables computers to mathematically predict outcomes or make classifications by finding patterns in large amounts of data,” Khalaf says. “It then learns to update these patterns or classifications over time as it sees new data.”
An early form surfaced in the 1990s, when IBM’s chess-playing Deep Blue computer evaluated 200 million chess positions per second as it defeated world chess champions.
“Winning these big games is very hard for humans, but easy for machines because they can crunch through a lot of data,” says Khalaf.
Machine learning is now coming to plant breeding, enabling faster and more precise varietal performance predictions, says Khalaf. In Inari’s case, it speeds products developed from gene editing (see “Gene Editing”) by reducing the number of edits that need to be tested in greenhouse and field trials, she adds.
“Machine learning is only as good as the data it sees,” says Khalaf. “It needs a lot of examples to understand and form patterns.”
It’s also essential to design the correct machine-learning algorithm for the specific task, she says. Algorithms are procedures for solving a mathematical problem that frequently involve repetition of one or more mathematical operations. They are often implemented and solved on computers.
Artificial intelligence and machine learning are no substitutes for human decision-making, either.
For example, computers trained with machine learning cannot differentiate beyond what they have been trained to do.
“You could use machine learning to train a computer to identify a cat,” says Khalaf. “If you showed it a picture of a dog, it could only understand it is not a cat.”
With the correct algorithms, though, machine learning can help make more accurate decisions in less time, she adds.
“We are using artificial intelligence to make selections from an almost limitless number of gene combinations,” says Bob Reiter, head of research and development for Bayer. “It has the potential to be a huge step change in terms of how breeding is done and the leveraging of technology — in particular digital agriculture — to create a new solution for our farmer-customers.”