The first herbicide-resistant weeds started showing up on Keith Miller’s farm about 18 years ago. He grows sorghum, wheat, alfalfa, and soybeans on 4,000 hectares in western Kansas. Amaranthus, often called pigweed, is his biggest problem.
Since the 1980s, crop protection companies have introduced only one herbicide with a new weed-killing mechanism. During this period, many weeds have evolved to survive the modes of action used by existing herbicides, creating huge problems for farmers. In response, crop protection companies are increasing R&D efforts to discover new modes of action. They’re using new tools, often borrowed from the pharmaceutical industry, to speed up that process. But even if new products do emerge, many experts warn that farmers should moderate their use of herbicides to avoid the evolution of resistance going forward.
Spraying weeds with the workhorse herbicide glyphosate does virtually nothing these days, Miller says. A combination of dicamba, 2,4-D, and glyphosate will sometimes work, but the weeds are becoming harder to control, and it’s starting to affect his bottom line.
“I’m having to change the crops that I produce because I can’t find herbicides that will take care of the weed problem,” he says.
The problematic weeds in Miller’s fields are the result of natural selection. A few popular herbicides—especially glyphosate, which is used on nearly all the corn, cotton, and soybeans grown in the US—are applied liberally to kill weeds on farms the world over. Over time, weeds have evolved ways to thwart the biochemical mechanisms, known as modes of action, that those herbicides use to kill them.
Miller says he wants chemical companies to create herbicides with modes of action that weeds haven’t evolved to tolerate. When he started farming in 1976, that would have been a reasonable expectation. Between 1952 and 1984, crop protection companies introduced an herbicide with a new mode of action roughly every other year, but then that rapid pace of discovery came to a grinding halt.
Agricultural chemical companies have introduced only a single herbicide with a new mode of action since 1984. During the same period, instances of herbicide-resistant weeds increased by more than 600%. These herbicide-resistant weeds can hurt a farm’s yield, and force farmers to spend more on large quantities of herbicide in an attempt to achieve some level of control.
As the cost of herbicide resistance increases, agricultural chemical companies are now dedicating more resources to the search for herbicides with new modes of action. They’re hoping new tools—including artificial intelligence and DNA-encoded libraries—as well as inspiration from human drugs and natural products, will help them end the drought. But even if products with new modes of action reach the market, experts warn, resistance will persist unless farmers commit to using a more diverse suite of weed management techniques.
The herbicide drought
Steve Duke, an herbicide researcher at the University of Mississippi, blames the lack of new modes of action on three main factors: the introduction of crops that were genetically modified to tolerate glyphosate, increased regulatory costs, and industry consolidation.
Monsanto, now owned by Bayer, first introduced Roundup Ready soybean seeds in 1996. Plants grown from these seeds are genetically modified to survive applications of glyphosate, the active ingredient in Roundup herbicide. Farmers can spray entire fields with glyphosate without harming their own crops. The Roundup Ready system worked so well that it didn’t make sense for chemical companies to try to discover something better.
“Some companies quit doing herbicide discovery,” Duke says. “Others reduced the amount of herbicide discovery they were doing dramatically.”
In a 2011 paper, Duke cited patent data showing that the number of patents issued in the US for herbicides dropped from more than 432 in 1997 to 65 by 2009. At the height of enthusiasm for glyphosate, Duke says, crop protection companies likely had herbicides with new modes of action in development but didn’t advance them because executives worried they wouldn’t be competitive. “People weren’t willing to take that risk,” he says.
At the same time, the cost of complying with regulations was rising. A 2018 study funded by the industry group CropLife International estimated that registration-related costs for a new active ingredient more than doubled between 1995 and 2014 globally. It also found that the introduction of active ingredients for herbicides peaked in the 1990s, with nearly 60 new products that decade. Fewer than 20 ingredients were introduced in the 2010s.
Some companies quit doing herbicide discovery.
Steve Duke, herbicide researcher at the University of Mississippi
Meanwhile, Ken Pallett, a researcher and consultant, says consolidation in the crop protection industry slowed innovation by reducing the diversity of research programs and the number of scientists working on herbicide discovery.
Pallett got caught up in this consolidation himself. There were more than 20 companies working on herbicide discovery when he got his first industry job, at Rhône-Poulenc, in the 1980s. Rhône-Poulenc merged with Hoechst to form Aventis in 1999, and a few years later, Aventis sold its crop science division to Bayer, where Pallett finished his corporate career. Today, just four companies—BASF, Bayer, Corteva Agriscience, and Syngenta—account for nearly 60% of global herbicide sales, according to the research firm AgbioInvestor.
“You might have two companies merge together, but they don’t keep . . . two herbicide research groups going. They normally combine them,” Pallett says. “It’s a way of cutting costs. You have everybody working at one site, which means adopting one approach.”
In most cases, herbicides work by binding to one of the many enzymes critical for plant survival and slowing down its activity. The decline in innovation isn’t due to a scarcity of enzymes to target, according to Josef Appel, vice president of global herbicide research at BASF.
The Herbicide Resistance Action Committee, an industry-funded group that fights herbicide resistance and defines each herbicide’s mode of action, recognizes 25 modes of action, generally categorized by the enzyme targeted. In a review of patents and academic literature, BASF scientists identified 230 enzymes that are critical for plant survival, meaning they could be good targets. For more than 60 of the enzymes, evidence indicated they could be inhibited by chemicals, making them even more attractive as targets.
At a recent event about searching for new modes of action, hosted by the US National Academies of Sciences, Engineering, and Medicine, BASF herbicide researcher Jens Lerchl noted that most of the recognized enzyme targets have large, deep pockets where herbicides can bind. They tend to share other physical properties as well. He argued that the best future herbicide targets are enzymes that share those characteristics.
While there are lots of enzymes to go after, Duke says identifying a good target isn’t enough. In addition to killing plants, the targeting molecule has to be environmentally safe, harmless to humans, and cheap enough to compete with products that are already on the market.
Back to the well
Up until the 1990s, Pallett says, chemical company researchers didn’t even bother to investigate the mode of action for most herbicides. If it killed weeds, they were satisfied. Even today, more than a dozen herbicides have an unknown mode of action. Understanding the mode of action was once viewed as a time-consuming academic exercise. That’s no longer the case.
More than 50 weed species have evolved resistance to glyphosate, according to the International Herbicide-Resistant Weed Database. What many farmers once saw as a miracle fix is becoming less effective, and they’re eager for a solution. In response, crop protection companies are increasing their research on new modes of action, hoping to find something that can break through resistance. Some have already seen early successes.
Last year, Mitsui Chemicals launched cyclopyrimorate, the only commercial herbicide with a new mode of action since the 1980s. Cyclopyrimorate bleaches and ultimately kills weeds in rice paddies by inhibiting an enzymatic pathway that produces plastoquinone, a molecule needed for photosynthesis.
Mitsui started working on the molecule that would eventually become cyclopyrimorate in 1982, but it abandoned the project in 1985. By then, many rice farmers were satisfied with herbicides that inhibit acetolactate synthase (ALS), an enzyme involved in the production of branched-chain amino acids in plants. The company didn’t restart its research program until weeds started to develop resistance to ALS inhibitors, in the late 1990s.
FMC also hopes to introduce a new mode of action in 2023 or 2024 with tetflupyrolimet, another rice paddy herbicide. FMC decided to make new modes of action a special focus for research starting in 2010, as cases of herbicide resistance continued to proliferate. Three years later, the company discovered the lead that would eventually become tetflupyrolimet. FMC researchers then spent 5 years making 1,500 analogs to improve it.
Tetflupyrolimet inhibits an enzyme that’s part of the pathway producing pyrimidine, a nucleotide precursor that is critical for plant growth. FMC is now investigating whether the herbicide could be useful in other crops, like sugarcane, wheat, soybeans, and corn.
Kathleen Shelton, FMC’s chief technology officer for R&D, says the company’s typical discovery process is like casting a wide net into an ocean of chemicals. The trick for scientists is to be strategic about where they throw the net.
The first step is picking the best chemicals to screen for herbicidal activity. FMC researchers could test an unimaginable number of chemicals, so they examine molecules’ structures to glean clues about the classes that might make good herbicides. Each year they ultimately choose about 50,000 to 60,000 molecules with properties that look promising. They then apply those molecules to tiny amounts of plant material to see if they have any effect. FMC tweaks the best candidates thousands of times to try to boost their performance.
Most existing herbicides were discovered using a version of this classic screening approach. Shelton says the biggest difference now is that FMC starts investigating the mode of action early, a task that can be both difficult and expensive.
“It doesn’t tell you that it’s going to be a successful commercial molecule, but it tells you that you have novelty,” she says.
Like FMC, BASF is putting extra effort into discovering herbicides with new modes of action—two-thirds of BASF’s herbicide R&D budget now goes toward that discovery. Appel says a product that uses an existing mode of action would have to be really special for BASF researchers to pursue it. “You cannot do a lot with known modes of action anymore,” he says.
Testing molecules on actual plant material provides a realistic perspective on whether they will work in the field, Appel says, but the company is starting to move away from that method in the early phases of discovery. As greenhouse screening becomes more expensive, BASF and other firms are turning to new tools that could help the discovery process move faster.
Copying the drugmakers
In the search for new modes of action, some companies are turning the herbicide discovery process on its head. Instead of starting with thousands of molecules and searching for an enzyme that one might affect, companies identify a crucial enzyme and search for a molecule that inhibits it.
AgPlenus, a subsidiary of Evogene that is partnering with Corteva and BASF on herbicide discovery, is going all in on this target-based approach. CEO Brian Ember argues that aiming for an enzyme that hasn’t been targeted previously should increase the chances of finding a new mode of action.
“You know going in . . . this is a pathway that’s never been worked on before,” Ember says. “The molecules are going to look different. The actual enzyme is different.”
No new herbicide modes of action were discovered between 1984 and 2018.
Global cases of herbicide resistance have increased significantly since the 1980s.
The number of US patents issued for herbicides dropped quickly after the introduction of glyphosate-tolerant crops.
The company applies artificial intelligence to predict which plant enzymes might make a good target and then uses X-ray crystallography to seek pockets on that protein where an herbicide could bind.
Once AgPlenus understands the enzyme’s structure, it runs billions of computer simulations to predict molecules that are likely to inhibit its activity. Later in the discovery process, the company also uses AI to predict how to tweak early herbicide candidates to improve them. Nir Arbel, Evogene’s chief product officer, says the computational screens aren’t 100% accurate, but they provide a good starting point. Most importantly, they allow the company to test many more molecules than is possible in typical screens.
Fifteen years ago, X-ray crystallography, a common tool in the pharmaceutical industry, would have been too expensive to use in crop protection, where profit margins aren’t as high, but the price has decreased, Ember says.
“Because of the value of pharma molecules, [drug companies] can spend a lot more money,” he says. “They build up the technologies. Then 10 or 15 years later, as the cost of the technology comes down, we copy it.”
The start-up Enko is borrowing another tool from the pharmaceutical industry to identify promising molecules: DNA-encoded libraries. The company, which is working with Bayer and Syngenta, dips the enzyme it wants to target into a soup of 140 billion compounds. Each compound has a unique DNA tag that describes how it was made.
As the enzyme stews in this molecular soup, some compounds may bind to it; those that don’t are washed away. At the end of the screening, researchers can amplify the DNA tags on the compounds that are left to identify which ones are bound to the enzyme.
BASF’s Appel says the ability to screen so many compounds with DNA-encoded libraries is a huge advantage, and his company is using the approach as well. But he cautions that hits from those screenings indicate only that a molecule binds to a target, not that it actually inhibits it. And hits do not indicate whether a molecule can get into a plant and move around inside it.
BASF uses AI to help predict whether hits from DNA-encoded libraries will actually work in a plant. But Appel says the approach still can’t replace “eyeballing by experienced herbicide chemists.”
Enko CEO Jacqueline Heard says tools like DNA-encoded libraries and AI should make it easier to search for new modes of action. In the past, she says, the high cost of greenhouse screening scared off companies from searching for new targets. It was safer to aim for enzymes disclosed in patents to increase the odds of finding something that works. “They’ve really worn out the targets,” she says.
Tools like AI and DNA-encoded libraries can also reduce the risk of searching for a new mode of action. Arbel says AgPlenus tries to predict whether early candidates are likely to clear all the hurdles that come with developing a new herbicide, such as the ability to move within a plant or environmental and human toxicity. “It’s not a guarantee, but it derisks the entire process,” he says.
New sources of inspiration
Franck Dayan, a weed researcher at Colorado State University, says it’s time for crop protection companies to look for new sources of inspiration in their search for innovative modes of action. The understandable preference of chemists for molecules that are easy to work with has severely limited the types of chemicals used to make herbicides, he contends.
“They ended up using basically the same chemical space over and over and over, leading to the same kind of targets over and over and over,” he says.
Dayan argues that chemists should work with more natural products. The molecules can be harder to crystallize and purify because they have more complex backbones, more chiral centers, and more oxygen and nitrogen molecules. But Dayan says natural products are a part of the chemical universe that has been relatively unexplored by herbicide chemists, and companies brave enough to go there will have better chances of finding new modes of action.
They ended up using basically the same chemical space over and over and over.
Franck Dayan, weed scientist, Colorado State University
FortePhest, a start-up with investment from BASF, is working on an herbicide inspired by a natural product and claims that it has a new mode of action. The natural product, m-tyrosine, is a noncoded amino acid produced by Festuca grasses, which release it into the soil to inhibit the growth of competitors. The structures of noncoded amino acids differ slightly from that of the 22 amino acids that are encoded in the genomes of living organisms.
When m-tyrosine is floating around a cell, it can be incorporated by mistake into a protein, leading to deformations. “These deformations are bad enough to kill the plant,” FortePhest CEO Alex Kozak says.
Other companies are looking to human drugs for inspiration. Oerth Bio, which spun out of the biotech company Arvinas, hopes to use proteolysis-targeting chimera (PROTAC) technology, which is being tested to treat cancer in humans, to develop herbicides.
Instead of inhibiting enzyme activity, as existing herbicides do, PROTAC molecules flag target proteins to be destroyed by a cell’s proteolysis system. The process is found in all plant and animal cells that break down old or misfolded proteins. “Whatever we do is going to be a novel mode of action because it’s not a classic inhibitor,” Oerth CEO John Dombrosky says.
Jason Speake, Oerth’s vice president of chemistry, says degrading proteins rather than stopping their activity also opens up lots of new targets. For example, he says, the company could target structural proteins that form the scaffolding for cells. Such proteins aren’t part of a critical biochemical pathway, but destroying them could still damage an unwanted plant.
The start-up Projini Agchem is trying to make herbicides that kill plants by disrupting interactions between proteins, another technology being used by the drug industry. Projini uses AI to predict where two proteins engage with each other and then tries to design molecules that will fit into that space and stop the interaction, similar to filling a keyhole with putty to prevent the key from getting in.
There are no herbicides that use this approach yet, but CEO Dotan Peleg says the company could target dozens, and maybe hundreds, of protein-protein interactions that are critical to a weed’s survival. The early-stage company says it has already shown that one of its proof-of-concept herbicide candidates can disrupt one of these critical interactions.
Projini’s cofounder and chief scientific officer, Itay Bloch, says the molecules used to disrupt protein interactions tend to be chemically different from classic enzyme inhibitors. That means the firm will be using molecules from a part of the chemical universe that has been largely untouched by crop protection companies, he says.
The future of weed control
Crop protection companies often present new chemical technology as the solution to herbicide resistance. And this is a convincing argument for many of the farmers that lived through the introduction of Roundup Ready crops. They saw a technology that, almost overnight, massively reduced the cost and complexity of weed management.
Sarah Lancaster, a weed researcher and extension specialist at Kansas State University, says new modes of action will certainly be useful tools, but she cautions that they aren’t a panacea.
“The worst thing that could happen would be for farmers to look at this new mode of action, or this new herbicide, as some sort of revolution,” she says. “They could very quickly lose that tool as well if it’s not managed carefully from the start.”
Even if crop protection companies can find new modes of action, she says, farmers need to use a wider variety of weed management techniques. That could include planting rows of crops close together, creating a shady canopy to slow down weed growth, or adjusting the timing of planting to reduce competition with weeds. It could also mean returning to tilling a limited amount of the soil, a practice that farmers have mostly abandoned because it can cause erosion. In short, responsible weed control of the future may look a lot like weed management plans used in the era before Roundup Ready crops.
Duke says farmers aren’t often enthusiastic about such techniques because they’re less effective, more costly, and more complicated than spraying a whole field with Roundup. Overuse of an herbicide might cause problems decades in the future, but he says most farmers are understandably more concerned about the current year’s crop.
However, Lancaster points out that many farmers want to pass on their land to subsequent generations. She says a weed management program that relies too heavily on a single herbicide could diminish the next generation’s ability to successfully operate a farm.
Miller, the farmer from Kansas, is the third generation in his family to farm his land, and he’s in the process of transitioning management to his daughter, son-in-law, and nephew. They each operate about 400 hectares of the farm right now, and they’re facing weeds that are much less likely to be controlled by herbicides.
The first report of herbicide resistance in Kansas appeared in 1976, the same year Miller started farming. Now, more than a dozen species of herbicide-resistant weeds can be found in the state.
Miller says he’ll try almost anything to kill his weeds. The farm grows a variety of crops, and he rotates them regularly to make the environment less hospitable for weeds. He’s thinking about buying a device that attaches to a combine and grinds up weed seeds. This summer, Lancaster has set up a test plot on one of Miller’s most problematic soybean fields to research the effect of electrocuting herbicide-resistant plants. “I guarantee that she’s going to have her hands full,” he says.
Miller is hopeful that some of those methods might help, but he says an herbicide with a new mode of action is what he really wants. He’s waiting for a silver bullet.