Herbicides as Weed Control Agents: State of the Art: I. Weed Control Research and Safener Technology: The Path to Modern Agriculture

The development of weed control and safeners and of screening tools to find them has evolved substantially over the past 100 years.

Abstract

The purpose of modern industrial herbicides is to control weeds. The species of weeds that plague crops today are a consequence of the historical past, being related to the history of the evolution of crops and farming practices. Chemical weed control began over a century ago with inorganic compounds and transitioned to the age of organic herbicides. Targeted herbicide research has created a steady stream of successful products. However, safeners have proven to be more difficult to find. Once found, the mode of action of the safener must be determined, partly to help in the discovery of further compounds within the same class. However, mounting regulatory and economic pressure has changed the industry completely, making it harder to find a successful herbicide. Herbicide resistance has also become a major problem, increasing the difficulty of controlling weeds. As a result, the development of new molecules has become a rare event today.

Modern industrial herbicide research begins with the analysis and definition of research objectives. A major part of this lies in the definition of economically important weeds in major arable crops (Kraehmer, 2012). Weed associations change slowly over time. It is important, therefore, to foresee such changes. Today’s weed associations result from events in the distant past. They are associated with the history of crops and the evolution of farm management. In Europe and the Americas, some large-acre crops such as winter oilseed rape and spring oilseed rape (canola), both derived from Brassica spp., and soybean (Glycine max) have attained their current importance only within the last 100 years. Other Old World crops, such as cereals, have expanded over a very long time span and were already rather widespread in Neolithic times (Zohary et al., 2012). The dominance of crop species in agricultural habitats only left room for weed species that could adapt to cultivation technologies. Changes in crop management and the global weed infestation have happened in waves. A major early factor in Europe was presumably the grain trade in the Roman period (Erdkamp, 2005). The Romans spread their preferred crops and, unintentionally, associated weed seeds throughout Europe, Asia, and Africa. A second wave of global vegetation change started in the 16th century after the discovery of the Americas. Crops and weeds were distributed globally by agronomists and botanists. Alien species started to spread on all continents. A third phase can be seen in the 19th century with the industrialization of agriculture and the breeding of competitive crop varieties. The analysis of weed spectra in arable fields grew from this historical background. Weeds are plants interfering with the interests of people (Kraehmer and Baur, 2013), which is why they have been controlled by farmers for millennia.

Chemical weed control began just about a century ago with a few inorganic compounds, such as sulfuric acid, copper salts, and sodium chlorate (Cremlyn, 1991). The herbicidal activity of 2,4-dichlorophenoxyacetic acid was detected in the 1940s (Troyer, 2001). Table I provides an overview of selected chemical families, selected representatives, and earliest usage reports according to Büchel et al. (1977) and Cremlyn (1991), Worthington and Hance (1991). Targeted herbicide research began in the 1950s. In the early days, herbicide candidates progressed from screens purely on the basis of their having biology that would satisfy farmers’ requirements. Mode of action (MoA) studies did not play a major role in the chemical industry prior to the 1970s. Analytical tools were developed and the rapid elucidation of plant pathways and in vitro-based screen assays were used from the 1980s onward. However, in the 1990s and beyond, ever-increasing regulatory and economic pressures have changed the situation of the industry completely, and to satisfy the new requirements, selection criteria beyond biological activity have needed to be applied. Herbicide resistance in weeds has developed into a more serious problem that now constrains the application of certain types of herbicides in some markets. Finally, the introduction of crops resistant to cheap herbicides and of glyphosate-resistant soybean, in particular, took value out of the market and resulted in an enormous economic pressure on the herbicide-producing industry. As a result of this changing and more difficult landscape, the development of new molecules is now a rare event.

Table I. History of chemical weed control innovations.

Post, Postemergence application; Pre, preemergence application, based on data from Cremlyn (1991), Worthington and Hance (1991), Büchel et al. (1977), Herbicide Resistance Action Committee (www.hracglobal.com), and others.

MoA, Target Site	Chemical Family	Examples	Use	Earliest Reports
Unspecific	Inorganic herbicides	H2SO4, Cu2SO4, FeSO4, NaAsO2	Total	1874
Uncouplers	Dinitrophenoles	dinitro-ortho-cresol	Post, dicots	1934
Auxins	Aryloxyalkanoic acid derivatives	2,4-Dichlorophenoxyacetic acid	Post, dicots in cereals	1942
Microtubule organization	Arylcarbamates	Propham, chloropropham	Pre, monocots in various crops	1946
Lipid synthesis	Chloroaliphatic acids	TCA, dalapon	Pre, monocots in various crops	1947
	Thiocarbamates	EPTC, triallate	Pre, monocots and dicots in various crops	1954
PSII	Arylureas	Monuron, diuron, isoproturon, linuron	Pre and Post, monocots and dicots in various crops	1951
	1,3,5-Triazines	Atrazine, simazine	Pre and Post, broad spectrum in corn	1952
	Pyridazines	Chloridazon	Pre, dicots in sugar beet	1962
	Uracils	Bromacil, terbacil, lenacil	Soil applied, broad spectrum in various crops	1963
	Biscarbamates	Phenmedipham	Post, dicots in sugar beet	1968
	1,2,4-Triazinones	Metribuzin	Pre in soybean	1971
Very-long-chain fatty acid biosynthesis	Chloroacetamides	Allidochlor, alachlor	Pre, monocots and dicots	1956
PSI	Bipyridyliums	Diquat, paraquat	Nonselective	1958
Protoporphyrinogen oxidase	Diphenyl ethers	Nitrofen, acifluorfen	Pre and Post, various crops	1960
	Oxadiazoles	Oxadiazon	Rice, nonselective	1969
Microtubule assembly	Dinitroanilines	Trifluralin, pendimethalin	Pre against monocots and dicots	1960
Cellulose biosynthesis	Nitriles	Dichlobenil	Plantations	1960
5-Enolpyruvylshikimate 3-phosphate synthase	Glys	Glyphosate	Post, nonselective	1971
Phytoene desaturase	Pyridazinones	Norflurazon	Pre and Post in cotton	1973
ACCase	Aryloxyphenoxy propanoates	Diclofop, fluazifop	Post, grasses	1975
	Cyclohexane diones	Alloxydim, sethoxydim	Post, grasses	1976
Gln synthetase		Glufosinate	Nonselective	1981
AHAS or ALS	Sulfonylureas	Chlorsulfuron, metsulfuron	Monocots and dicots in various crops	1982
	Imidazolinones	Imazapyr, imazethapyr	Nonselective or selective in soybean	1983
	Pyrimidinyl benzoates	Bispyribac sodium	Rice	1994
HPPD		Pyrazolynate, sulcotrione	Various crops, monocots and dicots	1984

This article is structured into three main topics. First, it provides an historic overview of the development of weed control history and of screening tools over the past 100 years. Thereafter, we concentrate on the use of MoA studies as a tool for optimizing chemical structures based upon knowledge of their receptors. Finally, we review the invention and use of safener technologies as a tool for improving the crop selectivity of herbicides. In a companion review (Kraehmer, et al., 2014), we address the serious challenges that farmers now face because of the evolution of herbicide resistance in weeds and the types of innovations that are urgently required.

AGRICULTURAL CHANGES IN THE PAST AND THEIR INFLUENCE ON WEED INFESTATION

Crops Grown on Arable Land within the Last 100 Years, Weed Management, and Changes in Weed Infestation

Crop management practices have had a major impact on weed infestation. Animals such as horses were used as production tools in the beginning of the last century. The horses required feeding, and as a result, for example, in 1920, oats (Avena sativa) were grown on more than 40 million acres in the United States (area harvested); today, the acreage amounts to only 1 million acres. It is not surprising, therefore, that wild oats, which were nearly impossible to control in the past, spread so quickly over many continents.

Table II illustrates a few other striking facts: corn (Zea mays) was already planted on 101 million U.S. acres in 1920 and on approximately 97 million acres in 2012. In sharp contrast, soybeans were cultivated on less than 500,000 acres in 1920 but on 76 million acres in 2012.

Table II. Acreage changes in U.S. crops during the last 100 years.

Values shown are million acres harvested. Source: U.S. Department of Agriculture Crop Production Historical Track Records, April 2013.

Crop	1920	1970	2012
Barley	7.5	9.7	3.2
Canola	<0.1	<0.1	1.6
Corn	101.4	66.1	97.2
Cotton	34.4	11.2	9.4
Hay	73.1	61.5	56.3
Oat	42.8	18.6	1.1
Potato	3.3	1.5	1.2
Rice	1.3	1.9	2.7
Rye	4.9	1.5	0.3
Soybean	<0.5	42.3	76.1
Sugar beet	0.9	1.5	1.2
Wheat	62.4	43.6	49
All crops	347	283	309

One very effective weed management tool is tillage. Mechanical weed management, however, is time consuming, labor intensive, and leads to a high energy consumption. Regrettably, it has also resulted in major erosion problems all over the world (Montgomery, 2007). Erosion rates on U.S. arable land declined considerably between 1982 and 1987: from 3.06 billion tons down to 1.73 billion tons, or from 7.3 tons per acre to 4.8 tons per acre (National Resources Inventory; http://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/nrcs143_012269.pdf). Changes in tillage practices were partly responsible for this achievement.

Weed infestation in spring crops and winter crops can differ considerably (Håkanson, 2003). The profound reduction of spring crops, and especially the reduction of the oat acreage when tractors replaced horses, resulted in a complete shift from wild oats (Avena fatua) as a major grass weed to blackgrass (Alopecurus myosuroides) and silky bentgrass (Apera spica-venti) in Europe (Kraehmer and Stuebler, 2012). The hard winters in Canada prevent the cultivation of winter crops in many agricultural areas. This is why winter annuals such as blackgrass do not play a major role there. Canada has a long tradition of conducting weed surveys. Despite major changes in the use of agrochemicals, the most dominant weeds have remained the same for decades (Leeson et al., 2005). The Weed Mapping Working Group of the European Weed Science Society endeavors to map the most common weeds of Europe in all major crops and to document changes in weed infestations (http://www.ewrs.org/weed_mapping.asp). One obvious result is that weed spectra change with cropping practices and environmental conditions and that some species are better adapted to the warmer climates in the Mediterranean area, such as Abutilon theophrasti or Sorghum halepense, whereas others are more frequent in northern Europe, such as Alopecurus myosuroides or Apera spica-venti. The weed species Chenopodium album and Echinochloa crus-galli are characterized by a wide ecological aptitude (Kraehmer, 2010) and can be found everywhere in Europe and in many crops.

BIOLOGICAL SCREENING

Glasshouse and Field Screening

In the early days, indicator species or model plants played a major role in herbicide discovery screening. Weed target species were often not available. Most herbicide screening in the agrochemical industry (Fig. 1) between the 1950s and 1980s was characterized by a protocol where solutions of test chemicals were sprayed over sets of plants in pots or over seeded bare soil in the case of preemergent screens.

Figure 1.

Plants in pots. Right, Untreated; left, experimental herbicide at 300 g active ingredient ha⁻¹.

The screening compounds were either predissolved in solvents such as acetone or they were formulated as wettable powders, emulsion concentrates, or suspension concentrates before dilution in water. The spraying process simulated the actual spraying situation in the field. Every company had its own specially designed spraying equipment. The screening process was divided into two to four initial selection steps. The initial, or plus-minus activity, tests started with high dose rates, generally between 1 and 10 kg active ingredient ha⁻¹. Compounds were sprayed in preemergence and postemergence trials. Later steps with lower dose rates and varying plant species followed to further refine activity and the weed spectrum. The results of these trials were usually visual phytotoxicity ratings on different scales, such as on a 0% to 100% rating scale, 0% meaning no phytotoxicity and 100% equal to complete control of plants. Fresh or dry weights were used for comparisons in special tests only. Symptomology was an essential and integral part of ratings. Each chemical class of compounds usually resulted in typical patterns of symptoms. Observations made in different screening indications (e.g. insecticide, acaricide, nematicide, or fungicide screens) were also taken into account during evaluation (Figs. 2 and 3).

Figure 2.

Typical symptoms of HPPD bleachers in an advanced screening stage.

Figure 3.

Typical auxin-transport inhibitor symptoms in a nematode screening assay.

Chemists usually prepared sample amounts of 3 to 5 g. This amount allowed the early testing of chemicals in all screening indications. Most midsize and large companies tested between 1,000 and 5,000 compounds per year. Information technology tools suitable to facilitate screening were not available before the 1980s, and structure-activity relationships had to be derived manually in long and time-consuming procedures. The last glasshouse step was a profiling procedure in which recommendations for field testing were derived. Often, special tests were carried out before or in parallel to field tests to check the soil-dependent performance of a compound, the influence of different formulations, potential carryover risks, and crop selectivity ranges. Depending on company size and resources, between 10 and 100 compounds were advanced into field testing per year. Compounds were tested in different parts of the world on plots between 1 and 10 m² with three to four replicates.

Two to 3 years in the field were usually sufficient to make informed development decisions. In the third quarter of the last century, it usually took between 4 and 6 years from the first synthesis of a compound until entry into the market. Later, the development process became more involved and took a greater investment of time to develop a new herbicide, as we will show later.

Early MoA Studies

The number of MoAs was quite limited between 1950 and 1970. Most commercial products were characterized as auxin-type herbicides, PSI or PSII inhibitors, or inhibitors of cell division. During the 1970s, inhibition of photosynthesis was typically tested at the whole-plant level using infrared gas analyzers in growth chambers. At the biochemical level, the so-called Hill reaction served as a tool for the identification of photosynthesis inhibitors in isolated chloroplasts (Arndt and Kötter, 1968). In the 1980s, Clark electrode measurements with isolated leaf cells and fluorescence emission assays with whole leaves provided further nondestructive tools for the characterization of PSII herbicides (Voss et al., 1984a, 1984b).

Plant tissue cultures coupled with HPLC analysis have been used since the 1970s to characterize herbicide-induced changes in metabolism. Cultures of aquatic plants such as Lemna minor have also long been used as an herbicide test system. Perennial weeds sometimes required a different approach. Systemic action of herbicides is required for the control of perennials such as couch (Elytrigia repens), johnsongrass (Sorghum halepense), or field bindweed (Convolvulus arvensis). Special translocation tests were used to check for this property. For example, Phaseolus spp. beans with two leaves provided some indication of systemic action when one of the leaves was treated. Translocated compounds then led to symptoms in the untreated leaf.

From the 1980s onward, special biochemical assays were developed for the characterization of target sites and for binding studies. Target enzymes of plant-specific pathways were preferred in order to avoid toxicological problems in mammals. We will come to this period in one of the following paragraphs.

Miniaturized Screening Assays

A perennial problem faced by the agrochemical industry has been the question of where to obtain new compounds in amounts sufficient for screening. At the end of the last century, an intensive exchange of compounds between nonagrochemical and agrochemical origins began. Universities and private institutes offered agrochemical companies their stocks and undertook the synthesis of additional desired compounds. Natural products were available for testing also. Since most third-party compounds were provided in low amounts, it drove the adoption of miniaturized herbicide discovery assays. These assays provide a basic initial test for phytotoxicity but give only limited discriminatory information to guide compound optimization. Seed germination assays in petri dishes were also quite commonly used. The first purpose of these assays was to find hits that would justify the synthesis of analogs and of samples with larger substance amounts. Tissue culture tests as described by Gressel et al. (1993) were also employed by a number of companies. Some companies have kept and improved such miniaturized assays until today, as published by Grossmann et al. (2012). Several assays were incorporated into high-throughput screening systems that allow the screening of one million compounds per year and more (Kraehmer, 2012). We will touch upon this screening approach in one of the following sections.

Glasshouse-to-Field Transfer

One disadvantage of early screening sets of the last century was the use of model plants such as onion (Allium cepa), carrot (Daucus carota), potato (Solanum tuberosum), and others. They were employed because the seed of target weeds was not available in the early herbicide screens. Today, specialized providers can deliver high-quality weed seed in guaranteed quantities. Nevertheless, even given the use of real target weeds, laboratory and glasshouse data are still not usually fully predictive of the performance of compounds in the field. Many parameters contribute to the variable transfer factors from glasshouse to field (Kraehmer and Russell, 1994). One of these parameters is the test species. Model plants such as Lemna spp., Arabidopsis (Arabidopsis thaliana), and Brachypodium distachyon, differ from weeds in many ways, including their specific uptake and translocation properties, metabolism, specific life cycle, and environmental requirements for growth. Lemna spp. are aquatic weeds in contrast to terrestrial weeds. Arabidopsis and B. distachyon are not serious competitors within arable crops under agricultural conditions. The same rules apply to many other species in miniaturized assays. General phytotoxicity principles can only rarely be used for the optimization of chemical structures or for structure-activity relationships. Too many positive results in an indicator assay often require repeated screening steps to reduce their numbers. Certain types of herbicides may not be detected at all if the indicator species is too different from the screen model species. For example, acetyl-coenzyme A carboxylase (ACCase)-inhibiting graminicides would not have been detected by screening against Arabidopsis. Similarly, the potential of a compound to control perennial weeds will not be evident from screens that only include annual species, and information about soil-plant interactions is absent from screens based upon liquid systems that do not include soil. The potential of a compound to control perennial weeds will often not be detected with annual species. Plant cells grown in tissue cultures are very different from whole plant cells because they are undifferentiated and can derive nutrients, often present in excess, from the culture medium and are buffered such that the influence of pH values on the uptake of compounds is masked. The actual value of a new compound is always based upon its performance against economically important weed species. It is highly questionable if all compounds controlling Lemna spp., Arabidopsis, and B. distachyon will, for example, automatically control herbicide-resistant Amaranthus spp. This is presumably why cells or tissues of target species have now found their way into physiological profiling assays (Grossmann et al., 2012).

Environmental Fate of Agrochemicals

Advances in the sensitivity of analytical tools in the 1970s and 1980s allowed the detection of trace levels of agrochemicals in groundwater, and through such analyses some chemicals proved to be persistent in the environment (Kraehmer, 2012). This led to a tightening of regulatory requirements and an increased need for agrochemical companies to test and characterize degradation rates of compounds in soil and water under aerobic and anaerobic conditions. The physicochemical parameters such as vapor pressure and the octanol-water partition coefficients of new compounds are now measured routinely, and increasingly sophisticated toxicology tests are carried out before field testing.

How Has Chemical Weed Control Changed Agriculture So Far?

Herbicide innovations have appeared in waves (Stuebler et al., 2008). The early auxin-type herbicides primarily controlled dicot weeds. Later on, monocots could be controlled with the advent of photosynthesis inhibitors, cell division inhibitors, and very-long-chain fatty acid biosynthesis inhibitors. Yields were considerably increased in all crops between 1940 and 2010 (Table III). Many factors contributed to this increase; one is definitely breeding, but herbicides also had a major impact. The value of herbicides was shown by Zimdahl (2004), who published data showing the influence of defined weed species on the yield of different crops.

Table III. Yields per acre of five U.S. crops between 1940 and 2010.

Source: U.S. Department of Agriculture Crop Production Historical Track Records, April 2013. b, Bushels; t, tons.

Crop	1940	1950	1960	1970	2010
Barley (b)	23	27.2	31	42.8	73.1
Corn (b)	28.9	38.2	54.7	72.4	152.8
Soybean (b)	16.2	21.7	23.5	26.7	43.5
Sugar beet (t)	13.4	14.6	17.2	18.7	27.7
Wheat (b)	15.3	16.5	26.1	31	46.3

Herbicides made it possible to control weeds in crops much more easily than before. Before the advent of herbicides, controlling weeds required hard physical labor such as hoeing. A particular example is weed control in sugar beet (Beta vulgaris), where high labor costs meant that farmers rapidly adopted herbicide technology once it became available. Another crop that could be cultivated more easily with the invention of new herbicides was winter oilseed rape. Broad-spectrum products such as paraquat and atrazine accelerated the use of reduced-tillage measures in agriculture and helped prevent soil erosion, which had become a major issue especially in the United States during the first half of the last century, as mentioned above. Selective postemergence grass control in some crops came in the mid-1970s. The ACCase inhibitors selectively controlled grass weeds in postemergence treatments, and the use of safeners even made their application in cereals possible. Glyphosate is a unique molecule with very specific properties. It allowed the farmer to kill weeds and to plant new seed within a few days after its application (similar to paraquat). It killed perennials and had a short soil half-life. Therefore, it was also regarded as one of the most effective products in plantation crops. Glufosinate was another nonselective molecule with a new MoA. It was strong against a few hard-to-control weeds such as Equisetum spp., and it appeared to be a bit safer to plantation crops than glyphosate. The introduction of acetolactate synthase (ALS) inhibitors drastically reduced the total amounts of agrochemicals applied per hectare. Most of them could be applied with a few grams per hectare, whereas many older chemicals required amounts in the kilogram range. Finally, 4-hydroxphenylpyruvate dioxygenase (HPPD) and phytoene desaturase inhibitors were ideal mixture partners for existing products and closed some evident gaps. No major new MoA has been found by the industry since the 1980s. As early as 1990, agrochemical market research indicated that the crop protection market was approaching maturity and that it was becoming increasingly difficult to discover new agrochemicals with significant advantages over existing products (Kraehmer and Drexler, 2009). The average total costs for the development of an herbicide had increased from $50 million to $250 million U.S. between the years 1975 and 1995 (Rüegg et al., 2007). Genetically modified crops entered the market in the second half of the 1990s. The high standard of existing products in the market and the high registration costs caused many chemical companies to give up their agrochemical business. Following numerous mergers and acquisitions, only a few companies with herbicide research capacity remained. The number of companies devoted to herbicide discovery was reduced from more than 40 in the 1970s to five to eight today. It is highly questionable, therefore, if farmers will experience many new innovations in the years to come, and yet there is an urgent need for new weed-control solutions. Several cropping systems in the Americas as well as in Europe are no longer sustainable without further herbicide innovations, as we will see in our final sections.

MOAs AND HERBICIDE DIVERSITY

There are several excellent reviews and books on the MoAs of herbicides (Seitz et al., 2003; Liu et al., 2009; Dayan et al., 2010; Krämer at al, 2012); therefore, there is no need to repeat what has been published recently. Instead, it is intended to give a short overview of previous and current trends in MoA research and the factors that affected the search for herbicides with novel MoAs.

Between 1960 and 1970, only one herbicide with a novel MoA, asulam, an inhibitor of dihydropteroate synthase, was discovered. In the early 1970s, this period of low innovation with respect to herbicides with novel MoAs was superseded by what, in retrospect, might be called the golden years of herbicide discovery. From 1971 to 1985, herbicides with eight novel MoAs were discovered, among them inhibitors of amino acid biosynthesis (glyphosate, glufosinate, and ALS inhibitors), lipid biosynthesis (ACCase inhibitors), and pigment biosynthesis (phytoene desaturase and HPPD inhibitors), MoAs that today still dominate the herbicide market (Figs. 4 and 5).

Figure 4.

Year of market introduction (Commercial) or publication (Experimental) of the first herbicide with a given MoA. The MoA of the corresponding herbicide might have been elucidated at a later date. The molecular MoAs of cinmethylin and oxaziclomefone are still not fully understood, despite articles describing biochemical effects attributed to these compounds. AdSS, Adenylosuccinate synthetase; AMPDA, AMP deaminase; Auxin, auxin herbicides; Cellulose, cellulose biosynthesis inhibitors; CytOx, cytokinin oxidase; DHPS, dihydropteroate synthase; DXR, 1-deoxy-d-xylulose 5-phosphate reductoisomerase; DXS, 1-deoxy-d-xylulose 5-phosphate synthase; EPSPS, 5-enolpyruvylshikimate 3-phosphate synthase; FPS, farnesyl-diphosphate synthase; GibB, gibberellic acid biosynthesis; GPAT, Gln phosphoribosylpyrophosphate amidotransferase; GS, Gln synthetase; GSAT, Glu semialdehyde aminotransferase; IGPD, imidazole glycerol phosphate dehydratase; IMDH, 3-isopropylmalate dehydrogenase; KARI, ketol acid reductoisomerase; LyCyc, lycopene cyclase; Microtubule assembly, inhibitors of microtubule assembly; Microtubule organization, inhibitors of microtubule organization; ObtDM, obtusifoliol demethylase; PDS, phytoene desaturase; PPO, protoporphyrinogen IX oxidase; PyrDH, pyruvate dehydrogenase; TrpS, Trp synthase; VLCFA, very-long-chain fatty acid biosynthesis; ZDS, ζ-carotene desaturase.

Figure 5.

Herbicide market share in 2010 according to MoA. For abbreviations, see Figure 4.

These herbicides not only had a profound impact on weed control in agriculture but also played a major role in expanding the understanding of fundamental plant processes through their ingenious use as molecular probes (Dayan et al., 2010). Many academic groups in the United States as well as in Europe embarked on this path, either studying the MoAs of herbicides or using herbicides to study plant metabolism.

From the mid-1980s until today, more than 140 new herbicide active ingredients were commercialized (Gerwick, 2010). Surprisingly, only two of these, the narrow-spectrum herbicides cinmethylin and oxaziclomefone, have an unknown and still not fully understood molecular MoA, despite publications describing biochemical effects attributed to these compounds. All the other new active ingredients target old (known) MoAs. The reasons for this lack of novelty have been reviewed recently and have been attributed to a number of mostly economic factors (Kraehmer et al., 2007; Duke, 2012). Two such factors are briefly discussed in the following paragraphs.

Due to their very low application rates in particular, ALS inhibitors easily outcompeted other chemical classes in the herbicide screening programs and drew significant synthesis capacity away from them. In addition to ALS inhibitors, significant synthesis capacity was directed toward one other chemically productive herbicidal MoA, protoporphyrinogen IX oxidase inhibitors, which controlled weeds with very low application rates. Protoporphyrinogen IX oxidase inhibitors had a very broad chemical scope too, but their commercial success was limited. The fast action of these herbicides prevented systemic action, selectivity for in-crop use was lacking, and the toxicological profile was often problematic.

Even though no major new MoA has been introduced into the market in the last 30 years, herbicide discovery has not come to a standstill. At least 14 target sites for herbicidal compounds have been discovered during this time period (Fig. 4). However, no compound interfering with any of these targets has made it to the market. Reasons for the lack of commercialization are diverse and include high application rates, incomplete weed spectrum, high costs of production, or a combination of any of these factors. Furthermore, limited scope for the chemical variation of the inhibitors for some of these targets (i.e. ketol acid reductoisomerase, AMP deaminase, and adenylosuccinate synthetase) hampered their optimization toward high bioefficacy and low field application rates.

The traditional method to discover a novel herbicide was by random screening of large numbers of synthetic compounds in the greenhouse. Based on this approach, at least one commercial product targeting each of today’s marketed MoAs was introduced before its MoA was elucidated. This random discovery approach was challenged in the mid-1990s by the introduction of novel research technologies in the pharmaceutical industry. Most big pharmaceutical companies had started to pick up molecular targets such as enzymes, receptors, etc., which became accessible by the genomic revolution and a deeper understanding of the molecular mechanisms of disease. Pharmaceutical companies tried to identify efficient inhibitors of these targets by using high-throughput in vitro screening. Progress in organic chemistry facilitated this approach, since combinatorial chemistry made it possible to prepare tens to thousands or even millions of compounds within a short time period. Most agrochemical companies also followed this strategy, especially after the full genome of Arabidopsis was published in 2000. Henceforward it became possible to validate putative herbicidal targets by genetic technologies on a large scale. The analysis of the phenotype of a plant in which a certain gene had been knocked out completely or the expression had been reduced to a significant extent by antisense RNA enabled the identification of so-called lethal targets. These targets, mainly enzymes, were subsequently subjected to high-throughput in vitro screening of chemical libraries to identify inhibitors. However, the transmission of the in vitro activity of compounds discovered this way into the greenhouse turned out to be an insurmountable hurdle. Therefore, not only have no herbicides with a novel MoA been discovered but also, to the authors’ knowledge, no compound with broad herbicidal activity at low application rates has been identified in this way.

In the past few years, the focus of herbicide research has once again shifted back toward random screening of synthetic compounds in the greenhouse and to the use of hits as starting points for targeted optimization processes. If a phytotoxic compound with interesting bioefficacy is identified, studies are undertaken to find the underlying MoA as soon as possible and to use this MoA information in rational chemical design approaches. Technological advances in molecular biology, such as gene expression profiling (transcriptomics), have also been adopted for this task (Eckes and Busch, 2007). When a plant is treated with an herbicide, vital processes of that plant are affected. This is reflected by distinctive, MoA-dependent changes in the transcriptome. By comparing the transcriptome of a plant treated with a phytotoxic compound from the research pipeline with a library of response profiles to compounds with known MoAs, it may be possible to classify the compound into one of the known MoAs. If the compound cannot be classified into an already existing MoA, it can be assumed that it has a new MoA, for which the profile might also provide clues.

The lack of herbicide innovation has triggered novel attempts to fight herbicide-resistant weed species. Monsanto is investigating the use of topically applied RNA molecules to induce a process called RNA interference in glyphosate-resistant weeds in order to counteract resistance to glyphosate. It remains to be seen if this approach will be successful. Countries that do not accept genetically modified crops must rely solely on the chemical industry to deliver novel herbicides with novel MoAs. To achieve this goal, previously discovered but not commercialized targets might be reconsidered in the light of new technological advances in drug discovery in the last 20 years, such as fragment (ligand)-based drug design. Alternatively, hits from greenhouse screening might be subjected to MoA elucidation using state-of-the-art omics technologies (Grossmann et al., 2012; Tresch, 2013) in order to identify novel starting points for herbicide discovery.

HERBICIDE SELECTIVITY VIA SAFENERS

For chemical weed control in fields of crops, the herbicide products that can be used must fulfill two contradictory objectives: control the weed plants but not injure the crop plants. Some herbicides provide these features innately (e.g. atrazine for weed control in corn). However, as a general rule, when herbicides are highly active against a wider range of weeds, the chances are much lower that they will also be highly crop selective. Crop sensitivity is itself a complex issue being influenced by many features such as crop variety, application timing, soil properties, or weather conditions. Therefore, some herbicides may be selective in some cases but not in others. Other herbicides may be selective in certain crops but not in others. For this reason, methods have been developed to increase crop selectivity. The two key selectivity technologies are herbicide tolerance traits (either from mutant selection or genetic modification) and safeners. Safeners (sometimes called antidotes or protectants) are chemicals that prevent herbicidal injury to crop plants without reducing weed control.

Safener Commercialization

Based on an accidental observation in 1947 by Otto Hoffmann of the Gulf Oil Company, the concept that certain chemicals could increase the tolerance of plants to herbicides was born (Hoffmann, 1953). Hoffmann had seen that tomato (Solanum lycopersicum) plants treated with 2,4,6-trichlorophenoxyacetic acid appeared to be protected from 2,4-dichlorophenoxyacetic acid vapor injury. Whether this was the first time an herbicide researcher had seen such effects is not known. However, in this case, Hoffmann realized that this may provide a useful tool to increase crop tolerance to herbicides. Gulf Oil initiated a research program that eventually led to the first commercial compound, 1,8-naphthalic anhydride (NA). When applied as a seed treatment to corn, this compound gave protection against various preemergence- and preplant-incorporated herbicides from the thiocarbamate class (e.g. ethyl dipropylthiocarbamate). Subsequent research in various agrochemical companies has subsequently provided nearly 20 commercial safeners (Hatzios and Hoagland, 1989; Davies and Caseley, 1999; Davies, 2001; Rosinger et al., 2012; Jablonkai, 2013).

The early phase of safener commercialization during the 1970s and early 1980s was dominated by seed treatment or soil-active (preemergence) safeners (Figs. 6 and 7). Seed treatment-applied safeners offer certain advantages and disadvantages. A key advantage is that the safening only influences the crop plant and not the weeds. Therefore, the safener does not need to be crop specific per se. However, a key disadvantage is that the treated crop may be sprayed subsequently with herbicides from other competitor companies, making value capture difficult. Conversely, preemergence tank-mix or coformulated safeners must be innately crop specific (i.e. with no antagonism on weeds) but offer significantly simplified safener technology for the farmer and manufacturer. The product could be used just as if the herbicide itself were selective. The manufacturer has better control of the performance of the product containing its herbicides.

Figure 6.

Seed treatment safeners.

Figure 7.

Preemergence tank-mix safeners.

In the late 1980s, a significant innovation leap occurred whereby “leaf-active” safeners were developed for coformulation with postemergence herbicides. Fenoxaprop-ethyl was launched in 1984. This molecule controlled grasses with postemergence selectively in broad-leafed crops, but it lacked sufficient selectivity in cereals, being moderately damaging to wheat (Triticum aestivum) and more severely damaging to barley (Hordeum vulgare). Therefore, a research project with the aim of a safener for postemergence use with fenoxaprop-ethyl in cereals was started in the 1980s. A new class of substituted triazole safeners was found from which fenchlorazole-ethyl was developed and introduced in 1992 as a postemergence coformulation product with fenoxaprop-ethyl (Bieringer et al., 1989). This product meant that fenoxaprop-ethyl could now be used for grass weed control in wheat and rye (Secale cereale). However, barley was not sufficiently safened by fenchlorazole-ethyl, so testing for further improved safeners was continued. Relatively quickly, a new class of safeners was identified with a modified central chemical scaffold (pyrazoline) from which mefenpyr-diethyl was developed (Hacker et al., 2000). This compound had a much stronger safening activity in cereals than fenchlorazole-ethyl, making it possible to also use fenoxaprop-ethyl in barley. Among other ACCase inhibitors was clodinafop-propargyl, which was launched in 1991. Like fenoxaprop, clodinafop had strong grass weed activity but lacked sufficient crop tolerance in cereals. Ciba-Geigy did not launch clodinafop without a safener. They used the highly effective safener cloquintocet-mexyl, which has a quite different (quinoline) scaffold from fenchlorazole or mefenpyr (Amrein et al., 1989). The structures of the above-mentioned ACCase inhibitors and safeners are shown in Figure 8.

Figure 8.

ACCase inhibitors and the associated safeners for postemergence use in cereals.

The safener innovation required to overcome the cereal injury meant that by the end of the 1990s, mefenpyr-diethyl and cloquintocet-mexyl were well established in the cereal market and the two safeners could be tested with new research compounds. At the end of the 1990s, iodosulfuron and mesosulfuron were introduced for use in cereals (Hacker et al., 1999, 2001). Both had selectivity limitations that could be overcome by using mefenpyr-diethyl. Cloquintocet-mexyl was also used as a safener for pinoxaden, an herbicide from the novel aryl-1,3-dione class of ACCase inhibitors (Hofer et al., 2006). More recently, the safening of mefenpyr-diethyl in cereals was used as the selectivity technology for the HPPD inhibitor pyrasulfotole (Schmitt et al., 2008) and the ALS inhibitor thiencarbazone. Since the basic patents for mefenpyr and cloquintocet have now expired, other companies have used them in their own safened products. Of particular note is a coformulation of the ALS inhibitor pyroxsulam with cloquintocet as a safener (Wells, 2008).

During the 1990s and 2000s, safener research was continued. One target was the extension of the fenoxaprop market into rice (Oryza sativa), which was not tolerant to this herbicide. Screening efforts identified the safener isoxadifen-ethyl (Fig. 9), which again retained the carboxylic acid ester of fenchlorazole-ethyl and mefenpyr-diethyl, but with a new heterocyclic isoxazole core. Although fenoxaprop in rice was the original target, early on in the research and development phase it was found that isoxadifen-ethyl could also safen postemergence against the sulfonylurea herbicide ethoxysulfuron in rice and foramsulfuron and iodosulfuron in corn (Collins et al., 2001; Pallett et al., 2001; Hacker et al., 2002). Therefore, isoxadifen now represents the current pinnacle of multicrop and multiherbicide safening.

Figure 9.

Structures and use of isoxadifen-ethyl and cyprosulfamide.

The most recently launched safener is cyprosulfamide (Fig. 9). Like isoxadifen-ethyl, it is strongly active postemergence in corn but has the significant advantage, for certain herbicides, of also having strong preemergence safening activity. Cyprosulfamide comes from a class of safeners (acyl sulfonamides) discovered during the mid-1990s. It is now used to support a significant number of different corn herbicides (Philbrook and Santel, 2008; Santel and Philbrook, 2008; Watteyne et al., 2009).

Safener MoA

Theoretically, safeners could increase crop tolerance in two main ways. They may reduce the amount of herbicide reaching the active site or interfere with the herbicide interaction at the target site. Both of these possibilities have further subdivisions (discussed below), and the MoA of safeners has been extensively investigated and reviewed (Davies and Caseley, 1999; Riechers et al., 2010; Rosinger et al., 2012).

Several studies found no significant interaction between safener and herbicide at the target site (Polge et al., 1987; Köcher et al., 1989). However, Walton and Casida (1995) reported competitive binding of the dichloroacetamide safener R-29148 and the herbicides EPTC and alachlor at a protein in corn extracts. This, together with good correlation between competitive binding of dichloroacetamide compounds and their safening activity, was taken as support that some safeners act as receptor antagonists for the herbicides. Another MoA might be the safener causing the crop to increase the activity of the herbicide target. Rubin and Casida (1985) found that the safener dichlormid caused an increase in ALS activity in corn. Milhomme and Bastide (1990) and Milhomme et al. (1991) also found an increase of ALS levels in corn treated with the safeners NA and oxabetrinil. However, Barrett (1989) could not find any enhancement of ALS activity in corn and sorghum (Sorghum bicolor) seedlings after treatment with the safeners NA, oxabetrinil, flurazole, and dichlormid. It is now felt that interactions between the safener and herbicide at the target site play, if anything, only a small role in safener MoA.

Regarding the possible effects of safeners on herbicide uptake, Davies and Caseley (1999) extensively reviewed the literature for herbicide/safener combinations and concluded that most cases showed no reduction in uptake. Subsequent studies carried out with the safener mefenpyr-diethyl and the sulfonylurea herbicides mesosulfuron-methyl and iodosulfuron-methyl sodium also found no effect on herbicide uptake (Köcher, 2005).

For postemergence cereal herbicides, which have long-distance transport between the treated leaf and the sensitive meristems, safening action may involve the inhibition of this transport. So far, no case is known in which a safener directly interferes with the long-distance translocation of these herbicides. However, the final proposed safener mechanism is enhanced herbicide metabolism/detoxification within the safened crop, which could indirectly reduce the amount of active herbicide reaching the target site. It is well understood that differences in herbicide metabolism play an important role in herbicide selectivity (Hatzios and Penner, 1982; Drobny et al., 2012). Therefore, it is not surprising that this was proposed as a possible safener mechanism as early as the 1970s and has been studied extensively since that time. Hatzios and Hoagland (1989) reviewed in detail the early studies, especially on the seed treatment safeners and preemergence safeners. This showed a large amount of evidence that safeners were indeed increasing the rate of herbicide metabolism in the safened crop plant. For more recent postemergence safeners, metabolism studies combined with gene expression analysis has strengthened the understanding of safener MoA. In wheat (but not the target grass weed), the level of fenoxaprop-ethyl and the free acid fenoxaprop declined (conversion to inactive metabolites) more rapidly when plants were treated with fenchlorazole-ethyl (Köcher et al., 1989; Yaacoby et al., 1991). For the safener cloquintocet-mexyl, the safening of wheat against clodinafop-propargyl was also reported to be based on enhanced detoxification (Kreuz et al., 1991; Kreuz, 1993). Mefenpyr-diethyl also enhanced the rate of conversion of fenoxaprop to nonphytotoxic products in wheat but not in target weeds (Hacker et al., 2000). Similarly, mefenpyr-diethyl enhanced the rate of metabolic degradation of iodosulfuron-methyl sodium and mesosulfuron-methyl in cereals but not in the grass weeds tested (Hacker et al., 1999, 2000, 2001). In corn, the postemergence safener isoxadifen-ethyl has also been shown to enhance the metabolism of foramsulfuron (Pallett et al., 2001; Hacker et al., 2002).

Herbicide metabolism in plants involves several enzyme-mediated steps (Hatzios, 1991; Van Eerd et al., 2003). The first step may sometimes be a conversion of an inactive proherbicide to the active herbicide. However, the metabolic steps relevant to safening are those that convert the active herbicide to inactive metabolites. Cytochrome P450 enzymes may catalyze the reduction, oxidation, or hydroxylation of the herbicide and thus provide functional groups for further metabolic steps. These often involve conjugation reactions between the first metabolite and glutathione, Glc, or amino acids that are catalyzed by multifunctional enzyme, glutathione S-transferases, or glycosyltransferases (Lamoureux et al., 1991; Armstrong, 1994; Marrs, 1996; Hatzios, 2001). Additionally, conjugation with malonate via malonyl-CoA has also been reported (Sandermann et al., 1997). The next step in herbicide detoxification is the transport of the conjugates into the vacuole, which may be catalyzed by ATP-binding cassette transporters (Tommasini et al., 1997). Once in the vacuole, the conjugates may be further degraded by peptidases with the removal of Gly and Glu. The speed of herbicide metabolism/detoxification, therefore, may depend on the level of activity of several key enzymes and, thus, the level of expression of genes that code for these enzymes. Advances in biotechnology over the past two decades have allowed increasingly detailed studies into the effect of safeners on gene expression. It has been shown that genes encoding for enzymes involved in herbicide detoxification are induced within a few hours of safener application (Kreuz et al., 1996; Cummins et al., 1997; DeRidder et al., 2002; Scalla and Roulet, 2002; Theodoulou et al., 2003; Zhang et al., 2007).

Gene expression studies have indicated parallels between the oxidative stress-related oxylipin pathway and safener signaling (Riechers et al., 2010). They have also indicated possible overlaps with the plant stress defense signaling pathway involving salicylic acid. For example, it could be shown that many safener-regulated genes are induced by salicylic acid (Behringer et al., 2011). Therefore, while the primary site of safener reception is still not known, it seems that several signaling pathways contribute to the full safener response in plants.

FUTURE PERSPECTIVES

There have been very few new safener structures patented in the past decade, suggesting that little research into new compounds is being conducted in the agrochemical industry. This is probably mainly due to the emergence of genetically modified herbicide tolerance as a major selectivity tool (e.g. glyphosate-tolerant Round-Up Ready cotton [Gossypium hirsutum], soybean, and corn) as well as the availability of strong existing safeners for use with newer herbicides. However, as the problem of weed resistance increases and the existing safeners reach the limit of their potential, stronger next-generation safeners may indeed be required for future selective weed control products. As biotechnology methods develop quickly, it is possible that the target site(s) of safeners will be identified in the not too distant future. It may then perhaps be possible to use rational design to identify new safener structures.

Glossary

MoA

mode of action

ACCase

acetyl-coenzyme A carboxylase

ALS

acetolactate synthase

HPPD

4-hydroxphenylpyruvate dioxygenase

NA

1,8-naphthalic anhydride