Increasing ecological heterogeneity can constrain biopesticide resistance evolution

Abstract


Biopesticides are important products for ecologically sustainable crop protection
The prevalence and adaptive capacity of insect pests cause huge problems for food security.
Insects consume as much as 20% of crops while growing or in storage [1], which represents a large fraction of the future improvements needed to feed the growing human population [2].
Despite considerable research, pest control methods continue to suffer reduced effectiveness due to pesticide resistance (See Glossary), leading to crop failures, economic losses, and food insecurity [2][3][4].New pest control technologies, including microbial biopesticides and other biological control agents, provide a welcome addition to the arsenal of crop protection methods [5].These biological agents are vital tools in integrated pest management (IPM) and can be used instead of synthetic products that have suffered resistance evolution or legislative restriction.Biocontrol provides attractive crop protection options due to minimal adverse effects on human health, promotion of ecosystem services, and compatibility with organic farming requirements [6].Globally, biopesticide use is increasing by almost 10% per year [7].Although the worldwide microbial biopesticide market was previously dominated almost exclusively by products based on Bacillus thuringiensis (Bt) (95% in the 1990s), the range of microbial bioinsecticides has increased substantially [8].In this article we focus on microbial biopesticides formulated from the living pathogens of insect crop pests.We argue that resistance management approaches need to be implemented for these crop protection products and propose new solutions.We place our argument into the context of other welldeveloped resistance management frameworks that are already implemented for synthetic insecticides and transgenic crops (such as those incorporating Bt pesticidal molecules).

Biopesticides present new opportunities for resistance management
To deliver ecologically sustainable crop protection the transition from synthetic pesticides to microbial biopesticides and other forms of biological control must accelerate; this will require new biological agents to come to market accompanied by increased adoption by farmers.However, expansions in microbial biopesticide usage will increase selection pressures on pests to develop resistance and therefore justifies careful consideration of approaches to proactively mitigate the risks of resistance evolution [9].Although resistance management for transgenic crops is well developed [10,11], until now resistance management for living biological control agents has been relatively neglected.
Notwithstanding early assertions that they would not incite resistance evolution [12], there is already considerable evidence that resistance to microbial biopesticides can evolve in the field and also in lab studies (Box 1).Following deployment of granulovirus based insecticides for control of codling moth (Cydia pomonella) in European apple orchards in the early 1990s, resistance developed by the mid-2000s requiring development of novel products with different viral strains [13].Perhaps the most famous viral biocontrol agent of all time, Myxoma virus, rapidly triggered resistance evolution in rabbits (Oryctolagus cuniculus) by selecting on pre-existing variation in immune system genes [14].Similarly, some major pest species have developed resistance to the most widely used bacterial biopesticide in the world: B. thuringiensis [15].Alarmingly, there is even recent evidence of substantially elevated resistance to classical biological control parasitoids under field settings [16].With this increasing evidence base, microbial biopesticides must be protected to avoid them suffering the same resistance fate as chemical pesticides.

How is resistance to microbial biopesticides different to other crop protection products?
In the case of synthetic pesticides and transgenic crops, resistance alleles frequently have binary effects on phenotype conferring orders of magnitude decrease in susceptibility to the agent.This is in part due to the relatively simply nature of molecular interactions between these control products and the pest molecules they target, which enables resistant phenotypes to arise from genetic changes at single or a small number of loci [17,18].[19,20].In contrast, resistance to living microbial biocontrol agents should involve more genes because living organisms are by necessity more complex than individual biomolecular compounds.It is worth noting for example that while resistance to living B. thuringiensis is rare, resistance to the specific insecticidal proteins produced by transgenic crops occurs more frequently [21].
Resistance to living organisms such as those in fungal biopesticides will therefore often be determined by multiple gene loci, where individual alleles may have only small effects on susceptibility [22]; as a consequence, susceptibility to such biopesticides typically varies continuously among individuals [23].
However, drawing general conclusions about resistance to all biopesticides is difficult because biopesticides and other biocontrol agents encompass a wide spectrum of natural enemies for which the evolutionary basis of pest resistance differs considerably.In Box 1 we place microbial biopesticides containing living agents in the context of a continuum of crop protection approaches for which the complexity of the genomic architecture of resistance varies from relatively simple (e.g., chemical insecticides and some transgenic crop varieties) through to complex (e.g., insects deployed in classical biological control).This variation in genetic complexity has profound consequences for resistance evolution.

Established resistance management strategies
The evolutionary genetic assumptions underlying classical resistance management theory are that resistance is usually genetically simple and underpinned by one or a few loci; alleles conferring resistance are rare (and therefore predominantly present in heterozygotes); and resistance alleles confer fitness costs in the absence of the pesticide, creating trade-offs to pesticide resistance [24].Crucially, while these assumptions are generally supported for synthetic pesticides [25], they probably do not hold for some classes of biopesticides (Box 1), which necessitates a different approach to microbial biopesticide resistance management.
Whilst the complexity of the genetic interactions between biopesticides and their hosts may reduce the risks of resistance evolution, we do not think that this complexity on its own is sufficient to prevent resistance evolution in many ecologically homogeneous agricultural landscapes.
Resistance management targeted toward synthetic insecticides and transgenic crops has a long pedigree in research and agricultural application.These resistance management strategies can be placed into three broad groups.
First are strategies that seek to limit the opportunities for resistance alleles to spread in spite of selection for resistance: in GM-crop systems, crop refuges encourage resistant individuals to mate with susceptible individuals to generate susceptible offspring; whereas pyramid-Bt varieties express multiple toxins with the aim that single step mutations will not confer resistance.
Second, many approaches aim to reduce the long-term intensity of selection for pesticide resistance, for example by minimising pesticide use through the adoption of IPM alternatives, or through temporal pesticide rotations where a single active ingredient is only used intermittently.These strategies rely on the principle that resistance alleles only confer high fitness in the presence of one chemical agent; then in the absence of that agent, costs of resistance cause allele frequencies to gradually decline.By cycling through pesticides with distinct modes of action farmers could keep resistance at a low level.Unfortunately, costs of resistance can be inconsistent across habitats [25], which can hamper their ability to constrain resistance evolution.
Third, a conceptually more attractive modification of standard pesticide rotations involves alternating between groups of pesticides that exhibit "negatively correlated cross-resistance" (hereafter NCC-R), in which alleles conferring resistance to one pesticide directly impair the ability to resist another, resulting in strong trade-offs [26].This approach differs fundamentally from a pesticide rotation because the management strategy is designed to drive down the frequency of resistance alleles (using an alternative pesticide) rather than simply relying on the general fitness costs of resistance to erode previous partial selective sweeps of resistance.Despite this theoretical promise, the ability of these trade-offs to prevent resistance evolution has not often been realised [27]: even if two pesticides confer NCC-R, the genetic associations that produce trade-offs can themselves evolve over time, and lead to positive cross-resistance (in which insects resistant to one pesticide are also resistant to others) [28].When these genetic associations involve a small number of loci, recombination to produce positive cross-resistance can happen relatively easily, meaning the efficacy of NCC-R in managing resistance can be short lived [29].However, the promise of NCC-R for generating variable selection is much greater for control methods for which resistance is under complex polygenic genetic control, such as for biopesticides containing living organisms.
Until now, strategies to manage pesticide resistance through heterogeneous selection pressures have principally sought to achieve it by creating diversity in the selective agents themselves, and in their presence or absence.Here, we suggest that inconsistent selection for resistance to a (potentially) single agent can be delivered by diversifying other aspects of the agricultural environment.

Why the evolutionary ecology of pathogens is particularly prone to inconsistent selection
Strong selection pressures do not always drive rapid evolutionary change [30,31].Natural host-parasite systems illustrate how variable selection can sustain genetic variation for infection susceptibility despite strong selection.Even though successful parasite defence must provide a major fitness advantage, host populations almost ubiquitously exhibit high genetic variability for parasite resistance traits [32].In some host-parasite systems coevolutionary interactions prevent resistance allele fixation through Red Queen Dynamics [33,34] (Note that in contrast to natural systems, coevolution between biopesticides or inundatively released biocontrol is impossible because the control agent is grown from stock in the lab, rather than cultivated).However, more generally, selection in the tangled bank of ecological systems is inconsistent due to environmental variation [35].Parasites can exert strong selection on hosts without driving fixation of resistance alleles because parasites are usually genetically diverse, host-parasite interactions are often mechanistically complex, and the outcomes of these interactions are frequently context-dependent.This context dependence has been quantified in the form of "genotype-by-environment" interactions (GEIs), in which the fitness of resistance alleles depends on the specific environment an organism inhabits (Box 2).
Unfortunately, most modern agricultural cropping systems are highly homogenous, which means that selection does not vary dramatically at a landscape scale (whole farms, and indeed farming regions, frequently specialise on growing a narrow range of crop plants).Yet, natural systems are far more diverse, meaning that multiple aspects of the environment vary continuously, including the ambient conditions, the nature and quality of food, the presence of symbiont, and the genotypes of competitors, pathogens and predators.It is this diversity that favours unique multilocus genotypes at many different loci depending on the precise ecological context.Can modern agricultural landscapes be engineered to similarly benefit from the power of GEIs, to sustain genetic diversity and prevent resistance evolution to biopesticides?

Successfully exploiting GEIs for pest resistance management
Our vision is to exploit GEIs to make biopesticide-based pest control more ecologically and evolutionarily sustainable.The orthodox framework for pesticide resistance management focusses on trying to delay evolution.While this approach limits pest adaptation, we instead advocate harnessing the evolution of pests using the variable selection pressures generated by heterogeneous landscapes.By keeping aspects of the pest landscape in sufficient flux, selection for resistance will not be directional at a landscape scale: as the agricultural habitat changes, the alleles favoured by selection will also change.This way local selection in any one generation will result in evolution that takes the population away from the optimum genotype to survive pest control measures in other distinct patches (or times) within the heterogeneous landscape.Importantly, our approach need not require sacrificing some of the crop as refuge, or foregoing pest control altogether, provided that the conditions under which control occurs are sufficiently diverse to prevent directional selection on a landscape scale.
The heterogeneity we call for to manage biopesticide resistance will require altered farm management at a landscape scale, but need not be substantially at odds with agricultural productivity.Both temporal rotations and spatial rotations of heterogenous landscape patches could be used by farmers to generate the required inconsistent selection (Box 3).The greater the difference between two habitats in which a pest lives, the more likely it is that multi-locus genotypes that promote performance in one habitat negatively affect performance in the second.Our approach will be maximally effective if habitat patches differ in as many ecological dimensions as possible.However, we recognise the tension between maximising heterogeneity and maximising farm efficiency.
An obvious way of generating heterogeneity is to alternate the species or strain of pathogen used in biopesticide products (as in chemical insecticide rotations, and consistent with IPM).
Trade-offs for resistance are nearly ubiquitous in host-pathogen interactions [36] and derive mainly from two sources.First, strong resistance specificity means that combatting one pathogen can make an organism more susceptible to others (mirroring concepts of NCC-R).
Second, investment in resistance may deprive organisms of the ability to invest in other life history traits like reproduction and growth.Strong GEI for pathogen resistance is much more likely than for chemical insecticides due to this specificity and the typically polygenic genetic basis of resistance to natural enemies [37,38].The biopesticide market is currently dominated by products containing a relatively narrow diversity of pathogen strains [39].Unlike synthetic insecticides, where the development of new products with novel modes of action is usually slow, the biological world provides us with an almost limitless array of natural pathogen strains which could be harnessed as biopesticides.
It is our opinion that microbial biopesticides offer a highly novel way to generate inconsistent selection for resistance.Farmers could alter other landscape dimensions (in addition to altering pest control methods) for which pest fitness traits are likely to be underpinned by complex multi-locus genotypes.Potential examples include the microbial community associated with crop plants, or the pest diet (determined by crop varieties or crop species in the case of polyphagous pests).Such environmental contexts are well known to change selection on resistance genes: costs of resistance to B. thuringiensis are environmentally dependent and vary depending on crop plant type [40], furthermore, exposure to additional pathogens may help sustain genetic variation in resistance to Bt insecticidal proteins [41].
The options for crop diversification to generate GEIs for biopesticide resistance may be greatest for polyphagous pests, not least because their interactions with different host plants are likely to involve many genes.Polyphagous pests are among the most notorious species for resistance evolution to chemical insecticides [42], perhaps due to prolonged coevolution with the diverse secondary compounds plants have evolved for their own defence [43].
Because polyphagous insects so readily evolve resistance to synthetic insecticides and transgenic crop varieties (e.g., see Table 1), biopesticides are particularly valuable control agents for these species.We see clear opportunities for generating fluctuating selection on these pests through GEIs if farmers diversify the crop species cultivated in the agricultural landscape (e.g., see Box 3).
Many insect pests are active dispersers generating considerable gene flow among populations.Therefore, the heterogeneity we endorse need only be coarse-grained across the farming ecosystem.Whilst the precise details will vary between pests, heterogeneity at the between-field or between-farm scale would probably be sufficient to forestall resistance evolution for most pests.Although there may be additional benefits to finer-scale heterogeneity (such as field margins, refuges or intercropping, which provide welldemonstrated ecological benefits [44,45]) these are unlikely to be necessary to manage resistance.Our approach of managing resistance evolution risks through crop heterogeneity may therefore mean that resistance management strategies could deliver the parallel ecological benefit of enhancing agricultural biodiversity to maximise ecosystem service delivery, further incentivising the diversification of agricultural landscapes [46,47].

Concluding remarks
Pesticide resistant insects are among the most important and expensive obstacles to food security.Conventional chemical pesticides will continue to face heightened regulation and scrutiny, resulting in fewer products on the market, and creating more opportunities for new biopesticides.It would be a mistake to continue to intensively overuse individual microbial biopesticide products, and thereby hasten resistance evolution.Instead, we must protect these emerging pest control products to avoid the same problems of resistance as chemical pesticides.From the industry perspective, it would be beneficial to create incentives for the development of novel products in parallel (rather than launching new products only once legislation or resistance has rendered previous products obsolete), and to alter licensing frameworks to make registration of new biopesticides more straightforward.Such actions will require care, especially in light of the highly variable and uncertain global pesticides market [49].Our proposal requires further research (see Outstanding Questions) and also presents some challenges in adjusting prevailing attitudes on the importance of diversity in the market and landscape.However, the promise of our approach justifies further effort: a landscape that does not sacrifice livelihoods, environment, or food quality, but that in its embrace of diversity makes for more resilient and evolutionarily sustainable food production.

Box 1. Comparing risks of biopesticide resistance evolution
Biopesticides include a wide range of active ingredients that differ substantially in mode of action and the biochemical complexity of their interactions with pests.These differences have important consequences for the risk of pests evolving resistance.In Figure I below, we organise control methods from biochemically simple (at the left) to biochemically complex (at the right), placing biopesticides (white columns) in the context of other crop protection approaches (grey columns).The first three table rows present the existence of evidence for insect resistance (in the field [3,13,15,16,50,51]; in laboratory selection studies [38,[51][52][53][54][55][56]; and standing genetic variation in insect populations that selection could act on [51,[57][58][59][60][61][62][63][64]).
Next, we present a general estimate of complexity of genetic architecture of resistance to different pest control measures.We emphasise that for commercially produced biocontrol agents, coevolution with the pest does not occur and therefore the genotype of the biocontrol organism remains relatively constant.Finally, we present our synthesis of the overall risk of pest resistance evolution for each crop protection technology.Resistance to biological agents is usually not a binary condition.Instead, due to the frequently polygenic nature of resistance, insect populations often tend to display a relatively continuous distribution of susceptibilities across different genotypes.Agents at the left-hand side of the figure have relatively simple molecular interactions with target pests and are at greater risk of eliciting resistance evolution than those agents towards the right.Biopesticides based on molecules derived from organisms (e.g., spinosad) may not require meaningfully more complex resistance mechanisms than synthetic pesticides [17], and therefore, any selection can rapidly drive heightened resistance.Insect resistance to viruses can sometimes be based principally on just one or two loci [65][66][67], whereas for other viruses more genes are implicated [61,68].Studies of resistance to bacterial and fungal infection in insects generally suggest a more complex genetic basis that can involve in excess of ten gene loci [22,62,69].One can imagine that resistance to macro-parasites such as nematodes and parasitoids is probably also generally genetically complex; nevertheless, the few existing studies have suggested simple genetics underpinning resistance [63,70,71].Even for predators, the genetics of resistance evolution can sometimes involve few genes of major effect, as illustrated by the famous case of industrial melanism in British peppered moths (Biston betularia) [72].Studies of the genetics of resistance are undertaken under simplified laboratory conditions and will therefore underestimate the number of loci involved in the field.Whilst clearly there is variation among different classes of biological enemies, on average we contend that the genetic basis of resistance is generally more complex than for synthetic insecticides.Therefore, the greater number of genes involved in resisting attack should make directional resistance evolution more difficult, because coadapted gene complexes tend to be broken up every time meiosis occurs [73].

Box 2. Genotype by environment interactions for pathogen resistance
There is widespread evidence that the ability of any one genotype to defend against pathogen infection depends on environmental parameters [31].Thus, the most effective genotype for parasite defence in one environment may not be the optimal genotype to survive infection for hosts exposed to a different set of environmental conditions (Figure II).This change in the relative fitness of resistance genotypes is a genotype by environment interaction (GEI) [74].For example, the optimum host genotype for pathogen defence often depends on the species of pathogen [75] or on the genetic strain of a given pathogen species [76]

Genotype-by-environment interactions:
A phenomenon in which the fitness of alleles depends on the environment in which those alleles are expressed.

Gene flow:
The introduction of new genetic material from one population to another through dispersal.
Integrated pest management: An ecosystem-based strategy for long-term crop damage reduction through techniques that seek to minimise economic and environmental risks.
Microbial biopesticides: Pesticides containing living microorganisms (viruses, fungi, bacteria, or nematodes) as the active ingredient in the formulation.

Mode of action:
The mechanism by which a pesticide attacks a pest.For synthetic pesticides, this is frequently one or a few target biomolecules, while for living biopesticides, there may be many modes of action that are not as easy to describe simply in biochemical terms.

Molecules derived from organisms:
The use of naturally occurring products e.g., pheromones, plant extracts/oils, or natural insect growth regulators, to control pests.
Pesticide resistance: An increased probability of pest survival and reproduction in the face of crop protection methods.To the extent that such abilities are heritable, repeated exposure to pesticides can lead to evolutionary changes in pest populations that collectively cause failures to achieve the expected level of control.

Pesticide rotation:
The alternating use of different pesticides (with distinct modes of action) in different areas of a farm or at different times to control target pests, in such a way that the pest population is not constantly exposed to the same pesticide.
Polyphagous pests: Pests that can feed on crops belonging to many diverse taxonomic groups.
Red Queen Dynamics: Continuous coevolutionary arms race between hosts and parasites, whereby parasites evolve to be more infectious to hosts and hosts evolve to keep pace in their ability to defend against infection.This process maintains genetic diversity for defence and attack genes in the host and parasite population respectively.
Refuge: An area of crops on a farm in which no pesticides are applied, serving to weaken selection for pesticide-resistance in pests and to serve as a reservoir for pesticide-susceptible alleles.
Resistance evolution: Improvement over time in the genetic propensity of a pest population to cope with pest control measures after repeated exposure to the control agent.
Parasitoids: Insects that use an insect host to develop, resulting in the death of the host.
Shifting mosaic: the sowing of alternating crop species through both space and time in a local area, in such a way that the landscape is both spatially and temporally diversified.

Figure I .
Figure I.The biocontrol continuum of resistance risks.Ticks denote the existence of evidence for a given factor;  indicates the absence of evidence, Speedometer dials show risks of resistance evolution (red = highgreen = low).
coinfection of the host by other pathogens[83].Manipulation of variables such as these in agricultural landscapes could be used to manage the threats of resistance evolution to microbial biopesticides used for crop protection.

FigureBox 3 .
Figure II.Four host genotypes are shown by different coloured lines; the resistance rank order of the genotypes varies between the two environments making selection for pathogen resistance inconsistent.

Table 1 :
Major agricultural arthropod pests that are polyphagous and the number of host plant genera they feed on[48].Polyphagous species may be particularly well suited for biopesticide resistance management by manipulating crop plant diversity in the landscape.
. Another major driver of GEIs for pathogen resistance in ectotherms is environmental temperature, where individual genotypes are best able to defend against infection only over a specific range of temperatures[77,78].Furthermore, the relative ability of host genotypes to defend against infection can strongly vary between different host diets[79].Nevertheless, the ability of environmental variables such as host diet and temperature to drive GEIs is apparently not universal[80,81].A further cause of inconsistent selection on host resistance genotypes is that fitness conferred by a particular host genotype can be dependent on the presence and genetic identity of symbiotic microbes within the host [82] and may also be influenced by