Colloquium Participants

Selection pressure ultimately drives resistance evolution

Disrupting features essential to life for pathogens or pests places enormous selection pressure on them to subvert or evade the treatments. Selection pressure is any force that reduces reproductive success in individuals from a population. Any individuals possessing a trait that mitigates the negative effects of a drug or pesticide are able to proliferate and become dominant. Such traits confer resistance, and frequency of the resistance trait in a population ultimately determines the point at which a treatment loses effectiveness and is no longer useful in controlling the target population. In an ideal therapeutic approach, a naïve population of pests or pathogens would be treated with a novel compound and all individuals would succumb. However, all populations have some level of pre-existing diversity, and some variants less vulnerable to the treatment agent usually exist before any exposure to the treatment. Even in a population arising clonally such as a cancerous tumor or an infection by a single virus, bacterium or fungal spore, every cell division provides an opportunity for genetic changes to occur via mutation. Normally these mutations, arising through mistakes in proof-reading during DNA replication or chromosome crossover events, might engender a fitness cost for the individual and its progeny compared to the rest of the population. But if the genetic mutation that confers resistance is not lethal in the absence of treatment, resistant individuals could survive as a small minority in the population before treatment begins. Once the treatment kills a large percentage of the susceptible portion of the population, these resistant individuals and their progeny flourish.

UNLESS THE FITNESS COST IS HIGH OR THE POPULATION SIZE IS SMALL, IT IS LIKELY THAT THERE WILL BE RESISTANT VARIANTS EXISTING IN A POPULATION, AND THESE RESISTANT VARIANTS WILL BE SELECTED FOR WHEN TREATMENTS ARE APPLIED.

For bacteria, viruses, and parasites, a fitness cost can mean that the resistant individual is less competitive in establishing an infection or growing within a host than its wild type counterparts. Other types of organisms, like plants, do not have such dual lifestyles, and so their fitness costs may involve reduced growth, survival, or reproduction. Although fitness costs and selection pressure manifest in individuals, the rate at which their effects become evident at the population level depends upon generation time, which varies across biological systems.

The fitness differential, then, between susceptible and resistant individuals in the different environments encountered by the organisms (e.g., with and without presence of a treatment) ultimately affect the spread of resistance. In absence of a treatment, such fitness differentials can vary greatly from species to species and for different mutations within a species, and some mutations conferring resistance may not even have costs associated with them. Fungi, for example, often seem to survive perfectly well with various chromosomal abnormalities, although survival may still be influenced by environmental characteristics. The take-home message is that unless the fitness cost is high or the population size is small, it is likely that there will be resistant variants existing in a population, and these resistant variants will be selected for when treatments are applied. Thus, the ideal scenario of a 100% vulnerable target population that can be eliminated in one treatment is unfortunately rare, especially when one considers that in most situations, many different populations of a pest or pathogen are exposed to treatments.

Sometimes, withdrawing the treatment that is generating the selective advantage for resistant individuals can allow the non-resistant population to return to dominance. Although not a universal principle, this approach worked in Malawi, when national policy dictated a removal of chloroquine to reduce malarial resistance to the drug. After a period of ten years, chloroquine was reintroduced and the parasite population was again susceptible. Success depends on the ability to withdraw the treatment completely and on the disappearance of the resistant individuals. An attempt to reduce chloroquine resistance in Latin America was unsuccessful because the drug was never fully abandoned and so selection for resistance persisted. Similar attempts with antivirals in HIV were unsuccessful as the resistant virus is never truly gone; HIV virions are able to persist in long-lived memory immune cells.

Common molecular mechanisms of drug resistance

The molecular mechanisms through which resistance can arise are often shared across biological systems.

• ▪ POINT MUTATIONS: Resistance can occur through point mutations — discrete substitutions, deletions, or insertions in the genetic code. Such mutations can often be found in the active site of the target protein, upon which the drug is designed to act. In some instances, a single point mutation can engender the resistance phenotype, but in the case of malarial resistance to the drug chloroquine, seven such mutations are necessary. Mutations, however, do not always occur in the drug target site, but can also be found in compensating pathways, effectively bypassing the drug's cellular effect.
• ▪ GENE AMPLIFICATION: Gene amplification is another common resistance mechanism frequently associated with cancers, and also found in pesticide and antifungal resistances. Through this mechanism, the organism or cells increase the overall copy number of the target gene such that sufficient amounts of the resulting protein are available to complete the biological task, despite a subset being hindered by the drug treatment. Gene amplification can also increase production of detoxifying enzymes. Amplified genes can be found scattered throughout the genome, and the mechanism underlying this mode of resistance is unknown. Further, it is unclear whether certain organisms are better able to amplify genetic material than others, or whether the trait is specific to certain genomic regions or DNA architecture.
• ▪ METABOLIC DEGRADATION: Pests and pathogens can also evolve mechanisms to metabolize toxic compounds, as is often seen in plants.
• ▪ EXCLUSION OR SEQUESTRATION: All organisms can pump toxins out of their cells through efflux channels, a common method among cancer cells and bacteria. Drug sequestration is another effective resistance mechanism; certain cancers develop resistance to the anti-angiogenic tyrosine kinase inhibitor sunitinib by accumulating the drug inside lysosomes. Both limited uptake and enhanced sequestration are mechanisms of some pesticide resistances.
• ▪ STRESS RESPONSES: Resistance can also emerge due to the effects of the drug itself if the method of action affects mutation rates. Similarly, resistance can arise through conditioning if the cellular response pathway activated by the drug triggers an up-regulation of compensatory changes.
• ▪ ECOLOGICAL AVOIDANCE: Finally, target organisms can physically avoid the drug or pesticide itself. For example, viruses and bacteria can enter latent phases of their lifecycles in which they do not replicate, and remain hidden to both the immune system and therapeutic approaches. Bacteria often grow in dense communities called biofilms, sometimes comprised of multiple species, and the interior environments of these biofilms experience reduced concentrations of antibiotics. Similarly, cancerous cells in the interior of a tumor face different challenges than their counterparts at the boundaries. Understanding three-dimensional tumor “ecology” will be an important development in oncology and the exploration of drug resistance in cancer.

Although the ways in which pests and pathogens develop resistance to treatment are extremely varied, there are significant commonalities across biological systems from plants to cancer to insects to viruses. Because different biological systems experience similar molecular, or proximate, causes of resistance, the means of counteracting those resistance mechanisms might also have common features, which will be explored later in this report.

Generation of genetic diversity through different reproductive strategies

Genetic diversity is generated in a population through mutation and reproduction. Reproduction can be either sexual or asexual, and some organisms such as fungi, malaria parasites, and some plant species are capable of both during their lifecycles. Asexual reproduction, such as that of viruses, bacteria, and cancer cells, generates diversity through mutations acquired during DNA (or RNA) replication. These mutations are then passed to progeny who, other than the mutations generated during replication, are genetically identical to their parents. Although asexual reproduction encompasses most of the viral and bacterial life cycle, these organisms do occasionally acquire genetic diversity through horizontal gene transfer, in which genetic material is passed between members of the same generation or between different species.

Other organisms such as fungi, parasites, plants, insects, and of course, humans reproduce sexually. Sexual reproduction allows for recombination of genes with every generation. Additionally, receiving genetic contributions from two parents means that each offspring inherits two copies of each gene that can be either dominant or recessive, further shaping the genetic landscape. Among agricultural pests, some weeds are capable of asexual reproduction, such as rooting following fragmentation of the parental plant. Others, such as dandelions, can produce seeds without flower fertilization; dissemination of these seeds via aerial routes can allow for rapid spread of clonal populations. Although some insects, such as aphids, reproduce asexually, many insect pests must reproduce sexually and therefore generate diversity through genetic recombination. Similarly, some plant species are obligate “out-crossers” either because they are comprised of unisexual individuals or their gametes are self-incompatible, and like interbreeding insects, require two individuals to colonize a new area. These different reproductive strategies affect how quickly resistance can emerge and how far resistant individuals can spread.

THE EXTENT OF GENETIC DIVERSITY IN A GROUP DEPENDS ON POPULATION SIZE, GENERATION TIME, AND MUTATION RATE.

The extent of genetic diversity in a group depends on population size, generation time, and mutation rate. First among these traits is the population size of the target, as population size is positively correlated with diversity and therefore the probability of emerging resistance. Cancer is believed to begin with a mutation in a single somatic cell that proliferates uncontrollably, leading to tumors, and sometimes acquires the ability to grow outside of its normal body site leading to metastases. In the case of bacterial, viral, protozoal, and fungal infections, the number of pathogens required to cause disease in an individual varies significantly depending on both the pathogen and the immune status of the infected individual. Sometimes that number can be very small. A 1938 study of malaria in rhesus macaques established that inoculation with a single sporozoite could result in infection and death. Artificial inoculation with ten arthospores of the fungal pathogen Coccidiodes immitus was sufficient to cause infection in monkeys.

Additional factors shaping resistance spread are generation time and mutation rate, which also vary widely across organisms. Some bacterial species, such as Staphylococcus aureus, undergo cell division every 30 to 60 minutes, whereas mosquitos enjoy a two week lifespan. Organisms with shorter generation times can potentially develop resistance more quickly than those whose generation times are orders of magnitude longer.

Box

ALTERNATIVE MECHANISMS OF GENERATING GENETIC DIVERSITY.

Mutation rates also vary among target organisms by orders of magnitude. Viruses like HIV and influenza exhibit extremely high mutation rates on the order of 10^-5 (one event for every 100,000 base pairs replicated), which effectively translates to about one mutation per replication event because the genomes of these viruses are only on the order of 10⁴ base pairs, much smaller than the human genome of 10⁹ base pairs. Many viruses can also generate sequence diversity through recombination or reassortment during replication. Mutation rates among bacteria are orders of magnitude lower, and in Plasmodium falciparum studied in vitro that value drops lower still, ranging from 10^-11 to 10^-20.

Significant debate exists in the cancer literature as to whether increased genetic instability is required to generate the multiple mutations that characterize cancer, or whether normal rates of mutation coupled with uncontrolled clonal expansion are sufficient for this process. Current evidence supports a hybrid model, in which normal mutation rates can occasionally result in a sequence discrepancy or chromosomal abnormality that itself predisposes that cell and its progeny to increased mutation frequencies.

Similar evidence exists in the microbial world that bacteria can increase their mutation rates under stress as a survival mechanism to generate additional diversity. If true, this suggests that evolutionarily, not only do antibiotics exert the selection pressure necessary to drive resistance phenotype proliferation, but may also trigger the target cells to increase their chances of developing such resistant traits. Indeed, mutation rates have been reported to increase by several orders of magnitude under starvation conditions; antibiotic stresses may also activate expression of stress-response error-prone DNA polymerases which can boost native mutation rates.

Ability of resistance to spread

While reproductive strategies, population size, generation times, and mutation rates play significant roles in generating genetic diversity and thus resistance emergence, transmission of the infection or spread of pests will shape the diversity profile over time. Cancers exhibit the lowest degree of resistance spread, as each case is confined to one individual and depends on the host's genetic code and mutational events arising with each somatic cell division. Thus, cancers, and therefore resistance (with the exception of a few contagious diseases in the animal world) cannot spread from individual to individual. In long-lasting infections like HIV and tuberculosis, resistant strains can emerge in response to antiviral or antibacterial treatment of a single individual, who can then transmit the resistant pathogen to others. The ability of bacteria to exchange genes through horizontal gene transfer, even between different species, means that resistance need not arise de novo, nor be spread only from patient to patient, but can also be acquired from the environment. Viruses that infect bacteria, called bacteriophages, can also act as carriers of genetic information throughout microbial communities. Genetic exchange can occur within a patient, such as in an established biofilm in the lung of a cystic fibrosis patient, or attached to an implanted device, or externally in soil, water, barns or hospitals. Sexual reproduction among malarial parasites, fungi, insects, and weeds can also generate substantial diversity and allow resistance to spread quickly. Malarial parasites exist in both haploid and diploid forms, so recombination occurs within the female mosquito host. When the mosquito takes her next blood meal, diverse sporozoites are released into the human bloodstream. Resistance genes and diversity spread quickly through human populations because the malaria parasites carried by mosquitos are polyclonal and additionally, humans can be biten by thousands of mosquitos across their lifespan. Resistance traits are likely to have greater mobility among plants and insects whose life cycles often require sexual reproduction and who enjoy dispersal mechanisms of both gametes and adults.

Failures and successes

Unfortunately, when drug resistance does occur, the results can be disastrous. Failures can result for a variety of reasons, from treatment strategy to drug design, but all suffer from the common feature of immense selection pressure driving resistance evolution.

SUCCESS DUE TO TREATMENT PLAN:

Success can also be due to treatment plan, and perhaps no approach better highlights the value of using evolutionary principles to design and implement a novel pest control strategy than the development of resistance management strategies for transgenic crops producing Bt toxins (Figure 2). Bt toxins are not harmful to humans and most other non-target organisms, but kill many key insect pests. Corn and cotton plants have been genetically engineered to produce Bt toxins. The insect pests ingest the toxins when feeding on such Bt crops, which reduces the need for insecticide sprays.

FIGURE 2.

In order to protect agricultural interests from insect pests, key crops such as corn and cotton have been genetically engineered to express Bt toxins which target the digestive tract of those pests. (C) Without the Bt toxin, over time the insects would (more...)

Proactively recognizing the threat of pest resistance to Bt crops, scientists used evolutionary principles to design resistance management strategies that have been implemented successfully in many cases. From the beginning, this process entailed collaboration among farmers, regulators, and scientists in academia and industry. The key element of most strategies for delaying resistance to Bt crops is a “refuge” of host plants that do not produce Bt toxins and thereby enable survival of susceptible pests. When Bt corn and cotton were first commercialized in the U.S. in 1996, EPA regulations required farmers to plant refuges of non-Bt corn and non-Bt cotton. The idea underlying the refuge strategy is that rare resistant adults emerging from Bt crops will be most likely to mate with the more abundant susceptible adults emerging from nearby refuges. If inheritance of resistance is recessive, the progeny of such matings between resistant and susceptible adults will die on Bt plants and resistance will be delayed substantially. Inheritance of resistance is recessive if the Bt plants produce a sufficiently high concentration of Bt toxins to kill heterozygous progeny produced by matings between homozygous resistant and homzygous susceptible adults. The strategy is sometimes called the “refuge/high dose strategy.” Therefore, one of the goals has been to manipulate gene expression so that the engineered plants produce high concentrations of Bt toxins.

As part of a regional program to eradicate pink bollworm, the U.S. state of Arizona has used a related strategy since 2006 to replace the host plant refuge with the release of sterile insects. The idea here is that the rare resistant insects surviving on Bt cotton will mate primarily with sterile insects, resulting in no fertile progeny. Under this plan, the EPA allows farmers to plant up to 100% Bt cotton and refuges of non-Bt cotton have become rare statewide. Billions of sterile pink bollworm moths have been released from airplanes over cotton fields each year, resistance has not evolved, and this insect that had plagued farmers for a century is now difficult to find in Arizona.

First-generation Bt plants that each produce a single Bt toxin are being largely replaced by second-generation Bt plants that produce two or more toxins. Compared with resistance to a single toxin, resistance is expected to evolve slower when two or more distinct toxins target the same pest. Another recent development is “refuge-ina-bag,” where non-Bt refuge seeds are mixed in the seed bag with Bt seeds to ensure that farmers sow the desired percentage of refuge seeds in each field. Despite some striking successes, several cases of resistance to Bt crops have emerged in the U.S. and abroad, sometimes reflecting failure to implement the refuge strategy.

Defining the goal of treatment: eradicate or contain?

If the only acceptable outcome of treatment is eradication of the pathogen or pest, drug design and treatment strategy may be approached differently than if maintaining the pathogen or pest at a manageable or economically acceptable level is adequate. Choosing the goal of eradication may have consequences in terms of reducing the useful lifetime of the drug or treatment because of the higher selection pressure. In cancer or infectious disease, the priority for patients and health care practitioners is patient survival, and this is almost always equated with eradication of the infectious agent or the cancer. HIV is a notable exception — complete cures are still sought but long term management is appropriately seen as a triumph. The agricultural industry traditionally adopts a management-driven strategy; a farmer may be willing to accept slightly lower yields to extend the lifetime of a insecticide or herbicide. Despite the differences in measurements of success across systems, there is much that practitioners in disparate biological systems can learn from one another; indeed, management approaches may result in better outcomes both in terms of patient lifespan and prevention of drug resistance.

Lessons learned from successful strategies

From an evolutionary standpoint, it is realistic to assume that resistance will always occur and so in any biological system, we must consider evolutionary principles for development of resistance prevention or management strategies. No matter how careful the drug design or target selection, enormous selection pressure will likely drive resistance eventually. From several success stories, the merits of drug combinations have been shown in diverse fields. Proper dosing is important because varied strategies can favor different types of resistance, such as dominant or recessive, and contribute to the speed at which resistance develops. Success stories also illustrate the importance of carefully designed and implemented treatment plans which account for evolutionary principles in the prevention of resistance evolution. Lessons from successful strategies, however, do not necessarily transcend boundaries between biological systems as each is defined by unique attributes that affect the frequency of resistant mutations and the potential spread of those mutations. Therefore, establishing standard guidelines for resistance prevention and management across all fields is not yet possible, but the principles discussed here may enable the formulation of system-specific recommendations and spark new approaches based on cross-system similarities.

A key question: is resistance already there?

The prevention of de novo resistance emergence and management of pre-existing resistance were the two key questions posed to the colloquium participants. The group agreed that once a resistance mutation is present in a target population, continued widespread application of the drug or pesticide will encourage proliferation of resistant individuals. Potential fitness costs engendered by the resistance mechanism could be negated if most environments encountered by the pathogens or pests are treated, thus shifting the evolutionary balance in favor of resistant individuals. Given the (1) frequency of resistance alleles in a population, the (2) selection coefficient describing the relative fitness of the mutant to the wild-type in treated and untreated evirtonments, the (3) relative abundance of treated and untreated environments, and the (4) rates of movement between those environments, classical genetics modeling can predict the rate of spread of the resistance mutation. If the resistant population grows beyond that which makes its extinction improbable, that population's continued success will be determined chiefly by selection. In some cases, the resistant population can completely overwhelm any remaining susceptible individuals, which is precisely the scenario that we wish to avoid. Therefore, if resistance pre-exists in a population, any treatment plan must preserve a drug-sensitive population to prevent complete takeover by the resistant strain.

RESISTANCE MECHANISMS IN ANTIBIOTICS ARE ALREADY WIDESPREAD IN THE ENVIRONMENT AND HAVE BEEN FOR MILLIONS OF YEARS. THE SELECTION PRESSURE OF ANTIBIOTIC USE IN MEDICINE AND AGRICULTURE DRIVES THE SPREAD OF RESISTANCE THROUGH HOSPITALS AND COMMUNITIES.

Maintenance of a drug sensitive cohort can be achieved through a variety of means. Back-mutation, in which the same mutation rates that gave rise to the initial resistance phenotype reverse the process, can lead to revertants more fit than their resistant brethren under non-selective conditions. An influx of sensitive individuals from alternative sources or inclusion of drug-free sanctuary sites or refuges as previously discussed can bolster sensitive populations. Additionally, the resistance mutations may cause a decreased fitness of the resistant individuals once the frequency of the mutation becomes too great in the total population.

In the case of antibiotics, resistance mechanisms are already widespread in the environment and have been so for millions of years. The selection pressure of antibiotic use in medicine and agriculture drives the spread of resistance through hospitals and communities. Resistance traits to drugs in other biological systems may not be so well-established or diverse, but nevertheless selection pressure will drive resistance evolution and spread in those systems as well.

Because resistant traits can arise through many different mechanisms, resistance prevention and management strategies need to incorporate evolutionary principles and accommodate the unique ecological attributes of each system. To effectively prevent resistance emergence or manage its presence, we can control features such as treatment plans, target selection, drug design and to some extent, transfer of resistance information.

Monotherapy versus combination therapy

Historically, across all biological systems, treatment with a single agent has been the norm. When a virus, cancer, or crop pest proved resistant to that agent, a secondary treatment was applied if available. This approach can be successful if resistance does not develop too quickly and if multiple treatments are available. However, HIV and cancer are particularly problematic because sequential application of drugs merely leads to sequential development of resistance to all drugs applied. Still, such a strategy can lengthen the patient's lifespan overall if the development of resistance is slow.

In the human health domain, the emergence of combination therapy as a means of HIV treatment has begun to alert other fields to the potential benefits of the practice. Already, tuberculosis infections are routinely treated with a spectrum of antibiotics and the World Health Organization recommends the highly effective 3-7 day Artimisinin Combination Therapy (ACT) for treatment of malarial infections. From an evolutionary perspective, de novo acquisition of resistance could be less likely through combination therapy because pathogens would need multiple mutations to overcome applied therapies. The probability of two independent mutations arising in the same organism is theoretically the product of the probability of each mutation arising individually. The multiplicative rather than additive effects greatly reduces the likelihood of dual resistance and increases the likelihood that the target population will be controlled by the drugs before resistance evolves. The second drug ensures that resistance to the first drug is essentially useless because the primary resistance trait cannot help in overcoming the effects of the secondary drug. Ultimately, resistance to the first drug should be lost from the population over time, unless its rate of loss is outpaced by the rate of acquisition of resistance to the second drug. Then the evolutionary advantage disappears and treatment with two drugs may only hasten resistance evolution to the second. Horizontal gene flow complicates combination therapy for bacterial infections, as several drug resistance mechanisms may be transferred at once into the target population from other species. Thus, combination therapy works best when frequencies of resistance are low or absent.

TO BEST PREVENT RESISTANCE FROM DEVELOPING, THE MECHANISMS OF ACTION AND THE PROCESSES TARGETED SHOULD BE DIFFERENT FROM ONE ANOTHER.

Furthermore, simply combining any two drugs does not always help unless the drugs act by sufficiently different mechanisms. If both drugs act on the same biochemical or regulatory pathway in the target organism, a single mutation could potentially arise bypassing that pathway and confer simultaneous resistance. In insect control, success is more often proportional to the difference in the ways that the insects are killed.

In practice, the preference for monotherapy over combination therapy across biological systems involves a trade-off between evolutionary and practical considerations. For herbicides, ideally the agricultural industry would want to use drug combinations to better manage infestations, but the expense and potential environmental impacts of such a practice may prohibit consistent use. Herbicides, therefore, are often rotated through fields in some cropping systems to prevent a single dominant phenotype from emerging. Such a practice temporaily reduces exposure to single compounds and can delay resistance evolution under some circumstances. Among the antifungal community, true monotherapy is typically sufficient to quell infections where clonal inoculation of the patient is the norm. Therefore, if the genotype of the fungus is known prior to treatment, application of a single drug can be the best approach. Any observed resistance is more likely to have been acquired from another individual or the environment than evolved within the current host.

What makes an ideal treatment combination?

Although this is a still an open research question, treatment combinations are often proposed to exhibit additive effects, wherein the dual effects of the treatments are greater than either one acting alone. To best prevent resistance from developing, the mechanisms of action and the processes targeted should be different from one another. Ideally, treatment combinations would target different pathways in the pathogen or pest, thus requiring multiple resistance mechanisms to bypass the treatments. Cross-resistance can be predictable in cancer, particularly if both drugs target the same pathway. Colloquium participants displayed optimism for the design of successful anti-fungal combinations in treating human infections. Although there are very few anti-fungal drugs available, there is depth of knowledge in what the drugs do and how they work.

Combinations that place the pathogen in an “evolutionary trap” are ideal, wherein resistance to one drug renders the pathogen susceptible to the action of the second, as was previously illustrated by the combination of the p53 vaccine and standard chemotherapy treatment of breast cancer. Combination HIV therapy using AZT and 3TC has also been effective in preventing dual resistance, as the mutation conferring 3TC resistance renders the virus more susceptible to AZT. Dual resistance can eventually develop, but multiple mutations become necessary to resist AZT when delivered in combination, and so overall resistance is delayed. According to a recent study, combination therapy “controls [HIV] viral replication because of steep, upwardly inflected dose-response curves for some drugs and synergies reflecting independent action for other[s]” (Jilek, et al., 2012).

For a successful combination therapy, both techniques need not be cytotoxic on their own if the addition of one compound increases the effectiveness of the other. For example, the anti-tuberculosis effects of weak acids, such as the antibiotic front-line drug pyrazinamide (PZA), can be enhanced by inhibitors of energy metabolism under aerobic conditions. However this enhancement is not observed under anaerobic conditions, highlighting that consideration of environmental variability is key during drug development and treatment design. Similarly, phage therapy could enhance the activity of antibiotics against bacterial infections. Indeed, both drugs might not need to be inhibitory to be maximally effective. One drug might be able to complement the host's immune system such that the target pathogen population, driven low, by another compound, can be managed by the host. In applying the “evolutionary trap” concept to the host immune systems, it has been hypothesized by HIV modeling that any drug to which a resistance phenotype elicits a potent immune response would be very difficult for the virus to combat.

Drug combinations can also be designed to boost the beneficial microbial background of the host. If developing resistance imposes a fitness cost on the pathogen, probiotics could be used to out-compete the resistant population. Such techniques could also be applied to plants, increasing their fitness against pathogens or insect pests. In humans, antibiotic combinations with the anti-diarrheal loperamide also have increased effectiveness.

Additionally, the two drugs need not necessarily be delivered simultaneously to be considered working in combination. Pairs can target different stages of the life cycle, such as drugs that target larval and adult insects. Combinations can also target different aspects of the pathogen's life cycle. In the case of malaria, drugs that act within the human could be combined with strategies to disrupt parasite-mosquito interactions.

An additional important feature of successful combinations is similar half-lives. The evolutionary goal is to force the pathogens to develop solutions to both treatments simultaneously. If one active ingredient decays much faster than the other, the pathogen or pest encounters a sequential monotherapy approach rather than a combination.

How should common evolutionary principles inform the development of optimal dosing strategies?

Once a treatment strategy has been determined, the dosing regimen assumes critical importance. The longstanding paradigm in many systems is to treat at the highest possible level. In cancer, oncologists begin with the maximally tolerated dose, with the goal being to kill as many cancerous cells as possible as quickly as possible. Such a strategy is also favored in agriculture, as farmers are advised to use high doses to discourage proliferation of low level resistance genes. Unfortunately, prohibitive herbicide costs often lead farmers to dilute these products and the lower doses encourage resistance evolution. Additionally, older and larger plants are generally less susceptible to treatment than younger seedlings, again highlighting the necessity for early and intense dosing. Similarly, because fungal infections are typically clonal in nature, if resistance is not detected upon diagnosis, physicians will proscribe high monotherapy doses. Evolutionarily, a high dose strategy is favored when high doses kill the target population with enough speed and efficiency to out-run mutation rates. Suboptimal doses have been shown to facilitate acquisition of resistance in diverse pathogens, including HIV, bacteria, the fungal pathogen Candida albicans, and malaria. Under reduced selection pressure, slightly resistant individuals are able to survive and accumulate even greater resistance capabilities. Low dose regimens also permit the survival of susceptible parental strains who have additional generations to develop resistance. This may account for the very common development of resistance in immunocompromised patients who are routinely given low-dose prophylactic antifungals. The level of dosing can also contribute to the speed at which resistance develops. For example, very low insecticide rates may hasten the evolution of resistance, as sub-lethal doses may induce stress pathways leading to enhanced mutation rates.

Short, high dose pulses of insectides are favored by modeling algorithms. The short duration of the treatments kills all but a few susceptible individuals, whose population is then allowed to rebound slightly before the next treatment. Between treatments, the preserved susceptible population should then be able to out-compete any resistant individuals if there is a resistance-related fitness cost.

While low dose regimens may allow the survival of weaker resistant phenotypes, maximum doses can also lead to resistance evolution; if the high doses fail to eliminate the target, the enormous selection pressure may allow resistance to emerge. Recent studies in malaria and cancer suggest that a “middle ground” approach may be successful at both managing the diseases and prolonging life. Intermittent, lower doses of malarial medication in mice were shown to be effective in keeping resistance down. Similarly, breast cancer cell lines were kept alive indefinitely through lower chemotherapy doses and no resistance was detected. Proponents of this idea cite evidence that decreased chemotherapy doses simultaneously decrease hypoxia and increase blood flow, thus allowing for better drug delivery than if higher doses were administered. The high toxicity of cancer drugs makes “metronomic therapy,” or periodic rather than continual dosing, additionally attractive because lower doses are tolerated at more frequent intervals. However, such malaria treatments in situ are impractical — while mouse-malaria models are encouraging from a resistance perspective, antimalarial drugs are not considered practical for distribution in developing nations if they must be taken for a course greater than three days, as compliance and cost issues negate the treatment benefits, as well as risk resistance emergence through incomplete treatment regimens.

Acceptance of the “lower, extended dose” or “maintenance” model will also require a dramatic paradigm shift among oncologists and the public at large, who strongly view cancer as a disease to be cured. While a cure may be a feasible outcome for certain cancers, many could potentially be managed to an asymptomatic background level through continuous low dose, combination therapy treatment. Such a scheme would not necessarily cure the cancer, but could relieve the overall disease burden, prevent resistance evolution, and most importantly, prolong life.

A “maintenance” model relies on tolerance of some level of target population which begs the question for each biological system: how much control is necessary to reduce the negative effects to an acceptable level? In cancer, this is a radical idea but laboratory experiments are promising. Even after a “cure,” one can often detect residual cells by molecular tests such as PCR. There does seem to be a correlation between the remaining cancerous cells detected and disease recurrence frequency, but quantification of cancerous cells could be used to establish treatment guidelines, much like viral load in HIV. However, in treating fungal infections, elimination is necessary once the fungus reaches the bloodstream. The same issue arises with bacterial sepsis although other types of infections offer more “wiggle room;” urinary tract infections, for example, characterized by fewer than 10³ cells/mL do not warrant treatment. Treatment of malarial infections also aims for complete clearance, except in cases of asymptomatic adults living in high risk areas, which occurs after more than 10 infections in the patient's lifetime. In plant pathogens, however, farmers have room to accept a moderate level of disease, and the same is often true for weed control and insect pests.

If the dosing of treatment is important either for elimination of the pest or management of its population, the timing of treatment is also critical; waiting until there's a heavy parasite, cancer cell, or viral load to begin treatment increases the potential for a higher degree of variation, and thus a higher likelihood of resistance evolution. However, treatment cannot commence until a diagnosis is made, which relies both on the timing of symptom development and confirmation by a health care professional. In some cases, the illness may not manifest until the target population has already reached high levels or the patient may delay in consulting with a doctor. To catch infections or infestations in their infancy, vigilance and surveillance must be maintained.

Consideration of the lifecycle of the target in treatment timing will increase the effectiveness of those treatments. Therapies should ideally be timed to apply treatment when its effect will be maximal. For example, dandelion weeds are at their weakest right after blooming, when food reserves in the roots are at their lowest. Tender young leaves are also most susceptible to herbicides in the spring. The early fall month of September can additionally be an appropriate time for herbicide treatment, because as the dandelion prepares for the upcoming winter months, it moves carbohydrates from the leaves to the roots, providing a chemical traffic flow that any applied herbicides can hijack to reach all of the plant body.

ACCEPTANCE OF THE “LOWER, EXTENDED DOSE” OR “MAINTENANCE” MODEL WILL ALSO REQUIRE A DRAMATIC PARADIGM SHIFT AMONG ONCOLOGISTS AND THE PUBLIC AT LARGE, WHO STRONGLY VIEW CANCER AS A DISEASE TO BE CURED.

Contingency planning

PATIENT COMPLIANCE AND DRUG RESISTANCE GO HAND IN HAND; INDIVIDUALS NEED TO BE AWARE OF HOW THEIR ACTIONS AFFECT THEIR OWN RISK OF DEVELOPING A RESISTANT POPULATION, BUT ALSO HOW THAT RESISTANCE CAN SPREAD TO THE COMMUNITY AS A WHOLE.

Because resistance can arise through treatment mis-use and despite the most carefully-designed treatment strategies, contingency plans need to be established. When the fall armyworm Spodoptera frugiperda developed resistance in only four years to Bt toxin-producing corn in Pueto Rico, the Bt crop was voluntarily withdrawn from the marketplace. Such withdrawal was in line with EPA recommendations stipulating that actions must be taken to thwart resistance to Bt crops. Education is effective and important for resistance prevention. For example, Bt cotton producing a single toxin remained effective for control of pink bollworm in Arizona but evolution of resistance to single-toxin Bt cotton has been observed in India. Arizona farmers complied with the refuge strategy mandated by the EPA to manage resistance in this pest; in contrast, refuges of non-Bt cotton were required in India but compliance was low. The lack of refuges likely promoted faster evolution of resistance in India compared with the U.S. Patient compliance and drug resistance also go hand in hand; individuals need to be aware of how their actions affect their own risk of developing a resistant population, but also how that resistance can spread to the community as a whole. Education about resistance risks and cancer progression may also lead to acceptance of cancer treatments as a life-long and life-prolonging route to halt growth of the disease, rather than looking for curative measures.

Principles of drug design for direct targets

For decades, identification of therapeutic agents relied on broad screening of chemical compounds for inhibitory or cytotoxic effects. With significant advances in biochemistry and molecular biology, drug design today takes a more targeted approach. Treatment target sites are selected because they are critical to the survival of the pathogen or pest. Finding such targets in bacteria or viruses is relatively less complicated than identifying targets in eukaryotic cells of organisms such as fungi, as the latter are more closely related to humans and thus have more similarity amongst their essential proteins. Thus, the compounds designed could elicit deleterious effects in the patient either through direct treatment applications or through environmental exposure in the case of herbicides and insecticides. Intelligent anti-cancer drug design is even more challenging because the genetic makeup of the cancerous cells is almost identical to that of the host; drug targets in these instances are often the mutated forms of normal human proteins which gave rise to the cancer in the first place.

Target-based approaches require knowledge of biological processes. Small molecules are sought that mimic the actual binding partners or protein receptors of the target, and either activate or inhibit its activity thus providing therapeutic benefit to the patient. Sometimes libraries of small molecules are screened for the desired effects but increasingly small molecules are designed in combination with computational modeling of the three-dimensional structure of the target and the target's structure bound to biologically-relevant interacting partners. The goal is to isolate a compound that competitively binds tightly to the target to the exclusion of the native binding partner or prevents a key conformational change. Such designed inhibitors are used in all of the disease targets that have been described, targeting among others: proteases, transcriptases, kinases, and synthases. Ideally, the chemical interactions between the compound and the corresponding target should be very specific, eliciting a much higher binding affinity than the native partner. To avoid drug resistance, the inhibitors should be designed to be as evolutionarily constrained as possible; this could involve mimicking an enzymatic transition state, as is often done, or interfering with a target that has multiple necessary binding partners. Targets with multiple binding partners are more evolutionarily constrained as co-evolution would have to occur within each of the partners in the complex for resistance to occur. Thus, the large number of primary and compensatory mutations needed to occur simultaneously would greatly decrease the probability of evolved resistance.

TO AVOID DRUG RESISTANCE, THE INHIBITORS SHOULD BE DESIGNED TO BE AS EVOLUTIONARILY CONSTRAINED AS POSSIBLE.

When much is known about the target enzyme or molecule, intelligent combinations of drugs can be designed against the same target, wherein development of resistance to one drug renders the second more potent through “negative cross-resistance” or an “evolutionary trap.” The two antagonists could even be within the same molecule as a bifunctional compound that attaches to the target protein in two places. In such case the protein would be forced to make multiple simultaneous changes to avoid the toxin. For proteases, active sites are commonly targeted, although targeting of allosteric sites, that is, sites other than the active site, could be considered as this could elicit broader effects on activity, such as inhibiting polymer formation. While the most commonly targeted molecules are proteins, membrane constituents vary enough between species that it could be possible to target lipids or other membrane components specific to target organisms. This strategy might be less prone to driving resistance than targeting proteins as membranes are more difficult to modify and linked to many diverse systems within the cell.

Drugs that affect multiple cellular targets can also stymy resistance evolution. The antimalarial artimisinin, derived from a Chinese herb qinghao, has been used for over two millennia in the treatment of fevers. The continued success of this drug could be attributed to its method of action, which relies in part on the generation of free radicals to disrupt multiple cellular targets and alter cellular redox cycling.

In selecting a host process or enzyme as a drug target, there was debate among the colloquium participants as to whether sequence conservation among species isolates was indicative of strong selection pressure and thus importance of the target to the host. When little biochemical or cell biological data is available, genomic and protein sequences may be the best available sources of information. While ideal treatment targets should be highly evolutionarily conserved, thus reducing the likelihood of compensatory resistance mutations occurring without accruing a considerable fitness cost, participants noted that both diversifying selection and directional selection can result from evolutionary force and that sequence diversity alone does not always accurately reflect selection pressures. In some cases, 100% sequence conservation could indicate an evolutionary constraint, or that the target itself has never been acted upon by the selection pressure of the immune system or other forces.

Indirect targeting

Counteracting alternate routes to resistance

Broad versus narrow spectrum treatments

When physicians are faced with the option of treating infections with either broad or narrow spectrum drugs, which is preferred from an evolutionary perspective of preventing resistance emergence? Broad-spectrum antibiotics have a clinical advantage in that they allow treatment when rapid and specific diagnosis is impractical. However, such drugs can also impose selection pressure against a wide variety of microorganisms, driving resistance more quickly than narrowly-focused drugs. Resistance that evolves in one species of bacteria can also be shared with others. The consensus among colloquium participants was that broad spectrum antibiotics should be used for as short a period of time as possible, followed by a more tailored approach as soon as the physician gains a better understanding of the pathogen at hand. An alternative option is to combine a multi-site drug with a narrow-spectrum one. This approach has shown success in fungicides. One compound hits multiple sites in the pathogen, disrupting multiple sites simultaneously, which is difficult to develop resistance against. Side effects associated with these drugs tend to discourage their use, which is why combination therapy with a very fungal-specific drug can be effective. Commercially, however, the development of broad spectrum drug therapies makes sense. Narrow spectrum drugs are a riskier venture because they are used on fewer patients and any resistance evolution negates the use of the compound. In the case of herbicides, narrow-spectrum compounds typically are not efficient to protect crops, as a single field will contain multiple weed species.

Using model systems to screen drug candidates for their propensity to elicit drug resistance

Before predictive modeling can be used to forecast the likely course of drug resistance evolution, an understanding of how resistance has developed previously is critical. Statistical modeling can help scientists learn from the field what happened in the past, and what factors were important in determining the outcomes. For example, a long-term, large-scale field study demonstrated that refuges can delay the evolution of resistance to insecticides. The study also determined the spatial scale at which pesticide-treated fields and refuges affected the evolution of resistance, which provided critical information for development of the refuge strategy to delay resistance.

Retrospective studies drawing from clinical data following patients over time can also be extremely informative. Longitudinal biopsies from cancer patients can provide not only progressive monitoring of features known to be important for specific drug resistance events, but a rich reservoir of data to be used in future studies when new molecular mechanisms of resistance have been identified. Similarly, measurement of pathogen profiles throughout the course of HIV or tuberculosis infections could reveal important trends. In studying the course of such infections, the patients themselves become the models. Cystic fibrosis patients could provide a wealth of data over long periods of time if sputum samples could be routinely characterized and tested for antibiotic resistance.

The aforementioned studies will provide clues as to spatial and temporal patterns of resistance evolution, but a mechanistic understanding of how and why resistance evolves also plays a significant role in shaping resistance prevention strategies. Powerful tools such as transcriptomics partnered with molecular and computational biology will enable a better understanding of the pathways to diseased states. In cancer biology, over a decade of research using gene expression arrays has established certain genes and systems as being up or down-regulated in cancerous cells, but identifying the causes behind these expression changes remains elusive. Often the mutations underlying the transcriptional changes are point mutations or gene amplifications. Colloquium participants proposed performing large-scale mutagenesis studies, using a variety of mutagens to capture a full spectrum of possible mutations and combinations of mutations. The resulting mutagenic data could then be combined with the well-established transcriptomics catalog to gain a deeper understanding of the regulatory processes involved in cancer development and drug resistance. The colloquium participants also stressed that a deeper and more thorough understanding of the underlying genetic diversity within individual biological systems is critical to the success of treatment plans. Mutation rates vary among species, and DNA sequence diversity also varies across the chromosomes of any one species. Deep sequencing reveals significant genetic diversity among environmental isolates; indeed, there are 60 var genes in malaria contributing to immune evasion, and across this data set at least 500 sequences have been identified and none are identical.

THE IDEAL TREATMENT SITUATION WOULD INVOLVE A DRUG, OR COMBINATION OF DRUGS, THAT IS HIGHLY LETHAL AND ALSO VERY RAPIDLY CLEARED FROM THE BODY OR ENVIRONMENT.

Challenges of model systems

To simplify environmental variables, model organism systems can be established in the laboratory to predict rates and mechanisms of resistance emergence. In such laboratory studies, the actual target organism is the ideal model, but this is not always possible. Malaria parasites are notoriously difficult to grow and genetic manipulations are tricky; determining which mutations confer the resistance pheno-types and thus mechanism of action of the drug is challenging. Alternatively, model laboratory organisms can be used as proxies for the actual pathogens. For example, the yeast Saccharomyces cerevisiae can be used as a model system for Candida and other fungal pathogens, but resistance mechanisms and underlying genetic diversity may not be conserved. Drosophila fruit flies are also more amenable hosts to malaria than mosquitos for high throughout laboratory studies but the interspecies interactions observed may not precisely mirror the natural ecology of the parasites.

Once a model species has been chosen, laboratory studies can first test the potency of the drug. If the compound is sufficiently potent to clear a diverse population of pathogens quickly, the window for the development of resistance is drastically narrowed. The ideal treatment situation would involve a drug, or combination of drugs, that is highly lethal and also very rapidly cleared from the body or environment. Combination therapies can also be tested using model systems approaches. Biochemists can begin with an initial drug whose effects are known, and then test a library of other chemicals to find synergistic properties. Proving synergy is not simple, but if established, the doses of one or both drugs could potentially be lowered, thus reducing any side effects. A drawback of this approach is that defining synergy as the sole endpoint might miss drug combinations which have no statistically-significant impacts in terms of killing or controlling the pathogen, but might offer a drastically decreased likelihood of resistance evolution. Distinguishing between a rare event, such as resistance to monotherapy, and an extremely rare event, such as resistance to combination therapy is also challenging. However, from an evolutionary standpoint, the path and time to drug resistance should be critical considerations in the selection of treatment plans. Screening for the time to resistance requires a large, diverse starting population. With some rapidly replicating species, the necessary diversity and population size can be generated quickly. In studying resistance evolution in cancer, artificially generating diversity through carcinogens is preferred to using genetically modified mice because a vast scope of possible mutants can be more readily created.

Mutagenesis approaches are recommended for quick screening and characterization of resistant phenotypes. For organisms with smaller genomes, mutagenesis approaches can almost be guaranteed to elicit mutations spanning the entire chromosome. However, the mutations obtained through artificial mutagenesis do not always reflect those that would occur in nature, such as gene amplifications. Artificially increasing mutation rates may also generate unrealistic results. When the target site is known, site-directed mutagenesis can be a very powerful tool. Laboratory studies could drive resistance evolution in the model organism through drug dosing, rather than artificially inducing mutations through artificial processes. Such an approach might take longer to achieve significant results, but those obtained will be more environmentally relevant. Once a resistant phenotype is discovered, it becomes possible to test for fitness differences between the resistant strain and the wild-type. Among bacteria, such fitness cost evaluations should be established for communities as well. Broad spectrum antibiotics are the norm during treatment of infections, but it is important to limit the overall impact of the drugs on the body's natural microbiota. Additionally, resistance can evolve in any of the commensal community members and spread via horizontal gene transfer, so it is necessary to know how drug treatments will affect other organisms and particularly whether resistance already exists in the environment. For example, the antibiotic vancomycin was thought to be “unresistable” until the entire resistance pathway was transferred from an environmental species; thus the path to resistance was always available in the environment, just not in the pathogens that were initially targeted.

TWO TYPES OF COMPUTATIONAL STUDIES CAN BE CONDUCTED: THE RESULTS OF “TRUTHED” STUDIES, BASED ON REAL WORLD PARAMETERS, CAN BE JUSTIFIED AGAINST ACTUAL CONDITIONS, WHILE THEORETICAL STUDIES ARE USEFUL TO SEE IF NOVEL APPROACHES WOULD BE EFFECTIVE AND CAN BE DESIGNED WITH CONCEPTUAL CLARITY.

Finally, computational modeling, though not always applicable to complex environmental situations, can be used to vary multiple theoretical constraints and thereby predict the emergence of resistance under changing conditions. Such studies have a distinct time advantage over traditional laboratory work because days, months, and even years can be condensed into simple algorithms. Two types of computational studies can be conducted: the results of “truthed” studies, based on real world parameters, can be justified against actual conditions, while theoretical studies are useful to see if novel approaches would be effective and can be designed with conceptual clarity. Variables such as generation time, mutation rate, population flow, drug half-life, and dosing schedule can all be tested in computer simulations to model resistance emergence and management strategies.

Colloquium participants cautioned against placing too much trust in modeling systems, however, as results can lead to both false negatives and positives. In one cited instance, antibiotic resistance was rapidly achieved in culture, although the drug works quite well in patients with urinary tract infections without apparent emergence of resistance. Alternatively, the model plant organism Arabidopsis thaliana was mutagenized to search for potential resistant varieties to glyphosate and none were identified, despite isolating seedlings resistant to two other tested herbicides at a frequency of about 10^-5. These promising results suggested that resistance to glyphosate would be extremely rare and yet many weed species have evolved resistance in the wild. Although glyphosate had been used for decades without weeds evolving resistance, the development of glyphosate-resistance crops resulted in much more intensive use of this herbicide and simplification of measures used to control weeds, thus contributing to a rapid increase in the number of glyphosate-resistance weeds. Finally, model systems cannot necessarily take into account the complexity of synergistic or negative relationships and their role in resistance. Laboratory systems are designed to minimize variables, and thus often rely on monocultures, clonal strains or inbred model organisms. For resistance prediction, these systems can greatly underestimate the strain diversity existing in nature.

Measuring and monitoring resistance in target populations

Vital to the prevention of drug resistance evolution is frequent and accurate surveillance of the target populations. If the mechanism of resistance is known, testing for that trait or underlying genetic sequence should be done prior to and throughout treatment. Particular importance is placed on screening for drug resistance in bacterial, malarial, and HIV infections before treatment, as front-end diagnostics are critical to drug and treatment selection. For more than 10 years, a few drops of blood and a simple PCR test have been sufficient to type a malarial infection for resistance; if resistance is determined, more expensive second-tier therapies are used. Prolonged illnesses such as cancer and HIV syndrome should also be monitored periodically during the treatment regime. However, even when the resistance mechanism is known, colloquium participants cautioned against relying on genotype and sequence information alone to track resistance in a population. Such an approach finds only the resistance mechanisms that are already known; screening for resistance phenotypes, while ultimately “trial by ‘treatment failure,’” has the advantage of uncovering new mutations.

If the resistance mechanism is unknown, phenotype screening is the only means of detecting emerging resistance. To augment the arsenal of resistance-detection tools, drug companies should be encouraged to invest in research that will identify resistance mechanisms prior to implementation if possible, so clinicians and farmers are prepared for any real-world emergence. However, even if resistance is not observed in the lab, it may still arise in situ so surveillance is critical.

THE PHYSICAL ENVIRONMENT OCCUPIED BY THE PATHOGENS AND PESTS IS SHARED BY EVERYONE, AND THE BURDEN OF MANAGING THEM SHOULD ALSO BE SHARED.

Only through rigorous diagnostics and patient surveillance can we effectively identify the emergence of drug resistance and trace its flow across time and geography. Global networks of hospitals and other primary care facilities will need to be established, as will uniform standards of resistance identification and reporting. Similarly, agricultural stakeholders should develop cohesive plans for sharing incidence reports of resistance among pests as mobility among seeds and pollen, plant pathogens, and insects can easily cause a local problem to devolve into a regional one. Oversight of drug treatment plans and herbicide/insecticide distribution should be handled by national or perhaps even international governing bodies to ensure that adopted practices minimize risk of resistance evolution and spread.

Individual benefits versus the common good: the tragedy of the commons

One final consideration is that treatment schemes can have effects beyond the individual patient, farm, or neighborhood being treated. In 1968 Garrett Hardin published an article outlining the “Tragedy of the Commons,” that is, the harm and destruction of shared community resources through seemingly rational and ultimately self-interested individual decisions. The commonly cited examples of over-grazing shared pasturelands to the point of field exhaustion or increased vehicular travel leading to gridlock highlight the concept well. Drug resistance management can also pose a ‘tragedy of the commons’ dilemma. Often, treatment strategies preferred by individuals have detrimental effects on their neighbors or even the global health community. The physical environment occupied by the pathogens and pests is shared by everyone, and the burden of managing them should also be shared.

For treatment regimes to last, they may need to be implemented such that the interests of the community are not forgotten. In agriculture, the dispersal radii of seeds and pollen, microbe propagules, and flight ranges of insects mean that resistance that evolves on one farm in response to treatment practices can potentially travel to a neighboring locale. In studying herbicide resistance, practices such as tillage, crop rotation, and periodic cessation of chemical treatments when infestation is minimal, and careful seed removal following harvest have been shown to contribute greatly to the longevity and effectiveness of the treaments. Incentives could be developed to encourage farmers to coordinate their practices with their neighbors, perhaps even executing scheduled rotation of herbicides and insecticides among them to mitigate resistance emergence after prolonged use in any one spot. In some cases, however, community benefits arise without such incentives, as seen with refuges used to delay resistance of European corn borer to Bt corn in the midwestern U.S. Most of the estimated $7 billion in economic benefits of Bt corn over 14 years were associated with planting of non-Bt corn refuges, because Bt corn reduced pest populations on nearby non-Bt corn and non-Bt corn seed was less expensive than Bt corn seed.

Mosquito control to combat malaria can also be seen as a community-based effort where the actions of each member either benefit or harm the group. Treatment of infectious diseases, however, will remain a constant battle for physicians, as long as they must make treatment decisions in the interest of an individual patient despite local and global issues of drug resistance. For example, the spread of antibiotic resistance genes through horizontal gene transfer is a real danger, particularly in hospital settings, and yet while drugs could potentially be designed to inhibit such gene flow through bacterial populations, only the community would be served, and not individual patients undergoing treatment; there is no financial incentive to develop such drugs, and no regulatory criterion that would favor them in clinical trials.

Encouragingly, manipulation of the target environment can also be applied directly to supplement treatment strategies. In certain cases, the environment can be altered to make drug resistance more costly for the pathogen. Temperature sensitivity is key for many resistant species, as increased temperature can render them more susceptible to treatment. This has been suggested for cancer, bacteria, fungi, and herbicide-resistant plants. Another proposed strategy includes increased use of natural enemies, as resistant genotypes could have reduced fitness relative to susceptible individuals when faced with predation. In the same vein, individuals resistant to Bt toxins could be more vulnerable to some plant defenses, and thus refuge plants could be manipulated to increase fitness costs that select against resistance.

Environmental manipulation could also be used to try to reduce movement or flow of resistance genes in a population. Through sexual education, distribution of condoms as preventative measures, and monitoring of blood banks, the spread of HIV and thus drug-resistant HIV has been lessened. With bacterial and other viral infections, the use of personal protective equipment and lifestyle modifications, along with quarantine if necessary, the spread of infectious, resistant agents can be halted. If a resistant population is truly isolated from other breeding partners, removal of the drug may result in a reemergence of the susceptible strain over time. To prevent the spread of herbicide or insecticide resistance genes in agriculture, farmers in neighboring areas should be encouraged to use different herbicides or insecticides than their neighbors, thus generating a mosaic of selection pressure.

Implementation and control are key for preventing resistance emergence and managing that resistance once it arises. However, despite the demonstated value of certain approaches, standard practices vary significantly across the globe. In the Netherlands, “A teams” comprised of a medical microbiologist, pharmacist, and clinician monitor every hospital infection. Prophylactic treatment is always given in advance of any open gut surgery, while other surgical procedures do not receive any additional antibiotic therapy. In contrast, antibiotic use prior to surgery is fairly routine in the United States because if the patient dies from a hospital-acquired infection, many insurance agencies are not bound to pay. This practice places pressure on hospitals to prevent infections, although rampant antibiotic use is not the answer from an evolutionary standpoint. Hospitals serve as hotbeds of resistance evolution. Antibiotics are very widely used in China, and the level of resistance there is the highest ever measured; in 2001, 89% of hospital-acquired infections involving S. aureus were resistant compared with 16% in the United States. Antibiotics were introduced as additives in feedstock almost as soon as they were discovered. This practice has broadly affected human health, as vancomycin resistance has flourished and pathogens like methicillin resistance S. aureus (MRSA) have emerged; both events seem to be products of the addition of antibiotics to livestock feed. Farm use of antibiotics is significantly less controlled than hospital use, and yet the drugs' mechanisms of actions are identical regardless of setting and yield the same paths to resistance.

A recent report in Science suggests a link between antibiotic resistance genes carried by soil microbiota and clinical isolates. In some instances, clusters of antibiotic resistance genes with >99% sequence identity between the two groups of microorganisms were identified, and were sometimes co-located with integrases or transposases for rapid horizontal gene transfer. Further, some soil isolates were taken from farmland fertilized with manure from antibiotic-treated livestock. Although the study does not prove directionality of gene flow, either movement of antibiotic resistance genes from soil to the clinic or vice versa is a troubling phenomenon, as the former establishes soil microbes as a direct source for pathogenic resistance genes while the latter raises the potential for transforming innocuous environmental species into pathogens.

Evolutionary opportunities

All resistance, regardless of the biological system, must be recognized as an evolutionary phenomenon. Our ability to combat cancer, in particular, would benefit from a shift towards viewing cancer as an evolutionary disease. A recent analysis of publications on cancer resistance and relapse since the 1980s showed that evolution terms were mentioned in only 1% of the abstracts. This finding is significant, as resistance to chemotherapies is often found in tumor samples taken before therapy and is driven by evolutionary pressures that kill the susceptible cells and allow the resistant ones to proliferate. Evolutionary thinking might prevent some of the pitfalls that arise when selection pressure is not considered during initial treatment plans. Re-conceptualizing cancer as a disease not to be cured, but to be managed, may also help in implementing treatments to minimize drug resistance and prolong life. Each biological system stands to improve treatment strategies by incorporating elements from others; the gift of “management” versus “eradication” from the herbicide and insecticide worlds may prove particularly valuable to other fields of study, as may a paradigm shift away from sequential drug treatment plans towards combination therapy approaches.

ALL RESISTANCE, REGARDLESS OF THE BIOLOGICAL SYSTEM, MUST BE RECOGNIZED AS AN EVOLUTIONARY PHENOMENON.

Ecological opportunities

While molecular and genetic techniques have enabled in-depth understanding of pathogens and identified specific means of combatting them, expanding our focus by taking into account the ecological roles of those species will greatly enhance the available tool kit. Drug developers will benefit from looking beyond the isolated laboratory setting, the cell, and the individual patient to consider the different types of environments encountered by pathogens or pests and how these can be manipulated or exploited to control outbreaks or infestations. In particular, malaria research could benefit if viewed as an ecological problem. Because of the multifaceted nature of the infection cycle, ecological approaches will reveal new places for therapies to intervene. Indeed, points of intervention exist throughout the parasite's life from its proliferation in the human host to its passage through the mosquito, which itself can be targeted at a multitude of life stages.

Additional opportunities for therapeutic inroads exist in the consideration of the local spatial dynamics of the target organisms. For drugs to be maximally effective, they need to encounter their targets equally; differences in doses experienced by the targets can determine whether the target is killed, slowed, or unharmed — which in turn affects resistance evolution in the survivors. Developing a greater understanding of biofilm dynamics and organization, for example, will enable more intelligent drug delivery systems so that individual cells are not shielded by their neighbors. Cancer cells in the center of tumors are likewise protected from chemotherapies and incorporation of tumor architecture into treatment plans is necessary.

Attention must also be given to the “ecology” of the patient as a whole. Immune and nutritional status and prior exposure to the pathogen may dictate specific courses of treatment. For example, data suggests that in Africa, 3 day quinine treatments were sufficient to cure malaria infections, while 7 days were required to achieve cures in Thailand. A high level of pre-existing immunity was thought to be present in the African patients, and that perhaps even vaccines which are not entirely preventative may still present an advantage in lessening the amount of drugs needed to fully fight infections. Such a practice could also lessen overall drug exposure, and thus the development of resistance.

In each of the above examples, drug resistance will clearly continue to be a prevalent problem unless steps are taken to include ecological characteristics and evolutionary pressures in the target selection, drug design, and treatment plan. Successful implementation of these plans will require cooperation from heath and regulatory agencies, drug companies, farming communities, hospitals and physicians, and individuals who understand that their actions have the potential to contribute to better outcomes for us all.

Management can prolong treatment effectiveness; every new treatment should have a management plan

After the initial rush of antibiotic discoveries, no new classes of antibiotics have been discovered since the 1970s. As resistance spreads across pathogens and environments, the development of new compounds becomes more urgent. To avoid the mistakes that led to emergent resistance among older drugs, colloquium participants stressed the need to make sure that access to any new drug is tightly controlled. For example, the group proposed that new drugs be given first to hospitals and placed under the control of the chief medical officer. Use of the drug should be restricted to prescriptions and only prescribed when accompanied with an accurate diagnosis. It was additionally advocated that the highest possible dose be given to patients to prevent resistance from developing.

What do we still need to know?

A common theme throughout the colloquium discussion addressed the need to better understand what specific drugs are available and how they work. Many old drugs have been abandoned because of toxicity risks, cost in production, or insufficient knowledge about the mechanism of action. By revisiting these drugs, experiments could uncover new target pathways or effective drug combinations. Further, if older drugs have been out of commission due to resistance concerns, sufficient time may have passed for new generations of targets to have regained susceptibility. These studies, along with continued efforts at identifying new drugs, will increase our arsenal of weapons against increasingly resistant pathogens and pests.

The mechanism underlying gene amplification is still an unknown, and yet represents a significant means of resistance evolution across multiple biological systems. Colloquium participants advocated further study of this phenomenon.

The propensity and speed of resistance evolution to different drugs should also be investigated through laboratory and computational studies; diversity of both domesticated lab strains of model organisms as well as environmental populations should be ascertained and incorporated into those experiments.