TARGETING EVOLUTION TO INHIBIT ANTIBIOTIC RESISTANCE

Abstract

Drug resistant bacterial infections have led to a global health crisis. Although much effort is placed on the development of new antibiotics or variants that are less subject to existing resistance mechanisms, history shows that this strategy by itself is unlikely to solve the problem of drug resistance. Here, we discuss inhibiting evolution as a strategy that, in combination with antibiotics, may resolve the problem. Although mutagenesis is the main driver of drug resistance development, attacking the drivers of genetic diversification in pathogens has not been well explored. Bacteria possess active mechanisms that increase the rate of mutagenesis, especially at times of stress, such as during replication within eukaryotic host cells, or exposure to antibiotics. We highlight how the existence of these pro-mutagenic proteins (evolvability factors) presents an opportunity that can be capitalized upon for the effective inhibition of drug resistance development. To help move this idea to move from concept to execution, we first describe a set of criteria that an “optimal” evolvability factor would likely have to meet to be a viable therapeutic target. We then discuss the intricacies of some of the known mutagenic mechanisms, and evaluate their potential as drug targets to inhibit evolution. In principle, and as suggested by recent studies, we argue that the inhibition of these and other evolvability factors should reduce resistance development. Finally, we discuss the challenges of transitioning anti-evolution drugs from the laboratory to the clinic.

INTRODUCTION

In 1859, Charles Darwin proposed that organisms evolve through natural selection. The process of natural selection acts upon beneficial phenotypic variations that, we now know, are generated through genomic changes. The coupling of mutagenesis and subsequent selection is responsible for the fascinating and beautiful diversity in living things that are found on planet earth. However, this same process of evolution is also continuously threatening human health by creating difficult-to-fight clinical problems.

In the era before Alexander Fleming’s discovery of the first antibiotic, in 1928, even simple infections that should be confined to an organ system could spread and become lethal. Antibiotics proved to be incredibly effective treatments, often helping to clear infections in a matter of days. However, it was also immediately obvious that the benefits could be temporary. In the wake of the tremendous advances made with antibiotics, the development of resistance to antibiotics arose, a problem which has now emerged as a global health crisis [1]. According to the World Health Organization, close to 800,000 people die from antibiotic resistant infections worldwide each year [2]. This number is on the rise, with predictions that roughly 10 million people per year will die from resistant infections by the year 2050. Therefore, it is imperative that we explore novel strategies which prevent resistance development in the first place.

Fundamentally, evolution is the mechanism responsible for the development of drug resistance. Despite this recognition, the majority of efforts to battle these problems have not dealt with evolution head-on, but rather are focused on new antibiotics or variants that are less subject to existing resistance mechanisms. However, history has shown that regardless of the nature or potency of drugs, resistance arises. As such, end-around strategies alone cannot solve the problem. Blocking the mechanisms that drive mutagenesis could remove the fuel from the fire of evolution. Here, we discuss the possibility of inhibiting evolution in the clinic as an underexplored strategy to prevent drug resistance development. In particular, we focus on bacterial infections, and the potential to inhibit the molecular mechanisms of mutagenesis that drive the evolution of antimicrobial resistance (AMR).

MUTAGENESIS IN ANTIMICROBIAL RESISTANCE

There are two general mechanisms by which genomic changes can result in drug resistance in bacteria: mutagenesis and horizontal gene transfer (HGT) [3]. Our focus on mutagenesis is based on the frequency with which it leads to resistance, as well as its major role in resistance associated with HGT. Chromosomal mutations that drive resistance often, although not always, arise within the genes that encode proteins being targeted. Mutagenesis in the target plays a major role in resistance development across broad categories of antibiotics. Such examples span mutations in topoisomerases with fluoroquinolones, alterations to RNA polymerase with rifampin, ribosomal mutations with translation inhibitors, and altered cell-wall penicillin binding proteins [4–6]. Resistance can also develop through gain-of-function mutations in genes encoding general transporters or efflux pumps, making drug export more efficient, or in regulatory elements that permit tolerance to the antibiotics [7–9]. In the case of HGT, plasmids or other types of DNA carrying antibiotic resistance genes are taken up by neighboring cells [10]. These mobile genetic elements often carry genes that can provide resistance to several different antibiotics in different antibiotic classes. Notably, these resistance elements are themselves subject to mutations, as evident in the modified version of β-lactamases that expand their spectrum of activity to encompass next generation cephalosporins or carbapenems [11]. Thus, blocking mutagenesis during treatment of infections could potentially inhibit the development of antibiotic resistance to a broad spectrum of antibiotics and in pathogens where resistance arises through mutations.

THE ORIGIN OF MUTATIONS

Studies in the mid-20^th century initiated by Salvador Luria and Max Delbruck suggested that mutations arise randomly, and that the occurrence of these mutations do not require exposure to selective conditions [12]. Although correct, a large body of work, some of which we discuss below, has since then has advanced this view by showing that mutagenesis is not a completely passive process and can be promoted by active mechanisms. These mechanisms that increase the rate of mutagenesis include the architecture of the genome, conflicts between replication and transcription, and exogenous changes in the environment (Fig. 1). As this revised view of the origins of mutations helps to frame strategies for potentially targeting evolution, we next briefly review some of the main mechanisms by which mutations can arise.

Figure 1. DNA transactions that contribute to mutagenesis.

Mutations in the genome arise from a variety of sources. Replication of DNA can occur with errors, and transcription is associated with higher levels of mutations. These DNA-associate transactions can come into conflict with one another or be influences by genomic features, such as structure or genomic location. The genome is also subject to various sources of damage, and stress-response pathways that are activated themselves be highly error prone.

The Replication Fork.

It is generally thought that the most common mechanism by which mutations occur is through errors attributable to the replisome and its interaction with proteins [7]. In other words, the fidelity of replication itself is assumed to be the main determinant of mutagenesis rates. In bacteria, the basal error rate of the replication process is quite low, roughly 10⁻⁹ to 10⁻¹⁰ mutations per base pair per round of replication [13]. This rate is determined by 1) the insertional fidelity of the replicative polymerase, DNA Pol III, 2) the efficiency of proofreading, and 3) mismatch repair (MMR). In E. coli, the error rate of DNA Pol III base selection is roughly 10–5 per nucleotide per generation, however, its 3’→5’ exonuclease activity increases the fidelity of the process by about 100-fold [14–16]. Finally, MMR, which can detect mismatched bases behind replication forks and facilitate their repair, improves the fidelity of the process by 10-to 100-fold [17].

Transcription.

Aside from the replication fork, although unappreciated, transcription has been shown to be mutagenic across organisms, from bacteria to yeast to human cells [18–20]. While the contribution of transcription to mutagenesis rates of a single gene is relatively low [21,22], the significance of this contribution could be quite meaningful given that 1) transcription is occurring frequently around the chromosome, 2) some important pathogens (such as Mycobacterium tuberculosis) gain resistance mutations under conditions where replication is not occurring, and therefore the mutations are unlikely to have originated from the replication fork [23], and 3) recent evidence showed that transcription-associated mutagenesis is critical for the rapid development of AMR (see Mfd below) [24]. Therefore, although the exact mechanism remains unknown, transcription should be considered a critical source of mutations and a major contributor to AMR development.

Endogenous/Exogenous Stressors and Repair.

Chemical changes in DNA and the resulting response to those changes can initiate and subsequently drive mutagenic DNA transactions. One common endogenous insult to DNA that increases the chances of obtaining mutations is the production of reactive oxygen species (ROS), which can arise for instance as a byproduct of metabolic processes [25]. This particular insult can also be produced exogenously, specifically during infections when the eukaryotic host cells bombard pathogens with ROS as a defense mechanism [26]. ROS are arguably one of the most common types of mutagens that generate genetic changes. Endogenously, aerobic cellular respiration oxidizes DNA through the production of superoxide (O2.-), hydrogen peroxide (H₂O₂) and hydroxyl radicals (·OH). These insults to DNA most commonly lead to the generation of 7,8-dihydro-8-oxoguanine (8-oxoG), which is corrected by a suite of repair enzymes that constitute the so called GO system in most bacteria [27]. However, this repair mechanism is not perfect and therefore 8-oxoG can increase mutation rates. The suite of stressors can also include other chemical agents, including alkylation, or photochemical reactions, such as the formation of thymine dimers. In addition to their direct role in introducing mutations, many of these genomic lesions lead to stalling of the replication fork or are recognized as DNA damage and activate repair pathways [28]. These triggered pathways often involve the use of non-replicative DNA polymerases, known as translesion (TLS) DNA polymerases, which have high error rates and thereby increase mutation rates [15,29].

Genomic Features.

When it comes to antibiotic development, genomic architecture, which is one of the less appreciated but potentially critical contributor to mutagenesis, should be considered. Not every gene mutates equally. Some genomic features that should be taken into account include 1) DNA secondary structures and sequence context which influence susceptibility to damage or polymerase fidelity [30,31], 2) repetitive sequences that are prone to insertion-deletions (indels) [32], 3) propensity for formation of pro-mutagenic R-loops, three stranded nucleic acid structures composed of an RNA:DNA hybrid [33,34], 4) high levels of transcription increasing mutation rates [35,36], and 5) gene orientation with lagging strand genes known to mutate more readily than leading strand genes [20,37]. These intrinsic genomic features can modulate mutagenesis through either direct or indirect downstream events including replication fork stalling, transcription-associated mutagenesis, translesion synthesis, or other (unknown or known) mutagenic repair mechanisms. Therefore, not only the mutational mechanisms, but also the nature of the genes that are mutated should be considered during the development of antibiotics. For instance, although speculative, given that the chance of resistance development will be much higher at those loci, genes that have a higher chance of gaining mutations are not optimal targets of antibiotics. Reciprocally, however, targeting mutagenesis may be particularly useful in such a situation.

DEFINING AN “IDEAL” ANTI-EVOLUTIONARY TARGET

It is clear that there are various mechanisms that can contribute to mutagenesis as described above. Which of these mutagenic processes, then, involve optimal targets for inhibiting evolution? Answering this question requires defining what constitutes an “ideal” evolvability target. Although the idea of targeting evolution has percolated for some time [3,38–40], the identification of ideal targets and dedicated efforts to antagonize them have, in our opinion, not received adequate attention. To help move these ideas forward, below we therefore offer a framework for defining desirable traits for anti-evolutionary therapy (Fig. 2). This framework can be used to help evaluate candidates that have either been identified or have not yet been uncovered.

Figure 2. Features of ideal anti-evolutionary targets.

Potential targets for slowing mutagenesis would ideally be associated with several features, some of which are distinct from conventional antibiotics (in bold) and others which are shared in common with conventional antibiotics (not bold). The fit of several candidate targets for anti-evolutionary therapy are highlighted. For each target, the assessment is based on inactivation of function. In the case of LexA this refers to proteolytic function and not DNA binding.

Synergism.

The most logical application of anti-evolutionary therapy would be in concert with a conventional antibiotic. Therefore, the mechanisms by which a potential anti-evolutionary agent and a conventional agent would work together are highly relevant. In the ideal scenario, inhibition of the anti-evolutionary target would not only slow the acquisition of resistance to the conventional agent, but also show synergy, permitting killing by a given conventional agent at lower levels of the antibiotic. While this is not strictly a requirement for the anti-evolutionary strategy, synergy could help better pave the pathway for drug development, as described in our concluding remarks.

Targeting non-essential pathways.

The ideal targets of conventional antibiotics are often characterized as being essential, with the view that disruption of essential pathways would result in pathogen killing. However, the same concept of essentiality is the driving force behind the evolution of resistance. This is simply due to the natural selection of randomly arising suppressors that survive the conditions that lead to the killing of the majority of the population. By contrast, the ideal targets for combating evolution would be ones where the fitness of the organism is not compromised by the inhibition of the target. In our concluding remarks we address the possibility of acquired resistance to anti-evolutionary agents, however, following the basic tenets of evolution the absence of fitness defects is anticipated to minimize the possibility of acquiring suppressors in the population that become resistant to the anti-evolutionary agent itself.

Targeting higher nodes.

Given the overall importance of genomic fidelity and adaptive pathways to bacterial survival, the networks involved in governing these responses can be complex and extensive. Targeting higher ‘nodes’ in the network, such as key regulators, may offer advantages over targeting more isolated downstream effectors. For example, shutting off transcription regulators such as RpoS (see below), which turn on multiple mutagenic polymerases, would be a better strategy than inhibiting mutagenic polymerases one at a time.

Molecular features of targets.

Identification of the best strategy to inhibit a given target will require a strong basic knowledge of the molecular functions and interactions of the candidate protein. In this regard, it is important to note that genetic validation of a target, for example via a knockout, by itself does not define the necessary features that a drug would need to target to mimic the knockout phenotype. In other words, if a protein has a large regulon, multiple interacting protein partners, or more than one enzymatic activity, it will be critical to know which of these features is of greatest significance with regards to its mutagenic activity so that small molecules targeting this feature can be designed with precision.

Uniqueness in and conservation across pathogens.

The ideal target should have limited functional redundancy. If the resistance-promoting pathways involve multiple similar targets in the pathogen, such as the redundant error-prone DNA polymerases, inhibitor design and effectiveness become a greater challenge. Ideally, anti-evolutionary drugs would also have breadth in their activity. In clinical circumstances, there is frequently a delay been onset of symptoms and isolation of the causative pathogen, or, at times, the pathogen may be altogether difficult to isolate. In these settings, broad spectrum conventional antibiotics are often employed. If the anti-evolution drug targets a highly conserved protein and therefore is effective in multiple pathogens, it could be used immediately, even before the infectious agent is identified.

Uniqueness from the host.

As with conventional antibiotics, the ideal anti-evolutionary agent should not impact any factors in the human hosts. While many proteins fundamental to DNA fidelity are common across phylogeny, those that are unique to prokaryotes would be particularly well-suited as anti-evolutionary targets.

POSSIBLE AND POTENTIAL CANDIDATE EVOLVABILITY FACTORS

Building on the framework for “ideal” anti-evolutionary targets, in the following section we present several candidates for this novel strategy. For each candidate, we consider their role in mutagenesis, progress on targeting the mechanism, and features that could either serve as advantages or present challenges to anti-evolutionary approaches. We then conclude this section by considering how screening-based approaches could be utilized to identify new targets for combating the evolution of antibiotic resistance.

SOS regulators: RecA and LexA

The bacterial DNA damage response, also known as the SOS response, offers two compelling candidates for an anti-evolutionary strategy: LexA and RecA [3]. In the absence of DNA damage, LexA acts as a repressor of DNA damage response genes. LexA/RecA-dependent genes appear to number less than ten in M. tuberculosis to over 40 in E. coli [41–43]. A common end product of various genomic stressors – including oxidative damage or metabolic stress – is the generation of single-stranded DNA (ssDNA) at stalled replication forks or during the processing of DNA lesions. The accumulation of ssDNA is sensed by RecA, which polymerizes to form an extended helical nucleoprotein filament (denoted RecA*) along the ssDNA. RecA* then transduces this “stress” signal to the regulator LexA, eliciting a latent serine protease activity within LexA that leads to its self-cleavage and degradation. The promoters liberated as part of the SOS response do not simply mediate high-fidelity repair; they also promote error-prone, pro-mutagenic repair, can accelerate genetic exchange via integrons, or promote tolerance pathways such as biofilm formation [44–49].

The SOS response can be genetically inactivated by either loss of RecA or by inactivation of LexA’s self-cleavage ability [50,51]. Such changes to the LexA/RecA axis result in a significant impairment of antibiotic-induced mutagenesis across most major classes of antibiotic agents [51]. In infection models, a bacterial strain harboring protease-deficient LexA fails to acquire resistance to ciprofloxacin or rifampin, supporting the viability of LexA as an anti-evolutionary target [50]. Beyond acquired resistance, SOS inactivation can also increase sensitivity to DNA damaging antibiotics, and even re-sensitize resistant bacteria to that antimicrobial agent [51,52]. Although some bacteria have added DNA damage responses that are LexA/RecA-independent, the fact that the SOS regulon almost invariable includes an error-prone polymerase makes the LexA/RecA axis an attractive target [53].

While the molecular interface of RecA* with LexA that governs SOS activation remains poorly understood, both have been successfully targeted with small molecules. DISARMERs, “drugs that inhibit SOS activation to repress mechanisms enabling resistance”, have been uncovered through various screening approaches [54–56]. Although RecA is likely a less ideal target than LexA (below), small molecules that inhibit RecA polymerization have been identified, with some classes appearing to slow fluoroquinolone-associated mutagenesis [56]. High-throughput screening for inhibitors of LexA cleavage has also revealed several classes of leads that can target the RecA*/LexA axis in vitro and prevent SOS activation in cells, including optimized leads that reduce mutagenesis and acquired resistance associated with antibiotic exposure [54,55].

Targeting either LexA or RecA has several potential advantages, including their place as high ‘nodes’ in the DNA damage response and a high degree of conservation across pathogens [57,58]. The target also permits dual mechanisms of action: SOS inactivation both abrogates acquired resistance and also increases the potency of conventional DNA damaging [51]. Assessing the viability of RecA versus LexA as targets suggests that LexA may be a more suitable target for DISARMERs. As recA null bacteria show compromised fitness relative to their wild-type partners [59–62], the drive towards resistance will likely be high. By contrast, strains with inactivated LexA compete effectively with the wild type strain in the absence of stress, decreasing the theoretical drive towards acquired resistance [51]. Furthermore, between the two targets, only LexA lacks human homologues, as host DNA repair pathways make use of numerous Rad51 family members with homology to RecA [63].

Despite these potential strengths as a target, from the aspect of “druggability” both RecA and LexA pose significant challenges. Inhibiting polymerization of RecA requires targeting a complex protein-protein interface, while its ATP binding site lacks uniqueness among proteins [64]. For LexA, targeting an intramolecular self-cleavage reaction where the substrate and enzyme are one and the same offers a sizable challenge and makes it less likely that competitive inhibitors will be viable agents [65].

Error-Prone DNA Polymerases

Error-prone TLS DNA polymerases offer another class of evolvability factors that could be targeted. As noted, most error prone polymerases can replace the high-fidelity Pol III at the replication fork upon stalling of the replication machinery or can be employed in repair pathways, including the SOS response or RpoS-mediated stress responses (detailed below) [66,67]. These TLS polymerases differ from Pol III in that they have open active pockets that decreases their ability to select the correct nucleotide [68]. The strongest evidence for a link between error-prone polymerases and antibiotic resistance comes from Mycobacterium tuberculosis, where deletion dnaE2, a component of the SOS-regulated error prone polymerase, resulted in a marked decrease in acquired resistance to rifampin in a mouse model [69]. This finding is of high clinical relevance given that rifampin is a frontline anti-tuberculosis therapeutic. Interestingly, human Y-family DNA polymerases appear to play a parallel role in acquired resistance to anti-cancer agents. While no specific inhibitors of bacterial translesion polymerases have been described, an anti-evolutionary small-molecule agent that targets human DNA Pol β has recently been reported and appears to improve the outcomes of chemotherapy in model systems [70].

However, applying our criteria for ideal features for anti-evolutionary strategies suggest challenges to targeting error-prone polymerases. On one hand, although the fitness of mutants has not been extensively evaluated in the absence of stress, deletions of error-prone polymerases generally appear to be well-tolerated making them reasonable targets [71]. Additionally, they offer discrete molecular targets, such as the dNTP binding pockets, which are likely druggable. However, as downstream effectors, these polymerases are not high nodes in adaptive pathways. Perhaps the most significant barrier, relates to their lack of uniqueness. Many pathogens harbor multiple error-prone DNA polymerases with overlapping functions. Furthermore, the shared features between bacterial and human error-prone polymerases offers an added impediment to targeting this class of proteins.

Sigma Factors

Transcription regulation in bacteria is generally achieved through differential regulation of sigma factors. Sigma factors direct RNA polymerase to specific promoters of genes that need to be expressed during various conditions. In the context of this discussion, the most relevant sigma factor to consider is RpoS and its homologues or analogs in other pathogens, as RpoS is not conserved across bacteria. RpoS is a stationary phase and stress response regulator that controls the expression of over 100 genes [72].

Two of the main stresses that turn on the RpoS-mediated gene expression program are oxidative stress and other types of DNA damage. When activated, RpoS turns on many genes important for defending the cell against ROS, which are of great relevance for pathogen adaptation and survival during infection of eukaryotic cells. Importantly, error prone polymerases that can increase mutation rates, as noted above, are within the RpoS regulon. A broad screen of factors that impact mutagenesis add support to the notion that RpoS could be considered an evolvability factor [73]. Edaravone, used to treat patients who have suffered strokes or have ALS, was shown to alter the frequency by which the RpoS program is activated in bacterial cells. However, these effects were indirect: edaravone is a general antioxidant and not a direct inhibitor of RpoS [74]. Therefore, although in principle reducing ROS could be a useful strategy to reduce mutagenesis, such an approach (e.g. edaravone treatment) will likely be highly detrimental to the clearance of infections given that host defenses rely on ROS production. Despite the promising results of this recent study, specific RpoS inhibitors that may be viable anti-evolution drugs are yet to be discovered.

Given that sigma factors are in many cases at a high node of gene regulation, if they promote mutagenesis and evolvability, they could be optimal targets of anti-evolution drugs. However, targeting these regulators, and in particular RpoS, is likely a double-edged sword. Although RpoS increases mutagenesis and therefore can likely promote evolution, RpoS is also (presumably) essential for survival of many stresses (e.g. ROS) that pathogens face during infection of the host cells and during exposure to antibiotics [75]. Therefore, inactivation of RpoS will likely put strong selective pressure on a given pathogen population, leading to the generation and selection of resistance against the anti-evolution drug itself. Furthermore, while some other candidate factors offer discrete molecular targets, it is not clear how small molecules should be employed to phenocopy the effects seen with deletion of rpoS. RpoS is also not conserved outside of proteobacteria. Additionally, little is known about the uniqueness of the target, which may pose significant challenges in efforts to derive broadly-acting agents with limited off target activity.

Mfd

Recently, the DNA translocase Mfd was discovered as an evolvability factor that is critical for rapid development of AMR [24,37,76–78]. Mfd interacts with stalled RNA polymerases, pushing them off of DNA [79]. For over four decades, Mfd has been thought of as a transcription-coupled repair factor [80]. This model was largely developed based on in vitro studies, which employed UV exposure, and examined the interaction of Mfd with UvrA, which is required for initiation of nucleotide excision repair (NER). However, cells lacking Mfd do not show any survival defects when exposed to various DNA damaging agents, including UV, which stands at odds with Mfd’s predicted DNA repair function [24]. In a few in vivo studies using exposure to UV damage, Mfd reduces mutagenesis during recovery [81]. However, in many other studies where exogenous DNA damage was not imposed on cells, the results were the opposite: when Mfd is absent, mutation rates go down, which showed that Mfd actually promotes basal levels of mutagenesis [24,37,76–78].

In a recent study, the role of Mfd in mutagenesis was extensively tested in a number of highly divergent species. In all cases, the presence of Mfd increased mutation rates [24]. Follow-up studies enforced the relevance of these findings for the development of AMR in culture and during infection of eukaryotic cells formation [24]. Importantly, the interaction of Mfd with RNA polymerase was required for its ability to accelerate resistance development, highlighting a critical role for transcription-associated mutagenesis in AMR development, and helping to isolate a possible target for rational design of small molecule antagonists.

Although the downstream mechanism of how exactly Mfd promotes mutagenesis is still unknown, this evolvability factor is particularly attractive because it meets the basic criteria for an ideal target. First, Mfd knockouts do not show fitness defects in culture or during infection of eukaryotic cells. Therefore, inhibition of Mfd is unlikely to promote the rise of suppressors within a population of pathogens during infection. Second, there are no proteins with similar functions that exist within bacteria. Third, Mfd is highly conserved across all pathogens and therefore, a single anti-evolution drug that inhibits Mfd could act as a broad-spectrum anti-evolution drug. Fourth, Mfd has multiple, theoretically druggable, critical domains, including an ATPase pocket, translocase domain, and other important scaffolds such as its RNA polymerase as well as UvrA interaction domains. While multiple molecular features make targeting Mfd a feasible undertaking, more work will be necessary to understand which of these domains should be a priority to target. Lastly, there is no homologue of Mfd in eukaryotes, although a functional and sequence divergent analogue, CSB, does exist [82]. It is unclear whether Mfd inhibition will permit synergy and increase susceptibility to antibiotics. This is one of the desirable features of an ideal target that in the case of Mfd, which needs to be further investigated.

Strategies for identification of new evolvability factors

Although promising and challenging candidates already exist for anti-evolutionary strategies, expanding the potential targets can help increase the likelihood of success with this approach. Conventional drug discovery efforts have often begun with identification of essential genes via genetic approaches, although many such targets end up failing on the path to drug development [83]. Identifying evolvability factors in bacteria, however, is much more challenging. As mutation is a relatively rare event, decreases in the rate or mutagenesis can be difficult to detect with sufficient throughput to uncover targets for anti-evolutionary approaches. As an added challenge, mutation is also stochastic, with variability from experiment to experiment as a defining feature.

Despite these challenges, several screening-based approaches have been used or could be adapted to identify new evolvability factors. In one such approach, using a colorimetric readout to detect within-colony mutations in β-galactosidase offered a means to identification of genes that mediate pro-mutagenic response in E. coli [73]. This approach notably supported a prominent role for SOS-, RpoS- and RpoE-associated stress responses, but also identified targets of unknown function. The generalizability of these targets in different pathogens, and their role in response to antibiotics as opposed to other forms of stress, could be particularly useful to more specifically identify new therapeutic targets. Clever selection strategies have also been employed to select for anti-mutator alleles that maintain plasmids at high fidelity [84], and analogous approaches could potentially be used to find new evolvability factors. Finally, non-genetic, proteomic-based approaches have been used to identify upregulated proteins under oxidative stress, and these targets may well be factors that mediate adaptive or pro-mutagenic responses [85].

It is interesting to consider how other novel technologies could be harnessed to help identify candidate evolvability factors. The direct visualization of step-wise antibiotic resistance in the “mega-plate” offers kinetic insights into the steps on the pathway to acquired resistance [86]. If such visualization platforms could be utilized for high throughput screening of genetic variants, factors that regulate critical steps in acquired resistance could be identified. Rather than tracking evolutionary pathways in a population, imaging at the single-cell level also offers the possibility of identifying features in variants that are able to escape from antibiotic stress and could inform new strategies for targeting evolution. Overall, the movement towards considering the fate of individual bacterium and their evolutionary trajectory is likely to help shed light on adaptive mechanism that could be future areas for exploration.

FUTURE CHALLENGES AND GOALS

Targeting mutagenesis is fundamentally about dealing head-on with the mechanisms by which antibiotic resistance arises. We have argued here that the clinical challenge we face today demands unconventional approaches, such as blocking evolution, and that there is significant opportunity for innovation in building on both known targets that regulate mutagenesis as well as discovering new ones. Nonetheless, it goes without saying that the challenge posed by such approaches are significant. In conclusion, we wish to highlight the conceptual challenges to anti-evolution strategies, as well as the logistical challenges posed by moving from concept to potential pharmacological reality.

The existence of previously mutated cells.

If mutations that lead to resistance are already present in the population of the pathogens being treated, prior to exposure to antibiotics, then the anti-evolution drug strategy by itself is unlikely to be effective. However, if the population is sensitive to a particular drug prior to therapy, then the antibiotic would be expected to benefit from being combined with the anti-evolution therapeutic. Sequencing studies are beginning to shed light on the extent of within-host genetic heterogeneity exists in bacterial infections, and the possibility of resistance existing prior to initiation of an anti-resistance agent poses perhaps the biggest concern for anti-evolution strategies [87–89]. From a clinical perspective, however, we can speculate that this potential problem will be more relevant for some infections than others. The multiplicity of infection will play a major role in this process: if only a few cells are necessary for the infection, then the chances of having a pre-existing mutant in the population will be quite low. Furthermore, the rate of expansion of the infectious population is also likely to be relevant to diversification. In practice, assessment of “baseline” number of pre-existing mutants for various pathogens and in various infections should be priority for study to best strategize about clinical scenarios that could benefit from targeting evolution.

HGT.

A second major conceptual challenge is the contribution of HGT-driven AMR, however this may also represent an opportunity. The current strategy for treating bacterial infections that are resistant to a particular antibiotic due to horizontally-transferred resistance genes is to use alternative antibiotics. Could the anti-evolution strategy help in these cases? The answer is arguably yes. Treatment of a resistant infection with a second antibiotic to circumvent the problem faces the same evolution-driven challenges as those described above. Mutations can drive resistance development to the alternative antibiotic as well, and the combined used of an anti-evolution drug could be quite beneficial to maintain susceptibility. Furthermore, many of the mechanisms by which HGT is activated involve the same genotoxic stress-responses that contribute to mutagenesis [44]. Thus, targeting pro-mutator pathways may in fact also slow HGT.

Although HGT could be a major roadblock to effectively combating AMR, there are some noteworthy cases where this mechanism does not seem to play a role in resistance development. In particular, there has not been any report of HGT-driven resistance development in Mycobacterium tuberculosis, the agent which causes tuberculosis and is arguably the most difficult infection to treat due to drug resistance [90].

Development of drugs.

Finally, moving anti-evolutionary approaches from concept to reality will also require strategic and logistical innovations in the drug-development process, specifically in the selection of drugs to partner together and in the regulatory pathways that could lead to their approval.

The most likely pathway towards advancement of an anti-evolutionary therapy will be in partnership with a conventional antibiotic. Knowledge of the mechanisms involved in resistance development to the conventional antibiotic could be used to determine the best strategy for partnership. For example, as fluoroquinolones are both potentiated by and less prone to resistance in SOS inactive strains, it would be logical to explore a partnership between fluoroquinolones and DISARMERs [51]. Similarly, if particular target genes of another antibiotic are prone to lagging strand mutagenesis, they may be ripe for combination with an Mfd inhibitor. One added possibility offered by anti-evolutionary strategies would be the revitalization of antibiotic candidates that were discarded due to concerns about acquired resistance. For example, effective agents targeting leucyl-tRNA synthetases have been developed, but stalled due to frequent acquired resistance in phase II trials [91]. Much like cilastatin can potentiate imipenem by altering its pharmacokinetics, an anti-evolutionary drug could partner with the tRNA synthetase inhibitors to advance them to clinical use. As has been recently expertly reviewed, while combining multiple agents in therapy presents opportunities, it also presents pharmacological challenges, including optimizing dosing and addressing combined toxicities [92].

Regulatory pathways may also need to be altered to accommodate novel anti-evolutionary approaches. The conventional path to approval of antibiotics involves easy-to-assess infections, such as pneumonia or skin and soft tissue infections, with clinical resolution of infection as an endpoint. Acquired antibiotic resistance is a highly concerning, but less common clinical endpoint, which makes it harder to design trials around. As noted above, the fact that we still do not know the situations where clinical failure is due to pre-existing resistance, or whether it develops on therapy makes it even more challenging to select the right trial design.

Development of resistance to the anti-evolution drugs.

It is possible that a pathogen may become resistant to the anti-evolution drug itself. However, shutting off the very mechanism that promotes mutagenesis will already decrease the chances of such mutations arising. Choosing targets that, when inactivated alone, do not lead to phenotypic defects is expected to decrease the likelihood of acquired resistance. In the long-term evolutionary time scale, selection (or lack thereof) that skews mutation rates can certainly occur. For example, patients who are colonized for decades with strains of Pseudomonas aeruginosa acquire hypermutator phenotypes[93]. However, such selection is less likely to be a problem during the short-term treatment periods in which antibiotics are used to treat acute infections. While studying resistance in culture is common for traditional anti-bacterial agents, it will be particularly important to evaluate fitness dynamics of cells in the presence of anti-evolutionary therapies to directly probe this model. Notably, while this framework for considering the risks of acquired resistance to therapy has been part of the rationale for other approaches to target non-essential genes such as anti-virulence approaches, the experimental evidence confirming a decreased risk of acquired resistance remains limited and remains a challenge that will need to be addressed [94–97].

The challenge posed by bacterial resistance to antibiotics are tremendous. The scope of the problem reflects the power of evolution as a force that can reshape genomes and offer pathogens a route to evade our best therapeutic strategies. Our hope is that, as a complement to the many creative approaches being taken to develop new conventional antibiotics, targeting the factor that accelerate diversification of genomes can be a means to potentiate our antimicrobial arsenal. Although targeting evolution will not be easy, with the framework we have offered here, we believe it is possible, and, further, that it is imperative.

Abbreviations:

HGT

Horizontal Gene Transfer

AMR

Antimicrobial Resistance

DISARMERs

Drugs that Inhibit SOS Activation to Repress Mechanisms Enabling Resistance

NER

Nucleotide Excision Repair

UV

Ultraviolet

ROS

Reactive Oxygen Synthesis

TLS

Translesion Synthesis

MMR

Mismatch Repair

SOS

Save Our Souls