Introduction
The application of penicillin as a therapeutic agent in 1942 ushered in the era of antimicrobial chemotherapy and marks a historic milestone in medicine. Many additional classes of antibiotics spanning a broad range of chemical structures and targets soon followed, forming the foundation of the current armory of antibiotics. All of these agents are efficacious because they inhibit processes that are essential for bacterial growth. Unfortunately, this also introduces extreme selection pressure for resistant bacteria, which has resulted in an unprecedented acceleration of bacterial evolution that has culminated in the emergence of resistance to every approved antibiotic, and in some cases, multidrug-resistant bacteria that are increasingly difficult to treat1, 2, 3. Despite improvements in the biological, chemical, informational and technological tools driving drug discovery, new antibiotics that target either historically validated or newly proposed essential bacterial targets have been disproportionately difficult to identify4. Thus, the rapid emergence of clinically significant resistance over the last six decades has been accompanied by a dearth of novel antimicrobial agents5.
Medicinal chemistry's inability to produce new antibiotics and thereby outpace the emergence of resistance has motivated the exploration of new strategies that target nonessential processes, and that therefore might be less susceptible to the evolution of bacterial resistance. One such strategy is to disarm bacteria by inhibiting the production or activity of virulence factors and allow the host immune system to clear the infection. This approach is reviewed by Hung in this issue. A second novel strategy is to directly inhibit the processes that cause antibiotics to fail. Approaches to this strategy have been inspired by recent advances in our understanding of bacterial evolution and pathogenesis, especially in the context of antibiotic-mediated stress and selection. With most of these approaches, the long-term goal is not the development of compounds that kill bacteria, but rather compounds that can be used as co-therapies to improve and preserve the efficacy of traditional antibiotics.
Although mutation and horizontal gene transfer have long been appreciated as important forces in evolution, it has recently become apparent that in some cases they may be accelerated by the use of antibiotics to the point where they could contribute to therapy failure on the timescale of an infection. In addition to bona fide resistance, under certain conditions bacteria have a remarkable ability to tolerate antibiotic therapy; some manage to survive apparently lethal conditions until therapy ceases or bona fide resistance is acquired by mutation and/or horizontal transfer. Here, we discuss the potential of augmenting traditional antibiotic therapy through the inhibition of these processes. Specifically, we discuss (i) the inhibition of induced mutation to prevent the de novo generation of antibiotic resistance during therapy, (ii) the inhibition of horizontal DNA transfer to prevent the spread of pre-existing antibiotic resistance and (iii) the inhibition of antibiotic tolerance in bacteria that are not heritably resistant. While academically fascinating, these strategies are relatively new and completely untested as practical approaches to combating bacteria. Nonetheless, it is our expectation that advances in microbiology and bacterial pathogenesis (and in supporting technologies) will create new possibilities for such approaches in the future, especially as more and more traditional antibiotics are rendered obsolete by bacterial resistance. Because of the early state of the field, we focus on the potential contribution of these phenomena to therapy failure, possible targets for intervention and potential limitations.
Inhibition of mutation
In the case of several important antibiotics, including fluoroquinolones and rifampicin, point mutation of chromosomally encoded proteins is the primary mechanism of resistance6, 7, and for some bacteria such as Mycobacterium tuberculosis point mutation seems to be the only mechanism of resistance to any antibiotic8, 9. In many cases where therapy fails, infections appear at first to be susceptible to an antibiotic but then develop resistance during therapy. Several different processes might contribute to this behavior (Fig. 1). First, spontaneous mutations produced during genome replication may generate resistant cells that are selected during therapy (Fig. 1b). Such mutations could occur either before the initiation of therapy or during successive rounds of growth following antibiotic-mediated purges. Both cases may contribute to the evolution of high-level resistance, as pre-existing strains with low-level resistance may preferentially survive and acquire additional mutations during successive cycles of growth. Deficiencies in methyl-directed mismatch repair (MMR) may accelerate the acquisition of resistance during these growth cycles by elevating spontaneous mutation rates up to 800-fold10, 11. Interestingly, although MMR-deficient mutants arise with very low frequency during exponential growth of laboratory reference strains, more than 1% of clinical Escherichia coli and Salmonella enterica isolates may harbor MMR deficiencies12. Similarly, as many as 30% of chronic Pseudomonas aeruginosa infections in the lungs of individuals with cystic fibrosis may be hypermutable, and these hypermutators are significantly more likely to exhibit antibiotic resistance13. This data suggests that constitutive hypermutators may contribute to the emergence of resistance in the clinic, despite the fact that they are generally less fit under stable conditions14, 15, 16.
Figure 1: Emergence of mutation-mediated resistance.
(a) Proposed mechanism for adaptive mutation. Stress produces single-stranded DNA, which templates RecA filament formation. The RecA filaments induce mutation by recombination or by inducing the cleavage of LexA, which de-represses the error-prone polymerases. Each bold 'T' indicates a possible point of intervention. (b) Diagram of bacteria (spheres) during the course of antibiotic treatment (red arrows). Sensitive bacteria (yellow) are killed, whereas pre-existing resistant mutants (red) are selected, causing treatment failure. (c) Sensitive bacteria are killed, but induced mutation produces resistant mutants during therapy, causing antibiotic failure. (d) Inhibition of induced mutation during antibiotic therapy prevents evolution of resistant mutants, and the infection is cleared.
Full size image (51 KB)Another process that may contribute to the emergence of resistance during therapy is adaptive or induced mutation (Fig. 1c), which allows for a regulated increase of mutation rates during stress without encumbering the bacteria with constitutively high mutation rates. Adaptive or induced mutation occurs in nongrowing or slowly growing cells under stressful conditions, and it has been studied extensively with a +1 frameshift mutation in a lactose metabolism operon in E. coli17, 18, 19, 20, 21. In exponentially growing cells, the frequency of frameshift mutations that allow growth on lactose is less than one in 109 cells; however, when this number of cells is plated on a minimal medium containing lactose as the sole carbon source, approximately 200 colonies form during a seven-day period18. Interestingly, these experiments argue that, at least under the conditions examined, induced mutation contributes more to adaptation than spontaneous mutation, considering that more than 200 cycles of death and regrowth would be needed for spontaneous mutation to produce the same number of adapted cells as the induced process.
Unlike spontaneous mutation, induced mutation depends on homologous recombination, as well as induction of the bacterial SOS response (Fig. 1a)22. The SOS response is most commonly triggered by single-stranded DNA, which accumulates as a result of either DNA damage or problematic replication, and which templates the filamentation of RecA. Filamented RecA effects recombinational DNA repair, but also induces the autoproteolysis of a family of transcriptional repressors, including LexA. LexA is a negative regulator of multiple genes involved in DNA repair and cell division, which in E. coli includes three nonessential polymerases: Pol II, Pol IV and Pol V. Other nonessential polymerases are regulated in a similar manner by the SOS response of other bacteria, for example, DnaE2 in M. tuberculosis (see below). Mutational analysis demonstrates that adaptive reversion of the mutant lac allele during carbon starvation requires a host of proteins that mediate recombination, including RecA, as well as Pol IV, which supports a model in which low-fidelity synthesis by Pol IV is responsible for the increase in mutation22.
Some antibiotics, including fluoroquinolones and 
-lactams, are known to activate the bacterial SOS response23, 24, and they also seem to induce mutation. For example, prolonged exposure of E. coli to inhibitory concentrations of the fluoroquinolone antibiotic ciprofloxacin increases mutation rates by more than three orders of magnitude23, 25. Similar to the frameshift mutations required for lac reversion, the point mutations conferring ciprofloxacin resistance require components of the SOS response, including the RecBCD helicase/nuclease and RecA, as well as the LexA-regulated polymerases23. In what might be a similar process, prolonged exposure of P. aeruginosa to the antibiotic tetracycline causes a 10,000-fold increase in the rate at which tetracycline resistance emerges26, although the genetic dependencies of this process have not been investigated.
Mouse models of M. tuberculosis infection provided the first in vivo evidence that induced mutation may contribute to the evolution of resistance during therapy27. Deletion of the SOS-regulated polymerase DnaE2 reduces the emergence of rifampicin resistance during therapy relative to a wild-type control without altering the sensitivity to the drug. Furthermore, the average survival time of mice infected with a dnaE2-defective strain is nearly double that of mice infected with wild-type M. tuberculosis. A provocative though speculative explanation of this data is that the mutations introduced by the SOS-regulated polymerase are required both for the efficient adaptation to host defense mechanisms and for the development of antibiotic resistance27; however, it is also possible that the polymerase contributes to fitness under these conditions.
The contribution of induced mutation to the evolution of resistance during the course of therapy, relative to pre-existing resistance and spontaneous growth-dependent mutation, remains to be unambiguously determined. The prevalence of MMR-deficient strains in clinical isolates and the results of computational studies suggest that elevated spontaneous mutation rates can be advantageous under stressful conditions28. However, the experiments with wild-type E. coli, P. aeruginosa and M. tuberculosis suggest that at least under some conditions, spontaneous mutation during multiple rounds of death and growth is unlikely to compete with the number of resistant cells generated by antibiotic-induced mutation. Reports of clinical isolates that have additional SOS polymerases that are not present in reference laboratory strains may further reflect a role for induced mutation in the clinic, although this could also reflect a need for additional repair functions in these settings29. Deconvoluting the roles of spontaneous and induced mutation during antibiotic therapy is further complicated by the fact that the in vitro experiments demonstrating high rates of induced mutation do not mimic the fluctuating antibiotic concentrations, nutrient conditions or immune responses encountered during infections. Therefore, additional animal models and more sophisticated in vitro systems such as the hollow fiber infection model30 are required to establish the potential clinical significance of induced mutation before the development of drugs that target the process are pursued. If such studies suggest that induced mutation contributes significantly to the evolution of resistance during therapy, the components of the process, perhaps most promisingly RecA, might be targeted for drug development. Such inhibitors could be co-administered with conventional antibiotics, thus preserving their utility by preventing the evolution of resistance (Fig. 1d).
Inhibition of horizontal DNA transfer
In addition to point mutation, resistance may also be mediated by enzymes that either modify the antibiotic (for example, -lactamases or aminoglycoside acetyltransferases), modify the target protein (ribosomal methylases) or reduce the intracellular concentration of the antibiotic (efflux pumps). These mechanisms of resistance are commonly disseminated between bacteria by horizontal DNA transfer. For Gram-negative bacteria, conjugal transfer (that is, horizontal DNA transfer involving direct cell-to-cell contact) is primarily responsible for the dissemination of antibiotic resistance between bacteria and is particularly problematic for the dissemination of plasmids carrying extended-spectrum -lactamases31. In Gram-positive bacteria such as Staphylococcus aureus, conjugal transfer is also a concern. In 2002 the first high-level vancomycin-resistant strain of methicillin-resistant S. aureus (MRSA) was isolated, and resistance was attributed to a conjugal plasmid encoding the vancomycin resistance cassette vanA32. Since this time, several additional cases of MRSA with high-level vancomycin resistance have been reported, all of them having acquired the vanA genes by conjugal transfer33.
In Gram-negative bacteria, conjugal DNA transfer requires type IV secretion systems (T4SSs), multiprotein machines composed of a type II pilus and a proteinaceous pore that spans both the inner and outer cell membranes, indicating that the inhibition of these systems might reduce horizontal transfer (Fig. 2a)34, 35, 36. Genes encoding T4SSs are often found on plasmids or occasionally within transposons, which enables these elements to effect their own cell-to-cell transfer37. In addition to their role in horizontal transfer, T4SSs are required to secrete certain toxins, and efforts to inhibit Gram-negative T4SS-mediated toxin secretion may have serendipitously produced inhibitors of horizontal transfer. Biochemical screens identified submicromolar inhibitors of Helicobacter pylori Cag
, a T4SS NTPase required for pore assembly and injection of the CagA toxin into epithelial cells38, 39, 40. Subsequent whole-cell assays and mouse gastrointestinal infection models have confirmed that the inhibitors prevent CagA toxin injection in vivo38. The absolute conservation of Cag-like NTPases in T4SSs involved in conjugal DNA transfer suggests that the identified CagA inhibitors might intervene in both toxin secretion and conjugal transfer. Additionally, while conjugal DNA transfer in Gram-positive bacteria is less well understood and analogous T4SSs have not been identified, Gram-positive conjugal plasmids do contain Cag homologs41. Cag homologs are also found in type IV pili, which have been implicated in the natural competence of several bacteria42. Thus, Cag-like NTPases may be required for multiple facets of horizontal DNA transfer, and their inhibitors may prevent resistance dissemination, in addition to toxin production, although further studies are required to validate this strategy.
Figure 2: Horizontal transfer of resistance and virulence factors.
(a) Stress produces single-stranded DNA that templates RecA filamentation, induces the SOS response and induces phage mobilization and phage encoded toxin production. Also shown is T4SS-mediated conjugal transfer of a resistance-encoding plasmid. Each bold 'T' indicates a possible point of intervention. (b) Diagram of bacteria (spheres) during the course of antibiotic treatment (red arrows). Horizontal transfer of resistance elements from commensal bacteria (pink) to pathogenic bacteria (yellow) during antibiotic therapy produces resistant pathogenic bacteria (red), causing antibiotic failure. (c) Toxin-encoding phage from pathogenic bacteria (spiky yellow spheres) infect commensal bacteria (pink), thereby amplifying phage and toxin production (phage particles, blue) and complicating therapy. (d) Inhibition of horizontal transfer during therapy prevents spread of resistance elements and toxin-encoding phage; the infection is cleared without complications.
Full size image (59 KB)In one of the few efforts specifically directed toward identifying inhibitors of conjugal DNA transfer, a high-throughput screen was developed that quantifies conjugation by detecting transfer of a plasmid-borne bioluminescence operon43. From a collection of 12,000 natural product fractions, this screen identified two long-chain fatty acids: linoleic acid and dehydrocrepenynic acid. Both compounds decrease the transfer frequency of the conjugal plasmid R388 by approximately two orders of magnitude at 1 mM. Though the identified compounds were toxic at the concentrations used, they showed a level of specificity when tested against a panel of conjugal plasmids that use different T4SSs, which suggests that their effects may not result entirely from bactericidal activity. Although (i) the targets of these fatty acids are not known, (ii) their potencies are limited and (iii) their structures are not ideal for drug development, this effort demonstrates the feasibility of whole-cell screening of small molecules for inhibitors of horizontal DNA transfer.
As with induced mutation, the inhibition of horizontal DNA transfer would only be useful if it contributes to the failure of therapy during an infection (Fig. 2b). Interestingly, analysis of the bacterial strains present in the first person diagnosed with high-level vancomycin-resistant MRSA suggested that the vancomycin-resistant strain arose from a sensitive strain that obtained the vanA cassette during therapy from commensal vancomycin-resistant enterococci32. The recently reported transfer of a mupirocin resistance cassette from Staphylococcus epidermidis to a strain of MRSA during mupirocin prophylaxis further emphasizes the potential impact of DNA transfer during therapy44. These examples demonstrate the feasibility of conjugal transfer during therapy, and a recent report of SOS-stimulated conjugal transfer of antibiotic resistance demonstrates that, at least in one case, antibiotic therapy may accelerate this process45. However, the frequency with which these events occur in the clinic is unclear and must be determined before the potential utility of this approach can be evaluated.
Recently, another strategy for combating plasmid-based resistance has been proposed that is predicated not on inhibiting conjugal transfer, but rather on the selective killing of cells that harbor the resistance-encoding plasmids46. This selective killing could be achieved by the inhibition of the plasmid-encoded antitoxin of toxin-antitoxin addiction modules that are used by plasmids to ensure their retention in the host bacterium. Though the inhibition of chromosomally encoded toxin-antitoxin systems has been proposed as a general antibiotic strategy47, 48, this approach would likely be plagued by resistance resulting from the deletion of the nonessential toxin gene. Because plasmids are typically present in multiple copies per cell, targeting plasmid-encoded toxin-antitoxin systems should not be susceptible to this facile resistance mechanism, and thus might represent an effective strategy to combat plasmid-mediated antibiotic resistance46. Importantly, this strategy would be efficacious regardless of plasmid transfer frequencies during therapy.
In addition to the dissemination of antibiotic resistance factors, horizontal DNA transfer also facilitates the spread of genes encoding virulence factors and toxins, and in some cases toxin production is linked to the transfer process. For example, in a mouse model of enterohemorrhagic E. coli (EHEC) infection, transfer of the gene encoding Shiga toxin, the primary virulence factor of EHEC, transformed commensal bacteria into toxin producers, thereby amplifying toxin levels49 (Fig. 2c). Both transduction and toxin transcription require activation of the bacterial SOS response, and accordingly a recA mutant of EHEC that is incapable of inducing the SOS response has been shown to be dramatically less virulent in a mouse infection model50. Interestingly, SOS-induced phage-mediated horizontal transfer of toxin genes in S. aureus has also been associated with increased toxin production51. Because of this and other data, it has been suggested that antibiotics known to induce the SOS response, such as fluoroquinolones, should not be used in the treatment of some infections52. Inhibitors of the SOS response might therefore reduce toxin production and horizontal transfer of virulence factors, and importantly, they could generally reduce concerns over the negative side effects of antibiotic therapy (Fig. 2d).
Inhibition of antibiotic tolerance
Resistance permits bacterial growth in the presence of an antibiotic53; however, it is not the only factor contributing to treatment failure. Indeed, quickly following the initial discovery of penicillin was the discovery that this drug is unable to completely sterilize a culture of sensitive staphylococci and instead leaves a small fraction of viable cells, regardless of the concentration of the drug used54. Unlike resistant cells, this small viable population does not grow in the presence of penicillin and produces a sensitive population upon removal of the antibiotic. This phenotype is referred to as antibiotic tolerance, or simply tolerance (Fig. 3). Tolerance was subsequently observed with other bacteria, such as S. aureus, E. coli and P. aeruginosa, that were treated with other antibiotics, such as -lactams, aminoglycosides, tetracyclines and fluoroquinolones, which suggests that it is a general phenomenon55, 56, 57. The clinical importance of tolerance is reflected by cases in which antibiotics failed to clear infections despite the absence of resistant bacteria, and clinical reports suggest that the contribution of tolerance to treatment failure and mortality in some infections can be as significant as the contribution of antibiotic resistance58, 59, 60.
Figure 3: Tolerance of antimicrobial therapy.
(a) Currently hypothesized mechanism for antibiotic tolerance55, 61. Antibiotics (red triangles) disrupt essential processes in actively growing cells (left), leading to cell death. These processes are downregulated via nonlethal mechanisms in persister cells (right), which prevents antibiotic-mediated death. (b) Diagram of bacteria (spheres) during the course of antibiotic treatment (red arrows) and after cessation of treatment (purple arrow). Sensitive bacteria (yellow) are killed, whereas spontaneously formed persister cells (blue) survive and regenerate a sensitive population after cessation of therapy. (c) Sensitive bacteria are killed, whereas persister cells that form in response to the immune system (blue oval) survive and regenerate a sensitive population after cessation of therapy. (d) Inhibition of persistence prevents cells from surviving antibiotic treatment, and the infection is cleared.
Full size image (54 KB)The mechanism of tolerance is not fully understood, but the current hypothesis is that it results from a metabolically dormant subpopulation of cells called persisters (Fig. 3a)55, 61. It is proposed that persisters are not sensitive to antibiotics because the functions that antibiotics inhibit or corrupt (for example, transcription, translation, DNA replication and cell wall biosynthesis) are already downregulated in a nonlethal manner. This hypothesis is supported by the observation that -lactams are only active against dividing bacteria and that many antibiotics have decreased potency against slowly growing cells62, 63. Furthermore, transient ectopic overexpression of proteins that inhibit macromolecular synthesis increases antibiotic tolerance55, 64. Finally, transcription of genes encoding metabolism and motility functions is downregulated in E. coli persister cells, whereas transcription of sulA, which encodes an inhibitor of cell division, is increased55, 65.
The regulation of persister cell formation remains unclear, but studies are beginning to identify genes that might be involved. In E. coli, P. aeruginosa and S. aureus, a small percentage of wild-type cells (approximately 1%) spontaneously enter the persister state during growth in stationary phase. The persister state is lost in many cells upon return to exponential growth, where persister frequencies decrease to between 10-4 and 10-6 (ref. 66). HipA was the first protein identified that affects the frequency of persister formation, and a mutant allele, hipA7, causes a significant increase in the frequency of persisters in an exponentially growing population67. HipA was subsequently shown to be part of a toxin-antitoxin system, HipBA (ref. 68). Toxin-antitoxin systems are generally thought to be involved in inhibiting protein and nucleic acid synthesis in response to stress, which provides a possible rationale for the role of HipA in persistence48, 64, 69, 70, 71, 72, 73. However, neither deletion of hipBA nor the simultaneous deletion of the other five chromosomally encoded E. coli toxin-antitoxin systems has an effect on persister cell frequency74, 75, 76, 77, and overexpression of the mutant hipA7 allele increases persister cell frequency without significantly altering macromolecular synthesis78. Thus, the inhibition of macromolecular synthesis by toxin-antitoxin systems is not generally required for persistence. Furthermore, transient exposure of E. coli to bacteriostatic antibiotics failed to increase persister cell frequency, which suggests that inhibition of bacterial macromolecular synthesis is also not sufficient to cause increased persister formation78. Notably, the recent discovery that HipA is a protein kinase suggests a regulatory role for this protein in persister formation79.
While hipA contributes to persistence in E. coli, the absence or poor conservation of this gene in other bacteria that show antibiotic tolerance, including P. aeruginosa as well as most Gram-positive bacteria, suggests the existence of other mechanisms. A clever selection for hyperpersistent E. coli strains from an overexpression library identified GlpD, which converts glycerol-3-phosphate (G3P) to dihydroxyacetone phosphate. Overexpression of GlpD generates a mild increase in the frequency of persisters, whereas deletion of glpD decreases the frequency of persisters in stationary phase approximately 50-fold. Analysis of proteins that interact with GlpD or that affect G3P metabolism identified PlsB, an essential enzyme that converts G3P to 1-acyl-G3P for subsequent phospholipid biosynthesis, and mutations in plsB cause a decrease in tolerance similar to deletion of glpD. These findings suggest a possible role for G3P metabolism in tolerance and provide potential insight into persistence in pathogens that lack hipBA80.
A separate study in E. coli demonstrated that disruption of the gene encoding PhoU, a negative regulator of phosphate metabolism, significantly reduces the number of ampicillin- and norfloxacin-tolerant cells in logarithmically growing or saturated cultures. Importantly, treating a saturated culture of 109E. coli cells lacking phoU with ampicillin or norfloxacin eliminates all viable cells within three days. In stark contrast, a significant number of wild-type cells remain viable following seven days of identical antibiotic treatment (over 105 c.f.u. and 104 c.f.u. for ampicillin and norfloxacin, respectively)76. Transcriptional profiling revealed that in addition to its known role in regulating phosphate metabolism, PhoU negatively regulates multiple facets of nutrient transport, energy metabolism and cell motility, which further links metabolism to tolerance. It should be noted that the minimum inhibitory concentration of ampicillin and norfloxacin is reduced by two- and four-fold, respectively, in E. coli lacking PhoU; therefore, further studies will be required to determine the contribution of the increased sensitivity to the reported decrease in persister frequency. Nonetheless because of its profound effect on antibiotic tolerance and its conservation in Gram-negative and Gram-positive bacteria, PhoU may provide a suitable target for antipersister cotherapies76.
In addition to the spontaneous generation of a subpopulation of persister cells during saturated growth, some pathogens enter dormant states to avoid immune detection, and these states can also be refractory to many antibiotics81(Fig. 3c). Thus, the mechanisms required for long-term survival in vivo may also present targets for the inhibition of antibiotic tolerance. The paradigm for long-term survival is the dormant state of M. tuberculosis, which can survive within a host for years and necessitates long-term antimicrobial therapy82. Over 40 genes have been identified that contribute to M. tuberculosis virulence and persistence83, and interestingly, many of these genes seem to be involved in adapting the pathogen's metabolism for survival in macrophages. For example, mouse infection models demonstrate that M. tuberculosis lacking the nonessential gene isocitrate lyase and M. bovis BCG lacking the nonessential gene nitrate reductase are less able to survive within activated macrophages84, 85. Isocitrate lyase is also required for Salmonella enterica persistence in macrophages, and nitrate reductase is essential for anaerobic growth of P. aeruginosa in the lungs of individuals with cystic fibrosis, which suggests that the genetic requirements for long-term survival in vivo might be conserved in some cases86, 87. Although these proteins seem to be important for survival, few studies have analyzed the potential synergistic effects that deletion of these genes may have with antibiotic therapy; deleting them may allow clearance of the bacteria over reduced periods of time (Fig. 3d). These studies should be facilitated by new screening methodologies such as the recently reported luminescence-based low-oxygen recovery assay (LORA), which is suitable for high-throughput screening to detect antibiotic activity against nonreplicating or slowly growing bacteria88.
Limitations to novel therapeutic strategies
Studies of mutation, horizontal gene transfer, and antibiotic tolerance have demonstrated that each of these processes can contribute to bacterial survival and adaptation and the evolution of new traits. The question is no longer whether these processes are important, but whether they contribute to the evolution of resistance on the timescale of a single infection. Although there is a scarcity of clinical data addressing this question, analyses of the genetic changes that occur both during acute therapy and during prolonged periods of infection are beginning to appear89, 90 and should help address this issue. It will also be critical to develop more sophisticated models that better mimic the clinical conditions under which antibiotic tolerance is important and under which resistance reproducibly emerges, such as P. aeruginosa infections in the lungs of people with cystic fibrosis, latent M. tuberculosis infections, S. aureus endocarditic infections, and E. coli urinary tract infections. These clinical and model studies must be used to determine in which cases, if any, a given process contributes to the failure of antibiotic therapy.
If the clinical relevance of any of these processes is confirmed, it must be determined whether suitable drug targets exist. Genetic studies have and are likely to continue to identify genes that contribute to mutation, horizontal transfer, and tolerance; however, they also reveal that these processes may be complex, context specific, and in some cases, redundant. For example, RecA seems to be required for induced mutation and phage mobilization, but downstream proteins of the SOS response, such as LexA, are required in some cases yet seem dispensable in others91; in general, induced mutation and spontaneous mutation may both contribute to the evolution of antibiotic resistance. Additionally, many genes modestly affect the ability of certain infections to persist over long periods of time, but few appear to be absolutely required, and their contributions may also be context specific. For example, M. tuberculosis requires nitrate reductase for long-term survival in the lungs and liver but not in the spleen84. The proteins involved in persister cell formation and maintenance may be similarly context dependent. Context specificity extends to virulence factors, which suggests that it may be a common feature of nonessential processes that contribute to infection. For example, P. aeruginosa requires a type III secretion system to establish acute bronchitis and sepsis92; however, expression of this same system must be downregulated to establish a chronic infection in the lungs of individuals with cystic fibrosis89, 93. Reciprocally, Bordetella bronchiseptica requires a type III secretion system to establish a chronic infection but becomes hyperlethal in immunodeficient mice when it is absent94. Thus, in general, choosing the optimal targets within nonessential pathways may require a better understanding of the specific infections under consideration than is required for the development of traditional antibiotics.
Once a target has been validated and small-molecule inhibitors have been identified, the rate at which resistance emerges becomes a critical factor for clinical success. It has often been suggested that inhibitors of nonessential processes will not be subject to the evolution of resistance, owing to the absence of suitable selection pressures. Indeed, this is frequently used to justify alternative approaches to antibacterial therapy despite the availability of hundreds of putatively essential genes identified by genomic studies that remain unexplored. In many cases this seems like a reasonable justification, especially when the corresponding gene knockout does not show a fitness defect or when resistance in a small percentage of the population is insufficient to confer a fitness advantage. However, this is not always expected to be the case. For example, inhibitors of mutation or horizontal transfer might engender selection pressure if the target proteins are involved in normal DNA replication or repair. This could be especially true in environments that damage DNA (for example, the oxidative environment of a macrophage). In addition, virulence inhibitors could engender strong selective pressure under some circumstances. For example, type III secretion systems are frequently required to disable the host immune system or for host cell invasion and subsequent clonal expansion95. In these settings, a small number of resistant clones would likely be strongly selected from a largely sensitive population. In cases where selective pressures exist, some of these strategies may be particularly susceptible to the evolution of resistance, as the targets are typically encoded by single genes96. Of course, strategies targeting the mutation process itself may by definition avoid this problem.
Pre-existing diversity, in the form of target variants that are resistant to inhibition or multiple proteins that perform redundant functions, is a significant concern even in the absence of a selection pressure, and recent difficulties in targeting essential proteins highlight these concerns4. For example, while MetRS was determined to be essential in laboratory reference strains of Streptococcus pneumoniae, MetRS inhibitors failed against clinical S. pneumoniae isolates because of the presence of a second, resistant variant of MetRS (ref. 97). In another example, inhibitors of the essential E. coli enoyl-ACP reductase FabI lacked activity against S. pneumoniae because this organism uses another enzyme, FabK, to catalyze the same reaction98. Because essential genes have at least some evolutionary freedom to diverge within and between species, nonessential genes may diverge to an even greater extent, because they may do so without immediate consequence. Such accelerated divergence may explain the extreme sequence and functional diversity of T4SSs99. Although it remains to be shown with respect to many of the novel targets proposed in this review, pre-existing diversity could be one of the most significant shortcomings of targeting nonessential processes.
The difficulties associated with these and other new therapeutic approaches against infectious bacteria should not be taken as evidence of their futility. Rather, they should reinforce the importance of carefully selecting the most biologically validated targets. Early evaluation of therapeutic relevance, susceptibility to resistance, potential spectrum, and synergy with traditional antibiotics will increase the likelihood that these approaches will overcome the associated challenges. In addition, developments in supporting technologies should facilitate these approaches. For example, advances in diagnosis should help overcome the limited spectrum that will likely be associated with nonessential targets.
Perhaps more than any other class of drugs, antibiotics have revolutionized health care. In the pre-antibiotic era, death from bacterial infections was common, but the introduction of antibiotics reduced many previously life-threatening diseases to nuisances, vanquished with nothing more than a pill. However, the dramatic evolution of resistance since the introduction of antibiotics now threatens our society with a postantibiotic era. Though it seems likely that new antibiotics will be found, at increasingly staggering costs, it seems certain that the bacteria will immediately begin to erode their efficacy by evolving resistance, as they have with every antibiotic developed so far. Partnering a traditional antibiotic with a drug that prevents bacteria from evolving resistance or from persisting during treatment might dramatically improve therapy and protect the efficacy of these precious drugs.
Sa3ms is intimately linked with the phage's life cycle
