Karl Drlica is a Member of the Public Health Research Institute and Research Professor of Microbiology, New York University School of Medicine, New York, N.Y.
Links to Other ASM Pages:
A Strategy for Fighting Antibiotic Resistance
Administering drugs to force pathogens into attaining two or more mutations to grow could restrict development of resistance
Karl Drlica
Antibiotic resistance among human pathogens has now entered the multidrug-resistant (MDR) phase. For example, some strains of Staphylococcus aureus, a major source of hospital infection, are resistant to all but one of the common antibiotics. And the pathogen responsible for many respiratory infections, Streptococcus pneumoniae, is rapidly losing its susceptibility to penicillin and other agents. Meanwhile, MDR tuberculosis, which is rampant within the Russian prison system and has arisen in other locations, threatens to become a global epidemic. Although scientists in academic settings and within the pharmaceutical industry are refining current antibiotics and searching for new ones, obtaining more potent compounds will have little long-term effect if we cannot protect them from being quickly neutralized by resistance.
Several strategies have been advocated to slow the development of resistance. One is to lower antibiotic consumption. Resistance to antimicrobial compounds correlates with use, and physicians are advised to avoid antibiotics when symptoms are likely to be due to viral infection. Experimental support for reducing consumption comes from studies in Iceland and Finland in which the prevalence of resistance declined when use was cut. In general, however, the results of such antibiotic conservation efforts have been disappointing.
Another strategy involves antibiotic cycling. This method, which is much like crop rotation, relies on resistant organisms being at a selective disadvantage when antibiotic pressure is removed. Such is not always the case. Moreover, compensatory mutations can arise that eliminate the disadvantage. Neither of these two approaches is likely to preserve our present antibiotic arsenal.
We have developed a third approach that entails new dosing strategies for antibiotics. The key idea, familiar to microbiologists but untested clinically, is that few mutants will be recovered if antibiotic concentrations require a cell to attain two or more concurrent mutations for growth in the presence of the drug (current dosing protocols generally allow a cell to grow after it develops only one resistance mutation). Dosing according to the two-mutation idea could help restrict the selection of resistant mutants. However, even if it proves effective, implementing this strategy would require refocusing some of the effort from finding new antibiotics to preserving existing ones. It may also require a shift in emphasis from individual to public health on the part of physicians and patients.
The Mutant Selection Window
Conceptually, selection of resistant mutants occurs within a specific drug concentration range called the mutant selection window. The lower boundary of the window is the lowest drug concentration that blocks the growth of the majority of drug-susceptible cells. This concentration can be approximated by the minimal inhibitory concentration for half the cells in the population (MIC50). The upper boundary of the window is the drug concentration that blocks the growth of the most resistant, single-step mutant. Above this boundary, cell growth requires two or more resistance mutations.
Because double mutations occur rarely, few mutants will be selectively amplified when drug concentrations exceed the upper boundary. For example, with fluoroquinolones the mutation frequency for resistance is less than 10-7, and so more than 1014 bacteria (107 x 107) would be required to find a cell with two concurrent fluoroquinolone-resistant target mutations. In a typical clinical case, bacterial populations may reach 1010 cells within an infected individual, but 1014 is unlikely. Thus resistance is expected to develop rarely when drug concentrations are above the upper boundary of the mutant selection window. This idea has led to the upper boundary being designated as the mutant prevention concentration (MPC).
Experimentally, we can demonstrate the mutant selection window by treating mycobacteria in vitro with fluoroquinolones. When cells are applied to a series of agar plates containing various concentrations of fluoroquinolone, the fraction of cells recovered drops sharply at drug concentrations that block wild-type growth (Fig. 1A). These inhibitory concentrations represent the lower boundary of the window. Further increases in fluoroquinolone concentration generate a plateau in colony recovery, followed by a second sharp drop. The plateau is due to colony formation by resistant mutants present in the population.
The second sharp drop occurs at the drug concentration that blocks the growth of mutants. We find that the MIC for the most resistant first-step mutant correlates with the concentration required to cause the second drop in mutant recovery. Fluoroquinolone concentrations higher than required to cause the second sharp drop, the concentrations that determine the MPC or upper boundary of the window, allow the recovery of no mutant even when more than 1010 cells are plated.
Some Pathogens Contain More Than One Drug Target
Bacterial pathogens other than mycobacteria often contain two fluoroquinolone targets, namely DNA gyrase and DNA topoisomerase IV. In such cases the requirement for two concurrent mutations can be met with drug concentrations that exceed the MIC of the less susceptible of the two targets. Since target preference can be changed by altering drug structure, it may be possible to identify compounds that attack both targets equally. Then MIC would equal MPC, and the mutant selection window would be closed. Structural changes in fluoroquinolones are gradually reducing the gap between MIC and MPC (Table 1).
The window between MIC and MPC can also be depicted with pharmacokinetic profiles in which serum or tissue drug concentration is expressed as a function of time after drug administration (Fig. 1B). In principle, the MPC serves as a dosing threshold: if serum/tissue drug concentration can be kept above the MPC during therapy, few, if any, mutants should be selectively amplified. For this idea to be clinically applicable, the drug concentration attained in relevant human tissues must exceed the MPC for clinical isolates. For some C-8-methoxy fluoroquinolones that inhibit Mycobacterium tuberculosis and S. pneumoniae, serum drug concentrations do exceed MPC (Fig. 2).
However, such is not the case for every pathogen-antibiotic combination. For example, we have readily measured MPC values for compounds such as isoniazid with M. tuberculosis, but the MPC exceeds the serum drug concentration that can be safely achieved. In other cases, such as rifampin tested against Escherichia coli, the MPC is not reached even at very high drug concentrations. For these two situations, any tolerable dosage produces a drug concentration that is either ineffective (below MIC50) or selective for mutant growth (within the mutant selection window). A compound whose concentration cannot be maintained above the MPC is expected to succumb to resistant mutants if applied as monotherapy; such agents need to be protected by being administered as a part of a combination therapy.
Strategies To Prevent Resistance Can Depend on How It Develops
Antibiotic resistance mutations fall into two general classes. In one, the mutations reduce susceptibility so much that no tolerable concentration of the drug can block growth. Within a bacterial population, individual organisms are either fully susceptible or fully resistant. The best known examples involve resistance determinants carried by mobile genetic elements. They enter a population horizontally, often after having achieved high-level resistance elsewhere. MPC-based monotherapy is not suitable for such cases even when the fraction of resistant mutants is initially very low. Preventing the enrichment of this type of mutant requires a strategy of multidrug therapy that pays heed to specific pharmacokinetic requirements.
For antibiotic resistance mutations within the second general class, the changes typically are not fully protective against the drug, and antibiotic concentrations can be found that block growth of mutant cells. Resistant populations develop stepwise through the gradual accumulation of mutations that separately confer low-to-moderate levels of resistance. When individual cells are tested, intermediate levels of resistance are observed. In these situations, MPC-based therapy is expected to restrict mutant enrichment.
An example of this second pattern is the development of penicillin resistance in S. pneumoniae. Treatment failure due to penicillin resistance was initially so uncommon that low-dose therapy was deemed adequate. Low-level-resistant mutants were gradually enriched and spread to fresh hosts. Then compounds with greater potency were required to cure infections, and mutants having even higher levels of resistance were selectively enriched. In some countries, penicillin is now ineffective against half of the S. pneumoniae isolates.
Stepwise development of resistance will occur most readily when a bacterial population contains a diverse array of resistant mutants associated with many different levels of susceptibility. With the fluoroquinolones, for example, more than 20 different resistance mutations have been found in the target gyrase genes of mycobacteria. In addition, mutants having nontarget resistance, presumably involving drug efflux, are so abundant at low quinolone concentrations that few target mutants are selected even though they are associated with higher levels of resistance. The same principles probably also apply to S. pneumoniae, suggesting that fluoroquinolone resistance is likely to develop gradually with this pathogen.
MPC Appears Better than MIC Breakpoints in Forestalling Clinical-Level Drug Resistance
Although the idea of gradual, stepwise enrichment of resistant mutants fits well with in vitro observations, its clinical relevance has not always been obvious. For one thing, resistance seems to arise suddenly among patients. For instance, the percentage of penicillin-resistant isolates of S. pneumoniae increased by a factor of 10 in only three years. Moreover, an explosive expansion of resistance is known to be due to clonal spread of pathogens such as S. aureus and S. pneumoniae.
Using the MIC to monitor resistance, as is often done, may distort the apparent rate that clinical resistance emerges because standard MIC-based testing usually examines fewer than 105 cells. An infected individual whose tissues contain 109 cells of a particular pathogen could harbor more than 104 resistant but undetected mutants. If the number of resistant mutants that can go undetected is near the threshold that overwhelms host defenses, then the time between infrequent detection of resistance and abandonment of a compound due to widespread loss of susceptibility would likely be very brief.
A threshold effect also helps explain why treatment failure is rarely attributed to the development of resistance during therapy. Until the threshold is reached, therapy will generally be successful. If the threshold is quickly crossed and resistant pathogens disseminate, most of the resistance-dependent treatment failures will arise because the pathogen had lost susceptibility before, rather than during, the round of infection being examined.
If the MIC is indeed insensitive to an early buildup of resistant mutants, MIC-based measurements would be inadequate for determining whether a compound is suitable for monotherapy against a particular pathogen. Then the common clinical practice of using MIC "breakpoints" for determining suitability would be providing a false sense of security. The practice may even contribute to the selective enrichment of antibiotic-resistant mutants by encouraging the use of marginal compounds as monotherapy. In principle, the MPC is a better measure of suitability for use in monodrug therapy with respect to restricting the selection of resistant mutants.
Dual- or Multidrug Combination Therapy Can Prevent Resistance
Conceptually, a combination treatment regimen containing two or more drugs of different classes should require at least two resistance mutations for the pathogen to grow. Because the simultaneous development of two such mutations could be expected only in a bacterial population of much greater size than is normally present within any individual, combination therapy with two distinct antibiotic types provides a way to reduce mutant selection using moderate concentrations of compounds that may individually have very high MPCs.
The best examples of combination therapy are found with tuberculosis. Because the standard agents used for treating patients with this disease cannot be dosed at concentrations that exceed the respective MPCs of these drugs, resistance readily arises when any of the agents is used in monotherapy. Dual-drug and often multidrug therapies thus are used routinely to reduce the number of treatment failures. Currently cases of drug resistance developing among tuberculosis patients are generally associated with the failure of patients to comply fully with therapy regimens. In effect, sporadic compliance creates the equivalent of repeated monotherapy punctuated by periods of bacterial population expansion. To assure patient compliance, a major effort has been placed on directly observed therapy (DOT), which significantly lowers the incidence of resistance-associated treatment failure.
Resistant tuberculosis occasionally emerges even with DOT and combination regimens. Thus, further refinement of this strategy is needed to enhance its effectiveness. One approach involves examining the pharmacokinetic profiles of compounds being used to treat patients with tuberculosis. Those profiles indicate that periods exist when only one antibiotic being administered is above its MIC while the others are below their MICs. For example, consider a clinical trial, reported in 1999 by Andrew Vernon and associates at the Centers for Disease Control and Prevention, Atlanta, Ga. HIV-1-positive pulmonary tuberculosis patients were first treated for two months with a standard four-drug regimen that consisted of isoniazid, rifampin, pyrazinamide, and ethambutol. Then they were divided into two groups for an additional four months of treatment. One group was given isoniazid and rifapentine, a long-acting derivative of rifampin, once a week. In this once-weekly rifapentine program, the total concentration of rifapentine in serum was above the MIC while isoniazid was below the MIC for nearly six days each week (Fig. 3A). Thus, for most of the treatment period, rifapentine acted as a monodrug therapy with its concentration falling in the mutant selection window. Under such circumstances, we would expect rifapentine/rifampin resistance to develop. Indeed, relapses occurred among about 10% of those patients, and 4 of 5 exhibited resistance.
The second group of patients received a twice-weekly isoniazid-rifampin regimen. In this protocol the total rifampin concentration in the serum dropped below the MIC roughly 17 hours after drug administration, and the isoniazid concentration dropped below the MIC about 8 hours later (Fig. 3B). The lack of effective antibiotic for several days after each dose allowed growth of Mycobacterium tuberculosis, and as with the first protocol, relapse occurred in about 10% of the cases. But since rifampin concentration dropped below its MIC before isoniazid concentration, no case of rifampin resistance was expected, and none of the three relapse cases produced rifampin-resistant bacilli.
While the number of cases in the trial was small and uncertainties exist concerning the relevant tissue concentration of the agents, the trial does call attention to what may happen when there are mismatches in antibiotic pharmacokinetics. The solution to the mismatch problem is conceptually straightforward: formulate and administer combination therapies in which pharmacokinetic profiles superimpose. Treatment failure due to other factors, such as lack of patient compliance with therapy regimens, will still occur. However, few new MDR strains should be generated as long as the equivalent of monotherapy is avoided. Such a result would strongly support the two-mutation idea for restricting the selection of resistance.
Implementing the Two-Mutation Strategy Faces Major Challenges
MPC can be measured for various pathogen-antibacterial agent combinations, and suitability for single or combination therapy can be assessed. However, developing additional evidence to convince practicing physicians to implement the two-mutation-based strategy for preserving antibiotic usefulness is a formidable challenge. For example, quantitative relationships between in vitro and in vivo susceptibility to specific drugs need to be established. Moreover, pharmacokinetics need to be measured at the sites of infection to refine estimates obtained from serum determinations and to create good matches for specific combination antibiotic regimens. And ultimately the strategy must reduce the enrichment of resistant mutants in patient populations.
Implementing this or any other strategy intended to halt the emergence of antibiotic resistance will require acceptance and widespread support from the pharmaceutical industry. Moreover, the industry's cooperation is needed to apply the strategy as an important criterion for developing new classes of antimicrobial compounds and for refining existing classes. While such refinements are pursued, incentives may be necessary to ensure that pathogens do not continue to build resistance to older members of such drug classes. One approach would be to shift the use of older compounds from monotherapy to combination therapy.
A simpler but economically problematic strategy is to remove older compounds from use in particular markets. If a specific compound is highly profitable and if its continued use would hasten the selection of resistant mutants that neutralize a newer, more active derivative developed by a competing company, it might be difficult to convince the first company to drop distribution of its established antibiotic.
Challenges also can be expected at the prescriber level because individual health often takes precedence over public health concerns. For example, even with difficult diseases such as tuberculosis, the chance of a successful outcome when combination drug regimens are carefully followed is better than 95%. For most other diseases, the odds are better even with single-drug therapy. Thus employing drug concentrations that fall within the mutant selection window is often of less immediate concern to physicians and their patients than is risking toxic side effects due to the high doses of drugs suggested for MPC-based strategies. But low dosing, which includes use of marginally effective compounds, eventually ruins the clinical utility of any antibacterial agent.
In summary, development of heritable antibiotic resistance within pathogens generally requires two events: resistant mutants must be generated, and then they must be selectively enriched. We cannot prevent the generation of resistant mutants. However, the mutant fraction of a susceptible population is usually small, often on the order of 1 in 106 to 108 cells. If we can keep the total pathogen population size low enough, resistant mutants will be absent for statistical reasons.
Even with large bacterial populations, the fraction of mutant cells is usually small enough to be removed by host defense systems if antimicrobial treatment eliminates wild-type cells. But repeated cycles of selective antibiotic pressure, coupled with population expansion, gradually erode the usefulness of an antibiotic. The repeated cycles can occur within a single host, as when tuberculosis patients take antibiotics sporadically, or the cycles can be spread over many hosts, as seen with streptococcal pneumonia (when many hosts are involved, population expansion occurs in each fresh host prior to administration of antibiotic). We can interfere with mutant selection by dosing so that tissue concentrations are above the mutant selection window, by using combination therapies with superimposed pharmacokinetics, and by blocking population expansion with sound infection control measures. A combination of these practices may allow us to delay the postantibiotic era.
SUGGESTED READING
Blondeau, J., X. Zhao, G. Hansen, and K. Drlica. 2001. Mutant prevention concentration (MPC) for fluoroquinolones with clinical isolates of Streptococcus pneumoniae. Antimicrob. Agents Chemother., in press.
Dong, Y., X. Zhao, J. Domagala, and K. Drlica. 1999. Effect of fluoroquinolone concentration on selection of resistant mutants of Mycobacterium bovis BCG and Staphylococcus aureus. Antimicrob. Agents Chemother. 43:1756-1758.
Dong, Y., X. Zhao, B. Kreiswirth, and K. Drlica. 2000. Mutant prevention concentration as a measure of antibiotic potency: studies with clinical isolates of Mycobacterium tuberculosis. Antimicrob. Agents Chemother. 44:2581-2584.
Iseman, M., and J. Sbarbaro. 1991. Short-course chemotherapy of tuberculosis. Am. Rev. Resp. Dis. 143:697-698.
Levin, B., V. Perrot, and N. Walker. 2000. Compensatory mutations, antibiotic resistance, and the population genetics of adaptive evolution in bacteria. Genetics 154:985-997.
Sindelar, G., X. Zhao, A. Liew, Y. Dong, J. Zhou, J. Domagala, and K. Drlica. 2000. Mutant prevention concentration (MPC) as a measure of fluoroquinolone potency against mycobacteria. Antimicrob. Agents Chemother. 44:3337-3343.
Vernon, A., W. Burman, D. Benator, A. Khan, and L. Bozeman. 1999. Acquired rifamycin monoresistance in patients with HIV-related tuberculosis treated with once-weekly rifapentine and isoniazid. Lancet 353:1843-1847.
Zhao, X., and K. Drlica. 2001. Restricting the selection of antibiotic-resistant mutants: a general strategy derived from fluoroquinolone studies. Clin. Infect. Dis., in press.
Zhao, X., C. Xu, J. Domagala, and K. Drlica. 1997. DNA topoisomerase targets of the fluoroquinolones: a strategy for avoiding bacterial resistance. Proc. Natl. Acad. Sci. USA 94:13991-13996.
Zhou, J.-F., Y. Dong, X. Zhao, S. Lee, A. Amin, S. Ramaswamy, J. Domagala, J. M. Musser, and K. Drlica. 2000. Selection of antibiotic resistance: allelic diversity among fluoroquinolone-resistant mutations. J. Infect. Dis. 182:517-525.
Email: webmaster@asmusa.org