Emergence and Costs of Drug Resistance
For other pathogens, extended antibiotic chemotherapy has been implicated in the evolution of multiple-antibiotic resistance (
143), an association likely to be replicated during
M. tuberculosis infections. The long duration of the DOTS regimen, together with the toxic side effects of the frontline drugs and the temptation to cease therapy as symptoms subside, often leads to patient noncompliance (
161). Furthermore, even where treatment schedules are adhered to, limited efficacy of one or more drugs can compromise combination therapy (
142); recent epidemiological evidence suggests that monotherapy, effective or actual, is common (
251).
M. tuberculosis has a long generation time and can adapt to prevailing growth conditions through a regulated shift to an alternative metabolic state (
60,
233,
241). These factors are considered key to the ability of the organism to establish latent asymptomatic infection, but they might also result in the selective activity of specific antibiotics against discrete subpopulations (Global Alliance for TB Drug Development [
http://www.tballiance.org ]). In addition, there is compelling evidence that infecting populations occupy diverse microenvironments within the host (
34,
35,
83,
120,
122), some of which might be recalcitrant to antibiotic penetration or refractory to activity (
78,
115). Significantly, the locally effective antibiotic concentration has been implicated in the evolution of low-level resistant variants during infection with other pathogens (
11) and has been identified as an important determinant of mutation rate. Furthermore, different mutation rates and genotypes are thought to arise at intervals along a spectrum of applied concentrations (
128,
260), a possibility with profound implications for the generation of diverse microbial populations in a single host (
10). The parallel evolution of a single founder population into heterogeneous, antibiotic-resistant subpopulations within isolated loci has been demonstrated in patients undergoing active treatment for TB (
122), for example.
The attachment of a fitness cost to resistance mutations (
6) has led to the assumption that removal of antibiotic selective pressure will favor reversion as a result of a competitive replicative disadvantage. However, there is evidence that evolution in the absence of the selective antibiotic preferentially results in the acquisition of compensatory mutations that ameliorate the cost of resistance (
5,
135,
136), rather than reversion. That is, the fitness cost more likely determines the stability and potential reversibility of the associated resistance mutation, with the ability to compensate genetically dictating the frequency of resistant mutants within a population. While the most fit mutants will be selected in a large population, lower-fitness, compensated mutants might become fixed during bottlenecks if they are formed at a higher rate than fitter, susceptible revertants (
136,
148), particularly where genetic linkage exists between selected and nonselected resistance markers (
80).
Information on the relative fitness of MDR TB isolates is limited (
65); however, there is evidence that compensatory mutations can restore reproductive potential in monoresistant
M. tuberculosis strains (
210,
215). In addition, while drug-resistant
M. tuberculosis strains more frequently possess low- rather than high-cost mutations (
202), studies investigating the effects of resistance on virulence (
16,
159,
181,
192,
202) have failed to establish a direct correlation (
50). Instead, relative fitness in vitro appears to depend not only on the particular resistance mutation but also on the specific assay (
151). Of course, there is the possibility that a resistance mutation might affect the ability of the pathogen to interact with the host environment and so might remain undetected in vitro (
5). Mutations impairing virulence, for example, such as deletion of
katG (
140), will not survive selection in areas of high transmission (
51).
The effects of resistance mutations on the fitness of
M. tuberculosis are crucial to epidemiological predictions of the spread of MDR isolates (
50). This concept has been further refined by recent evidence from mathematical models which suggests that, provided the relative fitness of an MDR strain remains above a defined threshold, a subpopulation of the low-fitness MDR strain will outcompete both the drug-sensitive strains and other, less fit MDR strains when confronted by a functioning TB control program (
19,
49). As a result, the distribution of fitness (
49) among circulating
M. tuberculosis strains might be considered a more accurate predictive measure of resistance emergence. This, in turn, has led to the proposal that DOTS regimens be supplemented with anti-MDR strategies to limit resistance amplification, as well as further transmission of MDR strains (
19,
49).
Coincident MDR TB prevalences and HIV infection rates (
252) add a further degree of complexity and are suggestive of a positive correlation between MDR and HIV seropositivity. Although the identification of HIV as an independent risk factor for MDR TB is contentious (
251), characteristic features of HIV/TB-associated clinical disease might favor the emergence of resistance. It has been suggested, for example, that disease outcomes in the treated TB patient are determined by a combination of host defense mechanisms and antimicrobial activity (
98). That is, a functional immune system might be required to potentiate drug activity, thus preventing the evolution of resistance. Although direct evidence is scant, the enhanced sterilizing activity of PZA in vivo, in contrast with its poor activity in vitro (
257), is suggestive of a synergistic interplay between drug and host (
236).
The large bacterial populations associated with
M. tuberculosis-infected immunocompromised individuals might provide an expanded subset for selection and transmission of rare mutation events. Furthermore, it has been suggested that the absence of a functioning immune response in those individuals might exacerbate the conditions implicated in the exposure of bacteria to monotherapy in immunocompetent patients (
93); for example, uncontrolled replication and dissemination could produce drug-inaccessible compartments, while drug absorption might be compromised by other HIV-associated chronic infections. It has also been proposed that drug-resistant strains of reduced fitness might undergo compensatory adaptation during passage through a population of immunocompromised individuals, ultimately restoring their capacity to infect immunocompetent hosts (
93). In general, it seems likely that increased TB incidence rates associated with high HIV prevalence will facilitate the spread of both susceptible and MDR strains (
55); while slower to emerge in immunocompetent individuals, MDR strains could result in huge burdens of disease in the future (
93). However, there is evidence to suggest that the impact of HIV on TB transmission (and therefore prevalence) is more complex and might depend on factors such as the duration of infectious period and the presence of a functioning TB control program (
56,
57).
Mutation Rates and the Role of Mutators
Based on in vitro measures of rates of mutation to single drug resistance in
M. tuberculosis (
64), the emergence of MDR TB appears to require a larger bacillary population than is usually present during infection. However, the assumption that the risk of acquiring multiple resistance equals the product of individual mutation rates is likely an oversimplification, considering the complex interplay of factors that might operate to increase mutation rates in vivo. Small subpopulations of mutators characterize commensal and pathogenic bacterial populations in vivo (
18,
67,
154,
180,
221,
232), consistent with the idea that elevated mutation rates may promote adaptation to the fluctuating host environment. However, the deleterious consequences of a constitutive mutator phenotype (
20) ensure that maintenance of the mutator allele depends on genetic linkage to the resultant beneficial mutation (
20,
232). In general, the selective advantage of a high mutation rate is transient, and regaining the wild-type genotype is essential to the long-term survival of the population (
94,
95).
Stable, acquisitive evolution is thought to depend on minimal disturbances to established bacterial pathways and host-pathogen interactions (
253), a concept consistent with the suggested adaptation of separate
M. tuberculosis lineages to particular host populations (
111). Instead, the long-term selection or counterselection of small-effect mutators is thought likely to exert greater influence on bacterial evolution (
221), perhaps explaining the failure to observe a mutator phenotype in
M. tuberculosis, despite its natural deficiency in several pathways associated with hypermutability in other organisms (
162). A possible exception is provided by the W-Beijing genotype, which is most frequently associated with emergence of MDR TB (
96). A high proportion of W-Beijing isolates contain mutations in genes required for elimination of damaged nucleotides (
mutT) and reversal of alkylation damage to DNA (
ogt) (
193). In vitro assays of mutation rates have so far failed to demonstrate increased mutagenesis in W-Beijing isolates (
246), although it is possible that the strains tested did not carry the characteristic “mutator” mutations. Furthermore, the prevalence of the W-Beijing lineage among MDR strains makes it tempting to speculate that the mutator phenotype might be manifest only under (stressful) in vivo conditions.
Inducible (environment-dependent) mutators, in contrast, increase global mutation rates specifically in response to applied stress (
158). Those cells that survive produce progeny cells with normal mutation rates, thereby reducing the risk of unchecked mutagenesis. Whereas the acquisition of a mutator phenotype is a random event, it has been proposed that inducible mutagenesis is an adaptive response that has evolved by second-order selection to modulate mutation rates while limiting the costs associated with a constitutive mutator phenotype (
158,
225,
226). That inducible mutator mechanisms are subject to selection has been inferred from the negative correlation between stress-induced mutagenesis and constitutive mutators (
17). This idea is further reinforced by the observation that where stresses are frequent or of long duration, inducible mutators are selected as efficiently as mutator alleles (
17). The variation in the strength, frequency, and nature of inducible mutagenesis mechanisms is thought to be reflective of the dynamic response of different pathogens to specific local environments (
225).
The role of stationary-phase or stress-induced mutagenesis in bacterial adaptation has been subject to considerable recent attention (
225). In particular, an association between inducible mutation pathways and the emergence of drug-resistant isolates of pathogenic bacteria has been described (
4,
18,
189,
197), which might be especially relevant to the generation of antibiotic and stress resistance mutations in
M. tuberculosis, whose microevolution within the host environment (as stated above) is driven by genetic rearrangement and point mutations (
105,
195). In most bacterial systems studied to date, adaptation to environmental stress is predicated on the activity of SOS-inducible, error-prone repair polymerases of the Y polymerase superfamily (
157,
220,
224,
255). Members of the Y family of DNA polymerases likely evolved to promote mutation avoidance and damage tolerance through a specialized ability to replicate across a variety of DNA lesions; however, the flip side of this ability is that the very properties enabling translesion synthesis are implicated in mutagenesis (
88). The
M. tuberculosis genome encodes two putative Y family polymerases of the DinB subclass (
178), but, unusually, neither is upregulated in response to DNA damage (
23,
66). Instead, their predicted physiological roles are fulfilled in
M. tuberculosis by a novel, damage-inducible C family polymerase, DnaE2, which is solely responsible for damage-induced base substitution mutagenesis (
23). Significantly, deletion of
dnaE2 results in damage hypersensitivity and eliminates damage-induced base mutagenesis in vitro and is associated with late-stage attenuation as well as reduced emergence of drug resistance mutations in a murine infection model (
23).
Coupled with the induction of
dnaE2 during stationary-phase infection in mice (
23), these observations suggest that genetically encoded antibiotic resistance mutations may arise as the result of DnaE2-mediated repair synthesis during persistent infection. According to this hypothesis, a range of host immune effectors and other environmental damaging agents, as well as endogenous oxidative and nitrosative metabolic stresses or antibiotics, might induce damage lesions. Stalled replication at a lesion induces
dnaE2 expression either as part of the mycobacterial SOS response or by unknown regulatory mechanisms analogous to novobiocin-mediated
dnaE2 induction (
25). Error-prone repair synthesis by DnaE2 might fix mutations in chromosomal DNA at the site of the damage, in some cases conferring antibiotic resistance which is then selected.