Introduction. Summary. Dmitri Shcherbakov, 1 Rashid Akbergenov, 1 Tanja Matt, 1 Peter Sander, 1,2 Dan I. Andersson 3 and Erik C.

Transcription

1 Molecular Microbiology (2010) 77(4), doi: /j x First published online 7 July 2010 Directed mutagenesis of Mycobacterium smegmatis 16S rrna to reconstruct the in vivo evolution of aminoglycoside resistance in Mycobacterium tuberculosismmi_ Dmitri Shcherbakov, 1 Rashid Akbergenov, 1 Tanja Matt, 1 Peter Sander, 1,2 Dan I. Andersson 3 and Erik C. Böttger 1,2 * 1 Institut für Medizinische Mikrobiologie, Universität Zürich, Zürich, Schweiz. 2 Nationales Zentrum für Mykobakterien, Universität Zürich, Zürich, Schweiz. 3 Department of Medical Biochemistry and Microbiology, Uppsala University, Uppsala, Sweden. Summary Drug resistance in Mycobacterium tuberculosis is a global problem, with major consequences for treatment and public health systems. As the emergence and spread of drug-resistant tuberculosis epidemics is largely influenced by the impact of the resistance mechanism on bacterial fitness, we wished to investigate whether compensatory evolution occurs in drug-resistant clinical isolates of M. tuberculosis. By combining information from molecular epidemiology studies of drug-resistant clinical M. tuberculosis isolates with genetic reconstructions and measurements of aminoglycoside susceptibility and fitness in Mycobacterium smegmatis, we have reconstructed a plausible pathway for how aminoglycoside resistance develops in clinical isolates of M. tuberculosis. Thus, we show by reconstruction experiments that base changes in the highly conserved A-site of 16S rrna that: (i) cause aminoglycoside resistance, (ii) confer a high fitness cost and (iii) destabilize a stem-loop structure, are associated with a particular compensatory point mutation that restores rrna secondary structure and bacterial fitness, while maintaining to a large extent the drug-resistant phenotype. The same types of resistance and associated mutations can be found in M. tuberculosis in clinical isolates, suggesting that compensatory evolution contributes to the spread of drug-resistant tuberculosis disease. Accepted 11 May, *For correspondence. boettger@imm. uzh.ch; Tel. (+41) ; Fax (+41) Equal contribution. Introduction In epidemiological terms, the single most important biological parameter that influences the spread of resistance is the fitness cost of drug resistance (Andersson, 2006). Depending on the genetic resistance determinant, drug resistance may frequently carry a biological cost, e.g. reduced growth rate in the absence of the drug. Based on laboratory in vitro studies it has been suggested that secondary-site mutations may reduce the biological fitness cost associated with a resistance determinant (Schrag et al., 1997; Reynolds, 2000; Maisnier-Patin et al., 2002; Maisnier-Patin and Andersson, 2004). However, the extent to which this type of compensatory evolution occurs in clinical isolates in vivo is still unclear. Drug resistance in Mycobacterium tuberculosis is an increasing and threatening public health problem (Donald and van Helden, 2009). Mathematical models have been used to estimate the epidemiological fitness cost of drug resistance in M. tuberculosis, pointing to the need for accurate experimental estimates of parameters such as fitness costs of resistance and in particular of whether compensatory mutations occur, as this would have significant epidemiological consequences (Luciani et al., 2009). Drug resistance in M. tuberculosis is unique as it is exclusively due to chromosomal mutational alterations affecting mainly either the drug target directly or enzymes involved in drug activation (Böttger and Springer, 2009). Previous studies on mycobacterial drug resistance have allowed some important conclusions with respect to the biological fitness of chromosomal drug resistance (Böttger and Springer, 2008): (i) drug resistanceconferring mutations can have a variable impact on bacterial fitness, which varies according to the specific resistance-conferring mutation (Böttger et al., 1998; Billington et al., 1999; Sander et al., 2002), (ii) low-cost resistance mutations are common and a significant selection exists in vivo for the corresponding resistance mutations, as testified by the observation of a significant inverse correlation between the frequency of particular resistance mutations in clinical strains and the fitness cost of the corresponding mutations as determined in in vitro fitness assays (Böttger et al., 1998; Billington et al., 1999; 2010 Blackwell Publishing Ltd

2 Drug resistance and fitness in M. tuberculosis 831 van Soolingen et al., 2000; Pym et al., 2002; Sander et al., 2002; Mariam et al., 2004). These studies have also established Mycobacterium smegmatis, a non-pathogenic mycobacterium amenable to genetic manipulations (Sander et al., 1996), as a suitable model system to study fitness-related aspects of resistance in M. tuberculosis, particularly for antibacterial agents affecting the ribosome. Early on it was hypothesized that compensatory mechanisms may be involved in the evolution of drug resistance in M. tuberculosis (Sherman et al., 1996), but firm genetic evidence for compensatory evolution playing a role in mycobacterial drug resistance in vivo is still lacking. In the past years we have studied resistance mechanisms to and drug target interactions of antibiotics that act by inhibiting protein synthesis, in particular 2-deoxystreptamine aminoglycosides (Böttger et al., 2001; Pfister et al., 2003; 2005; Hobbie et al., 2005; 2006a,b) compounds, which bind to the ribosome s small subunit decoding A-site and result in translation inhibition (Chambers and Sande, 1996). As the A-site is highly conserved in both sequence and structure between M. smegmatis and M. tuberculosis, we used genetic reconstructions in M. smegmatis to analyse aminoglycoside resistance mutations reported for clinical strains of M. tuberculosis (Sander et al., 1996; Alangaden et al., 1998; Suzuki et al., 1998; Krüüner et al., 2003; Maus et al., 2005; Feuerriegel et al., 2009; Jugheli et al., 2009). Based on the present results we can now provide plausible explanations for why certain resistance mutations have been selected in clinical strains of M. tuberculosis and others not. Results and discussion Ribosomal resistance because of point mutations in the decoding A-site of the small subunit has been recognized as a main cause of clinically acquired or laboratory generated resistance to kanamycin and amikacin in M. tuberculosis and other mycobacteria (Sander et al., 1996; Alangaden et al., 1998; Prammananan et al., 1998; 1999; Suzuki et al., 1998; Krüüner et al., 2003; Maus et al., 2005; Feuerriegel et al., 2009; Jugheli et al., 2009). Lowlevel aminoglycoside resistance in M. tuberculosis may be found in the absence of mutational target alterations (Meier et al., 1996; Alangaden et al., 1998; Suzuki et al., 1998; Krüüner et al., 2003; Böttger and Springer, 2009; Zaunbrecher et al., 2009). The A-site, the drug binding pocket, is composed of nucleotides of the 16S rrna (Fig. 1), and its sequence is conserved throughout the mycobacteria. To avoid the influence of an unknown or poorly defined genetic background, like those occurring in clinical isolates, we employed a well-defined model to reconstruct the evolution of drug resistance in clinical strains of M. tuberculosis. This model is based on a single rrna allelic strain of M. smegmatis, for which we have established a set of experimental techniques for genetic manipulation, and which as previously shown provides a relevant and adequate model to study the genetics and physiological effects of resistance mutations in M. tuberculosis (Sander et al., 1996; 2002; Böttger et al., 1998). In clinical aminoglycoside-resistant strains of M. tuberculosis the following alterations have been observed in 16S rrna: single-base changes A1408G, C1409U, G1491U and the double-base change C1409A/G1491U (Table 1). There is a clear bias in their frequencies among drug-resistant clinical isolates where the A1408G alteration is by far the most frequent (160/169). The C1409U and G1491U base changes are rare (6/169 and 1/169 respectively). Of particular importance is the observation that the C1409A/G1491U alteration is at least as frequent (2/169) as the single G1491U alteration. This is unexpected, as the statistical probability for two mutations occurring is much lower than that of a single mutational alteration, i.e. the stochastic probability for two mutations is the product of the probabilities for each single mutation. Noteworthy, in the rrna secondary structure residues 1409 and 1491 interact with each other by base pairing (Fig. 1). Each alteration, C1409U or G1491U, will disrupt this base-pair interaction, while base pairing is restored in the C1409A/G1491U double mutant. Mycobacterium smegmatis DrrnB was used to generate a complete set of isogenic mutants, which sample the various possible mutational alterations affecting rrna residues A1408, C1409, G1491 in a systematic manner (see Experimental procedures for details). Bacterial fitness was investigated by measuring bacterial generation times during growth in broth culture and in standard competition experiments to determine the mutationmediated cost per generation (cpg, defined as % difference in growth rate during competition). For both parameters of fitness, i.e. generation time in pure culture and cpg in competition culture, we observed an excellent correlation (Table 2). Drug susceptibility was determined by minimal inhibitory concentration (MIC) assays (Table 3). In general and compared with kanamycin, susceptibility to amikacin is less affected by alterations of rrna residues 1409 or Each mutational genotype was associated with a distinct phenotype in terms of both resistance level and fitness cost. (i) The A1408G base change conferred high-level drug resistance (RR, relative resistance to kanamycin > 1000) and carried a little, yet significant, fitness cost (cpg ). (ii) The C1409U alteration conferred low- to intermediate-level resistance (RR = 16) and was associated with a cpg of restoration of 1409/1491 base pairing by a G1491A alteration restored drug susceptibility and in part fitness (cpg ). (iii) The C1409G alteration conferred low- to intermediate-level resistance (RR = 32) and was associated with a very high

3 832 D. Shcherbakov et al. A C - G 1405 G - C U U C G A A A 1409 C - G G - C U - A C - G B Kanamycin A C D Fig. 1. Secondary and three dimensional structure of the bacterial decoding A-site. A. Secondary structure of the conserved bacterial A-site in the small subunit rrna subjected to site-directed mutagenesis (16S rrna numbering according to the Escherichia coli nomenclature); note, that the nucleic acid sequence of M. tuberculosis and M. smegmatis is identical in the A-site. B. Kanamycin A structure. C. Crystal structure of tobramycin complexed to a model oligonucleotide containing the decoding A-site (Hobbie et al., 2006b). D. Description of contacts between tobramycin and specific rrna nucleotides. Ring numbers (I III) and atom names are specified; W stands for water molecule; hydrogen bonds are shown as dashed lines. Tobramycin differs from kanamycin A in two substituents (kanamycin A 2 OH, 3 OH; tobramycin 2 NH, 3 H), which, however, do not affect the overall structure of the drug target complex (François et al., 2005). cpg of restoration of 1409/1491 base pairing by a G1491C base change largely restored drug susceptibility and partly improved biological fitness (cpg ). (iv) The G1491U alteration showed intermediate resistance (RR = 64) and a cpg of restoration of 1409/1491 base pairing by a C1409A base change little affected drug resistance but restored fitness to wild-type level (cpg ). (v) The G1491C alteration was associated with low- to intermediate-level drug resistance (RR = 32) and an excessive cpg of restoration of 1409/1491 base pairing by C1409G largely restored drug susceptibility and partly augmented fitness (cpg ). (vi) The G1491A base change conferred low-level drug resistance (RR = 2) and was associated with a cpg of restoration of 1409/ 1491 base pairing by C1409U restored drug susceptibility (RR = 1 2) and significantly increased biological fitness (cpg ). (vii) We have been unable to construct mutants with the single nucleotide alterations C1409A (which would result in an A1409 G1491 interaction), A1408U or A1408C, using RecA-driven homologous recombination and selective plating (Prammananan et al., 1999) or rrna operon plasmid exchange (Hobbie et al., 2007); for recombinant plasmids used see Table 4. This finding most likely reflects lethality of the corresponding sequence alterations, an interpretation supported by the

4 Drug resistance and fitness in M. tuberculosis 833 Table 1. Frequency of resistance mutations in aminoglycoside-resistant M. tuberculosis strains. Number of clinical strains Total (n = 169) 1408A G 26 a 8 b 13 c 11 d 3 e 34 f 65 g C U 1 a 1 f 4 g G U 1 d C A/1491G U 2 a 2 a. Suzuki et al. (1998). b. Sander et al. (1996). c. Alangaden et al. (1998). d. Maus et al. (2005). e. Krüüner et al. (2003). f. Feuerriegel et al. (2009). g. Jugheli et al. (2009). Clinical drug-resistant strains from Suzuki et al. (1998) were isolated from patients receiving kanamycin; clinical drug-resistant strains from Sander et al. (1996), Alangaden et al. (1998), Krüüner et al. (2003) and Maus et al. (2005) were isolated from patients receiving kanamycin or amikacin. Information on the patients drug regimen prior to isolation of the mutant clinical strains is not provided by Feuerriegel et al. (2009) and Jugheli et al. (2009). Table 2. M. smegmatis strains constructed: generation times and cost per generation. Strain Generation time a (hours standard deviation) Cost per generation b (% standard deviation) DrrnB (SZ380) c DrrnB rrs wt (SZ848) (SZ ) c DrrnB rrs 1408G (SZ461) (SZ ) DrrnB rrs 1491A (SZ463) (SZ ) DrrnB rrs 1491C (SZ468) (SZ ) DrrnB rrs 1491U (SZ505) (SZ ) DrrnB rrs 1409G (SZ508) (SZ508, 509) DrrnB rrs 1409U (SZ521) (SZ520, 521) DrrnB rrs 1409G/1491C (SZ605) (SZ605) DrrnB DrrnA attb: rrnb (SZ637) DrrnB DrrnA rrs wt (SZ768) (SZ ) DrrnB DrrnA rrs 1409U/1491A (SZ717) (SZ717, 719, 720) DrrnB DrrnA rrs 1409A/1491U (SZ763) (SZ763, 765, 766) a. Determined by OD readings. b. Determined by competition growth experiments. c. Recombinant strains chosen for analysis. Table 3. M. smegmatis strains constructed: drug susceptibility. MIC mg l -1 Strain Gentamicin Kanamycin A Amikacin DrrnB (SZ380) a DrrnB rrs wt (SZ ) DrrnB rrs 1408G (SZ ) > 1024 > 1024 > 1024 DrrnB rrs 1491A (SZ ) DrrnB rrs 1491C (SZ ) DrrnB rrs 1491U (SZ ) DrrnB rrs 1409G (SZ508, 509) DrrnB rrs 1409U (SZ520, 521) DrrnB rrs 1409G/1491C (SZ605) DrrnB DrrnA rrs wt (SZ ) DrrnB DrrnA rrs 1409U/1491A (SZ717, 719, 720) DrrnB DrrnA rrs 1409A/1491U (SZ763, 765, 766) a. Strains chosen for analysis.

5 834 D. Shcherbakov et al. Table 4. Plasmids used in this study. Plasmid Genotype rrna References pgem-teasy Promega pmv361 hyg Pfister et al. (2003) pz400 pmv361 hyg-rrs wt rrs wt, hyg Partial This study ph128 pmv361 hyg-1408g rrs A1408G, hyg Partial This study pz176 pmv361 hyg-1491a rrs G1491A, hyg Partial Pfister et al. (2005) pz178 pmv361 hyg-1491c rrs G1491C, hyg Partial Pfister et al. (2005) pz177 pmv361 hyg-1491u rrs G1491U, hyg Partial Pfister et al. (2005) pz191 pmv361 hyg-1409g rrs C1409G, hyg Partial Pfister et al. (2005) pz175 pmv361 hyg-1409u rrs C1409U, hyg Partial Pfister et al. (2005) pz179 pmv361 hyg-1409a rrs C1409A, hyg Partial This study ph150 pmih hyg rrnb wt rrs wt, hyg Complete rrna operon replacement Hobbie et al. (2007) ph297 pmih hyg rrnb 1409U 1491A rrs C1409U G1491A, hyg Complete rrna operon replacement This study ph154 pmih hyg rrnb 1409A 1491U rrs C1409A G1491U, hyg Complete rrna operon replacement This study ph151 pmih hyg rrnb 1408U rrs A1408U, hyg Complete rrna operon replacement This study ph152 pmih hyg rrnb 1408C rrs A1408C, hyg Complete rrna operon replacement This study ph153 pmih hyg rrnb 1409A rrs C1409A, hyg Complete rrna operon replacement This study absence of these sequences from prokaryotic or eukaryotic small subunit rrna To understand the development of aminoglycoside resistance in clinical M. tuberculosis strains, we made use of the above information where the impact on drug susceptibility and fitness cost was determined in M. smegmatis. By combining this information with the mutational alterations found in M. tuberculosis, we can suggest a plausible evolutionary path for resistance development in clinical M. tuberculosis strains that accounts both for the types of mutations observed and their clinical prevalence (Fig. 2). The single mutational alteration A1408G is by far the most frequent resistance mutation in clinical strains, most likely because this mutation confers highlevel drug resistance and confers the lowest fitness cost. The single, but rarely observed base changes C1409U and G1491U each confer an intermediate-level of resistance associated with a significant fitness cost (cpg for both mutations approximately 10%) however, these mutations impact on biological fitness considerably less than resistance mutations C1409G or G1491C (cpg for both mutations > 50%), which have not been observed in clinical M. tuberculosis strains. Introduction of C1409A into the G1491U mutant restored 1409/1491 base pairing and improved fitness to wild-type level, but is still associated with significant resistance to kanamycin. Of note, both of the C1409A/G1491U clinical strains were isolated from patients reportedly undergoing therapy with kanamycin (Suzuki et al., 1998). Thus, the C1409A alteration represents a bona fide compensatory mutation, which largely ameliorates the fitness cost of the primary resistance alteration G1491U but only marginally affects the kanamycin resistance level associated with this particular mutation. Interestingly, the C1409A/G1491U mutant has regained susceptibility to amikacin. Certain genetic alterations or compensatory mutations are not observed in clinical strains, as they are either: (i) lethal (C1409A, A1408U, A1408C), (ii) highly costly (C1409G, G1491C), or (iii) confer little (G1491A, C1409G/G1491C) or no (C1409U/G1491A) resistance; thus, no efficient selection operates for these genotypes. In contrast to the situation with the G1491U mutant, no compensatory mechanism seems to exist for the clinically observed C1409U mutant. Here, the corresponding sequence alteration restoring base pairing would be G1491A. However, a C1409U/ G1491A interaction shows wild-type drug susceptibility. Consequently, no selection pressure exists for this sequence alteration, readily explaining its absence as a compensatory mechanism in clinical isolates. In silico analyses of bacterial 16S rrna sequences indicate that is mostly, if not always a Watson- Crick base-pair interaction and that the dominant natural sequence variations of the interaction in bacteria are C1409 G1491 and A1409 U1491, but rarely, U1409 A1491 or if ever G1409 C1491 (data not shown). This finding is compatible with our experimental results that demonstrate that C G or A U interactions are biologically most fit, compared with U A or G C interactions, which are associated with a significant fitness deficit in our genetic reconstructions of the A-site (cpgs of 4.5 and 31.9 respectively). In this study, we have not determined the mutation rates of the different resistance mutations. However, we have shown previously for streptomycin that the different resistance mutations occur with a similar probability, suggesting that the underrepresentation of costly resistance mutations in clinical isolates is mediated by selective pressure operating under in vivo conditions (Böttger et al., 1998; Sander et al., 2002). For a resistance mutation that lowers fitness at least three possibilities exist for its future fate: (i) the mutation goes extinct, (ii) the mutation exists at a low frequency in the population and (iii) the mutation

6 Drug resistance and fitness in M. tuberculosis C lethal 1408U lethal wt A1408, C1409, G A lethal C 1491A 1491U 1409U 1409G 2 RR: 32 cpg: 67.1±5.7 3 RR: 2 cpg: 45.5±3.6 RR: 64 cpg: 11.7±5.9 RR: 16 cpg: 9.3±2.6 RR: 32 cpg: 50.4±1.3 rare clinical mutation rare clinical mutation 1408G 1491U/1409A RR: >1024 cpg: 5.0± C/1409G RR: 2-4 cpg: 31.9± A/1409U RR: 1-2 cpg: 4.5±3.1 RR: 16 cpg: 0.1± U/1491A RR: 1-2 cpg: 4.5± G/1491C RR: 2-4 cpg: 31.9±1.9 clinically the most prevalent mutation clinically observed rare compensatory mutation Fig. 2. Mutation development and compensatory evolution. 1 in green: genotypes observed in clinical drug-resistant M. tuberculosis strains; in red: genotypes not found in clinical strains; in blue: lethal genotypes. 2 RR (relative resistance): MIC kanamycin mutant divided by MIC kanamycin wild type; green observed in clinical strains, black not observed in clinical strains, red resistance too low (RR < 4) to provide a selective advantage and thus not found in clinical strains. 3 cpg (cost per generation): green observed in clinical strains, black not observed in clinical strains, red too costly (> 15% cpg) to be observed in clinical strains. is stabilized in the population by means of compensatory evolution. Laboratory studies have clearly demonstrated that the latter process occurs in a number of bacterial species in vitro (Maisnier-Patin and Andersson, 2004). A priori, compensatory evolution is unlikely to be important in maintaining drug resistance, as long as resistance mutations exist that confer very small or no fitness cost. Thus, available experimental evidence as well as theory suggests that for a spectrum of possible mutational resistance alterations each being associated with a distinct fitness cost, selection for those resistance mutations with the lowest cost would occur in vivo (Böttger et al., 1998; Billington et al., 1999; van Soolingen et al., 2000; Pym et al., 2002; Sander et al., 2002; Mariam et al., 2004; see Table 1). While bacterial reproduction has been postulated to be constrained by different environmental factors during in vitro and in vivo growth (Björkman et al., 2000), the dominance in clinical M. tuberculosis strains in vivo of a mutant genotype determined in vitro in the model M. smegmatis to be a low-cost mutation suggests that fitness determined in vitro adequately reflects growth in vivo for M. tuberculosis. While one cannot exclude a hypothetical condition in which a difference in fitness would become manifest, the finding of a low-cost resistance mutation as the dominant mutation in a range of heterogeneous clinical isolates, exposed to complex and fluctuating conditions and habitats, makes this possibility unlikely. The most prevalent low-cost resistance mutation A1408G is also the one that shows the highest drug resistance. Most likely, in vivo selection of a drug resistance mutation is a result of both its fitness cost associated and its resistance level conferred. Our model conditions measuring fitness in a laboratory medium are different from the physiological conditions as present in an infected human host. We also note some limitations, i.e. we can not formally exclude that certain resistance mutations, such as C1409G, became counter-selected in the diagnostic laboratory because of culture-based recovery of M. tuberculosis from clinical specimens, or that there was a positive

7 836 D. Shcherbakov et al. selection for the C1409A secondary-site compensatory mutation in the G1491U mutant under the laboratory conditions used for strain recovery and passage on antibiotic free media. Bacterial populations, even at clonal level, often exhibit a significant degree of genetic polymorphism. The possible impact of strain genetic background on the fitness cost of a defined resistance mutation is a matter of debate and most likely organism-specific. Recent studies suggest that in Enterobacteriaceae the same resistance mutation can have different fitness effects in different genetic backgrounds (Luo et al., 2005; Paulander et al., 2009). However, whether this also applies, as has been suggested (Gagneux et al., 2006), to M. tuberculosis is less clear. Instead for M. tuberculosis we suggest that, based on the finding of a significant correlation between the frequency of a resistance mutation as observed in genetically heterogeneous clinical strains and the fitness cost as estimated by generation time measurements of genetically constructed isogenic mutants in vitro (Tables 1 and 2), the strain genetic background in general plays a minor role in resistance-related fitness costs. An explanation for this observation could be that the high genetic similarity of various M. tuberculosis strains (Smith et al., 2009) reduces the potential for confounding epistatic mutational interactions, and thus the fitness of a resistant mutant is directly determined by the specific resistance mutation and less influenced by interactions with other mutations. Costly resistance mutations may become dominant in a bacterial population under various circumstances, for example: (i) under conditions of population bottlenecks during transmission and/or growth (Levin et al., 2000), (ii) for drugs for which seemingly cost-neutral resistance mutations do not exist (Nagaev et al., 2001; Nilsson et al., 2003) or (iii) if the rate of formation of high-cost and compensatory mutations is much higher than that for a low-cost mutation. Under these conditions compensatory evolution becomes relevant and is expected to impact the spread of resistant pathogens. Based on genetic reconstructions in M. smegmatis our data are suggestive of compensatory evolution in clinical aminoglycosideresistant strains of M. tuberculosis, as testified by the studies on the C1409A/G1491U mutation. Experimental procedures Bacterial strains and DNA techniques Mycobacterium smegmatis strains mc Sm S DrrnB and mc Sm S DrrnB DrrnA attb : rrnb were used for all experiments. These strains are variants of M. smegmatis with a single rrna operon in a chromosomal background devoid of any antibiotic resistance marker (strain construction to be described in detail elsewhere). Strategies used to generate the various isogenic mutants under study included: (i) transformation with plasmids carrying the rrna alteration of interest and subsequent integration into the chromosomal rrna operon by means of homologous recombination and (ii) gene replacement by plasmid exchange. For introduction of point mutations by RecA-mediated homologous recombination (Böttger et al., 2001; Pfister et al., 2005), approximately 1.0 kb large fragments of the rrn decoding region (16S rrna position 907 to ITS1 position 2367) were generated and mutations introduced by PCR mutagenesis as described previously (Pfister et al., 2005). The mutagenized rrn fragments were subcloned into pgem- Teasy vector (Promega), isolated by NcoI/SpeI digestion and ligated into the HpaI/SpeI site of pmv361 hyg (Sander et al., 1997). The complete rrn insert was checked by sequence analysis and the corresponding plasmids were transformed into M. smegmatis mc DrrnB (SZ380). RecA-mediated homologous recombination and selective plating (Prammananan et al., 1999; Pfister et al., 2003) was used to introduce the point mutation into the single functional rdna operon of M. smegmatis DrrnB. Nucleic acid sequencing was used to verify that the point mutation had been introduced into the single functional rrna operon by gene conversion and that additional mutations were absent in the part of the rrn fragment involved in homologous recombination. The strains were subsequently propagated on drug-free media. The following mutations were introduced into the single chromosomal rrna operon in this manner: A1408G, G1491A, G1491C, G1491U, C1409G and C1409U. Strain SZ605 with a C1409G/G1491C sequence alteration is a spontaneous mutant derived from a G1491C recombinant strain. As a control for generation time measurements, fitness assays and MIC determinations and to exclude a possible effect of the pmv361 vector with the partial rrn fragment, we transformed M. smegmatis DrrnB (SZ380) with pmv361 hyg carrying the wild-type sequence of the corresponding rrn fragment (pz400) resulting in strains SZ For a list of plasmids used and strains generated see Tables 4 and 5. Recombinant strains with the double mutations C1409U/ G1491A and C1409A/G1491U were generated by gene replacement techniques using an rrn plasmid-exchange strategy as described recently (Hobbie et al., 2007). Strain M. smegmatis DrrnB DrrnA attb : Gm-rrnB-sacB (SZ637) is a derivative of mc where the two endogenous rrn operons have been deleted by unmarked deletion mutagenesis and synthesis of rrn is provided by an integrative plasmid which has integrated at the chromosomal attb locus and which carries the complete rrnb operon under control of its own promoter; gentamicin and sacb are present as selectable markers for positive and negative selection respectively. Replacement of the wt rrn operon by plasmid exchange was done as described previously (Hobbie et al., 2007; 2008a,b) using vector pmih hyg rrnb (rrn derivatives of vectors pmih hyg and pmv361 hyg carry the identical vector backbone and differ only in the size of the rrn fragment pmih hyg and pmv361 hyg were derived from plasmid pmv361 by replacing the kanamycin resistance cassette with a hygromycin resistance marker). Vector pmih hyg rrnb carries the complete rrnb operon under control of its own promoter and the hygromycin resistance cassette as selectable marker. For rrn mutagenesis, approximately 2.8 kb fragments of the rrn decoding region (16S rrna pos. 309 to 23S rrna pos. 1267)

8 Drug resistance and fitness in M. tuberculosis 837 Table 5. Strains used in this study. Resistance marker References Recipient strain Plasmid used for transformation Strains SZ380 DrrnB; parental, unmarked single rrna allelic derivative of M. smegmatis mc 155 Kalapala et al., 2010 SZ a rrs wt pz400 SZ380 Hygromycin This study SZ a rrs 1408G ph128 SZ380 Hygromycin This study SZ a rrs 1491A pz176 SZ380 Hygromycin This study SZ a rrs 1491C pz178 SZ380 Hygromycin This study SZ a rrs 1491U pz177 SZ380 Hygromycin This study SZ508, 509 a rrs 1409G pz191 SZ380 Hygromycin This study SZ520, 521 a rrs 1409U pz175 SZ380 Hygromycin This study SZ605 a rrs 1409G/1491C pz178 SZ380 Hygromycin This study SZ637 DrrnB DrrnA attb : Gm-rrnB-sacB; parental, single rrna allelic derivative of Gentamicin E.C. Böttger and S.N. Hobbie, M. smegmatis mc 155 used for gene replacement studies by plasmid exchange unpublished SZ b rrs wt ph150 SZ637 Hygromycin This study SZ717, 719, 720 b rrs 1409U/1491A ph297 SZ637 Hygromycin This study SZ763, 765, 766 b rrs 1409A/1491U ph154 SZ637 Hygromycin This study a. Recombinant strains derived by vector-mediated integration into the attb site and subsequent introduction of the mutation into the genomic rrna locus by homologous recombination. b. Recombinant strains derived by gene replacement (plasmid exchange). were generated and mutations introduced by PCR mutagenesis. The mutagenized rrn fragments were subcloned into pgem-teasy vector (Promega) and isolated by EcoRV/NcoI digestion. The isolated fragments were ligated into the corresponding sites of vector pmih hyg rrnb. The complete rrn insert was checked by sequence analysis. Plasmid exchange at the attb locus is facilitated by a combined hygromycin/sucrose selection; successful rrn exchange was checked by PCR and sequence analysis. As a control for generation time measurements, fitness assays and MIC determinations and to exclude a possible contribution of the vector pmih hyg rrnb, we transformed M. smegmatis DrrnB DrrnA Gm-rrnB-sacB (SZ637) with pmih hyg carrying the wild-type rrn operon (ph150) resulting in strains SZ For a list of corresponding plasmid and strains generated see Tables 4 and 5. Minimal inhibitory concentrations and measurements of generation time Broth microdilution tests were performed in a microtitre plate format in triplicates for the mutants indicated as described previously (Hobbie et al., 2007). In brief, bacterial strains were cultured on Luria Bertani (LB) agar plates at 37 C. Freshly grown cultures were resuspended in LB broth supplemented with 0.05% of Tween 80 (LB-Tween), diluted to an absorbance at 600nm of and incubated in the presence of twofold serial drug dilutions. After incubation at 37 C for 72 h, the MIC was recorded as the lowest concentration of drug inhibiting visible growth. Growth experiments were done in triplicates for the mutants indicated in 200 ml Erlenmeyer flasks in a total volume of 50 ml of LB-Tween. Freshly, homogeneously grown cells were picked from an agar plate, resuspended thoroughly in 1 ml of LB-Tween and diluted for inoculation resulting in an A 600 of The measurements were started after overnight incubation at an A 600 of approximately 0.1. The A 600 was measured every 2 h for 12 h and generation times were calculated. Fitness assay The cost of a resistance mutation was determined by direct competition against the isogenic drug-susceptible parental strain as described previously (Sander et al., 2002). Equal densities (A 600 = 0.05) of drug-susceptible and drug-resistant strain were mixed and incubated in antibiotic-free LB-Tween broth; subculture was done every 12 h by diluting an aliquot 1:16 in fresh medium (total time period of competition growth was for 72 h). Serial dilutions were plated every 12 h on drug-free agar and on hygromycin containing agar and the number of colonies counted. The number of drug-resistant bacterial colonies was determined by counting the number of colonies on agar containing hygromycin; the number of drugsusceptible cells was calculated as the total number of bacterial cells counted on drug-free agar minus the number of drug-resistant bacterial cells. Resistance to hygromycin is conferred by the vector backbone used for mutant construction; the vectors themselves do not pose a considerable resistance cost as determined in parallel experiments. The

9 838 D. Shcherbakov et al. experiments were performed with either three independent strains carrying the same mutation or by three repetitions with the same strain, as indicated in Table 2. The difference in fitness between two competing strains at time t was computed by use of the following function: S t = ln[(r t/s t)/(r t -1/s t -1) 1/n ] where r t and s t denote the absolute number of drug-resistant and drug-susceptible cells at a given time t, respectively, and r t -1 and s t -1 denote the number of drug-resistant and drug-susceptible cells at a previous time point. S t is called the selection coefficient at time t. The quotient of the ratios of the cell numbers was standardized with the exponent 1/n in which n is the number of generation of the wild type in the time interval. The difference in fitness for a particular mutation was expressed as the average difference in fitness ( standard deviation) obtained for each time interval in independent experiments. Relative bacterial fitness at time t (fit t) was calculated as fit t = 1 + S t. The cpg was calucated as cpg = 1 - e St. Frequency of resistance mutations in clinical isolates The literature was searched for molecular biology-based analyses of mycobacterial drug resistance by use of the key words mycobacteria, aminoglycosides and drug resistance (note that the TB Drug Resistance Mutation Database is incomplete for references describing aminoglycoside resistance mutations; the database curator has been notified by ECB). Reviews were excluded as representing secondary source of information. The literature search resulted in the identification of seven references (Sherman et al., 1996; Alangaden et al., 1998; Suzuki et al., 1998; Krüüner et al., 2003; Maus et al., 2005; Feuerriegel et al., 2009; Jugheli et al., 2009) describing a total of 169 kanamycin-resistant strains of M. tuberculosis for which the mechanism of resistance had been characterized at the molecular level (Table 1). Acknowledgements This study was supported in part by grants from the University of Zurich and the European Community (PAR, FP7-HEALTH ). We thank S.N. Hobbie for experimental support, T. Janusic for expert technical assistance and S. Salas for help in manuscript preparation. References Alangaden, G.J., Kreiswirth, B.N., Aouad, A., Khetarpal, M., Igno, F.R., Moghazeh, S.L., et al. (1998) Mechanism of resistance to amikacin and kanamycin in Mycobacterium tuberculosis. Antimicrob Agents Chemother 42: Andersson, D.I. (2006) The biological cost of mutational antibiotic resistance: any practical conclusions? Curr Opin Microbiol 9: Billington, O.J., McHugh, T.D., and Gillespie, S.H. (1999) Physiological cost of rifampin resistance induced in vitro in Mycobacterium tuberculosis. Antimicrob Agents Chemother 46: Björkman, J., Nagaev, I., Berg, O.G., Hughes, D., and Andersson, D.I. (2000) Effects of environment on compensatory mutations to ameliorate costs of antibiotic resistance. Science 287: Böttger, E.C., and Springer, B. (2008) Tuberculosis: drug resistance, fitness, and strategies for global control. Eur J Pediatr 167: Böttger, E.C., and Springer, B. (2009) Mycobacterium tuberculosis: drug resistance and genetic mechanisms facts, artifacts and fallacies. In HIV and Tuberculosis: A Deadly Liaison. Kaufmann, S.H.E., and Walker, B. (eds). Weinhem: Wiley VCH, pp Böttger, E.C., Springer, B., Pletschette, M., and Sander, P. (1998) Fitness of antibiotic-resistant microorganisms and compensatory mutations. Nat Med 4: Böttger, E.C., Springer, B., Prammananan, T., Kidan, Y., and Sander, P. (2001) Structural basis for selectivity and toxicity of ribosomal antibiotics. EMBO Rep 2: Chambers, H.F., and Sande, M.A. (1996) Antimicrobial agents the aminoglycosides. In Goodman and Gilman s the Pharmacological Basis of Therapeutics, 9th edn. Goodmann, L.S., Limbird, L.E., Milinoff, P.B., Gilman, A.G., and Hardmann, J.G. (eds). New York: McGraw-Hill, pp Donald, R., and van Helden, P.D. (2009) The global burden of tuberculosis combating drug resistance in difficult times. N Engl J Med 360: Feuerriegel, S., Cox, H.S., Zarkua, N., Karimovich, H.A., Braker, K., Rüsch-Gerdes, S., and Niemann, S. (2009) Sequence analyses of just four genes to detect extensively drug-resistant Mycobacterium tuberculosis strains in multidrug-resistant tuberculosis patients undergoing treatment. Antimicrob Agents Chemother 53: François, B., Russell, R.J.M., Murray, J.B., Aboul-ela, F., Masquida, B., Vicens, Q., and Westhof, E. (2005) Crystal structures of complexes between aminoglycosides and decoding A site oligonucleotides: role of the number of rings and positive charges in the specific binding leading to miscoding. Nucleic Acids Res 33: Gagneux, S., Davis Long, C., Small, P.M., Van, T., Schoolnik, G.K., and Bohannan, B.J. (2006) The competitive cost of antibiotic resistance in Mycobacterium tuberculosis. Science 312: Hobbie, S.N., Pfister, P., Bruell, C., Westhof, E., and Böttger, E.C. (2005) Analysis of the contribution of individual substituents in 4,6 aminoglycoside-ribosome interaction. Antimicrob Agents Chemother 49: Hobbie, S.N., Pfister, P., Bruell, C., Sander, P., François, B., Westhof, E., and Böttger, E.C. (2006a) Binding of neomycin-class aminoglycoside antibiotics to mutant ribosomes with alterations in the A-site of 16S rrna. Antimicrob Agents Chemother 50: Hobbie, S.N., Bruell, C., Kalapala, S., Akshay, S., Schmidt, S., Pfister, P., and Böttger, E.C. (2006b) A genetic model to investigate drug-target interactions at the ribosomal decoding site. Biochimie 88: Hobbie, S.N., Kalapala, S.K., Akshay, S., Bruell, C., Schmidt, S., Dabow, S., et al. (2007) Engineering the rrna decoding site of eukaryotic cytosolic ribosomes in bacteria. Nucleic Acids Res 35: Hobbie, S.N., Akshay, S., Kalapala, S.K., Bruell, C., Shcherbakov, D., and Böttger, E.C. (2008a) Genetic analysis of interactions with eukaryotic rrna identify the

10 Drug resistance and fitness in M. tuberculosis 839 mitoribosome as target in aminoglycoside ototoxicity. Proc Natl Acad Sci USA 105: Hobbie, S.N., Bruell, C., Akshay, S., Kalapala, S.K., Shcherbakov, D., and Böttger, E.C. (2008b) Mitochondrial deafness alleles confer misreading of the genetic code. Proc Natl Acad Sci USA 105: Jugheli, L., Bzekalav, N., de Rijk, P., Fissette, K., Portaels, F., and Rigouts, L. (2009) High level of cross-resistance between kanamycin, amikacin, and capreomycin among Mycobacterium tuberculosis isolates from Georgia and a close relation with mutations in the rrs gene. Antimicrob Agents Chemother 53: Kalapala, S.K., Hobbie, S.N., Böttger, E.C., and Shcherbakov, D. (2010) Mutation K42R in ribosomal protein S12 does not affect susceptibility of Mycobacterium smegmatis 16S rrna A-site mutants to 2-deoxystreptamines. PLoS One (in press). Krüüner, A., Jureen, P., Levina, K., Ghebremichael, S., and Hoffner, S. (2003) Discordant resistance to kanamycin and amikacin in drug-resistant Mycobacterium tuberculosis. Antimicrob Agents Chemother 47: Levin, B.R., Perrot, V., and Walker, N. (2000) Compensatory mutations, antibiotic resistance and the population genetics of adaptive evolution in bacteria. Genetics 154: Luciani, F., Sisson, S.A., Jiang, H., Francis, A.R., and Tanaka, M.M. (2009) The epidemiological fitness cost of drug resistance in Mycobacterium tuberculosis. Proc Natl Acad Sci USA 106: Luo, N., Pereira, S., Sahin, O., Lin, J., Huang, S., Michel, L., and Zhang, Q. (2005) Enhanced in vivo fitness of fluoroquinolone-resistant Campylobacter jejuni in the absence of antibiotic selection pressure. Proc Natl Acad Sci USA 102: Maisnier-Patin, S., and Andersson, D.I. (2004) Adaptation to the deleterious effects of antimicrobial drug resistance mutations by compensatory evolution. Res Microbiol 155: Maisnier-Patin, S., Berg, O.G., Liljas, L., and Andersson, D.I. (2002) Compensatory adaptation to the deleterious effect of antibiotic resistance in Salmonella typhimurium. Mol Microbiol 46: Mariam, D.H., Mengistu, Y., Hoffner, S.E., and Andersson, D.I. (2004) Effect of rpob mutations conferring rifampin resistance on fitness of Mycobacterium tuberculosis. Antimicrob Agents Chemother 48: Maus, C.E., Plikaytis, B.B., and Shinnick, T.M. (2005) Molecular analysis of cross-resistance to capreomycin, kanamycin, amikacin, and viomycin in Mycobacterium tuberculosis. Antimicrob Agents Chemother 49: Meier, A., Sander, P., Schaper, K.J., Scholz, M., and Böttger, E.C. (1996) Correlation of molecular resistance mechanisms and phenotypic resistance levels in streptomycinresistant Mycobacterium tuberculosis. Antimicrob Agents Chemother 40: Nagaev, I., Björkman, J., Andersson, D.I., and Hughes, D. (2001) Biological cost and compensatory evolution in fusidic acid-resistant Staphylococcus aureus. Mol Microbiol 40: Nilsson, A.I., Berg, O.G., Aspevall, O., Kahlmeter, G., and Andersson, D.I. (2003) Biological costs and mechanisms of fosfomycin resistance in Escherichia coli. Antimicrob Agents Chemother 47: Paulander, W., Maisnier-Patin, S., and Andersson, D.I. (2009) The fitness cost of streptomycin resistance depends on rpsl mutation, carbon source and rpos (sigma-s). Genetics 183: Pfister, P., Hobbie, S.N., Vicens, Q., Böttger, E.C., and Westhof, E. (2003) The molecular basis for A-site mutations conferring aminoglycoside resistance: relationship between ribosomal susceptibility and X-ray crystal structures. ChemBioChem 4: Pfister, P., Hobbie, S.N., Bruell, C., Corti, N., Vasella, A., Westhof, E., and Böttger, E.C. (2005) Mutagenesis of 16S rrna C1409-G1491 base pair differentiates between 6 OH and 6 NH 3+ aminoglycosides. J Mol Biol 346: Prammananan, T., Sander, P., Brown, B.A., Frischkorn, K., Onyi, G.O., Zhang, Y., et al. (1998) A single 16S ribosomal RNA substitution is responsible for resistance to amikacin and other 2-deoxystreptamine aminoglycosides in Mycobacterium abscessus and Mycobacterium chelonae. J Infect Dis 177: Prammananan, T., Sander, P., Springer, B., and Böttger, E.C. (1999) RecA-mediated gene conversion and aminoglycoside resistance in strains heterozygous for rrna. Antimicrob Agents Chemother 43: Pym, A.S., Saint-Joanis, B., and Cole, S.T. (2002) Effect of katg mutations on the virulence of Mycobacterium tuberculosis and the implication for transmission in humans. Infect Immun 70: Reynolds, M.G. (2000) Compensatory evolution in rifampinresistant Escherichia coli. Genetics 156: Sander, P., Prammananan, T., and Böttger, E.C. (1996) Introducing mutations into a chromosomal rrna gene using a genetically modified eubacterial host with a single rrna operon. Mol Microbiol 22: Sander, P., Prammananan, T., Meier, A., Frischkorn, K., and Böttger, E.C. (1997) The role of ribosomal RNAs in macrolide resistance. Mol Microbiol 26: Sander, P., Springer, B., Sturmfels, A., Prammananan, T., Kappler, M., Pletschette, M., and Böttger, E.C. (2002) Fitness cost of chromosomal drug resistance conferring mutations. Antimicrob Agents Chemother 46: Schrag, S.J., Perrot, V., and Levin, B.R. (1997) Adaptation to the fitness costs of antibiotic resistance in Escherichia coli. Proc R Soc Lond B 264: Sherman, D.R., Mdluli, K., Hickey, M.J., Arain, T.M., Morris, S.L., Barry, C.E., 3rd, and Stover, C.K. (1996) Compensatory ahpc gene expression in isoniazid-resistant Mycobacterium tuberculosis. Science 272: Smith, N.H., Glyn Hewinson, R., Kremer, K., Brosch, R., and Gordon, S.V. (2009) Myths and misconceptions: the origin and evolution of Mycobacterium tuberculosis. Nat Rev Microbiol 7: van Soolingen, D., de Haas, P.E., van Doorn, H.R., Kuijper, E., Rinder, H., and Borgdorff, M.W. (2000) Mutations at amino acid position 315 of the kat G gene are associated with high-level resistance to isoniazid, other drug resistance, and successful transmission of Mycobacterium tuberculosis in the Netherlands. J Infect Dis 182:

11 840 D. Shcherbakov et al. Suzuki, Y., Katsukawa, C., Tamaru, A., Abe, C., Makino, M., Mizuguchi, Y., and Taniguchi, H. (1998) Detection of kanamycin-resistant Mycobacterium tuberculosis by identifying mutations in the 16S rrna gene. J Clin Microbiol 36: Zaunbrecher, M.A., Sikes, D.R., Jr, Metchock, B., Shinnick, T.M., and Posey, J.E. (2009) Overexpression of the chromosomally encoded aminoglycoside acetyltransferase eis confers kanamycin resistance in Mycobacterium tuberculosis. Proc Natl Acad Sci USA 106: