FQ resistance in M. tuberculosis
has been attributed mostly to the modification of the gyrA
QRDR region with mutations at codon 90 and 94 most commonly associated with drug resistance 
. Recent studies have identified mutations such as A74S and T80A which are outside of the defined QRDR in clinical isolates alone or in combination with other QRDR mutations, but the effect of mutations outside the traditional QRDR region on FQ resistance is not clear 
. Although gyrB
mutations in FQ-resistant M. tuberculosis
isolates have recently been reported 
, these mutations need to be assessed genetically to enhance our understanding of FQ resistance and possibly improve the molecular testing of FQ-resistant strains.
To date, the best evidence of involvement of gyrB
mutations in FQ resistance was generated from the use of in vitro
measurements for various FQs toward purified gyrase containing wild type or mutant forms of GyrA or GyrB 
. Many GyrB mutants exhibit higher IC50
values; however it is not known how high the IC50
needs to be to confer resistance to FQs. It would be difficult to standardize these assays to allow one to correlate the IC50
with the exact MIC for each FQ. Additional evidence for the contribution of mutations within gyrB
to FQ resistance is based on identification of gyrB
mutations in phenotypically FQ-resistant isolates 
. However, it is often difficult to compare FQ-resistant isolates from various studies due to the differences in testing procedures (method and medium for drug susceptibility testing (DST)) and the definitions of resistance. In many cases, the FQ-resistant strains harbored not only gyrB
mutations but also gyrA
mutations known to confer resistance and could also contain other unidentified mutations 
To circumvent these concerns, we introduced specific mutations into the chromosomal copy of either gyrA
of fully susceptible M. tuberculosis
strains using a specialized phage system and determined the MIC for various FQs. A strain was considered resistant to a FQ if the MIC was greater than the recommended testing critical concentration by the proportion method suggested by CLSI 
. The allelic exchange system used in this study has been successfully employed by Vilcheze et al
to transfer inhA
S94A mutation to M. tuberculosis
and M. bovis
BCG to unambiguously demonstrate that S94A indeed confers resistance to isoniazid and ethionamide antibiotics. In addition, Starks et al
used this same system to prove that embB
codon 306 confers ethambutol resistance in M. tuberculosis
. One caveat to the phage system is that the efficiency and preferred site of recombination can vary between constructs. Additionally, the location of the desired mutation within the allelic exchange substrate can also affect the efficiency of its integration into the chromosome. Fortunately, we were able to introduce the desired mutations into multiple genetic backgrounds for most mutations analyzed in this study. However, in a few cases, especially for gyrA
, this proved more difficult, and we were ultimately unable to introduce a small number of mutations into one or the other background. We did not experience this problem with the gyrB
mutations probably due to the amount of homologous DNA surrounding the mutations which was much greater as compared to the gyrA
In this study, we identified two mutations, G247S and A384V, located outside of the gyrA
QRDR in FQ-susceptible and resistant isolates. Neither mutation affected the MIC for any of the FQs tested when transferred into a wild type background. These mutations may be naturally occurring polymorphisms and do not play any role in FQ resistance. In contrast to an earlier report 
, the A74S mutation only slightly increased the MIC for OFX and MXF instead of exhibiting high-level resistance. The discordance between the studies could be due to different genetic backgrounds of the strains, and the clinical strain in the earlier report could possibly harbor additional mutations that also affect the MIC level. The A74S mutation did act synergistically with the D94G mutation and increase the MIC 2–8 fold over either mutation alone. Neither the T80A nor A90G mutation conferred resistance either alone or in combination. In fact, strains harboring the A90G mutation were hypersusceptible to the FQs as demonstrated with in vitro
enzymatic assays 
. Thus, with the exception of A74S/D94G combination, mutations outside of the gyrA
QRDR tested in this study did not lead to FQ resistance.
The QRDR region of gyrA
is well defined, and the majority of mutations found in this region confer resistance to FQs, albeit not at the same level (unpublished data). The same is not true for gyrB
, and recently it was proposed to expand the QRDR region of gyrB
to include amino acids 500–540 () 
. More than 15 mutations have been identified in this region alone and several mutations have been identified that are located outside of the QRDR of gyrB
. The list of gyrB
mutations continues to grow as more groups analyze gyrB
of FQ-resistant isolates. Until now, no functional genetic studies in M. tuberculosis
were completed to definitively determine if a specific mutation was able to confer resistance as determined by standardized DST methods.
We analyzed 19 different mutations located either within or outside of the gyrB QRDR to determine their role in FQ resistance. The mutations located outside of the QRDR either had no effect or only slightly increased the MIC levels for the FQs tested. However, this increase was not sufficient to be considered resistant by our testing method. The mutations located within the QRDR of GyrB exhibited an array of MIC levels and conferred resistance to various FQs. D533A was the only mutation located within the QRDR of GyrB that did not exhibit any significant increase in the MIC level for the four FQs. The double mutation, N538T + T546M, was also not sufficient to confer resistance to any of the FQs.
Several substitutions were analyzed at residues 500, 538, 539, and 540 within the QRDR of GyrB, and different substitutions at a single residue did not confer the same level or pattern of resistance. For instance, the D500A mutation did not confer resistance to any of the FQs tested while D500H and D500N conferred resistance to OFX and LVX. The MIC for OFX and LVX was 2 fold higher for these two mutations as compared to D500A. Based upon the structural model of the M. tuberculosis
gyrase in complex with a nicked dsDNA substrate and a quinolone in this and other studies 
, D500 lies in the quinolone binding pocket (QBP) and the aliphatic part of this glutamate residue likely interacts with the alkyl or cycloalkyl group at R1 of the FQs (). The carboxyl group of D500 may be involved in hydrogen bonding interactions with the nearby residues and the DNA base that stacks onto the quinolone ring, fitting snugly between the DNA and the drug. The small and nonpolar alanine substitution of D500 would presumably be involved in a similar interaction while being readily accommodating to substitutions as large as a cyclopropyl at R1 of the FQs. In contrast, the larger side chains of histidine and asparagine that bear a positive charge may alter the geometry and electrostatics of the binding pocket and disfavor FQ binding. These mutations conferred resistance to LVX and OFX which have the same R1 group (the active form of OFX is LVX) and did not confer resistance to MXF or CIP which both have a cylcopropyl group at R1.
A structural model of M. tuberculosis gyrase inhibition.
Residues 538–540 of GyrB form a part of the QBP and interact with the R7 group of FQs. The N538D mutation is one of the more common GyrB mutations found in FQ-resistant isolates 
. In this study, this mutation conferred resistance to all four FQs tested and increased the MIC levels greater than any other single mutation in gyrB
. The side chain amine of this asparagine appears to make a hydrogen bond as a donor with the phosphoribose moiety of the nucleotide that stacks with the quinolone ring. In addition, this side chain can also interact sterically with the R7 group in case of a methyl piperazinyl or a comparable sized substitution, fitting snugly between the drug and the DNA. The mutation of this asparagine to an aspartic acid may remove the electron donating character and introduce a rather unfavorable electrostatic repulsion of the DNA backbone thus disfavoring the quinolone binding and resulting in higher levels of resistance. The lysine substitution at this position increased the MIC for all FQs but was only resistant to MXF. The fold increase of the MIC for MXF was greater than the other FQs. The longer side chain of a lysine residue is not easily accommodated in this region, and since the R7 group of MXF is much larger (azabicyclo) compared to the piperazine and methylpiperazine of CIP, OFX and LVX, this may explain why N538K is resistant only to MXF. A threonine at position 538 (N538T) does not alter the electrostatics of the pocket or its geometry as much as the other substitutions and does not lead to resistance for the FQs tested. Based on our modeling and the high structural conservation of this region 
, the side chain of residue T539 points into the solvent; consistently, substitutions at this position (T539N and T539P) did not reproducibly confer resistance. The T539P mutation did confer low-level resistance to MXF for one round of testing. The nonpolar character of the solvent-exposed proline or the higher rigidity of the backbone introduced by this substitution may be structurally perturbing to the backbone, affecting the flanking residues (N538 and E540), both of which interact directly with the drug.
The E540V mutation conferred resistance to all four FQs in this study which was also reported for a clinical isolate with the same mutation in recent studies 
. This glutamate appears to interact with the drug through hydrophobic interactions between its aliphatic portion and a hydrophobic group at R7 (e.g the methyl of the methyl piperazinyl) whereas the carboxylate likely forms a salt bridge with R521 thereby positioning it for several direct interactions with both DNA and the drug. The valine substitution at this position likely perturbs this arginine coordination. These alterations are sufficient to confer resistance to multiple classes of FQs. Interestingly, the substitution of glutamic acid for aspartic acid (E540D) conferred high-level resistance to MXF and did not significantly alter the MIC for the other three FQs. This substitution may still accommodate the salt bridge with the arginine and would only slightly shorten the side chain, likely affecting interactions with R7. The aspartate residue seems to be able to form hydrogen bonds with CIP, OFX, and LVX which have very similar R7 groups (piperazine and methylpiperazine) but not with the larger R7 group (azabicyclo) of MXF. In line with these structural observations, the effect of the aspartate substitution at this position is relatively subtle.
Mutations located outside of the QRDR of gyrB were not capable of conferring resistance alone. However, we identified a unique double mutation, R485C + T539N, with one mutation located within and one outside of the QRDR that conferred resistance to all four FQs, and the MIC levels were usually higher than any single gyrB mutation. Based on structural modeling (), the T539N mutation is located within the QBP as described above. The R485C mutation is located at the GyrA-GyrB interface and not near the quinolone binding site of the gyrase complex. The arginine may be important for interactions with GyrA to properly position GyrB relatively to DNA and thus substitutions at this position could affect the quinolone binding pocket allosterically. Together, these two mutations may destabilize the QBP significantly enough to confer resistance. Based on these data, the structural modeling appears to explain the results of the MIC levels and resistance patterns observed for the GyrB mutations investigated in this study.
Based on current literature, up to 58% of FQ-resistant M. tuberculosis
strains have no identified mutation in the QRDR region of gyrA
and possibly possess an alternate mechanism of resistance 
. Factors such as decreased cell wall permeability, active efflux pumps, and drug sequestration or inactivation have been proposed to account for FQ resistance in these isolates 
. In a previous study, we sequenced the gyrA
QRDR of 98 FQ-resistant isolates and identified mutations within this region in more than 80% of these isolates. The cause for FQ resistance in the remaining isolates was unknown. Recent publications have suggested a link between mutations within gyrB
and FQ resistance, and in the present study, we have generated gyrB
mutations in well-studied genetic backgrounds and demonstrated conclusively that certain mutations within the GyrB QRDR do lead to FQ resistance. Subsequently, we identified FQ-resistance conferring gyrB
mutations in several of the WT-gyrA
isolates from our previous study. During the course of the present study we came to appreciate that the level of gyrA
sequence heterogeneity among specimens isolated from individuals is relatively high with many individuals having both wild-type and mutant gyrA
sequences. Since, in our hands, Sanger sequencing can only detect sequences that make up greater than 25–50% of the population we chose to enrich for the resistant population in the remaining WT-GyrA, WT-GyrB isolates by growing them on FQ containing media prior to sequencing. Consequently, we identified mutations within gyrA
in all remaining isolates. Based on these results, we believe that the majority, if not all, FQ resistance in M. tuberculosis
can be attributed to single nucleotide polymorphisms (SNPs) in gyrA
. Importantly, several rapid molecular tests have been developed to assess FQ resistance in M. tuberculosis
utilizing mutations within the GyrA QRDR as markers for FQ resistance 
, and based on the current study, inclusion of the QRDR of gyrB
in rapid molecular testing which would detect specific substitutions in gyrB
would provide a more complete picture of FQ resistance. Unfortunately, FQ resistance imparted by GyrB QRDR mutations seems to be more complex than is the case for GyrA mutations, and the genetic background appears to have some effect on resistance. Molecular assays that analyze the QRDR region of gyrB need to determine the exact mutation since not all mutations confer resistance and the pattern of cross-resistance varies among the mutations.
Most laboratories performing DST for M. tuberculosis only test at the critical concentration recommended for their specific testing method. Strains harboring mutations leading to a higher MIC level than wild-type strains but equal to or slightly less than the critical concentration would test susceptible with conventional testing. Data presented in this study demonstrates that this would be the case for many of the gyrB mutations. Unfortunately, the importance of these mutations in patient care is unknown. Clinical evidence establishing the efficacy of treatment of individuals infected with strains harboring these types of mutations with various FQ is lacking. However, molecular assays could identify these mutations that result in borderline resistance levels and alert clinicians to possible treatment complications, and in some cases, the genetic information could be useful in tailoring the treatment regimen for the patient.