A Fluoroquinolone Resistance Protein from Mycobacterium tuberculosis That Mimics DNA

See allHide authors and affiliations

Science  03 Jun 2005:
Vol. 308, Issue 5727, pp. 1480-1483
DOI: 10.1126/science.1110699


Fluoroquinolones are gaining increasing importance in the treatment of tuberculosis. The expression of MfpA, a member of the pentapeptide repeat family of proteins from Mycobacterium tuberculosis, causes resistance to ciprofloxacin and sparfloxacin. This protein binds to DNA gyrase and inhibits its activity. Its three-dimensional structure reveals a fold, which we have named the right-handed quadrilateral β helix, that exhibits size, shape, and electrostatic similarity to B-form DNA. This represents a form of DNA mimicry and explains both its inhibitory effect on DNA gyrase and fluoroquinolone resistance resulting from the protein's expression in vivo.

Increasing resistance to two bactericidal compounds that act on rapidly growing Mycobacterium tuberculosis, isoniazid and rifampicin, is driving the search for new therapies. Fluoroquinolones exert their powerful antibacterial activity by interacting with DNA gyrase and DNA topoisomerase IV (1). They bind reversibly to the enzyme-DNA complex and stabilize the covalent enzyme tyrosyl-DNA phosphate ester, which is normally a transient intermediate in the topoisomerase reaction. Hydrolysis of this linkage leads to the accumulation of double-stranded DNA fragments and is the bactericidal consequence of fluoroquinolone treatment. Newer fluoroquinolones, including moxifloxacin and gatifloxacin, exhibit powerful in vitro activity against mycobacteria (2, 3), and they can reduce multidrug treatment regimens from 6 to 4 months when substituted for isoniazid (4). Resistance to fluoroquinolones remains rare in clinical isolates of M. tuberculosis (5), but it has been increasing as their use in the treatment of multidrug-resistant M. tuberculosis infections increases (6). High-level resistance to fluoroquinolones in laboratory strains of M. tuberculosis and M. smegmatis (7, 8) is known to result from amino acid substitutions in the putative fluoroquinolone binding region of the M. tuberculosis gyrA-encoded A subunit of DNA gyrase (7, 8). This is the only type II topoisomerase encoded in the M. tuberculosis genome (9) and thus is the unique target for fluoroquinolones in this organism (10, 11).

Genetic selection for fluoroquinolone resistance in M. smegmatis identified a new resistance mechanism (12). The mfpA-encoded protein, when expressed from a multicopy plasmid, resulted in low-level resistance (a 4- to 8-fold increase in the minimum inhibitory concentration) to ciprofloxacin and sparfloxacin. The sequence of MfpA revealed it to be a memberofthe “pentapeptide repeat” family of bacterial proteins (13), in which every fifth amino acid is either a leucine or phenylalanine. M. tuberculosis contains a 183–amino acid MfpA homolog (MtMfpA), encoded by the Rv3361c gene, that is 67% identical to the 192-residue M. smegmatis MfpA protein. A second member of the pentapeptide repeat family is the McbG protein responsible for resistance to microcin B17 in Escherichia coli (14). Microcin B17 also inhibits DNA gyrase (15), although by a different mechanism of action than fluoroquinolones (16). A third member of the pentapeptide repeat family is the plasmid-encoded Qnr protein, originally identified in quinolone-resistant strains of Klebsiella pneumoniae (17). This plasmid-encoded protein protects DNA gyrase against fluoroquinolone inhibition (18), and similar proteins have been identified in fluoroquinolone-resistant clinical isolates of Enterobacteriae in Japan (19) and Europe (20).

The M. tuberculosis Rv3361c open reading frame was amplified by polymerase chain reaction from M. tuberculosis strain H37Rv genomic DNA and ligated into a pET28a plasmid. Expression in E. coli strain BL21(DE3), induced by isopropyl-β-d-thiogalactopyranoside and transformed with the plasmid, yielded extracts from which the protein could be purified to homogeneity using nickel-nitriloacetate chromatography.

The expressed MfpA protein was tested for its effect on both adenosine triphosphate (ATP)–dependent DNA supercoiling and ATP-independent relaxation reactions catalyzed by E. coli DNA gyrase. MtMfpA inhibited both reactions in a concentration-dependent manner (Fig. 1, A and B). The apparent median inhibitory concentration (IC50) values were calculated to be ∼1.2 μM (based on an active dimer, see below) for both reactions. To distinguish between indirect effects on catalysis or the direct interaction of MfpA with gyrase, we performed surface plasmon resonance experiments using standard amine coupling of MfpA to carboxymethyl sensor chips. The signal was saturable with increasing concentrations of gyrase and allowed us to calculate a dissociation constant (Kd) value of 460 nM from the ratio of kon and koff values of ∼103 M–1 s–1 and ∼10–4 s–1, respectively (Fig. 1C). These values are in approximate agreement with the IC50 values obtained for gyrase inhibition, indicating that MfpA interacts directly with DNA gyrase.

Fig. 1.

(A and B) Inhibition of supercoiling and relaxing activity of E. coli DNA gyrase by MfpA. (A) Lane 1, relaxed plasmid pBR322 alone; lane 2, relaxed pBR322 containing 5 μM MfpA; lane 3, 3 units of gyrase; lanes 4 to 8, 3 units of gyrase and 1, 2, 3, 5, and 8 μM MfpA, respectively. (B) Lane 1, supercoiled pBR322; lane 2, supercoiled pBR322 containing 5 μM MfpA; lane 3, supercoiled pBR322 with 25 units of gyrase; lanes 4 to 8, 25 units of gyrase containing 1, 2, 3, 5, and 8 μM MfpA, respectively. nc, l, and sc represent nicked circular, linear, and supercoiled forms, respectively. Gyrase assays were performed as described elsewhere (SOM text). (C) BIAcore sensor-grams of DNA gyrase binding to MfpA. DNA gyrase, at 12 μM (line 1), 6 μM (line 2), 1.5 μM (line 3), 0.75 μM in duplicate (line 4), 0.375 μM in duplicate (line 5), 0.188 μM (line 6), and 0.094 μM (line 7), and the buffer alone (line 8) were injected to immobilized MfpA as described, and the sensograms were recorded. Black lines show experimentally recorded values; red lines are a fit of the data to a 1:1 Langmuir model.

MfpA was crystallized by vapor diffusion under oil, and both native and selenomethionine-substituted proteins were crystallized in several space groups that diffracted to 2.0 to 2.7 Å. Diffraction data on selenomethionine-substituted protein crystals in space group P3221 were collected at three wavelengths (table S1). Higher resolution data from the native protein in the P21 crystal form were added to extend the phases and improve the quality of the maps (fig. S1). The final structure was refined to 2.0 Å (Table 1 and table S2).

Table 1.

Data collection and refinement statistics for MfpA. Statistics for the highest bin are in parentheses. RMS, root mean square.

Data collection and refinement statistics
Data statistics
Space group P21
Unit cell (Å/°) a = 53.8, b = 31.0, c = 96.8, β = 93.2
Maximum resolution (Å) 2.0 (2.0 to 2.07)
Completeness (%) 98.8 (95.8)
Rsym (%) 3.2 (16.1)
Mean I/σ(I) 29.6 (7.4)
Redundancy 4.4 (3.5)
Refinement statistics
Model A1 to A183, B1 to B180, 205 H2O, 1 SO4
Rwork/Rfree (%) 17.7 (16.6)/21.8 (22.2)
RMS deviations from ideal Bond (Å)/angle (°) 0.021/1.89
Average B-factor (Å2) Protein/nonbonded 18.7/28.9

MfpA is a dimer in solution (21) and in the crystal, with the C-terminal α helices interacting to generate the dimer. The MfpA monomer is almost entirely composed of a right-handed β helix (Fig. 2A) (residues 2 to 165 out of 183 residues) that has eight complete coils, each with four nearly equivalent sides, giving the core of the structure a quadrilateral appearance (Fig. 2B). The coils are stacked upon each other with only a slight left-handed twist. Each of the sides is made up of one of the pentapeptide-repeating units, with the middle hydrophobic residue (i) and the first small polar or hydrophobic residue (i–2) pointing inwards and the remaining residues (i–1, i+1, and i+2) pointing outwards (Fig. 2C and fig. S2). There is extensive hydrogen bonding interaction between the peptide backbone atoms of neighboring coils, including in the turns, although only the i–1 residue is consistently in a full parallel β strand interaction. Every 20 residues, the right-handed β helix completes a revolution and travels ∼4.8 Å along the helical axis. The core of the β helix is devoid of water but is not entirely hydrophobic in nature. Where there are violations of the hydrophobic nature of the i residue, a compensatory polar residue is positioned nearby, to which it hydrogen bonds. For example, the side chains of Thr24 and His44 (i position residues in coils 2 and 3, respectively) are on consecutive coils on the same face and form a hydrogen bond. His44 also forms a hydrogen bond with the side chain Ser42 at the i–2 position. Where there is a small polar residue at position i–2, its side chain typically points into the corner of the quadrilateral and forms hydrogen bond(s) with backbone amides or carbonyls of its own adjacent turn or with that of turns below its position. Where the small polar/hydrophobic residue at position i–2 rule is broken, this requires compensatory changes nearby to allow for a larger residue. For example, Asn97 and Leu102 are allowed at the i–2 position because the helical axis tilts here, creating a larger separation between coils 4 and 5. This tilt is caused by the presence of Pro81 between faces 4 and 1, which leads to a disruption of the hydrogen bonding between coils 4 and 5 and a 12° change in the helical axis of coils 1 to 4 and 5 to 8. There are several tightly bound water molecules that accommodate the open hydrogen bonds thus created.

Fig. 2.

MfpA structural fold illustrations. (A) Stereoview of the Cα trace of the MfpA dimer, shown with the monomers colored from blue (N terminus) to red (C terminus). Every tenth Cα is shown as a small sphere. Every 20th residue is labeled. (B) The Cα trace, from the N terminus (blue) to C terminus (red), viewed down the helical axis of an MfpA monomer. (C) Stick representation of residues 2 to 81 (coils 1 to 4), viewed down the helical axis, colored by atom type (carbon, gray; oxygen, red; nitrogen, blue; sulfur, orange).

Both the N- and C termini of the β helix are capped by tryptophan residues in the i position (Trp4 and Trp154). The 20 C-terminal residues appear as a two-turn (α1) and a three-turn (α2) helix, with the former occupying the place of the face 3 β strand of coil 8. The C-terminal α2 helices interact in an antiparallel manner to generate the molecular twofold axis and a hydrophobic dimer interface that is observed in all four crystal forms. The dimer is rod-shaped and highly asymmetric, with a length of ∼100 Å and a diameter of 27 Å at the N termini and 18 Å at the dimer interface. Although the main chain atoms form coils of nearly square quadrilaterals when viewed down the long axis, the outward-facing side chains of the i–1, i+1, and i+2 residues produce a protein surface with a more cylindrical shape when viewed in this direction. All of the charged residues (19 Arg, 1 Lys, 18 Asp, and 7 Glu residues) are located at these positions, generating a dimer with an overall charge of –10. However, the charge distribution is not uniform, and there is a distinct negative potential on faces 1 and 2 along the length of the molecule (Fig. 3A). The right-handed helical nature of the fold, the dimensions and shape of the dimer, and the negative electrostatic surface potential suggest that MfpA might be mimicking a 30–base pair segment of B-form duplex DNA and could be capable of interacting directly with DNA gyrase. No structures in the Protein Data Bank share significant similarity to MfpA.

Fig. 3.

Electrostatics of MfpA/GyrA and a model of their interaction. (A) Surface representation of the MfpA dimer, showing either faces 1 and 2 (front) or faces 3 and 4 (back). (B) Molecular model of dimeric MfpA (green trace) bound to dimeric GyrA59 (surface) in two orthogonal orientations. Surfaces are colored according to electrostatic projections at a level of +5 (blue) and –5 (red) kT/e, where k is Boltzmann's constant, T is the absolute temperature, and e is the magnitude of the electron charge.

The structures of the MfpA protein and the N-terminal domain of the E. coli gyrase A subunit (GyrA59) (22) could be readily docked, without steric clashes, to provide electrostatic complementarity between the highly cationic “saddle” at the gyrase A2 dimer interface, thought to be the position where DNA binds and is cleaved, and the highly anionic surface of the MfpA dimer (Fig. 3B). The MfpA dimer extends across the entire GyrA dimer, providing an explanation for its powerful inhibition of gyrase activity and suggesting that MfpA will compete with B-form DNA for the gyrase surface. Because fluoroquinolones bind only to the DNA gyrase-DNA complex (23), MfpA binding to DNA gyrase prevents the formation of this complex and provides a molecular explanation for the resistance phenotype. Additional support for this mechanism comes from the recent report that Qnr from K. pneumoniae competes with DNA for DNA gyrase (24).

DNA mimicry by proteins has been reported for the interaction of TAFII230 with the TATA box-binding protein (25) and in the structure of the highly acidic 107–amino acid residue HI1450 protein from Haemophilus influenzae (26). The HI1450 structure bears some overall structural similarity to the gyrI-encoded DNA gyrase inhibitor (27) that protects cells from microcin B17 (also referred to as SbmC), the structure of which has also been solved by crystallographic methods (28). The structure of the bacteriophage T7 Ocr protein (29) reveals a dimer with a surface anionic charge that has also been suggested to mimic the surface charge distribution of DNA. However, MfpA is folded into a structure that is itself a right-handed helix with a size, shape, and charge distribution markedly reminiscent of B-form DNA. It appears likely that the other members of this large bacterial family of pentapeptide repeat proteins (Protein family database: pf00805) (30) will adopt a similar overall fold.

The physiological role that might be played by the MfpA family of proteins in the various organisms in which they are found is not yet clear. The negative surface potentials and DNA-like proportions of other pentapeptide repeat proteins suggest that they may be a general class of inhibitors of DNA binding proteins, with properties analogous to MfpA inhibition of DNA gyrase. A single report has appeared in which the transcript level of mfpA (Rv3361c) has been altered by treatment of M. tuberculosis cultures (31), and its upstream and downstream neighbors (Rv3360 and Rv3362) have been reported as nonessential (32). In M. tuberculosis, expression of MfpA may be coordinated with cell replication to provide DNA topological assistance when needed, but maintain a condensed chromosome and prevent undesired topological changes during periods of replicative senescence. Viewed in this context, mechanisms that would either control expression of MfpA or modulate its activity would be likely. Finally, the core of the right-handed quadri-lateral β-helix structure appears robust enough to allow for surface amino acid substitutions that could tailor specificity and could provide a platform for the rational design of proteins that specifically target DNA-binding proteins of known structure.

Supporting Online Materials

Materials and Methods

Figs. S1 and S2

Tables S1 and S2

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article