Structural Basis of 5-Nitroimidazole Antibiotic Resistance

5-Nitroimidazole-based antibiotics are compounds extensively used for treating infections in humans and animals caused by several important pathogens. They are administered as prodrugs, and their activation depends upon an anaerobic 1-electron reduction of the nitro group by a reduction pathway in the cells. Bacterial resistance toward these drugs is thought to be caused by decreased drug uptake and/or an altered reduction efficiency. One class of resistant strains, identified in Bacteroides, has been shown to carry Nim genes (NimA, -B, -C, -D, and -E), which encode for reductases that convert the nitro group on the antibiotic into a non-bactericidal amine. In this paper, we have described the crystal structure of NimA from Deinococcus radiodurans (drNimA) at 1.6 Å resolution. We have shown that drNimA is a homodimer in which each monomer adopts a β-barrel fold. We have identified the catalytically important His-71 along with the cofactor pyruvate and antibiotic binding sites, all of which are found at the monomer-monomer interface. We have reported three additional crystal structures of drNimA, one in which the antibiotic metronidazole is bound to the protein, one with pyruvate covalently bound to His-71, and one with lactate covalently bound to His-71. Based on these structures, a reaction mechanism has been proposed in which the 2-electron reduction of the antibiotic prevents accumulation of the toxic nitro radical. This mechanism suggests that Nim proteins form a new class of reductases, conferring resistance against 5-nitroimidazole-based antibiotics.

dimetridazole (DMZ), and tinidazole (TNZ). MTR is extensively used in the treatment of anaerobic infections caused by Trichomonas vaginalis, Entamoeba histolytica, Enterococcus species, Giardia lamblia, Clostridium species, and Bacteroides (1)(2)(3)(4) and is also a critical ingredient of modern multidrug therapies for Helicobacter pylori eradication regimes used to control ulcers (5). The mode of action for the 5-Ni antibiotics, as illustrated in Fig. 1, has been shown to be similar in different pathogens (6 -8). The inactive prodrug enters cells by simple diffusion and is then reduced in a 1-electron reduction into the toxic compound, the short-lived radical anion R-N⅐O 2 Ϫ . This reaction is mediated by ferredoxin, which receives an electron from the pyruvate-ferredoxin oxidoreductase (PFOR) complex via conversion of pyruvate to acetyl coenzyme A (9). The resulting nitro radical anion probably causes DNA strand breaks, DNA helix destabilization, unwinding of DNA, and finally cell death (1,2,10,11), and damage to other vital cell systems is also possible (6).
The success of such drugs depends on the reductive activation of the nitro group on the 5-Ni drug, which is controlled by the redox system of the target cell. As a consequence, species with altered, absent, or elevated redox potential pathways are resistant to 5-Ni drugs (see Ref. 12 and references therein). For H. pylori, the most convincing data regarding MTR resistance relate to inactivation of the RdxA gene, which encodes an oxygen-insensitive NADPH nitroreductase (13). Still, resistance has been found in H. pylori strains with an intact RdxA gene (14). In T. vaginalis, a reduced amount of available ferredoxin as an electron acceptor/donor is thought to be responsible for drug resistance (15), but strains with knock-out ferredoxin genes are not resistant under aerobic or anaerobic conditions (16). Therefore, it is likely that multiple pathways lead to both activation and resistance of the MTR and other 5-Ni drugs.
The 5-Ni resistance of some of Bacteroides fragilis strains was shown to be mediated by specific genes, named Nim, located either on the chromosome (NimB) or on small mobilizable plasmids, e.g. pIP417 (NimA), pIP419 (NimC), and pIP421 (NimD) (17)(18)(19)(20). A fifth Nim gene, NimE, was discovered that confers resistance to high MTR concentrations in strains from Bacteroides thetaiotaomicron, B. fragilis, and Bacteroides ovatus (21). The enzymatic activity of the Nim gene products was deduced by comparing the metabolism of a 5-Ni-susceptible strain with the same strain harboring a plasmid containing the NimA sequence from B. fragilis (bfNimA) (22). In the sensitive strain, the classic reduction of DMZ to its nitro radical anion was observed in agreement with the general scheme (see Fig. 1). However, in the resist-ant strain, DMZ was reduced to its amine derivative (R-NH 2 ) through a low redox potential reaction. The addition of pyruvate, which can act as an electron donor through the PFOR complex ( Fig. 1), increased the drug uptake (22). Hence, it was proposed that Nim proteins are 5-Ni reductases, which possibly use ferredoxin as the electron donor (22).
Following the increasing availability of new bacterial genome sequences, it emerges that Nim homologues are present in other genera of bacteria, including Deinococcus radiodurans, B. fragilis, Helicobacter hepaticus, Clostridium sp., Salmonella typhi, Streptomyces avermitilis, as well as in Archaea Methanosarcina sp. (Fig. 2). Although the underlying physiological function of the Nim homologues in these organisms is unknown, it seems likely that the Nim gene family is ancient and widespread. More importantly, because of its unstable nature on mobilizable plasmids, Nim gene-based resistance poses a real threat to the existing applications of 5-Ni drugs. Indeed, a recent study identified seven Bacteroides-resistant strains with minimum inhibitory concentrations of Ͼ32 g/ml, all containing Nim genes (23).
In this paper, we present the first crystal structure of a Nim enzyme, the NimA from D. radiodurans (drNimA, DR0842) at 1.6 Å resolution as well as its complexes with the MTR antibiotic, covalently bound pyruvate and covalently bound lactate. drNimA shares 28% sequence identity and 54% sequence homology with bfNimA, and should thus be representative of the Nim family as a whole. Taken together with previous studies on NimA from B. fragilis and comparison with other closely related enzymes, our observations suggest that Nim proteins do indeed function as reductases. We propose a mechanism for the reduction of 5-Ni compounds by Nim enzymes that leads to generation of non-toxic derivatives and confers resistance against these antibiotics.

EXPERIMENTAL PROCEDURES
Cloning, Protein Expression, and Purification of DR0842-The gene DR0842 (drNimA, 21.9 kDa), was cloned from genomic D. radiodurans DNA into the Gateway destination vector pDEST17 (Invitrogen) by the company Protein'eXpert SA, Grenoble, France. The correct sequence with an amino-terminal hexahistidine tag (sequence MSYYHHHH-HHLESTSLYKKAG) has been confirmed by DNA sequencing.
E. coli BL21(DE3)pLysS cells (Novagen) transformed with pD-EST17-drNimA were grown at 37°C in rich broth 2ϫYT medium (16 g/liter bacto-tryptone, 10 g/liter bacto-yeast extract, 10 g/liter NaCl) with 100 mg/liter ampicillin and 34 mg/liter chloramphenicol, and at A 600 of 0.5, the cells were induced with 1 mM isopropyl-␤-D-thiogalactopyranoside for 4 h at 37°C. The harvested cells were resuspended in lysis buffer (20 mM Tris-HCl, pH 7.2, 500 mM NaCl, 10 mM imidazole) supplemented with DNaseI, Lysozyme, and Complete EDTA-free Protease Inhibitor Mixture (Roche) and lysed by sonication, and the soluble lysate was applied to nickel-nitrilotriacetic acid resin (Qiagen). The protein was eluted with a linear gradient of imidazole at a concentration of from 10 to 500 mM. Fractions containing drNimA were de-salted (HiTrap desalting column, Pharmacia Corporation) and loaded onto a MonoQ column (Amersham Bioscience) and further eluted with a NaCl gradient (0 -1 M) in which drNimA eluted at around 0.1 M NaCl. Fractions with drNimA were pooled and further purified by analytical gel filtration (Superdex 200, Pharmacia Corporation) in a buffer consisting of 10 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol, 1 mM EDTA. The peak fractions were concentrated to 20 mg/ml. Electrospray mass spectrometry of the purified drNimA confirmed the molecular mass of ϳ24 kDa (drNimA 21.9 kDa ϩ His-tag 2.5 kDa).
Oligomeric State and Cross-linking-An additional gel filtration run (in 25 mM Tris-HCl, pH 8.0, 100 mM NaCl) was also carried out with drNimA and the standard molecular mass markers, albumin (67 kDa) and chymotrypsin (25 kDa), to correlate the elution volume of drNimA to the molecular mass of the protein in solution. One run with only the molecular mass marker, ovalbumin (43 kDa), was also performed. All four proteins were at concentrations of ϳ1 mg/ml.
Another analytical gel filtration run (Superdex 200 (Pharmacia Corporation) in 50 mM Hepes, pH 7.5, and 1 mM EDTA) with the pure drNimA protein was performed to change the buffer from Tris to Hepes, and this elution curve is given in Fig. 4a. The main peak from this run was used for the cross-linking experiments. A series of 0, 0.25, 0.50, 1.0, and 5.0 mM cross-linker, ethylene glycol-bis (succinic acid N-hydroxysuccinimide ester) (Sigma) in Me 2 SO was added to the protein and left for 20 min on the bench, and then Tris was added to a final concentration of 60 mM. The series was analyzed on a 12% SDS-PAGE gel (see Fig. 4b).
Crystallization, Soaking, Data Collection, and Structure Solution-Initial screening for suitable crystallization conditions was carried out by the hanging drop method using standard commercial screening solutions. The final crystallization conditions at 4°C had 4-l hanging drops consisting of a 1:1 mixture of protein (6 mg/ml) and reservoir solution with 0.65-0.9 M sodium acetate and 0.1 M sodium cacodylate, citrate, or MES buffered at pH 5.5-6.0. The crystals grew as rosettes with plate-like "fingers," which were cracked, and the resulting plates with an approximate size of 100 ϫ 30 ϫ 5 m 3 were used for data collection. For structure solution purposes, a mercury derivative (drNimA-Hg) was prepared by soaking the crystal in a solution with 1 mM ethyl mercury thiosalisylate, 0.2 M sodium acetate, 0.1 M sodium cacodylate, pH 6.0, and 30% polyethylene glycol 4000 at 4°C for 20 min.
The complex structures described in this paper were prepared by soaking the native crystals in solution containing 0.2 M sodium acetate, 0.1 M sodium cacodylate, pH 6.0, and 30% polyethylene glycol 4000 and 1, 5, or 10 mM MTR or TNZ (both from Sigma) at 4°C. As will be seen, soaking for 2 h in 10 mM MTR produced a structure with drNimA in complex with MTR (drNimA-MTR), whereas soaking for 20 h in 1 mM TNZ yielded a complex with covalently bound pyruvate (drNimA-Pyr), and soaking for 23 h in 1 mM MTR produced a complex with drNimA and covalently bound lactate (drNimA-Lac).
All diffraction data were collected at the European Synchrotron Radiation Facility, Grenoble, France, using crystals cooled to 100 K. Cryo-protection for the native crystals was effected by leaving the crystals for ϳ20 s in a solution with 0.2 M sodium acetate, 0.1 M sodium cacodylate, pH 6.0, and 30% polyethylene glycol 4000. Crystals of the mercury derivative and of the various complexes could all be frozen straight from their soaking solutions. The crystals of drNimA (including the substrate soaks) showed great sensitivity to ambient temperatures; therefore all manipulations described above and flash freezing in liquid nitrogen were preformed in a cold room maintained at 4°C.
All crystals belong to the space group C2, and the unit cell dimensions for the native crystal were a ϭ 99.86 Å, b ϭ 38.95 Å, c ϭ 59.81 Å, and ␤ ϭ 114.25°(see Table I for further details). All data were integrated using MOSFLM, scaled with SCALA, and structure factors FIG. 1. Schematic activation mechanism of 5-Ni drugs in anaerobic bacteria, here illustrated with metronidazole. Pyruvate is oxidized into acetyl coenzyme A by the pyruvate-ferredoxin oxidoreductase (PFOR) complex, and further, PFOR reduces ferredoxin (Fd), which finally reduces metronidazole in a single electron transfer into the toxic free radical. Reduced Fd level in T. vaginalis has been related to resistance (15). This figure has been adapted from Ref. 15. ox, oxidized state; red, reduced state.

TABLE I Statistics from the data collections
The numbers in parentheses represent values in the highest of 10 resolution shells, and the resolution limits for these are indicated.  obtained using TRUNCATE software (24). The structure of drNimA was elucidated using the single isomorphous difference with anomalous scattering technique, with initial experimental phases based on the single ethyl mercury thiosalisylate derivative obtained to 2.1 Å using the software program SOLVE (25). Phase improvement and extension to 1.6 Å resolution were carried out with RESOLVE (26), and ARP/wARP software (27) was then used to produce an initial model for drNimA (192 of the final 224 residues). There is one monomer in the crystallographic asymmetric unit resulting in a solvent content of 43% and a Matthews coefficient of 2.2 Å 3 /Da. Phases were improved by iterative cycles of refinement in the REFMAC program (28) interspersed with rounds of manual rebuilding in O software (29), during which solvent molecules were incorporated. The structures of the drNimA complexes described here were obtained by using the final model of the native structure (drNimA) stripped for solvent molecules as the starting point for the refinement, and model improvement was carried out as for the native structure. Full details of the data collections and the results of all refinements are given in Tables I and II.  Analytical gel filtration and cross-linking experiments clearly indicate that drNimA is a homodimer (Fig. 4). This homodimer is, in the crystals, formed by the crystallographic 2-fold axis. When the dimer is being formed, it buries a surface area of 2314 Å 2 that is 18% of the area in a monomer. In the dimer, the long ␤-strands (␤4, ␤5, ␤6) of one monomer grip onto the ␤-barrel of the other monomer forming a ␤-propeller with ten strands (Fig. 3, c and e) with approximate dimensions of 75 ϫ 47 ϫ 43 Å 3 . The two drNimA monomers are held together by 10 hydrogen bonds, including a salt bridge between Arg-38 and the carboxyl group of Glu-91, some aromatic interactions, and several water-mediated interactions at the dimer interface.

Structure
Native Structure-During the refinement of the drNimA structure, a flat, X-shaped portion of difference electron density was observed in a solvent-exposed pocket on the monomermonomer interface (Fig. 5a). This density is at hydrogen-binding distance to the absolutely conserved residue His-71 (Fig. 2). To form an interpretation of this electron density, several possible molecules were tested for suitability. Only molecules satisfying the constraints of size and shape placed by the difference electron density, and also conforming to the chemical environment available, were considered. Ultimately, the most satisfactory explanation of the residual difference density was obtained by the incorporation of a pyruvate moiety into the model. The assignment of a pyruvate molecule here seems to be reasonable, because all three oxygen atoms are involved in hydrogen binding networks (Fig. 5b), and the hydrophobic methyl group (of the pyruvate) is facing Phe-140 from one molecule and Phe-98Ј from the second monomer (Fig. 5b, where the "Ј" implies residues from the second monomer in the homodimer). The refined pyruvate moiety fits within the observed electron density (Fig. 6a), it has reasonable bond lengths and bonds angles, and no difference density was observed when the structure refinement was completed.
In the final native structure, the pyruvate is located between the ␤2 and ␤3 strands, and O-1 in the carboxyl group is 2.38 Å away from His-71 N⑀-2. The three oxygen atoms of the pyruvate are hydrogen bound to the four water molecules W-1, W-2, W-3, and W-5. The amino acids Val-139 and Leu-107Ј, along with Phe-140 and Phe-98Ј, also contribute to the binding site (Figs. 3f and 6a).
Comparison of Sequence and Structural Homologues-To search for sequence and structural homologues, the drNimA sequence and coordinates were used with the BLAST (30) and DALI servers (31). BLAST identified a number of sequence homologues in the data base, which included sequences from the Clostridium species, Bacteroides species, H. hepaticus and S. typhi, and Archaea (Methanosarcina mazei). The sequence identity of drNimA toward the sequences included in Fig. 2 is 14 -23%, and the homology is 46 -55%. Two sequence motifs were found in the Conserved Domain Database (32): flavin mononucleotide (FMN) binding and pyridoxamine 5Ј-phosphate oxidase.
The DALI server confirmed these results by identifying other proteins with similar ␤-barrel fold as drNimA, which are all FMN-binding proteins. They include the human pyridoxine 5Ј-phosphate oxidase (PNPO) (Protein Data Bank (PDB) code 1nrg, Z-score 4.0, root mean square deviation 2.15 Å over 97 C␣ atoms), the FMN-binding protein from Desulfovibrio vulgaris (PDB code 1axj, Z-score 8.2, r.m.s.d. 2.3 Å over 83 C␣ atoms), and ferric reductase from Archaeoglobus fulgidus (PDB code 1i0r, Z-score 5.3, r.m.s.d. 2.5 Å for 83 C␣ atoms). Comparison of drNimA with a monomer of these structures shows that drNimA contains some unique structural elements, namely the orientation of helix ␣2 and ␣3 and the extension of ␣4, ␣5, and ␣6 (compare Fig. 3, a and b, which are in the same orientation). In drNimA, the helices ␣2 and ␣3 are involved in the pyruvate binding site, whereas the strand extensions stabilize the homodimer (Fig. 3a). Superposition of these enzymes reveals that the location of the active sites are all on the same side of the barrel, and this region in drNimA contains His-71, which appears to be important for enzymatic activity of the Nim enzymes.
Interestingly, the PNPO structure displays a similar dimer organization (33) to that observed in drNimA. Although the sequence identity between the two enzymes is low (15% identity and 44% homology), the ␤-barrel folds and location of the barrels in the dimers are very similar, as shown in Fig. 3, c and d, of the drNimA and PNPO dimers. Further, the MTR binding site (as described under "Complex Structures") overlaps with the isoalloxazine rings of the FMN (compare Fig. 3, a-d).
Thus it appears that drNimA shares a structural scaffold with a broad family of ␤-barrel-containing enzymes in which the ␤-barrel fold seems important for the electron transfer abilities of these enzymes. Because drNimA has this fold and His-71 overlaps with the active sites of the other enzymes, this supports the hypothesis that drNimA is a reductase in which His-71 is important. Remarkably, the two active sites of both drNimA and PNPO are composed of residues from both monomers and are located at the same place at the dimer interfaces.
Complex Structures-To gain further insight into the antibiotic binding site of the Nim enzymes, drNimA crystals were soaked to obtain substrate-bound states of the structure, and here three structures are presented. Refinement statistics for all structures discussed are summarized in Table II. In the second structure (drNimA-MTR; 2-h soak with MTR), a MTR molecule could clearly be distinguished sitting in a cleft at the monomer-monomer interface, with a water accessible surface of 42 Å 2 . The antibiotic is sandwiched between Pro-56 and Tyr-111Ј (Fig. 3f), two residues that are highly conserved FIG. 3. a, ribbon diagram of the drNimA monomer with the active site His-71, Pro-56, Tyr-111, the pyruvate, and the antibiotic metronidazole, all depicted as ball-and-stick atoms. The structure is color-coded from blue to red when going from the amino terminus to the carboxyl terminus. Secondary structure elements are labeled corresponding to the sequence alignment (Fig. 1). The acetate ion inside the barrel is also included in in the sequence alignment (Fig. 2). The aromatic ring of Tyr-111Ј and the imidazole ring of MTR are almost parallel, separated by 3.2 Å (between the plane of the two rings; see Fig. 3f), indicating that -orbital interactions between the two rings systems are important. Although Pro-56 is strictly conserved, Tyr-111 is substituted in some sequences by a phenylalanine, suggesting that an aromatic ring at position 111 is required to interact with the imidazole ring of the drug.
The structural comparison of drNimA and drNimA-MTR shows several differences between the two structures. Both the pyruvate and the MTR are now found at the interface of the ␤-barrels, sandwiched between the two protein monomers, and interact with residues from both monomers. The O-1 atom in the carboxyl group of the pyruvate is moved slightly closer toward His-71 N⑀-2 (compare Fig. 6, a and b). The binding of MTR results in other minor structural changes in and around the active sites: a water molecule (W-75) is displaced, Pro-56 and Tyr-111Ј are moved further away from each other, and residues Asn-20 to Ser-23, Ser-113, and Ser-171 to Asn-175 are displaced. The RMS displacements between drNimA and drNimA-MTR (main chain residues 2-195) are 0.269 Å 2 .
In the third structure (drNimA-Pyr; 20-h soak with TNZ), no antibiotic could be found bound to the protein; however, we could clearly distinguish the pyruvate molecule covalently attached to His-71 N⑀-2. In the finally refined structure, a well defined His-71-Pyr residue was therefore included, with a bond the figure. b, shown is the PNPO monomer labeled with secondary structure elements according to Musayev et al. (33), and the flavin (FMN) and pyridoxal-5Ј-phosphate (PLP) are also depicted. The orientation and color coding are the same as for the drNimA in panel a). Shown also are the drNimA (c) and the PNPO dimers (d). The active residues are shown in both panels, which are in the same orientations and have same sizes. The location of the two ␤-barrels and the two active sites are indicated with arrows and boxes. e, the drNimA dimer folded as a ␤-propeller as shown by the magenta and green monomers. Residues His-71, Pro-56, Tyr-111, the pyruvate, and the antibiotic metronidazole are all depicted as ball-and-stick atoms for the two active sites. f, a close view down one of the antibiotic binding sites in the drNimA dimer. Residues involved are color-coded in magenta and green according to the monomer of origin. The pyruvate and the antibiotic metronidazole are shown in white and some hydrogen bonds are given. All figures were made with MOLSCRIPT (39) and Raster-3D (40) .   FIG. 4. a, the elution profile of the gel filtration run (Superdex 200) of drNimA in 50 mM Hepes, pH 7.5, and 1 mm EDTA. The elution volume after injection was 14.56 ml for drNimA, whereas the molecular mass markers came off at 13.5 ml for albumin (67 kDa), 14.6 ml for ovalbumin (43 kDa), and 16.9 ml for chymotrypsin (25 kDa). The elution peak for chymotrypsin was well separated from drNimA when the proteins were run together (data not shown). b, a 12% SDS-PAGE gel with Coomassie Blue staining. Lane 1 indicates the sample used without cross-linker. Lanes 2 and 3 are drNimA after adding 0.25 and 0.50 mM cross-linker. Lane 4 indicates the molecular mass markers. We could clearly see that the main ϳ24-kDa band without cross-linker (lane 1) became weaker when cross-linker was added, and a second band at ϳ45 kDa appeared (lanes 2 and 3), which fit well with the molecular mass of a drNimA homodimer. Together these experiments clearly show that the drNimA protein is a homodimer in solution.
In the fourth structure (drNimA-Lac; 23-h soak with MTR), the modeling of a His-71-Pyr residue left a significant peak of difference electron density (F o Ϫ F c ). After consideration of the chemical environment and shape of the density, (most of) this electron density could be described when a His-71-Lac residue was used in the refinement (Fig. 6d). However, the final refinement showed some positive F o Ϫ F c density overlapping with the His-71-Pyr residue in drNimA-Pyr (Fig. 6d). It was therefore concluded that the full reduction of pyruvate into lactate was not achieved. Finally, we chose to include only a His-71-Lac residue (with full occupancy) in the drNimA-Lac structure, because that was the major conformation of this residue.

DISCUSSION
The 5-Ni-based drug, MTR, is still one of the most effective drugs against infections caused by anaerobic bacteria, particularly when treating those caused by e.g. T. vaginalis and Bacteroides (1,2,34). 5-Ni drugs are inactive prodrugs that enter the cell by simple diffusion and are thereafter reduced in a 1-electron reduction by ferredoxin into its toxic radical anion (R-N⅐O 2 Ϫ ) (6) as shown in Fig. 1. Resistance toward these compounds has been encountered in several bacteria, based on an altered or deficient reduction mechanism by which the prodrug could not be transformed into the toxic radical (13,15,16). In B. fragilis-resistant strains, the resistance mechanism appears to be related to a specific family of genes named Nim. Nim family proteins are thought to be 5-nitroimidazole reductases that transform the nitro group of the prodrug into its non-toxic amine derivative (22).
The structure of drNimA from D. radiodurans presented here is the first available structure of a Nim family protein. The dimeric structure has the same basic ␤-barrel scaffold used by other enzymes involved in electron transfer (33,(35)(36)(37) but displays a simpler active site in which the antibiotic and the pyruvate interact with residues from both monomers.
Putative Mechanism and Action of Pyruvate-Consistent with previous evidence (22), the four structures presented allow us to propose how drNimA might function as a reductase and reveal the essential role played by pyruvate in the reaction mechanism. In the native state of the protein, a pyruvate ion was found within hydrogen-binding distance to the conserved His-71. Upon binding of MTR, the pyruvate moiety initially became more tightly bound to His-71, and finally a modified His-71-Pyr residue forms the oxidation product of His and pyruvate with the release of two electrons and one H ϩ (Fig. 7, Step ❶). The electrons can be shuttled via a water molecule (e.g. W-1) to the antibiotic. Thus, the first step of the antibiotic reduction becomes R-NO 2 3 RNϭO (Fig. 7, Step ❷), similar to other bacterial nitro reductases (38). Importantly, this 2-electron reduction avoids formation of the toxic nitro radical anion (Fig. 1) and leads to resistance against the antibiotic.
In this proposed mechanism, the drNimA-MTR structure is an intermediate structure, which appears in between the native state and the drNimA-Pyr somewhere along Step ❶ (Fig.  7). The reduction of pyruvate into lactate, from the drNimA-Pyr structure to the drNimA-Lac structure, may provide an idea as to how the enzyme is recycled back to its native state. To reduce pyruvate into lactate, 2H ϩ and 2e Ϫ are needed. In vivo, putative electron sources are available for the Nim enzymes (e.g. NAD(P)H), but in the drNimA-Lac crystal structures, water is one possible electron donor (H 2 0 3 1 ⁄2O 2 ϩ 2H ϩ ϩ 2e Ϫ ) that could explain how pyruvate was reduced to lactate. To recycle the pyruvate/lactate cofactor, the lactate to the His-71 bond must be broken, but our crystallographic study does not give any details on this process.
The refined structures of drNimA show that His-71 and pyruvate can be oxidized into a His-71-Pyr residue (and release H ϩ ϩ 2e), and then the His-71-Pyr residue can further be reduced into His-71-Lac, as in the drNimA-Lac structure, by use of 2H ϩ and 2e Ϫ . The enzyme can therefore perform both oxidation and reduction of the cofactor pyruvate, and indeed both are needed to function as an enzyme with consecutive rounds of catalysis. Pyruvate is used (e.g. in the PFOR complex) and can be synthesized by anaerobic bacteria (9) and should therefore be available for use by the Nim enzymes. Our proposed mechanism might therefore occur in the anaerobic bacteria for which the 5-Ni drugs are the main drug targets.
Consequences for Antibiotic Resistance in Other Bacteria-During this study, other Nim homologues were found in several other bacteria, including S. typhi, Clostridium tetani, S. avermitilis, H. hepaticus, and the Archaea Methanosarcina sp. (see Fig. 2). Although the exact function of these proteins in their organisms is not known, the sequence alignment strongly suggests that they possess the same structural scaffold as drNimA (Fig. 2). Nim gene-based resistance has so far only been encountered in the Bacteroides (17)(18)(19)(20). Nevertheless, the mobile nature of Nim genes, localized on mobilizable plasmids (17)(18)(19)(20), poses a real threat of 5-Ni drug resistance spreading among bacteria occupying close niches. Bacteroides and Helicobacter are indeed exposed to an increasing selection pressure toward resistance against 5-Ni drugs, and gene transfer from these species is not theoretically impossible.
In this crystallographic study, we shed light on some of the structural aspects of a novel reductase mechanism that could confer bacterial resistance against the 5-nitroimidazole-based drugs that are used daily by humans and animals against anaerobic bacterial infections. Although conformation of the proposed mechanism awaits further biochemical and structure studies, the structure of drNimA reported here should provide a first step toward the design of new or modified 5-Ni drugs. a ⌺ h ͉͉F obs ͉ Ϫ ͉F calc ʈ / ⌺ h ͉F obs ͉, where ͉F obs ͉ and ͉F calc ͉ are observed and calculated structure factor amplitudes for all reflections (R-factor) and the reflections applied in the test R free set (reflection not used in the structure refinement), respectively.

FIG. 7. Proposed antibiotic resistance mechanism.
Step ❶, this is from the native drNimA structure to the covalently bound pyruvate structure (drNimA-Pyr), an oxidation of His-71 and pyruvate into a His-71-Pyr residue, a reaction that releases 2e Ϫ and H ϩ .
Step ❷, the released electrons can further be used to reduce the antibiotic. Because the antibiotic gets 2e Ϫ , it prevents formation of the toxic bactericidal radical R-N⅐O 2 Ϫ as given in Fig. 1. Our drNimA-MTR structure seems to be an intermediate, which is located somewhere along Step ❶ in between the native drNimA and the drNimA-Pyr complex.