Rifamycin antibiotic resistance by ADP-ribosylation:Structure and diversity of ArrJennifer Baysarowich*, Kalinka Koteva*, Donald W. Hughes†, Linda Ejim*, Emma Griffiths*, Kun Zhang*,Murray Junop*, and Gerard D. Wright*‡

*M.G. DeGroote Institute for Infectious Disease Research, Department of Biochemistry and Biomedical Sciences, McMaster University,Hamilton, ON, Canada L8N 3Z5; and †Department of Chemistry, McMaster University, Hamilton, ON, Canada L8S 4M1

Edited by Christopher T. Walsh, Harvard Medical School, Boston, MA, and approved January 28, 2008 (received for review December 19, 2007)

The rifamycin antibiotic rifampin is important for the treatment oftuberculosis and infections caused by multidrug-resistant Staphy-lococcus aureus. Recent iterations of the rifampin core structurehave resulted in new drugs and drug candidates for the treatmentof a much broader range of infectious diseases. This expanded useof rifamycin antibiotics has the potential to select for increasedresistance. One poorly characterized mechanism of resistance isthrough Arr enzymes that catalyze ADP-ribosylation of rifamycins.We find that genes encoding predicted Arr enzymes are widelydistributed in the genomes of pathogenic and nonpathogenicbacteria. Biochemical analysis of three representative Arr enzymesfrom environmental and pathogenic bacterial sources shows thatthese have equally efficient drug resistance capacity in vitro and invivo. The 3D structure of one of these orthologues from Mycobac-terium smegmatis was determined and reveals structural homol-ogy with ADP-ribosyltransferases important in eukaryotic biology,including poly(ADP-ribose) polymerases (PARPs) and bacterial tox-ins, despite no significant amino acid sequence homology withthese proteins. This work highlights the extent of the rifamycinresistome in microbial genera with the potential to negativelyimpact the expanded use of this class of antibiotic.

crystallography � resistome � rifampin � ribosyltransferase

The successful treatment of infectious diseases is under con-stant threat as a result of the small number of new antibiotics

being brought to the clinic and the relentless pressure of evolvingresistance to antibiotics (1). Antibiotic resistance manifests itselfthrough a number of mechanisms, including drug efflux andtarget alteration; however, it is the exquisite power and speci-ficity of enzymatic drug inactivation that is perhaps the apogeeof bacterial response to antibiotics (2). These enzymes use anumber of strategies to overcome the toxic properties of anti-biotics, including destruction of the antibiotic for example byhydrolysis, as well as drug modification by kinases, acyltrans-ferases, adenylyltransferases, and glycosyltransferases. To thisimpressive arsenal of antibiotic countermeasures, research byDabbs and colleagues has added ADP-ribosylation (3–10).

ADP-ribosylation is widely recognized as an important mech-anism of protein posttranslational modification with numerousbiological outputs ranging from modulation of microbial viru-lence to influencing eukaryotic cell-cycle progression (11).These bacterial and eukaryotic protein (ADP-ribosyl)trans-ferases (ARTs) share similar substrates (all proteins), catalyticmechanisms, and general overall 3D fold (12, 13). In all cases, theADP-ribosyl donor is NAD�, and enzyme action transfersthe ADP-ribose unit to a susceptible amino acid residue on thetarget protein with loss of nicotinamide.

The 3D structures of several protein ADP-ribosyltransferaseshave been reported including chicken poly(ADP-ribose) poly-merase (PARP)-1 (14), rat ART2.2 (15), mosquitocidal toxinfrom Bacillus sphaericus (16), diphtheria toxin (17), cholera toxin(18), domain III of Pseudomonas aeruginosa exotoxin A (19),iota-toxin from Clostridium perfringens (20), pertussis toxin (21),C2-toxin (22), and C3-toxin (23) from Clostridium botulinum.

Although well established as a protein modification mecha-nism of significant importance, small molecule ADP-ribosylation is unknown with one exception: a reported ADP-ribosylation of the antibiotic rifampin that results in resistance inbacterial pathogens (5, 6, 9, 10). Rifampin is a semisyntheticderivative of the natural product rifamycin B derived over 40years ago from Amycolatopsis mediterranei (Fig. 1) (24). Rifa-mycins inhibit the �-subunit of bacterial DNA-dependent RNApolymerase (RpoB) by obstructing the exit tunnel for nascentRNA (25). Rifampin is used primarily in the combination drugtreatment of infections caused by Mycobacterium tuberculosisand is a front line drug in control of this disease (26). Increas-ingly, rifamycins also find use for the treatment of drug-resistantStaphylococcus aureus (27).

Consistent with the lack of evidence for plasmid-mediatedhorizontal gene transfer in mycobacteria, rifampin resistance inM. tuberculosis, including extensively drug-resistant tuberculosis,results from point mutations in rpoB that lower affinity for thedrug (28). In the opportunistic pathogen Mycobacterium smeg-matis, however, rifampin is an ineffective drug because of thepresence of a chromosomally encoded rifampin ADP-ribosyltransferase (Arr-ms). Arr-ms is a small enzyme (�16kDa) with no sequence similarity to known protein ADP-ribosyltransferases such as ART, PARP, and bacterial toxins (6,29). Paradoxically, a homologue of Arr, Arr-2 (55% identicalwith Arr-ms), also is present in several transposons and integronsfound in Gram-negative pathogenic bacteria such as P. aerugi-nosa (30), Escherichia coli (31), Klebsiella pneumoniae (32), andAcinetobacter baumannii (33). This distribution of arr-2 in Gram-negative pathogens is unexpected because rifampin is not usedto treat infections caused by these bacteria, and therefore theselective pressure for maintaining this resistance cassette wouldbe predicted to be low unless the patient was receiving rifampintreatment for other infections and as a result this has selected for‘‘bystander’’ resistance in Gram-negative bacteria.

The rifamycin class of antibiotics has been revisited in recentyears as a scaffold worthy of investigation for the developmentof new drugs (Fig. 1). Rifaximin, a semisynthetic rifamycin, wasapproved in 2003 for prophylactic use in the treatment oftraveler’s diarrhea (34). Rifalazil, an analogue with antichlamy-

Author contributions: G.D.W. designed research; J.B., K.K., D.W.H., L.E., E.G., K.Z., and M.J.performed research; J.B., K.K., D.W.H., M.J., and G.D.W. analyzed data; and J.B., M.J., andG.D.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The crystal structure has been deposited in the Protein Data Bank,www.pdb.org (PDB ID code 2HW2).

‡To whom correspondence should be addressed at: M. G. DeGroote Institute for InfectiousDisease Research, Department of Biochemistry and Biomedical Sciences, McMasterUniversity, 1200 Main Street West, Hamilton, ON, Canada L8N 3Z5. E-mail: wrightge@mcmaster.ca.

This article contains supporting information online at www.pnas.org/cgi/content/full/0711939105/DC1.

© 2008 by The National Academy of Sciences of the USA

4886–4891 � PNAS � March 25, 2008 � vol. 105 � no. 12 www.pnas.org�cgi�doi�10.1073�pnas.0711939105

dial activity (35), is in Phase III trials. This rediscovery of therifamycin antibiotics continues apace, and new semisyntheticrifamycin derivatives are in various stages of development (e.g.,refs. 36–38).

With the growing importance of rifamycin antibiotics inantimicrobial therapy, there is opportunity for the selection ofpreexisting resistance elements such as Arr-2 and for opportu-nistic pathogens with chromosomal Arr-like elements and theirsubsequent mobilization and transfer to human pathogens. Here,we report the characterization of the enzymatic efficacy ofseveral Arr orthologues and their activity with a panel ofrifamycins and the 3D structure of Arr-ms from M. smegmatis incomplex with rifampin. The structure reveals similarity to pro-tein ADP-ribosyltransferases, despite a lack of amino acidsequence similarity, and provides another example with proteinkinases (39) and protein acetyltransferases (40, 41) of thereengineering of important protein modifying enzyme scaffoldsfamiliar to eukaryotic biology in antibiotic resistance.

ResultsDiversity and Distribution of arr in the Resistome. Using the Arr-2protein sequence as a query, we identified a number of possibleArr orthologues in microbial genome and metagenome data-bases. All of these similar proteins were predicted to be 15–18kDa and showed �55% sequence similarity. Sequence alignmentrevealed a broad distribution of predicted Arr orthologuesthroughout microbial genera [supporting information (SI) Fig. 5in SI Appendix]. Many, but not all, of these arr-like genes areeither on mobile genetic elements (e.g., arr-2 in Gram-negativepathogens) or are clustered in the genome with gene sequences(e.g., predicted transposase encoding genes) that suggest historicacquisition by horizontal gene transfer (SI Fig. 6 in SI Appendix).

The sequence search identified arr homologues in environ-mental bacteria, including opportunistic pathogens such as M.smegmatis, Stenotrophomonas maltophilia, and Burkholderiacenocepacia, as well as a number of nonpathogenic bacteria,including the actinomycetes Streptomyces coelicolor and Coryne-bacterium glutamicum, the anaerobic firmicute Desulfitobacte-rium halfniense, and the �-proteobacterium Rhodopseudomonaspalustris. Furthermore, the number of complete and partialarr-like sequences derived from environmental metagenomestudies from marine and terrestrial samples speaks to the

expansive distribution of Arr-like proteins in the environmenteven in the absence of obvious human pressure.

Arr-like genes therefore are widespread in a diversity ofbacterial genomes. Homology, however, is not equivalent withorthology. We selected three representative genes for furtherstudy of their associated enzymatic activities: arr-2 from P.aeruginosa, which is identical to the gene found on severaltransposons and integrons isolated from a variety of clinicalisolates of Gram-negative pathogens; arr-ms, from the genomeof the opportunistic pathogen M. smegmatis; and arr-sc, found inthe genome of the benign soil bacterium S. coelicolor.

Characterization of Arr Enzymatic Activity. Each of the predictedArr orthologues was successfully expressed and purified from E.coli (SI Fig. 7 in SI Appendix). Using an HPLC-based enzymeassay with mass spectrometric detection, we confirmed that allthree enzymes efficiently catalyzed ADP-ribosylation of ri-fampin in the presence of NAD� as a cosubstrate (Fig. 2).NADH and NADP� were not substrates for the enzymes.

Using this HPLC assay, we compared the steady-state kineticparameters of the enzymes (Table 1). Remarkably, regardless ofsource, all of the enzymes displayed very similar kinetic behav-iors and constants and are equally effective antibiotic resistanceenzymes. The kcat/Km ratios measure the efficiency of theenzyme to catalyze the reaction at low substrate concentrations;these were essentially identical, varying only 3-fold. This ratio ismost important as the antibiotic begins to enter the cell andtherefore serves as an indicator of effectiveness as a druginactivation mechanism (42).

The Arr-2 enzyme efficiently modified rifamycin antibioticswith diverse naphthyl substituents, including the parent naturalproduct rifamycin SV (no substituent) and the semisyntheticdrugs rifampin (antituberculosis drug), rifabutin (for treatment

Fig. 1. Structures of rifamycin antibiotics of clinical importance. The rifa-mycin antibiotics consist of an aromatic naphthyl core and a polyketidederived ansa-bridge.

Fig. 2. Arr-2 catalyzes rifampin ADP-ribosylation. HPLC analysis of rifampinADP-ribosylation by Arr-2.

Baysarowich et al. PNAS � March 25, 2008 � vol. 105 � no. 12 � 4887

MIC

ROBI

OLO

GY

of Mycobacterium avium complex), and rifaximin (prophylactictreatment of traveler’s diarrhea) (SI Table 4 in SI Appendix; seeFig. 1 for chemical structures). Arr therefore has a broad abilityto inactivate both ‘‘old’’ and new rifamycin antibiotics.

The in vivo efficiency of this inactivation was assessed bydetermination of the minimal inhibitory concentrations (MICs)of rifamycin antibiotics required to arrest cell growth. In thepresence of recombinant arr genes in E. coli, MIC values of allof the tested antibiotics were greatly increased, and underconditions where the genes were induced, MICs all increased to�512 �g/ml (Table 2). For comparison, MIC of wild-type M.smegmatis expressing arr-ms is 32 �g/ml (27), well above break-point concentrations of this drug (�2 �g/ml). Therefore, Arr isa potent rifamycin resistance element with broad antibioticsubstrate specificity in vitro and in vivo.

The structure of the ADP-ribosyl-rifampin product was de-termined by NMR and mass spectrometric analysis of thepurified compound. Mass spectra confirmed the predicted mo-lecular mass of the product (SI Table 4 in SI Appendix). Througha series of 1D and 2D NMR experiments, the regiospecificity ofmodification was determined to be on the hydroxyl linked to C23in the ansa-bridge (see SI Experimental Procedures in SI Appen-dix). This result is consistent with the findings of a previousreport of Arr-ms activity (5). We further determined the abso-lute conformation of the ADP-ribosylated product and foundthat the reaction occurs with inversion of configuration at C1 ofthe ribose of NAD� (�3 �).

Crystal Structure of Rifampin-Bound Arr-ms. One of the intriguingfeatures of the rifampin-inactivating Arr family of ARTs is theirsmall size (�16 kDa). Previously characterized protein modify-ing ARTs are much larger, typically �40 kDa. In addition, theArr family shares no obvious amino acid sequence conservationwith any of the dozens of known or proposed ARTs.

We determined the structure of Arr-ms in complex withrifampin to 1.45 Å (PDB ID code 2HW2). The final modelcontained no disordered regions and was refined to R and Rfree

values of 13.4 and 18.5, respectively. SI Fig. 8 in SI Appendixprovides an Fo � Fc omit map of bound rifampin. Complete datacollection and model refinement statistics are listed in SI Table9 in SI Appendix.

Arr-ms contains a single domain comprised of two antiparallel�-sheets and two �-helices situated to one side, forming a secondlayer in the overall �� fold (Fig. 3). The first �-sheet, encom-passing four strands (�1-4-7-2) is arranged almost perpendicularalong the edge of the sheet to the second three-stranded �-sheet(�3-6-5). Helix 1 (�1) is situated against one side of the extended�-sheet region, making closer contact with the first four-stranded �-sheet. Together, �1 and �2 form the walls of a deep

Table 1. Steady-state kinetic parameters of Arr enzymes

Enzyme Substrate Km, mM KI, mM kcat, s�1 kcat/Km, s�1�M�1

Arr-ms Rifampin 1.42 � 0.38 1.07 � 0.35 2.84 � 0.57 2.00 � 103

NAD� 0.18 � 0.04 1.58 � 104

Arr-sc Rifampin 2.80 � 0.80 0.20 � 0.07 7.34 � 1.90 2.62 � 103

NAD� 0.31 � 0.11 2.37 � 104

Arr-2 Rifampin 0.74 � 0.34 0.55 � 0.30 4.48 � 1.49 6.05 � 103

NAD� 0.18 � 0.05 2.49 � 104

Table 2. MICs of rifamycins against E. coli-harboringArr orthologues

Arr and PlasmidVector

MIC, �g/ml

IPTG Rifabutin RifampinRifamycin

SV Rifaximin

pET28a(�) � 8 4 64 8pET28a(�) � 8 8 64 16pET28a(�)-Arr-2 � �512 512 �512 �512pET28a(�)-Arr-2 � �512 �512 �512 �512pET22b(�) � 16 8 32 16pET22b(�) � 16 8 64 16pET22b(�)-Arr-ms � 64 128 512 �512pET22b(�)-Arr-ms � �512 �512 �512 �512pET15b(�) � 16 8 32 16pET15b(�) � 16 8 32 16pET15b(�)-Arr-sc � 64 128 512 �512pET15b(�)-Arr-sc � 512 512 �512 512

IPTG, isopropyl �-D-thiogalactopyranoside.Fig. 3. Structure of rifampin-bound Arr-ms. Orthogonal views of Arr-ms incomplex with rifampin (yellow) and the dipeptide glycylglycine (purple).

4888 � www.pnas.org�cgi�doi�10.1073�pnas.0711939105 Baysarowich et al.

rifampin-binding cleft. Helix 2 is present at the C terminus ofArr-ms and participates in an overall wrapping of the C-terminalregion around the core domain of Arr-ms. Given the limitedcontact that �2 makes with the rest of the protein, it is likely thatthis helix would be unstructured in the absence of rifampin.Rifampin makes significant contact not only with �1 and �2 butalso with the extended loop region connecting �5 and �6. This�5–�6 loop serves as a lid, partially covering rifampin oncepositioned into its deep binding cleft. The most remarkablefeature of the interaction between rifampin and Arr-ms is thatdespite burying 620 Å2 of surface area, not one H bond or ionicinteraction exists between rifampin and a side chain of Arr-ms.Two H bonds are formed with main chain atoms from N86 andK91, and an additional six water-mediated interactions also arepresent. The remaining direct interactions are entirely hydro-phobic in nature. SI Table 10 in SI Appendix gives a complete listof rifampin–Arr-ms interactions. The apparent plasticity ofinteractions observed between rifampin and Arr-ms suggest thatArr-ms may be able to act on other substrates in addition torifampin.

Comparison of Arr-ms with Other ADP-Ribosyl Transferases. Repre-sentative crystal structures of ART family members from botheukaryotes (PARPs and ARTs), and prokaryotes (ADP-ribosylating toxins) have been determined (reviewed in refs. 12,13, and 43). ARTs can be classified into two subfamilies basedon 3D structure: class I, characterized by PARP-1 and diphtheriatoxin structures, and class II, characterized by ART2 and VIP2toxin structures. Each class builds on a common NAD�-bindingcore of 5 �-strands arranged as two adjoining sheets. Distin-guishing class features include the presence of distinct motifs(H-Y-E or R-S-E) and the addition of a variable number ofconserved �-helices above and below the core �-sheet structure.

Comparative structural analysis using the program Dali (44)revealed limited but significant similarity with three previouslydetermined structures: PARP catalytic fragment from chicken(Gallus gallus; PDB ID code 1A26), exotoxin A domain III fromP. aeruginosa (PDB ID code 1AER), and a putative ART fromthe archaeon Aeropyrum pernix (1WFX). Only a small amount ofsequence (�80 residues) within these proteins could be alignedwith Z scores ranging from 3 to 5 and rmsd values �3 Å.Nevertheless, as shown in Fig. 4, Arr-ms could be structurallyaligned with the conserved core region from other ARTs.Comparison of topology (SI Fig. 9 in SI Appendix) further

suggested that Arr-ms belongs to the class I subfamily of ARTs.As demonstrated in SI Fig. 9 in SI Appendix, Arr-ms representsa minimal ART unit, containing only the essential elementsnecessary for NAD� binding and catalysis. For simplicity, wehereafter refer to the Arr-ms labeling of topology in makingcomparison to other ART family members.

Class I ARTs contain a highly conserved H-Y-E motif;however, standard sequence alignment algorithms fail to revealthe presence of such a motif in Arr-ms. Using a structure-basedsequence alignment (SI Fig. 10 in SI Appendix), we identified tworesidues (H19 in �1 and Y49 in �3) structurally conserved acrossArr and remaining class 1 ARTs. H19 and Y49 make up part ofthe so-called H-Y-E motif (43) and are predicted to contributedirectly to NAD� binding. We mutated both His-19 and Y49 toAla, and these showed no in vitro activity and reduced MIC levelsof rifampin to control levels of susceptibility (Table 3), consistentwith a vital role in enzyme activity.

Previous studies have shown that only a single residue, thecatalytic glutamic acid present on the front edge of �6, isabsolutely conserved in all structures of known functional ARTs.This residue is essential for stabilizing the predicted oxocarbe-nium ion in the transition state (13, 45). Remarkably, this residueis missing from Arr-ms. Instead an aspartic acid residue (Asp-84)located within the substrate-binding loop connecting the lessconserved fifth �-strand to �6, appears to fulfill this function inArr-ms. There is no obvious structural need for this to haveoccurred; however, positioning the key catalytic residue in thesubstrate-binding loop would be an effective way of preventingformation of a functional active site before substrate binding.Mutation of Asp-84 to Ala abolished both in vitro ADP-ribosylation and in vivo rifampin resistance as predicted if thisresidue fulfills the role of the invariant Glu in protein ARTs.

DiscussionADP-ribosylation of rifampin in M. smegmatis cell-free extract(3, 5) and by purified Arr-2 (this work) occurs on the hydroxylgroup at C23 of the antibiotic and with inversion of configurationat C1 of the NAD�-derived ribose. The C23 hydroxyl groupinteracts with the amide backbone of Phe-394 in the crystalstructure of rifampin bound to the Thermus aquaticus RNApolymerase (25), providing the molecular basis for Arr-mediatedresistance. The substrate specificity of Arr spans old antibioticssuch as rifampin and the natural product rifamycin SV as well asnew agents such as rifaximin. The location of modification on theansa-bridge of the antibiotic at position 23 and the tolerance ofsubstitutions on the naphthyl ring indicate that it is very likelythat several new rifamycin derivatives being considered forclinical use also will be susceptible to Arr-catalyzed inactivation.

Substrate binding of ARTs is perhaps the least understoodaspect of ART function. For class I ARTs, the loop connecting�5 and �6 has been suggested to function in substrate recogni-tion. Our structure of rifampin-bound Arr-ms provides struc-tural evidence to support this claim because the �5–�6 loop ofArr-ms does indeed form the lid to the rifampin-binding pocket.A significant difference between Arr and other ARTs is thepresence of the C-terminal helix �2, which serves as a platformfor rifampin binding and physically blocks access to the active site

Table 3. In vitro rifampin susceptibility of Arr-ms mutants

Construct in E. coli MIC, �g/ml

pET22b(�) (negative control) 16Arr-ms (positive control) 256His-19–Ala 4Tyr-49–Ala 4Asp-84–Ala 8

Fig. 4. Structural alignment of the core NAD� binding cleft conservedamong ADP-ribosyl transferase enzymes. The structure of Arr-ms (yellow) incomplex with rifampin is shown aligned in stereo with the structurally con-served NAD� binding cleft of domain III of P. aeruginosa exotoxin (green),PARP catalytic fragment (purple), and a probable 2�-RNA phosphotransferase(blue) (PDB ID codes 1AER, 1A26, and 1WFX, respectively). The structure ofdomain III of P. aeruginosa exotoxin was solved in complex with �-methylene-thiazole-4-carboxamide adenine dinucleotide shown here in light blue tohighlight the NAD� binding site.

Baysarowich et al. PNAS � March 25, 2008 � vol. 105 � no. 12 � 4889

MIC

ROBI

OLO

GY

by large substrates such as proteins. As noted above, none of thedirect contacts mediated by Arr-ms side chains involves a H bondto rifampin. The functional significance of the highly conservedlysine residues (K90 and K91) within the substrate-binding loopis not apparent. These residues may be important for Arr activitydirected toward other uncharacterized substrates. The absenceof direct interactions of antibiotic with the enzyme, the broadsubstrate specificity, and the similarity of kinetic specificityconstants between enzymes suggests that Arr enzymes likelyhave other functions in the cell, possibly as small molecule ARTs,and that rifamycin antibiotic resistance is a secondary or fortu-itous activity, which would be consistent with unexpected broaddistribution of arr in microbial genomes in the absence ofobvious antibiotic selection.

The structure of Arr-ms, along with site-directed mutagenesisstudies and literature precedent for other ARTs, is consistentwith a molecular mechanism where Asp-84 stabilizes a devel-oping oxocarbenium transition state enabling attack of thehydroxyl group at position 23 of the antibiotic to attack C1 of theribose (Scheme 1). This action would result in inversion ofstereochemistry in the ADP-ribosylated product at this position,which we confirmed experimentally. This mechanism predictsthat transition state inhibitors of ARTs that take advantage ofthe predicted structure and electronic configuration of thetransition state, e.g., ref. 46, could find use as inhibitors of Arrand perhaps overcome resistance in vivo.

Recently, new rifamycins with carbamate derivatization atposition 25 have been reported that maintain antibiotic activityin the presence of Arr (38). The crystal structure of rifampinbound to Arr-ms provides the logic for evasion of inactivation bythese new compounds because the bulky carbamate derivativesreported would be unable to bind productively to the enzyme,thereby avoiding modification. Additional modifications of theansa-bridge at adjacent positions should prove equally effectivein avoiding Arr-mediated resistance.

Significance. The Arr family of proteins is widely distributed inmicrobial communities. Although this mechanism is not yethighly prevalent in pathogenic bacteria, the growing number ofnew rifamycin antibiotics either already in the clinic or indevelopment for the treatment of nontuberculosis infectionsprovides the perfect opportunity to select for Arr-mediatedresistance. This study of the distribution, activity, structure, andmechanism of Arr provides an early warning that enzymaticresistance to rifamycin antibiotics is prevalent and mobilizable.The 3D structure and function studies presented here providethe molecular rationale for strategies to overcome Arr such as

through alteration of the ansa-bridge of the antibiotic (e.g., ref.38) and the development of Arr inhibitors that have the potentialto overcome this threat to the next generations of rifamycinantibiotics. Furthermore, this work provides another example ofthe exploitation of important protein scaffolds by evolutionresulting in antibiotic resistance. As we begin to more fullyappreciate the depth of the resistome, it is likely that furtherexamples of this replication of protein structure and mechanismwill emerge.

MethodsOverexpression and Purification of Arr Enzymes. Arr enzymes were expressedas hexa-His fusions and purified by Ni-affinity and ion exchange chromatog-raphy (see SI Experimental Procedures in SI Appendix for further detail).

Analytical Procedures. Reverse-phase (RP)-HPLC and liquid chromatography/electrospray mass spectrometry (LC/ESI-MS) were used to confirm the presenceof ADP-ribosyl rifampin before assay development. RP-HPLC was performedwith a C18 column (3 �m, 120 Å, 4.6 � 150 mm). The activity of rifampinADP-ribosyl transferase enzymes Arr-ms, Arr-sc, and Arr-2 was analyzed byusing a dual solvent system consisting of 0.05% trifluoroacetic acid (TFA) inwater (vol/vol) and 0.05% TFA in acetonitrile (vol/vol). LC/ESI-MS data wereobtained by using an Agilent 1100 Series HPLC system and an Applied Biosys-tems QTRAP LC/MS/MS system. For LC/ESI-MS analysis, formic acid was used asa counter ion instead of TFA.

Assay Development and Kinetics. Progress curves were generated for each ofthe Arr enzymes at three different substrate concentrations and used todetermine the linear initial rate range of the reaction. Reactions were pre-pared in 100 �l of 50 mM Hepes, pH 7.5, 1 nmol of Arr, and 0.1 mM, 0.5 mM,or 1.0 mM NAD� or rifampin and stopped by the addition of methanol to 50%(vol/vol). Product formation was analyzed by RP-HPLC as described above. Acalculated conversion factor based on the area under the curve (AUC) of afixed concentration of rifampin at 340 nm was used to convert the productAUC into units of molarity.

Steady-state kinetic parameters for rifampin and NAD� were determinedfor the Arr-ms and Arr-2 enzymes by varying substrate at six concentrationsfrom 0.05 to 2.0 mM, and for the Arr-sc enzyme by varying substrate at fiveconcentrations from 0.05 to 1.0 mM. Reactions were prepared in 100 �l of 50mM Hepes, pH 7.5, 1 nmol of Arr, 0.5 mM NAD�, and varied rifampin concen-trations to determine steady-state kinetics for rifampin, and in 100 �l of 50mM Hepes, pH 7.5, 1 nmol of protein, 0.5 mM rifampin, and varied concen-trations of NAD� to determine steady-state kinetics for NAD�. All reactionswere set up in duplicate. Reaction mixes were incubated for 5 min, stopped bythe addition of methanol to 50% (vol/vol), and analyzed by HPLC for productformation as described above.

Arr Structure Determination. Arr-ms was crystallized in the presence of ri-fampin by using hanging drop-vapor diffusion. Single-wavelength anomalousdispersion (SAD) phasing was performed with selenomethionine-derivatized

Scheme 1. Proposed mechanism of Arr-catalyzed rifampin inactivation.

4890 � www.pnas.org�cgi�doi�10.1073�pnas.0711939105 Baysarowich et al.

Arr (see SI Experimental Procedures in SI Appendix for further detail of crystalgrowth, data collection, and model refinement).

Site-Directed Mutagenesis of Arr-ms. Site-specific mutations H19A, Y49A, andD84A were introduced into the arr-ms gene by using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). Oligonucleotide primers are listed in SITable 11 in SI Appendix. The pET22b (�) vectors carrying the amplified genesfor the arr mutants were used to transform E. coli BL21 (DE3) cells. Mutantswere confirmed by DNA sequencing.

MIC Determinations. The MICs were determined by broth dilution in Mueller–Hinton broth (MHB). Colonies of E. coli BL21 (DE3) cells were picked from fresh

overnight plates and mixed with 0.85% saline to an OD625 of 0.08–0.1.Aliquots of 100 �l of the saline solution were mixed into 18.9 ml of MHB. Afterthis, 95 �l of the mixture was added to each well of a sterile 96-well microtiterplate containing serial dilutions of rifampin at concentrations ranging from512 to 0 �g/ml. Each well contained 5 �l of rifampin dilution. The plate wasincubated overnight at 37°C and assessed for bacterial growth.

ACKNOWLEDGMENTS. We thank Dr. D. Boehr and Mr. M Crouch for prepa-ration of plasmids for expression of Arr-ms and Arr-sc and Drs. D. Rowe-Magnus (Sunnybrook Health Sciences Centre, Toronto) and J. Liu (University ofToronto, Toronto) for gifts of plasmids. This work was supported by GrantsMOP-53209 (to M.J.) and MT-13536 (to G.D.W.) from the Canadian Institutesof Health Research and the Canada Research Chairs Program (G.D.W.).

1. Alekshun MN, Levy SB (2007) Molecular mechanisms of antibacterial multidrug resis-tance. Cell 128:1037–1050.

2. Wright GD (2005) Bacterial resistance to antibiotics: Enzymatic degradation andmodification. Adv Drug Deliv Rev 57:1451–1470.

3. Morisaki N, et al. (2000) Structures of ADP-ribosylated rifampicin and its metabolite:Intermediates of rifampicin-ribosylation by Mycobacterium smegmatis DSM43756. JAntibiot (Tokyo) 53:269–275.

4. Imai T, et al. (1999) Identification and characterization of a new intermediate in theribosylative inactivation pathway of rifampin by Mycobacterium smegmatis. MicrobDrug Resist 5:259–264.

5. Quan S, et al. (1999) ADP-ribosylation as an intermediate step in inactivation ofrifampin by a mycobacterial gene. Antimicrob Agents Chemother 43:181–184.

6. Quan S, Venter H, Dabbs ER (1997) Ribosylative inactivation of rifampin by Mycobac-terium smegmatis is a principal contributor to its low susceptibility to this antibiotic.Antimicrob Agents Chemother 41:2456–2460.

7. Andersen SJ, Quan S, Gowan B, Dabbs ER (1997) Monooxygenase-like sequence of aRhodococcus equi gene conferring increased resistance to rifampin by inactivating thisantibiotic. Antimicrob Agents Chemother 41:218–221.

8. Tanaka Y, et al. (1996) Different rifampicin inactivation mechanisms in Nocardia andrelated taxa. Microbiol Immunol 40:1–4.

9. Morisaki N, et al. (1995) Structure determination of ribosylated rifampicin and itsderivative: New inactivated metabolites of rifampicin by mycobacterial strains. JAntibiot (Tokyo) 48:1299–1303.

10. Dabbs ER, et al. (1995) Ribosylation by mycobacterial strains as a new mechanism ofrifampin inactivation. Antimicrob Agents Chemother 39:1007–1009.

11. Walsh CT (2006) Posttranslational Modification of Proteins (Roberts, Englewood, CO),pp 317–330.

12. Yates SP, Jorgensen R, Andersen GR, Merrill AR (2006) Stealth and mimicry by deadlybacterial toxins. Trends Biochem Sci 31:123–133.

13. Holbourn KP, Shone CC, Acharya KR (2006) A family of killer toxins: Exploring themechanism of ADP-ribosylating toxins. FEBS J 273:4579–4593.

14. Ruf A, Rolli V, de Murcia G, Schulz GE (1998) The mechanism of the elongation andbranching reaction of poly(ADP-ribose) polymerase as derived from crystal structuresand mutagenesis. J Mol Biol 278:57–65.

15. Mueller-Dieckmann C, Ritter H, Haag F, Koch-Nolte F, Schulz GE (2002) Structure of theecto-ADP-ribosyl transferase ART2.2 from rat. J Mol Biol 322:687–696.

16. Reinert DJ, Carpusca I, Aktories K, Schulz GE (2006) Structure of the mosquitocidal toxinfrom Bacillus sphaericus. J Mol Biol 357:1226–1236.

17. Bell CE, Eisenberg D (1996) Crystal structure of diphtheria toxin bound to nicotinamideadenine dinucleotide. Biochemistry 35:1137–1149.

18. Zhang RG, et al. (1995) The three-dimensional crystal structure of cholera toxin. J MolBiol 251:563–573.

19. Li M, Dyda F, Benhar I, Pastan I, Davies DR (1996) Crystal structure of the catalyticdomain of Pseudomonas exotoxin A complexed with a nicotinamide adenine dinucle-otide analog: Implications for the activation process and for ADP ribosylation. Proc NatlAcad Sci USA 93:6902–6906.

20. Tsuge H, et al. (2003) Crystal structure and site-directed mutagenesis of enzymaticcomponents from Clostridium perfringens iota-toxin. J Mol Biol 325:471–483.

21. Stein PE, Boodhoo A, Armstrong GD, Cockle SA, Klein MH, Read RJ (1994) The crystalstructure of pertussis toxin. Structure (London) 2:45–57.

22. Schleberger C, Hochmann H, Barth H, Aktories K, Schulz GE (2006) Structure and actionof the binary C2 toxin from Clostridium botulinum. J Mol Biol 364:705–715.

23. Han S, Arvai AS, Clancy SB, Tainer JA (2001) Crystal structure and novel recognitionmotif of rho ADP-ribosylating C3 exoenzyme from Clostridium botulinum: Structuralinsights for recognition specificity and catalysis. J Mol Biol 305:95–107.

24. Floss HG, Yu TW (2005) Rifamycin-mode of action, resistance, and biosynthesis. ChemRev 105:621–632.

25. Campbell EA, et al. (2001) Structural mechanism for rifampicin inhibition of bacterialRNA polymerase. Cell 104:901–912.

26. Van Scoy RE, Wilkowske CJ (1999) Antimycobacterial therapy. Mayo Clin Proc 74:1038–1048.

27. Ellis MW, Lewis JS, II (2005) Treatment approaches for community-acquired methicillin-resistant Staphylococcus aureus infections. Curr Opin Infect Dis 18:496–501.

28. Musser JM (1995) Antimicrobial agent resistance in mycobacteria: molecular geneticinsights. Clin Microbiol Rev 8:496–514.

29. Alexander DC, Jones JR, Liu J (2003) A rifampin-hypersensitive mutant reveals differ-ences between strains of Mycobacterium smegmatis and presence of a novel transpo-son, IS1623. Antimicrob Agents Chemother 47:3208–3213.

30. Tribuddharat C, Fennewald M (1999) Integron-mediated rifampin resistance inPseudomonas aeruginosa. Antimicrob Agents Chemother 43:960–962.

31. Naas T, Mikami Y, Imai T, Poirel L, Nordmann P (2001) Characterization of In53, a class1 plasmid- and composite transposon-located integron of Escherichia coli which carriesan unusual array of gene cassettes. J Bacteriol 183:235–249.

32. Arlet G, Nadjar D, Herrmann JL, Donay JL, Lagrange PH, Philippon A (2001) Plasmid-mediated rifampin resistance encoded by an arr-2-like gene cassette in Klebsiellapneumoniae producing an ACC-1 class C beta-lactamase. Antimicrob Agents Che-mother 45:2971–2972.

33. Houang ET, Chu YW, Lo WS, Chu KY, Cheng AF (2003) Epidemiology of rifampinADP-ribosyltransferase (arr-2) and metallo-beta-lactamase (blaIMP-4) gene cassettesin class 1 integrons in Acinetobacter strains isolated from blood cultures in 1997 to2000. Antimicrob Agents Chemother 47:1382–1390.

34. Huang DB, DuPont HL (2005) Rifaximin—A novel antimicrobial for enteric infections.J Infect 50:97–106.

35. Rothstein DM, Shalish C, Murphy CK, Sternlicht A, Campbell LA (2006) Developmentpotential of rifalazil and other benzoxazinorifamycins. Expert Opin Investig Drugs15:603–623.

36. Murphy CK, Karginova E, Sahm D, Rothstein DM (2007) In Vitro activity of novelrifamycins against Gram-positive clinical isolates. J Antibiot (Tokyo) 60:572–576.

37. Kim IH, et al. (2007) Synthesis and antibacterial evaluation of a novel series ofrifabutin-like spirorifamycins. Bioorg Med Chem Lett 17:1181–1184.

38. Combrink KD, et al. (2007) New C25 carbamate rifamycin derivatives are resistant toinactivation by ADP-ribosyl transferases. Bioorg Med Chem Lett 17:522–526.

39. Hon WC, et al. (1997) Structure of an enzyme required for aminoglycoside resistancereveals homology to eukariotic protein kinases. Cell 89:887–895.

40. Wybenga-Groot L, Draker Ka, Wright GD, Berghuis AM (1999) Crystal structure of anaminoglycoside 6�-N-acetyltransferase:defining the GCN5-related N-acetyltransferasesuperfamily fold. Structure (London) 7:497–507.

41. Wolf E, Vassilev A, Makino Y, Sali A, Nakatani Y, Burley SK (1998) Crystal structure ofa GCN5-related N-acetyltransferase: Serratia marcescens aminoglycoside 3-N-acetyltransferase. Cell 94:439–449.

42. Radika K, Northrop DB (1984) Correlation of antibiotic resistance with Vmax/Km ratio ofenzymatic modification of aminoglycosides by kanamycin acetyltransferase. Antimi-crob Agents Chemother 25:479–482.

43. Otto H, Reche PA, Bazan F, Dittmar K, Haag F, Koch-Nolte F (2005) In silico character-ization of the family of PARP-like poly(ADP-ribosyl)transferases (pARTs). BMC Genom-ics 6:139.

44. Holm L, Sander C (1993) Protein structure comparison by alignment of distancematrices. J Mol Biol 233:123–138.

45. Parikh SL, Schramm VL (2004) Transition state structure for ADP-ribosylation of eu-karyotic elongation factor 2 catalyzed by diphtheria toxin. Biochemistry 43:1204–1212.

46. Zhou GC, et al. (2004) Inhibitors of ADP-ribosylating bacterial toxins based on oxacar-benium ion character at their transition states. J Am Chem Soc 126:5690–5698.

Baysarowich et al. PNAS � March 25, 2008 � vol. 105 � no. 12 � 4891

MIC

ROBI

OLO

GY