Cell, Vol.

104, 901912, March 23, 2001, Copyright 2001 by Cell Press

Structural Mechanism for Rifampicin Inhibition of Bacterial RNA Polymerase

Elizabeth A. Campbell,* Nataliya Korzheva, Arkady Mustaev, Katsuhiko Murakami,* Satish Nair,* Alex Goldfarb, and Seth A. Darst* * The Rockefeller University 1230 York Avenue New York, New York 10021 Public Health Research Institute 455 First Avenue New York, New York 10016 Mutations conferring Rif resistance (RifR) map almost exclusively to the rpoB gene (encoding the RNAP subunit) in every organism tested, including E. coli (Ezekiel and Hutchins, 1968; Wehrli et al., 1968b; Heil and Zillig, 1970) and M. tuberculosis (Ramaswamy and Musser, 1998; Heep et al., 2000). Comprehensive genetic analyses have provided molecular details of amino acid alterations in conferring RifR (Figure 1; Ovchinnikov et al., 1983; Lisitsyn et al., 1984a, 1984b; Jin and Gross, 1988; Severinov et al., 1993; Severinov et al., 1994). High-resolution structural studies of the Rif-RNAP complex should lead to insights into Rif binding, the mechanism of inhibition, and also the mechanism by which mutations lead to RifR. This could shed light on the transcription mechanism itself, as well as provide the basis for the development of drugs that selectively inhibit bacterial RNAPs but are less prone to single amino acid substitutions giving rise to resistance. The recent determination of the crystal structure of core RNAP from Thermus aquaticus (Taq; Zhang et al., 1999) has opened the door to further studies of RNAP structure, function, and interactions with substrates, ligands, crystal strucand inhibitors. Here we describe the 3.3 A ture of Taq core RNAP complexed with Rif. The structure explains the effects of Rif on RNAP function. In combination with a model of the ternary transcription complex (Korzheva et al., 2000) and biochemical experiments, the data indicate that the predominant effect of Rif is to directly block the path of the elongating RNA transcript at the 5 end when the transcript becomes either 2 or 3 nt in length. Results Rifampicin Inhibition of Taq RNAP From a biochemical perspective, the interaction of Rif with RNAP has been extensively characterized using E. coli RNAP, which served as a prototype for bacterial pathogens (Honore et al., 1993; Nolte, 1997; Ramaswamy and Musser, 1998; Drancourt and Raoult, 1999; Heep et al., 1999; Morse et al., 1999; Padayachee and Klugman, 1999; Wichelhaus et al., 1999). We investigated the inhibition of Taq RNAP by Rif to assess this system as a structural model for Rif-RNAP interactions. Sequence comparisons in the four distinct regions of rpoB that harbor RifR mutations indicate a very high level of conservation among prokaryotes. Between E. coli, Taq, and M. tuberculosis, the sequences are 91% identical over 60 residues (93% conserved), explaining the broad spectrum of Rif activity. Nevertheless, among the 23 positions with single amino-acid substitutions that confer RifR in either E. coli or M. tuberculosis, 5 of these (Taq 387, 395, 398, 453, and 566; Taq numbering will be used throughout unless otherwise specified) are substituted in Taq (Figure 1). In contrast, there is a relatively low level of conservation between prokaryotes and eukaryotes within these regions (Figure 1), explaining the lack of Rif activity against eukaryotic RNAPs.

Summary Rifampicin (Rif) is one of the most potent and broad spectrum antibiotics against bacterial pathogens and is a key component of anti-tuberculosis therapy, stemming from its inhibition of the bacterial RNA polymerase (RNAP). We determined the crystal structure of Thermus aquaticus core RNAP complexed with Rif. The inhibitor binds in a pocket of the RNAP subunit deep within the DNA/RNA channel, but more than 12 A away from the active site. The structure, combined with biochemical results, explains the effects of Rif on RNAP function and indicates that the inhibitor acts by directly blocking the path of the elongating RNA when the transcript becomes 2 to 3 nt in length. Introduction Each year, there are 810 million new cases of tuberculosis (TB), which is the leading cause of death in adults by an infectious agent (Raviglioni et al., 1995; Shinnick, 1996). With TB near epidemic proportions in some parts of the world and the rapid increase in multidrug-resistant strains of Mycobaterium tuberculosis, the World Health Organization declared TB to be a global public health emergency (Raviglioni et al., 1995). Rifampicin (Rif; Sensi et al., 1960; Sensi, 1983) is one of the most potent and broad spectrum antibiotics against bacterial pathogens and is a key component of anti-TB therapy. The introduction of Rif in 1968 greatly shortened the duration of TB chemotherapy. Rif diffuses freely into tissues, living cells, and bacteria, making it extremely effective against intracellular pathogens like M. tuberculosis (Shinnick, 1996). However, bacteria develop resistance to Rif with high frequency, which has led the medical community in the United States to commit to a voluntary restriction of its use for treatment of TB or emergencies. The bactericidal activity of Rif stems from its highaffinity binding to, and inhibition of, the bacterial DNAdependent RNA polymerase (RNAP; Hartmann et al., 1967). The essential catalytic core RNAP of bacteria (subunit composition 2) has a molecular mass of around 400 kDa and is evolutionarily conserved among all cellular organisms (Archambault and Friesen, 1993).
To whom correspondence should be addressed (e-mail: darst@ rockefeller.edu).

Cell 902

Figure 1. The Rif-Resistant Regions of the RNAP Subunit The bar on top schematically represents the E. coli subunit primary sequence with amino acid numbering shown directly above. Gray boxes indicate evolutionarily conserved regions among all prokaryotic, chloroplast, archaebacterial, and eukaryotic sequences (labeled AI at the top; Allison et al., 1985; Sweetser et al., 1987). Red markings indicate the four clusters where RifR mutations have been identified in E. coli (Ovchinnikov et al., 1983; Lisitsyn et al., 1984a, 1984b; Jin and Gross, 1988; Severinov et al., 1993; Severinov et al., 1994), denoted as the N-terminal cluster (N), and clusters I, II, and III (I, II, III). Directly below is a sequence alignment spanning these regions of the E. coli (E.c.), T. aquaticus (T.a.), and M. tuberculosis (M.t.) RNAP subunits. Amino acids that are identical to E. coli are shaded dark gray, and those that are homologous (ST, RK, DE, NQ, FYWIV) are shaded light gray. Mutations that confer RifR in E. coli and M. tuberculosis are indicated directly above (for E. coli) or below (for M. tuberculosis) as follows: for deletions, for insertions, and colored dots for amino acid substitutions (substitutions at each position are indicated in single amino acid code in columns above or below the positions). Color coding for the amino acid substitutions (for reference to subsequent figures) is as follows: yellow, residues that interact directly with the bound Rif (see Figure 4); green, residues that are too far away from the Rif for direct interaction (see Figure 5); purple, three positions that are substituted with high frequency (noted as a % immediately below the substitutions) in clinical isolates of RifR M. tuberculosis (Ramaswamy and Musser, 1998). Below the three prokaryotic sequences is a sequence alignment of three eukaryotic sequences with shading as above. The dots indicate a gap in the alignment.

A plate assay (see Experimental Procedures) showed that Taq cells were unable to grow on media supplemented with 50 g/ml Rif (data not shown). For in vitro studies, Taq RNAP holoenzyme was reconstituted using Taq core RNAP (purified from Taq cells; Zhang et al., 1999) and recombinant Taq A (overexpressed and purified from E. coli; Minakhin et al., 2001b). The enzyme initiated, elongated, and terminated transcripts efficiently from a template containing the T7A1 promoter and the tR2 intrinsic terminator (Figure 2a; Nudler et al., 1994) at 37C using the dinucleotide CpA as the initiating primer. The major RNA products, a trimeric abortive transcript (CpApU), a 105 nt terminated transcript (Term), and a 127 nt runoff transcript (Run off), were the same as those produced by E. coli RNAP (Figure 2a, lanes 1 and 8). Since E. coli 70 is totally inactive when combined with Taq core RNAP in this assay (Minakhin et al., 2001b), the possibility of trace contamination with E. coli 70 does not affect the conclusions from this assay for Taq RNAP. Quantitatively, the two RNAPs responded very differently to Rif; the Ki (estimated from the Rif concentration where the production of long transcripts was inhibited by 50%) for E. coli RNAP was about 0.1 M while for Taq RNAP it was about 10 M, a 100fold difference in sensitivity (the Rif sensitivity of the thermophilic RNAP decreased at higher assay temperatures; data not shown). Qualitatively, however, both

RNAPs responded the same way, with an increase in the production of the trimeric product and a concurrent precipitous drop in the production of the long transcripts (Figure 2a). Mustaev et al. (1994) used chimeric Rif-nucleotide compounds to measure the distance between the initiating nucleotide binding site (the i-site) and the Rif binding site. By varying the linker between the Rif and the nucleotide and testing for maximal transcription initiation activity, the optimal length was found that allowed binding of each moiety in its respective site. This experiment was used to compare the disposition of the Rif and i-sites in E. coli and Taq RNAP. In both cases, optimal initiation activity was observed when the linker comprised five (CH2) groups (Figure 2b). Thus, in spite of the fact that Taq RNAP requires a 100-fold higher concentration of Rif for inhibition, we conclude that Taq RNAP binds Rif and is inhibited through the same biochemical mechanism as E. coli RNAP, and the disposition of the Rif site with respect to the universally conserved active site is identical. We conclude that Taq RNAP can serve as a model for Rif interactions with other RNAPs. Rif-RNAP Structure Determination and Refinement Tetragonal crystals of Taq core RNAP (Zhang et al., 1999) were incubated overnight in stabilization buffer with 0.1 mM Rif, followed by a 30 min. soak in cryo-

Structure of the Rifampicin-RNA Polymerase Complex 903

Figure 2. Rif Inhibition of Taq RNAP (a) Autoradiographs showing the radioactive RNA produced by Taq (lanes 17) and E. coli (lanes 813) RNAP holoenzymes transcribing a template containing the T7 A1 promoter and the tR2 terminator, analyzed on a 15% polyacrylamide gel and quantitated by phosphorimagery. In the absence of Rif (lanes 1 and 8), the major RNA products from each RNAP correspond to a trimeric abortive product (CpApU), a 105 nt terminated transcript (Term), and a 127 nt runoff transcript (Run off). Lanes 27 and 913 show the effects of increasing concentrations of Rif. The quantitated results are shown on the right, where the amounts of each product (normalized to 100% for the Run off and Term transcripts in the absence of Rif, and for CpApU at the highest concentration of Rif) are plotted as a function of Rif concentration. (b) The distance between the bound Rif and the initiating substrate (i-site) of E. coli and Taq RNAP holoenzymes was measured using chimeric Rif-nucleotide compounds as previously described (Mustaev et al., 1994). Rif-nucleotide compounds (Rif-(CH2)n-Ap) with different linker lengths, n (indicated above each lane), were bound to RNAP, then extended in a specific transcription reaction with -[32P]UTP by the RNAP catalytic activity. The products were analyzed on a 23% polyacrylamide gel, visualized by autoradiography, and quantitated by phosphorimagery. The quantitated results are shown directly below, where the product yield (as % activity normalized to 100% at the highest level) is plotted as a function of the Rif-nucleotide linker length (n).

Cell 904

solution (without Rif) before flash freezing. During this procedure, the crystals took on a deep orange color, confirming the binding of Rif. The same results were obtained with cocrystals grown in the presence of 0.1 mM Rif, suggesting that Rif binding does not cause significant conformational changes in the RNAP. The Taq core Rif-RNAP crystals were isomorphous with the native Taq core RNAP crystals (Zhang et al., 1999). Strong electron density was observed in difference Fourier maps for the Rif (Figure 3a), which occupies a shallow pocket between structural domains 3 and 4 (Figure 3b) that is surrounded by the known RifR mutations (Figure 1; Zhang et al., 1999). The electron difference density also indicated shifts and/or ordering of several residues interacting directly with Rif, including Q390, L391, Q393, D396, H406, R409, and L413 (Figure 4). Only very small shifts in localized regions of the protein backbone were indicated (not shown). The Rif X-ray crystal structure (Brufani et al., 1974) was easily placed into the difference density. Subsequent refinements resulted in small shifts of the ansa chain (Figure 3c) to better fit the density. Multiple rounds of manual rebuilding against (2|Fo| |Fc|) maps and refinement resulted in the current model (see Experimental Procedures and Table 1). Overall Structure As expected from the fact that all mapped RifR mutants occur in rpoB (Figure 1), Rif makes contacts only with the RNAP subunit in a close complementary fit to its binding pocket deep within the main DNA/RNA channel. Clearly, Rif does not bind directly at the RNAP active site (Figure 3b). The closest approach of Rif to the active site, defined as the distance between the active site . Mg2 and Rif C38 (Figure 3c), is 12.1 A Detailed Interactions A large number of Rif derivatives have been investigated for antimicrobial activity. In general, modification of the ansa bridge, or modifications that alter the conformation of the ansa bridge, reduce activity. Other structural features of the antibiotic that are particularly critical for activity include the napthol ring with oxygen atoms (O1 and O2) at C1 and C8, and unsubstituted hydroxyls (O10 and O9) at C21 and C23 (Figure 3c; Brufani et al., 1974; Lancini and Zanichelli, 1977; Arora, 1981, 1983, 1985; Sensi, 1983; Arora and Main, 1984). Most Rif modifications that retain activity involve substitutions at C3 of the napthol ring, which have only modulatory effects on in vitro activity. These results are explained by the structural details of the Rif-RNAP complex (Figures 4 and 5). A cluster of hydrophobic residues (L391, L413, G414, I452) line one wall of the Rif binding pocket and make van der Waals contact with the napthol ring and the methyl group at C7. One end of the binding pocket (the bottom in Figure 4) is formed by Q390. The alkyl chain of Q390 makes van der Waals contact with Rif C28 and C29, while the polar head group may interact with O5. Protein groups are positioned to make hydrogen bonds with each of the four critical hydroxyls of Rif: R409 with O1, Q393 and S411 with O2, and D396 and H406 with O10. O9

and O10 are also in position to interact with the backbone amide and carboxyl of F394, respectively. Also positioned to make a potential hydrogen bond with the backbone amide of F394 is O8. D396 contributes to the binding interface in several ways. In addition to forming a potential hydrogen bond with O10, it forms the top end of the binding pocket (in Figure 4) by making van der Waals contact with C18 C21, and C31. Moreover, the negative charge of D396 may be important for neutralizing the positive charges of two nearby side chains, R405 (not shown) and R409 away. The charge neutraliza(Figure 4), each about 6 A tion might be important for the binding of the relatively apolar Rif. Most RifR mutants at 396 substitute a large, bulky group that would likely interfere with Rif binding and would not have the correct geometry for hydrogen bonding O10 (Y), or else apolar groups (V, G, or A) with no hydrogen bonding ability. One of these mutants, D396V (position 516 in E. coli), was among the original, strong RifR mutants mapped by Ovchinnikov et al. (1983), pointing to the importance of this residue in forming the Rif binding interface. Another mutant identified in E. coli, however (D396N), is isosteric with Asp and would likely maintain the hydrogen bond with O10. Nevertheless, this substitution yields weak RifR (Lisitsyn et al., 1984a), which may be caused by the loss of negative charge at this position. Rif has a partial charge, localized at N4 (Figure 3c). A negatively charged residue, E445, is situated nearby and may contribute to the Rif binding site by neutralizing this charge. This is not likely to be a strong effect, as many Rif derivatives with equal or stronger activity than Rif do not have this partial charge. E445 is the only residue close enough to Rif to be involved in potentially direct interactions (Figure 4) for which a RifR mutant has not been reported. However, this residue is universally conserved as either Glu or Asp in a segment of region D that is invariantly present in prokaryotes, chloroplast, archaebacteria, and eukaryotes (Allison et al., 1985; Sweetser et al., 1987), pointing to its importance for the basic function of RNAP. Thus, of the 12 residues that are close enough to Rif to make direct interactions (including backbone interactions with F394; Figure 4), 11 mutate to a RifR phenotype. The 12th position, E445, is highly conserved so that its substitution would likely be lethal and consequently not be detectable as RifR mutations. Twelve additional positions have been identified at which substitution gives rise to RifR (Figure 1). These residues surround the Rif binding pocket but do not make direct interactions with the antibiotic (Figure 5). In every case, the RifR mutations involve replacement by a different sized amino acid side chain (almost always substituting a small residue with a more bulky one), or else involve adding or removing a Pro residue. These substitutions would likely affect the folding or packing of the protein in the local vicinity of the substituted residue, causing distortions of the Rif binding pocket. Mechanism of RNAP Inhibition by Rif The effects of Rif on RNAP in each stage of the transcription cycle have been probed using detailed kinetic analy-

Structure of the Rifampicin-RNA Polymerase Complex 905

Figure 3. Rif-RNAP Cocrystal Structure (a) Stereo view of the Rif binding pocket of Taq core RNAP, generated using O (Jones et al., 1991). Carbon atoms of the RNAP subunit are of the Rif), while carbon atoms of the inhibitor are orange. Oxygen atoms are red, nitrogen atoms are blue, cyan or yellow (residues within 4 A and sulfur atoms are green. Electron density, calculated using (|FoRif Fonat|) coefficients (Rif denotes the Rif-RNAP cocrystal, native denotes the native core RNAP crystal), is shown (orange) for the Rif only (contoured at 3.5 ), and was computed using phases from the final refined RNAP model with the Rif omitted. (b) Three-dimensional structure of Taq core RNAP in complex with Rif, generated using GRASP (Nicholls et al., 1991). The backbone of the RNAP structure is shown as tubes, along with the color coded transparent molecular surface (, cyan; , pink; , white; the subunits are behind the RNAP and are not visible). The Mg2 ion chelated at the active site is shown as a magenta sphere. The Rif is shown as CPK atoms (carbon, orange; oxygen, red; nitrogen, blue). (c) Structural formula of Rif. Features of the structure discussed in the text are color coded (ansa bridge, blue; napthol ring, green). The four oxygen atoms critical for Rif activity (Brufani et al., 1974; Lancini and Zanichelli, 1977; Arora, 1981, 1983, 1985; Sensi, 1983; Arora and Main, 1984) are shaded with red circles.

ses. Rif has essentially no effect on specific promoter binding and open complex formation (Hinkle et al., 1972; McClure and Cech, 1978). A small increase (about 2-fold) in the apparent Km for initiating substrate binding in the enzymes i-site (the 5 nucleotide) was observed, but the binding of the incoming nucleotide substrate in the i1 site (the 3 nucleotide) and the formation of the first phosphodiester bond were largely unaffected (McClure and Cech, 1978). The dominant effect of Rif

binding on RNAP activity was a total blockage of synthesis of the second (when transcription was initiated with a nucleoside triphosphate) or third (when transcription was initiated with a nucleoside di- or monophosphate) phosphodiester bond (McClure and Cech, 1978). Since synthesis of the first and second phosphodiester bonds can occur in the presence of Rif, the antibiotic does not interfere with substrate binding, catalytic activity, or the intrinsic translocation mechanism of the RNAP. After

Cell 906

Figure 4. Detailed Interactions of Rif with RNAP (a) Stereo view of the Taq RNAP Rif binding pocket complexed with Rif, generated using RIBBONS (Carson, 1991), showing residues that interact directly with the inhibitor. The view is from the top of the RNAP in Figure 3b, above the subunit, looking down through to the Rif, but with obscuring parts of removed. The backbone of the subunit is shown as a cyan ribbon. Side chains (and backbone atoms of F394) of Rif are shown. Carbon atoms are orange (Rif), magenta (three residues substituted in M. tuberculosis RifR clinical of residues within 4 A isolates with high frequency, see Figure 1), or yellow; oxygen atoms are red; nitrogen atoms are blue. Potential hydrogen bonds between protein atoms and Rif are shown as dashed lines. (b) Schematic drawing of RNAP subunit interactions with Rif, modified from LIGPLOT (Wallace et al., 1995). Residues forming van der Waals interactions are indicated, those participating in hydrogen bonds are shown in a ball-and-stick representation, with hydrogen bonds depicted as dashed lines. Carbon atoms of the protein are black, while carbon atoms of Rif are orange. Oxygen atoms are red and nitrogen atoms are blue.

RNAP has synthesized a long transcript and entered the elongation phase, it becomes totally resistant to Rif. These properties led to the proposal that Rif inhibits RNAP through a simple steric block of the path of the

elongating RNA at the 5 end (McClure and Cech, 1978). Whether Rif directly blocked the path of the RNA or if blockage was an indirect effect due to a conformational change in the RNAP induced by Rif binding could not

Structure of the Rifampicin-RNA Polymerase Complex 907

Table 1. Crystallographic Data and Structural Model Diffraction Data Parameter ) Resolution range (A Rmergea (%) Completeness (%) I/I No. of reflections No. of unique obs. Structural Model Protein Subunitb I II Total Refinement Rcryst (%) Rfree (%)
a b

Total 303.3 7.7 86.1 10.7 75,420 214,453

Outer Shell 3.423.3 34.4 71.1 1.7 6173 11,549

No. of Residues Mr (kDa) 170.7 124.4 34.9 34.9 11.6 376.5 28.1 35.9 Sequence 1525 1119 313 313 99 3369 Model 1139 1114 223 229 98 2803 Regions Modeled 331, 69155 (poly-Ala), 452523, 5361241, 12501410, 14141497 21115 6228 3231 198

Rmerge |Ij I|/Ij. Also included in the model was one Mg2 and Zn2 ion. (Zhang et al., 1999), and one Rif molecule (Brufani et al., 1964).

be distinguished. Alternatively, others have proposed that Rif exerts its effect allosterically by decreasing the affinity of the RNAP for short RNA transcripts (Schulz and Zillig, 1981). The Rif-RNAP crystal structure explains the results described above and strongly supports the simple steric block mechanism (McClure and Cech, 1978). Rif directly abuts the base of a loop that comprises the C-terminal part of conserved region D (residues 443451, shaded red in Figure 5; D loop II in Korzheva et al. [2000]), and a cluster of RifR mutants, Rif cluster I (Figure 1), flanks this region. Modeling suggests that this loop, which contains several nearly universally conserved residues, participates in forming the binding site for the base pair at 1 in the transcription complex (Korzheva et al., 2000), so effects of Rif on the Km for the initiating substrate are not surprising. However, Rif does not directly contact the end of this loop. In addition, conformational changes of the protein in this region are not indicated from the structural data, consistent with the observation that the effect of Rif on this region is small. The principal effect of Rif is seen in the context of a model of the transcriptionally active ternary complex (Korzheva et al., 2000) containing RNAP, DNA template, and RNA transcript (Figure 6). In this figure, only the RNAP active site Mg2 and the 9 bp RNA/DNA hybrid (from 1 to 7) from the ternary complex model are shown. The rest of the RNAP and nucleic acids are omitted for clarity. Also shown is the atomic model of Rif as it would be positioned in its binding site on the subunit. It can be seen that the two substrate nucleotides, at 1 (green) and 1, are not directly affected by the presence of Rif, so that RNAP can bind and catalyze the formation of a phosphodiester bond between the two substrates in the presence of the antibiotic. With a

transcript length of 3 nt, however, the 5 phosphates of the 5 nucleotide (at 2) sterically clash with Rif, and the nucleotides further upstream (3 to 5) severely clash with Rif. At the same time, Rif does not interfere with the DNA (gray). Thus, the structure, in combination with the ternary complex model, explains the biochemical data on the mechanism of Rif inhibition, provides strong support for the proposal that Rif sterically blocks the path of the elongating RNA transcript at the 5 end, and indicates that the blockage is a direct consequence of Rif binding in its site. The model even suggests why transcripts initiated with nucleoside triphosphates are blocked after the first phosphodiester bond, while transcripts initiated with nucleoside di- or monophosphates are blocked after the second phosphodiester bond. In the model, the nucleoside monophosphate in the transcript at the 2 position clashes only slightly with Rif, while the presence of a 5 triphosphate at the 2 position would extend into Rif. The confluence of the structural and biochemical data also lends support to the ternary complex model of Korzheva et al. (2000). Core RNAP can bind a preformed minimal nucleic acid scaffold of RNA/DNA oligonucleotides (Figure 6b, top) to yield functional ternary elongation complexes (Korzheva et al., 2000). We performed order of addition experiments using this system in order to assess whether Rif and RNA binding were competitive (Figure 6b). The DNA component of the scaffold was annealed with varying lengths of RNA transcript, and the effect of Rif, added before or after the oligonucleotides, on the sequence-dependent extension of RNA by one nucleotide (radioactively labeled CTP) was assayed at room temperature. In the case of E. coli core RNAP in the absence of Rif, the RNA transcript was extended with nearly equal efficiency regardless of its length within a range of 37 nt (Figure 6b, lanes 1115). When

Cell 908

Figure 5. Rif Binding Pocket and RifR Mutants Stereo views of the Taq RNAP Rif binding pocket complexed with Rif. The view (the same in [a] and [b]), rotated approximately 180 about the horizontal axis from the view of Figure 4a, is as if one was in the middle of the main RNAP channel in Figure 3b (between and , looking up towards the Rif, with the subunit behind). (a) The backbone of the subunit is shown as a cyan ribbon, but with a highly conserved segment of region D (443451, see text) colored red. Side chains (and backbone atoms of F394) of residues where substitutions confer RifR (see Figure 1) are shown. Carbon atoms are orange (Rif), magenta (three residues substituted in M. tuberculosis RifR clinical isolates with high frequency, see Figure 1), yellow (other residues that interact directly with Rif, as in Figure 4), or green (all other RifR positions). Oxygen atoms are colored red; nitrogen atoms are blue. Generated using RIBBONS (Carson, 1991). (b) The subunit is shown as a cyan molecular surface, with a highly conserved segment of region D colored red, and surface-exposed RifR of the Rif) or green. Generated using GRASP (Nicholls et al., 1991). positions colored yellow (within 4 A

Rif was added prior to the nucleotide scaffold, the RNAP was unable to extend any of the RNA oligos, regardless of length (lanes 15), indicating that Rif occupied its site and blocked the extension and/or binding of all of the transcripts. When the scaffold was added prior to Rif addition, Rif was able to occupy its site and block the extension of the 3 nt transcript (lane 6), but had no effect on the extension of the longer transcripts (lanes 710), presumably because Rif could not access its binding site blocked by the longer RNA transcripts (Figure 6a). These results are consistent with the early data that Rif inhibits the RNA extension from 2 to 3 nt if the 5 nucleoside is tri-phosphorylated, but inhibits extension from 3 to 4 nt if the 5 nucleoside is mono- or di-phosphorylated (McClure and Cech, 1978) since the synthetic RNA oligos lack 5 phosphates. Similar experiments were performed with Taq core RNAP (Figure 6b, lanes 1630). In the absence of Rif, the efficiency of transcript extension was strongly dependent on the transcript length (lanes 2630). Exten-

sion of the shortest transcripts was barely detectable, suggesting that, unlike E. coli RNAP, Taq core RNAP does not bind and stabilize the short, intrinsically unstable RNA/DNA hybrids. In the presence of Rif, a generalized inhibition of transcript extension was observed regardless of the order of addition or of the transcript length (lanes 1625). We explain these results by the low binding affinity of Taq core RNAP for both Rif and for short RNA transcripts compared with E. coli core RNAP. The low affinities imply fast off rates, which would allow equilibrium to be established between the Rif and scaffold binding during the time of the assay. Discussion In summary, we have shown that, although Taq RNAP is relatively insensitive to Rif, at sufficiently high concentrations, the antibiotic binds and inhibits the enzyme. Inhibition of Taq RNAP occurs through the same biochemical mechanism as E. coli RNAP, and the disposi-

Structure of the Rifampicin-RNA Polymerase Complex 909

Figure 6. Mechanism of RNAP Inhibition by Rif (a) The RNAP active site Mg2 (magenta sphere) and the 9 bp RNA/ DNA hybrid (from 1 to 8) from a model of the ternary elongation complex (Korzheva et al., 2000) are shown. The RNAP itself and the rest of the nucleic acids are omitted for clarity. The incoming nucleotide substrate at the 1 position is colored green, the 1 and 2 positions, which can be accommodated in the presence of Rif, are colored yellow. The RNA further upstream (3 to 8), which cannot be accommodated in the presence of Rif, is colored pink. The template strand of the DNA is colored gray. Also shown is a CPK representation of Rif as it would be positioned in its binding site on the subunit (carbon atoms, orange; oxygen, red; nitrogen, blue). The Rif is partially transparent, illustrating the RNA nucleotides at 3 to 5 that sterically clash. Generated using GRASP (Nicholls et al., 1991). (b) The structure of the minimal scaffold systems with RNA lengths from 37 nt (labeled above the RNA chain; Korzheva et al., 2000). The results are presented below as autoradiographs of the radioactive RNAs produced by E. coli (lanes 115) or Taq (lanes 1630) core RNAPs transcribing the minimal scaffolds with the indicated lengths

tion of the Rif site with respect to the active site is identical to E. coli RNAP and presumably other prokary X-ray otic RNAPs (Figure 2). We determined the 3.3 A crystal structure of Taq core RNAP complexed with Rif. The inhibitor bound with a close complementary fit in a pocket between two structural domains of the RNAP subunit. Only small, local conformational changes of both the inhibitor and the protein were observed. The binding site is deep within the main RNAP channel, but the closest approach of the inhibitor to the RNAP active (Figure 3b). The Rif binding site Mg2 is more than 12 A pocket is surrounded by the 23 known positions where amino acid substitutions confer RifR (Figure 5). Twelve of these residues are close enough to interact directly with the Rif (Figure 4). Predominant are van der Waals interactions with hydrophobic side chains near the napthol ring of Rif, and potential hydrogen bond interactions with five polar groups of Rif (two on the napthol ring and three on the ansa bridge), four of which have been shown to be essential for Rif activity. The remaining RifR mutants are one layer removed from the Rif itself, and are likely to affect Rif binding through small structural distortions of the Rif binding pocket. The structure explains the effects of Rif on RNAP function determined from detailed biochemical and kinetic studies. In combination with a model of the ternary transcription complex, the structure strongly suggests that the predominant effect of Rif is to directly block the path of the elongating RNA transcript at the 5 end when the transcript becomes either 2 or 3 nt in length, depending on the 5 phosphorylation state of the 5 nucleotide (Figure 6). In this view, Rif binds its site in the RNAP holoenzyme either before or after the binding of the DNA template and formation of the open complex, functions of RNAP that are not affected by the presence of Rif. Next, the two nucleotide substrates bind their sites in the RNAP active site. The initiating nucleotide substrate binds the RNAP i-site with a small, approximately 2-fold increase in the apparent Km due to the presence of Rif, while the second nucleotide binds in the i1 site with little effect by Rif. More or less normally, the RNAP then catalyzes the formation of a phosphodiester bond between the two nucleotides. If the initiating nucleoside bears a 5 triphosphate, the subsequent translocation of the RNAP attempts to move the 2 nt RNA transcript upstream such that the i1 nucleotide occupies the i-site (1 position), and the i-site nucleotide moves into the 2 position (Figure 6a). The movement of the 5 nucleotide into the 2 position, however, results in a severe steric clash with the Rif. The molecular details of ensuing events are unclear, but in the end, the RNAP remains at the same template position, the 2 nt transcript is released, and the

of RNA (X ) and analyzed on a 23% polyacrylamide gel. Lanes 110 and 1625 demonstrate the effect of Rif inhibition of transcription when it was bound by RNAP either before (lanes 15 and 1620) or after (lanes 610 and lanes 2125) addition of the scaffold. Lanes 1115 and 2630 show elongation of the same scaffolds in the absence of Rif. The RNA with the critical length of 3 nt, which cannot be elongated by E. coli RNAP in the presence of Rif regardless of the order of Rif and scaffold addition (lanes 1 and 6), is colored red. The RNAs of 47 nt (colored green) were extended by E. coli RNAP when added before Rif (lanes 610).

Cell 910

futile cycle begins again. If the 5 nucleoside contains a di- or a monophosphate at its 5 end (or if its unphosphorylated), then after the synthesis of the first phosphodiester bond, the RNAP can translocate normally and the steric clash of the transcript with Rif occurs during the translocation of the 3 nt transcript following the synthesis of the second phosphodiester bond. The relative insensitivity of Taq RNAP to Rif is likely due to amino acid substitutions in Taq RNAP compared with other, more Rif-sensitive RNAPs. The 12 residues close enough to interact directly with Rif are identical between E. coli, Taq, and M. tuberculosis (marked yellow in Figure 1). Among the 11 positions that do not directly interact with Rif but likely affect Rif binding indirectly, 5 are substituted in Taq RNAP (387, 395, 398, 453, and 566; Figure 1). Although the effect of these residues on the structure of RNAP is difficult to assess with only one available structure, we can venture some suppositions. Three of these positions, 387, 398, and 453, contain amino acids that are not dramatically different in overall size from their E. coli and M. tuberculosis counterparts and we predict that these residues are not the origin of the Taq RNAP insensitivity to Rif. Position 566 is highly conserved among all RNAPs as either Lys or Arg (the homologous position is Arg in both E. coli and M. tuberculosis) but is Thr in Taq RNAP. This substitution is unlikely to be the main determinant of the Taq RNAP Rif insensitivity, however, since Minakhin et al. (2001b) mutated Taq Thr-566 to Arg, but this had little effect on the RifR of the enzyme assayed at 45C. This leaves position 395, which is highly conserved as a hydrophobic residue among all bacterial RNAPs. In E. coli and M. tuberculosis, this position is a Met, but in Taq, it is a Lys. Taq Lys-395 appears to participate in buried salt bridges with Asp-124 and Asp-133 that may contribute to thermostability of the protein. This nonconservative substitution (Lys for Met) could affect the local path of the polypeptide backbone, and is immediately adjacent to Phe-394, the backbone amide and carboxyl of which appear to be involved in important interactions with the Rif (Figure 4). All but one of the residues that are close enough to Rif to participate in direct interactions are known to mutate to strong RifR (Figure 4). However, additional residues could be important for the formation of the Rif binding pocket, but not revealed as RifR mutants if they are necessary for basic RNAP function. As mentioned above, the four regions of the subunit that harbor RifR mutants are highly conserved among prokaryotes (Figure 1), but the much weaker homology with archaebacterial and eukaryotic RNAPs, combined with the fact that so many RifR mutations have been discovered, indicate that these regions are not critical to RNAP function in vivo. Nevertheless, some RifR mutations do have profound functional effects (Landick et al., 1990; Jin and Gross, 1991), and E. coli strains with RifR RNAP have been shown to be at a competitive disadvantage to wt E. coli in the absence of Rif (Jin and Gross, 1989). The clinical success of Rif proves that the bacterial RNAP is an excellent target for antimicrobials. The structure and available genetic and biochemical data suggest that the design of modified versions of Rif to overcome the effects of RifR mutations, while perhaps leading to incremental improvements, may be futile in the long run

because of the apparently small functional penalties of mutating this region of the RNAP, and the variety of amino acid positions and mutations that result in RifR (Figure 1). Somewhat encouraging, however, are the findings from clinical isolates of RifR M. tuberculosis. Although the RifR mutations are spread over 15 positions of rpoB, 77% of all the mutations isolated involved substitutions at one of only two positions, corresponding to Taq 406 and 411, and an additional position (Taq 396) accounts for a combined 86% of all the mutants. An important conclusion from this study emerges from the inhibition mechanism of Rif, a simple steric block of transcription elongation. Thus, the powerful effects of Rif do not stem from the details of its chemical structure, and do not involve interference with the catalytic activity of the RNAP, for instance by mimicking substrates or transition states of the polymerization reaction. Such an inhibitor would act on features that are highly conserved between prokaryotes and eukaryotes, rendering it useless as an antimicrobial agent. Rather, the effects of Rif depend only on its ability to bind tightly to a relatively nonconserved part of the structure, disrupting a critical RNAP function by virtue of its presence. Decades of functional studies (Chamberlin, 1993; Mustaev et al., 1997; Korzheva et al., 1998; Nudler, 1999), and more recent structural evidence (Mooney and Landick, 1999; Zhang et al., 1999; Cramer et al., 2000; Korzheva et al., 2000), indicate that cellular RNAPs operate as complex molecular machines, with extensive interactions with the template DNA, product RNA (Korzheva et al., 2000), and other regulatory molecules. It seems likely that many distinct sites exist where the tight binding of a small molecule would disrupt critical features of the functional mechanism.
Experimental Procedures Purification and Crystallization Native Taq core RNAP was purified and crystallized as described previously (Zhang et al., 1999). The crystals were then soaked in stabilization solution (2 M (NH4)2SO4, 0.1 M Tris-HCl, pH 8.0, and 20 mM MgCl2) with 0.1 mM Rif for at least 12 hr. The crystals were then prepared for cryo-crystallography by soaking in stabilization solution containing 50% (w/v) sucrose for 30 min before flash freezing in liquid nitrogen. Diffraction data was collected at the APS beamline SBC 19ID using 0.3 oscillations, and processed using DENZO and SCALEPACK (Otwinowski, 1991). Structure Determination The native core RNAP structure (Zhang et al., 1999) was used as a starting model for rigid body refinement and positional refinement against the observed amplitudes from the Rif-RNAP complex crystal (FoRif:) using CNS (Adams et al., 1997), yielding an initial R factor of 0.354 (Rfree 0.41, where the identical reflections was set aside as for the Rfree determination of the native structure) for data from 100 resolution. An initial Fourier difference map, calculated using 3.2 A |FoRif Fonat| amplitude coefficients and using phases calculated from the native core RNAP structure (nat), clearly revealed density for the Rif molecule (Figure 3a). Multiple rounds of manual rebuilding against (2|Fo| |Fc|) maps using O (Jones et al., 1991), and refinement using CNS (Adams et al., 1997), resulted in the current model (Table 1). At later stages of the refinement, The Rif X-ray crystal structure (Brufani et al., 1974) was easily placed into the difference density. Included in the model is the recently determined sequence of the Taq subunit (Minakhin et al., 2001a), modeled earlier as a polyalanine chain (Zhang et al., 1999). Still missing from the model is a 300 amino acid, nonconserved domain inserted between conserved regions A and B of the subunit (Zhang et al., 1999).

Structure of the Rifampicin-RNA Polymerase Complex 911

Assays Taq cells were tested for sensitivity to Rif on solid media. Plates containing 3% bactoagar and 1/5 dilution of Luria broth were poured with and without 50 g/ml of Rif. Cells from frozen stock were then streaked onto plates and incubated at 65C for 2 days and assessed for growth. The transcription assay comparing Rif inhibition of E. coli and Taq RNAPs (Figure 2a) was performed as described (Nudler et al., 1994). Briefly, 0.1 pmol of purified Taq core RNAP (Zhang et al., 1999) was incubated with Taq A (Minakhin et al., 2001b) in 20 l of transcription buffer (40 mM Tris-HCl, pH 7.9, 40 mM KCl, 5 mM MgCl2) for 15 min at 37C to form holoenzyme. Rif was added to the final concentrations indicated in Figure 2a and incubated another 5 min at 37C, followed by addition of 0.15 pmol of T7A1 promoter fragment and incubation for 5 min at 37C. RNA synthesis was initiated by the addition of CpA primer (100 M), NTPs (25 M each), and -[32P]UTP (0.3 M), and the reaction was stopped after incubation for 10 min at 37C. The assay for E. coli RNAP holoenzyme was the same except the CpA primer was added to a concentration of 10 M. Radioactive RNA products were analyzed on a 15% polyacrylamide sequencing gel. Assays for extension of the Rif-nucleotide compounds (Figure 2b) were carried out as described (Mustaev et al., 1994) with minor modifications. After binary complex formation, transcription reactions were started by the addition 10 M Rif-(CH2)n-A compound, with the n indicated in Figure 2b, and -[32P]UTP (0.3 M). The reactions were incubated for 2 min at room temperature for E. coli RNAP and 3 min at 55C for Taq. Under these conditions, the reaction was not complete, and the yield of the Rif-(CH2)n-ApU depended on the linker length. Radioactive RNA products were analyzed on a 23% polyacrylamide sequencing gel. Transcription reactions on the minimal scaffold system shown (Figure 6b) were performed as described (Korzheva et al., 2000) with minor modifications. The RNA and DNA components of the scaffold (100 pmol of each) were mixed in 100 l of transcription buffer at 45C and the mixture was allowed to cool to room temperature over 30 min. RNAP/scaffold complexes were formed by incubation of the annealed scaffold (10 pmol) with a molar equivalent of core RNAP (either E. coli or Taq) preincubated with Rif (100 M for E. coli, 200 M for Taq) for 10 min, to form the RNAP/scaffold complex. Extension of the RNA oligonucleotide was assayed by the addition of -[32P]CTP (0.3 M) and 5 min incubation at room temperature. In Figure 6b, lanes 15 and 1620, RNAP was preincubated with Rif (100 M for E. coli RNAP, 200 M for Taq) for 10 min. In lanes 610 and 2125, the RNAP/scaffold complexes formed in the absence of Rif were incubated with Rif (concentrations as above) for 10 min. Finally, in lanes 1115 and 2630, the RNAP or RNAP/scaffold complex was not exposed to Rif. Radioactive RNA products were analyzed on a 23% polyacrylamide sequencing gel. Acknowledgments We are indebted to A. Joachimiak, S. L. Ginell, and F. J. Rotella at the Advanced Photon Source Structural Biology Center for support during data collection. Use of the Argonne National Laboratory Structural Biology Center beamlines at the Advanced Photon Source was supported by the U. S. Department of Energy, Office of Biological and Environmental Research, under Contract No. W-31-109ENG-38. We thank V. Nikiforov and J. McKinney for invaluable discussions. E. C. was supported by a Kluge postdoctoral fellowship and a National Research Service Award (NIH GM20470). K. M. was supported by a Human Frontiers Sciences Program postdoctoral fellowship and a Norman and Rosita Winston postdoctoral fellowship. This work was supported by NIH grants GM49242 and GM30717 to A. G., and GM53759 and GM61898 to S. A. D. Received December 13, 2000; revised January 31, 2001. References Adams, P.D., Pannu, N.S., Read, R.J., and Brunger, A.T. (1997). Cross-validated maximum likelihood enhances crystallographic

simulated annealing refinement. Proc. Natl. Acad. Sci. USA 94, 5018 5023. Allison, L.A., Moyle, M., Shales, M., and Ingles, C.J. (1985). Extensive homology among the largest subunits of eukaryotic and prokaryotic RNA polymerases. Cell 42, 599610. Archambault, J., and Friesen, J.D. (1993). Genetics of RNA polymerases I, II, and III. Microbiol. Rev. 57, 703724. Arora, S.K. (1981). Structural investigations of mode of action of drugs III. Structure of rifamycin S iminomethyl ether. Acta crystallogr. B37, 152157. Arora, S.K. (1983). Correlation of structure and activity in ansamycins: molecular structure of sodium rifamycin SV. Mol. Pharmacol. 23, 133140. Arora, S.K. (1985). Correlation of structure and activity in ansamycins: structure, conformation, and interactions of antibiotics rifamycin S. J. Med. Chem. 28, 10991102. Arora, S.K., and Main, P. (1984). Correlation of structure and activity in ansamycins: moleculr structure of cyclized rifamycin SV. J. Antibiot. 37, 178181. Brufani, M., Cerrini, S., Fidelli, W., and Vaciago, A. (1974). Rifamycins, an insight into biological activity based on structural investigations. J. Mol. Biol. 87, 409435. Carson, M. (1991). RIBBONS 2.0. J. App. Crystallogr. 24, 958961. Chamberlin, M.J. (1993). New models for the mechanism of transcription elongation and its regulation. Harvey Lect. 88, 121. Cramer, P., Bushnell, D.A., Fu, J., Gnatt, A.L., Maier-Davis, B., Thompson, N.E., Burgess, R.R., Edwards, A.M., David, P.R., and Kornberg, R.D. (2000). Architecture of RNA polymerase II and implications for the transcription mechanism. Science 288, 640649. Drancourt, M., and Raoult, D. (1999). Characterization of mutations in the rpoB gene in naturally rifampin-resistant Rickettsia species. Antimicrob. Agents and Chemotherapeutics 43, 24002403. Ezekiel, D.H., and Hutchins, J.E. (1968). Mutations affecting RNAP associated with rifampicin resistance in Escherichia coli. Nature London 220, 276277. Hartmann, G., Honikel, K.O., Knusel, F., and Nuesch, J. (1967). The specific inhibition of the DNA-directed RNA synthesis by rifamycin. Biochim. Biophys. Acta 145, 843844. Heep, M., Beck, D., Bayerdorffer, E., and Lehn, N. (1999). Rifampin and rifabutin resistance mechanism in Helicobacter pylori. Antimicrob. Agents and Chemotherapeutics 43, 14971499. Heep, M., Rieger, U., Beck, D., and Lehn, N. (2000). Mutations in the beginning of the rpoB gene can induce resistance to rifamycins in both Helicobacter pylori and Mycobacterium tuberculosis. Antimicrob. Agents and Chemotherapeutics 44, 10751077. Heil, A., and Zillig, W. (1970). Reconstitution of bacterial DNA-dependent RNA polymerase from isolated subunits as a tool for the elucidation of the role of the subunits in transcription. FEBS Lett. 11, 165168. Hinkle, D.C., Mangel, W.F., and Chamberlin, M.J. (1972). Studies of the binding of Escherichia coli RNA polymerase to DNA. IV. The effect of rifampicin on binding and on RNA chain initiation. J. Mol. Biol. 70, 209220. Honore, N., Bergh, S., Chanteau, S., Doucet-Populaire, F., Eiglmeier, K., Garnier, T., Georges, C., Launois, P., Limpaiboon, T., Newton, S., et al. (1993). Nucleotide sequence of the first cosmid from the Mycobacterium leprae genome project: structure and function of the Rif-Str regions. Mol. Microbiol. 7, 207214. Jin, D.J., and Gross, C.A. (1988). Mapping and sequencing of mutations in the Escherichia coli rpoB gene that lead to rifampicin resistance. J. Mol. Biol. 202, 4558. Jin, D.J., and Gross, C.A. (1989). Characterization of the pleiotropic phenotypes of rifampin-resistant rpoB mutants of Escherichia coli. J. Bacteriol. 171, 52295231. Jin, D.J., and Gross, C.A. (1991). RpoB8, a rifampicin-resistant termination-proficient RNA polymerase, has an increase Km for purine nucleotides during transcription elongation. J. Biol. Chem. 266, 1447814485.

Cell 912

Jones, T.A., Zou, J.-Y., Cowan, S., and Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta crystallogr. A47, 110119. Korzheva, N., Mustaev, A., Nudler, E., Nikiforov, V., and Goldfarb, A. (1998). Mechanistic model of the elongation complex of Escherichia coli RNA polymerase. Cold Spring Harb. Symp. Quant. Biol. 63, 337345. Korzheva, N., Mustaev, A., Kozlov, M., Malhotra, A., Nikiforov, V., Goldfarb, A., and Darst, S.A. (2000). A structural model of transcription elongation. Science 289, 619625. Lancini, G., and Zanichelli, W. (1977). Structure-activity relationships in rifamycins. In Structure-Activity Relationship in Semisynthetic Antibiotics, D. Perlaman, ed. (New York: Academic Press), pp. 531600. Landick, R., Stewart, J., and Lee, D.N. (1990). Amino acid changes in conserved regions of the beta-subunit of Escherichia coli RNA polymerase alter transcription pausing and termination. Genes Dev. 4, 16231636. Lisitsyn, N.A., Gurev, S.O., Sverdlov, E.D., Moiseeva, E.P., and Nikiforov, V.G. (1984a). Nucleotide substitutions in the rpoB gene leading to rifampicin resistance of E. coli RNA polymerase. Bioorg Khim 10, 127128. Lisitsyn, N.A., Sverdlov, E.D., Moiseyeva, E.P., Danilevskaya, O.N., and Nikiforov, V. (1984b). Mutation to rifampicin resistance at the beginning of the RNA polymerase beta subunit gene in Escherichia coli. Mol. Gen. Genet. 196, 173174. McClure, W.R., and Cech, C.L. (1978). On the mechanism of rifampicin inhibition of RNA synthesis. J. Biol. Chem. 253, 89498956. Minakhin, L., Bhagat, S., Brunning, A., Campbell, E.A., Darst, S.A., Ebright, R.H., and Severinov, K. (2001a). Bacterial RNA polymerase subunit and eukaryotic RNA polymerase subunit RPB6 are sequence, structural, and functional homologs and promote RNA polymerase assembly. Proc. Natl. Acad. Sci. USA, 98, 892897. Minakhin, L., Nechaev, S., Campbell, E.A., and Severinov, K. (2001b). Recombinant Thermus aquaticus RNA polymerase-A new tool for structure-based analysis of transcription. J. Bacteriol. 183, 7176. Mooney, R.A., and Landick, R. (1999). RNA polymerase unveiled. Cell 98, 687690. Morse, R., OHanlon, K., Virji, M., and Collins, M.D. (1999). Isolation of rifampin-resistant mutants of Listeria monocytogenes and their characterization by rpoB gene sequencing, temperature sensitivity for growth, and interaction with an epithelial cell line. J. Clin. Microbiol. 37, 29132929. Mustaev, A., Zaychikov, E., Severinov, K., Kashlev, M., Polyakov, A., Nikiforov, V., and Goldfarb, A. (1994). Topology of the RNA polymerase active center probed by chimeric rifampicin-nucleotide compounds. Proc. Natl. Acad. Sci. USA 91, 1203612040. Mustaev, A., Kozlov, M., Markovtsov, V., Zaychikov, E., Denissova, L., and Goldfarb, A. (1997). Modular organization of the catalytic center of RNA polymerase. Proc. Natl. Acad. Sci. USA 94, 6641 6645. Nicholls, A., Sharp, K.A., and Honig, B. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons, proteins structure. Funct. Genet. 11, 281296. Nolte, O. (1997). Rifampicin resistance in Neisseria meningitidis: evidence from a study of sibling strains, description of new mutations and notes on population genetics. J. Antimicrob. Chemother. 39, 747755. Nudler, E. (1999). Transcription elongation: structural basis and mechanisms. J. Mol. Biol. 288, 112. Nudler, E., Goldfarb, A., and Kashlev, M. (1994). Discontinuous mechanism of transcription elongation. Science 265, 793796. Otwinowski, Z. (1991). Isomorphous Replacement and Anomalous Scattering, W. Wolf, P.R. Evans, and A.G.W. Leslie, eds. (Daresbury, UK: Science and Engineering Research Council, Daresbury Laboratory). Ovchinnikov, Y.A., Monastyrskaya, G.S., Guriev, S.O., Kalinina, N.F., Sverdlov, E.D., Gragerov, A.I., Bass, I.A., Kiver, I.F., Moiseyeva, E.P.,

Igumnov, V.N., et al. (1983). RNA polymerase rifampicin resistance mutations in Escherichia coli: sequence changes and dominance. Mol. Gen. Genet. 190, 344348. Padayachee, T., and Klugman, K.P. (1999). Molecular basis of rifampin resistance in Streptococcus pneumoniae. Antimicrob. Agents and Chemotherapeutics 43, 23612365. Ramaswamy, S., and Musser, J.M. (1998). Molecular genetic basis of antimicrobial agent resistance in Mycobacterium tuberculosis: 1998 update. Tuber. Lung Dis. 79, 329. Raviglione, M.C., Snider, D.E., Jr., and Kochi, A. (1995). Global epidemiology of tuberculosis. Morbidity and mortality of a worldwide epidemic. JAMA 273, 220226. Schulz, W., and Zillig, W. (1981). Rifampicin inhibition of RNA synthesis by destabilization of DNA-RNA polymerase-oligonucleotide complexes. Nucleic Acids Res. 9, 68896906. Sensi, P. (1983). History of the development of rifampin. Rev. Infect. Dis. 3, 402406. Sensi, P., Greco, A.N., and Ballotta, R. (1960). Isolation and properties of rifomycin B and rifomycin complex. Antibiot. Ann. 19591960, 262270. Severinov, K., Soushko, M., Goldfarb, A., and Nikiforov, A. (1993). Rifampicin region revisited. New rifampicin-resistant and streptolydigin-resistant mutants in the beta subunit of Escherichia coli RNA polymerase. J. Biol. Chem. 268, 1482014825. Severinov, K., Soushko, M., Goldfarb, A., and Nikiforov, V. (1994). RifR mutations in the beginning of the Escherichia coli rpoB gene. Mol. Gen. Genet. 244, 120126. Shinnick, T., ed. (1996). Current Topics in Microbiology and Immunology (New York: Springer-Verlag). Sweetser, D., Nonet, M., and Young, R.A. (1987). Prokaryotic and eukaryotic RNA polymerases have homologous core subunits. Proc. Natl. Acad. Sci. USA 84, 11921196. Wallace, A.C., Laskowski, R.A., and Thornton, J.M. (1995). LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Engineering 8, 127134. Wehrli, W., Neusch, J., Knu sel, F., and Staehelin, M. (1968b). Action of rifamycin on RNA polymerase from sensitive and resistant bacteria. Biochem. Biophys. Res. Commun. 32, 284288. Wichelhaus, T.A., Schafer, V., Brade, V., and Boddinghaus, B. (1999). Molecular characterization of rpoB mutations conferring crossresistance to rifamycins on methicillin-resistanc Staphylococcus aureus. Antimicrob. Agents and Chemotherapeutics 43, 28132816. Zhang, G., Campbell, E.A., Minakhin, L., Richter, C., Severinov, K., and Darst, S.A. (1999). Crystal structure of Thermus aquaticus core resolution. Cell 98, 811824. RNA polymerase at 3.3 A Protein Data Bank ID Code The Rif-RNAP coordinates have been deposited in the Protein Data Bank under ID code 1I6V.