Structures of the Escherichia coli ribosome withantibiotics bound near the peptidyl transferasecenter explain spectra of drug actionJack A. Dunklea, Liqun Xiongb, Alexander S. Mankinb, and Jamie H. D. Catea,c,1

aDepartments of Molecular and Cell Biology and Chemistry, University of California, Berkeley, CA 94720; bCenter for Pharmaceutical Biotechnology,University of Illinois, Chicago, IL 60607; and cPhysical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720

Edited by Peter Moore, Yale University, New Haven, CT, and approved July 22, 2010 (received for review June 18, 2010)

Differences between the structures of bacterial, archaeal, andeukaryotic ribosomes account for the selective action of antibiotics.Even minor variations in the structure of ribosomes of differentbacterial species may lead to idiosyncratic, species-specific interac-tions of the drugs with their targets. Although crystallographicstructures of antibiotics bound to the peptidyl transferase centeror the exit tunnel of archaeal (Haloarcula marismortui) and bacter-ial (Deinococcus radiodurans) large ribosomal subunits have beenreported, it remains unclear whether the interactions of antibioticswith these ribosomes accurately reflect those with the ribosomesof pathogenic bacteria. Here we report X-ray crystal structures ofthe Escherichia coli ribosome in complexes with clinically importantantibiotics of four major classes, including the macrolide erythro-mycin, the ketolide telithromycin, the lincosamide clindamycin,and a phenicol, chloramphenicol, at resolutions of ∼3.3 Å–3.4 Å.Binding modes of three of these antibiotics show important varia-tions compared to the previously determined structures. Biochem-ical and structural evidence also indicates that interactions oftelithromycin with the E. coli ribosome more closely resemblesdrug binding to ribosomes of bacterial pathogens. The presentdata further argue that the identity of nucleotides 752, 2609,and 2055 of 23S ribosomal RNA explain in part the spectrumand selectivity of antibiotic action.

ribosome structure ∣ erythromycin ∣ telithromycin ∣ clindamycin ∣chloramphenicol

Antibiotics are small organic molecules synthesized by fungiand bacteria that can inhibit the growth of other microorgan-

isms (1). The ribosome is a major target of antibiotics, whichaffect nearly all steps of protein synthesis (2). The peptidyl trans-ferase center (PTC) of the ribosome is inhibited by a chemicallydiverse group of compounds including lincosamides and pheni-cols (Fig. 1 A–C) (3). Despite the chemical dissimilarities of thesecompounds they share overlapping mechanisms of inhibition bypreventing proper orientation of tRNA in the PTC and interfer-ing with peptide bond formation. Another important group ofdrugs, macrolides and their modern ketolide derivatives, bindin the exit tunnel of the large ribosomal subunit and inhibit ex-trusion of the nascent peptide, leading to peptidyl tRNA drop-off(Fig. 1 B and C) (4–6).

Bacterial pathogens have become resistant to antibiotics thatinhibit protein synthesis over decades of clinical use of thesecompounds. One of the major mechanisms of resistance is basedon altering ribosomal RNA (rRNA) in the drug binding site. Forexample, resistance to the macrolide erythromycin is oftenmediated by mutations of nucleotide A2058 in 23S rRNA locatedin the site of drug action or by chemical modification of this nu-cleotide by Erm methyltransferase, a gene that is often acquiredby bacterial pathogens (Fig. 1B) (7, 8). Remarkably, similarmutations often provide resistance to multiple protein synthesisinhibitors that bind to the overlapping sites in ribosome. Forexample, mutations of 23S rRNA nucleotides 2058 or 2059 canprovide resistance to macrolides, lincosamides, streptogramin B,

or ketolides, yielding the so called MLSBK phenotype. Mutationof nucleotide 2057 can provide resistance to the MLSBK antibio-tics plus chloramphenicol (Fig. 1B) (8, 9). Other mutations (atpositions 2452, 752, and 2611) described in several species canconfer resistance to varying subsets of these compounds (Fig. 1B)(9, 10).

Structural studies have greatly advanced our understanding ofthe inhibitory mechanisms of antibiotics that bind in the PTC orexit tunnel. However, uncertainty concerning the interactions ofthese compounds with rRNA and resistance phenotypes persist,due to many significant disagreements between the reports forantibiotics bound to either the Deinococcus radiodurans orHaloarcula marismortui 50S ribosomal subunits (11). These dif-ferences include the conformation of the macrolide ring, the con-formation of the alkyl-aryl arm of telithromycin, the orientationof the pyrrolidinyl moiety of clindamycin and the two nonover-lapping binding sites observed for chloramphenicol (6, 12–15).Notably, neither H. marismortui nor D. radiodurans are closelyrelated to pathogenic bacteria.

To address the differences that persist between the H. maris-mortui and D. radiodurans structural data, we solved structures offour antibiotics bound to theEscherichia coli ribosome: erythromy-cin, telithromycin, clindamycin and chloramphenicol, at resolu-tions of ∼3.3 Å–3.4 Å (Table S1). Because the E. coli ribosomehas rRNA sequences similar to bacterial species of medical inter-est, these data give a more accurate picture of the interactions be-tween antibiotics and the large ribosomal subunit of pathogenicbacteria. Together with biochemical data probing the interactionsof antibiotics with the PTC and exit tunnel, the present structuresreveal how rRNA sequence differences contribute to the spectrumof activity for antibiotics and offer new clues as to why these com-pounds do not inhibit cytoplasmic eukaryotic ribosomes.

ResultsThe Structure of Erythromycin Bound to the E. coli Ribosome.The ori-ginal macrolide antibiotic in clinical use since the 1950s, erythro-mycin is composed of a 14-membered macrolactone ring, withcarbohydrates at positions 3 and 5 (Fig. 1A). The desosaminesugar at position 5, which contains a dimethyl amine that iscrucial for binding to the ribosome, makes contact with A2058

Author contributions: J.A.D., A.S.M., and J.H.D.C. designed research; J.A.D. and L.X.performed research; J.A.D., L.X., A.S.M., and J.H.D.C. analyzed data; and J.A.D., A.S.M.,and J.H.D.C. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The coordinates for the structural models have been deposited in theProtein Data Bank, www.pdb.org [PDB ID codes 3OFO, 3OFP, 3OFR, 3OFQ (70S ribosomein complex with erythromycin), 3OAQ, 3OAR, 3OAS, 3OAT (70S ribosome in complex withtelithromycin), 3OFX, 3OFY, 3OFZ, 3OG0 (70S ribosome in complex with clindamycin), and3OFA, 3OFB, 3OFC, 3OFD (70S ribosome in complex with chloramphenicol)].

See Commentary on page 17065.1To whom correspondence should be addressed. E-mail: jcate@lbl.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1007988107/-/DCSupplemental.

17152–17157 ∣ PNAS ∣ October 5, 2010 ∣ vol. 107 ∣ no. 40 www.pnas.org/cgi/doi/10.1073/pnas.1007988107

Dow

nloa

ded

by g

uest

on

May

20,

202

0

in 23S rRNA, the most commonly mutated nucleotide in resistantbacteria. The 14-atom macrolactone ring of erythromycin servesas the scaffold for several semisynthetic compounds in clinicaluse, with various appendages attached to the ring. The macrolac-tone ring was initially reported in two different conformations,“folded-in” and “folded-out” when bound to D. radioduransand H. marismortui 50S subunits, respectively (12). However,the existence of the folded-in conformation for erythromycinwhen bound to the ribosome has been questioned because a pu-tatively lower energy folded-out conformation of erythromycinexists in the crystal structure of the free compound (6).

In the structure of erythromycin bound to the E. coli 70S ribo-some, we observed difference electron density for the drug in ex-cellent agreement with its position bound to the H. marismortui50S ribosomal subunit containing a G2058A mutation (6) andwith its conformation in the crystal structure of the free com-pound (Fig. S1) (16). A slight movement of the antibiotic relativeto its position when bound to the G2058A mutant H. marismortui50S subunit is visible, possibly due to a movement of rRNA helixH73 and the adjacent nucleotides in the E. coli ribosome, relativeto their position in the H. marismortui 50S subunit (Fig. S1B). Inspite of this spatial translocation, the drug maintains its contactswith A2058, which involves a hydrogen bond between the deso-samine hydroxyl and the N1 atom of A2058, and tight packing ofthe hydrophobic face of the lactone ring against nucleotides 2611and 2057 in the peptide exit tunnel wall. Notably, H. marismortuicontains a G-C base pair C2057-G2611 in the opposite polarity ofthe E. coli base pair G2057-C2611, but this seems not to affect theinteraction with erythromycin significantly (Fig. 1C and Fig. S1B).

A Key Moiety of Telithromycin Forms Species-Specific Contacts torRNA. Telithromycin belongs to the family of ketolide antibioticsthat represent the newest generation of macrolides. In telithro-mycin, a carbonyl group replaces the C3 cladinose sugar (Fig. 1A),which in macrolides is necessary for ribosome stalling and regu-lating the induction of resistance genes (17). Similar to otherclinically relevant ketolides, telithromycin contains an alkyl-arylarm attached to a carbamate heterocycle that involves the C11and C12 positions of the ketolide macrocycle. This moiety in-creases the affinity of the ketolide scaffold for the ribosome byseveral hundred-fold, demonstrating that it is an important phar-macophore (18). The alkyl-aryl arm of telithromycin was seen intwo distinct conformations in prior ribosome crystal structures. Itwas folded back over the macrolactone ring when bound to theG2058A H. marismortui 50S subunit, or interacting with rRNAfurther down the peptide exit tunnel when bound to the D. radio-durans 50S subunit (Fig. 2) (6, 13). Importantly, neither structurecan easily explain telithromycin resistance mutants at residueU2609, telithromycin protection of A752 in RNA footprinting

Fig. 1. Binding sites for antibiotics in the PTC and peptide exit tunnel. (A)The chemical structures of erythromycin (a macrolide), telithromycin (a keto-lide), chloramphenicol (a phenyl propanoside), and clindamycin (a lincosa-mide) are shown. (B) An overview of the antibiotic binding sites within the50S subunit. Erythromycin (green), telithromycin (pink), clindamcyin (purple),and chloramphenicol (orange) are shown as stick models. Ribbons denotethe sugar phosphate backbone of 23S rRNA (gray) with nucleotides of inter-est colored light blue, the acceptor ends of A-site tRNA (yellow) and P-sitetRNA (red). The location of the peptide exit tunnel is labeled “exit,” andan icon indicating the point of view is shown on the right. (C) The secondarystructure of the 3′ region of 23S rRNA showing elements of the PTC and theadjacent peptide exit tunnel. Nucleotides that are divergent between E. coliand H. marismortui are shown in red. D. radiodurans diverges from E. coliat the 2057-2611 base pair and at nucleotides 752 and 2586. RibosomalRNA helices emanating from this region are marked with dotted lines.

Fig. 2. Telithromycin bound to the E. coli ribosome. A comparison of theconformations reported for telithromcyin bound to the ribosome. 23S rRNAfor E. coli is shown in gray. Telithromycin models from H. marismortui (gold),D.radiodurans (cyan), and E. coli (pink) are shown. Nitrogens in the alkyl-arylarm are also shown for reference.

Dunkle et al. PNAS ∣ October 5, 2010 ∣ vol. 107 ∣ no. 40 ∣ 17153

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

May

20,

202

0

experiments, or the fact that deletion of A752 or mutations in itsvicinity lead to resistance (9, 19–22).

In the E. coli ribosome, in contrast to H. marismortui orD. radiodurans, nucleotides A752 and U2609 form a base pairthat bridges domains II and V in 23S rRNA. Notably, in the struc-ture of telithromycin bound to the E. coli ribosome, the alkyl-arylarm stacks on the A752-U2609 base pair, a conformation notobserved in prior structures (Fig. 2 and Fig. S2A). By contrast,the position of the macrolactone ring remains in essentiallythe same conformation observed when telithromycin is boundto the G2058A H. marismortui 50S subunit (Fig. 2) (6, 13).The A752-U2609 base pair, which exists in E. coli and many othereubacteria but not inH. marismortui orD. radiodurans (Fig. S2B),provides a surface for the entire face of the alkyl-aryl arm to en-gage in a stacking interaction that likely favors drug binding.

Stacking of the alkyl-aryl arm of telithromycin on the A752-U2609 base pair would also likely lead to protection of A752 fromchemical probes by stabilizing the base pair interaction. To testthis model, we probed E. coli ribosomes in solution by RNAfootprinting for telithromycin protections of this base pair. Wealso examined Halobacterium halobium (a close relative ofH. marismortui), D. radiodurans and Staphylococcus aureus ribo-somes. Binding of telithromycin to E. coli or S. aureus ribosomesprotected A752 from chemical modification (Fig. 3). By contrast,binding of telithromycin to D. radiodurans and H. halobium ribo-somes did not protect the corresponding nucleotide, althoughboth compounds were bound in all cases (Fig. 3). Given the se-quence conservation of the A752-U2609 base pair among manyeubacteria, the structural and biochemical data obtained heresuggest that the interactions of telithromycin with the E. coli ri-bosome likely reflect those that occur when telithromycin bindsthe ribosomes of medically relevant eubacterial species. Further-more, the interaction between the alkyl-aryl arm and the A752-U2609 base pair helps to explain why deletion of A752 or muta-tions of U2609 provide resistance to telithromycin (9, 20, 22).

Clindamycin Sterically Clashes with A-site tRNA. Clindamycin is asemisynthetic derivative of the lincosamide class of compoundsused clinically to treat gram-positive bacterial infections. Struc-tural models of clindamycin bound to both the G2058A H. mar-ismortui andD. radiodurans 50S ribosomal subunits (6, 12) agreedroughly in the placement of the galactose sugar of clindamycinbut differed in the positioning of the propyl pyrrolidinyl moietyof the antibiotic, the portion that is hypothesized to interfere withA-site tRNA positioning (6). In the models of clindamycin boundto the H. marismortui and D. radiodurans 50S subunits, thepyrrolidinyl propyl group is rotated by 180° in one structure re-lative to the other (Fig. 4A). Structural studies of the ribosomecontaining aminoacylated A-site and P-site tRNAs, or oligonu-

cleotide mimics of the tRNAs, show that the pyrrolidinyl propylgroup, as modeled in theH. marismortui complex, would interferewith the positioning of A-site aminoacyl-tRNA (23). Unbiaseddifference electron density maps for clindamycin bound to theE. coli ribosome show density for the propyl group consistent withits position in theH. marismortui structure (Fig. 4 A and B). Thus,clindamycin and other lincosamides likely interfere with A-siteaminoacyl-tRNA binding, as proposed (6).

The location of the galactose ring, which forms numeroushydrogen bonds with A2058, A2059, G2505, and A2503 in boththe G2058AH. marismortui andD. radiodurans 50S subunit struc-tures, overlaps with the binding site of the desosamine sugar ofmacrolides and ketolides, explaining why lincosamides share re-sistance mutations with these antibiotic families. In the structureof the E. coli ribosome with clindamycin bound, difference elec-tron density for the galactose ring agrees with the prior structuralmodels for placement of the sugar (Fig. 4). The interactions ofclindamycin with 23S rRNA nucleotides A2058 and A2503 inE. coli are identical to those observed with the correspondingH. marismortui nucleotides. However, nucleotides 2504–2507are shifted relative to their positions in the G2058A H. marismor-tui 50S subunit, leading to alternative interactions with clindamy-cin (Fig. 4C). The shift probably is due to the presence of A2055in H. marismortui versus C2055 in E. coli, which stacks againstU2504 in the structure. This stacking interaction, together withdifferences in the base pair partners of G2505 and U2609, altersthe conformations that nucleotides in this region adopt in E. coliribosomes. In the unbiased difference electron density map, astrong positive peak indicates a shift of G2505 and U2506 towardthe drug molecule, with U2506 packing perpendicular to the pro-pyl group of clindamycin. Additional packing with C2452 leads todesolvation of the hydrophobic propyl group (Fig. S3). Thealtered position of these nucleotides in E. coli ribosomes relativeto their position in H. marismortui 50S subunits allows the brid-ging amine of clindamycin to hydrogen bond to the ribose O4′ ofG2505, while preserving contacts with A2058, A2059 and A2503(Fig. 4 C and D). Thus, a comparison of the structures of clinda-mycin bound to the G2058AH. marismortui and E. coli ribosomessuggests that clindamycin forms a similar number of hydrogenbonds to rRNA in the E. coli structure, with increased van derWaals contacts with the ribosome.

Chloramphenicol Positioning In and Near the 50S Subunit A-site. Priorstructural studies identified one binding site for chloramphenicolin the D. radiodurans 50S subunit and a second distinct site in theH. marismortui 50S subunit (12, 15). The differing positions ofchloramphenicol are probably due to the fact that wild typeH. marismortui is resistant to chloramphenicol, possibly due tothe low affinity of the drug for its primary binding site (10). This

Fig. 3. Telithromycin and erythromycin footprints on 23S rRNA reveal species-specific binding modes. Interactions of antibiotics erythromycin (E) or telithro-mycin (T) in solution with ribosomes prepared from archaea (H. halobium) or three different bacterial species (D. radiodurans, E. coli, or S. aureus) as revealedby dimethylsulfate (DMS) probing. (Upper) Protections of A2058 and A2059 by both antibiotics to all the tested ribosomes. (Lower) Protection of A752 fromDMSmodification by telithromycin in E. coli or S. aureus ribosomes, but not in either H. halobium or D. radiodurans ribosomes. Lanes on the gels are labeled asfollowing: A, an A-specific sequencing reaction; K, unmodified ribosome; 0, ribosome modified with DMS in the absence of antibiotics; E and T, ribosomemodified with DMS after preincubation with 50 μM of erythromycin or telithromycin, respectively.

17154 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1007988107 Dunkle et al.

Dow

nloa

ded

by g

uest

on

May

20,

202

0

intrinsic tolerance of archaea to chloramphenicol is likely due torRNA sequence differences between archaea and bacteria; i.e. innucleotides C2055 of H73, and U2609, and C2610 adjacent to thebase of rRNA helix H73 (Fig. 1C).

In the structure of chloramphenicol bound to the E. coli ribo-some, we did not observe any difference electron density in thechloramphenicol binding site identified in theH. marismortui 50Ssubunit (24). By contrast, clear positive difference electron den-sity is located at the site of chloramphenicol binding reported intheD. radiodurans 50S subunit, but this density is not entirely con-sistent with the reported orientation of chloramphenicol (Fig. 5 Aand B) (12). In the E. coli ribosome complex, the nitrobenzenemoiety of chloramphenicol overlaps with the position that wouldbe occupied by the pyrrolidinyl propyl group of clindamycin(Figs. 1B, 4C, and 5B). When compared to the structure ofclindamycin bound to the E. coli ribosome, the base of U2506 isrotated such that the hydrophilic nitro group in chloramphenicolpoints into solvent. The nitrobenzene ring of chloramphenicol isstacked on C2452 (Fig. 5B) and positions one chlorine atom of

chloramphenicol in a similar position to the chlorine in clindamy-cin (Fig. 4C). The other chlorine atom in chloramphenicol is in aposition to contact the exocyclic amine of A2062, a residue thatwhen mutated results in chloramphenicol resistance (10).Finally, the amine of chloramphenicol hydrogen bonds to thenonbridging phosphate oxygen of G2505 (Fig. 5B).

Because rRNA sequences in the region of chloramphenicolbinding are conserved in both E. coli and D. radiodurans, se-quence divergence cannot explain the different orientations ofthe antibiotic. Unless the second shell rRNA residues dramati-cally affect chloramphenicol binding, there is likely one principalmode of interaction between chloramphenicol and eubacterialribosomes (25). The unbiased difference electron density inthe E. coli structure determination provides a much clearer viewof details of chloramphenicol orientation than is provided by theelectron density used to model chloramphenicol bound to the D.radiodurans 50S subunit (compare Fig. 5A with Fig. 1C of ref. 12).Furthermore, the structural model of chloramphenicol boundto the E. coli ribosome is at higher resolution and with a lowerRfree value (25%) when compared to the D. radiodurans model(Rfree ¼ 32%) (Table S1). Finally, the interactions between chlor-amphenicol and rRNA in the orientation in the present modelare more chemically favorable due to additional stacking interac-tions that should facilitate drug binding. Notably, in the presentstructural model of chloramphenicol bound to the E. coli 70Sribosome, the nitrobenzene ring stacks on C2452 in the sameorientation that the methoxyphenyl ring of the related compoundanisomycin stacks on C2452 when bound to the H. marismortui50S subunit (26).

Fig. 4. Clindamycin bound to the E. coli ribosome. (A) The conformations ofclindamycin bound to the ribosome. The structural model of clindamycinbound to the D. radiodurans 50S subunit has the pyrrolydinyl propyl group(marked with an asterisk, and showing nitrogen atoms for reference) rotatedby 180 degrees. (B) Unbiased difference electron density (Fobs-Fcalc) for clin-damycin bound to the E. coli ribosome. The electron density is contoured at3.5 standard deviations from themean. (C) E. coli 23S rRNA is superpositionedwith H. marismortui 23S rRNA to reveal that some nucleotides (2504–2507)are shifted in space due to the sequence difference at 2055, C in E. coli,A in H. marismortui. E. coli nucleotides are in gray, H. marismortui in yellow.(D) A diagram of the interactions of clindamycin with the H. marismortui(Left) and E. coli (Right) ribosomes. Dashed lines represent hydrogen bonds,and curves represent key van der Waals interactions.

Fig. 5. Chloramphenicol bound to the E. coli ribosome. (A) Unbiased differ-ence electron density (Fobs-Fcalc) contoured at 3.5 standard deviations fromthe mean for chloramphenicol bound to the E. coli ribosome (Left). At right,the structure of chloramphenicol in the context of E. coli rRNA is shown(orange) compared to the structure of chloramphenicol reported boundto D. radiodurans ribosomal subunits (cyan). (B) Chloramphenicol (orange)bound to the A-site cleft of the E. coli ribosome. Hydrogen bonds areshown as black dashes. Nucleotides that are sites of resistance mutationsto chloramphenicol are labeled in red.

Dunkle et al. PNAS ∣ October 5, 2010 ∣ vol. 107 ∣ no. 40 ∣ 17155

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

May

20,

202

0

DiscussionThe rRNA of the PTC is highly conserved throughout the 3domains of life, explaining why many antibiotics that bind inor near the PTC can inhibit a wide range of species. However,many antibiotics that bind in this region are selective. For exam-ple erythromycin does not bind to eukaryotic or archaeal ribo-somes except at extremely high concentrations. The identity ofnucleotide 2058, adenosine in eubacteria, guanosine in eukar-yotes and archaea, is thought to be a major contributor to thisphenomenon (6). Conversely, wild type eubacteria are resistantto anisomycin whereas archaea and eukaryotes are sensitive(26, 27).

Our data reveal several rRNA residues that can account forthe selectivity of antibiotic action. In the E. coli ribosome, 23SrRNA residues A752 and U2609 form a base pair that interactswith a key element of telithromycin, its extended 11,12 alkyl-arylarm (Fig. 3 A and B). Our data suggest that disruption of this basepair would decrease the affinity of telithromycin for the ribosome,consistent with mutation U2609C or deletion of A752 that lead tolow level telithromycin resistance (20, 22). In D. radiodurans andH. marismortui, this base pair does not exist, leading to a differentconformation of the alkyl-aryl side chain (6, 13) and likelydecreasing telithromycin affinity for the ribosome (Fig. 4).

Nucleotide C2055 in eubacteria is an adenosine in archaea andeukaryotes (Table S2) (28), and also seems to serve as an impor-tant determinant for the spectrum of action of A-site inhibitorssuch as chloramphenicol and clindamycin, as well as other anti-biotics (29, 30). Our structures reveal that alteration of A2055,which stacks on U2504 in the H. marismortui 50S subunit struc-tures, to C2055 in E. coli leads to a displacement of U2504 alongwith G2505, U2506, and C2507 (Figs. 1B and 4C). Thus the se-quence difference at C2055 results in a change in the position offour universally conserved PTC nucleotides (Table S2). The struc-tures of the E. coli ribosome show that chloramphenicol and clin-damycin both interact with these nucleotides. G2505 hydrogenbonds to clindamycin whereas U2506 desolvates the propyl groupof clindamycin (Fig. 4 C and D). G2505 forms hydrogen bonds tochloramphenicol, interactions that are not possible in H. maris-mortui, given the conformation of this residue in H. marismortuiribosomal subunits (Fig. 4B). These structural data explain whyarchaeal and eukaryotic ribosomes bind lincosamides and chlor-amphenicol poorly. In fact, difference electron density for chlor-amphenicol is not observed in the eubacterial binding site evenwhen H. marismortui ribosomal subunits are soaked in millimolarconcentrations of the compound (15). In addition, the mutationC2055A in eubacterial ribosomes increases the minimal inhibi-tory concentration (MIC) of chloramphenicol and clindamycin(31). Furthermore, the C2055A mutation in M. smegmatis ribo-somes leads to resistance to pleuromutilin antibiotics, which alsobind the eubacterial PTC (32).

The proposed role of 23S rRNA nucleotide 2055 as an anti-biotic resistance spectrum determinant through its interactionwith A-site cleft nucleotides 2504 and 2506–2507 suggest thatmutations of these residues should also appear in mutants resis-tant to chloramphenicol and lincosamides. Despite the sensitivelocation of these nucleotides in the PTC and their universalconservation, mutations of nucleotide 2504 confer resistanceto chloramphenicol (31, 33) and increase the MIC of clindamcyin(31). The lack of posttranscriptional modification (pseudouridi-nylation) at this position renders bacteria hypersensitive toseveral peptidyl transferase inhibitors (34).

Mutation of G2058A in archaeal ribosomes leads to a ∼104-fold improvement in macrolide binding to H. marismortui orH. halobium 50S subunits, suggesting that the identity of a singlenucleotide could explain the insensitivity of archaeal and eukar-yotic ribosomes toward macrolides (6, 35). However, in S. cere-visiae ribosomes carrying the G2058Amutation in their ribosomalRNA were still not inhibited by erythromycin, suggesting addi-

tional phylogenetic differences contribute to resistance (36).Taken together with phylogenetic data, the present structuraland biochemical data suggest that, in addition to position2058, divergence of rRNA at positions 752, 2609, 2610, and2055 also contribute to the species specificity of PTC and peptideelongation inhibitors. The availability of high-resolution struc-tures of the ribosome and ribosomal subunits from archaeaand divergent bacteria will now make it possible to probe theseand other sequence determinants of PTC and peptide elongationinhibitors.

Experimental ProceduresRibosome Purification, Crystallization, and Antibiotic Soaking Experi-ments. Ribosomes were purified from MRE600 E. coli cells asdescribed previously (37). Ribosome crystals were grown asdescribed previously (38). For antibiotic soaking experiments,the cryoprotection buffer was supplemented with 48 μM of teli-thromcyin or 100 μM of either erythromycin, chloramphenicol orclindamycin. Erythromycin, chloramphenicol and clindamycinwere obtained from Sigma. Telithromycin was a gift from CempraPharmaceuticals. The telithromycin stock solution in 50% aceticacid was diluted ∼800-fold in cryoprotection buffer for soaks.Erythromycin and chloramphenicol stock solutions were made inethanol, and diluted 100-fold to working concentrations in cryo-protection buffer. Clindamycin was dissolved in DMSO and di-luted 100× in cryoprotection buffer for soaking experiments.In all four cases incubation of the crystals with antibiotic tookplace overnight. Crystals were then flash frozen in liquid nitrogen.

X-ray Diffraction Experiments, Model Building, and Figure Prepara-tion. X-ray diffraction data were measured at beamlines 8.3.1and 12.3.1 of the Advanced Light Source, Lawrence BerkeleyNational Laboratory, using 0.1–0.3 ° oscillations at 100 K andrecorded on an ADSC Q315 detector. X-ray reflections were re-duced and scaled using HKL2000. Difference electron density(Fobs-Fcalc) maps were calculated using the phenix.refine moduleof the PHENIX software suite (39). Antibiotics were positionedbased on the unbiased difference electron density obtained ineach experiment. If necessary, changes in the coordinates forrRNA nucleotides dictated by the difference electron densitywere made using Coot (40) and phenix.refine was used for posi-tional and atomic displacement parameter (ADP) refinement ofthe model. The crystals contain two copies of the ribosome in theasymmetric unit, with molecule 1 (PDB coordinates 3OAT,3OFR, 3OFC, 3OFZ for the 50S subunit in complex with telithro-mycin, erythromycin, chloramphenicol, and clindamycin, respec-tively) exhibiting lower ADP values and clearer electron densitythan molecule 2. Whereas difference electron density for antibio-tics was observed in both molecule 1 and molecule 2 of the crystalin each experiment, coordinates for antibiotics were only mod-eled into molecule 1. Ribosomal RNA sequence alignments weremade using MUSCLE (41) or obtained from the ComparativeRNA Website (28). Coordinate superpositions were performedin PyMOL using the C1′ atoms of domain V of 23S rRNA,excluding regions where insertions or deletions occur. Figureswere made using PyMOL (42).

Preparation of Ribosomes and Ribosomal Subunits for RNA Probing.Ribosomes and large ribosomal subunits were isolated fromE. coli (strain JM109) or S. aureus (strain RN4220) using stan-dard protocols described in ref. 43. Large ribosomal subunitsprepared from D. radiodurans were a gift of Francheschi andSkripkin (Rib-X, Inc). To increase affinity of macrolides forthe archaeal large ribosomal subunits, G2058 in the 23S rRNAgene (E. coli numbering) of the single ribosomal operon ofH. halobium was mutated to A. This mutation was introducedtogether with the selective anisomycin resistance mutation

17156 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1007988107 Dunkle et al.

Dow

nloa

ded

by g

uest

on

May

20,

202

0

C2452U using the procedure described in ref. 44. Haloarchaealribosomes were isolated as previously described (45).

Chemical Probing of the Interaction of Erythromycin and Telithromy-cin with Large Ribosomal Subunits. Chemical probing of interac-tions of erythromycin and telithromycin with bacterial andhaloarchaeal large ribosomal subunits was carried out using pro-cedures described in ref. 46 and in ref. 47 with ribosomes anddrugs present at 200 nM and 50 μM, respectively. Ribosomalsubunits, vacant or complexed with antibiotics, were modified

with dimethylsulfate and the distribution of modified residueswas analyzed by primer extension.

ACKNOWLEDGMENTS. We thank K. Frankel, S. Classen, and G. Meigs for helpwith data measurement at the SIBYLS and 8.3.1 beamlines at the AdvancedLight Source; P. Afonine and P. Adams for help with Phenix refinement; andS.-H. Kim, P. Adams, and J. Holton for useful crystallographic discussions.This work was funded by the National Institutes of Health GrantGM65050 (J.H.D.C.) and National Cancer Institute Grant CA92584 for theSIBYLS and 8.3.1 beamlines, by the National Science Foundation GrantMCB-0824739 (A.S.M.), and by the Department of Energy Grant DE-AC0376SF00098 for the SIBYLS and 8.3.1 beamlines.

1. Cundcliffe E (2000) Antibiotic biosynthesis: Some thoughts on “why” and “how”.The Ribosome:Structure, Function, Antibiotics and Cellular Interactions, eds RA Gar-rett, A Liljas, AT Matheson, PB Moore, and HF Noller (ASM Press, Washington, DC),pp 409–417.

2. Sohmen D, Harms JM, Schluenzen F, Wilson DN (2009) SnapShot: Antibiotic inhibitionof protein synthesis II. Cell 139 212.e211.

3. Polacek N, Mankin AS (2005) The ribosomal peptidyl transferase center: Structure,function, evolution, inhibition. Crit Rev Biochem Mol Biol 40:285–311.

4. Vester B, Douthwaite S (2001) Macrolide resistance conferred by base substitutions in23S rRNA. Antimicrob Agents Chemother 45:1–12.

5. Mankin AS (2008) Macrolide myths. Curr Opin Microbiol 11:414–421.6. Tu D, Blaha G, Moore PB, Steitz TA (2005) Structures of MLSBK antibiotics bound

to mutated large ribosomal subunits provide a structural explanation for resistance.Cell 121:257–270.

7. Poehlsgaard J, Douthwaite S (2005) The bacterial ribosome as a target for antibiotics.Nat Rev Microbiol 3:870–881.

8. Ettayebi M, Prasad SM, Morgan EA (1985) Chloramphenicol-erythromycin resistancemutations in a 23S rRNA gene of Escherichia coli. J Bacteriol 162:551–557.

9. Canu A, et al. (2002) Diversity of ribosomal mutations conferring resistance to macro-lides, clindamycin, streptogramin, and telithromycin in Streptococcus pneumoniae.Antimicrob Agents Chemother 46:125–131.

10. Mankin AS, Garrett RA (1991) Chloramphenicol resistance mutations in the single23S ribosomal RNA gene of the archaeon Halobacterium halobium. J Bacteriol173:3559–3563.

11. Wilson DN, Harms JM, Nierhaus KH, Schlunzen F, Fucini P (2005) Species-specificantibiotic–ribosome interactions: Implications for drug development. Biol Chem386:1239–1252.

12. Schluenzen F, et al. (2001) Structural basis for the interaction of antibiotics with thepeptidyl transferase centre in eubacteria. Nature 413:814–821.

13. Berisio R, et al. (2003) Structural insight into the antibiotic action of telithromycinagainst resistant mutants. J Bacteriol 185:4276–4279.

14. Hansen JL, et al. (2002) The structures of four macrolide antibiotics bound to the largeribosomal subunit. Mol Cell 10:117–128.

15. Hansen JL, Moore PB, Steitz TA (2003) Structures of five antibiotics bound at thepeptidyl transferase center of the large ribosomal subunit. J Mol Biol 330:1061–1075.

16. Stephenson G, Stowell J, Toma P, Pfeiffer R, Byrn S (1997) Solid-state investigations oferythromycin A dihydrate: Structure, NMR spectroscopy, and hygroscopicity. J PharmSci 86:1239–1244.

17. Vazquez-Laslop N, Thum C, Mankin AS (2008) Molecular mechanism of drug-dependent ribosome stalling. Mol Cell 30:190–202.

18. Hansen LH, Mauvais P, Douthwaite S (1999) The macrolide-ketolide antibiotic bindingsite is formed by structures in domains II and V of 23S ribosomal RNA. Mol Microbiol31:623–631.

19. Xiong L, Shah S, Mauvais P, Mankin AS (1999) A ketolide resistancemutation in domainII of 23S rRNA reveals the proximity of hairpin 35 to the peptidyl transferase centre.Mol Microbiol 31:633–639.

20. Garza-Ramos G, Xiong L, Zhong P, Mankin A (2001) Binding site of macrolide antibio-tics on the ribosome: New resistance mutation identifies a specific interaction ofketolides with rRNA. J Bacteriol 183:6898–6907.

21. Xiong L, Korkhin Y, Mankin AS (2005) Binding site of the bridged macrolides in theEscherichia coli ribosome. Antimicrob Agents Chemother 49:281–288.

22. Novotny GW, Jakobsen L, Andersen NM, Poehlsgaard J, Douthwaite S (2004) Ketolideantimicrobial activity persists after disruption of interactions with domain II of 23SrRNA. Antimicrob Agents Chemother 48:3677–3683.

23. Voorhees RM, Weixlbaumer A, Loakes D, Kelley AC, Ramakrishnan V (2009) Insightsinto substrate stabilization from snapshots of the peptidyl transferase center of theintact 70S ribosome. Nat Struct Mol Biol 16:528–533.

24. Hansen JL, Moore PB, Steitz TA (2003) Structures of five antibiotics bound at thepeptidyl transferase center of the large ribosomal subunit. J Mol Biol 330:1061–1075.

25. Nierhaus D, Nierhaus KH (1973) Identification of the chloramphenicol-bindingprotein in Escherichia coli ribosomes by partial reconstitution. Proc Natl Acad SciUSA 70:2224–2228.

26. Blaha G, Gurel G, Schroeder SJ, Moore PB, Steitz TA (2008) Mutations outside theanisomycin-binding site can make ribosomes drug-resistant. J Mol Biol 379:505–519.

27. Hummel H, Böck A (1987) 23S ribosomal RNA mutations in halobacteria conferringresistance to the anti-80S ribosome targeted antibiotic anisomycin. Nucleic AcidsRes 15:2431–2443.

28. Cannone JJ, et al. (2002) The comparative RNA web (CRW) site: An online database ofcomparative sequence and structure information for ribosomal, intron, and otherRNAs. BMC Bioinformatics 3:2.

29. Davidovich C, et al. (2007) Induced-fit tightens pleuromutilins binding to ribosomesand remote interactions enable their selectivity. Proc Natl Acad Sci USA104:4291–4296.

30. Gürel G, Blaha G, Moore PB, Steitz TA (2009) U2504 determines the species specificityof the A-site cleft antibiotics: the structures of tiamulin, homoharringtonine, andbruceantin bound to the ribosome. J Mol Biol 389:146–156.

31. Long KS, et al. (2009) Single 23S rRNA mutations at the ribosomal peptidyl transferasecentre confer resistance to valnemulin and other antibiotics in Mycobacterium smeg-matis by perturbation of the drug binding pocket. Mol Microbiology 71:1218–1227.

32. Schlünzen F, Pyetan E, Fucini P, Yonath A, Harms JM (2004) Inhibition of peptidebond formation by pleuromutilins: the structure of the 50S ribosomal subunit fromDeinococcus radiodurans in complex with tiamulin. Mol Microbiol 54:1287–1294.

33. Blanc H, Wright CT, Bibb MJ, Wallace DC, Clayton DA (1981) Mitochondrial DNA ofchloramphenicol resistant mouse cells contains a single nucleotide change in theregion encoding the 3′ end of the large ribosomal RNA. Proc Natl Acad Sci USA78:3789–3793.

34. Toh SM, Mankin AS (2008) An indigenous posttranscriptional modification in theribosomal peptidyl transferase center confers resistance to an array of proteinsynthesis inhibitors. J Mol Biol 380:593–597.

35. Böttger EC, Springer B, Prammananan T, Kidan Y, Sander P (2001) Structural basis forselectivity and toxicity of ribosomal antibiotics. EMBO Rep 2:318–323.

36. Bommakanti AS, Lindahl L, Zengel JM (2008)Mutation from guanine to adenine in 25SrRNA at the position equivalent to E. coli A2058 does not confer erythromycinsensitivity in Sacchromyces cerevisae. RNA 14:460–464.

37. Schuwirth BS, et al. (2005) Structures of the bacterial ribosome at 3.5 A resolution.Science 310:827–834.

38. Zhang W, Dunkle JA, Cate JH (2009) Structures of the ribosome in intermediatestates of ratcheting. Science 325:1014–1017.

39. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-molecular structure solution. Acta Crystallogr D 66:213–221.

40. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. ActaCrystallogr D 60:2126–2132.

41. Edgar RC (2004) MUSCLE: Multiple sequence alignment with high accuracy and highthroughput. Nucleic Acids Res 32:1792–1797.

42. Delano WL The PyMOL User’s Manual (DeLano Scientific, San Carlos, CA).43. Moazed D, Noller HF (1989) Interaction of tRNA with 23S rRNA in the ribosomal A,

P, and E sites. Cell 57:585–597.44. Mankin AS, Zyrianova IM, Kagramanova VK, Garrett RA (1992) Introducing mutations

into the single-copy chromosomal 23S rRNA gene of the archaeon Halobacteriumhalobium by using an rRNA operon-based transformation system. Proc Natl AcadSci USA 89:6535–6539.

45. Tan GT, DeBlasio A, Mankin AS (1996) Mutations in the peptidyl transferase center of23S rRNA reveal the site of action of sparsomycin, a universal inhibitor of translation.J Mol Biol 261:222–230.

46. Merryman C, Noller HF (1998) Footprinting and modification-interference analysis ofbinding sites on RNA. RNA:Protein Interactions, A Practical Approach, ed CWJ Smith(Oxford University Press, Oxford).

47. Mankin AS (1997) Pactamycin resistance mutations in functional sites of 16S rRNA.J Mol Biol 274:8–15.

Dunkle et al. PNAS ∣ October 5, 2010 ∣ vol. 107 ∣ no. 40 ∣ 17157

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

May

20,

202

0