Ribosome protection by antibiotic resistance ATP-binding cassette protein

MsrE Rescues AZM Affected Translation.

The msr genes have mostly been identified in staphylococci, streptococci, and enterococci, and have recently spread to P. aeruginosa (18). We have previously observed that the exogenous expression of MsrE protein from its putative promoter significantly increases the azithromycin (AZM; a second-generation derivative of erythromycin) resistance of P. aeruginosa laboratory strains. Furthermore, the expression of MsrE from the arabinose-inducible promoter P_BAD confers AZM resistance to Escherichia coli in a dose-dependent manner (SI Appendix, Table S1). The induction of MsrE expression by 0.2% arabinose increases the minimum inhibitory concentration of AZM by 16-fold compared with the uninduced condition. Overexpression of MsrE does not significantly affect the fitness of E. coli, as revealed by the comparison of growth curves of E. coli/P_BAD-msrE in the presence of 0.2% arabinose and 0.2% glucose (which reduces leaky expression from the arabinose-inducible promoter) (SI Appendix, Fig. S2). Moreover, purified MsrE protein rescues the AZM-inhibited translation process in a dose-dependent fashion, as shown by our in vitro transcription/translation assay (SI Appendix, Fig. S3A). These results clearly support a ribosomal protection mechanism for MsrE and ARE ABC-F proteins in general, which is also in agreement with the recent bacteriological and biochemical studies on VgaA and LsaA proteins (ARE ABC-F members conferring resistance to lincosamides, pleuromutilins, and streptogramin A) (SI Appendix, Fig. S1) (20, 22).

Structure of Ribosome-Bound MsrE.

Purified MsrE protein can bind to Thermus thermophilus ribosome with a stoichiometry of 1:1 (SI Appendix, Fig. S3B). The addition of AZM, tRNA^fMet, and its corresponding mRNA did not significantly affect the binding stoichiometry. Furthermore, the complex was reconstituted in the presence of ATP or its nonhydrolyzable analog AMP-PNP (SI Appendix, Fig. S3B), indicating that ATP hydrolysis is not required for MsrE ribosome binding.

We next determined the cryo-EM structure of the MsrE-ribosome complex with an mRNA and its cognate P-tRNA at 3.6-Å resolution (Fig. 1A and SI Appendix, Fig. S4). Local resolution analysis revealed that the PTC/NPET of the 50S and the decoding center region of the 30S, as well as the tRNA anticodon stem loop and MsrE domain linker region, could be visualized at approximately 3.5-Å resolution. The local resolution was lower for the globular domains of the MsrE protein and the acceptor stem of the tRNA (Fig. 1 B and C), indicating higher structural flexibility.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Cryo-EM structure of the MsrE-ribosome complex. (A) Electron density of the overall complex. Ribosome 50S and 30S subunits are shown in pale orange and cyan, respectively. The electron density of 50S is made partially transparent to reveal the densities corresponding to MsrE protein (green) and tRNA (purple). (B) Local resolution of MsrE and tRNA shown in same orientation as in A. (C) Cryo-EM electron density of the MsrE domain linker extended loop and tRNA acceptor stem region.

As no structures of ARE ABC-F proteins are available, the ribosome-bound MsrE model is of great interest. The MsrE protein has a needle-like overall structure, with two ABC transporter domains (ABC1 and ABC2) carrying highly conserved nucleotide-binding sites (NBSs) (residues 37–44 and 329–336) assembled as its base and a domain linker (residues 180–279) that form two long crossed helices connected by an extended loop) assembled as the needle and tip (Fig. 2A). Both NBSs reveal electron density for the bound nonhydrolyzable ATP analog (AMP-PNP) (Fig. 2 B and C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Structure of the ribosome-bound MsrE protein. (A) MsrE protein shown in cartoon with secondary structure elements of ABC1 (magenta), ABC2 (yellow), and domain linker (green) labeled. The nonhydrolyzable ATP analog AMP-PNP is shown bound to NBSs NBS1 and NBS2. (B and C) Close-up views of the MsrE NBS1 (B) and NBS2 (C) with the cryo-EM electron density map shown for AMP-PNP.

Despite structural similarity to the ribosome-bound ATP form of the non-ARE ABC-F protein EttA (25), ATP hydrolysis appears to play a different role in the functioning of MsrE. ATP hydrolysis is essential for the release of EttA from the ribosomes, as revealed by the finding that its deficient mutant (double substitutions E188Q/E470Q of the catalytic residues) affects cell growth by inhibiting protein synthesis through trapping ribosomes (26). In contrast, the cell growth of E. coli is not significantly affected by overexpression of the corresponding ATP hydrolysis deficient MsrE mutant (E104Q/E413Q) (SI Appendix, Fig. S2). However, the MsrE E104Q/E413Q mutant in vitro ribosome binding (SI Appendix, Fig. S3B) and in vivo AZM resistance (SI Appendix, Table S2) efficiencies are significantly reduced, indicating that ATP hydrolysis is crucial for mediating macrolide resistance.

Analyzing the aforementioned mutations one by one reveals that both residues contribute to antibiotic resistance, but the one in NBS1 (E413Q) has a more significant effect (SI Appendix, Table S2). In line with this, ATPase activity has been reported to be crucial for ARE ABC-F protein VgaA, since mutation of the catalytic glutamines results in abolished antibiotic resistance in vivo (27) and rescue activity in the in vitro translation system (20). It was recently reported that the hydrolytically inactive Staphylococcus haemolyticus VgaA_LC and Enterococcus faecalis LsaA mutants inhibit peptidyl transferase activity of the ribosome in a reconstituted translation system (28).

The long and flexible domain linker is a conserved feature in the ARE ABC-F family and likely provides an explanation for the lack of available structures. The sole identified crystal structure of ABC-F protein is the E. coli EttA (non-ARE) mentioned above (26), which has a significantly shorter linker region. Furthermore, the EttA crystal structure reveals a dimer formation mediated by the linker that might have facilitated the crystallization, yet the monomer state is favored in solution and is likely the active form of EttA (26). MsrE likely also functions as a monomer in cells, as revealed by molecular weight analysis. Surprisingly, in our structure, the entire backbone of the MsrE domain linker can be traced, demonstrating two crossed helices, longer α5, and shorter α6 (referred to as αL and αS, respectively) connected by an extended loop (Fig. 2A), with the majority of the amino acid side chains visualized in the electron density map (Fig. 1C).

MsrE Binds to Ribosomal E-Sites.

MsrE is held in a position known as the ribosomal exit (E) site (Fig. 3A). The ABC1 domain of MsrE faces the L1 stalk of 50S, whereas the ABC2 domain contacts the 30S head (h41 and h42 of 16S rRNA and the r-protein S7), the r-protein L5 in the P-site finger region of 50S, and the elbow of P-tRNA (Fig. 3B). In contrast to the few ribosome contacts observed for the two ABC domains, the elongated domain linker establishes extensive contacts with ribosomes as it stretches in parallel with the acceptor arm of P-tRNA toward the NPET (Fig. 3C). The region at the interface between ABC1 and the linker (foremost the residues Glu191 and His194) contacts the H68 of 23S rRNA, which is involved in ribosomal intersubunit bridge as well as E-site formation and is important for translation activity (29) (Fig. 3C). Starting from the middle of domain linker and heading toward the terminal loop, MsrE αS helix (Lys264-Thr250 region) forms extensive contacts with H74 (A2435-U2438 and C2064-C2065 regions), H80 (G2252-G2253), and H93 (C2601) of 23S rRNA (mainly the backbone) (Fig. 3C). From the P-tRNA side, the residues Pro275 (conserved in the Msr subfamily) and Glu276 at the C terminus of the domain linker interact with the P-site tRNA elbow region (D-loop nucleotides 17–19) (Fig. 3C). The residues Arg217, Lys216, and Gln214 in αL directly contact the acceptor stem of P-tRNA (Fig. 3C). The orientation and detailed interactions of the extended loop in the MsrE domain linker with ribosome are discussed below.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Structure of the MsrE-ribosome complex. (A) Overall structure of the MsrE-ribosome complex. The 50S and 30S subunits are shown in orange and cyan, respectively, and the tRNA and mRNA are in purple and blue, respectively. (B) Interactions of the MsrE ABC1 domain with the 50S subunit L1-stalk (23S rRNA helices H68 and H76) and of the ABC2 domain with the 30S subunit (16S rRNA helices h41 and h42 and r-protein S7), r-protein L5, and tRNA. (C) MsrE domain linker helical region interactions with 50S subunit 23S rRNA and P-site tRNA.

Taken together, the foregoing findings indicate that MsrE interacts with the ribosome mainly through nonspecific interactions, in agreement with its cross-species specificity. Therefore, MsrE can likely function on drug-affected ribosomes across a variety of bacterial species, making it an intriguing target for investigations of antiresistance strategies.

MsrE binding causes dramatic conformational changes in both the ribosome and the P-tRNA. Overall, the 30S subunit of the MsrE-bound ribosome is rotated counterclockwise with respect to the 50S subunit by 2.5 Å, with its head rotated by 3.9° (SI Appendix, Fig. S5A) compared with the post-peptidyl transfer state ribosome (30, 31). Concurrently, MsrE affects the positioning of the P-tRNA; the acceptor stem shifts by 30 Å toward a site usually occupied by the acceptor stem of the fully accommodated A-tRNA (Fig. 4A and SI Appendix, Fig. S5B), and the anticodon stem loop moves by approximately 5 Å coupled to the ribosome rotation (SI Appendix, Fig. S5B) but is seen to maintain the codon–anticodon interaction with its cognate mRNA (SI Appendix, Fig. S5C). Other notable changes include a more open positioning of the L1 stalk and displacement of the N-terminal extension of the r-proteins L27 and L16 from their usual positions in close proximity to the PTC to accommodate the MsrE protein (Fig. 4A). Since the N-terminal extension of L27 is known to be crucial for tRNA binding and ribosome functioning (32⇓–34), the MsrE-induced conformation is likely transient, and the original state is restored on MsrE dissociation from the ribosome to resume normal translation.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Major conformational changes in ribosomes surrounding the MsrE domain linker. (A) Comparison of the acceptor stem of the P-site tRNA (purple), the N-terminal domain of r-protein L27 (forest green), and the r-protein L16 (violet) in the MsrE-ribosome complex and in the post-peptidyl transferase state ribosome (gray) (30). Conformational changes are highlighted with arrows. (B) The shifted 3′ CCA of MsrE-ribosome complex tRNA interacts with the MsrE extended loop. The extended loop residues form bilateral interactions with both the amino acid arm of P-tRNA and H89 of 23S rRNA located at the PTC.

While both MsrE and EttA belong to the ABC-F family and bind to ribosome E-sites, their interactions with the ribosomes vary significantly. MsrE has the elongated domain linker but lacks the insertions called “arm” and “toe” observed in EttA (SI Appendix, Fig. S6A). The EttA arm makes additional contacts with the L1 stalk, resulting in an even more open form than that seen in the present structure (SI Appendix, Fig. S6B). Note that EttA arm has been shown to restrict the ribosome and tRNA dynamics required for translation elongation in response to the availability of ATP (25). The EttA toe region interacts with the r-protein L5 and positions it away from the 30S (SI Appendix, Fig. S6B). The different positions of MsrE and EttA in the ribosome likely reflects their diverse functions.

MsrE Extended Loop Displaces AZM from Ribosome.

As mentioned above, the extended loop of the MsrE domain linker binds deep into the PTC/NPET region and causes deformation of the r-protein L16, the N-terminal extension of the r-protein L27, and the acceptor stem of P-tRNA (Fig. 4A). Unexpectedly, we observed the positions of the acceptor stems of both classical A- and P-tRNA occupied by the domain liker of MsrE (SI Appendix, Fig. S5B). The shifted P-site tRNA acceptor stem is likely stabilized by interactions with MsrE residues Asp224–Lys226 (Fig. 4B). Furthermore, the residues Arg241, Leu242, and His244 at the tip insert deep even beyond the PTC, with their side chains approaching the macrolide-binding site (Fig. 5 A–C). In particular, the Leu242 clashes (within 1.8 Å) with AZM when our structure is compared with that of AZM bound to the ribosome (2, 3) (Fig. 5C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Structure and function of the MsrE domain linker extended loop. (A) The MsrE domain linker extended loop projects into the PTC/NPET region. The 50S subunit is shown in surface representation, and the P-tRNA acceptor stem is shown in cartoon representation. (B) Orientation of the MsrE extended loop and surrounding key PTC residues. (C) Conformational changes in the PTC and macrolide-binding site in the MsrE-ribosome complex compared with the post-peptidyl transfer state ribosome (gray) (30). Conformational changes are indicated by arrows. For reference, the AZM-binding mode is shown based on AZM-ribosome X-ray crystal structure (3). (D) Effect of WT and mutant MsrE proteins on E. coli-derived in vitro transcription/translation assay inhibited by AZM. Results are means of three independent repeats; error bars represent SDs. The uninhibited condition served as a standard. EQ2, MsrE(E104Q/E413Q) mutant. (E) Effect of MsrE on AZM displacement from ribosomes. Ribosomes treated with AZM were incubated with WT and mutant MsrE proteins preincubated with AMP-PNP and ATP, respectively. Ribosomes were pelleted through a sucrose cushion, and AZM was quantified using mass spectrometry. Results are means of three MS assays; error bars represent SDs. The no MsrE addition condition served as a standard.

The 23S rRNA residue A2062, located in the NPET, is believed to relay the drug-induced stalling signal to the PTC, as mutations of this nucleotide can eliminate stalling (35). The base of A2062 protrudes into the tunnel lumen in drug-free ribosomes (30), whereas it moves closer to the tunnel wall as macrolide and nascent peptide chain fill the tunnel (9, 35). Such a movement would clash with MsrE His244 residue, however (Fig. 5C). Instead, A2062 nucleotide is reoriented in the tunnel lumen in the presence of MsrE compared with that in drug-free ribosomes (Fig. 5C).

On AZM binding to ribosomes, the nucleotide U2506 was observed to shift toward AZM and to be involved in its binding, corroborating the role of U2506 in macrolide drug action (2, 3). In contrast, the U2506 in the present complex is seen to undergo a significant conformational change away from both the PTC and the macrolide-binding site (Fig. 5C). Furthermore, the neighboring U2504, which has been implicated in determining the species-specificity of several PTC A-site–binding antibiotics (e.g., tiamulin) (36), shifts away by approximately 2 Å (Fig. 5C). Simultaneously, the next nucleotide, A2503, whose identity was found to be critical for programmed translational arrest (8, 9) and whose C8 methylation by Cfr methyltransferase is known to cause resistance to some macrolides (37, 38), is also deformed from AZM-binding positioning (SI Appendix, Fig. S7A). These conformational changes in the A-site side of PTC appear to be caused by MsrE extended-loop tip residues, especially Arg241 (Fig. 5C). As for the P-site side, U2585, a key player in PTC activity, is displaced away from PTC (>90° flip and a 5-Å movement), consequently forming a stacking interaction with A2602, which is also rearranged from its conformation in ribosomes with AZM (2, 3) and without AZM (30) (Fig. 5C).

The inherent flexibility of the universally conserved residues U2506 and U2585 at active sites is essential for their pivotal role in peptide transfer (30, 39, 40). For instance, the 180° flip of U2585 caused by the ErmCL nascent chain and macrolide interplay in the NPET prevents the stable binding and accommodation of A-site tRNA, leading to inhibition of peptide bond formation and the translation of downstream macrolide resistance protein ErmC (41). The movement of U2585 caused by macrolide erythromycin binding is apparently accompanied by repositioning of A2602 (16). The function of the universally conserved A2602, essential for peptidyl-tRNA hydrolysis to release nascent peptide, is highly dependent on its positioning (42). Here the residue Lys233 in MsrE is likely involved in stabilizing the reorientation of A2602 (Fig. 5B).

Our mutagenesis study on MsrE revealed that mutation of Arg241, Leu242, or His244 to Ala results in a significantly diminished ability of MsrE to confer AZM resistance in vivo (SI Appendix, Table S2) and to rescue AZM-affected translation in vitro (Fig. 5D). This finding demonstrates their crucial role in mediating macrolide resistance, which is consistent with our structural information. Furthermore, truncation of this extended loop region (residues Lys216–Lys254, Δ loop) or even mutation of only the two residues (Arg241Ala/His244Ala) completely abolished its ability to mediate AZM resistance (Fig. 5D and SI Appendix, Table S2). This finding demonstrates that the extended loop region, particularly the two residues Arg241 and His244, is essential for displacing AZM from its binding site through strong interactions with ribosomes. Interestingly, Lys233, Arg241, Leu242, and Lys246 are universally conserved in the Msr subfamily, whereas His244 is found only in MsrE and MsrD, substituted by small residues Ala and Ser in MsrA and MsrC, respectively (SI Appendix, Fig. S1). The Arg241Ala and His244Ala mutations or even loop deletion had no significant effect on the ribosome binding efficiency (SI Appendix, Fig. S3B), but did render the MsrE protein unable to recover the AZM-affected translation in vitro (Fig. 5D).

Finally, we tested whether binding of MsrE could displace AZM and release it from the ribosome. Quantification of AZM copelleting with ribosomes revealed that incubation of the AZM-ribosome complex with MsrE indeed reduced the amount of AZM associated with ribosomes (Fig. 5E). Furthermore, our data demonstrate the importance of the loop residues Arg241 and His244, as mutating either to Ala resulted in complete loss of AZM release activity (Fig. 5E). Curiously, when the lesser binding efficiency of the ATP hydrolysis-deficient mutant E104Q/E413Q (SI Appendix, Fig. S3B) is taken into account, its drug displacing effect is comparable to that of WT MsrE (Fig. 5E). Therefore, the role of ATP hydrolysis in MsrE resistance activity is likely due to its importance in turnover.

Regardless of the size of the macrolactone ring, all macrolides are oriented in the ribosomal tunnel in a similar manner involving a hydrogen bond between their desosamine hydroxyl and the N1 atom of A2058 in the 23S rRNA (2, 3). Furthermore, the binding of macrolides to ribosomes is stabilized by the tight hydrophobic packing of the lactone ring against the conserved U2611 and A2057, as well as the ionic interaction between desosamine and the phosphate group of G2505. In addition to the aforementioned notable changes for several crucial PTC residues on MsrE binding (Fig. 5C), some minor changes are observed in 23S rRNA (U2609 and U2611 on one side and A2058 and A2057 on the other side) and r-protein L22 (Arg90 at the tip of elongated hairpin) involved in formation of the macrolide-binding site in the tunnel (SI Appendix, Fig. S7A). As a result, the tunnel around the macrolide- binding site widens by approximately 1.5–2 Å.

Taken together, the MsrE extended loop mediated allosteric relay of changes to the PTC and NPET in the vicinity of the macrolide-binding site likely contributes to the dislodgement of AZM from the tunnel. Consistent with this notion, many nucleotide modifications mediating antibiotic resistances do not involve direct interactions with the drug, but instead involve reshaping of the binding pocket to release the drug.

It appears that the composition and length of the extended loop correlate with antibiotic resistance profiles of ARE ABC-F proteins (SI Appendix, Fig. S1). Indeed, detailed mutational analysis of VgaA linker residues identified a short stretch of residues 212–220 whose composition determines the efficiency as well as the specificity of antibiotic resistance (27, 43). Based on sequence alignment (SI Appendix, Fig. S1), this short stretch corresponds to the region Arg241-Leu242, the importance of which can be rationalized by our structural information. The diverse nature of ARE ABC-F domain linker extended loops correlating with the differences in their drug specificities suggests that bacteria have evolved a plethora of mechanisms to protect the ribosome from PTC- and NPET-targeting antibiotics.