Structural and functional analysis of the Mycobacterium tuberculosis MmpS5L5 efflux pump presages increased bedaquiline resistance

Abstract

Bedaquiline, an antitubercular drug that targets ATP-synthase, is a key component of a new oral drug regimen that has revolutionized the treatment of multidrug-resistant tuberculosis. Clinical bedaquiline resistance in Mycobacterium tuberculosis has rapidly emerged, primarily due to mutations in the transcriptional repressor Rv0678 that result in upregulation of the resistance-nodulation-division (RND) efflux pump MmpS5/MmpL5 (MmpS5L5). Here, to understand how MmpS5L5 effluxes bedaquiline, we determined the structure of the MmpS5L5 complex using cryo-electron microscopy, revealing a trimeric architecture distinct from the canonical tripartite RND efflux pumps of gram-negative bacteria. Structure prediction modeling in conjunction with functional genetic analysis indicates that it uses a periplasmic coiled-coil tube to transport molecules across the cell wall. Structure-guided genetic approaches identify MmpL5 mutations that alter bedaquiline transport; these mutations converge on a region in MmpL5 located in the lower portion of the periplasmic cavity, proximal to the outer leaflet of the inner membrane, suggesting a route for bedaquiline entry into the pump. While currently known clinical resistance to bedaquiline is due to pump upregulation, our findings that several MmpL5 variants increase bedaquiline efflux may presage the emergence of additional modes of clinical resistance.

Keywords: MmpS5L5 drug efflux pump; Mycobacterium tuberculosis; bedaquiline; drug resistance.

Fig. 1.

Architecture of the trimeric MmpS5L5–AcpM complex. (A) Cryo-EM density map of the MmpS5–MmpL5∆CC–AcpM complex. (B) Structural model of MmpS5–MmpL5∆CC–AcpM complex. Periplasmic domains PD1 (pink) and PD2 (orange) are highlighted. (C) Slice view of the TM domains of MmpS5L5 (Left) and AcrB (Right). TM helix numbers are indicated.

Fig. 2.

MmpL5’s coiled-coil domain forms a 13 nm alpha-helical tube. (A) Structural model of the AlphaFold2 predicted structure of MmpS5L5 trimers, colored according to per-residue pLDDT value (Left) and by subunit (Right). (B) Tomogram of Mtb mc²6206. Inset, averaged cell wall density with the AlphaFold2 model of MmpS5L5 and the experimentally determined MCE1 complex [PDB: 8FEF (40)] overlaid with their TM domains on the inner membrane. IM—inner membrane, GL—granular layer, PG/AG—peptidoglycan/arabinogalactan, MM—mycomembrane. (C) AlphaFold2 model of MmpL5’s coiled-coil domain, with residues colored according to position in the heptad repeat (gabcdef). (D) Top view down the axis of the tube. The interior methionine residues are highlighted in red. A conserved disulfide bond is colored blue.

Fig. 3.

MmpS5 spans three MmpL5 protomers and is essential for efflux activity. (A) MmpS5’s TM interacts with TM helix 8 of MmpL5. Beta strands 2, 3, 5, and 8 of MmpS5’s Ig-like domain interact across the interface formed between MmpL5 protomers.(B) MIC values of bedaquiline, clofazimine, and PBTZ-169 show that MmpS5 is necessary for MmpL5 to efflux drugs. Paralogs MmpS1, MmpS2, and MmpS4 are unable to compensate for MmpS5. (C) Anti-FLAG immunoblot showing MmpL5 expression in all strains.

Fig. 4.

MmpL5 activity variants converge on a region of MmpL5 in a lower portion of the periplasmic cavity. (A) Plot of R_eq against bedaquiline concentration on a log₁₀ axis. Points represent the mean value of three technical replicates. Error bars represent SD. K_d values and Hill coefficient (h) were calculated using a specific binding with Hill slope model. (B) Bar chart showing the bedaquiline MIC values for the indicated strains. Open circles indicate technical replicates. (C) Alignment of MmpL5/4/2/1 at the selected positions. Amino acids for each paralog are shown. Amino acid changes chosen for mutation are indicated in red boxes. (D) MIC values for the selected amino acid substitutions are colored according to fold change in MIC versus the complemented MmpS5L5 strain. N.D = Not determined. (E and F) Upper, Summary diagram showing strategy for isolating suppressor/activity variants. (G) Experimental structural model of MmpL5 indicating residues that alter bedaquiline transport. Fold changes are given alongside the indicated substitution. Green indicates positive fold change in bedaquiline MIC, red indicates a negative fold change in bedaquiline MIC. (H) Overlay of a protomer from MmpS5L5 trimer structure with the AlphaFold2 model of MmpL5. Arrows indicate displacement of MmpL5 relative to the AlphaFold model. (I) A nonprotein density in the TM1–4 cleft of MmpL5 that is present in both “apo” and bedaquiline incubated samples. No density is observed in this cleft in the monomer structure.

Fig. 5.

Model for bedaquiline entry and transport. (A) Experimental structural model of MmpL5, with residues that alter the activity of MmpS5L5 indicated. PD1/PD2—Periplasmic domain 1 or 2, respectively. Residues identified as being functionally important for bedaquiline efflux are indicated with red spheres. (B) The model for MmpS5L5 mediated efflux of bedaquiline and other hydrophobic antitubercular drugs. Hydrophobic drugs like bedaquiline partition from the membrane into the periplasmic cavity of MmpL5 via TMs 1–4. The essentiality of the coiled-coil domain and its unique methionine-lined lumen, suggests that substrates are transported via this conduit. The length of this domain is too short to cross the entire cell wall. Strong sequence conservation of a hydrophobic surface at the tip of the coiled-coil domain suggests that an unknown outer membrane factor facilitates further substrate transport across the mycomembrane.