MmpL3 is a lipid transporter that binds trehalose monomycolate and phosphatidylethanolamine

Significance

The unique architecture of the mycobacterial cell wall plays a predominant role in Mycobacterium tuberculosis pathogenesis. Mycobacterial membrane protein Large 3 (MmpL3) is essential and required for transport of trehalose monomycolates (TMMs) across the cell membrane for cell-wall biosynthesis. Here we report a crystal structure of Mycobacterium smegmatis MmpL3 that indicates a plausible pathway for TMM translocation. Native mass spectrometry suggests that the monomeric MmpL3 transporter is able to specifically bind TMM and phosphatidylethanolamine lipids. Overall our data suggest a mechanism for TMM transport involved in mycobacterial cell-wall biogenesis.

Abstract

The cell envelope of Mycobacterium tuberculosis is notable for the abundance of mycolic acids (MAs), essential to mycobacterial viability, and of other species-specific lipids. The mycobacterial cell envelope is extremely hydrophobic, which contributes to virulence and antibiotic resistance. However, exactly how fatty acids and lipidic elements are transported across the cell envelope for cell-wall biosynthesis is unclear. Mycobacterial membrane protein Large 3 (MmpL3) is essential and required for transport of trehalose monomycolates (TMMs), precursors of MA-containing trehalose dimycolates (TDM) and mycolyl arabinogalactan peptidoglycan, but the exact function of MmpL3 remains elusive. Here, we report a crystal structure of Mycobacterium smegmatis MmpL3 at a resolution of 2.59 Å, revealing a monomeric molecule that is structurally distinct from all known bacterial membrane proteins. A previously unknown MmpL3 ligand, phosphatidylethanolamine (PE), was discovered inside this transporter. We also show, via native mass spectrometry, that MmpL3 specifically binds both TMM and PE, but not TDM, in the micromolar range. These observations provide insight into the function of MmpL3 and suggest a possible role for this protein in shuttling a variety of lipids to strengthen the mycobacterial cell wall.

Tuberculosis (TB) is the leading cause of mortality as a result of an infectious agent, exceeding both malaria and HIV (1, 2). In 2016, there were an estimated 10.4 million new cases of active TB and 1.7 million patients died from active TB infection (2). The causative agent, Mycobacterium tuberculosis (Mtb), has now infected more than one-third of the world’s population. The emergence of multidrug-resistant TB (MDR-TB) presents an increasingly difficult therapeutic challenge, so much so that MDR-TB is now the main cause of death due to antimicrobial resistance. Unfortunately, the lethality of TB combined with its multidrug-resistant capacity has now transformed this long-neglected disease into a global health priority.

The unique architecture of the mycobacterial cell wall plays a predominant role in Mtb pathogenesis. This complex cell-wall structure supports membrane stability and provides a barrier against the host environment and antibiotics, thereby contributing to the intrinsic resistance of the organism. The outer membrane of Mtb is characterized by the abundance of very long-chain mycolic acids (MAs). MAs are transported across the inner membrane as trehalose monomycolates (TMMs) and then are either covalently linked to the arabinogalactan-peptidoglycan layer as mycolyl arabinogalactan peptidoglycan (mAGP) or incorporated into trehalose dimycolates (TDM), which comprises the majority of the outer leaflet of this membrane. The outer leaflet also contains other noncovalently associated lipids, such as phthiocerol dimycocerosates and sulfolipids (3). This unique construction makes the outer membrane of Mtb very rigid and extremely impermeable to a wide range of compounds, including many antibiotics (3). Additionally, the surface-exposed lipids are immunomodulatory and play a crucial role in host–pathogen interactions (4–9).

TMM biosynthesis and its transport are essential for Mtb, as it is the precursor of both mAGP and TDM in the outer leaflet of the cell envelope. TMMs are synthesized in the cytoplasm via a highly conserved and well-characterized pathway (10), targeted by the first-line anti-TB drug isoniazid (11). However, exactly how these TMMs are transported across the membrane from the cytosol to participate in mycobacterial cell-wall remodeling is not well understood. Recent work demonstrated that mycobacterial membrane protein Large (MmpL) transporters are critical for mycobacterial physiology and pathogenesis by shuttling fatty acids and lipid components to the mycobacterial cell wall. Mtb possesses 13 MmpL transporters (12). These membrane proteins share sequence similarity with the resistance-nodulation-cell division (RND) superfamily of transporters (13). Although MmpL3 is absolutely essential in vivo, there is strong evidence that MmpL4, MmpL5, MmpL7, MmpL8, MmpL10, and MmpL11 are all required for full virulence as well as for the growth of mycobacteria in mouse lungs (14–18). Interestingly, of the 13 MmpLs, only MmpL3 is shown to export TMM (18, 19). Using a spheroplast system, a recent study strongly indicated that MmpL3 is a TMM flippase (20). This biochemical analysis is in line with genetic studies that determined MmpL3 expression is necessary for Mtb survival and that depletion of MmpL3 in Mycobacterium smegmatis results in the accumulation of TMMs concomitant with a reduction in levels of mAGP and TDM (18, 19). Thus, MmpL3 has now been shown to be essential for the biosynthesis of mAGP and TDM, as well as their incorporation into the mycobacterial cell wall, with both processes critical for mycobacterial replication and viability (21, 22). It is worth noting that Mtb mmpL3 can rescue the viability of the M. smegmatis mmpl3 null mutant (18). This result suggests that the two mmpL3 orthologs can substitute each other to function. Collectively, studies from both Mtb and M. smegmatis have established the importance of MmpL3 in membrane biosynthesis and TB viability as well as being a target of several potent anti-TB agents.

As MmpL3 is absolutely essential for mycobacterial cell-wall biogenesis, we decided to elucidate the molecular mechanism of lipid translocation across the membrane via this transporter. We here present a crystal structure of M. smegmatis MmpL3 to a resolution of 2.59 Å. Unexpectedly, a phosphatidylethanolamine (PE) molecule was found to be associated within a partially buried region of this protein. Using native mass spectrometry, we observed that MmpL3 is capable of binding PE and TMM, but not TDM, with dissociation constants within the micromolar range. Our work provides molecular insights into the mechanism of lipid transport via the MmpL3 membrane protein and suggests a possible role for MmpL3 in shuttling different lipids across the membrane for cell-wall biogenesis.

Results

Overall Structure of the M. Smegmatis MmpL3 Transporter.

To obtain the structural information of MmpL3, we cloned the full-length M. smegmatis MmpL3 transporter, which contains 1,013 amino acids, into the Escherichia coli expression vector pET15b, with a 6xHis tag at the C terminus to generate pET15bΩmmpL3. This MmpL3 protein was overproduced in E. coli BL21(DE3)ΔacrB cells and purified using Ni²⁺-affinity and Superdex 200 size exclusion columns. Native mass spectrometry (native MS) shows that the MmpL3 transporter is monomeric in detergent solution (Fig. 1A). To further confirm the oligomerization state of this protein in its native environment, we cloned and overproduced MmpL3 in M. smegmatis mc²155 cells. Again, the native mass spectra indicate that this purified MmpL3 protein is a monomer in detergent solution (Fig. 1B).

Fig. 1.

Mass spectra of purified MmpL3 proteins. (A) Mass spectrum of full-length MmpL3 expressed in E. coli indicates that the protein exists as a monomer. Calculated and observed masses, including the C-terminal 6×His tag, are 110,222 and 110,246 Da. (B) Mass spectrum of full-length MmpL3 expressed in M. smegmatis. In addition to the monomeric protein in solution, the presence of degraded protein bands indicates that the purified protein is unstable. The two observed masses are 110,371 and 91,990 Da. (C) Mass spectrum of a 5-d-old sample of full-length MmpL3 expressed and purified from E. coli. The spectrum depicts three major degraded species in the solution sample. The observed masses of these species are 87,998, 84,842, and 83,359 Da, corresponding to residues 1–806, 1–776, and 1–763 of the protein, respectively. (D) Mass spectrum of the MmpL3₇₇₃ protein. The spectrum indicates that this protein exists as a monomer in solution. Observed and calculated masses in this case are 85,950 and 85,925 Da. The mass observed from the adduct peaks is 86,660 Da, which corresponds to the MmpL3₇₇₃-PE complex.

Crystals of MmpL3 were obtained using vapor diffusion. The best crystal diffracted X-rays to a resolution of 3.30 Å (SI Appendix, Fig. S1 and Table S1). However, these crystals were difficult to reproduce, and most of the protein samples could not be crystallized. We then harvested several of these protein crystals and washed them with the crystallization precipitant solution. SDS-PAGE analysis indicated that the MmpL3 protein within these crystals was ∼25 kDa smaller than the full-length protein. Since the C-terminal region (residues 733–1,013) of MmpL3 contains a large number of prolines (52 proline residues), it is likely that this C-terminal sequence is proteolytically clipped and quite possibly unstable in solution.

We then used native MS to determine the size of the M. smegmatis full-length MmpL3 protein isolated from E. coli. Surprisingly, the data indicate that the protein sample contains three major fragments (Fig. 1C). These three fragments are shorter than the full-length protein, which is composed of 1,013 residues. The observation of these three protein fragments suggests that the full-length protein may be clipped by proteases. A detailed analysis indicates that these three fragments contain residues 1–806, 1–776, and 1–763 of MmpL3. However, only a small quantity of the fragment 1–763 was detected. These protein fragments possess the full transmembrane region, but the C-terminal end was absent. Since it has been reported that this C-terminal domain is not essential for the function of MmpL3 (23), we decided to remove this C-terminal sequence and produce the truncated MmpL3 transporter, designated to be MmpL3₇₇₃ (residues 1–773), to improve crystal quality for structural determination. Our native MS data indicate that the MmpL3₇₇₃ protein also exists as a monomer, which presents as a single protein fragment suitable for crystallization (Fig. 1D).

The best MmpL3₇₇₃ crystal diffracted X-ray to 2.59 Å resolution (Fig. 2). We determined its structure using single anomalous dispersion (SI Appendix, Fig. S2 and Table S1). We then used the structure of MmpL3₇₇₃ as the template to resolve the structure of full-length MmpL3, where we found that the C-terminal residues 753–1,013 were missing. The conformation of this MmpL3 structure is identical to that of MmpL3₇₇₃. Superimposition of the two structures gives a root-mean-square-deviation (r.m.s.d.) of 0.6 Å (for 722 Cα atoms) (SI Appendix, Fig. S3). Although MmpL3 belongs to a subclass of the RND superfamily, its 3D topology is unique and very different from the existing structures of RND transporters, including AcrB (24, 25), MexB (26), CusA (27, 28), MtrD (29), CmeB (30), and HpnN (31). Thus, the available structural information of RND proteins cannot be used to understand the function of this transporter.

Fig. 2.

Structure of the M. smegmatis MmpL3 transporter. (A) Secondary structural topology of the MmpL3₇₇₃ monomer. The topology was constructed based on the crystal structure of MmpL3₇₇₃. The TMs are colored blue. The periplasmic loops 1 and 2 are in cyan and green, respectively. The CD of MmpL3 is colored red. (B) Ribbon diagram of a monomer of MmpL3₇₇₃ viewed in the membrane plane. The TMs, periplasmic loops 1 and 2, and CD are colored slate, cyan, green, and red, respectively. The bound DDM is depicted as yellow spheres, whereas the bound PE is depicted as pink spheres. The periplasmic loops 1 and 2 cross over each other to form the periplasmic domains 1 and 2 (PD1 and PD2). (C) The MmpL3₇₇₃ monomer forms a channel spanning the outer leaflet of the inner membrane up to the periplasmic domain. The orientation of this MmpL3 molecule has been rotated by 110° (as shown in the figure) compared with the orientation of B. The channel (colored gray) was calculated using the program CAVER (https://www.caver.cz/). The TMs, periplasmic loops 1 and 2, and CD are colored the same as in B. The bound DDM is depicted as yellow sticks, whereas the bound PE is depicted as pink sticks.

The MmpL3₇₇₃ molecule consists of 12 transmembrane helices (TMs 1–12) and two periplasmic loops (loops 1 and 2), which create the periplasmic domain (Fig. 2 A and B). Loop 1 is located between TMs 1 and 2, and loop 2 is found between TMs 7 and 8. These two loops contribute to generate subdomains PD1 and PD2 in the periplasm. Based on our structural information, the proline-rich C-terminal residues 733–1,013 should form a cytoplasmic domain (CD) of MmpL3.

The N-terminal and C-terminal halves of MmpL3₇₇₃ are assembled in a twofold pseudosymmetrical fashion. These two halves can be superimposed with a r.m.s.d. of 2.6 Å (for 294 Cα atoms) (SI Appendix, Fig. S4). PD1 is composed of four α-helices and three β-strands. The majority of the PD1 amino acids come from loop 1. However, residues 437–448 of loop 2 also contribute to form helix α5 of PD1. PD2 constitutes three α-helices and three β-strands. Like PD1, the PD2 amino acids arise mainly from loop 2, but residues 49–60 of loop 1 participate in the formation of helix α1 of this periplasmic subdomain. The cross-over of these two periplasmic loops allows for the two subdomains, PD1 and PD2, to be spatially adjoined within the periplasm. This structural feature is in good agreement with the finding that the two periplasmic loops closely interact with each other (32). There is a long flexible linker of 12 residues connecting the C-terminal end of TM1 and α1 of PD2. Similarly, a long, 15-residue flexible loop is found to link the C-terminal end of TM7 and α5 of PD1 together. The presence of these long linkers suggests that the periplasmic domains of MmpL3 are quite flexible in nature. The TMs are membrane-embedded, but both TM2 and TM8 are significantly longer and protrude into the periplasmic region. These two TMs directly tether PD1 and PD2, respectively, and form part of the periplasmic structure of the protein.

The MmpL3₇₇₃-Phosphatidylethanolamine Complex.

Surprisingly, the MmpL3₇₇₃ molecule forms a channel-like cavity originating from the outer leaflet of the inner membrane up to the periplasmic domain (Fig. 2C). The beginning of this cavity is generated by a hydrophobic pocket created by TMs 7–10. This pocket opens to the outer leaflet of the inner membrane and periplasmic space. However, the majority of this cavity is found at the center of the periplasmic domain surrounded by the secondary structures of PD1 and PD2. This large space potentially constitutes a binding site for TMM.

Unexpectedly, two large extra electron densities were found within the structure of MmpL3₇₇₃. One of these densities was observed in the pocket surrounded by TMs 7–10. The shape of this extra density is compatible with an n-dodecyl-β-d-maltoside (DDM) detergent molecule (Fig. 3A). This is not surprising as we solubilized, purified, and crystallized the MmpL3₇₇₃ transporter in solutions containing DDM detergent. Within 4.5 Å of the bound DDM, there are 13 residues, including L422, S423, L424, N524, A527, Q554, I557, F561, L564, P565, A568, L600 and I636, that provide hydrophobic and electrostatic interactions for anchoring this detergent molecule (Fig. 3A). Among these residues, the conserved leucine L564 has been observed to be important for the function of MmpL3 (33). Interestingly, within the vicinity of this binding site, we also find the conserved amino acids Q309, D555, and P630. These three residues are located at the beginning of the channel formed by MmpL3. It has been shown that these three amino acids are important for the activity of this transporter in M. smegmatis (23).

Fig. 3.

The DDM- and PE-binding sites of the MmpL3-PE complex. (A) The F_o − F_c electron density map of bound DDM in MmpL3. The bound DDM is shown as a stick model (yellow, carbon; red, oxygen). The F_o − F_c map is contoured at 3σ. Residues involved in DDM binding are shown as green sticks. The secondary structural elements of MmpL3₇₇₃ are colored light brown. (B) The F_o − F_c electron density map of bound PE in MmpL3. The bound PE is shown as a stick model (magenta, carbon; red, oxygen; blue, nitrogen). The F_o − F_c map is contoured at 3σ. Residues involved in PE binding are shown as green sticks. The secondary structural elements of MmpL3₇₇₃ are colored light brown.

The second extra density is observed within the large space between PD1 and PD2. The shape of this electron density resembles a large lipid molecule with two elongated hydrocarbon backbones (Fig. 3B). Presumably, this extra electron density belongs to a fortuitous ligand that we copurified and cocrystallized with the MmpL3₇₇₃ protein. Consistent with this observation, our MS data of purified MmpL3₇₇₃ indicate a charge state series corresponding to an adduct mass of ∼700 Da (Fig. 1D). We then used liquid chromatography coupled with mass spectrometry to identify this unknown bound ligand. The result indicates that this fortuitous ligand is PE [16:0–17:1(9Z)] (SI Appendix, Fig. S5A). Interestingly, these PE lipids are also observed to bind in the full-length MmpL3 protein expressed in both E. coli (SI Appendix, Fig. S5B) and M. smegmatis (SI Appendix, Fig. S5C). Because of the presence of these DDM and PE molecules, the conformation of our structure of MmpL3 should represent its ligand-bound form.

PE binds at the central cavity formed by PD1 and PD2 of the periplasmic domain. Two extended loops (residues 61–68 and residues 449–456) that both run across the subdomains PD1 and PD2 are found to sandwich the bound PE. To secure the binding, residues 40–44 of the elongated loop connecting TM1 and PD2, as well as residues 426–429 that belong to the other elongated loop connecting TM7 and PD1, participate to form the bottom of this large cavity. The binding of PE is extensive. Several conserved amino acids are found to surround the wall of this central cavity. Within 4.5 Å of the bound PE molecule, there are at least 32 amino acids involved in the binding. These residues are Q40, S41, F43, Y44, D64, T66, S67, V70, V109, T121, M125, F134, S136, D144, L171, L174, A175, Q421, I427, S428, E429, F445, F452, R453, T454, P456, R501, P502, A503, N504, Q517, and T549, which provide electrostatic and hydrophobic interactions to bind PE (Fig. 3B). Among them, the conserved residue Q40 has been reported to be crucial to the function of this transporter (23).

The Proton-Relay Network.

MmpL3 is a proton-motive-force–dependent transporter that functions via an antiport mechanism. Coupled with the movement of substrates toward the periplasm, protons have to flow into the cytoplasm to energize this translocation process. Within the transmembrane region, a hydrogen bond is formed between the conserved residues D256 and Y646. Likewise, the nearby conserved residues Y257 and D645 also form a hydrogen-bonded pair. Residue S293, which is also conserved, is within 3.7 Å away from D645 and interacts with this aspartate through electrostatic interaction (SI Appendix, Fig. S6). Within the vicinity of these residues, the crystal structure suggests that the conserved charged residues E647 and K591 also form another hydrogen bond between them. Taken together, these data suggest that these residues may form an important network for proton transfer within MmpL3. This is in good agreement with previous studies that have emphasized the importance of many of these residues to the function of this transporter (23), perhaps by creating a proton-relay network for energy coupling.

PE Binding to MmpL3.

As our structure of MmpL3 contained a bound PE, we wanted to further characterize this MmpL3–PE interaction. To achieve this, we performed a lipid-binding experiment using native MS. A solution containing 5 μM delipidated MmpL3₇₇₃, 200 mM ammonium acetate (pH 8.0), and 0.05% (wt/vol) lauryldimethylamine N-oxide was used for the experiment. Equimolar amounts of MmpL3₇₇₃ and PE were mixed together, and then the mass spectra under our optimized conditions were recorded (Fig. 4A). The experimental result depicts a charge state series for a second higher mass species at ∼15% intensity of the peak corresponding to MmpL3₇₇₃. The 700-Da mass difference is consistent with the binding of one molecule of PE to MmpL3 (Fig. 4A).

Fig. 4.

Determination of dissociation constants for the binding of PE and TMM to MmpL3₇₇₃. (A) Mass spectra recorded for solutions of MmpL3₇₇₃ with increasing concentrations of PE. At 5 μM PE, a charge state series is observed (light brown), corresponding to bound PE, which increases in intensity as the PE concentration is increased to 80 μM. A second PE-binding peak (gray) emerges at concentrations above 20 μM. (B) Plot of relative fractional intensity of lipid-binding peaks over the total peak intensity versus PE concentration (Methods), yielding a curve for the first binding event and linear-like fit for the second, consistent with nonspecific PE binding. Each data point and SD are calculated from the average of five observed charge states in three independent experiments. Error bars represent SDs (n = 3). (C) Mass spectra recorded for solutions of MmpL3₇₇₃ with increasing concentrations of TMM. At 2.5 μM TMM, a charge state series is observed (orange), corresponding to bound TMM, which increases in intensity as the TMM concentration is increased to 40 μM. A second TMM-binding peak (purple) emerges at concentrations above 20 μM. (D) Plot of the relative fractional intensity of lipid-binding peaks over total peak intensity versus TMM concentration (Methods), yielding a curve for the first binding event and a linear-like fit for the second, consistent with nonspecific TMM binding. Each data point and SD are calculated from the average of five observed charge states in three independent experiments. Error bars represent SDs (n = 3).

Subsequently, we incubated the MmpL3₇₇₃ protein with increasing concentrations of PE. At concentrations above 20 μM, we observed an increase in the intensity of the one PE-bound species, as well as an additional charge state series that corresponds to the binding of a second molecule of PE to MmpL3₇₇₃. At 80 μM of PE, the predominant species is the lipid-bound form of MmpL3₇₇₃. Based on our mass spectral data, we extracted and plotted the relative intensity of PE-bound forms as a function of lipid concentration. The data suggest a dissociation constant (K_d) of 19.5 ± 6.3 μM for the first PE binding (Fig. 4B).

TMM Binding to MmpL3.

Since there is strong evidence that MmpL3 is a TMM transporter, we decided to perform TMM-binding experiments to probe the affinity of MmpL3 to TMM. Similar to the binding experiments described above, equimolar concentrations of MmpL3₇₇₃ and TMM were mixed together to acquire mass spectra under the conditions similar to those for PE binding. As expected, the data suggest a charge state series for a second higher mass species at ∼25% intensity of the peak corresponding to MmpL3₇₇₃. The mass difference of 1,425 Da is consistent with the binding of one molecule of TMM to MmpL3₇₇₃, indicating that MmpL3 is capable of recognizing TMM (Fig. 4C).

To further define the affinity of TMM for MmpL3, we incubated the protein with increasing concentrations of TMM. As with PE binding, at concentrations higher than 20 μM TMM, we observed evolution of an additional charge state series, corresponding to binding of a second molecule of TMM to MmpL3₇₇₃. Again, we extracted and plotted the relative intensity of TMM-bound forms with respect to TMM concentration, suggesting a K_d of 3.7 ± 1.3 μM for the first TMM binding (Fig. 4D).

MmpL3 Does Not Bind TDM.

As MmpL3 is able to specially bind TMM, we questioned if this transporter can bind TDM as well. We purified glycolipids from M. smegmatis mc²155 cells. MS analysis indicates that the two main species of this glycolipid sample are TDM and TMM (Fig. 5). We then added this sample to the purified MmpL3₇₇₃ membrane protein and recorded the mass spectra. In addition to the MmpL3₇₇₃ and MmpL3₇₇₃-PE peaks, the MS data show a new charge state series corresponding to an adduct mass of ∼1,428 Da, which is the mass of TMM. Surprisingly, our data show that only short chain TMM lipids (∼1,428 Da) bind specifically to MmpL3₇₇₃ even though different longer chain lengths of TMM lipids are present in the solution (red Inset in Fig. 5). The result is consistent with the binding data that MmpL3₇₇₃ selectively interacts with TMM. Of specific interest, the spectra did not present any other adduct peaks that correspond to the TDM bound species, demonstrating that MmpL3₇₇₃ does not bind TDM.

Fig. 5.

Analysis of different components of purified M. smegmatis glycolipids and their binding to MmpL3₇₇₃. Mass spectrum of MmpL3₇₇₃ with purified M. smegmatis glycolipids shows a preferential binding to TMM (orange charge state series) but not to TDM. Purified lipid fraction has both TMM (red Inset) and TDM (blue Inset) at the lower m/z region.

To further test the interaction between MmpL3 and TDM, we performed binding experiments with the purified TDM lipid (Sigma-Aldrich). Again, we did not observe the event of TDM binding to the MmpL3 protein (SI Appendix, Fig. S7). Collectively, our data indicate that TMM, but not TDM, is preferentially bound by the MmpL3 transporter.

Other Potential MmpL3 Substrates.

In addition to PE, our lipidomic analysis indicates that diacylglycerol (DAG) and other phospholipids, including phosphatidylglycerol (PG), phosphatidylinositol (PI), and cardiolipin (CDL), are bound by full-length MmpL3 purified from M. smegmatis (SI Appendix, Figs. S5 and S8). To elucidate if these lipids are potential substrates for MmpL3, we used native mass spectrometry to detect the direct binding of PG or CDL to MmpL3. We observed that both PG and CDL specifically interact with MmpL3 (SI Appendix, Fig. S9), indicating that these lipids are potential substrates for the MmpL3 transporter.

Discussion

MmpL3 Functions as a Monomer.

The translocation of TMM across the cytoplasmic membrane is an essential step in mycobacterial cell-wall biogenesis. This process is absolutely required for the biosyntheses of mAGPs and TDMs that subsequently allow the cell envelope of Mtb to form a very rigid and extremely impermeable layer to a wide range of antimicrobial compounds. However, exactly how TMMs are shuttled across the membrane for cell-wall biosynthesis remains largely unknown. This void significantly hampers progress in understanding the molecular mechanisms of Mtb cell-wall biogenesis.

Our crystallographic and native MS data strongly indicate that the M. smegmatis MmpL3 transporter functions as a monomer. Its oligomerization state and 3D structure are very distinct from all other known structures belonging to the RND superfamily of membrane proteins. For example, the well-studied hydrophobic and amphiphilic efflux subfamily of RND transporters, including AcrB (24, 25), MexB (26), MtrD (29), and CmeB (30), exist as trimers within the cytoplasmic membrane of Gram-negative bacteria. They have large, periplasmic domains that confer substrate specificity and interact with periplasmic adaptors and outer membrane channels to form tripartite efflux complexes. Likewise, crystal structures of the heavy metal efflux subfamily of RND transporters, such as CusA (27, 28, 34) and ZneA (35), also suggest a trimeric oligomerization, where each protomer was crystallized around a crystallographic pseudo threefold axis. Recently, another subfamily of the RND-superfamily transporters, termed hopanoid biosynthesis-associated RND (HpnN), from Burkholderia multivorans was crystallized (31). Surprisingly, this HpnN transporter presents as a dimer, which shuttles hopanoid lipids from the cytoplasmic membrane to the outer membrane of B. multivorans for cell-wall remodeling. In the present study, we observed that MmpL3, which is affiliated with the MmpL subfamily of RND-superfamily transporters, operates as a monomer. Based on the MmpL3 structure, it is likely that members of the MmpL-family transporters may constitute as a monomer to translocate substrates. Our data indeed pose a possibility that different subfamilies of these RND membrane proteins may assemble into various oligomerization states to function.

Proposed Mechanism for TMM Translocation.

We propose that the monomeric MmpL3 transporter takes up TMM from the outer leaflet of the cytoplasmic membrane via a channel constituted by the transporter. A cleft surrounded by TMs 7–10 forms the entrance of this channel, which spans the outer leaflet of the cytoplasmic membrane and up to the central cavity of the periplasmic domain between PD1 and PD2. The TMM molecule could be shuttled through the channel to reach the periplasmic domain of the protein. This bound lipid could then be released to the inner leaflet of the outer membrane, where biosyntheses of mAGP and TDM take place (Fig. 6). Proton transfer via the proton-relay network mediates the energy needed for substrate translocation. The processes of proton import and substrate export may be coupled within the transport cycle.

Fig. 6.

Proposed mechanism for TMM translocation via MmpL3. This schematic diagram indicates that the MmpL3 transporter is capable of picking up a TMM molecule from the outer leaflet of the cytoplasmic membrane. This TMM molecule will pass through the channel formed by MmpL3 and arrive at the periplasmic lipid-binding site. The TMM moiety will then be exported to the inner leaflet of the outer membrane for the biosynthesis of TDMs and mAGPs.

MmpL3 Specifically Binds a Variety of Lipids.

The unique composition of lipids that make up the cell envelope is a predominant feature of Mtb and other mycobacteria. These lipids constitute ∼40% of the dry cell mass (36), with the organization of these lipids between the inner and outer membranes significantly contributing to its impermeability to biocides and antimicrobials. In addition to being integral components of the mycobacterial cell wall, Mtb is capable of releasing PE, PI, and CDL into the infected organism (37, 38). These mycobacterial lipids may play a role in the infectivity and persistence of mycobacteria within the host (38). Therefore, mycobacterial lipid biogenesis and cell-wall biosynthesis are regarded as important targets for many first-line antimycobacterial agents (39). Using native MS, we observed that MmpL3 binds to a variety of lipids, including TMM, PE, PG, PI, DAG, and CDL, indicating that this transporter is probably a promiscuous lipid-binding protein.

Our data also suggest that MmpL3 likely contributes to mycobacterial cell-envelope biogenesis by active translocation of several important lipid components to the cell wall. The ability of several potential therapeutics to block the binding and transport of these lipids via MmpL3 implicates this membrane protein as an important TB drug target. Our study is a significant advance in the field as it defines the structure of MmpL3 in the context of lipid substrates. While our paper was under review, Zhang et al. (40) reported the T4 lysozyme-MmpL3 fusion protein structure as well as structures of this protein with several presumed inhibitors of MmpL3 (SI Appendix, Fig. S10). Combined, our studies will ultimately inform an era in structure-guided drug design for the critical cell-envelope transporter MmpL3 and other virulence-associated MmpL proteins.

Methods

All protein crystals were grown at 25 °C using vapor diffusion. A 2-μL protein solution containing 20 mg/mL MmpL3 or MmpL3₇₇₃ in 20 mM Na-Hepes (pH 7.5) and 0.05% (wt/vol) DDM was mixed with 2 μL of reservoir solution containing 25% PEG 400, 0.1 M sodium acetate (pH 5.4), and 0.05 M magnesium acetate. The resultant mixture was equilibrated against 500 μL of the reservoir solution. The crystallization conditions for SeMet-MmpL3₇₇₃ were the same as those for MmpL3₇₇₃. Cryoprotection of these crystals was achieved by raising the PEG 400 concentration to 30%. A full description of methods for protein expression and purification, X-ray data collection and analysis, native mass spectrometry, lipidomic analysis, and lipid binding are described in SI Appendix, SI Materials and SI Methods.