Spatially distinct and metabolically active membrane domain in mycobacteria

Significance

Mycobacterium is a family of bacteria that includes a number of dangerous pathogens. Arresting the growth of mycobacteria may be possible through the disruption of control points that regulate cell envelope biosynthesis. We demonstrate that Mycobacterium smegmatis possesses a spatially distinct biosynthetic membrane domain enriched in the polar growth region of the cell. This membrane domain may act as an organizing center to spatiotemporally coordinate biosynthetic activities during growth in live cells. Thus, our findings provide an important insight into the potential regulatory mechanisms of lipid metabolism in mycobacteria.

Abstract

Protected from host immune attack and antibiotic penetration by their unique cell envelope, mycobacterial pathogens cause devastating human diseases such as tuberculosis. Seamless coordination of cell growth with cell envelope elongation at the pole maintains this barrier. Unraveling this spatiotemporal regulation is a potential strategy for controlling mycobacterial infections. Our biochemical analysis previously revealed two functionally distinct membrane fractions in Mycobacterium smegmatis cell lysates: plasma membrane tightly associated with the cell wall (PM-CW) and a distinct fraction of pure membrane free of cell wall components (PMf). To provide further insight into the functions of these membrane fractions, we took the approach of comparative proteomics and identified more than 300 proteins specifically associated with the PMf, including essential enzymes involved in cell envelope synthesis such as a mannosyltransferase, Ppm1, and a galactosyltransferase, GlfT2. Furthermore, comparative lipidomics revealed the distinct lipid composition of the PMf, with specific association of key cell envelope biosynthetic precursors. Live-imaging fluorescence microscopy visualized the PMf as patches of membrane spatially distinct from the PM-CW and notably enriched in the pole of the growing cells. Taken together, our study provides the basis for assigning the PMf as a spatiotemporally distinct and metabolically active membrane domain involved in cell envelope biogenesis.

Tuberculosis, caused by the infection of Mycobacterium tuberculosis (Mtb), is a disease that claims about 1.5 million human lives annually (1). The thick, lipid-laden cell envelope of mycobacteria is composed of a plasma membrane, peptidoglycan-arabinogalactan layer, and mycolate outer membrane that are crucial for pathogenicity (2–4). The cross-sectional structure of the five or more distinct layers that form static mycobacterial cell envelope has been elucidated (2, 5). However, any mechanism for lateral elongation of a multilayered structure has not been defined. Most models posit that key components are synthesized inside the plasma membrane, with subsequent transport to outer layers of elongating cell wall (CW) (6). In mycobacteria, the elongation of the cell envelope is restricted to the polar region of the cell (7–10), suggesting the presence of spatiotemporal control mechanisms to supply cell envelope biosynthetic intermediates to this region (6). Indeed, a recent study demonstrated that key biosynthetic enzymes of the peptidoglycan-arabinogalactan-mycolic acid core structure are specifically enriched in the subpolar region of mycobacterial cells (11).

We previously reported membrane compartmentalization in Mycobacterium smegmatis (Msmeg) (12). Density gradient fractionation of mycobacterial lysate revealed a distinct fraction containing plasma membrane free of the CW (PMf) in addition to a fraction containing the classical plasma membrane tightly associated with the CW (PM-CW). Both of these membranes are composed of major phospholipids such as phosphatidylethanolamine (PE), phosphatidylinositol (PI), and cardiolipin (12). However, the PMf fraction is enriched in specific enzymes related to the biosynthesis of PE and PI mannosides (PIMs) (12), implying its distinct role in phospholipid metabolism. To synthesize phospholipids, mycobacteria use the cytidine diphosphate-diacylglycerol (CDP-DAG) pathway (13), in which phosphatidic acid (PA) is activated to CDP-DAG, and then converted to phosphatidylserine and PI (14). PE is produced from phosphatidylserine by phosphatidylserine decarboxylase (Psd), and this enzyme activity is enriched in the PMf (12). PIMs are made by sequential additions of mannose onto a PI, and AcPIM2 and AcPIM6, containing two and six mannose residues, respectively, are two major products. Although these mature PIM species are distributed in both the PM-CW and PMf, the enzymatic activities from PI to AcPIM2 and from AcPIM2 to AcPIM6 are enriched in the PMf and the PM-CW, respectively (12). Furthermore, PimB′, the mannosyltransferase that mediates the second mannose addition (15, 16), is specifically associated with the PMf (17). In addition, polyprenol-phosphate-mannose (PPM) is a lipidic mannose donor critical for the synthesis of mannose-containing glycolipids such as AcPIM6, lipomannan, and lipoarabinomannan. PPM synthase, composed of Ppm1 and Ppm2, is essential for survival (18), and its activity is also enriched in the PMf (12). Hence, biosynthetic reactions critical for cell envelope biosynthesis are associated with the PMf. However, whether these enzymes are specifically bound to the PMf and whether the PMf is a spatially distinct membrane in vivo remained undetermined. In the current study, we combined large-scale analytical methods with live-cell imaging. Our data reveal the broader composition of a spatiotemporally distinct membrane domain, demonstrating colocalization of enzymes with the products of lipid pathways in which they operate. These data support the idea that the PMf is an organizing center for the biosynthesis of specific metabolites in mycobacteria.

Results

The PMf Is a Multifunctional Membrane.

To broadly understand the protein composition of the PMf, we conducted a comparative proteomic analysis and identified a total of 240 and 626 proteins enriched in the PMf and the PM-CW, respectively (Dataset S1). We used the DAVID gene functional classification tool (19) to reveal enrichment of the transport and metabolic machineries of inorganic ions, amino acids, and carbohydrates in the PM-CW (SI Appendix, Fig. S1). The PM-CW was also enriched in enzymes involved in protein trafficking, energy metabolism, and signal transduction (SI Appendix, Table S1). Our proteomic analysis extends the key conclusion of our previous study that the PM-CW is the classical plasma membrane tightly bound to the CW. In contrast, the PMf was enriched in proteins involved in metabolism of specific cell envelope components (SI Appendix, Fig. S1). As predicted, we observed PimB′ (MSMEG_4253), Psd (MSMEG_0861), and Ppm1 (MSMEG_3859) in the pool of 240 PMf-enriched proteins, validating our analysis. Only eight of 240 PMf proteins were predicted to have transmembrane (TM) domains (SI Appendix, Table S2) (20), suggesting that most PMf-associated proteins are peripheral membrane proteins.

Five of the newly identified PMf-associated proteins were chosen for further analysis based on known or predicted function, abundance, protein size, and the presence of a TM domain (SI Appendix, Materials and Methods). They were GlfT2 (MSMEG_6403, UDP-galactosyl transferase), Gtf1 (MSMEG_0389, glycosyltransferase), a geranylgeranyl reductase (MSMEG_2308), PyrD (MSMEG_4198, dihydroorotate dehydrogenase), and a putative membrane protein (MSMEG_1944), of which the last two have predicted TM domains. We expressed these proteins with a C-terminal HA tag in wild-type Msmeg (see SI Appendix, Fig. S15A for the vector design). Western blotting of density gradient fractions showed that all the selected proteins were enriched in the PMf fractions, colocalizing with an endogenous PMf marker PimB′ (Fig. 1A and SI Appendix, Fig. S2A).

Fig. 1.

GlfT2-HA and PimB′ are bound to the same PMf membrane. (A) Western blot of density gradient fractions of GlfT2-HA-expressing cell lysate. GlfT2-HA, 74.9 kDa, anti (α)-HA; PimB′ (PMf marker), 41.4 kDa, α-PimB′; and MptA (PM-CW marker), 54.3k Da, α-MptA. (B) Co-IP of GlfT2-HA and PimB′ and its disruption by mild detergent. Input is the equivalent amount of lysate used in co-IP experiment. HES (Hepes at pH 7.4, EDTA, NaCl) and HESD (HES with mild detergents) indicate buffers used for IP. (C) Negative-stain immunogold TEM of the PMf from GlfT2-HA expressing cells illustrates colocalization of GlfT2-HA (5 nm gold, open arrowheads) and PimB′ (10 nm gold, solid arrowheads). (Scale bar, 100 nm.) (D) Quantification of immunogold-labeled PMf vesicles treated with (as in C) or without (as a negative control) primary antibodies. None, no gold particle detected on vesicle; Both, at least one 5-nm and one 10-nm gold particle detected; HA, one or more 5-nm gold particles detected; and PimB′, one or more 10-nm gold particles detected. n = 60.

Further analysis of GlfT2 and PyrD showed that both GlfT2-HA and PyrD-HA coimmunoprecipitated with PimB′ from crude lysate (Fig. 1B and SI Appendix, Fig. S2B). The coimmunoprecipitation (IP) of PimB′ was dependent on the HA epitope tag (SI Appendix, Fig. S2C). MptA (MSMEG_4241), a PM-CW mannosyltransferase involved in the biosynthesis of lipomannan and lipoarabinomannan (17), was not pulled down under the same conditions, and mild detergents disrupted the co-IP of PimB′, consistent with the idea that the PMf membrane mediates these protein interactions. Previously, negative-stain transmission electron microscopy (TEM) revealed vesicle-like structures in the PMf fractions (12), termed PMf vesicles. Immunogold TEM revealed both GlfT2-HA and PimB′, detected by anti-HA and anti-PimB′ antibodies, respectively, on the same PMf vesicles, reinforcing the co-IP results (Fig. 1C). Of 60 randomly chosen vesicles, 44 vesicles were detected by at least one of the antibodies, and 10 vesicles were detected by both (Fig. 1D). Control experiments established the specificity of this detection (Fig. 1D and SI Appendix, Fig. S3 A–E). Taken together, these data indicate the PMf is a multifunctional membrane bound by a specific set of proteins with known functions.

The Refined Proteome Indicates the Roles of the PMf in Lipid Metabolism and Cell Envelope Biogenesis.

Many PMf proteins were categorized as having known functions in DNA replication and protein translation, rather than membrane biogenesis (SI Appendix, Fig. S1). In addition, GlnA1 (MSMEG_4290), an abundant glutamine synthetase in mycobacteria (21) that forms a ∼600-kDa homo-dodecamer (22), was highly enriched in the PMf (Dataset S1). However, when GlnA1-HA was expressed, it did not colocalize with nor coimmunoprecipitate PimB′ (SI Appendix, Fig. S2 A, D, and E). Thus, large cytoplasmic protein complexes, such as GlnA1 and those involved in DNA replication and protein translation, contaminated the PMf fraction, likely as a result of overlapping sedimentation properties.

We established an epitope tag-based purification system to remove contaminants by creating a transgenic Msmeg strain with the endogenous glfT2 gene replaced with an mCherry (mC) fusion gene, HA-mC-glfT2 (SI Appendix, Fig. S15B). GlfT2 is an essential galactosyltransferase for arabinogalactan biosynthesis (23–25). Transgenic and wild-type strains grew at similar rates (SI Appendix, Fig. S4A), suggesting that HA-mC-GlfT2 replaces the function of the wild-type protein. After confirming the specific PMf localization of HA-mC-GlfT2 (SI Appendix, Fig. S4B), we purified the PMf by density gradient fractionation, followed by anti-HA IP, and performed comparative proteomics on this and an identically treated wild-type sample. Using a stringent criterion of 10-fold enrichment in the PMf from HA-mC-GlfT2-expressing cells, we detected 309 PMf-associated proteins (Dataset S1). Among them, 117 were present in the initial PMf proteome, including previously characterized PimB′, PyrD, Ppm1, Gtf1, Psd, geranylgeranyl reductase, and MSMEG_1944. Importantly, we no longer detected GlnA1 or other suspected cytoplasmic contaminants. Instead, the PMf was reaffirmed as a membrane enriched in proteins involved in cell envelope biogenesis and lipid metabolism, as well as transport and metabolism of secondary metabolites, amino acids, and inorganic ions (SI Appendix, Fig. S4C). Notably, we found acyltransferase MSMEG_2934 involved in AcPIM2 biosynthesis, consistent with previous findings that the early steps of PIM biosynthesis take place in the PMf (12). In addition, we found putative DAG kinases (MSMEG_4335 and MSMEG_1920; the orthologs of Mtb Rv2252 and Rv3218) (26) and 1-acylglycerol-3-phosphate O-acyltransferase (MSMEG_4248; the ortholog of Mtb Rv2182c) (27, 28) enriched in the PMf. These enzymes function in two independent pathways of PA biosynthesis, indicating the possibility that the PMf is the site of PA production, a hypothesis investigated in detail later. Collectively, proteomic analysis indicated the PMf as a specialized membrane–protein complex with defined and interrelated metabolic functions.

Comparative Lipidomics Supports Specialized Biosynthetic Function of the PMf.

Selective association of lipid biosynthetic enzymes with the PMf suggested that the lipid composition of the PMf and PM-CW might be different. Using an established normal phase HPLC-mass spectrometry (MS) platform and mycobacteria-specific ion databases (29, 30), we completed comparative lipidomics analysis of the PMf and PM-CW. Biological quadruplicate analyses detected 11,079 molecular events, which represent linked m/z, retention time, and intensity values (Fig. 2A). After aligning datasets as paired events with equivalent mass and retention time values and calculating ion intensity ratios for all pairs, we enumerated the molecular events that changed in intensity above a high threshold value (twofold, Benjamini-Hochberg corrected P < 0.05). We found 642 and 796 events that were significantly overrepresented in the PMf and PM-CW, respectively (Fig. 2A and Datasets S2 and S3). Thus, 13% of all lipids met this stringent change criterion, demonstrating broad differences in lipid composition of the two fractions.

Fig. 2.

Lipidomic profiling of the PMf and PM-CW by MS reveals shared and distinct lipids. (A) Volcano plot of ions detected after HPLC-MS. Individual ions (circles) detected in the positive and negative ion modes were combined and plotted based on fold-change and statistical significance in PM-CW versus PMf, using paired samples (n = 4). The horizontal dashed line indicates P < 0.05 (Benjamini-Hochberg corrected P value); the vertical dashed lines indicate twofold change. The intensity of individual points indicates their relative ion intensity. All lipids indicated by arrows in A were subjected to CID-MS to identify ions of interest, with five representative lipids shown in B–F. PE (35:0) and PI were detected in both fractions (B and C). Monoacyl PI trimannoside (AcPIM3) was PM-CW-enriched (D). TAG (52:1) and PA (34:2) were PMf-enriched (E and F). The chemical structures and fragments detected after tandem MS are shown.

Identification of Altered Lipids.

The m/z and retention time values embedded in changed events could be used to identify the biochemical identities of key lipids corresponding to the changed events. Because the large number of changed events (1,438) precluded identification of all lipids by collision-induced dissociation MS (CID-MS), we used recently validated methods to prioritize events for biochemical analysis (31, 32). These criteria included detection of multiple isotopes of the same molecular ion (M) or mass intervals characteristic of alkane series with more than one member detected. We further focused on the abundant mycobacterial membrane phospholipids in the MycoMass and MycoMap databases (29), using their reported m/z and retention time values for preliminary identifications of matching ions (Fig. 2A). Tentative identifications obtained by database matching were validated using CID-MS analysis for at least one member of each class (Fig. 2 B–F; SI Appendix, Figs. S5–S9 and Table S3, indicated by CID-MS Y; SI Appendix, Table S4). Other members of the indicated class were assigned when nearly identical retention times and mass variations typical of acyl chain length or unsaturation were detected (SI Appendix, Table S3, indicated by CID-MS N).

Major phospholipid species such as PI (35:0; i.e., 35 carbon chain with no unsaturated bond) and PEs (32:2, 34:1, 34:2, and 35:0) (33) were equally distributed between the PM-CW and the PMf (Fig. 2 A and B and SI Appendix, Fig. S5 and Tables S3 and S4), consistent with previous TLC analysis (12). Consistent with the previous observations that the biosynthetic steps after AcPIM2 take place in the PM-CW (12), AcPIM3 species were enriched in the PM-CW fraction (Fig. 2 A and D and SI Appendix, Fig. S7 and Table S3). We also found three acyl forms of Ac₂PIM2 enriched in the PM-CW, suggesting that inositol acylation to produce tetra-acyl PIMs may take place in the PM-CW as well. In contrast, the PMf is enriched in lipid species that are likely involved in core glycerolipid metabolism. Notably, we found four PA species (32:1, 34:0, 34:1, 34:2) enriched in the PMf (Fig. 2 A and F and SI Appendix, Fig. S9 and Tables S3 and S4). This correlates with the enrichment of PA-producing enzymes in the PMf proteome, as described earlier. Similarly, TAGs (51:1, 52:1) were enriched in the PMf (Fig. 2 A and E), and we found a putative TAG synthetase (Tgs) (MSMEG_0290) enriched in the PMf (Dataset S1). In contrast, DAG, the substrate for Tgs, was found in both the PMf and the PM-CW, although only one acyl species (34:1) could be reliably detected (Discussion). Taken together, despite having a similar major phospholipid composition to the PM-CW, the PMf is uniquely enriched in specific lipid species that correspond to the enzymatic content of the fraction.

The PMf Exists as Discrete Patches in Live Msmeg Cells.

In addition to the HA-mC-GlfT2 strain described earlier, we created a second strain to visualize the PMf by live-imaging fluorescence microcopy. In this strain, endogenous Ppm1 was replaced with the fusion protein Ppm1-mNeonGreen (mNG)-cMyc (SI Appendix, Fig. S15B). Ppm1 is the essential catalytic subunit of PPM synthase (18) and was identified in the refined PMf proteome (Dataset S1). The transgenic strain grew at a rate comparable to wild type, implying that the fusion protein is functional (SI Appendix, Fig. S10A). Furthermore, the fusion protein localized to the PMf during density gradient fractionation (SI Appendix, Fig. S10B). We visualized the PMf using these strains expressing either HA-mC-GlfT2 or Ppm1-mNG-cMyc, and showed that the PMf formed foci of variable size and intensity throughout the cells, with particularly intense foci at the poles (Fig. 3A). To determine whether the fluorescent patches were artificial aggregates induced by fusion proteins (34), we expressed GlfT2 fused with a monomeric mTurquoise2 (mT) in the HA-mC-GlfT2-expressing strain (35). We confirmed PMf localization by density gradient fractionation (SI Appendix, Fig. S11A) and showed that fluorescent patterns of mT-GlfT2-FLAG paralleled HA-mC-GlfT2 patterns (SI Appendix, Fig. S11B). Previously reported autofluorescence from wild-type Msmeg (36) was negligible compared with the fluorescence from mT-GlfT2-FLAG expressed in wild-type cells (SI Appendix, Fig. S12). These data rule out common artifacts that sometimes occur with protein engineering within bacterial cells and corroborate a method for tracking key PMf markers.

Fig. 3.

Live imaging of the PMf markers in Msmeg, showing discrete patches of the PMf. (A) HA-mC-GlfT2 (Left) and Ppm1-mNG-cMyc (Right) expressing cells showing punctate PMf foci often enriched at the poles of the cells. (B) Coexpression of PimE-GFP-FLAG (PM-CW) and HA-mC-GlfT2 (PMf) demonstrates distinct fluorescent patterns. (C) Two PMf-associated proteins (HA-mC-GlfT2 and Ppm1-mNG-cMyc) colocalize in single cells. In all panels, arrowheads indicate polar foci. (Scale bar, 5 µm.) Two representative cells are shown in each panel.

To compare the fluorescent patterns of the PMf and PM-CW, we introduced a PimE-GFP-FLAG expression vector in the HA-mC-GlfT2-expressing strain. PimE is a PM-CW-associated enzyme involved in AcPIM6 biosynthesis (37), although it was not identified in the PM-CW proteome, likely because it is a highly hydrophobic membrane protein (Dataset S1). We confirmed that PimE-GFP-FLAG was enriched in the PM-CW-containing fraction of density gradients (SI Appendix, Fig. S11C). In striking contrast to HA-mC-GlfT2 fluorescence, PimE-GFP-FLAG revealed annular fluorescent patterns, in addition to apparent septal fluorescence, consistent with the PM-CW as the classical plasma membrane tightly associated with CW (Fig. 3B).

To examine the colocalization of the two PMf proteins, HA-mC-GlfT2 and Ppm1-mNG-cMyc, we generated another transgenic strain in which both endogenous GlfT2 and Ppm1 were replaced with HA-mC-GlfT2 and Ppm1-mNG-cMyc, respectively. This transgenic strain grew at a similar rate to wild type, implying functional proteins (SI Appendix, Fig. S10A). We confirmed PMf localization of both proteins by density gradient fractionation (SI Appendix, Fig. S10 C and D). We then showed overlapping in vivo localization of these two proteins (Fig. 3C), with the Pearson colocalization coefficient of 0.89 ± 0.06 (n = 20), which was substantially higher than the coefficient of 0.37 ± 0.16 (n = 20) for HA-mC-GlfT2 (PMf) and PimE-GFP-FLAG (PM-CW) shown in Fig. 3B. Thus, the PMf is spatially distinct from the plasma membrane in live Msmeg cells.

Enrichment of the PMf Is Spatiotemporally Correlated to Actively Growing Poles.

Mycobacteria grow from their polar ends, where we found the strongest enrichment of HA-mC-GlfT2 and Ppm1-mNG-cMyc foci (Fig. 3). Time-lapse microscopy of the strain expressing both PimE-GFP-FLAG and HA-mC-GlfT2 showed that PimE-GFP-FLAG stably associated with the plasma membrane, whereas HA-mC-GlfT2 foci continuously associated with the growing poles (Fig. 4A and Movie S1). In an alternative approach, we stained the preexisting CW with a fluorescent amine-reactive dye, washed the dye, allowed the cells to grow, and visualized the growth of the unlabeled poles (9). We found that the unlabeled growing poles possessed HA-mC-GlfT2 foci (SI Appendix, Fig. S13). These data indicated that the PMf-associated proteins are enriched in the growth pole of the cell.

Fig. 4.

The spatiotemporal correlation of the PMf and polar CW elongation. (A) Frame shots of time-lapse microscopy (Movie S1) of the PimE-GFP-FLAG/HA-mC-GlfT2 dual-expressing cells illustrate the stable annular fluorescence of the PM-CW and the continuous association of the PMf foci at the growing cell poles. Time (minutes) represents the real time since the start of the recording. (Scale bar, 10 μm.) (B) Two representative time-lapse movies of HA-mC-GlfT2-expressing Msmeg cells during the recovery phase after transient exposure to DCS (see Movie S2 for sample 1) show the correlation of newly created ectopic polar growth and the PMf foci. The time indicates when the frame shots were taken in relation to the first frame in real time. Arrows, branch point; arrowheads, polar ectopic growth. (Scale bar, 15 μm.)

To further test the temporal correlation of the PMf with polar cell envelope elongation, we induced ectopic polar growth and abnormal branching of the HA-mC-GlfT2-expressing cells by transiently treating the cells with d-cycloserine (DCS). We found a fraction of cells initiating growth at ectopic poles during the recovery from DCS-induced growth arrest (Movie S2). Importantly, HA-mC-GlfT2 was observed at the site of bifurcation during branching initiation (Fig. 4B) and continuously enriched at the ectopic, DCS-induced growth pole. Thus, our data indicate PMf localization to the poles of mycobacterial cells in two models of growth.

Discussion

Organized lipid microdomains in live mycobacteria were reported in 1999 (38), using fluorescent lipid probes. More recently, cardiolipin enrichment was observed at the septa and poles of actively growing cells with the fluorescent dye 10-N-nonyl acridine orange (39). These studies suggested that the mycobacterial membrane is not entirely homogenous. However, the physiological significance of these observations remained unclear.

In this study, we presented three independent lines of evidence supporting the existence of functionally specialized membrane domain in mycobacteria. First, proteomic analysis identified more than 300 putative proteins that specifically associate with the PMf. Identification of PimB′, Psd, and Ppm1 verified our previous observations (12). In addition, PimB′, and Ppm1 are essential enzymes (16, 18), indicating the critical role of the PMf as a site of essential lipid biosynthetic reactions. We confirmed the PMf association of six proteins by density gradient fractionation and performed further analyses on GlfT2, Ppm1, and PyrD. GlfT2 does not contain predicted TM domains, but rather, its crystal structure implies peripheral membrane association (40). Extending previous studies showing that GlfT2 cofractionates with both membrane and CW (24), our data demonstrated stable and specific binding of GlfT2 to the PMf membrane. In contrast, the membrane association of Ppm1 is mediated by the formation of a heterodimer with the membrane protein Ppm2 (41), which may dictate PMf localization. We did not detect Ppm2 in our PMf proteome, likely because of the highly hydrophobic nature of Ppm2. The PMf association of GlfT2 and Ppm1 predicts that the galactan precursor and PPM, the products of these enzymes, may be found in the PMf lipidome. However, they are transient biosynthetic intermediates and could not be identified by our current analysis. PyrD is a family 2 quinone-dependent dihydroorotate dehydrogenase (42) involved in pyrimidine biosynthesis. It is predicted to be essential in Mtb (43) and is hypothesized to interact with the PMf membrane through a predicted TM domain. Whereas precise molecular mechanisms of how these different proteins associate with the PMf membrane remain to be determined, our data indicate that the PMf harbors multiple essential biosynthetic enzymes.

Second, we established that the PMf has a distinct lipid composition from that of the classical plasma membrane. We confirmed our previous observation that PE and PI, major phospholipid species, are equally distributed in both membrane fractions, suggesting that the PMf and the classical plasma membrane are formed by certain ubiquitous phospholipids. In contrast, all four detectable acyl forms of PA are enriched in the PMf, reinforcing the interpretation of the PMf as a metabolically active membrane with specialized function (SI Appendix, Fig. S14). In addition to the PA-producing enzymes, the PMf proteome was enriched in a putative glycerol phosphate acyltransferase (MSMEG_4703) (Dataset S1) that mediates the committed step of PA biosynthesis, further suggesting that glycerolipid biosynthesis initiates in the PMf (SI Appendix, Fig. S14). Interestingly, a DAG species was present in both membrane fractions, suggesting the subsequent reaction of TAG synthesis can take place in either membrane. Among eight Msmeg homologs of Mtb Tgs enzymes (44), we found MSMEG_0290 enriched in the PMf, consistent with the enrichment of TAG species in the PMf. However, two other Tgs homologs, MSMEG_5242 and MSMEG_1882, were found in the PM-CW, suggesting that TAGs can be made in both membranes but does not accumulate in the PM-CW. Thus, our data may offer a spatially dynamic perspective on the TAG metabolism in mycobacteria (SI Appendix, Fig. S14). Regarding the PIM biosynthesis, we could not identify intermediates such as PIM1 and PIM2 in the lipidome, likely because of their low abundance. Nevertheless, we identified AcPIM2 biosynthetic proteins such as acyltransferase MSMEG_2934 and PimB′ in the PMf proteome (SI Appendix, Fig. S14), confirming the previous studies suggesting that AcPIM2 is produced in the PMf (12). The absence of the first mannosyltransferase PimA in the PMf proteome was expected because it is an amphipathic enzyme transiently associating with membrane by weak ionic interactions (12, 45). Because AcPIM2 is found in both membranes, AcPIM2 produced in the PMf must be distributed to the PM-CW. Although further studies are needed to characterize dynamic features of individual metabolic pathways, our current study clearly demonstrates the distinct lipid composition of the PMf with correlative association of specific enzymes.

Finally, we showed that the PMf forms spatially distinct patches with particular enrichment at the poles of live cells. We further showed temporal correlation of PMf markers with the growth of cell poles. Taken together, the current study provided evidence for the PMf as a spatiotemporally distinct membrane domain in which specific biosynthetic reactions take place. Although enriched at the pole, two observations suggest the PMf is not simply a purified cell pole. First, fluorescent patterns of the PMf membrane domain indicate it distributes throughout the cell as discrete patches in addition to polar enrichment. We speculate that these less intense nonpolar patches are involved in cell envelope maintenance, rather than elongation. Second, some pole-associated proteins are not enriched in the PMf. Most notably, DivIVA, a well-known pole-associated protein (11, 46), was enriched in the PM-CW, rather than the PMf (Dataset S1). Taken together, we conclude that the PMf is a membrane domain spatially distinct not only from classical plasma membrane but also from some known polar structures. We suggest that the PMf acts as an organizing center for multiple biosynthetic enzymes that are critical for cell envelope biogenesis in mycobacteria.

Materials and Methods

Plasmids were made using standard molecular biology techniques (SI Appendix, Fig. S15). Growth of mycobacterial strains, preparation of lysates, and fractionation of lysates by sucrose density gradient were as described (12), with modifications in SI Appendix. Full descriptions of other methods are given in SI Appendix.