Tuberculostearic Acid Controls Mycobacterial Membrane Compartmentalization

ABSTRACT The intracellular membrane domain (IMD) is a laterally discrete region of the mycobacterial plasma membrane, enriched in the subpolar region of the rod-shaped cell. Here, we report genome-wide transposon sequencing to discover the controllers of membrane compartmentalization in Mycobacterium smegmatis. The putative gene cfa showed the most significant effect on recovery from membrane compartment disruption by dibucaine. Enzymatic analysis of Cfa and lipidomic analysis of a cfa deletion mutant (Δcfa) demonstrated that Cfa is an essential methyltransferase for the synthesis of major membrane phospholipids containing a C19:0 monomethyl-branched stearic acid, also known as tuberculostearic acid (TBSA). TBSA has been intensively studied due to its abundant and genus-specific production in mycobacteria, but its biosynthetic enzymes had remained elusive. Cfa catalyzed the S-adenosyl-l-methionine-dependent methyltransferase reaction using oleic acid-containing lipid as a substrate, and Δcfa accumulated C18:1 oleic acid, suggesting that Cfa commits oleic acid to TBSA biosynthesis, likely contributing directly to lateral membrane partitioning. Consistent with this model, Δcfa displayed delayed restoration of subpolar IMD and delayed outgrowth after bacteriostatic dibucaine treatment. These results reveal the physiological significance of TBSA in controlling lateral membrane partitioning in mycobacteria.

ABSTRACT The intracellular membrane domain (IMD) is a laterally discrete region of the mycobacterial plasma membrane, enriched in the subpolar region of the rod-shaped cell. Here, we report genome-wide transposon sequencing to discover the controllers of membrane compartmentalization in Mycobacterium smegmatis. The putative gene cfa showed the most significant effect on recovery from membrane compartment disruption by dibucaine. Enzymatic analysis of Cfa and lipidomic analysis of a cfa deletion mutant (Dcfa) demonstrated that Cfa is an essential methyltransferase for the synthesis of major membrane phospholipids containing a C 19:0 monomethyl-branched stearic acid, also known as tuberculostearic acid (TBSA). TBSA has been intensively studied due to its abundant and genus-specific production in mycobacteria, but its biosynthetic enzymes had remained elusive. Cfa catalyzed the S-adenosyl-L-methionine-dependent methyltransferase reaction using oleic acid-containing lipid as a substrate, and Dcfa accumulated C 18:1 oleic acid, suggesting that Cfa commits oleic acid to TBSA biosynthesis, likely contributing directly to lateral membrane partitioning. Consistent with this model, Dcfa displayed delayed restoration of subpolar IMD and delayed outgrowth after bacteriostatic dibucaine treatment. These results reveal the physiological significance of TBSA in controlling lateral membrane partitioning in mycobacteria. IMPORTANCE As its common name implies, tuberculostearic acid is an abundant and genus-specific branched-chain fatty acid in mycobacterial membranes. This fatty acid, 10-methyl octadecanoic acid, has been an intense focus of research, particularly as a diagnostic marker for tuberculosis. It was discovered in 1934, and yet the enzymes that mediate the biosynthesis of this fatty acid and the functions of this unusual fatty acid in cells have remained elusive. Through a genome-wide transposon sequencing screen, enzyme assay, and global lipidomic analysis, we show that Cfa is the long-sought enzyme that is specifically involved in the first step of generating tuberculostearic acid. By characterizing a cfa deletion mutant, we further demonstrate that tuberculostearic acid actively regulates lateral membrane heterogeneity in mycobacteria. These findings indicate the role of branched fatty acids in controlling the functions of the plasma membrane, a critical barrier for the pathogen to survive in its human host. by mycobacteria. Also, mycobacterial membranes have evolved structural adaptations that are fundamentally different from eukaryotic cells and model eubacterial organisms based on the number and nature of cell envelope layers present. The mycobacterial cell envelope consists of five biochemically distinct layers: the plasma membrane, the peptidoglycan layer, the arabinogalactan layer, the mycomembrane, and the capsule (1)(2)(3)(4). The high impermeability of the mycobacterial cell envelope is attributed to long-chain mycolic acids of the outer mycomembrane. However, the inner plasma membrane may also contribute to the regulation of the permeability of mycobacterial cellular response (5). Indeed, our recent study suggests that mycobacteria have a rapid stress response mechanism to remodel the plasma membrane after exposure to membrane-fluidizing chemicals (6). The main building blocks of the plasma membrane are phospholipids, of which cardiolipin, phosphatidylethanolamine (PE), phosphatidylinositol (PI), and PI mannosides (PIMs) are the major phospholipid species. Fine-tuning the composition of phospholipid headgroups and hydrocarbon chains ensures the overall integrity of the plasma membrane (7).
Further, recent work makes clear that the mycobacterial plasma membrane is laterally heterogenous and actively regulated. Density gradient fractionation of cell lysate revealed two physically separable fractions with distinct densities and content. One, containing plasma membrane free of cell wall components, is called the intracellular membrane domain (IMD). A second denser fraction, containing plasma membrane tightly associated with cell wall components, is called the PM-CW (8,9). The proteome and lipidome of the IMD are distinct from those of the PM-CW; the IMD harbors enzymes that are important for active growth and homeostasis, suggesting that the IMD is a biosynthetic hub in the bacteria (8). Consistent with a role in active growth, the IMD is enriched in the subpolar region, where rod-shaped mycobacteria grow and elongate (10)(11)(12).
In nongrowing stationary-phase cells, the IMD is spatially reorganized and delocalized from the subpolar regions to more proximal columnar parts of the cell (13). IMD delocalization was also observed under nutrient starvation and upon cell wall-targeted antibiotic treatment. Furthermore, membrane-targeted perturbations induced by fluidizers such as benzyl alcohol and dibucaine disrupted the subpolar enrichment of the IMD (14,15). Thus, emerging data suggest that mycobacteria have a mechanism to spatiotemporally coordinate the IMD in response to stress and different growth conditions.
To gain insight into the molecular mechanism of plasma membrane partitioning in mycobacteria, we screened for genes that are critical for Mycobacterium smegmatis to remain viable during and recover from dibucaine treatment. Dibucaine is a topical anesthetic which preferentially inserts into the liquid-ordered phase of a membrane to fluidize the membrane and perturb the lateral membrane organization (16). Here, we used unbiased whole-organism genetic and lipidomic screens and targeted gene deletion to discover that cyclopropane-fatty-acyl-phospholipid synthase (encoded by cfa) controls the membrane recovery response and mycobacterial growth. Mechanistic investigation demonstrates that Cfa plays an essential role in producing an abundant and Mycobacterium-characteristic lipid known as tuberculostearic acid (TBSA) that is distributed among the major membrane phospholipids and controls membrane compartmentalization and growth of M. smegmatis.

RESULTS
Dibucaine is bacteriostatic and transiently delocalizes IMD from the subpolar regions. In our previous study, using recombinant M. smegmatis strains expressing IMDassociated proteins such as MurG-Dendra2, mCherry-GlfT2, and Ppm1-mNeonGreen, we demonstrated that the treatment of M. smegmatis with 200 mg/mL dibucaine for 3 h disrupts the subpolar localization of these IMD-associated proteins (14,15). Using a previously established strain, in which mCherry-GlfT2 and Ppm1-mNeonGreen are expressed from the endogenous loci (8), we first measured bacterial growth and IMD marker dispersion over time during dibucaine treatment. In a 9-h window, we observed no growth and no viability decline (Fig. 1A). Consistent with previous observations (14,15), subpolar localizations of both mCherry-GlfT2 and Ppm1-mNeonGreen were progressively diminished  (Fig. 1B). DivIVA, a pole-associated non-IMD control membrane protein, was unaffected by dibucaine (Fig. 1C). Selective effects on known IMD markers suggested that dibucaine specifically disrupts the IMD. Although the subpolar enrichment of IMD was diminished, the IMD was still biochemically detectable after sucrose density gradient fractionation (Fig. 1D). However, the distribution of IMD marker proteins, Ppm1-mNeonGreen and PimB9, became more diffuse, again suggesting IMD disruption. MptC, a PM-CW marker, was unaffected. We then used a 3-h dibucaine pulse followed by wash and chase to examine recovery by growth rate (optical density at 600 nm [OD 600 ]). Dibucaine showed little effect on growth, and cells resumed a normal rate of growth almost immediately (Fig. 1E). These results show that dibucaine disrupts IMD polarization and attenuates cell replication, but these bacteriostatic effects dissipate when dibucaine is removed.
Cfa protects against dibucaine stress. To discover the genes mediating resistance to dibucaine treatment, we treated a transposon (Tn) library of the M. smegmatis strain expressing mCherry-GlfT2 and Ppm1-mNeonGreen with dibucaine for 3 h. We compared Tn insertion frequencies in dibucaine-treated cells with those in vehicle-treated cells. The volcano plot showed five genes significantly diminished in frequency after dibucaine treatment ( Fig. 2A, Table 1). Among them, cfa (MSMEG_6284) was depleted approximately 4fold in dibucaine-treated cells in comparison to vehicle-treated cells, and it was the most significant genetic change detected (P value, 0.0018), suggesting that cfa is important for either surviving during or recovering from dibucaine treatment.
The cfa gene encodes a putative S-adenosyl-L-methionine (SAM)-dependent methyltransferase involved in cyclopropane fatty acyl phospholipid synthesis and forms a putative operon with the upstream MSMEG_6283 gene annotated as a flavin adenine dinucleotide (FAD)-binding domain protein (Fig. 2B). The cfa operon structure is widely conserved among mycobacteria, and Cfa is a highly conserved protein. For example, the amino acid identity of M. smegmatis Cfa to the ortholog in the key pathogen, Mycobacterium tuberculosis, is 78% while the identity to that in Mycobacterium leprae, which is likewise pathogenic and has otherwise undergone massive restriction in its genome size compared to other mycobacteria (17), is 74% (Fig. 2C). In Escherichia coli, heterologous expression of this operon from Mycobacterium chlorophenolicum resulted in the production of 10-methyl octadecanoic acid, also known as tuberculostearic acid (TBSA), from oleic acid (18). Moreover, one study reported the presence of TBSA in all 61 strains of M. tuberculosis complex and 47 strains of nontuberculous mycobacteria except Mycobacterium gordonae (19). This operon is apparently absent in the syntenic region of the M. gordonae genome ( Fig. 2D and E). Furthermore, the identity of the M. gordonae homolog (gene ID: 2930227449) closest to M. smegmatis Cfa was only 30%, implying that it is unlikely to be the ortholog. These observations together suggest the potential role of Cfa in TBSA synthesis, but its enzyme activity has not been demonstrated, and its in vivo role has not been genetically tested in mycobacteria. Further, any physiological function of Cfa in mycobacterial cells remains unknown.
Cfa is a SAM-dependent methyltransferase. We first tested the suggested SAMdependent methyltransferase activity of Cfa. We cloned cfa into an expression vector, produced with a polyhistidine tag (Cfa-6ÂHis) in E. coli, and purified the protein to near homogeneity by affinity chromatography (Fig. 3A). We combined eluates 5 to 9 and used the combined eluates as a source of purified enzyme. To test the enzyme examined before and after dibucaine treatment. Representative images from biological duplicates are shown. n = 36 (before) or 37 (3 h). (D) Sucrose gradient fractionation of cell lysates of the strain expressing mCherry-GlfT2 and Ppm1-mNeonGreen. The strain was treated with or without dibucaine. Ppm1-mNeonGreen (indicated by asterisks) was visualized by in-gel fluorescence after SDS-PAGE. The fluorescence intensity of each band was quantified and shown in a bar graph. Anti-PimB9 antibody sometimes detects a cytoplasmic protein migrating slightly lower than PimB9, and we do not know the nature of this protein (8). PimB9 and MptC (indicated by asterisks) were visualized by Western blotting using rabbit anti-PimB9 and anti-MptC antibodies. PimB9, an IMD marker; MptC, a PM-CW marker, which was unaffected by dibucaine treatment. Representative results from biological duplicates are shown. (E) Recovery from dibucaine treatment. The same strain expressing mCherry-GlfT2 and Ppm1-mNeonGreen was treated with dibucaine for 3 h, washed, and recovered in a fresh Middlebrook 7H9 medium. The growth recovery was monitored by OD 600 . Time 0 corresponds to the beginning of the recovery period. Representative results from biological duplicates are shown. activity, we incubated the enzyme with SAM and phosphatidylglycerol (PG) carrying palmitic acid (C 16:0 ) at sn-1 and oleic acid (C 18:1 ) at sn-2 as acyl moieties (designated PG C 16:0 / C 18:1 ). The SAM-dependent enzyme activity was measured by the production of S-adenosyl-L-homocysteine (SAH) using a commercially available methyltransferase assay system. As shown in Fig. 3B, we detected a robust production of SAH from SAM only when the purified enzyme was added to the reaction mixture containing SAM and PG C 16:0 /C 18:1 . To test the substrate specificity, we next compared PG C 16:0 /C 18:1 with PG C 16:0 /C 16:0 , which does not carry an oleic acid moiety. While the enzyme activity increased with increasing concentrations of PG C 16:0 /C 18:1 , PG C 16:0 /C 16:0 did not serve as a substrate at all concentrations tested (Fig. 3C). Taken together, our data support the previous prediction that Cfa is a SAM-dependent methyltransferase that utilizes a lipid containing an oleic acid moiety as a substrate. Cfa detectably alters PIMs. To determine the in vivo function of cfa, we next obtained a Dcfa mutant from the Mycobacterial Systems Resource and complemented it with an expression vector for Cfa-Dendra2-FLAG (Dcfa L5::cfa-dendra2-flag, or Dcfa c for short). When grown in a standard rich medium (Middlebrook 7H9) without dibucaine stress, the mutant did not show any significant growth defects (Fig. 4A). Cfa was previously identified as an IMD-associated protein by high-throughput proteomic and fluorescence microscopy analyses (8,20). Indeed, under fluorescence microscopy, Cfa-Dendra2-FLAG was specifically enriched in the subpolar region with sidewall patches (Fig. 4B), suggesting the IMD association of the protein. In our previous study, native Cfa protein was lost from the IMD proteome when the IMD membrane vesicles were purified by immunoprecipitation (8), potentially implying weak association with the IMD. Consistent with this observation, Cfa-Dendra2-FLAG was identified in both the IMD and cytosolic fractions by density gradient fractionation analysis (Fig. 4C).

Role of Tuberculostearic Acid in Mycobacteria mBio
Next, we extracted lipids and analyzed them with high-performance thin-layer chromatography (HPTLC). Major membrane phospholipids (Fig. 4D), including PI, PE, and  cardiolipin (CL), as well as glycopeptidolipids, and trehalose dimycolate ( Fig. 4E), were equivalent in density after cfa knockout. Interestingly, tetraacylated species of PIMs, Ac 2 PIM2 and Ac 2 PIM6, increased after cfa knockout ( Fig. 4F). This outcome of altered PIM pools after gene deletion matches the separately observed response of PIMs to membrane fluidizers (6,14). Combined, these results, as well as the emergence of cfa  Role of Tuberculostearic Acid in Mycobacteria mBio from a membrane fluidizer-based screen (see Fig. 2A), provided a hint for a candidate function of Dcfa in control of membrane fluidization in some way that involves membrane lipids. Global lipidomics of cfa mutants. TBSA is a major fatty acid among M. smegmatis phospholipids (21,22); in particular, PIMs exclusively carry TBSA at the sn-1 position of glycerol (23). If Cfa is involved in TBSA synthesis, the lack of cfa will result in changes in fatty acid composition, which may not be detected by low-resolution HPTLC analysis of bulk lipids. We therefore conducted an unbiased high-performance liquid chromatography mass spectrometry (HPLC-MS)-based analysis of total lipid extracts from the wild type, Dcfa, and Dcfa c in biological quadruplicate. This lipidomics platform broadly measures named phospholipids, neutral lipids, and many hundreds of unnamed lipids that can be tracked across genetically altered samples based on their accurate mass retention time values (Fig. 5A). Among 2,442 identified ion chromatograms, 366 ions were significantly (corrected P , 0.05; fold change, .2) enriched in Dcfa, while 453 were enriched in wild type and Dcfa c. Thus, in contrast to the narrow spectrum of PIM changes seen with HPTLC, the more sensitive MS-based lipidomics approach demonstrated a broader scope of lipidic change. These many lipid changes represented a potential explanation for the observed effects of cfa on recovery from membrane disorder, as well as an opportunity to discover cfa as a putative lipid-modifying gene.
Comparative lipidomics query for loss of TBSA. To identify the lipids for which cfa is essential, we sought targeted analysis of lipids downregulated by cfa deletion. First, we formulated a lipidomic query based on the arithmetic difference between the mass of oleic acid (m/z 282.256) and TBSA (m/z 298.287), which are the proposed precursors and products, respectively, of Cfa. If Cfa is essential for TBSA (C 19:0 ) biosynthesis, substitution with oleic acid (C 18:1 ) would reduce the overall mass of any TBSA-containing lipid by CH 4 (m/z 16.031) (Fig. 5B), and the two lipids would nearly coelute. Indeed, 62 lipid molecular species pairs met these criteria in the comparison of Dcfa versus the wild type and Dcfa c (Fig. 5C), ranging from m/z 450 to 1,800 and in a retention time range of 10 to 33 min. This retention time range matches the general retention time range seen for phospholipids (24). Overall, this pattern suggested broad substitution of C 18:1 for C 19:0 in many types of polar lipids after deletion of cfa.
Identification of cfa-dependent lipids. Next, we sought to identify key cfa-dependent lipids from these ion pairs. The M. smegmatis lipidome is not yet annotated, but we could use mass-based annotations for lipids shared with the M. tuberculosis lipidome (24,25) and confirm chemical structures with collision-induced dissociation mass spectrometry (CID-MS). Given the large number of lipids affected by cfa deletion, we simplified the analysis by taking advantage of the phenomenon whereby structurally similar lipids cluster into groups with similar mass and retention time ( Fig. 5B and  C). Then, we implemented a strategy to chemically solve one compound in each cluster based on embedded mass values and then deduce the structures corresponding to all ions in one group. For example, one member of a cfa-dependent ion pair matched the expected mass of PI (m/z 851.566) and comprised TBSA and palmitic acid (PI C 19:0 / C 16:0 ). This lipid clustered together with a PI compound matching the mass of a chain length variant (PI C 19:0 /C 18:0 ), a saturation variant (PI C 19:0 /C 16:1 ), and 6 isotopes of these molecules (Fig. 5C, inset). We ruled in the PI structure with CID-MS that detected glyceryl-inositol-phosphate and the defining fatty acyl fragments (Fig. 6A). The strong signal of PI C 19:0 /C 16:0 in ion chromatograms (;600,000 counts) was consistent with the known high abundance of this PI species compared with other PI species (8,26,27) (Fig. 6A). Ion chromatograms also demonstrated the complete loss of PI C 19:0 /C 16:0 in Dcfa and its restoration through complementation, establishing the essential role of cfa in its biosynthesis. Importantly, while cfa was essential for the TBSA-containing form of PI, deletion increased the production of PI formed from C 18:1 oleic acid, which is the putative precursor of TBSA (Fig. 6A). All outcomes could be explained if cfa encodes the essential enzyme for converting C 18:1 oleic acid to C 19:0 TBSA.
Extending this analysis of putative cfa-dependent ion pairs to other major membrane phospholipids, other cfa-regulated lipid pairs were identified as PE (m/z 716.5326) and AcPIM2 (m/z 1397.8695) by CID-MS ( Fig. 6B and C). Similar to PI species, the cfa dependence of ion chromatograms of PE and AcPIM2 variants containing C 19:0 fatty acids was complete and genetically complementable, demonstrating an essential role for cfa. Also, both PE and AcPIM2 showed strong increases in forms containing C 18:1 fatty acids, which ruled out the possibility of a general block of PE and PIM biosyntheses and instead indicated the defect in the biosynthesis of TBSA-containing lipids. We note that PI C 34:1 and PE C 34:1 were depleted in Dcfa c ( Fig. 6A and B). We speculate that the complementation resulted in an overproduction of Cfa, leading to the depletion of PI and PE carrying C 18:1 . In summary, the deletion of cfa is associated with a selective defect in TBSA incorporation into many mycobacterial lipid families (Fig. 5 and 6), including at least three major membrane phospholipids, while leaving the larger total pools of these phospholipids containing other fatty acids intact ( Fig. 4D and F).
Effects of cfa deletion on mycobacterial growth and membrane integrity. To determine the physiological function of Cfa in live cells, we next tested if the mutant is more susceptible to dibucaine treatment. We treated Dcfa with 200 mg/mL dibucaine for 3 h and examined the viability of the mutant. Dcfa did not show any increased sensitivity to dibucaine, and CFU did not decline from pre-to post-treatment (Fig. 7A). We next examined the recovery after dibucaine treatment in a pulse-chase experiment and found that the growth rate of Dcfa was slower than that of the wild type initially, but the mutant recovered during culture in dibucaine-free media after 15 to 18 h (Fig. 7B). Thus, Dcfa is defective in recovering from, but not in surviving during, dibucaine treatment under the conditions tested. We tested another membrane fluidizer, benzyl alcohol, and found that Dcfa was not defective in recovering from this membrane fluidizer (Fig. 7C), suggesting that the effect of dibucaine on the mycobacterial plasma membrane is distinct from that of benzyl alcohol.
Depletion of TBSA and accumulation of cis-monounsaturated fatty acids in membrane phospholipids predict increased membrane fluidity, which could explain the slower recovery from membrane-fluidizing stress. At the extreme, excess membrane fluidity can compromise the permeability barrier such that compounds can orthogonally transit through the plasma membrane (28,29). To test if the plasma membrane of Dcfa is more permeable, we used the membrane-impermeable DNA staining dye TO-PRO-3 and the membrane potential probe 3,39-diethyloxacarbocyanine iodide [DiOC 2 (3)]. As a positive control, heat-killed cells became TO-PRO-3 positive and lost membrane potential (Fig. 8A). In contrast, cells lost membrane potential upon treatment with carbonyl cyanide m-chlorophenyl hydrazone (CCCP), a membrane potential uncoupler, while they remained TO-PRO-3 negative (Fig. 8A). We compared the wild type and Dcfa with or without dibucaine treatment in five experiments, which reproducibly found that both the wild type and Dcfa maintained the membrane potential even after dibucaine treatment ( Fig. 8A and B). While subpopulations (10 to 60%) of both wild-type and mutant cells became TO-PRO-3-positive after dibucaine treatment ( Fig. 8A and C), there was no statistically significant difference between the strains (Fig. 8C). With these data taken together, Dcfa is defective in recovering growth after membrane-fluidizing treatments, but Role of Tuberculostearic Acid in Mycobacteria mBio plasma membrane integrity is not substantially compromised to allow orthogonal transit of measured molecules. Membrane domain partitioning is altered by cfa deletion. Our results so far showed that, in the absence of dibucaine challenge, Dcfa is not overtly defective in growth or its plasma membrane integrity. Since Dcfa was initially identified as a mutant sensitive to dibucaine, which alters membrane compartmentalization, we wondered if Dcfa is defective in recovering from dibucaine-induced disruption of membrane partitioning. Biochemically, both the IMD and PM-CW were purified from Dcfa under normal growth conditions (Fig. 9A). However, after 3 h of dibucaine treatment, the IMD marker PimB9 became diffuse and less enriched in the IMD fractions (Fig. 9A) like the wild type (see Fig. 1D). We also observed more diffuse distribution of the PM-CW protein MptC spanning to the lighter density region (Fig. 9A), compared with the wild type (see Fig. 1D), suggesting more severe impact of dibucaine on membrane partitioning in Dcfa. We then examined by microscopy the recovery of the subpolar IMD enrichment over time after dibucaine treatment. In wildtype cells, subpolar enrichment of the IMD marker mCherry-GlfT2 was partially restored after a 3-h recovery and fully restored after 6 h ( Fig. 9B and C). In contrast, for Dcfa, the subpolar enrichment of mCherry-GlfT2 required 12 h for full recovery ( Fig. 9B and C). Thus, Dcfa is delayed in recreating subpolar membrane partitioning after dibucaine treatment.

DISCUSSION
Tuberculostearic acid (TBSA) has been studied for decades as an abundant Mycobacteriumcharacteristic lipid. TBSA is produced by most mycobacterial species, where it amounts to 10 to 20% of fatty acids (30,31). Because TBSA is not produced by humans, it has been proposed as a diagnostic test for tuberculosis (32)(33)(34)(35)(36)(37)(38). The structure of TBSA was determined in 1934 (39), and it was first synthesized in 1947 (40). TBSA biosynthesis from oleic acid was proposed in 1962 (41), which frames the interpretations of the cellular lipidomics performed here.
There are seven and nine paralogs of cfa in M. smegmatis and M. tuberculosis, respectively, and at least two other genes have also been proposed to mediate the biosynthesis of TBSA in mycobacteria (42,43). Our study shows that cfa (MSMEG_6284 in M. smegmatis), rather than the other two previously proposed genes, encodes the SAM-dependent methyltransferase needed for the biosynthesis of many TBSA-containing lipids, including three major families of membrane phospholipids: PI, PE, and PIMs (23,32,44). The defects of Dcfa that are restored with cfa complementation unequivocally demonstrate the necessary and sufficient role of Cfa in TBSA synthesis. Given the prior work by Machida et al. (18), we speculate that the FAD-binding domain protein is the oxidoreductase, which mediates the second step of the reaction, where cis-9,10-methylene octadecanoic acid produced by Cfa is reduced to TBSA. Subcellular location of Cfa remains an open question. Prior studies suggested that Cfa is an IMD-associated protein (8,20), and our fluorescence microscopy analysis supported  Role of Tuberculostearic Acid in Mycobacteria mBio on density gradient fractionation (45). Since Cfa binds both SAM, a soluble substrate, and oleic acid, a lipidic substrate, there may be physiological reasons why Cfa is found in both cytoplasm and the IMD. However, we acknowledge that the localization of Cfa reported in this study may be affected by the attachment of Dendra2, a large fluorescent protein.
Additionally, whereas we successfully used PG C 16:0 /C 18:1 as a Cfa substrate in an enzymatic assay, we do not know if the in vivo substrates of Cfa are phospholipids such as PG or biosynthetic intermediates such as oleoyl-CoA, which may be available in the cytoplasm bound by an acyl-CoA binding protein. Additional studies are needed to determine if and how Cfa associates with the plasma membrane, how it catalyzes the first step of TBSA biosynthesis, how the membrane association contributes to its catalytic activity, and how synthesized TBSA distributes to both the IMD and PM-CW. Through comparative lipidomics, unbiased examination of all ionizable lipids altered through gene deletion in live mycobacterial cells is now possible (46). Using this platform, we revealed a dominant pattern of CH 4 loss among polar lipids in Dcfa. Mechanistically, we strongly favor the interpretation that cfa deletion causes loss of these lipids through ablation of the shared TBSA pool, rather than some adjunct effect of Cfa on overall PI, PE, and PIM synthesis. Lipidomics data demonstrated both the ablation of TBSA-containing phospholipids and increased C 18:1 -containing PE, PI, and AcPIM2. The opposite effects of cfa deletion based on the fatty acid present ruled out a general block in synthesis of these phospholipid products, and it strongly supports that cfa deletion ablated the downstream product (TBSA) and increased the immediate upstream product (oleic acid). Also, the opposite regulation of distinct PI and PE acyl forms by cfa can resolve the otherwise apparent contradiction between preserved total PI and PE pools observed in HPTLC and the complete loss of TBSA-containing PI and PE in MS-based lipidomics.

Role of Tuberculostearic Acid in Mycobacteria mBio
In Mycobacterium phlei, the fatty acid composition shifts from TBSA to cis-unsaturated fatty acids such as oleic and linoleic acids when grown at a lower temperature (31). This observation implies that tilting the balance to cis-unsaturated fatty acid from TBSA helps the cells to maintain membrane fluidity under a colder growth temperature, where the membrane becomes more ordered. Therefore, TBSA may be involved in homeoviscous adaptation, a stress response mechanism (47,48), in which TBSA makes the plasma membrane more rigid than cis-unsaturated fatty acids. Our data are consistent with the importance of TBSA in homeoviscous adaptation as the lack of TBSA and the accumulation of oleic acids in Dcfa would make the mutant's plasma membrane more disordered, explaining the reason why Dcfa is more vulnerable to dibucaine, a membrane-fluidizing molecule. We propose that TBSA biosynthesis is important for creating a lipid environment resilient to external threats that increase the fluidity of the mycobacterial plasma membrane.
Branched-chain fatty acids are important for membrane homeostasis, and their depletion makes bacterial cells defective in membrane fission during L-form proliferation and fluidity maintenance (49)(50)(51). While these studies used iso and anteiso fatty acids in other bacteria, we propose that TBSA in mycobacteria could similarly support membrane integrity. However, our data indicate that the defect in synthesizing TBSA correlates with the slower kinetics of membrane domain partitioning and not with other functions. First, Dcfa grows like the wild type, indicating that there are no gross impacts on growth under laboratory culture conditions. Second, we used the membrane potential probe DiOC 2 (3) and demonstrated that Dcfa can create a proton gradient across the plasma membrane for oxidative phosphorylation. Third, we tested the membrane permeability of Dcfa using the fluorescent DNA-binding dye TO-PRO-3 and did not observe any significant defects relative to the wild type with or without dibucaine challenge. These observations suggested that the lack of Cfa and the resultant changes in the ratio of oleic acid to TBSA do not have significant effects on overall membrane integrity. In contrast, we found that there was a delay in subpolar IMD enrichment in Dcfa during recovery from dibucaine treatment. This delay coincided with delayed recovery of growth after dibucaine treatment. Benzyl alcohol has a similar disruptive effect on membrane partitioning (15), and yet the recovery of growth of Dcfa after the treatment with this membrane fluidizer was no different from that of the wild type. Dibucaine has been suggested to insert into the ordered membrane domain preferentially, while benzyl alcohol preferentially inserts into disordered regions (16,52). Thus, we speculate that the molecular mechanisms of disrupting mycobacterial membrane partitioning are different between these two membrane fluidizers. Although TBSA-containing phospholipids are distributed in both the IMD and the PM-CW (8), they could have more dominant roles in maintaining the integrity of the PM-CW, which, we speculate, is more ordered than the IMD (53). Being consistent with this model, we observed changes in subcellular fractionation patterns of not only IMD but PM-CW proteins in Dcfa upon dibucaine challenge. Overall, our results, combined with known roles of branched-chain fatty acids, support the critical role of TBSA in membrane partitioning. Nevertheless, we acknowledge a less likely scenario that the membrane partitioning defect of Dcfa is not a direct consequence of the deficient TBSA biosynthesis. Determining the precise mechanisms of how TBSA controls membrane partitioning at the molecular level is an important direction for future research.
We recently reported inositol acylation of PIMs as a response to membrane fluidization conserved in M. tuberculosis (6). Our current study implies a potential role of TBSAcontaining lipids, which are widely conserved in mycobacteria, including M. tuberculosis, in homeoviscous adaptation. One transposon mutagenesis study predicts cfa to be essential in M. tuberculosis (54). Furthermore, the cfa ortholog in Mycobacterium bovis BCG is critical for persistence during passage in bovine lymph node (55). These studies illuminate the potential importance of homeoviscous adaptation in M. tuberculosis, an obligate human pathogen that does not experience temperature fluctuations. Thus, our studies suggest the intriguing hypothesis that there are yet-to-be-defined host factors that affect the pathogen's membrane fluidity during host-pathogen interactions.

MATERIALS AND METHODS
Growth and treatment with membrane fluidizers. M. smegmatis mc 2 155 was grown at 37°C in Middlebrook 7H9 broth supplemented with 11 mM glucose, 14.5 mM NaCl, and 0.05% (vol/vol) Tween 80. Dibucaine (Sigma-Aldrich) was added to a log-phase culture at the final concentration of 200 mg/mL, and an equivalent volume of water was used as a vehicle control. After a 3-h incubation at 37°C, the culture was washed with phosphate-buffered saline (PBS) containing 0.05% Tween 80 (PBST) three times and resuspended in the same volume of Middlebrook 7H9 broth for recovery. Benzyl alcohol treatment was identical to dibucaine treatment except that the treatment was for 1 h at the final concentration of 100 mM, as described previously (14). CFU were determined by serially diluting cell culture using Middlebrook 7H9 broth and spotting 5 mL on Middlebrook 7H10 agar supplemented with 11 mM glucose and 14.5 mM NaCl. The agar plates were incubated at 37°C for 2 to 3 days before the number of microcolonies was determined.
Transposon sequencing. We prepared a transposon mutant library using UMycoMarT7 phage as before (15), grew the library of cells to a log phase, and treated the cells with dibucaine (or water as the vehicle control) for 3 h. The cells were washed to remove dibucaine and recovered in Middlebrook 7H9 broth for ;16 h. Genomic DNA was purified, sheared, and barcoded. Transposon insertion sites were then amplified by a nested PCR. To prepare the library for high-throughput DNA sequencing, we used the KAPA library preparation kit (Kapa Biosystems) and TruSeq adapters (Illumina) as before (56). The library was sequenced by 100-bp paired-end sequencing using the Illumina HiSeq 3000 platform. Identified genes were compared between water-or dibucaine-treated samples using TRANSIT as described in the literature (57,58). Library sequencing yielded 5 million unique transposon-inserted-sequences, which covered over 35% of the possible TA sites in the genome. Cfa purification and methyltransferase assay. We amplified the cfa gene from M. smegmatis genomic DNA using a forward primer (59-CAC CGC ATC CAT GAC CAC ATT CAA AGA ACG CGA GAC GTC CAC AGC GG-39) and a reverse primer (59-GGC GGA CCA CCA GGG CCG CA-39) and cloned into pET101 TOPO vector using the Champion pET101 directional TOPO expression kit (Invitrogen). The plasmid was introduced into E. coli BL21 (Invitrogen). The transformed E. coli strain was grown at 30°C for 6 h, and the production of Cfa-6ÂHis was induced for 6 h with 1 mM isopropyl b-D-1-thiogalactopyranoside (Fisher). Cells were washed, resuspended in a lysis buffer containing 1 mg/mL lysozyme (Fisher), 50 mM HEPES-NaOH (pH 7.4), 200 mM NaCl, 0.33 mM phenylmethanesulfonyl fluoride, 0.33 mM dithiothreitol, and lysed by sonication. The lysate was centrifuged to remove cell debris, and the supernatant was applied to 250 mL bed volume of Nickel NTA agarose resin (GoldBio) for affinity column chromatography. The crude lysate was loaded onto the column, and the flowthrough was collected. The resin was washed twice with 1 mL of wash buffer (100 mM HEPES-NaOH [pH 7.5], 500 mM NaCl, and 10 mM imidazole) supplemented with 0.05% Tween 20 and four times with 1 mL of wash buffer (without Tween 20). The bound protein was eluted twice with 200 mL of 100 mM HEPES-NaOH (pH 7.5) containing 75 mM imidazole, once with 200 mL of 100 mM HEPES-NaOH (pH 7.5) containing 125 mM imidazole, and 10 times with 200 mL of 100 mM HEPES-NaOH (pH 7.5) containing 250 mM imidazole. Eluate fractions 5 to 9 were combined, loaded onto an Amicon Ultra-4 centrifugal filter unit (EMD Millipore), and washed three times using 4 mL of 20 mM HEPES-NaOH (pH 7.5).
Live cell imaging. Cells were grown to log phase (OD 600 , 0.5 to ;1.0), and 5 or 10 mL of cell culture was placed on a 1% (wt/vol in water) agar pad on a glass slide, and fluorescent protein localization was visualized using a Nikon Eclipse E600 microscope (100Â objective; numerical aperture [NA], 1.30) equipped with an ORCA-ER cooled charge-coupled-device camera (Hamamatsu) and Openlab software 5.5.2 (Improvision). All fluorescence images were taken at the exposure of 4 s with a gain of 4. Fluorescence intensity profiles were quantified as before (15). Briefly, cell shape was contoured using Oufti (60), and each cell was divided into 100 sections along the long axis. Average relative fluorescence intensity was calculated using MATLAB with published scripts (61) and plotted along the normalized cell length.
Density gradient fractionation, SDS-PAGE, and immunoblotting. Subcellular fractionation was performed as before (8,9). Briefly, cells were grown to a log phase (OD 600 , 0.5 to ;1.0), harvested, and lysed by nitrogen cavitation. The cell lysate was placed on top of a sucrose gradient (20 to 50%, wt/vol) and fractionated by sedimentation at 35,000 rpm for 6 h on an SW-40Ti rotor (Beckman-Coulter) at 4°C. SDS-PAGE and immunoblotting of gradient fractions were as previously described (8,45). Rabbit anti-PimB9 and anti-MptC antibodies were raised previously (63), affinity-purified, and used at 1.0 and 1.1 mg/mL, respectively. Mouse anti-FLAG M2 antibody was from Sigma-Aldrich and used at 1 mg/mL. Horseradish peroxidase-conjugated donkey anti-rabbit (Cytiva) or sheep anti-mouse antibody (Sigma-Aldrich) was used at a 4,000Â dilution as a secondary antibody, and the protein bands were visualized by chemiluminescence. Ppm1-mNeonGreen was visualized by in-gel fluorescence imaging. Both luminescence and fluorescence were detected using either an ImageQuant LAS 4000mini or an Amersham ImageQuant 800 system (Cytiva).
Lipid extraction, HPTLC, and lipidomics. In quadruplicates, each strain was harvested at a log phase (OD 600 , 0.5 to ;1.0), and the wet cell pellets were sequentially treated with 20 volumes of chloroform/ methanol (2:1, vol/vol) relative to cell mass (i.e., 20 mL per g wet pellet), 10 volumes of chloroform/methanol (1:1, vol/vol), and 10 volumes of chloroform/methanol (1:2, vol/vol). A 10% volume of the combined organic phase was set aside for HPTLC analysis, and the remainder was subjected to lipidomic analysis. For HPTLC analysis, combined organic phase was dried under a stream of nitrogen gas, and the dried lipids were resuspended in 5 volumes of chloroform/methanol (1:1, vol/vol) relative to the original pellet weight (i.e., 5 mL per g wet pellet). Then, 10 mL of lipid extracts was spotted onto an HPTLC silica gel 60 sheet (Merck) and chromatographed using chloroform/methanol/13 M ammonia/1 M ammonium acetate/water (180:140:9:9:23, vol/vol/vol/vol/vol) as a mobile phase. PIMs were visualized by spraying the HPTLC plate with an orcinol spray reagent (16 mM orcinol, 2 M H 2 SO 4 , and 14 M ethanol in water) and baking at 120°C. Phospholipids were visualized using a molybdenum blue spray reagent (Sigma-Aldrich). For the purification of glycopeptidolipids (GPLs) and trehalose dimycolate (TDM), dried lipids were further purified by chloroform/methanol/water (8:4:3, vol/vol/vol) phase partitioning. The organic phase containing GPLs and TDM was dried under a nitrogen gas stream, and the dried lipids were resuspended in chloroform/methanol/water (9:1:0.1, vol/vol/vol). Lipids were separated on an HPTLC plate using a mobile phase containing chloroform/methanol/water (9:1:0.1, vol/vol/vol). GPLs were visualized using an orcinol spray reagent. Lipidomic analysis and CID-MS were performed as previously described (8) using an Agilent 1260 Infinity LC system and 6546 quadrupole time of flight (QTOF) mass spectrometer. Dried lipids were dissolved in hexane/isopropanol (70:30, vol/vol) at 1 mg/mL by dry weight and were separated using a normal-phase Inerstil Diol column (GL Sciences, Tokyo, Japan) with 0.1% formic acid and 0.05% aqueous ammonia added to all solvents. HPLC/MS data were analyzed using MassHunter (Agilent) and R for statistical analysis and data visualizations. Raw data and R code are available upon request.
Flow cytometry. A 2-mL aliquot of log-phase cells was treated with 200 mg/mL dibucaine for 3 h at 37°C with shaking. Cells were then treated with 100 nM TO-PRO-3 (Thermo Fisher Scientific) and 30 mM DiOC 2 (3) (TCI Chemicals). After a 15-min incubation, cells were centrifuged at 2,000 Â g for 5 min, and the pellet was resuspended in 2% formaldehyde in PBS for fixation. Fixed cells were washed again and resuspended in PBS for flow cytometry analyses using a three-laser (405 nm, 488 nm, and 640 nm) Dual LSRFortessa instrument (BD Biosciences). As in a previous report (64), DiOC 2 (3) was excited at 488 nm with red emission detected through one filter set (a long-pass [LP] filter [600 nm] and a band-pass [BP] filter [610 nm/20 nm]), and green emission was detected through another filter set (LP filter, 505 nm; BP filter, 525 nm/50 nm). TO-PRO-3 was excited at 640 nm and detected through a BP filter (670 nm/ 30 nm). As a positive control for membrane permeability (TO-PRO-3 labeling), cells were heat-killed for 1 h at 65°C. As a positive control for the disruption of membrane potential [DiOC 2 (3) labeling], cells were incubated with 25 mM carbonyl cyanide m-chlorophenyl hydrazone (CCCP, Sigma-Aldrich) for 1 h. Data were analyzed using FlowJo 10.0 (FlowJo LLC, BD). The red/green ratio of DiOC 2 (3) was determined using the derived parameter function comparing the DiOC 2 (3) red and green median fluorescence parameters.