Role of LmeA, a Mycobacterial Periplasmic Protein, in Maintaining the Mannosyltransferase MptA and Its Product Lipomannan under Stress

Mycobacteria differentially regulate the cellular amounts of lipoglycans in response to environmental changes, but the molecular mechanisms of this regulation remain unknown. Here, we demonstrate that cellular lipoarabinomannan (LAM) levels rapidly decline under two stress conditions, stationary growth phase and nutrient starvation, while the levels of another related lipoglycan, lipomannan (LM), stay relatively constant. The persistence of LM under stress correlated with the maintenance of two key mannosyltransferases, MptA and MptC, in the LM biosynthetic pathway. We further showed that the stress exposures lead to the upregulation of lmeA gene expression and that the periplasmic protein LmeA plays a key role in maintaining the enzyme MptA and its product LM under stress conditions. These findings reveal new aspects of how lipoglycan biosynthesis is regulated under stress conditions in mycobacteria.

(rv0817c) was introduced into the deletion mutant instead of M. smegmatis lmeA, it was able to restore the LM/LAM phenotype of M. smegmatis ΔlmeA (Fig. 1A). LmeA does not have an impact on the biosynthesis of structurally related PIM species, and as expected, M. tuberculosis LmeA showed no effect on their biosynthesis (Fig. 1B). These data suggested that Rv0817c is functionally equivalent to M. smegmatis LmeA. Therefore, M. tuberculosis LmeA, which has been reported to be critical for growth and is upregulated upon mouse infection (24)(25)(26), is a functional ortholog, suggesting its involvement in mannan elongation in M. tuberculosis.
M. smegmatis lmeA is upregulated upon stress. The upregulation of M. tuberculosis lmeA (rv0817c) during mouse infection implies that expression of this gene may be a response to environmental stress. Given that Rv0817c is orthologous to M. smegmatis LmeA, we speculated that LmeA may respond to stress conditions in M. smegmatis as well. To examine if lmeA gene expression is upregulated upon stress, we extracted RNA from cells under stress and examined the transcript levels of lmeA in M. smegmatis strains (the wild type [WT], ΔlmeA mutant, and ΔlmeA::P native lmeA-HA strain) by quantitative real-time PCR (qRT-PCR). As a model stress condition, we first tested the stationary growth phase and found that lmeA was induced 2.5-fold by 72 h in WT cells compared to levels at log phase ( Fig. 2A). In addition, the ΔlmeA mutant complemented with the lmeA gene, which carries its upstream 165 bp as a putative endogenous  promoter region (ΔlmeA::P native lmeA-HA) (23), showed a similar trend of upregulation, although the upregulation was not statistically significant.
The cellular levels of LAM rapidly decline in stationary phase, but the levels of LM remain relatively constant regardless of the growth phase (21,27). Because LmeA plays a role in mannan biosynthesis (23), we speculated that the deletion of lmeA may affect LM maintenance during stationary phase. As with previous studies, we noted a 2.5-fold decrease in LAM levels in the WT by the 72-h time point, and similar declines in LAM levels were observed in the ΔlmeA and ΔlmeA::P native lmeA-HA strains in the stationary phase ( Fig. 2B and D; see also Fig. S1A and S1B in the supplemental material). In contrast, the LM levels of WT cells were not significantly different between log and stationary phases ( Fig. 2B and C; Fig. S1A and S1B). Strikingly, LM levels in the ΔlmeA mutant were 3-fold lower than those in the WT at the 48-h time point (Fig. 2C). The observed decline in LM levels in the ΔlmeA mutant is specific, as there were no specific changes in the levels of other lipids, such as phospholipids and PIMs, that can be attributable to the lack of LmeA (Fig. S1C). Overall, the loss of LmeA not only makes mannan elongation defective but also impacts the maintenance of LM levels during stationary phase.
LmeA is critical for the maintenance of MptA during stress. Since LmeA is involved in a step of LM biosynthesis at or near the MptA-mediated mannan elongation (23), we speculated that the lack of LmeA may affect MptA abundance, leading to the decreased amount of LM during stationary phase. To examine this possibility, we examined the MptA protein levels in anti-MptA Western blots at each time point (Fig. 3A). The abundance of MptA was maintained at similar levels during the stationary phase in WT cells. In contrast, in the ΔlmeA mutant, the protein level was reduced at both the 48-and 72-h time points, and these reductions in MptA abundance were alleviated in the complemented strain. To confirm that this decrease is not a general reduction in protein abundance, we examined changes in the abundance of another mannosyltransferase, MptC, which mediates the ␣1,2 mono-mannose addition. The levels of MptC were comparable among the three strains. There were no specific changes in its abundance that can be attributed to the lmeA gene deletion (Fig. 3B). We therefore conclude that LmeA is necessary for the maintenance of MptA protein.
Is the reduced abundance of MptA protein in ΔlmeA mutant stationary-phase cells due to downregulation of mptA transcription in the mutant? To answer this question, we determined the relative levels of the mptA transcript by qRT-PCR. While there were no significant differences in the levels of expression of mptA between the three strains, a general decrease was observed in mptA transcript levels at 48 h in comparison to levels in the log-phase cells (Fig. 3C). After this time point, the transcript levels of mptA were stable, with no further decrease. The result that mRNA levels were comparable between the WT and ΔlmeA mutant suggests that the cellular MptA level declines at the protein level during stress conditions in the absence of LmeA.
Next, we used starvation as another condition to characterize stress on LM/LAM biosynthesis. We starved M. smegmatis strains (the WT, ΔlmeA mutant, and ΔlmeA:: P native lmeA-HA strain) in phosphate-buffered saline (PBS) for 24 h. We found an increase in lmeA transcripts upon starvation, which was also observed in the ΔlmeA:: P native lmeA-HA strain (Fig. S2A). As with the trend of transcriptional upregulation in the complemented strain, we detected a higher level of LmeA-HA protein when the ΔlmeA::P native lmeA-HA strain was starved (Fig. S2B). As observed in stationary-phase cells, there was a substantial depletion of LAM upon starvation, while LM remained relatively constant in the WT (Fig. 4A, B, and C; Fig. S3A and S3B). As with the observed depletion of LM in the stationary-phase ΔlmeA mutant, LM abundance in the ΔlmeA mutant during starvation was at ϳ50% of the level in the WT. The abundances of other membrane phospholipids appeared similar among the three strains (Fig. S3C). Furthermore, MptA protein levels were ϳ20 times lower in the ΔlmeA mutant than in the WT, a more dramatic fold change than that of stationary phase. In contrast, changes in MptC levels were much milder, being less than 2-fold ( Fig. 4D and E). The ΔlmeA::P native lmeA strain was able to partially complement the phenotype, although MptA levels were 3.5-fold less than that in the WT. These changes in MptA are at the protein level because the mptA transcript did not show specific changes in the ΔlmeA mutant (Fig.  4F). Overall, starvation is another stress condition where lmeA gene expression is upregulated, and the presence of LmeA is critical for the maintenance of MptA and LM.
Given that lmeA transcription was upregulated upon stress exposure and that the deletion of lmeA resulted in decreased abundances of LM both in stationary phase and under starvation, we considered the possibility that the overexpression of LmeA may impact LM abundance even during active growth. We created an episomal expression vector for LmeA-hemagglutinin (HA), in which gene expression is driven by the strong heat shock protein 60 (HSP60) promoter (28). In this LmeA overexpression strain (LmeA OE), LmeA-HA protein levels were ϳ10-fold higher than those in the ΔlmeA:: P native lmeA-HA strain, where lmeA-HA transcription is driven by the putative endogenous promoter (Fig. 5A). In response to the overexpression of LmeA-HA, MptA levels show an increase of ϳ40% in LmeA OE compared to those in the WT ( Fig. 5B and C). Concomitantly with the increase in MptA, LM levels were increased by ϳ80% (Fig. 5D and E; Fig. S4A and S4B). In contrast, in the ΔlmeA mutant, the MptA levels decreased by ϳ1.9-fold compared to those in the WT even in the log phase. The impact of LmeA OE is specific to LM, as it does not show a significant impact on LAM levels and does not impact the abundance of PIMs and other phospholipids (Fig. S4C). Taken together, these data suggest that LmeA not only prevents the reduction of MptA under stress conditions but also helps to enhance the production of LM during active growth.
The expression level of LmeA correlates with antibiotic resistance. Previously, by use of mptA knockdown and mptC deletion mutants, LM/LAM were shown to play important roles in conferring intrinsic resistance to various antibiotics (10,21). Given the significant impacts of lmeA deletion and overexpression on LM/LAM as shown above and previously (23), we examined the impact of these changes in LM/LAM on antibiotic resistance. We determined the MIC 90 s of five antibiotics that target either periplasmic PG biosynthesis (ampicillin, cefotaxime, and vancomycin) or cytoplasmic protein synthesis (clarithromycin and erythromycin) ( Table 1). The ΔlmeA mutant was more sensitive to all five antibiotics than the WT. In particular, the MIC 90 s of cefotaxime and erythromycin for the ΔlmeA mutant were Ͼ10and 8-fold reduced in comparison to those for the WT, respectively. The complementation of the mutant (ΔlmeA::P na -tivelmeA) restored resistance to all antibiotics, but the resistance to vancomycin, cefotaxime, and erythromycin was only partially restored. As we demonstrated previously that antibiotic sensitivity is dependent on the growth medium (9), we further tested sensitivities to these antibiotics in a medium different from Middlebrook 7H9, our standard growth medium. As shown in Table 2, WT M. smegmatis grown in M63 medium became notably more resistant to vancomycin than strains grown in Middlebrook 7H9. The MIC 90 s of clarithromycin and erythromycin also became higher in M63 than in Middlebrook 7H9. Nevertheless, the increased sensitivities to antibiotics found  in the ΔlmeA mutant were reproducible in M63, and the complemented strain restored antibiotic resistance at least partially (Table 2). Taken together, the ΔlmeA mutant is more sensitive to various antibiotics, likely due to its truncated mannan backbone of LM and LAM compromising the cell envelope integrity. LmeA OE, the strain overexpressing LmeA in the WT background, showed enhanced LM abundance compared with that in the WT. We wondered if the increased abundance of LM might impact the antibiotic resistance. As shown in Table 1, LmeA OE showed up to an ϳ2-fold increase in resistance to vancomycin and clarithromycin (Table 1), indicating that the amount of LM correlates with antibiotic resistance.

DISCUSSION
The cellular amount of LAM decreases as M. smegmatis enters stationary phase (21,27). In contrast, the amount of LM remains relatively stable during the growth phase shift. The molecular mechanisms behind the selective maintenance of cellular LM levels remain unknown. In this study, we examined the role of a periplasmic protein, LmeA, under stress conditions. Our data are consistent with a regulatory mechanism in which the lmeA gene is upregulated under stress and in which the increased level of LmeA protein ensures the maintenance of the mannosyltransferase MptA and its product LM.
To support our model, we tested nutrient starvation as another stress model in addition to stationary phase and demonstrated that LAM selectively decreases under both stress conditions (Fig. 6). The decrease in the cellular amount of LAM cannot be explained by the depletion of the LM precursor because LM as well as the biosynthetic mannosyltransferases MptA and MptC remain relatively stable under these conditions. We further showed that the maintenance of MptA under stress conditions requires LmeA. MptA is a key mannosyltransferase that mediates the mannan core elongation, and the defective maintenance of MptA in the ΔlmeA mutant correlated with a reduction in LM levels. Taken together, LmeA appears to function as a regulator in the stress response, preventing the decline of MptA levels and thereby maintaining LM levels.
Since there is no difference in mptA mRNA levels between the ΔlmeA mutant and the WT, LmeA maintains the cellular level of the mannosyltransferase MptA, likely by acting at the posttranscriptional level. How does LmeA prevent the decline of MptA  (29). Furthermore, LmeA also carries a Chorein-N domain, which is also involved in lipid transfer (29). Indeed, we have previously shown that M. smegmatis LmeA can bind to phospholipids (23). Although it is beyond the scope of the current study, we speculate that the full elongation of the mannan chain is coupled with the transfer of LM across the periplasmic space. Such a concurrent process of mannan elongation and lipid transfer would explain why LM intermediates with a short mannan chain accumulate in the ΔlmeA mutant. We further speculate that, in the absence of LmeA, MptA may be stalled with partially elongated LM intermediates and may be subjected to degradation, particularly during stress conditions. The role of LmeA under stress exposure may be further inferred by the location of the gene in the genome. The gene encoding LmeA appears to form an operon with a downstream gene encoding a putative thioredoxin, ThiX (MSMEG_5786 in M. smegmatis and Rv0816c in M. tuberculosis H37Rv), and this operon structure is widely conserved in mycobacteria. ThiX has a predicted signal peptide, suggesting that it is a periplasmic thioredoxin. In Gram-negative bacteria, periplasmic thioredoxins are critical for withstanding redox stress (30). Furthermore, upstream of lmeA is glnR (MSMEG_ 5784/Rv0818), which is involved in responding to environmental stresses, such as redox stress and nitrogen starvation (31,32). glnR forms an apparent operon with mshD, the final enzyme for the synthesis of mycothiol (33), which is a key molecule to regulate the reduced environment in the cytoplasm. These surrounding genes suggest that lmeA is situated in a potential regulon for the redox stress response. How LM/LAM might be important for resisting redox stress remains unclear. However, cell envelope rearrangement is critical for mycobacterial cells when shifting from an actively growing phase to a more stressful nongrowing phase (4). LmeA may play an integral role in regulating the cell envelope rearrangement through facilitating the action of the MptA mannosyltransferase.
Extending previous studies (21,27), our current study shows that LAM rapidly depletes under nutrient starvation, whereas the abundance of LM does not change significantly. These differences in stress response suggest that LM and LAM play functionally different roles during stress exposures. It remains to be examined if LAM is degraded or shed from the cell under these stress conditions. In either case, LM may play a more vital role in the stabilization of the cell envelope under stress conditions. Indeed, LmeA OE showed an increase in LM but not LAM (Fig. 5B), and this increase in LM was correlated with increased antibiotic resistance (Table 1). Rather subtle changes in the structure and abundance of LM and LAM have a significant impact on cell envelope integrity. We propose that LmeA is a stress response protein that plays a critical role in the biosynthesis and trafficking of LM/LAM, thereby impacting their cell envelope integrity and antibiotic resistance.
For the starvation condition, the M. smegmatis strains were grown in Middlebrook 7H9 at 30°C with shaking at 130 rpm until a logarithmic growth phase (OD 600 ϭ 0.6 to 1.0). The cells were then centrifuged for 10 min at 3,220 ϫ g, resuspended in PBS, and incubated for additional 24 h under the same conditions.
Expression vectors. Two vectors, the Mtb_LmeA expression vector (pMUM105) and LmeA overexpression vector (pMUM063), were constructed as detailed below. For pMUM063, lmeA was excised from pMUM054 (23) by digestion with Van91I and XbaI and ligated into pYAB017 (34), which was digested with Van91I, XbaI, and SphI. pYAB017 has an oriM gene and does not integrate into the mycobacterial chromosome. After ligation, the plasmid was electroporated into WT M. smegmatis to create the LmeA overexpression strain. For pMUM105, the gene encoding Rv0817c was PCR amplified from M. tuberculosis genomic DNA using primers A367 and A368 (see Table S1 in the supplemental material). The fragment was then ligated into pDO23A (35) through Gateway BP cloning, creating plasmid pMUM089. The attB-Rv0817c region of this plasmid was then amplified with primers A444 and A445. The PCR fragment was then inserted into pMUM082, which was linearized by ClaI and NdeI, through Gibson assembly. The final plasmid construct, pMUM105, was electroporated into M. smegmatis ΔlmeA.
Cell lysis and protein analysis. Cell pellets were lysed by bead beating as previously described (23). The protein concentration of each lysate was determined with a bicinchoninic acid (BCA) assay kit (ThermoFisher Scientific) and adjusted to 1.2 mg/ml. The lysate (12 l) was mixed with reducing sample loading buffer, denatured on ice for 30 min, loaded onto a 12% SDS-PAGE gel, and electrophoresed at 150 V for 70 min. Western blotting was performed as previously described (21) using rabbit anti-MptA, rabbit anti-MptC, or mouse anti-HA (Millipore-Sigma) at a 1:2,000 dilution as a primary antibody. After incubation with horseradish peroxidase-conjugated anti-rabbit or anti-mouse IgG as a secondary antibody at a 1:2,000 dilution (GE Healthcare.), protein bands were visualized by chemiluminescence and the image was captured with an ImageQuant LAS 4000 mini (GE Healthcare).
Purification of RNA. RNA was extracted and purified from the cell lysate as previously described (37). Briefly, cell culture was centrifuged at 3,220 ϫ g for 10 min. The wet pellets (ϳ400 mg) were then frozen at Ϫ80°C overnight and then resuspend in 1 ml of TRIzol (ThermoFisher Scientific). Cells were lysed with 0.5 ml of zirconia beads using a BeadBug microtube homogenizer (Benchmark Scientific) at 4°C with beating at 4,000 rpm for 1 min, followed by resting on ice for 2 min. The bead homogenization was performed three times in total and followed by a 5-min incubation at room temperature. The samples were then centrifuged for 30 s at 10,000 ϫ g at 4°C. The purified RNA was then concentrated using an RNA Clean and Concentrator kit by following the manufacturer's protocol (Zymo Research). Two micrograms of RNA was used to synthesize cDNA using Maxima H minus reverse transcriptase (ThermoFisher Scientific) and 1 M random hexamers (Promega) in accordance with the manufacturer's instructions.
Quantitative PCR. For each 25-l reaction mixture, 12.5 l of the Roche FastStart Universal SYBR green master mix (Millipore Sigma) was mixed with cDNA and primers to achieve the final concentrations of 100 ng/l and 400 nM, respectively (Table S1). The quantitative PCR was performed on a Stratagene Mx3005P apparatus (Agilent Technologies) with the following program: 95°C for 10 min, followed by 40 cycles of incubations at 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s. Data were analyzed to determine the 2 -ΔΔCT (where CT is the threshold cycle) using the MxPro software and evaluated for statistical significance through GraphPad Prism version 8.
Antibiotic sensitivity. Frozen aliquots of the WT, ΔlmeA, ΔlmeA::P native lmeA-HA, ΔlmeA::P hsp60 lmeA-HA, and LmeA OE strains were prepared from log-phase cultures (OD 600 ϭ 0.5 to 1.0) grown in Middlebrook 7H9, and the numbers of CFU were determined. In 96-well microtiter plates, antibiotics were serially diluted in 100 l of either 7H9 or M63 and mixed with 100 l of cells from the frozen stocks to achieve the final density of 5.0 ϫ 10 3 CFU/ml. The plates were incubated in a humidity chamber at 37°C for Middlebrook 7H9 or 30°C for M63. After a 24-h incubation, 20 l of 0.015% (wt/vol) resazurin solution was added to initiate colorization. After an additional 8 h for Middlebrook 7H9 or 13.5 h for M63 incubation, the plates were read on a spectrophotometer at 570 and 600 nm. The percent difference in cell viability between antibiotic-treated and control cells was calculated as described before (9). The MIC 90 values were calculated using OriginPro 9.1 data analysis software.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
This work was supported by a biomedical research grant (RG-414805) from the American Lung Association and a research grant from the Pittsfield Anti-Tuberculosis Association to Y.S.M.
We thank Sandra Petersen and Kelsey Clements for help with qRT-PCR. We also thank Ian Sparks, Whitney Blocker McTigue, Takehiro Kado, and Breanna Pasko for discussion and critical reading of the manuscript.