A FACS-based screen for MOMP assembly factors

We selected the PorH porin as a model MOMP [14] because its C-terminus was previously shown to be detectable at the cell surface [15]. To monitor PorH assembly in the MM and its surface exposure, a multicopy plasmid encoding PorH with a 6x-His tag on the C-terminus (PorH-His) was constructed. This tagged variant of the porin was as functional as a corresponding untagged porH construct in complementing the kanamycin resistance phenotype of ΔporH cells (S1 Fig). To test for surface exposure of the His tag, cells producing PorH-His were incubated with a commercially available anti-His antibody labeled with Alexa Fluor 647 and imaged using fluorescence microscopy. We observed strong signal at the cell surface, indicating that we can detect surface exposed PorH that has been assembled into the MM (Fig 1A). As a control, we also probed PorH-His surface exposure in cells in which cmt1 has been replaced with a zeocin-resistance cassette (cmt1::zeo). In the absence of Cmt1, PorH-His cannot be O-mycoloylated, leading to a dramatic decrease in PorH levels in the MM [7–9, 13]. As anticipated, cmt1::zeo cells expressing PorH-His showed no surface fluorescence (Fig 1A). We conclude that the PorH-His fusion provides a suitable probe to track MOMP assembly in the MM.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. FACS-based Tn-seq screen for MOMP assembly factors.

(A) Wild-type (WT) or cmt1::zeo cells expressing PorH-His were stained with an anti-His Alexa Fluor 647 antibody and visualized by fluorescence microscopy. (B) FACS plots of WT cells expressing PorH-His (high fluorescent gate), cmt1::zeo cells expressing PorH-His (low fluorescent gate), and the high-density Tn library all expressing PorH-His. Cells were grown until-mid-log, stained with anti-6x His Alexa Fluor 647, and washed three times in 1X PBS. WT and cmt1::zeo cells were used to define the gates for sorting. Gates represent approximations and percentages represent the proportion of the total cell population that are within the experimentally sorted gates. (C) Ranked gene plot of log₂-fold value ratios of transposon insertions enriched in the low fluorescent population versus high fluorescent population. Each dot represents an individual gene. (D) Transposon insertion profiles of sigD, marP, and rip1. Each line signifies a site of transposon insertion, and the height of the line represents the number of reads. Data visualization made using Artemis [16].

https://doi.org/10.1371/journal.pgen.1011127.g001

To screen for mutants that impair PorH assembly in the MM, the porH-His encoding plasmid was transformed into a previously constructed high-density transposon library [17]. Pooled transformants were then grown to mid-log, incubated with the fluorescent anti-His antibody, and sorted into “high fluorescent” and “low fluorescent” populations by FACS (Fig 1B). Wild-type and cmt1::zeo cells harboring the PorH-His plasmid served as controls to define the high fluorescent and low fluorescent gates for sorting, respectively (Fig 1B). Following collection of the high fluorescent (~94% of cells) and low fluorescent (~6% of cells) populations, the genomic DNA of both populations was isolated along with that of the unsorted library. The transposon insertion profiles in each population were then determined by transposon-sequencing (Tn-seq) [18]. We focused on genes that were enriched for insertions in the low fluorescent relative to the high fluorescent population because they were likely to encode factors required for the proper assembly of PorH-His at the cell surface (Fig 1C and S1 Table). As an indication that the screen was working as intended, the cmt1 (cgp_0413) gene was identified as one of the top hits required for PorH-His surface exposure (Fig 1C and S1 Table).

The σ^D envelope stress response is required for proper MOMP surface exposure

Other top hits in the screen in addition to cmt1 were sigD (cgp_0696), marP (cgp_0356), and rip1 (cgp_2207) (Fig 1C and 1D, and S1 Table). The sigD gene encodes an alternative ECF sigma factor that is thought to control an envelope stress response (Fig 2A) [19–21]. It is conserved across the Mycobacteriales order along with its membrane-anchored anti-sigma factor RsdA [22–24]. Similar to the σ^E envelope stress response in E. coli, activation of the σ^D system is thought to involve regulated intramembrane proteolysis (RIP) of the anti-sigma factor by site-1 and site-2 proteases [23,24]. In Mtb, the RseP family member Rip1 functions as the site-2 protease for several anti-sigma factors, including RsdA [23,24]. The site-1 protease for the σ^D response has not been identified. However, the marP gene that was a hit in the screen along with sigD and rip1 is conserved in Mtb and encodes a homolog of the E. coli site-1 protease DegS. MarP is therefore an attractive candidate for the site-1 protease of the σ^D response. Thus, the screen for factors involved in PorH surface display identified many of the components of the σ^D signaling pathway, suggesting that this regulatory system is required for proper MOMP assembly in the MM.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The σ^D pathway is required for MOMP surface exposure.

(A) Model of proposed σ^D envelope stress response activation pathway. Under basal conditions, RsdA sequesters σ^D at the inner membrane and prevents expression of the σ^D regulon. In response to a stress signal, MarP, Rip1, and ClpXP sequentially cleave RsdA to release σ^D to activate the σ^D regulon. (B-D) Flow cytometry of WT cells or σ^D pathway mutants expressing PorH-His. Cells were grown to mid-log and stained with anti-His Alexa Fluor 647. A representative replicate is displayed as a histogram and scaled using unit area. (C) An ectopic complement of sigD was expressed from a P_sod-riboE1 promoter in cells lacking the native copy of sigD expressing PorH-His. Cells were grown with 1mM theophylline for sigD induction prior to staining and analysis. (D) An ectopic complement of cmt1 was expressed from a P_sod-riboE1 in sigD null cells expressing PorH-His. Cells were grown-until mid-log with 1mM theophylline for induction of cmt1 prior to staining and analysis.

https://doi.org/10.1371/journal.pgen.1011127.g002

To confirm the role of the σ^D response in PorH-His transport, we constructed mutants with deletions of sigD and rsdA. The PorH-His plasmid was transformed into these deletion backgrounds and cells were incubated with the fluorescent anti-His antibody. Surface exposure of PorH-His was then analyzed by flow cytometry and compared to wild-type (high fluorescent) or cmt1::zeo cells (low fluorescent). The sigD mutant exhibited an intermediate PorH-His transport phenotype in which fluorescence of the antibody treated cells had labeling intensities between the wild-type and cmt1::zeo cell populations (Fig 2B). This phenotype was complemented by ectopic expression of sigD (Fig 2C). Conversely, cells deleted of the anti-sigma factor rsdA, which causes constitutive σ^D pathway activity, showed enhanced PorH-His surface exposure as indicated by their labeling profile shifting to greater intensities than wild-type cells (Fig 2B).

As porH and cmt1 are both in the σ^D regulon [22,25], we investigated if the impaired surface exposure of PorH-His in ΔsigD cells was due to reduced expression of these genes. We expressed a second copy of the cmt1 from an ectopic location in the sigD deletion strain to increase Cmt1 levels. Surprisingly, when the second copy of cmt1 was expressed at levels that correct the phenotype of cmt1::zeo cells (S2 Fig), there was no restoration of PorH-His surface exposure in the flow cytometry profile of ΔsigD cells (Fig 2D), but rather a strong decrease. Why overexpression of cmt1 has this effect on ΔsigD cells remains unclear. We also tested whether the defect in PorH-His surface exposure in ΔsigD cells was due to changes in PorH expression by assaying protein levels using immunoblot analysis (S3 Fig). We detected native levels of ProtX (S3A and S3B Fig) that fluctuate with σ^D activity, as ProtX is a member of the σ^D regulon [22]. We also detected multiple proteoforms of Por-His that likely correspond to MOMPs at various stages of processing/modification, such as pre- and post-cleavage of a secretion signal and/or pre- and post-O-mycoloylation. All of these forms were drastically reduced in cmt1::zeo cells (S3B and S3C Fig), in agreement with previous studies (S3B and S3C Fig) [7,8,13]. We observed a slight but repeatable decrease in the upper proteoforms(s) of PorH-His upon deletion of sigD (S3 Fig). However, there appeared to be no changes to lower proteoform of PorH-His, which we believe corresponds to the mature protein, in ΔsigD cells. Thus, inactivation of the σ^D pathway reduces the amount of potential precursor forms of PorH-His, but not mature protein. We conclude that the PorH-His surface exposure phenotype resulting from the inactivation of the σ^D pathway is not due to lowered levels of Cmt1 or mature PorH, suggesting a role for members of the regulon in the processing of PorH and its assembly in the MM.

Blocking arabinogalactan or mycolic acid synthesis activates the σ^D pathway

Previous studies in Mtb and Cglu observed elevated expression of specific genes in the σ^D regulon upon treatment with isoniazid and phenol [26,27], suggesting that defects in mycolic acid activates the pathway. To further investigate the signals that activate the σ^D pathway, we constructed a reporter that fused the promoter of the σ^D-controlled gene cgp_2320 [22] to lacZ (P_{cgp_2320}::lacZ). In wild-type cells carrying the reporter, β-galactosidase activity was comparable to that observed in ΔsigD, ΔmarP, and Δrip1 cells, indicating that cgp_2320 promoter is only weakly active if at all in unperturbed cells (Fig 3A). Importantly, deletion of the anti-sigma factor rsdA resulted in elevated levels of β-galactosidase activity, indicating that the P_{cgp_2320}::lacZ reporter is sensitive to changes in σ^D regulation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The σ^D pathway is activated by envelope stress in a MarP- and Rip1-dependent manner.

β-galactosidase assays were performed in biological triplicate. The average activity is graphed with the error bar representing standard deviation. (A) β-galactosidase assay of a panel (left) of treatments or genetic backgrounds that target different aspects of the Mycobacteriales cell envelope (right). Cultures were grown, when indicated, with either 0.25x MIC (“low”, 1.25μg/mL) or 0.5x MIC (“high”, 2.5μg/mL) of the indicated antibiotic. Significance was determined using one-way ANOVA comparing to WT as a control (*** p < 0.0001). (B) β-galactosidase assay of the indicated strains carrying either an empty vector (EV) or plasmid expressing ectopic PorH-His from a P_tac promoter induced with 1mM IPTG. (C) σ^D pathway mutants were grown overnight in the presence (green) or absence (black) of 0.25x MIC of EMB (1.25μg/mL) and σ^D reporter activity was measured by β-galactosidase assay. The displayed numbers denote fold-change in mean reporter activity between the–EMB and + EMB conditions indicated. (D) Investigation of HA-RsdA cleavage under stress conditions (left). Immunoblot analysis (right) of σ^D pathway mutants lacking native rsdA with an ectopic HA-tagged complement expressed from a P_tac promoter (no induction). Cells were grown overnight in the presence or absence of 0.25x MIC (1.25μg/mL) EMB. A strain carrying an empty vector (EV) served as a control.

https://doi.org/10.1371/journal.pgen.1011127.g003

To identify conditions that activate the σ^D pathway, we exposed the reporter strain to a variety of chemical treatments and genetic perturbations that disrupt different aspects of cell envelope biogenesis (Fig 3A). Both ethambutol (EMB) treatment to disrupt AG synthesis and deletion of pks, a homolog of Mtb pks13 that encodes an enzyme required for mycolic acid synthesis, resulted in strong induction of the reporter (19.1-fold and 10.8-fold increases, respectively for EMB and Δpks) (Figs 3A and S4). No effect was observed upon treatment with ampicillin (Amp) to disrupt PG synthesis or deletion of cmt1 to block protein mycoloylation (Fig 3A). To further test perturbations to PG biogenesis, reporter activity was monitored in strains lacking the SEDS-type PG synthase RodA or either of the two class A PBP-type PG synthases (PBP1a and PBP1b encoded by ponA and ponB, respectively), but no induction was observed (S4 Fig). Given that the analogous E. coli σ^E response is induced by unfolded outer membrane proteins that can be artificially generated by overproducing them, we also tested the effect of porH overexpression (S5 Fig) on reporter activity. No induction was observed when the MOMP was overproduced (Fig 3B). We conclude that the σ^D pathway is responsive to defects in the biogenesis of the AG layer and mycolic acids, but not defects in PG synthesis or MOMP assembly.

MarP and Rip1 are required for RsdA cleavage and σ^D pathway activation

With conditions that induce the σ^D response identified, we next investigated the role of the MarP and Rip1 proteases in pathway activation. In other characterized ECF sigma factor systems that employ RIP, blocking proteolysis of the anti-sigma factor through deletion of the site-1 or site-2 proteases prevents sigma factor activation. We therefore monitored the P_{cgp_2320}::lacZ reporter in ΔmarP, and Δrip1 cells treated with EMB. These strains exhibited only 1.9-fold or 4-fold activation upon EMB treatment as compared to the 15-fold increase in σ^D activity after wild-type cells were treated with EMB (Fig 3C). Similarly, deletion of marP and rip1 strongly reduced reporter activity in cells deleted for pks compared to wild-type cells (S6A Fig).

In addition to the σ^D transcriptional reporter, we also monitored processing of the RsdA anti-sigma factor in wild-type cells and mutants lacking MarP and/or Rip1. For these experiments, a functional N-terminal HA-tagged version of RsdA (HA-RsdA) (S7 Fig) was produced as the sole source of RsdA. Importantly, the gene encoding HA-RsdA was expressed under the control of a σ^D-independent promoter. Immunoblotting with anti-HA antibodies detected a band of the expected size of full-length HA-RsdA in untreated wild-type cells (Fig 3D). This band was undetectable following EMB treatment, consistent with the activation of proteolytic processing (Fig 3D). In ΔmarP cells, elevated levels of HA-RsdA were detected in untreated cells, and these levels were unaffected by EMB treatment, indicating that proteolytic processing is blocked (Fig 3D). By contrast, levels of full-length HA-RsdA were similar to wild-type in untreated cells lacking Rip1, but a faint band of ~35kDa was specifically detected in this background (Fig 3D). This ~35kDa HA-RsdA fragment accumulated to high levels following EMB treatment while the full-length protein was no longer detectable (Fig 3D). Results in ΔmarP Δrip1 double mutant cells mirrored those of the ΔmarP single mutant (Fig 3D). Thus, loss of MarP appears to block HA-RsdA processing completely whereas Rip1 inactivation appears to block the processing of a smaller HA-RsdA fragment produced by an initial cleavage step. These results combined with those using the transcriptional reporter support a model in which MarP functions as the site-1 protease and Rip1 as the site-2 protease for the RsdA cleavage reactions that release and activate σ^D.

Disruption of primary arabinan chain formation activates the σ^D response

EMB inhibits polymerization of the arabinan chains of AG from the precursor decaprenyl-arabinose (DPA) [28–32]. Thus, EMB treatment results in the formation of the galactan chain of the AG without its arabinan branches where mycolic acids are attached. Another consequence of EMB treatment is the accumulation of the DPA donor in the outer leaflet of the inner membrane [32], which would sequester decaprenyl lipid carrier needed for the biogenesis of other surface glycopolymers like PG and other lipoglycans [33, 34]. To investigate whether it is the loss of the arabinan chain or the accumulation of DPA that activates the σ^D response, we monitored the activity of the P_{cgp_2320}::lacZ reporter in a collection of mutants blocked at various stages of arabinan chain biogenesis (Fig 4B, 4C, 4E and 4F). Such mutants are possible to construct in Cglu because the MM and arabinan branches of the AG are not essential in this organisms unlike other commonly studied members of Mycobacteriales like Msmeg and Mtb where they are required for growth.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The σ^D pathway responds to defects in the primary polymerized arabinan formation.

For β-galactosidase assays, graphs represent the average reporter activity with standard deviation of three biological replicates. (A) Pathway for DPA biosynthesis. The membrane displayed is the inner membrane. The red square denotes phosphoribose, the asterisk is empty decaprenyl carrier, and the green circle represents arabinan residues. DPPR = decaprenylphosphoryl-5-phosphoribose, DPR = decaprenylphosphoryl-ribose, DPA = decaprenylphosphoryl-D-arabinose. (B-C, E-F) Measurement of σ^D reporter activity by β-galactosidase assay in the indicated strains. Significance was determined using one-way ANOVA comparing to WT as a control (**** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant) (C) β-galactosidase activity in cells grown overnight in the absence (black) or in the presence of either of 0.25x MIC EMB (1.25μg/mL) (green) or 0.25x MIC TCA1 (1.25μg/mL) (checked) or both (green/checked). (D) Model of AG biosynthesis on the outer leaflet of the inner membrane. The gray polygon is galactan core, the green circles are arabinan residues, and the green circles with “X” are terminal arabinan residues. Chemical treatments or genetic disruptions that activate the σ^D pathway in red. The hypothesized molecule that is sensed is boxed in red.

https://doi.org/10.1371/journal.pgen.1011127.g004

Synthesis of the DPA precursor of the arabinan chain occurs on the cytoplasmic face of the inner membrane (Fig 4A). UbiA (Cgp_3189) loads phosphoribose-diphosphate onto the decaprenyl carrier to form decaprenylphosphoryl-5-phosphoribose (DPPR) [35,36]. DPPR is then converted to decaprenylphosphoryl-5-ribose (DPR) by Cgp_3190/Cgp_3193 [37]. The essential enzyme DprE1 (Cgp_0238) and redundant enzymes DprE2/Cgp_1680 then convert DPR into DPA [38,39], which is then flipped to the periplasmic face of the inner membrane by an unidentified transporter. After the DPA donor is flipped, a set of arabinofuranosyltransferases add arabinan residues to the galactan core (Fig 4D) [40–42]. First, AftA (Cgp_0236) adds the priming arabinan residue [40] and Emb, the target of the EMB drug [29,30,43], polymerizes the arabinan residues into a linear chain [28]. AftC (Cgp_2077) initiates branching of the growing chain [42] and AftD (Cgp_3161) elongates these branches [41]. Finally, AftB (Cgp_3187) caps the non-reducing ends of the growing AG molecule, generating the sites for modification with mycolic acid to form the inner leaflet of the MM [44,45].

Blocking DPA synthesis either by deletion of ubiA or treatment with the DprE1 inhibitor TCA1 [46] resulted in strong σ^D activation (Fig 4B and 4C). Co-treatment of cells with both EMB and TCA-1 resulted in reporter activity that was equivalent to either treatment alone (Fig 4C). As expected, disruption of emb resulted in σ^D activity as did inactivation of aftA to prevent arabinan chain initiation (Fig 4E and 4F). By contrast, inactivation of the arabinofuranosyltransferases that decorate the primary arabinan chain (aftC,D, or B) did not result in a significant increase in σ^D activity (Fig 4F). Thus, these results suggest that the σ^D signaling pathway does not respond to a reduction in the free pool of the decaprenyl lipid carrier or to elaborations in AG branch chaining but instead responds to the loss of the primary arabinan chain itself.

In addition to their use in forming the arabinan chain of the AG layer, DPA precursors are transferred to an inner membrane lipoglycan called lipomannan (LM) to form lipoarabinomannan (LAM) [47]. The arabinosyltransferase(s) responsible for LM/LAM formation are currently unknown. LMs and LAMs play critical role(s) in the cell envelope as evidenced by the fitness defects upon deletion of LAM biosynthetic enzymes [48–51]. In Mtb and Msmeg, the Emb homolog EmbC catalyzes the transfer of arabinan residues to LAM and is inhibited by EMB [52]. Thus, it is formally possible that activation of the σ^D pathway by EMB treatment is due to defective LAM biosynthesis. To investigate this possibility, we monitored σ^D activity in strains deleted for the genes encoding the mannosyltransferases MptA (Cgp_2385) [50], MptB (Cgp_1766) [49], or MptC (Cgp_2393) [51] needed for LM and LAM biogenesis. σ^D activity was unaffected in these mutants, indicating that defective LM/LAM production is not a signal sensed by this stress response pathway (S8 Fig). Based on the overall analysis of the AG and LM/LAM pathways, we conclude that the σ^D response is induced by inhibiting the biogenesis of the primary arabinan chain of the AG layer. Accumulation of DPA, defects in addition of branches to the arabinan chain, or the inhibition of LM/LAM biogenesis do not appear to be monitored by the response.

Evidence that the σ^D pathway monitors levels of mature mycolic acids in the inner membrane

A common feature of the σ^D activation conditions we identified, EMB treatment and deletion of pks, is that they both prevent proper MM biogenesis [44]. EMB inhibits the polymerization of the arabinan chains of AG used for the covalent attachment of mycolic acids to form the MM whereas inactivation of Pks completely blocks the synthesis of mycolic acids to prevent MM biogenesis. Thus, blocking either an early or late step required for MM biogenesis results in σ^D activation. We were therefore curious whether blocking any step in mycolic acid synthesis or transport activates the σ^D response.

Mycolic acid biosynthesis (Fig 5A) begins with the formation of long chain fatty acids by fatty acid synthase I enzymes to generate acyl-CoA molecules that will become the α branch and the meromycolate chains of mycolic acids [1,53]. These molecules are activated by either the acyl-CoA carboxylase PccB (Cgp_3177, homolog of Mtb AccD4) [54] or the acyl-AMP ligase FadD2 (Cgp_3179, homolog of Mtb FadD32) [55]. The two activated fatty acids are then condensed into a myolic acid molecule by Pks in a Claisen-like condensation reaction [56–58]. Pks activity requires 4’-phosphopantetheinylation by the enzyme PptA (Cgp_2171, homolog of Mtb PptT) [59]. Once formed, the mycolic acid molecule is linked to trehalose by Pks [56]. Trehalose can be imported into the cell by a dedicated transporter complex [60] or synthesized in the cytoplasm from glucose by either the OtsAB [61], TreS, or TreYZ biosynthetic pathways [62,63]. The trehalose mycolate is then reduced by CmrA (Cgp_2717) to form trehalose monocorynomycolate (TMCM) in the inner leaflet of the inner membrane [64,65]. TMCM is further modified by the proteins MmpA (Cgp_3165) and TmaT (Cgp_3163), and these modifications appear to be required for proper TMCM transport [66,67]. AhfA (Cgp_0475) has also been shown to be required for TMCM synthesis in Cglu, but its role in the process remains unclear [68].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The σ^D pathway responds to the absence of mature mycolic acid in the inner membrane.

(A) Pathway for mycolic acid biosynthesis. Proteins in red denote steps that activate the σ^D pathway when deleted. The molecule boxed in red is hypothesized to be sensed by the σ^D pathway. (B-D) Data shown is the average of three replicates and error bars show standard deviation. Significance was determined using one-way ANOVA comparing to WT as a control (** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant). (B) Reporter analysis by β-galactosidase assay in a panel of mycolic acid biosynthetic mutants. (C) β-galactosidase assay of cells treated with the vehicle control (DMSO) or 0.025mM AU1235 [67] in control strains or strains lacking cmpL4. (D) σ^D reporter activity as assayed by β-galactosidase activity in the indicated cmpL mycolic acid transport mutants.

https://doi.org/10.1371/journal.pgen.1011127.g005

Deletion of genes encoding the enzymes involved in the early steps of mycolic acid synthesis (pccB, fadD2, pptA) all induced the P_{cgp_2320}::lacZ reporter to levels comparable to the deletion of pks (Fig 5B). Deletion of ahfA also induced the response, further supporting its connection with mycolic acid synthesis (Fig 5B). Inactivation of the first enzymes in two of the three major trehalose biosynthetic pathways in Cglu [62,69], otsA (cgp_2907) and treY (cgp_2323), individually did not induce the reporter (S9 Fig). However, a double mutant lacking both enzymes, which has been shown to deplete trehalose from the cell wall [62], resulted in activation of the σ^D reporter (S9 Fig). Inactivation of the trehalose-mycolate reducing enzyme CmrA also activated the response but defects in the additional modifying enzymes (ΔtmaT and ΔmmpA) did not (Fig 5B). Thus, blocking the formation of TMCM prior to the final two modification steps triggers the σ^D pathway.

We next tested whether blocking the transport of TMCM to the MM might also serve as an inducing signal for the σ^D response. In Mtb, a single transporter called MmpL3 has been implicated in the transport of mycolic acids across the inner membrane [70–72]. In Cglu, two partially redundant transporters, CmpL1 (Cgp_3174) and CmpL4 (Cgp_0284), are thought to perform the same function. The MM is formed in mutants lacking a single transporter but not in a double mutant. Only CmpL1 is sensitive to the MmpL3 inhibitor AU1235 [73,74]. We therefore tested whether this compound would induce the σ^D reporter in cells in which cmpL4 was deleted where AU1235 should effectively block all remaining TMCM transport by CpmL1. Surprisingly, no induction was observed in ΔcmpL4 cells treated with AU1235 (Figs 5C and S10). Similarly, no induction was observed in a mutant deleted for both cmpL1 and cmpL4 or a mutant deleted for all four genes in Cglu encoding MmpL-like proteins (Fig 5D). We therefore infer that the σ^D pathway is not responding to defects in MM assembly per se in these instances but rather that it senses the lack of TMCM in the inner membrane.

MarP is the site-1 protease of the σ^D pathway

Prior work investigating the σ^D response in mycobacteria revealed that RsdA is the anti-sigma factor and that the site-2 protease that helps release σ^D from RsdA to activate it is Rip1 [23,24]. However, the identity of the site-1 protease that initiates the RIP cascade has remained unknown. Our results provide strong evidence that MarP serves this function. Mutants lacking MarP were hits in our screen along with those inactivated for sigD and rip1 (Fig 1D). Cells deleted for marP are also defective in the activation of a σ^D responsive promoter upon exposure to inducing conditions such as EMB treatment or the inactivation of pks (Figs 3C and S6). Furthermore, immunoblotting revealed that RsdA processing is blocked in ΔmarP cells following EMB treatment (Fig 3D). Thus, the likely signaling cascade for σ^D activation involves the sequential processing of RsdA by MarP and Rip1 followed by its final cleavage by ClpXP to release σ^D to transcribe its regulon (Fig 2A) [75].

A potential sensing mechanism monitoring mycolic acid in the inner membrane

To investigate the signals leading to induction of the σ^D pathway, we tested its response to defects in MM assembly. Deletion of genes encoding early steps in the synthesis of mycolic acid preceding formation of mature trehalose monomycolate (TMCM) (ΔfadD2, ΔpccB, Δpks, ΔpptT, ΔotsAΔtreY, and ΔcmrA) activated the σ^D pathway (Figs 5B and S9). Surprisingly, however blocking the modification or transport of TMCM needed for MM assembly (ΔtmaT, ΔmmpA, ΔcmpL1/4, and AU1235 treatment) failed to stimulate the response (Fig 5B–5D). Thus, the RIP cascade leading to σ^D activation appears to be held inactive as long as trehalose-linked mycolic acids are present in the inner membrane. This observation suggests a potential role for these lipids as repressors of RsdA processing, potentially through their association with the anti-sigma factor or the regulatory proteases. Notably, a previous study identified Rip1 as a potential target of O-mycoloylation due to the presence of a conserved O-mycoloylation motif in the transmembrane region of the protein [7]. Serine-serine signatures that are found in this O-mycoloylation motif are also present in both RsdA and MarP, suggesting an attractive potential mycolic acid sensing mechanism involving responsiveness to the lipidation status of one or more proteins in the RIP cascade.

A second activation signal involving defects in AG biosynthesis?

In addition to defects in mycolic acid synthesis, inhibition of AG biosynthesis also serves as an inducer of the σ^D response. Our results suggest that the molecule in the AG biogenesis pathway that is monitored by the σ^D pathway is the primary arabinan chain of AG molecules. Indeed, both genetic backgrounds (ubiA::kan, emb::kan, and aftA::kan) and chemical perturbation (TCA1 and EMB treatment) that truncate AG molecules results in pathway activation (Fig 4B and 4E). We note, however, that the possibility remains that the aftA::kan mutation is polar on the downstream gene, emb. In contrast, inhibiting branching and terminal capping of AG did not activate the response (Fig 4F). Thus, disruption of MM biogenesis is not the ultimate inducer of the σ^D response. One possible connection between the inducing properties of the mycolic acid and AG defects might be via a feedback mechanism through which defects in synthesis of the primary arabinan chain of AG causes the down-regulation of TMCM synthesis. However, prior studies have shown that although blocking AG biosynthesis via EMB treatment results in a decrease of cell-wall bound lipids, it does not decrease mycolic acid production. Rather, EMB treatment leads to unchanged or, in some studies, increased TMCM biosynthesis [76–80]. Thus, we believe that the mycolic acid and AG activating signals are distinct and that the σ^D response monitors the status of multiple aspects of the envelope. How the σ^D pathway might sense the AG defects is not clear, but an attractive possibility is that RsdA might use its long, disordered periplasmic domain of approximately 160 amino acids to detect AG defects in a similar manner to which the disordered domain of the anti-sigma factor RsgI in Bacillus subtilis senses cell wall defects to induce the σ^I response [81]. Alternatively, activation may be induced by a buildup of galactan core in the inner membrane or possible side effects of EMB treatment and/or mutants defective in AG biogenesis on inner membrane integrity.

Potential role of the σ^D response in maintenance of envelope homeostasis

The regulons controlled by alternate sigma factors often directly address the defects that promote pathway activation to re-establish envelope homeostasis. The best characterized example of this homeostatic response is the σ^E system in E. coli. Defects in outer membrane protein assembly and mislocalized lipopolysaccharide cause σ^E activation [82,83]. Accordingly, the σ^E regulon includes genes encoding periplasmic chaperones, the outer membrane protein assembly machinery (BAM complex), and components of the lipopolysaccharide transport machinery (Lpt system) [84,85] to ameliorate the problems that stimulated induction of the response.

Previous studies have shown that the σ^D regulon contains a number of genes related to mycolic acid biosynthesis (fadD2, pks, pccB), cell wall homeostasis (lppS, putative glycosyltransferases), and MM integrity (porA, porH, mycoloyltransferases) [22,25,26]. Under conditions of reduced TMCM in the inner membrane that we have shown activate the σ^D pathway, increased expression of mycolic acid biosynthetic genes may enable to cell to rapidly increase TMCM production to respond to a perceived deficit. Similarly, the alteration of peptidoglycan structure by increasing the level of the crosslinking enzyme LppS may circumvent defects in AG biosynthesis that promote σ^D activation. Indeed, previous work has suggested that in the absence of normal peptidoglycan crosslinks AG is not properly ligated onto peptidoglycan [86]. We therefore propose that the σ^D regulon, like the σ^E response in E. coli, is similarly poised to fix problems in envelope assembly that promote its induction.

A potential role for σ^D regulon members in MOMP assembly

Multiple regulatory components of the propose σ^D pathway were hits in our screen for MOMP assembly factors. Additionally, activation of the σ^D response was found to alter the abundance of different proteoforms of PorH, which is suggestive of defects in proper MOMP processing or modification needed for assembly into the MM. Thus, just as the E. coli σ^E regulon includes components of the BAM machinery needed for β-barrel protein assembly in the outer membrane, we think it is likely that the σ^D regulon in Cglu includes components of the MOMP processing and assembly pathway that are yet to be identified. We anticipate that the MOMP surface exposure assay developed as a part of this study will be useful in testing this hypothesis and defining this important aspect of envelop assembly in corynebacterial and their relatives.

Bacterial strains & growth conditions

C. glutamicum MB001, a prophage-free strain derived version from ATCC 13032, was used for all experiments [87]. C. glutamicum strains (S2 Table) were grown in brain heart infusion (BHI) medium (BD) or brain heart infusion medium supplemented with 9.1% sorbitol (BHIS) at 30°C with aeration, as indicated. Ectopic expression constructs were induced either with 1mM theophylline when expressed from attB(P_sod-riboE1) [88,89] or relied on leaky expression (no IPTG) from an inducible P_tac promoter when expressed from a pTGR5-derived replicating plasmid [90], as indicated. E. coli DH5α(λ_pir) cloning strain (NEB) was propagated at 37°C in LB (1% tryptone, 0.5% yeast extract, 0.5% NaCl) with aeration unless harboring pCRD206 derivatives, in which cells were cultured at 30°C. C. glutamicum strains were grown with 15μg/mL kanamycin, 3.5μg/mL chloramphenicol, or 150μg/mL zeocin, when appropriate. E. coli strains were grown in 25μg/mL kanamycin, 25μg/mL chloramphenicol, or 25μg/mL zeocin, when necessary. Chemical inhibitors were used at the following concentration: EMB high: 2.5μg/mL, EMB low: 1.25μg/mL, ampicillin high: 2.5μg/mL, ampicillin low: 1.25μg/mL, AU1235: 0.025mM, TCA1: 1.25μg/mL.

Plasmid construction

Plasmids were constructed using isothermal assembly (ITA) (S3 Table) and transformed into E. coli DH5α(λ_pir) (NEB) competent cells by heat shock (42°C for 40 seconds). Primers used in plasmid construction are listed in S4 Table and were purchased from IDT or Genewiz. gBlocks were purchased from IDT using IDT’s codon optimization tool for Corynebacterium glutamicum. To construct pCRD206 derivatives [91] to delete genes by allelic exchange, 500-750bp fragments encoding homology regions upstream and downstream of the gene of interest leaving a small scar fragment of the coding region to maintain the reading frame were assembled into the pCRD206 by ITA. pCRD206 derivatives [91] to disrupt genes by allelic exchange were constructed in the same manner, however a zeocin-resistance cassette was encoded within the scar fragment. The pEMH25 empty vector was constructed by replacing the lacI/P_tac::eGFP fragment from the published pTGR5 vector [90] with the P_sod promoter and native sod RBS from the C. glutamicum genome using ITA. Subsequently, the kanamycin resistance cassette was replaced with a chloramphenicol resistance cassette using ITA and the region downstream of the promoter was deleted using site-directed mutagenesis using KLD enzyme mix (NEB). The pEMH120 empty vector was constructed from a pTGR5 [90] derivative that is chloramphenicol resistant by removing the native BamHI site, inserting a new BamHI site downstream of the P_tac promoter, and deleting the eGFP open reading frame using site-directed mutagenesis with KLD enzyme mix (NEB). Ectopic complementation vectors were cloned onto pSEC1 (derivative of pK-PIM integrating vector, P_sod-riboE1, kanamycin-resistant or zeocin-resistant) [88,89] or pEMH120. All genes were amplified from MB001 gDNA. Tags (6x His or HA) were introduced using site-directed mutagenesis with KLD enzyme mix (NEB).

Colony PCR in C. glutamicum

Colony PCR was performed using either Q5 (NEB) or Sapphire master mix (Fisher Scientific, Clonentech Labs) from cells grown on solid agar. For Q5 colony PCR, cells were lysed in 10μL of 2mg/mL lysozyme for 30 minutes at 37°C. Lysis was completed by adding 10μL of nuclei lysis buffer (Promega) and diluting with 100μL of ddH₂O. PCR reactions were set up with 6.25μL of Q5 master mix, 5.25μL of ddH₂O, 0.625μL of forward and reverse primer (10μM), and 0.5μL of lysate. For Sapphire colony PCR, cells were lysed in 5μL of 1mg/mL lysozyme for 30 minutes at 37°C. Lysis was completed by adding 5μL of nuclei lysis buffer (Promega) and 180μL of ddH₂O. PCR reactions were set up with 5μL of Sapphire master mix, 3.7μL of ddH₂O, 0.5μL of forward and reverse primer (10μM), and 0.3μL of lysate.

Strain construction

C. glutamicum competent cells were prepared as previously described [17,92]. Briefly, strains were sub-cultured from an overnight culture into transformation media (BHI supplemented with 91g sorbitol, 0.1% Tween 80, 0.4g isoniazid, 25g glycine) and grown until in mid-log (approximately 3 hours at 30°C or overnight at 18°C). Cells were washed two times in 10% glycerol spinning at 3,000rpm and resuspended in 10% glycerol. Gene deletion or disruption was performed using allelic exchange via sacB counterselection with pCRD206 derivative plasmids [91] or by recombineering, as described in more detail below. Integration of pK-PIM derivatives (pSEC1 variants) [89] was validated by colony PCR.

Recombineering in C. glutamicum

Gene deletions in C. glutamicum by allelic exchange

Allelic exchange via sacB counterselection was performed as previously described [91]. The pCRD206 derivative was transformed into the appropriate strain via electroporation and recovered at 25°C on BHI(S) Kan15 to allow for plasmid replication. Transformants were restreaked to BHI(S) Kan15 at 30°C to isolate integrants. Integrants were grown overnight at 25°C in 1mL BHI(S) and plated on BHI + 10% sucrose agar at 30°C to select against sacB encoded on the pCRD206 vector. Candidates were further screened by colony PCR with primers that anneal upstream and downstream of the gene of interest to identify isolates harboring the deletion.

Tn-seq screen and fluorescence activated cell sorting (FACS)

The high-density transposon library was previously generated [17]. Tn library competent cells were prepared as previously described [17,92]. pEMH27 was transformed into 10 aliquots of Tn library competent cells by electroporation, as previously described [17,92]. Transformed cells were recovered for 1 hour at 30°C and plated undiluted on BHI Cam3.5 at 30°C. Plates were grown for 24 hours at 30°C and 16 hours at 25°C. Cells were scraped and frozen in 15% glycerol. A frozen aliquot of the Tn library harboring pEMH27 (OD₆₀₀ = 28) was thawed on ice. Approximately 7 x 10⁷ cells (5μL) of the thawed library was used to sub-culture 5mL of BHIS Cam3.5 and grown at 30°C for 4.5 hours. Sub-cultures of control strains (MB001 pEMH27 and cmt1::zeo pEMH27) were made by inoculating 50μL of an overnight culture in 5mL BHIS Cam3.5 and grown at 30°C for 4.5 hours. After growth, and OD₆₀₀ equivalent to 0.1 was pelleted for 3 minutes at 12,000rpm. The pellet was resuspended in 20μL of 1X phosphate buffered saline (PBS) and anti-6x His Alexa Fluor 647 (abcam, cat: ab237337) was added at a dilution of 1:100. Cells were stained in the dark for 1 hour. The treated cells were washed three times in 1X PBS, spinning for 3 minutes at 5000 x g between washes. The final cell pellet was resuspended in 1mL of 1X PBS. FACS was performed on a Beckman Coulter MoFlo Astrios EQ high speed cell sorter using the APC optical configuration (640nm laser, 671/30 bandpass filter). Gates were defined using MB001 pEMH27 cells to gate the “positive/high fluorescent” population and cmt1::zeo pEMH27 cells were used to define the “negative/low fluorescent” population. After sorting, 1.797 million cells from the positive/high fluorescent and 107.68 thousand cells from the negative/low fluorescent populations were obtained. The populations were diluted 1:10 in 1X PBS and the entire negative/low fluorescent population and approximately one third of the positive/high fluorescent populations were plated to BHI Cam3.5 at 30°C. Colonies from the positive/high fluorescent population were scraped and stocked after 24 hours of growth at 30°C. Colonies from the negative/low fluorescent population were scraped and stocked after 64 hours at 30°C.

Tn-seq library preparation

Aliquots of the final populations were thawed and the gDNA was extracted using the Wizard Genomic DNA Purification Kit (Promega). Briefly, populations were lysed by: (1) resuspending a cell pellet in 540μL of 50mM EDTA and 60μL of 10mg/mL lysozyme at 37°C for 30 minutes, (2) addition of 600μL of Nuclei Lysis Buffer (Promega), and (3) incubation at 80°C for 5 minutes. Cell lysates were treated with 3μL of 4mg/mL RNase solution and incubated at 37°C for 30 minutes. Proteins were precipitated through the addition of Protein Precipitation Solution (Promega) followed by vigorous vortexing. Samples were incubated on ice for 5 minutes and centrifuged at 16,000 x g for 10 minutes at 4°C. The supernatant was added to 600μL isopropanol and inverted until a white mass of DNA was visible. Samples were centrifuged for 5 minutes at 16,000 x g and the resulting pellet was washed with 70% ethanol, dried, and rehydrated with DNA Rehydration Solution (Promega) overnight at 4°C. The genomic DNA (gDNA) was cleaned using the Zymo Genomic DNA Clean and Concentrator kit using the manufacturer specifications for gDNA. The cleaned gDNA was sheared to generate fragments smaller than 600bp and enriched around 300bp. Briefly, 4μg of gDNA was diluted with 10mM Tris-Cl pH = 8.0 to a final volume of 200μL in low adhesion tubes. The gDNA was sonicated using the following specifications: 20% amplitude, 15 seconds on/15 seconds off duty cycle, 6 minute sonication time (total 12 minutes run time). Proper shearing of the gDNA was confirmed by gel electrophoresis. The sheared DNA was cleaned using AMPure XP beads (Agencourt) following manufacturer specifications. 3’-poly-dC tails were added to the sheared gDNA and filtered through Performa DTR gel filtration cartridges (Edge Biosystems) to remove small molecules. The transposon junctions were amplified using a nested PCR protocol using Easy-A cloning enzyme (Agilent) in which the 3’-poly-dC DNA fragments were amplified, (1) using a forward primer that recognizes the 3’-poly-dC tails (PolyG-1^st-1) and a reverse primer that recognizes the Tn5 site (Tn5-1^st-1) and, (2) using a forward primer that recognizes the Tn5 transposon (Tn5-2^nd-1) and the amplified 5’-poly-gC tail to add a library-specific barcode. The resulting PCR reactions were quantified using Qubit (ThermoFisher Scientific) and equivalent concentrations of the libraries were pooled in a final volume of 30μL. The size range of the pooled PCR products were checked by gel electrophoresis and fragments ranging from 250-500bp were isolated from the gel using a Qiagen Gel Extraction kit. The concentration of the gel-extracted, pooled DNA was measured using Qubit. The size of the DNA fragments and molarity was calculated using the Agilent High Sensitivity D1000 ScreenTape System.

Tn-seq analysis

The pooled and prepared libraries were run on a MiSeq sequencer (Illumina) and the resulting reads were trimmed using trimmomatic [94] and mapped to the MB001 genome using bowtie 1.0.0 [95]. The further analysis pipeline was performed as described previously [48]. Transposon insertion profiles were visualized using Artemis software [16]. All NGS data from this study are deposited under BioProject PRJNA1113594.

Flow cytometry

Cells were grown to mid-log (OD600 ~ 0.3–0.8) with aeration in the appropriate media at 30°C. An equivalent to 1mL of an OD₆₀₀ = 0.3 of cells were gently pelleted (2 minutes at 15,000rpm) and resuspended in 60μL of 1X PBS. anti-6x His Alexa Fluor 647 (Abcam) was added at a dilution of 1:100. Cells were stained in the dark for 45–60 minutes. Samples were washed two times in 100μL of 1X PBS and resuspended in a final volume of 3mL 1X PBS for flow cytometry analysis. Samples were analyzed on a BD LSRII flow cytometry machine using the APC optical configuration (637nm laser, 660/20 bandpass filter).

Immunoblot analysis

The OD600 of overnight cultures was measured and an equivalent of 1mL of cells at an OD₆₀₀ = 2 of cells was pelleted. The pellets were resuspended in 50μL of lysis buffer (per sample: 0.5μL benzonase/universal protease, 0.5μL 1M MgSO₄, 1uL of protease inhibitor cocktail, 10μL of 10mg/mL lysozyme, and 8μL of H₂O). Samples were lysed at 37°C for 1 hour, diluted 1:2 in sample buffer containing β-mercaptoethanol (BME), and boiled for 10 minutes. PorH-His immunoblots were performed on 16.5% tris-tricine gels (BioRad) using anti-His (mouse) primary antibody (GenScript) at a dilution of 1:3,000. RsdA cleavage immunoblot analysis was performed on TGX 10% gels (BioRad) using anti-HA (mouse) primary antibody (Abcam) at a dilution of 1:3,000. In all cases, anti-mouse HRP secondary antibody (Rockland Biosciences) was used at a dilution of 1:3,000.

β-galactosidase assays

The OD₆₀₀ of overnight cultures was measured and an equivalent to 1mL at an OD₆₀₀ = 0.75 of cells was pelleted. The pellets were resuspended in 1mL of Z-buffer (0.06M Na₂HPO₄. 7H₂O, 0.04M NaH₂PO₄. H₂O, 0.01M KCl, 0.001M MgSO₄, 0.05M BME, pH = 7.0) and incubated at 30°C for 5 minutes. Lysis was completed by adding 5μL of toluene and 20μL of 10mg/mL lysozyme and incubating at 30°C for 15 minutes. A 200μL volume of 4mg/mL O-nitrophenyl-β-D-galactopyranoside (ONPG) solution made up in Z-buffer was added to begin the reaction and the samples were incubated at 30°C. The reaction was quenched with 500μL of 1M Na₂CO₃. OD420 and OD₄₂₀ values were read on a Biochrom Ultrospec 2100 pro. Miller units were calculated using the following equation: 1 Miller unit = 1,000 x (OD₄₂₀/(time x volume x OD₆₀₀)). All raw data for β-galactosidase assays are included as S5 Table.

6TMR-tre fluorescence assay

Incorporation of 6TMR-tre into the mycomembrane was measured as described previously [68]. Strains were sub-cultured 1:100 in BHIS at 30°C until they reached mid-log (~4 hours). Cells were grown in the presence of a DMSO vehicle control or 0.025mM AU1235, where appropriate. An equivalent to 1mL of cells at an OD₆₀₀ = 0.3 was pelleted and resuspended in either BHIS or BHIS supplemented with a 1:100 dilution of 10mM 6TMR-Tre stock solution and incubated at 25°C in the dark for 30 minutes. Samples were washed twice in 1X PBS and resuspended in a final volume of 1mL 1xPBS. Samples were aliquoted into black-walled, clear bottom 96-well plates (Corning). The OD600 of each sample and 6TMR-Tre incorporation was measured at an excitation of 532nm and emission of 580nm on an Tecan Infinite M Plex. Measurements were made in technical duplicate and biological triplicate. All raw data for the 6TMR-tre assay is detailed in S6 Table.

Fluorescence microscopy

Overnight cultures of Cglu strains were sub-cultured 1:100 in BHI and grown until mid-log (3–4 hours). Cells were concentrated in in 1X PBS and spotted onto 2% M9 agarose pads (. Fluorescent microscopy images were acquired on a Nikon Ti2-E inverted widefield microscope utilizing a motorized stage, a perfect focus system, and a 1.45 NA Plan Apo x100 Ph3 DM objective lens with Cargille Type 37 immersion oil. Images were obtained with Lumencore SpectraX LED Illumination using the mCherry channel with Chroma 49008 filters and an Andor Zyla 4.2 Plus sCMOS camera (65 nm pixel size) with Nikon Elements acquisition software (v5.10). All images were rendered for publication using Fiji and brightness/contrast was normalized between samples to wild-type cells.