Mycobacterium tuberculosis Pst/SenX3-RegX3 Regulates Membrane Vesicle Production Independently of ESX-5 Activity

ABSTRACT Mycobacterium tuberculosis releases membrane vesicles (MV) that modulate host immune responses and aid in iron acquisition, although they may have additional unappreciated functions. MV production appears to be a regulated process, but virR remains the only characterized genetic regulator of vesiculogenesis. Here, we present data supporting a role for the M. tuberculosis Pst/SenX3-RegX3 signal transduction system in regulating MV production. Deletion of pstA1, which encodes a transmembrane component of the phosphate-specific transport (Pst) system, causes constitutive activation of the SenX3-RegX3 two-component system, leading to increased protein secretion via the specialized ESX-5 type VII secretion system. Using proteomic mass spectrometry, we identified several additional proteins hyper-secreted by the ΔpstA1 mutant, including LpqH, an MV-associated lipoprotein. Nanoparticle tracking analysis revealed a 15-fold increase in MV production by the ΔpstA1 mutant. Both hyper-secretion of LpqH and increased MV release required RegX3 but were independent of VirR, suggesting that Pst/SenX3-RegX3 controls MV release by a novel mechanism. Prior proteomic analysis identified ESX-5 substrates associated with MV. We therefore hypothesized that MV release requires ESX-5 activity. We constructed strains that conditionally express eccD5, which encodes the predicted ESX-5 transmembrane channel. Upon EccD5 depletion, we observed reduced secretion of the ESX-5 substrates EsxN and PPE41, but MV release was unaffected. Our data suggest that ESX-5 does not affect vesicle production and imply that further characterization of the Pst/SenX3-RegX3 regulon might reveal novel mechanisms of M. tuberculosis vesicle biogenesis.

transporting membranous material through their complex outer cell walls. In the last decade, however, MV production has been described for many of these organisms, including the mycobacteria (2), although the mechanisms of MV release are not well understood (3).
Mycobacterium tuberculosis, the etiologic agent of tuberculosis in humans, produces MV derived from the inner membrane that modulate the host immune response (2,4). MV contain glycolipids, including lipoarabinomannan (LAM), that inhibit CD4 ϩ T cell activation (4) and lipoproteins, such as LpqH, that act in a TLR2-dependent manner to elicit the production of inflammatory cytokines and chemokines, including interleukin 1␤ (IL-1␤), IL-6, IL-12, tumor necrosis factor alpha (TNF-␣), and CXCL1 (2). LpqH can also regulate CD4 ϩ T cell activation and macrophage major histocompatibility complex (MHC) class II expression (5,6). It is unclear whether these MV-induced responses are beneficial or detrimental to bacterial survival. In mice pretreated with MV via the intratracheal route, these inflammatory responses impaired control of subsequent M. tuberculosis infection (2). However, when MV were administered systemically via subcutaneous injection, they induced a level of protection similar to that produced by the Mycobacterium bovis BCG vaccine (7). Furthermore, an M. tuberculosis mutant that produces more MV was hyper-inflammatory, inducing greater production of TNF-␣ and IL-6 from infected primary human macrophages, and this mutant was attenuated compared to wild-type (WT) M. tuberculosis (8). M. tuberculosis MV-null mutants might be used to conclusively determine the role of either MV-induced inflammation or MV-mediated inhibition of T cell function in pathogenesis, but such mutants have yet to be described.
MV production is an active and regulated process, since MV release can be induced by iron limitation (9), but the mechanisms that control MV biogenesis remain largely uncharacterized. The only genetic element currently known to affect M. tuberculosis MV production is virR (locus tag rv0431); disruption of virR resulted in increased MV release (8). VirR is a cytosolic protein that associates with the inner membrane and was proposed to act as part of a higher-order complex to regulate MV production through an unknown mechanism. Genetic screens in the model Gram-negative organism Escherichia coli revealed that numerous unique mutations cause increased MV release (10,11), suggesting that virR is unlikely to be the only regulator of MV production in M. tuberculosis.
In addition to producing MV, M. tuberculosis interacts with the host via several protein secretion systems. Among these are five specialized type VII secretion systems, collectively referred to as the ESX systems (12). ESX-1, -3, and -5 each play an important role in pathogenesis. ESX-1 mediates the escape of M. tuberculosis from the phagosome and helps activate the inflammasome (13,14). Furthermore, ESX-1 is required to trigger type I interferon responses, which are dependent on the cytosolic DNA sensor cyclic GMP-AMP synthase (cGAS) (14,15). ESX-3 is essential for bacterial growth; it plays a role in mycobactin-mediated iron acquisition (16) but also has an iron-independent role in virulence (17). ESX-5 is important for inducing the inflammatory response and IL-1␤ production, as well as causing caspase-dependent cell death (18). These responses likely aid bacterial survival within the host (18). EccD 5 , the protein predicted to form the ESX-5 transmembrane channel, is required for full pathogenicity in macrophages and severe combined immunodeficiency (SCID) mice (19)(20)(21). We recently reported that ESX-5 activity is regulated by the Pst/SenX3-RegX3 system in response to extracellular inorganic phosphate (P i ) availability (22).
The Pst (phosphate-specific transport) system is a high-affinity, low-velocity, ABCtype transport system. We have previously shown that the Pst system controls gene expression in response to extracellular P i availability through an interaction with the SenX3-RegX3 two-component signal transduction system (23). This system, comprising a membrane-bound sensor histidine kinase (SenX3) and DNA binding response regulator (RegX3), is inhibited by the Pst system when P i is abundant. When P i is limiting, however, inhibition by the Pst system is relieved, leading to activation of SenX3-RegX3. Deletion of pstA1, which encodes a transmembrane component of the Pst system, causes constitutive activation of SenX3-RegX3 regardless of P i availability (23). We demonstrated that in addition to controlling expression of genes involved in the P i scavenging response, RegX3 directly binds and regulates transcription of esx-5 genes (22,23). The ΔpstA1 mutant exhibits a RegX3-dependent increase in esx-5 transcription and hyper-secretion of the ESX-5 substrates PPE41 and EsxN (22).
Here, we describe that in addition to affecting protein secretion via ESX-5, the ΔpstA1 mutant hyper-secretes the lipoproteins LpqH and PstS1, which are released from M. tuberculosis in association with MV (2,8,24). The ESX-5 substrates EsxN and PPE41 have also previously been detected in association with MV by proteomic approaches (2,24). We therefore hypothesized that the Pst/SenX3-RegX3 system regulates MV production and that MV release depends on the ESX-5 system. We show that while the ΔpstA1 mutant does hyper-secrete MV in a RegX3-dependent manner, MV release occurs independently of ESX-5 activity. Furthermore, our data indicate that MV hyper-secretion from the ΔpstA1 mutant is independent of VirR. Our data suggest that genes under the control of RegX3 impact MV release from ΔpstA1 bacteria by a novel mechanism.

RESULTS
The ⌬pstA1 mutant hyper-secretes the lipoproteins LpqH and PstS1. We previously demonstrated that ESX-5 secretion system activity is regulated in response to P i availability by the Pst/SenX3-RegX3 signal transduction system (22). A ΔpstA1 mutant, in which the RegX3 response regulator is constitutively activated, hyper-secretes the ESX-5 substrates EsxN and PPE41 under standard P i -replete culture conditions (22). We observed that the ΔpstA1 mutant has a secreted protein profile distinct from that of the wild-type (WT) control (Fig. 1A). To preliminarily characterize these differences in protein secretion, we selected four protein bands at molecular weights of approximately 75, 50, 37, and 25 kDa that appeared to be hyper-secreted by the ΔpstA1 mutant and subjected them to proteomic mass spectrometry (MS) analysis. Multiple proteins were identified in each band; the complete data are provided in Table S1 in the supplemental material.
We obtained antibodies against several proteins identified by mass spectrometry (LpqH, PstS1, and KatG) and tested secretion of these proteins by Western blotting. We detected increased secretion of the lipoproteins LpqH and PstS1 by the ΔpstA1 mutant compared to their levels in the WT control, though there was no change in the abundance of these proteins in the cellular fraction (Fig. 1B). As in our previous report, we observed increased secretion of the ESX-5 substrate EsxN from the ΔpstA1 mutant compared to its level in the WT control. EsxN was undetectable in the secreted WT fraction when 5 g of protein was loaded, but we previously reported that EsxN secretion is induced at least 8-fold in the ΔpstA1 mutant (22). The catalase-peroxidase KatG was detected only in the cellular fraction of both strains, and as such, it was excluded from further analyses (Fig. 1B). GroEL2 served as a loading control for cell-associated protein and indicated that cellular lysis did not contaminate the secreted protein fraction. ModD, a protein secreted by the general Sec system, served as a control for equivalent loading of secreted proteins; we detected ModD as a doublet, consistent with its glycosylation (25). These data corroborated the mass spectrometry identification of LpqH and PstS1 as being hyper-secreted by the ΔpstA1 mutant.
The ⌬pstA1 mutant overproduces membrane vesicles containing the lipoproteins LpqH and PstS1. The LpqH and PstS1 lipoproteins were previously described to be associated with M. tuberculosis MV (2,8,24). Based on their lipid composition, mycobacterial MV are derived from the inner membrane and contain lipoproteins and other protein cargo (2,24). Since both LpqH and PstS1 were hyper-secreted by the ΔpstA1 mutant, we sought to determine whether these proteins were MV associated. We isolated MV from culture filtrates by ultracentrifugation and performed Western blotting experiments to analyze the distribution of LpqH and PstS1. Extracellular LpqH and PstS1 were localized exclusively to the MV fraction, and both proteins were more abundant in MV released from the ΔpstA1 mutant than in the WT control ( Fig. 2A). Complementation of the ΔpstA1 mutation restored the MV-associated release of LpqH and PstS1 to levels comparable to those in the WT control.
To quantify MV release, we conducted nanoparticle-tracking analysis (NTA) on culture supernatants collected from WT, ΔpstA1, and ΔpstA1 pMVpstA1 bacteria using a NanoSight instrument. We observed a statistically significant 15-fold increase in MV  release from ΔpstA1 bacteria compared to that in WT bacteria but no difference in MV release between the WT and ΔpstA1 pMVpstA1 strains (Fig. 2B). The accompanying size analysis of MV isolated from WT and ΔpstA1 bacteria revealed that both strains produced particles that fell within the 40-to 250-nm-diameter range previously observed for M. tuberculosis MV (2) ( Fig. 2C and D). To further confirm these results, purified MV from these strains were imaged via transmission electron microscopy (TEM). We observed vesicular structures with the predicted size of MV from both WT and ΔpstA1 mutant bacteria ( Fig. 2E and F). Taken together, these data support the finding that the ΔpstA1 mutant releases an increased number of MV, and these vesicles are enriched for both the LpqH and PstS1 lipoproteins.
Overproduction of membrane vesicles by the ⌬pstA1 mutant requires RegX3. Many of the phenotypes that we have previously reported for the ΔpstA1 mutant are dependent on RegX3 (22,23,26,27). To determine whether constitutive activation of RegX3 also contributes to overproduction of MV by the ΔpstA1 mutant, we performed additional Western blotting and NTA on the ΔpstA1 ΔregX3 and ΔpstA1 ΔregX3 pN-DregX3 strains. Hyper-secretion of the ESX-5 substrates EsxN and PPE41 was abolished in the ΔpstA1 ΔregX3 mutant (Fig. 3A), as previously described (22). EsxN and PPE41 were detected only in the secreted protein fraction despite previous reports from proteomic analyses that these proteins are MV associated (2,24). This is likely due to the lower sensitivity of Western blots than of proteomic mass spectrometry. Both LpqH and PstS1 were detected in the vesicle fraction from the ΔpstA1 ΔregX3 mutant but at reduced levels compared to those in the ΔpstA1 strain (Fig. 3A). In all cases, complementation of the ΔregX3 deletion with pNDregX3 restored hyper-secretion (Fig. 3A).
We again used NTA to quantify MV release from the regX3 deletion mutant and complemented strains. Size analysis revealed that all strains produced MV of the appropriate diameter, with the majority clustering between 50 and 100 nm (Fig. S1)  The results were normalized to numbers of CFU per milliliter determined from a control culture grown in Sauton's medium with Tween 80 and plated on 7H10 medium. Data are means Ϯ standard deviations for three independent cultures. **, P Ͻ 0.01; ****, P Ͻ 0.0001. n.s., the difference was not significant. (C) The transcript abundance of lpqH, pstS1, and virR relative to that of sigA were determined by quantitative RT-PCR for the WT and ΔpstA1 strains grown to mid-logarithmic phase in 7H9 complete medium. Results are the means Ϯ standard deviations for three independent experiments.

M. tuberculosis Membrane Vesicle Regulation
ΔregX3 strain, NTA revealed that vesicle release from the ΔpstA1 ΔregX3 mutant returned to WT levels ( Fig. 3B). When the regX3 deletion was complemented, we observed a significant increase in MV release of more than 10-fold compared to MV release in the ΔpstA1 strain. We hypothesize that this increase is due to overexpression of regX3 from the complementation vector, as previously reported (23). We observed no significant difference in abundance of the lpqH or pstS1 transcripts between the ΔpstA1 mutant and WT bacteria (Fig. 3C), consistent with no apparent change in LpqH or PstS1 production in the cellular protein fraction (Fig. 3A). These results demonstrate that increased release of LpqH and PstS1 associated with MV is not due to changes in their expression. Taken together, these results indicate that some other factor, or factors, regulated by RegX3 influence MV production. Increased vesicle production by ⌬pstA1 bacteria is not dependent on VirR. virR was previously implicated as a regulator of MV production; transposon disruption of virR led to increased MV biogenesis (8). We analyzed the transcription of virR to determine whether changes in virR expression contribute to increased MV release. While the difference was not statistically significant, we observed a 1.7-fold decrease in the virR transcript level in the ΔpstA1 mutant compared to that in the WT control ( Fig.  3C). We investigated whether decreased virR transcript abundance could account for increased MV release from the ΔpstA1 strain. Western blotting revealed similar levels of VirR production in the cellular fractions of WT and ΔpstA1 bacteria, suggesting that the modest differences in virR transcript abundance do not affect VirR protein production (Fig. 4A). VirR associates with MV and is more abundant in MV from the ΔpstA1 strain, consistent with enhanced MV release (Fig. 4A).
We took two approaches to determine whether MV production by the ΔpstA1 mutant depends on virR. We overexpressed FLAG-tagged VirR (pvirR) (8) to restore virR expression. Western blots of the WT pvirR and ΔpstA1 pvirR strains revealed a dramatic overproduction of VirR (Fig. 4A). We also constructed a deletion in virR, which we predicted would not affect MV production by the ΔpstA1 mutant if both PstA1 and VirR act in the same MV biogenesis pathway. To avoid polarity on adjacent essential genes, we deleted amino acid residues 85 to 103 of VirR, which correspond to the core globular structure of the protein (8). Western blotting confirmed loss of VirR production in both the ΔvirR and the ΔpstA1 ΔvirR strains (Fig. 4A).
The overexpression of virR did not dramatically alter LpqH or PstS1 abundance in the MV fraction. However, VirR was much more abundant in MV from both the WT pvirR and the ΔpstA1 pvirR strains (Fig. 4A), indicating that VirR may become overrepresented in MV when it is overproduced. Both VirR overexpression strains had slightly increased MV production compared to that of the corresponding parental controls, but this was statistically significant only for the ΔpstA1 pvirR strain (Fig. 4B). Notably, ΔpstA1 pvirR bacteria still produced more MV than WT bacteria, demonstrating that virR overexpression does not rescue the MV hyper-secretion phenotype of the ΔpstA1 strain (Fig. 4B).
The ΔvirR strain had slightly more LpqH and PstS1 in the MV fraction than the WT control (Fig. 4A), suggesting increased MV production as previously reported (8). Both LpqH and PstS1 were also more abundant in the MV fraction from the ΔpstA1 ΔvirR mutant than in either single mutant or WT bacteria (Fig. 4A). NTA confirmed that the ΔvirR strain produced 3-fold-more MV than the WT strain, although this increase was not statistically significant (Fig. 4B). NTA also confirmed that the ΔpstA1 ΔvirR strain produced 4-fold-more MV than the ΔpstA1 mutant and 63-fold-more MV than the ΔvirR mutant, both significant increases (Fig. 4B). Taken together, these results indicate that overproduction of MV by the ΔpstA1 strain is independent of VirR. Furthermore, the mechanisms driving increased MV production in the ΔpstA1 and ΔvirR strains appear to synergize with each other, as a strain harboring deletions of both of these genes produced significantly more MV than either single mutant.
Construction of conditional eccD 5 depletion mutants. We previously demonstrated that the ΔpstA1 mutant exhibits increased ESX-5 activity that is dependent on the RegX3 response regulator (22). Furthermore, others have reported detection of ESX-5 substrates within MV (2,24). We therefore sought to determine whether increased MV production by the ΔpstA1 mutant was due to increased ESX-5 activity. We attempted to delete eccD 5 , which encodes a core component of the ESX-5 secretion system that is predicted to form an inner membrane pore through which substrates are secreted (21,28). We constructed a ΔeccD 5 allelic-exchange vector and introduced it into both the WT Erdman and ΔpstA1 mutant strain backgrounds. After obtaining cointegrates with the plasmid integrated at the eccD 5 locus, we screened sucroseresistant colonies for the ΔeccD 5 deletion by PCR. All the colonies screened (130 with the WT background, 169 with the ΔpstA1 background) retained the WT eccD 5 allele, suggesting that eccD 5 is essential for the replication of M. tuberculosis in the standard Middlebrook 7H9 and 7H10 media that we used.
To create strains for conditional expression of eccD 5 , we first introduced a plasmid (pTIC-eccD 5 ) in which eccD 5 transcription is under the control of a tetracycline-inducible (Tet-ON) promoter. We then attempted to delete the chromosomal copy of eccD 5 in the WT pTIC-eccD 5 and ΔpstA1 pTIC-eccD 5 strains. All strain construction steps were conducted in the presence of anhydrotetracycline (ATc) to stimulate eccD 5 transcription from pTIC-eccD 5 . We screened 8 and 16 sucrose-resistant, hygromycin (Hyg)-sensitive isolates in the WT and ΔpstA1 mutant backgrounds, respectively. Among these isolates, we identified 2 ΔeccD 5 pTIC-eccD 5 mutants and 2 ΔpstA1 ΔeccD 5 pTIC-eccD 5 mutants that we herein refer to as eccD 5 Tet-ON and ΔpstA1 eccD 5 Tet-ON, respectively. These data support the contention that eccD 5 is essential for the replication of M. tuberculosis under standard in vitro culture conditions.
We analyzed the growth of the eccD 5 Tet-ON and ΔpstA1 eccD 5 Tet-ON strains compared to that of the parental controls in liquid medium and on agar plates. Both eccD 5 Tet-ON strains had a growth defect in liquid medium compared to the growth of the respective parental control even in the presence of ATc to induce eccD 5 transcription ( Fig. S2A and B). The eccD 5 Tet-ON strains also produced smaller colonies on agar plates regardless of ATc treatment (data not shown). Although eccD 5 transcription was induced in the eccD 5 Tet-ON strains in the presence of ATc, there was still detectable transcript in no-ATc controls (Fig. S2C), and we observed no difference in EccD 5 protein abundance between treatment conditions (Fig. S2D). Additionally, PPE41 was secreted even in the absence of ATc (Fig. S2D). Overall, our data suggest that leaky expression from the Tet-inducible promoter in the absence of ATc is sufficient to support the normal activity of the ESX-5 secretion system in the eccD 5 Tet-ON strains.
We therefore turned to construction of strains harboring tetracycline-repressible eccD 5 (eccD 5 Tet-OFF). We cloned eccD 5 under the control of the tetO-4C5G operator (29) on a vector that integrates at the same site as pTIC10a. This pGMCH eccD 5 Tet-OFF vector was introduced into the eccD 5 Tet-ON strains, and integrants in which pGMCH eccD 5 Tet-OFF replaced pTIC-eccD 5 by a plasmid swap were selected on 7H10 containing hygromycin. We recovered over 15-fold-more colonies from electroporation with pGMCH eccD 5 Tet-OFF than with an empty hygromycin resistance (Hyg r ) plasmid control. The majority of pGMCH eccD 5 Tet-OFF transformants were hygromycin resistant and kanamycin (Kan) sensitive, while isolates transformed with the empty Hyg r plasmid were resistant to both hygromycin and kanamycin, indicating that they retained the pTIC-eccD 5 plasmid. These data provide further evidence that eccD 5 is essential for M. tuberculosis replication in vitro. Strains in which the pGMCH eccD 5 Tet-OFF plasmid replaced pTIC-eccD 5 were confirmed by PCR and designated eccD 5 Tet-OFF and ΔpstA1 eccD 5 Tet-OFF. We also attempted construction of dual-control strains in which both eccD 5 transcriptional repression and EccD 5 protein degradation would occur upon addition of ATc due to the addition of a DASϩ4 degradation tag on EccD 5 (29). However, attempts to introduce the plasmid encoding DAS-tagged EccD 5 by plasmid swap produced Ͻ10 colonies, all of which were resistant to both hygromycin and kanamycin, suggesting that EccD 5 -DASϩ4 is nonfunctional.
The eccD 5 Tet-OFF strains in both the WT and ΔpstA1 backgrounds had statistically significant growth defects (P Ͻ 0.05) in liquid medium containing ATc beginning at day 3 compared to the growth of the no-drug controls ( Fig. 5A and C). These defects were maintained throughout the incubation period until day 15, when they dropped just below the level of statistical significance. Similarly, we recovered fewer viable CFU from cultures of both eccD 5 Tet-OFF strains at days 3 to 12 when ATc was added ( Fig. 5B and  D). These differences did not quite achieve statistical significance for the eccD 5 Tet-OFF strain (Fig. 5B). Addition of ATc significantly reduced the growth of the ΔpstA1 eccD 5 Tet-OFF strain at days 6 and 9, but by day 15, we observed a significant increase in the number of viable CFU recovered from these cultures (Fig. 5D). Both eccD 5 Tet-OFF strains produced smaller colonies when grown on 7H10 medium with ATc (Fig. S3). Importantly, the addition of ATc did not affect the growth kinetics or colony morphology of the WT or ΔpstA1 parent strain (Fig. S3 and S4). Quantitative reverse transcription-PCR (qRT-PCR) confirmed that addition of ATc to the eccD 5 Tet-OFF and ΔpstA1 eccD 5 Tet-OFF strains resulted in significant 2.7-fold and 3.6-fold repression of eccD 5 , respectively (Fig. 5E). This repression did not dramatically alter the expression of other genes within the ESX-5 locus, such as espG 5 , or the expression of lpqH itself (Fig. 5E). To confirm that reduced eccD 5 transcription altered EccD 5 protein production and ESX-5 secretion system activity, we performed Western blotting. EccD 5 production was reduced 2.4-fold and 3.5-fold in the eccD 5 Tet-OFF and ΔpstA1 eccD 5 Tet-OFF strains, respectively, in the presence of ATc (Fig. 6A). These results suggest that eccD 5 expression is not fully repressed by ATc in the Tet-OFF strains and that sufficient EccD 5 remains to support growth, even when the cultures contain ATc.
We observed typical hyper-secretion of both EsxN and PPE41 from the ΔpstA1 eccD 5 Tet-OFF strain in the absence of ATc, but secretion of these proteins was abrogated when ATc was added (Fig. 6A). Treatment of the ΔpstA1 eccD 5 Tet-OFF strain with ATc also caused an increase in PPE41 associated with the cellular fraction (Fig. 6A), indicating that PPE41 becomes trapped within the cell when ESX-5 secretion is prevented. We were able to faintly detect PPE41 only in the MV fraction of the untreated ΔpstA1 eccD 5 Tet-OFF strain, further supporting our finding that the majority of secreted EsxN and PPE41 is not MV associated. Other secreted proteins, such as the antigen 85 (Ag85) complex and ModD, were not affected by the repression of eccD 5 (Fig. 6A). Taken together, these analyses confirm that the Tet-OFF strains function as expected and demonstrate that repression of eccD 5 results in reduced secretion through ESX-5.
The ESX-5 secretion system is not required for membrane vesicle release. We predicted that repression of eccD 5 transcription in the ΔpstA1 eccD 5 Tet-OFF strain would prevent MV hyper-secretion. MV were isolated from eccD 5 Tet-OFF and ΔpstA1 eccD 5 Tet-OFF strains as described above but with the addition of ATc to the indicated cultures during the final growth phase. Western blots for LpqH and PstS1 demonstrated that MV were hyper-secreted from the ΔpstA1 eccD 5 Tet-OFF strain in similar amounts whether or not ATc was added to the cultures (Fig. 6A). NTA confirmed that there was no significant difference in the numbers of MV released between samples with and without ATc treatment or between the Tet-OFF strains and the corresponding parental control (Fig. 6B). These findings refute our initial hypothesis

M. tuberculosis Membrane Vesicle Regulation
® and instead indicate that ESX-5 secretion system activity is not required for MV release from M. tuberculosis.

DISCUSSION
M. tuberculosis produces MV, but the molecular mechanisms and regulatory processes controlling their release are poorly characterized. We demonstrate that the ΔpstA1 mutant hyper-secretes MV containing the lipoproteins LpqH and PstS1 in a RegX3-dependent manner. Because the ΔpstA1 mutant also exhibits a RegX3dependent increase in the activity of the ESX-5 secretion system and ESX-5 substrates had previously been identified within MV, we initially hypothesized that MV release and ESX-5 activity were connected. However, the overproduction of MV from the ΔpstA1 eccD 5 Tet-OFF strain was not affected by repressing ESX-5 secretion. Our data demonstrate that the Pst/SenX3-RegX3 system regulates the production of MV independently of ESX-5 activity.
Our data suggest that one or more factors regulated by RegX3 causes increased MV release from the ΔpstA1 mutant. Previous work implicated virR as a regulator of MV production (8), but we observed only modest, insignificant changes in virR expression in ΔpstA1 bacteria compared to its expression in the WT. Furthermore, deletion of virR in the ΔpstA1 background resulted in an even greater increase in MV production, suggesting that these genes promote MV production through separate mechanisms that are capable of synergizing with each other. Overexpression of virR in the ΔpstA1 background also resulted in a slight increase in MV production, providing further support for this hypothesis and suggesting that dramatic overproduction of VirR may also influence MV production. Together, these findings indicate that MV production is mediated by a novel mechanism in ΔpstA1 bacteria. Over 60 genes are dysregulated in the ΔpstA1 mutant in a RegX3-dependent manner (23), and any one or more of these genes may be involved in MV release. Our future studies will seek to identify the basis for increased MV release by the ΔpstA1 mutant to provide mechanistic insight into MV biogenesis.
By carefully examining the RegX3 regulon, we developed several hypotheses to explain increased MV production from the ΔpstA1 mutant. Several highly upregulated genes in the ΔpstA1 mutant are involved, or predicted to be involved, in the production of lipoproteins and glycolipids. One of the most upregulated genes, lppF, was overexpressed 20-fold in the ΔpstA1 mutant and is predicted to encode a lipoprotein (23). While LppF has not been detected in association with MV, several other lipoproteins, including LpqH, LppX, LprA, LprG, and LprF, are highly abundant (24). Additionally, rv0557 (mgtA) transcript levels are almost 14-fold higher in ΔpstA1 bacteria (23). The Corynebacterium glutamicum ortholog of this gene is required for production of a novel glycolipid, 1,2-di-O-C 16 /C 18:1 -(␣-D-mannopyranosyl)-(1¡4)-(␣-D-glucopyranosyluronic acid)-(1¡3)-glycerol (ManGlcAGroAc 2 ), that contributes to the lipomannan (LM) pool (30). In M. tuberculosis, inactivation of rv0557 results in reduced cell wall LAM and LM (31). As both LAM and LM are incorporated in MV (2,4), it is possible that changes in the production of these lipoglycans influence MV release. Two genes encoding putative acyltransferases (rv3027c and rv3026c) are also upregulated in the ΔpstA1 mutant 10-and 7-fold, respectively (23). These genes may contribute to increased lipid production, resulting in more MV biogenesis. Finally, rv1491c is also upregulated 3-fold in the ΔpstA1 mutant (23) and encodes a DedA family protein. In E. coli, deletion of genes encoding DedA family proteins causes cell division defects and altered membrane phospholipid composition (32). We hypothesize that Rv1491c has similar roles in M. tuberculosis membrane biogenesis and thus influences MV production. Additionally, Rv1491c is encoded near Rv1488, a putative membrane protein that interacts with VirR (8). It is possible that both Rv1491c and Rv1488 play roles in MV biogenesis. We intend to explore these hypotheses as part of our future studies.
We initially hypothesized that ESX-5 would be involved in MV release based in part on prior detection of the ESX-5 substrates PPE41 and EsxN in MV by mass spectrometry (2, 24), yet we were generally unable to detect these proteins in MV by Western blotting. We observed the majority of PPE41 and EsxN only in the secreted protein fraction and confirmed that their secretion requires EccD 5 . Our results indicate that PPE41 and EsxN are secreted primarily as soluble protein via ESX-5.
Our data also suggest that the ESX-5 core component EccD 5 is essential for M. tuberculosis viability. We were unable to delete eccD 5 unless a complementing copy of the gene was provided in trans, and we also could not efficiently replace the complementing copy of eccD 5 with an empty vector. Similarly, both eccB 5 and eccC 5 , which encode two additional core components of ESX-5, could not be deleted without a complementing copy of the appropriate gene provided in trans, suggesting that they are also essential for M. tuberculosis viability (33). Conserved ESX-5 components eccC 5 and mycP 5 also appear to be essential for the growth of the mycobacterial species M. marinum and M. bovis BCG (34). Others have reported deletion (20) or disruption (35) of eccD 5 , in contrast to our data. It is possible that these mutants were still viable due to compensatory mutations. Indeed, eccC 5 could be deleted in M. marinum mutants deficient for production of the virulence lipid phthiocerol dimycocerosate (PDIM) (34), and spontaneous loss of PDIM production by M. tuberculosis has previously been reported (36). Overall, our data support a growing body of literature suggesting that ESX-5 secretion in general and the EccD 5 core component in particular are essential for mycobacterial growth in vitro.
To circumvent the essentiality of EccD 5 , we generated conditional eccD 5 Tet-OFF strains in both the WT and ΔpstA1 backgrounds. Addition of ATc to these strains repressed eccD 5 transcription but caused only moderate growth defects either in liquid medium or on solid agar. EccD 5 production was also decreased in both eccD 5 Tet-OFF strains when ATc was added, although protein was still detectable by Western blotting. We hypothesize that the residual EccD 5 produced by these strains was sufficient to support some growth. Nevertheless, depletion of EccD 5 led to decreased ESX-5 activity, as indicated by reduced secretion of PPE41 and EsxN. Collectively, these data indicate that native EccD 5 is necessary for M. tuberculosis viability and that lower eccD 5 expression results in decreased growth and reduced protein secretion via ESX-5. Future work will use these eccD 5 Tet-OFF strains to identify additional secreted substrates of the ESX-5 system and to investigate the role that ESX-5 secretion plays in pathogenesis.

MATERIALS AND METHODS
Strains and culture conditions. M. tuberculosis Erdman and ΔpstA1, ΔpstA1 pMVpstA1, ΔpstA1 ΔregX3, and ΔpstA1 ΔregX3 pND-regX3 strains were previously described (23). Construction of the ΔvirR, ΔpstA1 ΔvirR, virR overexpression, eccD 5 Tet-OFF, and ΔpstA1 eccD 5 Tet-OFF strains are described below. Bacteria were routinely cultured in Middlebrook 7H9 liquid medium (Difco) supplemented with 10% albumin-dextrose-saline (ADS), 0.5% glycerol, and 0.1% Tween 80 or on Middlebrook 7H10 agar medium (Difco) supplemented with oleic acid-albumin-dextrose-catalase (OADC; BD Biosciences) and 0.5% glycerol. Strains containing the pvirR plasmid were grown in the presence of 50 g/ml hygromycin B (Sigma). Frozen stocks were prepared by growing cultures to mid-logarithmic phase, adding glycerol to a 15% final concentration, and aliquoting for storage at Ϫ80°C. Proteomic analysis. M. tuberculosis WT and ΔpstA1 bacteria were grown in Sauton's medium without Tween 80, and the secreted protein fractions were isolated as previously described (22). Total secreted proteins (10 g) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The gel was fixed for 30 min in 40% ethanol-10% acetic acid and then stained with Imperial protein stain (Thermo Scientific) and destained in water. Bands of interest from three SDS-PAGE gel lanes of the ΔpstA1 mutant corresponding to 75, 50, 37, and 25 kDa were excised from the gel, pooled, cut intõ 2-by 2-mm cubes, and processed by in-gel trypsin digestion. Gel slices were washed twice for 15 min at room temperature in wash buffer (1:1 acetonitrile-100 mM NH 4 HCO 3 ) and then treated with 100% acetonitrile until the pieces shrank and turned white and semiopaque. Proteins were reduced with 10 mM dithiothreitol (DTT) in 10 mM NH 4 HCO 3 for 1 h at 56°C and then alkylated by treatment with 55 mM iodoacetamide in 100 mM NH 4 HCO 3 for 30 min at room temperature in the dark. Gel slices were washed twice with wash buffer and then washed briefly with 100% acetonitrile before rehydration in digestion buffer (50 mM NH 4 HCO 3 , 5 mM CaCl 2 , 12.5 ng/l trypsin) at 4°C for 15 min, followed by overnight incubation in 50 mM NH 4 HCO 3 , 5 mM CaCl 2 at 37°C. Peptides were extracted from the gel slices with 50% acetonitrile, 0.3% formic acid for 15 min and then with 80% acetonitrile, 0.3% formic acid for 15 min. Peptides were incubated at Ϫ80°C for 30 min, dried in a speed vac, and stored at Ϫ80°C prior to analysis by liquid chromatography-tandem mass spectrometry (LC/MS-MS) on an LTQ Orbitrap Velos mass spectrometer (Thermo, Fisher). The MS-MS data were compared with sequences in an M. tuberculosis proteomic database using Scaffold V4 software.
Purification of His 6 -tagged EccD 5 1-131. The portion of eccD 5 encoding a predicted soluble cytoplasmic domain from residues 1 to 131 (EccD 5 1-131) was amplified from M. tuberculosis Erdman genomic DNA by PCR with primers His-eccD5_F1 and His-eccD5_R1 and cloned into plasmid pET28bϩ between the NdeI and HindIII restriction enzyme sites to generate pET28-His 6 EccD 5 , encoding EccD 5 1-131 with an N-terminal His 6 tag. The pET28-His 6 EccD 5 plasmid was introduced into E. coli BL21(DE3), and the protein was purified by Ni 2ϩ -nitrilotriacetic acid (NTA) affinity chromatography (Qiagen) as previously described for PPE41-His 6 (37,38). Briefly, purified protein was bound to the column under native conditions in 20 mM HEPES buffer, 300 mM NaCl, pH 7.8, and eluted in 20 mM HEPES buffer, 500 mM NaCl, pH 7.8, containing 50 to 150 mM imidazole. Purified His 6 EccD 5 was concentrated through a 5-kDa-cutoff Amicon Ultra centrifugal filtration unit (Millipore), and then contaminant proteins were removed by fast-protein liquid chromatography (FPLC) using a BioLogic DuoFlow apparatus (Bio-Rad). Protein was bound to an Enrich Sec 650 (Bio-Rad) column and eluted in phosphate-buffered saline (PBS) using an isocratic flow at a rate of 500 l/min. Protein which eluted at 0.02 absorbance unit was collected and pooled.
Preparation of membrane vesicles. Thirty-milliliter M. tuberculosis cultures were grown in Sauton's medium as previously described for analysis of secreted proteins (22). MV were isolated from culture supernatants as previously described, with minor modifications (2). Briefly, cells were pelleted by centrifugation (4,700 ϫ g, 15 min, 4°C), and supernatants were sequentially filtered through 0.45-m and 0.22-m Millex syringe filters (Millipore). The supernatants were centrifuged (4,000 ϫ g, 15 min, 4°C) to remove remaining cellular debris and then concentrated to approximately 1 ml using 100-kDa-cutoff Amicon Ultra centrifugal filtration devices (Millipore). The filtrates were further processed to concentrate secreted proteins using a 5-kDa-cutoff Amicon Ultra centrifugal filtration device, as previously described (22). The 100-kDa filter retentates were then centrifuged at 100,000 ϫ g (75 min, 4°C) to pellet MV. The MV pellets were resuspended in 500 l PBS containing Complete EDTA-free protease inhibitors (Roche).
Western blotting. Cellular and secreted proteins were isolated from M. tuberculosis cultures grown in Sauton's medium as previously described, with minor modifications (22). Briefly, bacteria were pelleted and culture supernatants were filter sterilized before MV were purified and secreted proteins were concentrated as described above. Pellets were resuspended in PBS containing Complete EDTA-free protease inhibitors (Roche) and subjected to bead beating to lyse the cells before they were filtered to remove intact cells. Cellular and secreted proteins were quantified via a bicinchoninic acid (BCA) assay (Pierce BCA protein assay kit; ThermoFisher). Five micrograms of cellular and secreted proteins was separated by SDS-PAGE and transferred to 0.2-m nitrocellulose membranes (Bio-Rad) as previously described (22). The protein concentration in MV preparations was below the limit of detection of the BCA assay, so 20 l of MV samples, corresponding to 1.2 ml of culture, was separated by SDS-PAGE. Membranes were blocked in PBS with 0.1% Tween 20 (PBST) containing 5% nonfat milk powder at room temperature for 1 to 2 h. Membranes were washed in PBST and then probed overnight at 4°C with primary antisera diluted in PBST containing 2.5% nonfat milk powder. Primary antisera were used at the following dilutions: mouse anti-KatG, 1:1,000; mouse anti-LpqH, 1:500; mouse anti-PstS1, 1:1,000; mouse anti-GroEL2, 1:1,000; rabbit anti-ModD, 1:25,000; rabbit anti-PPE41, 1:1,000; rabbit anti-EsxN 1:1,000; rabbit anti-EccD 5 , 1:1,000; and rabbit anti-VirR, 1:10,000. Membranes were again washed in PBST before incubation for 1 to 2 h at room temperature with the appropriate secondary antibody (either goat anti-rabbit or rabbit anti-mouse antibody conjugated to horseradish peroxidase; Sigma) diluted 1:30,000 (1:5,000 when used to detect VirR) in PBST containing 2.5% nonfat milk powder. Membranes were washed again in PBST, and bands were detected using SuperSignal West Pico chemiluminescent substrate (Thermo Scientific). Blots were imaged using an Odyssey Fc imaging system (LI-COR), and protein abundance was analyzed using Image Studio software (LI-COR).
Vesicle quantification. Filter-sterilized culture supernatants from M. tuberculosis strains grown in Sauton's medium without Tween 80 were analyzed with a NanoSight instrument to determine vesicle numbers and sizes. Culture filtrates were diluted in a 1-ml final volume of 1ϫ PBS (Corning) to a concentration acceptable for analysis by the NanoSight NS300 (Malvern Instruments, United Kingdom). Triplicate videos of each sample were taken at 24.5°C in light scatter mode using the equipped 532-nm green laser and a syringe pump. Particle displacement was detected with a camera level of 14, and analyses were performed using NanoSight 3.0 or 3.1 software and a threshold of between 3 and 5. Triplicate video statistics were averaged for each sample. A control culture of each strain grown in Sauton's medium with Tween 80 was serially diluted and plated on 7H10 medium in duplicate. Colonies were counted after 3 weeks of incubation at 37°C to determine numbers of CFU per milliliter, and the number of particles per milliliter was normalized to this value to determine the number of vesicles per CFU.
EM of purified vesicles. All reagents were electron microscopy (EM) grade and were purchased from Electron Microscopy Supply Co. Transmission electron microscopy (TEM) was used to visualize MV purified from 600-ml cultures grown as described above. Formvar-coated copper grids were floated on 20-l drops of purified MV solutions for 20 min. The grids were washed in water and then floated on 1% uranyl acetate for 30 s. The grids were washed again before being imaged on a FEI Tecnai Spirit Bio-Twin.
Construction of virR deletion and overexpression strains. Sequences of all primers used for cloning and strain construction are provided in Table S2 in the supplemental material. A vector for the deletion of base pairs 252 to 306 of virR was constructed in the allelic-exchange plasmid pJG1100, which contains the aph (kanamycin [Kan] resistance), hyg (hygromycin [Hyg] resistance), and sacB (sucrose sensitivity) markers (36). Regions of the M. tuberculosis genome~900 bp 5= and 3= of the deletion site were amplified by PCR with primers virR_F2/virR_R6 and virR_F6/virR_R3. The virR_R6 primer to amplify the 5= region of virR was designed with an AvrII restriction site in frame with the virR start codon, and the virR_F6 primer used to amplify the 3= region of virR was designed with an AvrII restriction site in frame with the stop codon. The resulting construct encodes a copy of virR lacking 54 bp. The PCR products were cloned into PCR2.1-TOPO (Invitrogen) and sequenced. The 5= and 3= homology regions were subsequently removed from the pCR2.1 vector by digestion with PacI/AvrII or AvrII/AscI, respectively, and cloned in pJG1100 digested with PacI/AscI. The resulting pJG-ΔvirR vector was confirmed by sequencing. Plasmids for the overexpression of FLAG-tagged virR (pvirR) were previously described (8) and generously provided by Carl Nathan.
The pJG-ΔvirR and pvirR vectors were introduced into WT and ΔpstA1 bacteria by electroporation as previously described (23). Transformants containing pvirR were selected by plating them on complete 7H10 containing 50 g/ml Hyg. The presence of pvirR was confirmed by PCR with the primer pair pDE43_F/pDE43_R. Transformants containing pJG-ΔvirR were selected by plating them on complete 7H10 containing 50 g/ml Hyg and 15 g/ml Kan. Integration of pJG-ΔvirR at the virR locus was confirmed using primers virR_F4/PJGR and PJGF/virR_R4. Transformants that had integrated the plasmid were grown in complete 7H9 medium before serial dilution and plating on complete 7H10 medium containing 2% sucrose to counterselect the pJG-ΔvirR plasmid. Sucrose-resistant isolates were screened by PCR with the primer pair virR_F4/virR_R4 to ensure loss of the pJG-ΔvirR plasmid, and the 54-bp deletion was confirmed by sequencing.
Cloning of eccD 5 deletion and conditional expression vectors. A vector for deletion of eccD 5 was constructed in the allelic-exchange plasmid pJG1100. Regions of the M. tuberculosis genome~800 bp 5= and 3= of eccD 5 were amplified by PCR with primer pairs eccD5F1/eccD5R1 and eccD5F2/eccD5R2. The eccD5R1 primer to amplify the 5= region was designed with an AvrII restriction site in frame with the eccD 5 translation start codon. The eccD5F2 primer to amplify the 3= region was designed with an AvrII restriction site in frame with the eccD 5 stop codon. To avoid polarity of the final construct upon expression of mycP 5 , which overlaps the eccD 5 gene at the 3= end, the eccD5R2 primer was designed 5= of the predicted mycP 5 start codon. The resulting construct encodes the first 5 amino acids of EccD 5 fused in frame to the last 16 amino acids of EccD 5 . The PCR products were cloned in pCR2.1-TOPO (Invitrogen) and sequenced. The 5= and 3= homology regions were subsequently removed from the pCR2.1 vector by digestion with PacI/AvrII or AvrII/AscI, respectively, and cloned in pJG1100 digested with PacI/AscI. The resulting pJG-ΔeccD 5 vector was confirmed by sequencing.
For tetracycline-inducible expression, eccD 5 was cloned in pTIC10a, a derivative of the integrative pTIC6 vector (39), which contains a codon-optimized Tet repressor expressed under the control of the constitutive mycobacterial groEL promoter, the aph Kan resistance marker, and the P smyc -tetO mycobacterial promoter and Tet operator 5= of the multicloning site. Full-length eccD 5 was amplified from M. tuberculosis Erdman genomic DNA by PCR with primers Tet-eccD5F and Tet-eccD5R, cloned in pCR2.1-TOPO (Invitrogen), and sequenced. The eccD 5 insert was removed by digestion with HindIII and EcoRI and cloned in similarly digested pTIC10a. The resulting pTIC-eccD 5 vector was confirmed by sequencing.
A vector for tetracycline-repressible expression of eccD 5 was constructed as previously described using Gateway vectors generously provided by Dirk Schnappinger (40). eccD 5 was amplified using primers eccD5_P1, which adds an attB2 site upstream of the eccD 5 promoter sequence, and eccD5_P4, which adds an attB3 site downstream of the eccD 5 stop codon, before being cloned into the pDO23A entry vector via a BP Gateway reaction (Gateway BP clonase II enzyme mix; ThermoFisher) to create pEN23A-eccD 5 . The final pGMCH-T38S38-750-eccD 5 vector was constructed via an LR Gateway reaction (Gateway LR clonase II enzyme mix; ThermoFisher) using plasmids pDE43-MCH, pEN41A-T38S38, pEN12A-P750, and pEN23A-eccD 5 and confirmed by sequencing. A similar strategy was used to construct a vector for the expression of eccD 5 -DAS containing the DASϩ4 degradation tag (29). eccD 5 -DAS was amplified using primers eccD5_P1, eccD5_P2, and eccD5_P3. The eccD5_P2 and eccD5_P3 primers were used in sequential PCRs to add the DASϩ4 tag with a stop codon and introduce an attB3 site as described previously (40). Gateway reactions to clone eccD 5 -DAS into pDO23A and to construct the final pGMCH-T38S38-750-eccD 5 -DAS plasmid were performed as described above, and the final plasmid was confirmed by sequencing.
Construction of a conditional eccD 5 depletion strain. The pTIC-eccD 5 vector was introduced into WT and ΔpstA1 mutant M. tuberculosis by electroporation as described previously (23), and transformants were selected by plating them on complete 7H10 agar containing 15 g/ml Kan. The presence of the plasmid in transformants was verified by PCR with the primer pair pTseqF/Q95R1. The resulting WT pTIC-eccD 5 and ΔpstA1 pTIC-eccD 5 strains were subsequently electroporated with pJGΔeccD 5 , and transformants were selected on 7H10 agar containing 50 g/ml Hyg and 15 g/ml Kan. Integration of the pJG-ΔeccD 5 vector at the eccD 5 locus in transformants was confirmed by PCR using primers Q94F1/mycP5R1 and Rv1794_3=F/Q96R1 to detect integration via the 5= and 3= regions of homology, respectively. Isolates in which the plasmid integrated site specifically were then grown in 7H9 supplemented with 15 g/ml Kan and 50 ng/ml anhydrotetracycline hydrochloride (ATc; Sigma) to induce eccD 5 expression from pTIC-eccD 5 and allow excision of pJGΔeccD 5 . The culture was serially diluted and plated on 7H10 agar containing 15 g/ml Kan, 100 ng/ml ATc, and 2% sucrose to counterselect the pJGΔeccD 5 plasmid. Sucrose-resistant isolates were patched to 7H10 medium containing 50 g/ml Hyg to identify those that had excised pJGΔeccD 5 . Hyg-sensitive colonies were grown in 7H9 medium containing 15 g/ml Kan and 50 ng/ml ATc and tested by PCR for the ΔeccD 5 mutation using primers Q94F1 and Q96R1. These strains were designated eccD 5 Tet-ON and ΔpstA1 eccD 5 Tet-ON.