Nα-Acetylation of the virulence factor EsxA is required for mycobacterial cytosolic translocation and virulence

The Mycobacterium tuberculosis virulence factor EsxA and its chaperone EsxB are secreted as a heterodimer (EsxA:B) and are crucial for mycobacterial escape from phagosomes and cytosolic translocation. Current findings support the idea that for EsxA to interact with host membranes, EsxA must dissociate from EsxB at low pH. However, the molecular mechanism by which the EsxA:B heterodimer separates is not clear. In the present study, using liposome-leakage and cytotoxicity assays, LC-MS/MS–based proteomics, and CCF-4 FRET analysis, we obtained evidence that the Nα-acetylation of the Thr-2 residue on EsxA, a post-translational modification that is present in mycobacteria but absent in Escherichia coli, is required for the EsxA:B separation. Substitutions at Thr-2 that precluded Nα-acetylation inhibited the heterodimer separation and hence prevented EsxA from interacting with the host membrane, resulting in attenuated mycobacterial cytosolic translocation and virulence. Molecular dynamics simulations revealed that at low pH, the Nα-acetylated Thr-2 makes direct and frequent “bind-and-release” contacts with EsxB, which generates a force that pulls EsxB away from EsxA. In summary, our findings provide evidence that the Nα-acetylation at Thr-2 of EsxA facilitates dissociation of the EsxA:B heterodimer required for EsxA membrane permeabilization and mycobacterial cytosolic translocation and virulence.

deaths (1,2). It is believed that after the Mtb-containing aerosolized droplets are inhaled into the lung, Mtb is encountered by alveolar macrophages and internalized into the phagosome, where Mtb manages to survive through arresting phagosome maturation, including inhibition of vATPase-mediated acidification (3)(4)(5)(6). Recent compelling evidence supports that Mtb penetrates the phagosome and translocates into the cytosol (termed cytosolic translocation), where Mtb replicates and undergoes cell-to-cell spreading (7). The ability of Mtb to arrest phagosome maturation and to translocate from the phagosome to the cytosol has been attributed, at least in part, to the type VII secretion system, named ESX-1 and the secreted virulence factors EsxA (ESAT-6) and EsxB (CFP-10). The Mtb mutants with either gene deletions or defects in secretion of EsxA and/or EsxB were not able to translocate into the cytosol and showed significant reduction in host-cell lysis and cell-to-cell spreading (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19).
In our previous studies, we have found that Mtb EsxA exhibits a unique membrane-permeabilizing activity that is not present in the homologous EsxA ortholog from nonpathogenic Mycobacterium smegmatis (Ms) (20). EsxA undergoes pH-dependent conformational changes, inserts into the membrane, and forms a membrane-spanning complex (21). The essential role of EsxA membrane-permeabilizing activity in phagosome rupture and cytosolic translocation is further confirmed by a recent study in which single-residue mutations at Gln-5 of EsxA up-or down-regulated the membrane-permeabilizing activity and consequently up-or down-regulated mycobacterial cytosolic translocation and virulence in cultured cells and in zebra fish (22).
The genes encoding EsxA and EsxB are located in the same operon within the ESX-1 locus. EsxA and EsxB are co-expressed and co-secreted as a heterodimer (23). Our earlier study has demonstrated that EsxA, but not EsxB, has the membrane-permeabilizing activity, and EsxB is believed to function as a chaperone (20). Current studies support a model that the heterodimer is dissociated at low pH to allow EsxA to penetrate the membranes (24). However, the data regarding to the heterodimer dissociation are conflicting. The native heterodimer extracted from Mtb culture filtrate was found to be dissociated at low pH (24). Surprisingly, however, the studies using the recombinant proteins prepared from Escherichia coli suggest that the heterodimer was not dissociated by acidification. This is evidenced by one of our earlier studies that the heterodimer prepared from E. coli (hereafter termed "Ec-heterodimer") was inactive in membrane disruption. In the absence of lipid membranes, EsxA formed aggregates in the acidic solution because of increased solvent-exposed hydrophobicity. In contrast, the heterodimer showed little aggregation at pH 4.0, suggesting that EsxB remains bound to EsxA at low pH and prevents EsxA from forming aggregates, which otherwise would be observed if EsxA was released from EsxB (20). Our data are consistent with an earlier CD analysis showing that the Ec-heterodimer is not dissociated at low pH (25).
We hypothesized that the mycobacteria-produced proteins contain unique features (e.g. post-translational modifications) that are required for heterodimer dissociation at low pH. In line with this hypothesis, the native EsxA protein isolated from the culture filtrate of Mtb was displayed as multiple spots in 2-D SDS-PAGE, and some of the spots contained a N ␣ -acetylation at residue Thr-2 (26). Moreover, the heterodimer produced from a Ms strain was found to have a N ␣ -acetylation on the Thr-2 residue of EsxA (27). Interestingly, EsxB preferred to bind the nonacetylated EsxA, but not the acetylated form in a 2-D overlay assay (26). Deletion of the N ␣ -acetyltransferase in Mycobacterium marinum (Mm) disrupted the homeostasis of EsxA N ␣ -acetylation and attenuated the virulence (28). Together, these studies suggest that the N ␣ -acetylation of EsxA plays an important role in mycobacterial virulence through facilitating heterodimer dissociation at low pH. In the present study, we have obtained the evidence showing that the N ␣ -acetylation at Thr-2 of EsxA is required for EsxA membrane permeabilization, mycobacterial cytosolic translocation and virulence through facilitating heterodimer dissociation.

The Ms-produced Mtb heterodimer, but not Ec-produced heterodimer, disrupted liposomal membrane at low pH
We hypothesized that Ms-heterodimer, but not Ec-heterodimer, dissociates at low pH and permeabilizes the liposomal membrane. The membrane-permeabilizing activity of Ms-heterodimer and Ec-heterodimer was tested with the ANTS/DPX fluorescence dequenching assay. As expected, Msheterodimer permeabilized the membrane at low pH, whereas Ec-heterodimer was not active (Fig. 1A). 4-Chloro-7-nitroben-zofurazan (NBD-Cl) only reacts with free N-terminal ␣-amino group in nonacetylated proteins and emits fluorescence, but it does not react with N ␣ -acetylated proteins because of a lack of free N-terminal amino group. Thus, we used NBD-Cl to test the states of N ␣ -acetylation for Ms-heterodimer and Ec-heterodimer. As expected, the Ms-heterodimer exhibited a significantly lower NBD-Cl fluorescence, compared with Echeterodimer, indicating that Ms-heterodimer, but not Ec-heterodimer, is N ␣ -acetylated (Fig. 1B).

The mutations at Thr-2 abolished the membranepermeabilizing activity of the Ms heterodimers through blocking separation of EsxA and EsxB
The Gln and Ala residues have been used to functionally mimic acetylation of amino group of an internal Lys residue, whereas an Arg serves as a nonacetylated control (29). Thus, we generated T2A, T2Q, and T2R mutations and tested the effects of these mutations on the heterodimer membranepermeabilizing activity.
Unexpectedly, all of the mutations abolished the Ms-heterodimer's membrane-permeabilizing activity (Fig. 2, A and B). The result suggests that either the mutations blocked the heterodimer separation or abolished the EsxA membrane-permeabilizing activity. To test whether the mutations abolished the membrane-permeabilizing activity, we purified the EsxA proteins containing the same mutations from E. coli and applied them to ANTS/DPX dequenching assay. The result showed that the mutations did not affect EsxA membrane-permeabilizing activity (Fig. 2, C and D), suggesting that the mutations at Thr-2 blocked the heterodimer separation at low pH.

The mutations that block EsxA and EsxB separation do not have N ␣ -acetylation
To validate the acetylation state of EsxA WT and the mutants, we developed a protocol to isolate EsxA and EsxB from the Ms-heterodimer proteins (Fig. 3A). Then the presence of N ␣ -acetylation on the EsxA proteins was measured by NBD-Cl (Fig. 3B). Consistent with the results in Fig. 1B, Ec-EsxA(WT) had a significantly higher fluorescence signal than Ms-EsxA(WT). Similar to Ec-EsxA(WT), the Ms-EsxA mutants (T2A, T2Q, and T2R) emitted significantly higher fluorescence

EsxB preferred to bind nonacetylated EsxA to inhibit the membrane-permeabilizing activity
An earlier study has shown that EsxB preferred to bind nonacetylated EsxA rather than acetylated EsxA in a 2-D overlay assay (26). Thus, we hypothesize that EsxB will prefer to inhibit the membrane-permeabilizing activity of the nonacetylated Ms-EsxA(T2A) than that of the acetylated Ms-EsxA(WT). First, we tested and confirmed that the proteins Ms-EsxA(WT) and Ms-EsxA(T2A) that were isolated from the heterodimers had similar membrane-permeabilizing activity to Ec-EsxA(WT), which once again confirms that the states of N ␣ -acetylation do not affect membrane-permeabilizing activity (Fig. 4, A and B). Then Ms-EsxA(WT) and Ms-EsxA(T2B) were incubated with EsxB at a series of EsxB/EsxA molar ratios. As expected, EsxB preferred to inhibit Ms-EsxA(T2A) rather than Ms-EsxA(WT) in membrane permeabilization, especially at the lower EsxB/ EsxA ratios (Fig. 5, A-C).

The mutations without N ␣ -acetylation attenuated mycobacterial virulence and inhibited cytosolic translocation
Here, we investigated the effects of the T2X mutations in mycobacterial pathogenesis. The genes carrying T2X mutations were expressed in the Mm⌬EsxA:B strain, in which the endogenous esxB-esxA operon was deleted. We found that the T2X mutations did not affect the expression and secretion of EsxA and EsxB in the Mm strains (Fig. S1). As expected, the Mm strains carrying the nonacetylated mutations T2A, T2Q, and T2R had a significantly lower cytotoxicity than the strain carrying the acetylated mutation T2S (Fig. 6A). Mm WT and Mm⌬EsxA:B were used as the positive control and negative control, respectively. Furthermore, using the previously established CCF4-FRET assay, we found that the nonacetylated mutations T2A, T2Q, and T2R abolished mycobacterial cytosolic translocation, whereas T2S maintained a similar activity as the WT (Fig. 6B). The data are consistent to a previous report indicating that deletion of a N ␣ -acetyltransferase in Mm disrupted the homeostasis of EsxA acetylation and attenuated the virulence (28).

Molecular dynamic simulation detects frequent "bind-andrelease" contacts between the acetylated Thr-2(Ac) and EsxB
The reported solution structure of EsxA:B heterodimer does not have the N ␣ -acetylation on Thr-2, and the Thr-2 residue is distal from the contact interface between EsxA and EsxB. It is not clear how the acetylation at Thr-2 affects the heterodimer separation at low pH. Thus, we performed (MD) simulation on the heterodimers with and without N ␣ -acetylation of Thr-2 at pH 7 and 4, respectively (Movie S1 and Fig. 7). At pH 7 the nonacetylated Thr-2 comes in a close vicinity of EsxB but is unable to make a direct contact to EsxB (Fig. 7, A and C). Compared with nonacetylated Thr-2, the acetylated Thr-2(Ac) moves further away from EsxB at pH 7 (Fig. 7, B and D). Interestingly, at pH 4 the nonacetylated N-terminal loop of EsxA has no direct contact with EsxB (Fig. 7, E and G), but the acetylated N-terminal loop is able to make direct contacts with EsxB in a frequent bind-and-release mode (Movie S1 and Fig. 7, F and H). The electrostatic force was calculated between two sets of residues as demonstrated in Fig. S2 using DelPhiForce program. A contact is defined as two residues that contain at least a pair of atoms within 4 Å distance. The magnitude of the force is 43.959 pN, and its direction is shown by the orange arrow (Fig. S2). Therefore, when the N terminus of EsxA is moving away from EsxB, it pulls EsxB away in the direction of the arrow. Hence, the MD simulation data support that the acetylated Thr-2 residue plays a significant role in the dissociation process of the complex.

Discussion
EsxA and EsxB are co-expressed and co-secreted as a heterodimer in mycobacteria. The roles of ESX-1, EsxA, and EsxB in mycobacterial cytosolic translocation and virulence have been confirmed in a series of studies (7, [17][18][19]. Earlier biochemical studies have demonstrated that EsxA has a pH-dependent membrane-permeabilizing activity, whereas EsxB appears to function as a chaperone for EsxA (20). Current studies support a model that the EsxA:B heterodimer is dissociated at low pH, which allows EsxA to permeabilize the membranes (24). However, the mechanism of the heterodimer separation is not known. For the first time, the present study tested the heterodimers containing EsxA with or without N ␣ -acetylation and obtained the evidence that the N ␣ -acetylation at Thr-2 of EsxA facilitates the heterodimer separation at low pH, which allows EsxA to permeabilize liposomal membrane in vitro, as well as mediate mycobacterial phagosome escape and cytosolic translocation in mycobacteria-infected macrophages.
As discussed above, the essentiality of EsxA and EsxB in mycobacterial pathogenesis has been well-documented in a series of reports. Genetic manipulations that either deleted the gene of esxA or esxB or abolished the secretion of EsxA and EsxB have attenuated mycobacterial virulence and inhibited the phagosome rupture, cytosolic translocation, and cell-to-cell spreading (7, 10,[17][18][19]. Moreover, the biochemical characterizations have demonstrated that EsxA possesses a unique membrane-permeabilizing activity that is not present in its ortholog in nonpathogenic M. smegmatis (20). Thus, it is reasonable to believe that during infection, the secreted EsxA exerts its membrane-permeabilizing activity to penetrate the phagosome membranes and facilitate mycobacterial cytosolic translocation. We have reported that the mutations at the Gln-5 residue of EsxA (e.g. Q5V and Q5K) have resulted in up-or downregulation of EsxA membrane-permeabilizing activity in vitro. Moreover, these mutations up-or down-regulated the mycobacterial virulence and cytosolic translocation accordingly, demonstrating the specific and accurate correlation between EsxA membrane-permeabilizing activity and mycobacterial virulence, as well as the ability to penetrate phagosome membrane (22). Once again, the present study provides new evidence that the N ␣ -acetylation at Thr-2 of EsxA is required for mycobacterial virulence and cytosolic translocation through facilitating the heterodimer separation.
Because Thr-2 has no any contact with EsxB as shown in the reported solution structure of EsxA:B heterodimer, how N ␣ -acetylation on Thr-2 affects heterodimer separation had  N ␣ -Acetylation of EsxA facilitates heterodimer separation become a puzzle. Here, the MD simulation result provides an interesting model, in which the acetylated Thr-2(Ac) has frequent bind-and-release contacts with EsxB at low pH, generating a dragging force to pull EsxB away from EsxA. N ␣ -Acetylation is a common modification in eukaryotes and plays important roles in protein-protein interaction, protein activity and stability, and cell growth and cell cycle, etc. (30,31). Although N ␣ -acetylation in eukaryotes has been well-studied, little is known about it in bacteria, including mycobacteria. Over 100 proteins in Mtb have been found to be N ␣ -acetylated, including EsxA (32), and protein acetylation has been correlated to pathogenesis (33,34). Currently, three N ␣ -acetyltransferases in E. coli, RimI, RimL, and RimJ, have been identified to acetylate ribosomal proteins at Ser-18, Leu-12, and Ser-5, respectively (34 -36). A biochemical study has found that RimI in Mtb has a relaxed substrate specificity, but little is known about the physiological substrates of the enzyme (37), which warrants further investigation.
A recent study has shown that the recombinant EsxA does not lyse cell membranes, and the lytic activity previously attributed to EsxA is due to residual ASB-14 detergent in the preparation (38). In fact, we had the similar observations that addition of the recombinant EsxA protein to the surface of lung epithelial cell lines WI-26 and A549 did not lyse the cells. Moreover, Conrad et al. (40) showed that blocking phagosomal acidification by bafilomycin did not decrease the ESX-1-mediated phagosomal permeabilization, suggesting that acidification is not required for membrane permeabilization. Recently, Lienard et al. (39) employed a collection of Mm ESX-1 transposon mutants, including the mutants that disrupt EsxA secretion, to infect macrophages and showed that the factors independent of EsxAB play a role in cytosolic translocation. It is not clear how the discrepancy arises and what is the broken link between the EsxA pH-dependent membrane-permeabilizing activity in liposomal membrane and the ability of mycobacteria to rupture phagosome membrane during infection. Other factors from mycobacteria and host cells, even including properties of target membranes, may be involved in this process, which warrants further investigation.

Generation of T2X mutations on the Mtb esxA gene for expression in E. coli, Ms, and Mm
For expression in E. coli-Using the previously reported plasmid pET22b-esxA-His 6 as the template (20 -22, 41), the mutations T2A, T2Q, T2R, and T2S were introduced into the esxA gene by PCR using the primers listed in Table S1. All of the mutations were confirmed by DNA sequencing. The resulting plasmids were transformed into BL21 (DE3) cells for expression. The cells were grown at 37°C while shaking at 250 rpm until A 600 reached 0.6 -0.8. Protein expression was induced by adding 1 mM isopropyl-␤-D-1thiogalactopyranoside for 3-8 h at 37°C. The cells were harvested, and the proteins were purified as previously described (20 -22, 41). Typical yield averaged between 30 and 60 mg/liter of culture.
For expression in Ms-The pMyNT plasmid containing the Mtb esxB-esxA operon (27) was used as the template. The mutations T2A, T2Q, T2R, and T2S were introduced by overlapping PCR using the primers listed in Table S1. All of the mutations were confirmed by DNA sequencing. The pMyNT plasmids carrying various T2X mutations were electroporated into Ms mc2155 strain (voltage, 2,500 V; capacitance, 25 microfarads; resistance, 1,000 ⍀). The Ms cultures were grown at 37°C overnight or until A 600 reached 2.0. Protein expression was induced by adding 0.2% (w/v) acetamide for 12-16 h. The cells were harvested. The soluble heterodimer proteins were purified with immobilized metal ion affinity chromatography by passing through a Ni 2ϩ column, followed by a size-exclusion N ␣ -Acetylation of EsxA facilitates heterodimer separation chromatography as previously described (27,(42)(43)(44)(45). Typical yield averaged between 20 and 40 mg/liter of culture.
For expression in Mm-The T2X mutations (T2A, T2R, T2Q, and T2S) were generated by site-directed mutagenesis (Agilent QuikChange kit) using the pMH406 plasmid containing esxB-esxA operon as a template. The mutations were confirmed by DNA sequencing. The mutated plasmids were electroporated into Mm⌬EsxA:B as previously described (44).

Liposome leakage assay
The liposome leakage assay was performed as previously described (20,21,41,46,47). Briefly, 20 mg of 1,2-dioleoyl-snglycero-3-phosphocholine was dried with nitrogen air and left in a vacuum overnight. The samples were rehydrated with 1 ml of the buffer (5 mM HEPES, 50 mM 8-aminonapthalene-1,3,6 trisulfonic acid (ANTS), and 50 mM p-xylene-bis-pyridinium bromide (DPX)). The suspension was subjected to six freezethaw cycles and extruded via a 0.2-m membrane filter 20 times. The liposomes were then desalted to remove excess ANTS and DPX using a HiTrap desalting column. The desalted liposomes were mixed with 150 mM NaCl, 100 mM NaAc at pH 4.0. Three separate reaction mixtures were generated and individually excited at 350 nm, and emissions were recorded at 520 nm in an ISS K2 phase modulation fluorometer. 100 g of the tested protein was injected into the solution after ϳ30 s of the assay, and fluorescence was observed.

Isolation of Mtb EsxA proteins from the Ms-produced Mtb EsxA:B heterodimer
The Mtb EsxA:B heterodimer purified from Ms was incubated in a solution containing 6 M guanidine at 4°C overnight. The proteins were then passed through a HisTrap column (GE Healthcare). The His-tagged EsxB protein was bound to the column, and the untagged EsxA was collected in flow through. The His-tagged EsxB protein was eluted by an imidazole gradient. Both EsxA and EsxB were subjected to an extensive dialysis using a 3000 molecular weight cutoff membrane. The samples were then concentrated and passed through gel filtration for a complete buffer exchange.

Detection of N ␣ -acetylation by NBD-Cl
NBD-Cl only reacts with free N-terminal ␣-amino group in nonacetylated proteins and emits fluorescence, and it does not react with N ␣ -acetylated proteins because of the lack of free N-terminal amino group (48). Proteins (6 M) were incubated with 0.5 mM NBD-Cl at room temperature. At different hours of postincubation, the samples were subjected to fluorescence measurement with excitation at 460 nm and emission at 535 nm.

Sample preparation for mass spectrometry
The concentration of four protein samples, EsxA protein purified from E. coli and the EsxA, EsxA(T2A), and EsxA proteins purified from Ms were determined via bicinchoninic acid assay (BCA) according to manufacturer instructions (Pierce BCA protein assay kit, catalog no. 23227). Then 100 g of each protein sample was used for protein digestion via FASP protein digestion (Expedeon, catalog no. 44250), using either trypsin (catalog no. Sigma, catalog no. T6567) or pepsin (Sigma, catalog no. P7012). In summary, 100 g of each sample was resuspended in 200 l of 12.48 M urea Tris-HCl solution (urea solution). Then 10 mg of DL-DTT (Sigma, catalog no. D0632-5G) was added to each sample and placed in a nutating mixer for 45 min. The samples were then transferred to a 30-kDa filter and centrifuged for 15 min at 14,000 rpm. 200 l of urea solution was added to each spin filter and centrifuged at 14,000 ϫ g for 15 min and repeated once more. Then 100 l of iodoacetamide 1ϫ solution (iodoacetamide prepared at a 1:10 ratio in urea solution) was added to each spin filter and incubated without mixing for 20 min in the dark. Afterward, spin filters were centrifuged at 14,000 rpm for 10 min. The spin filters were washed twice by adding 100 l of urea solution and centrifuging at 14,000 rpm for 15 min. Urea was removed from the samples by adding 100 l of 50 mM ammonium bicarbonate solution to spin filters and centrifuged at 14,000 rpm for 15 min. This step was repeated twice for a total of three times. Spin filters were transferred to new spin filter tubes, and 100 l of digestion solution at 0.02 g/l (either with trypsin or pepsin) was added to each sample and incubated at 37°C for 18 h. 200 l of 0.1% formic acid water solution was added to each sample and spin filters containing resultant peptides and centrifuged at 14,000 ϫ g for 10 min. Filtrate containing digested proteins were frozen at Ϫ80°C for 2 h and lyophilized for 12 h. The samples were resuspended with 100 l of 0.1% formic acid water solution at a final concentration of 1 g/l.

LC-MS/MS and bioinformatic method
Resultant complex peptide mixtures were analyzed via technical duplicates by LC-MS/MS for 2 h to 1 day with the QE Orbitrap (Thermo Fisher Scientific) along with the Dionex UltiMate 3000 RSLCnano UHPLC system (Thermo Fisher Scientific) using the two digestive protocols (trypsin or pepsin). The samples were loaded in line onto a C18 PicoChip column (75-m inner diameter ϫ 15-m tip packed with 10.5 cm of Reprosil PUR C18 3 m 120 Å; 25 m ϫ 50 cm fused-silica tail, New Objective) that was previously equilibrated with solvent A (95% water, 5% acetonitrile, 0.1% formic acid) and solvent B (5% water, 95% acetonitrile, 0.1% formic acid). The column was conditioned for 10 min with a flowrate of 0.5 l/min with 95% solvent A, 5% solvent B. The sample was injected and loaded onto column and conditioned for 10 min with solvent A. Elution of peptides was completed by running a linear gradient to 40% solvent B for 95 min, followed by 10 min of 95% solvent B. The sample was re-equilibrated for 10 min with 5% solvent B. Full-scan spectra were collected via Xcalibur (Thermo Fisher Scientific). Two blank injections were run in between samples using a 30-min interval with seesaw washes using 5-80% ACN gradients. QE Orbitrap settings are as follows: full MS resolution of 70,000, AGC target of 1e6, scan range from 400 to 1600 m/z; MS/MS were run with a resolution of 17,500, AGC target of 2e6, scan range from 200 to 2000, and 3 m/z isolation window. The spectra were searched using Proteome Discoverer 2.1.1.21 (Thermo Fisher Scientific) and filtered via Sequest HT with an estimated false-discovery rate of 0.01 against sequences from Ms M1552, E. coli BL21, human, bovine, human keratin, and porcine trypsin. A 20-ppm precursor and a 0.02 fragment N ␣ -Acetylation of EsxA facilitates heterodimer separation mass tolerance were used. Cysteine carbamidomethylation, methionine, oxidation, and acetylation were set fixed and variable modifications, respectively. The output files were then manually analyzed and collectively used to generate representative data.

Western blotting
Mm recombinant strains with respective EsxA insertions were cultured in 7H9 medium and grown to mid-log phase. They were washed with PBS and transferred to Sauton's medium while normalizing all cultures to A 600 ϭ 0.8. The bacteria were cultured for 2 days until harvest. The bacterial cells were collected by centrifugation. The proteins in the culture supernatant were precipitated by TCA. The bacterial cells were resuspended in 1 ml of PBS containing a mixture of protease inhibitors (Thermo Fisher Scientific) and sonicated at 30% amplitude for five cycles of 30-s pulse and 60-s rest. The culture filtrates and total bacterial lysates were applied to SDS-PAGE and transferred transferring onto polyvinylidene difluoride membrane. Western blots were performed to detect EsxA using anti-EsxA antibody (sc-57730, Santa Cruz). As controls, Ag85 (secreted in culture filtrate) and GroEL (only in cell lysate) were also detected by anti-Ag85 (NR-13800, BEI) and anti-GroEl antibodies (NR-13813, BEI), respectively.

Live/dead cytotoxicity assay
RAW264.7 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum with penicillin and streptomycin (100 units/ml) at 37°C and 5% CO 2 . Raw 264.7 macrophages were plated in a 24-well plate with a density of 5 ϫ 10 5 /well for infection on the following day. The Mm strains were prepared with a single cell preparation protocol as previously described (22,49). RAW264.7 cells were infected with various Mm strains at a multiplicity of infection (MOI) of 10 for 1 h. The macrophages were washed three times with PBS to remove free mycobacteria and incubated for another 3 h. The macrophages were stained using calcein-AM and ethidium homodimer (Life Sciences) for 30 min, enabling visualization under a fluorescence microscope for green cells (live) and red cells (dead). The numbers of dead cells were quantified from dozens of random fields from each sample.

CCF-4 FRET assay
Mycobacterial cytosolic translocation was measured by CCF-4 FRET assay as previously described (17,22,50). Briefly, RAW264.7 cells were plated in triplicate in a 6-well plate at a density of 2.5 ϫ 10 6 cells/well. The macrophages were then incubated with CCF4-AM according to the manufacturer's protocol (Liveblazer B/G loading kit, Life Sciences). The cells were infected with recombinant Mm strains at a MOI of 10 for 2 h. Following infection, the macrophages were washed three times using PBS. Dulbecco's modified Eagle's medium with 10% fetal bovine serum was added to the cells and incubated for ϳ2 days. The samples were then excited at 409 nm, and the emissions were measured at 450 and 535 nm. The blue/green ratio was calculated as I 450 /I 535 .

Molecular dynamic simulation
The structure of EsxA:B heterodimer was downloaded from Protein Data Bank with Protein Data Bank code 1WA8 (51). DelPhiPka web server (52) was used to obtain the protonation states of ionizable residues at pH 4 and pH 7 and assign the respective states with visual molecular dynamics (VMD) (53). N ␣ -Acetylation of Thr-2 was performed on VMD after removal of Met-1. The four structures, i.e. nonacetylated and acetylated at pH 4 and pH 7 of EsxA:B heterodimer, were then solvated in water box with TIP3 (54) water model and ionized with 150 mM NaCl in VMD. The final systems were then simulated with the MD simulation program NAMD (55). Each simulation was performed for 20 ns employing force field CHARMM27 (56). The temperature was set as 300 K, and the pressure was 1 atm. The snapshots from the simulations were taken to study the behaviors of the N-terminal loop of EsxA with and without N ␣ -acetylation of Thr-2 at pH 4 and pH 7.

Data availability statement
All data are contained within the manuscript.