An N-acetyltransferase required for ESAT-6 N-terminal acetylation and virulence in Mycobacterium marinum

ABSTRACT N-terminal protein acetylation is a ubiquitous post-translational modification that impacts diverse cellular processes in higher organisms. Bacterial proteins are also N-terminally acetylated, but the mechanisms and consequences of this modification in bacteria are poorly understood. The major virulence factor EsxA (ESAT-6, early secreted antigen, 6 kDa) was one of the first N-terminally acetylated proteins identified in bacteria. EsxA is conserved in mycobacterial pathogens, including Mycobacterium tuberculosis and Mycobacterium marinum, a non-tubercular mycobacterial species that causes tuberculosis-like disease in ectotherms. However, the enzyme responsible for EsxA N-terminal acetylation has been elusive. Here, we used genetics, molecular biology, and mass-spectroscopy-based proteomics to demonstrate that MMAR_1839 (renamed Emp1, ESX-1 modifying protein, 1) is the putative N-acetyltransferase (NAT) solely responsible for EsxA acetylation in M. marinum. We demonstrated that ERD_3144, the orthologous gene in M. tuberculosis Erdman, is functionally equivalent to Emp1. We previously quantified widespread N-terminal protein acetylation in pathogenic mycobacteria (C. R. Thompson, M. M. Champion, and P. A. Champion, J Proteome Res 17:3246–3258, 2018, https:// doi: 10.1021/acs.jproteome.8b00373). We identified at least 22 additional proteins that require Emp1 for acetylation, demonstrating that this putative NAT is not dedicated to EsxA. Finally, we showed that loss of emp1 did not prevent phagosomal escape but resulted in a significant reduction in macrophage cytolysis by and cell-to-cell spread of M. marinum during infection. Collectively, this study identified a NAT required for N-terminal acetylation and pathogenesis in Mycobacterium. IMPORTANCE N-terminal acetylation is a protein modification that broadly impacts basic cellular function and disease in higher organisms. Although bacterial proteins are N-terminally acetylated, little is understood how N-terminal acetylation impacts bacterial physiology and pathogenesis. Mycobacterial pathogens cause acute and chronic disease in humans and in animals. Approximately 15% of mycobacterial proteins are N-terminally acetylated, but the responsible enzymes are largely unknown. We identified a conserved mycobacterial protein required for the N-terminal acetylation of 23 mycobacterial proteins including the EsxA virulence factor. Loss of this enzyme from M. marinum reduced macrophage killing and spread of M. marinum to new host cells. Defining the acetyltransferases responsible for the N-terminal protein acetylation of essential virulence factors could lead to new targets for therapeutics against mycobacteria.

N-terminal acetylation is the covalent addition of an acetyl group to the α amino group of the N-terminal amino acid of a protein by N-acetyltransferases (NATs) (25)(26)(27).NATs can irreversibly acetylate the iMet (initiator Met, following deformylation) or the first amino acid following iMet cleavage (28,29).EsxA was one of the first bacterial proteins recognized to be N-terminally acetylated (30).The role of EsxA N-terminal acetylation has been linked to protein interaction in vitro (30,31), and phagosomal lysis (31) using Mycobacterium marinum, an established model for studying the mycobacterial ESX-1 system (32).Importantly, the deletion of esx-1 genes in M. marinum is functionally complemented by the expression of orthologous genes from M. tuberculosis, demon strating a conserved function (33).
Bacterial genomes encode several putative NATs, which are part of the GNAT (GCN-5-related NAT) family (26,47).It is impossible to predict if the putative NAT acetylates Lys residues (KATs), small molecules, antibiotics, and/or protein N-termini (26).There are 27 predicted NATs in M. tuberculosis, 23 of which are conserved in M. marinum (43).The best characterized NAT is RimI, which N-terminally acetylates the S18 rRNA protein and functions as a generalist KAT in Escherichia coli, Salmonella, and in Mycobacterium (47)(48)(49)(50)(51).The individual NATs responsible for N-terminal acetylation of specific mycobacterial proteins are unknown, limiting our understanding of how this modification impacts virulence and physiology.
The EsxA NAT or NATs have remained elusive.Changes in EsxA acetylation are difficult to assess because the amino acid composition of EsxA yields a large N-terminal tryptic fragment with poor chromatographic performance by mass-spectroscopy-based proteomics (52)(53)(54).Top-down approaches do not rapidly assign all N-terminal isoforms of EsxA (52)(53)(54).
We sought to identify the NAT responsible for N-terminally acetylating EsxA to further understand N-terminal acetylation in Mycobacterium.Based on the conservation of EsxA and the putative NATs between M. marinum and M. tuberculosis, we hypothesized that we could leverage the use of M. marinum to identify the conserved EsxA NAT.To test this hypothesis, we used an N-terminal acetyl-EsxA antibody coupled with a knockout M. marinum strain collection to identify the EsxA NAT.We measured EsxA acetylation in the presence and absence of the putative NAT using western blot analysis, MALDI, and label-free quantitative (LFQ) mass spectrometry.We tested ESX-1 function and mycobac terial virulence using in vitro systems and a macrophage model of infection.

In vitro discrimination between the EsxA and acetyl-EsxA N-termini
We obtained and characterized a polyclonal antibody synthesized against an acetylated N-terminal EsxA peptide (Fig. 1A, Ac-EsxA NT ) for use as a tool to identify the NAT(s) responsible for N-terminal acetylation of EsxA.We performed a dot blot to determine if the Ac-EsxA antibody specifically recognized the acetylated N-terminus of EsxA.The Ac-EsxA antibody specifically produced a signal where the Ac-EsxA NT peptide was spotted on the nitrocellulose (red) but not where the unacetylated peptide was spotted (Fig. 1B, EsxA NT , outline).In contrast, a commercially available EsxA antibody raised against EsxA NT peptide specifically produced a signal where the unacetylated EsxA peptide was spotted onto the nitrocellulose (green), but not where the Ac-EsxA NT peptide was spotted.From these data, we conclude that the Ac-EsxA antibody discrimi nates between the acetylated and unacetylated forms of the EsxA N-terminal peptide.Moreover, the commercial EsxA antibody specifically recognizes the unacetylated form of EsxA.
We next tested if the Ac-EsxA antibody could detect the acetylated version of the EsxA protein (Ac-EsxA) in a complex mixture.We collected cell-associated and secreted protein fractions from the wild-type (WT) and ΔesxBA M. marinum strains.We separated the proteins by charge using neutral pH urea Triton polyacrylamide gel electrophore sis (NUT-PAGE), which separates acetylated and unacetylated proteins (55).NUT-PAGE followed by western blot analysis allowed for the separation of several protein species (i-iii) detected by the EsxA antibody in the protein fractions collected from the WT M. marinum strain (Fig. 1C).All three of these species were absent from the protein fractions generated from the ΔesxBA strain, which fails to produce EsxA protein.Notably, only species iii, which is the most negatively charged EsxA species, was reliably detected by the Ac-EsxA antibody.From these data, we conclude that we can separate and detect acetylated EsxA from a complex mixture of M. marinum proteins.

N-terminal acetylation of EsxA and other proteins is dependent on MMAR_1839
The EsxA proteins from M. marinum and M. tuberculosis are identical through the 15th amino acid and 92% identical overall (Fig. S1).Therefore, we reasoned that the NAT responsible for acetylating EsxA would be highly conserved.We identified the five most highly conserved GNAT proteins between the two species (Table 1).We generated unmarked deletions of each putative NAT gene in M. marinum and confirmed the deletion of each gene by PCR (Fig. S2) and targeted DNA sequencing.
We hypothesized that Ac-EsxA would be absent from M. marinum strains lacking an EsxAspecific NAT.We collected cell-associated proteins from M. marinum strains lacking each of the five most conserved NATs.We measured EsxA and Ac-EsxA in these strains using western blot analysis, as compared to proteins generated from the WT and ΔesxBA strains.Both the EsxA and Ac-EsxA proteins were detected in lysates generated from the WT M. marinum strain (Fig. 2A, lane 1), consistent with prior studies demonstrating that both species exist in the WT strain (30,43,(56)(57)(58)(59).Both EsxA species were lacking from the ΔesxBA strain (lane 2), demonstrating the specificity of both antibodies to EsxA.The EsxA and Ac-EsxA proteins were present in the lysates generated from the ΔMMAR_1067, ΔMMAR_4519, or ΔMMAR_1882 strains (lanes 3, 5, and 7).The ΔMMAR_1968 strain lacks the ArgA NAT and is auxotrophic for arginine (60).The ΔMMAR_1968 strain exhibited reduced growth compared to the WT and complemented strains (Fig. S3).Addition of L-arginine to the growth media restored the growth of the ΔMMAR_1968 strain (Fig. S3), allowing detection of both Ac-EsxA and EsxA protein from the ΔMMAR_1968 lysates (Fig. 2B).The EsxA protein (Fig. 2A, lane 6), but not the Ac-EsxA protein, was detected in the lysate from the ΔMMAR_1839 strain.From these data, we conclude that MMAR_1839 is required for EsxA acetylation in M. marinum.We renamed MMAR_1839, ESX-1 modifying protein-1 (Emp1).
To determine if Emp1 was required for the acetylation of additional mycobacterial proteins, we performed LFQ mass spectrometry to measure the relative changes in acetylation and protein levels in the Δemp1 strain as compared to the WT and comple mented strains (Δemp1/comp, Data set S1, Raw and Trimmed Data tabs S1A and S2B; the raw data have been uploaded to MassIVE due to size requirements and can be found at  ftp://MSV000091442@massive.ucsd.edu).EsxA levels were comparable in all three strains (Fig. 2C; gray bars, Data set S1; Table S1C).Ac-EsxA was detected in the WT and the complemented strains but not in the Δemp1 strain.From these data, we conclude that Emp1 is the only NAT responsible for EsxA acetylation in M. marinum.
We next performed k-means (62) clustering analysis for each N-terminally acetylated protein in the data set, using the biological replicate with the best coverage (Data set S1; Table S1C).This approach systematically identified patterns between the WT, ∆emp1, and complementation strains across every protein.The variables considered for the clustering were from the following LFQ area ratios: Δemp1:Δemp1/comp, Δemp1:WT, and Δemp1/comp:WT.The proteins were clustered into three groups, using 25 random starting points.We reasoned that proteins that clustered with EsxA were potential acetylation targets of Emp1, as they also exhibited loss of acetylated intensity in the ∆emp1 strain, which was restored upon complementation.The acetylation intensity patterns of proteins identified from the clustering were compared across all biological replicates.
This approach revealed a cluster of 22 proteins whose N-terminal acetylation followed a similar pattern to Ac-EsxA.All 23 proteins, including EsxA, exhibited undetectable levels of N-terminal acetylation in the Δemp1 strain, and restoration of acetylation in the Δemp1/comp strain, similar to the WT strain (Fig. 2E; EsxA highlighted in pink, Data set Table S1E).Functional analysis using annotations in Mycobrowser (63) revealed that proteins dependent on Emp1 for acetylation are predicted to function in lipid metab olism or intermediary metabolism and respiration, and virulence (Fig. 2E).More than half are annotated as essential in vitro in M. tuberculosis.Finally, comparing the N-termi nal sequences of the putative Emp-1 targets against the entire M. marinum proteome revealed a strong negative bias for basic residues within these first 10 amino acids (Fig. 2F, upper), likely due to the use of trypsin for the mass-spectrometry-based proteomics, which cleaves after Lys and Arg.Consequently, those peptides are underrepresented in the first 10 amino acids of the N-termini because they would not be observed due to their small size (43,54).There was also a strong preference for threonine, with a mild preference for serine at the second amino acid position, consistent with our prior work (43).Comparison of the N-terminal sequences of the putative Emp1-targets against the M. marinum N-terminal acetylome (43) revealed a strong preference for threonine and glutamic acid at the second and third amino acid positions and a significant underrepre sentation of alanine at the sixth amino acid position (Fig. 2F, lower).Together, these data demonstrate that Emp1 is required for the N-terminal acetylation of EsxA and at least 22 other proteins in M. marinum.

Emp1, and therefore EsxA acetylation, is dispensable for EsxA/EsxB secretion from M. marinum
The identification of Emp1 allowed us to test the role of EsxA acetylation on ESX-1 function in the presence of the WT esxA gene in M. marinum.It was previously suggested that EsxA acetylation impacted the interaction between EsxA and its binding partner, EsxB (30).The EsxA-EsxB interaction is required for ESX-1 function because it is required for protein stability and for targeting the EsxA-B pair to the ESX-1 system (6,56,64).If N-terminal acetylation of EsxA was required for EsxA-EsxB interaction then we would expect a loss of EsxB protein from the Δemp1 strain.We generated cell-associated and secreted protein fractions from M. marinum strains.EsxA, Ac-EsxA, and EsxB were produced (Fig. 3A, lane 1) and secreted (lane 7) from the WT M. marinum strain.EsxA and Ac-EsxA levels were reduced in the ΔeccCb 1 strain (lane 2), consistent with the reduced levels of EsxA in the absence of secretion (6,65).Neither EsxA, Ac-EsxA, nor EsxB was secreted from the ΔeccCb 1 strain (lane 8).Although EsxA and EsxB were produced in the Δemp1 strain, Ac-EsxA was not detected (lane 3).EsxA and EsxB were secreted from the Δemp1 strain (lane 9).Constitutive expression of the emp1 gene restored the production (lane 4) and secretion of Ac-EsxA (lane 10).Likewise, expression of the orthologous gene from M. tuberculosis Erdman (ERD_3144) restored the production and secretion of Ac-EsxA from the Δemp1 strain (lanes 5 and 11).Finally, we mutated the predicted active site of Emp1 (W223A), which would render the enzyme unable to bind Ac-CoA (66).Expression of emp1W223A in the Δemp1 strain did not restore the production or secretion of Ac-EsxA (lanes 6 and 12).Together, these data demonstrate that Emp1 is required for the acetylation of EsxA.Contrary to the existing models (31), the loss of EsxA acetylation did not result in a loss of the EsxA or EsxB protein, suggesting that the acetylation state of EsxA is dispensable for EsxA/EsxB interaction and secretion from M. marinum during in vitro growth.Our data support that Emp1 likely functions as a NAT in M. marinum and is functionally conserved in M. tuberculosis.
We sought a more sensitive approach to verify that EsxA was completely unacetylated in the Δemp1 strain and rule out cross-talk by other putative NATs in the absence of emp1.We previously demonstrated that both Ac-EsxA and EsxA are resolved in proteins washed from the surface of M. marinum colonies using whole-colony MALDI-TOF-MS
The ΔeccCb 1 strain is a lysis control because this strain produces but does not secrete EsxA and EsxB (6,17).Both EsxA species and EsxB were significantly diminished from the proteins isolated from the surface of the ΔeccCb 1 strain (6, 58).Therefore, the observed peaks were due to the secretion of EsxA and EsxB to the cell surface.Proteins isolated from the surface of the Δemp1 strain resulted in a single EsxA peak which corresponded to the unacetylated EsxA protein and a peak for EsxB.The Ac-EsxA peak was completely abrogated.Expression of the WT emp1 gene, but not the emp1W223A gene, restored the peak corresponding to the Ac-EsxA protein.From these data, we conclude that deletion of the emp1 gene results in a complete loss of Ac-EsxA in M. marinum, demon strating that Emp1 is solely responsible for the acetylation of EsxA in vivo.The absence of acetylation in the W223A active-site mutant of Emp-1 (Fig. 3B) demonstrates that functional Emp-1 is required for the acetylation of EsxA.Moreover, our findings show that EsxA and EsxB are secreted from M. marinum independently of EsxA-N-terminal acetylation during in vitro growth.M. marinum lyses red blood cells in a contact-dependent, ESX-1-dependent manner (17).EsxA is required for the hemolytic activity of M. marinum, likely because it is required for the secretion of the majority of the ESX-1 substrates (5,7,33,67).M. marinum strains that secrete all of the known ESX-1 substrates, including the late substrates, EspE and EspF, are hemolytic (17,68).The ΔesxA M. marinum strain is non-hemolytic (5).Because the acetylation of EsxA depends on Emp1, we reasoned that if EsxA acetylation was required for EsxA-mediated secretion, the Δemp1 strain would have altered hemolytic activity.WT M. marinum lysed sheep red blood cells (sRBCs), while the ΔeccCb 1 strain (which fails to secrete ESX-1 substrates) exhibited significantly reduced hemolytic activity (Fig. 3C, P < 0.0001, relative to the WT strain).Water and phosphatebuffered saline (PBS) (cell-free) were used as positive and negative controls for hemolysis, respectively.The activity of the ΔeccCb 1 strain was not significantly different from the PBS control (P < 0.9999).The hemolytic activities of the Δemp1 and the Δemp1 complemented strains were not significantly different from the WT strain (P = 0.4837 and P = 0.9998) or each other (P = 0.2689).From these data, we conclude that Emp1 is dispensable for the hemolytic activity of M. marinum.Because ESX-1 mediates hemolysis, the data suggest that the acetylation of EsxA is also dispensable for hemolysis and is consistent with the secretion of EsxA and EsxB from the Δemp1 strain (Fig. 3B).Finally, because additional ESX-1 substrates required for hemolysis depend upon EsxA for secretion, our data suggest that the secretion of additional ESX-1 substrates occurs independently of EsxA acetylation.

Emp1 is required for macrophage cytolysis and cell-to-cell spread of M. marinum
It was previously reported that N-terminal acetylation of EsxA was required for ESX-1 function during infection (31).We reasoned that if acetylation of EsxA was required for ESX-1 function then the Δemp1 M. marinum strain would phenocopy the ΔeccCb 1 strain during macrophage infection (6,69,70).We infected RAW 264.7 cells with M. marinum at a multiplicity of infection (MOI) of 0.2 and measured colony-forming units over time.The WT M. marinum strain grew in the macrophages, while the ΔeccCb 1 strain was attenuated for growth (Fig. 4A, P < 0.0001, relative to all of the additional strains).Growth of the Δemp1 and the Δemp1 complemented strains was not significantly different from the WT strain.We conclude that the Δemp1 strain is not attenuated for growth during macrophage infection, distinct from ΔeccCb 1 strain.
ESX-1-dependent phagosomal damage allows M. marinum translocation to the macrophage cytoplasm (Fig. 4B).It was previously reported that the N-terminal acetylation of EsxA was required for ESX-1-dependent phagosomal lysis (31).To test if the Δemp1 strain escapes the phagosome, we adopted a reporter used to reflect (Continued on next page) phagosomal escape and cytoplasmic bacteriolysis in Listeria monocytogenes and in M. tuberculosis (9,71).Bacterial escape from the phagosome and subsequent bacteriolysis in the macrophage cytoplasm releases a reporter plasmid, allowing plasmid translocation to the nucleus and transcription of the luc+ gene, causing luminescence.Phagosomal bacteria do not cause luminescence because the reporter plasmid does not access the cytoplasm.We adapted the reporter plasmid by adding an episomal mycobacterial origin and introduced it into M. marinum.We optimized the assay to distinguish between the cytoplasmic WT M. marinum strain and the phagosome-localized ΔeccCb 1 strain (P < 0.0001, Fig. S3B).Expression of the eccCb 1 gene in the ΔeccCb 1 strain signifi cantly restored luminescence levels.These data confirm that the bacteriolysis reporter distinguishes between phagosomal and cytoplasmic M. marinum strains and measures cytoplasmic M. marinum bacteriolysis.
We infected iIFNAR −/− macrophages at an MOI of 20 and measured luminescence at 20 and 24 hours post infection (Fig. 4C).We reasoned that if acetylation of EsxA was required for phagosomal escape then the Δemp1 M. marinum strain would be retained in the phagosome, similar to the ΔeccCb 1 strain.At 20 hours post infection, the Δemp1 strain resulted in luminescence levels similar to the ΔeccCb 1 strain (P = 0.6191), significantly lower than the WT and complemented strain (P = 0.0003 vs WT, P = 0.0001 vs Δemp1/comp).However, by 24 hours post infection, the Δemp1 strain resulted in significantly higher levels of luminescence than the ΔeccCb 1 strain (P < 0.0001), although still significantly lower than both the WT (P = 0.0264) and complemented (P = 0.0195) strains.From these data, we conclude that the Δemp1 strain is not retained in the phagosome like the ΔeccCb 1 strain, suggesting that the N-terminal acetylation of EsxA is not essential for phagosomal escape.Our data suggest that the Δemp1 strain is either released from the phagosome less efficiently than the WT strain or exhibits delayed or reduced bacteriolysis in the macrophage cytoplasm compared to the WT strain.
Release of M. marinum into the cytoplasm promotes macrophage cytolysis through ESX-1-independent mechanisms (Fig. 4B) (4,11,12,(72)(73)(74)(75). To test if Emp1 was required for macrophage cytolysis, we infected RAW 264.7 cells with M. marinum at an MOI of 5 and measured uptake of the membrane impermeable dye, ethidium homodimer 1 (EthD-1).Infection with WT M. marinum resulted in significant cytolysis, as reflected by EthD-1 uptake, compared to the uninfected cells (Fig. 4D, P < 0.0001).The ΔeccCb 1 strain was significantly less cytolytic than the WT strain (P < 0.0001), similar to the uninfected control.The Δemp1 strain was significantly less cytolytic than the WT strain (P < 0.0001) but significantly more cytolytic than the ΔeccCb 1 strain (P < 0.0001).Constitu tive expression of the emp1 gene in the Δemp1 strain restored cytolysis to levels similar to the WT strain.Representative images of macrophage cytolysis are shown in Fig. S5.To confirm that these strains were not attenuated due to the spontaneous loss of the outer lipid phthiocerol dimycocerosates (PDIM), we performed thin-layer chromatogra marinum strains in RAW 264.7 cells, 24 hpi, MOI = 5.The data include at least three biological replicates each in technical triplicate, with 10 fields selected for each well.Each data point is the number of red cells per field.Significance was determined using an ordinary one-way ANOVA (P < 0.0001) followed by a Sidak's multiple comparison test.The significance shown is compared to the WT strain, with additional comparisons discussed in the text.****P < 0.0001.(E) Spread of M. marinum strains in a RAW 264.7 monolayer, at 24, 48, and 72 hpi, MOI = 1.The data include four biological replicates, each in technical triplicate, with 10 fields taken from each well, with the exception of the 24-hour timepoint.Each data point is the bacterial counts per field as measured by counting GFP "particles" using ImageJ.Significance was determined using an ordinary one-way ANOVA followed by a Tukey's multiple comparison test.****P < 0.0001.Black asterisks represent the comparison to the WT strain from the same timepoint.***P = 0.0001 and **P = 0.0060.The gray asterisks represent the comparison between the 24 timepoint with the 48-or 72-hour timepoints for each strain.***P = 0.0080.If no asterisks are listed, the P value is not significant.phy (TLC) analysis.All of the strains produced PDIM similar to the WT strain (Fig. S6).From these data, we conclude that Emp1 is required for macrophage cytolysis.
Macrophage cytolysis precedes cell-to-cell spread of M. marinum (Fig. 4A).We tested if Emp1 was required for M. marinum to spread between macrophages.To test the ability of M. marinum to spread, we introduced a plasmid that expresses green fluorescent protein (GFP) from a tet-ON promoter (the promoter is turned on by the addition of ATc) into the WT, ΔeccCb 1 , Δemp1, and complemented strains.We measured GFP expression in the M. marinum cells grown under laboratory conditions to confirm that GFP was expressed in all strains (Fig. S7).We infected RAW 264.7 cells with M. marinum at an MOI of 1 and measured the spread of the GFP-labeled M. marinum cells across the macrophage monolayer.Following infection and treatment to kill the extracellular bacteria, we overlaid the monolayer with 0.8% agarose to prevent the movement of the macrophages.We observed GFP signal using microscopy every 24 hours for 72 hours post infection.As shown in Fig. 4E, infection of RAW 264.7 cells with WT M. marinum resulted in detectable spread of M. marinum across the monolayer, as reflected by increasing GFP signal over time (black violin plots, P < 0.0001 between the 24 hour and 72 hour timepoints).The ΔeccCb 1 strain exhibited significantly less spread than the WT strain at 48 and 72 hours post infection (P < 0.0001), and there was no significant difference between the GFP signals generated from the ΔeccCb 1 infections over the course of the experiment (pink violin plots).The lack of spread of the ΔeccCb 1 strain was consistent with phagosomal retention, attenuated growth during infection and reduced macrophage cytolysis (Fig. 4B through D).Interestingly, although the Δemp1 strain grew to WT levels during infection (Fig. 4A), it exhibited significantly less spread than the WT strain at 48 and 72 hours post infection (P = 0.0060 and P = 0.0001), with no significant difference between the GFP signals generated over the course of the experiment.These data are consistent with the reduced macrophage cytolysis measured in Fig. 4D.Expression of emp1 in the Δemp1 strain (Δemp1/comp) restored cell-to-cell spread (purple violin plots, P < 0.0001 at 72 hours relative to the 24 hours timepoint, P = 0.0080 at 48 hours relative to the 24 hours timepoint) to levels significantly higher than the WT strain at 72 hpi (P < 0.0001).Representative images are shown in Fig. S8.These data support that emp1 is required for spread of M. marinum between macrophages.Overall, our data support that the attenuation of the Δemp1 strain occurs downstream of ESX-1 function.

DISCUSSION
In this study, we demonstrated that Emp1, a predicted NAT, is required for the N-termi nal acetylation of EsxA and other mycobacterial proteins.We demonstrated functional conservation between the M. tuberculosis and M. marinum emp1 genes.We found that Emp1 is solely responsible for the acetylation of EsxA and other mycobacterial proteins.We demonstrated that Emp1 is dispensable for ESX-1-dependent secretion and hemolysis and for growth in macrophages during infection.However, Emp1 was required for WT levels of cytoplasmic translocation or survival, macrophage cytolysis, and cell-to-cell spread of M. marinum.Collectively, this study identified a NAT required for N-terminal acetylation in Mycobacterium and provided insight into the requirement of N-terminal acetylation of EsxA and other proteins for mycobacterial virulence in the macrophage.
Although ~10%-15% of the mycobacterial proteome is likely N-terminally acetyla ted (43), little is known about the NAT enzymes responsible for N-terminal acetyla tion in Mycobacterium.Prior studies aimed at understanding N-terminal acetylation have focused on EsxA.The initial study demonstrating N-terminal acetylation of EsxA suggested that EsxA acetylation impacted the interaction with its binding partner, EsxB (30).If this were the case, we would have expected a loss of EsxA and EsxB protein in the Δemp1 strain, similar to the ΔesxA strain.Instead, EsxA and EsxB were made and secreted from M. marinum in the Δemp1 strain.Aguilera et al. mutated the Thr residue at the second position of EsxA, reporting a loss of cytoplasmic translocation by FRET assay and reduced macrophage cytolysis (31).They found no significant difference in the FRET assay between the phagosome associated ΔesxA strain and the strains with unacetylated EsxA variants, leading to the conclusion that N-terminal acetylation of EsxA was required for phagosomal lysis and macrophage cytolysis (31).In agreement with this study, abrogation of EsxA acetylation through the deletion of emp1 resulted in a significant reduction of macrophage cytolysis.However, we do not attribute the reduced cytolysis to a loss of ESX-1 function for several reasons.First, our prior work demonstrates that EsxA is required for ESX-1-dependent secretion and hemolytic activity because it is required for the secretion of ESX-1 substrates (5,67).We have reported attenuated and non-hemolytic M. marinum strains that secrete WT levels of EsxA and EsxB (5,67,69).If N-terminal acetylation was essential for EsxA function, we would have expected a loss of both protein secretion from and hemolytic activity of the Δemp1 strain.Instead, neither secretion nor hemolysis were dependent on Emp1.Second, if unacetylated EsxA was nonfunctional, we would have expected the Δemp1 strain to phenocopy the ΔeccCb 1 strain during macrophage infection.Instead, Δemp1 strain grew similarly in macrophages to the WT strain, and the Δemp1 strain accessed the cytoplasm.Our findings are consistent with the Δemp1 strain undergoing reduced bacteriolysis.Decreased bacteriolysis would cause less bacterial DNA in the cytoplasm, reducing pyroptosis, which is a cause of cytolysis (9,71,76), consistent with our other data.We suspect changing the second residue of EsxA to modulate acetylation abrogated EsxA function, causing a loss of secretion, a lack of phagosomal escape, and reduced cytolysis.In our study, the unacetylated EsxA protein retains its WT sequence and clearly promotes ESX-1-dependent secretion and virulence, supporting that N-terminal acetylation is dispensable for ESX-1 function under the conditions tested in this study, moving the field of type VII secretion forward.It remains unclear why EsxA is N-terminally acetylated, but it is formally possible that EsxA N-terminal acetylation may contribute to cytolysis downstream of phagosomal lysis.
We propose that the Emp1-dependent N-terminal acetylation of another protein or proteins is required for macrophage cytolysis, downstream of ESX-1 function.For example, Emp-1 dependent N-terminal acetylation may stabilize select mycobacterial proteins, consistent with N-terminal acetylation in higher organisms (25,44).Alterna tively, a specific Emp1 target may promote macrophage cytolysis and cell-to-cell spread.Regardless of the mechanism, Emp1 is a bacterial enzyme specifically required for macrophage cytolysis and cell-to-cell spread but not for the growth of the pathogen.Therapeutics that target Emp1 activity may be a new way to stop mycobacterial infections by limiting cell-to-cell spread of mycobacterial pathogens.
In general, M. marinum strains that are hemolytic are also cytolytic.The Δemp1 strain is hemolytic but not cytolytic.We previously published an M. marinum strain that was non-hemolytic but retained cytolytic activity, demonstrating the two activities can be separable.One reason for the disconnect between growth and cytolysis could be due to the permissive nature of RAW cells for mycobacterial growth, and this may not reflect growth in other macrophage types.
Here, we advance the field of mycobacterial physiology by identifying a NAT that promotes N-terminal acetylation, contributing to the basic understanding of this fundamental protein modification in bacteria.The framework we provide is widely accessible and applicable to any system and will facilitate the identification of NATs required for the N-terminal acetylation of specific protein targets in any organism.Importantly, we showed that an antibody against an acetylated N-terminal peptide could discriminate between acetylated and unacetylated N-termini.The generation of similar antibodies for additional N-terminally acetylated proteins could be used in any system to demonstrate N-terminal acetylation and identify the responsible NAT.We will continue to address which of the additional putative NATs contribute to N-terminal acetylation in Mycobacterium as well as their breadth of function and specificity.Furthermore, because Emp1 is required for a subset of the N-terminally acetylated mycobacterial proteins, we will focus on understanding the specificity of Emp1.
One limitation of this study is that we were unable to show the Emp1 was sufficient and necessary for the acetylation of EsxA in vitro.This would support the hypothesis that Emp1 directly acetylates EsxA at its N-terminus.We expressed the Emp1 and the Emp1W223A versions in E. coli with the goal of purification from a heterologous host.Despite trying different temperatures, additives, and vectors, we were unable to generate and isolate soluble forms of the proteins.Instead, we expressed Emp1 in E. coli and incubated the resulting lysate to acetylate a series of EsxA N-terminal peptides.Finally, we tried co-expressing emp1 and either esxA or esxBA in E. coli and measuring EsxA acetylation using western blot analysis.We were unable to observe acetylation using these approaches.We are uncertain why we are unable to produce functional Emp1 protein in vitro or in E. coli, while we can express and purify functional NATs from E.coli and Salmonella typhimurium (RimI).We suspect that Emp1 requires additional, unidentified cofactors or environmental cues for function that are specific to Mycobacte rium.
Overall, this study contributes a fundamental understanding of the conserved mechanisms and underlying N-terminal protein acetylation in pathogenic mycobacteria and identifies the NAT solely responsible for EsxA acetylation in M. marinum, opening new avenues of study aimed at further understanding this protein modification and its function in bacteria.

MATERIALS AND METHODS
M. marinum strains were grown as described previously (5).Plasmids were constructed using FAST Cloning or restriction cloning and maintained in E. coli as described (5,69,77).M. marinum strains were constructed using allelic exchange (5,69,77).N-ter minal acetylation was measured using dot blot and NUT-PAGE followed by western blot analysis.Protein production and secretion were measured using western blot analysis as previously described (5).Site directed mutagenesis of the emp1 gene was performed as in references (56,57).Whole colony MALDI mass spectrometry to measure surface-associated EsxA, Ac-EsxA, and EsxB was performed as in reference (58).LFQ mass spectrometry was used to measure N-terminal acetylation from M. marinum whole cell lysates, similar to references (70,78).Hemolytic activity of M. marinum was meas ured against sRBCs as described (5).Thin-layer chromatography was used to confirm PDIM production in the Δemp1 strain (79).iIFNAR −/− macrophages were used with a luciferase reporter to indirectly measure release of M. marinum from the phagosome and survival in the macrophage cytoplasm.RAW264.7 cells were used as an infection model to measure growth of M. marinum during infection and macrophage cytolysis as described previously (5).Spread of M. marinum across the RAW264.7 monolayer was analyzed based on these studies (16,17).Bioinformatic analysis and statistical analysis were performed using Prism and R Studio.Detailed Materials and Methods are available in the Supplementary Material.

FIG 1
FIG 1 The Ac-EsxA antibody specifically recognizes Ac-EsxA in complex mixtures.(A) The EsxA and Ac-EsxA N-terminal peptides.The "C" is not native to the EsxA protein.NT subscript = N-terminal.(B) Dot blot of EsxA N-terminal peptides.Twenty micrograms of each peptide was spotted on nitrocellulose and immunoblotted with the ⍺EsxA and ⍺Ac-EsxA antibodies.The image is representative of three independent replicates.(C) NUT-PAGE of secreted and cell-associated proteins from WT and ΔesxBA M. marinum strains.Twenty micrograms of protein was loaded in each lane.The image is representative of at least three independent biological replicates.

FIG 2
FIG 2 MMAR_1839 (Emp1) is required for the acetylation of EsxA and other proteins.(A and B) Western blot analysis of cell-associated proteins from the indicated M. marinum strains.Ten micrograms of protein was loaded in each lane.In panel B, 2 µM of L-arginine was added to the culture media.Both images are representative of at least three biological replicates.RpoB is a control for loading.(C) MS Analysis of relative abundance of acetylated and non-acetylated N-terminus of EsxA.Label-free quantitative (LFQ) proteomics intensity of the EsxA N-terminal peptide from WT, Δemp1, Δemp1/complemented strains.The normalized intensity was transformed by 10 4 to convert the Log2 values to positive integers.Propagated error was performed on technical triplicates.(D) K-means clustering of all N-terminally acetylated proteins observed from bottom-up proteomics in WT, Δemp1, Δemp1/complemented strains.Shown is the cluster that contained EsxA.(E) Functional analysis from Mycobrowser of the 22 proteins that clustered with EsxA from the k-means analysis from M. marinum.For conserved hypothetical proteins, if the closest ortholog in M. tuberculosis was annotated, that annotation was used instead.(F) ICE Logo from the protein (Continued on next page)

FIG 3
FIG 3 Emp1 is dispensable for EsxA and EsxB stability and secretion from M. marinum.(A) Western blot analysis of cell-associated and secreted proteins from M. marinum in the presence and absence of Emp1.(B) Whole colony MALDI-TOF MS.Spectra generated by whole colony MALDI-TOF for M. marinum strains are shown.The labeled peaks correspond to EsxA (9,915 m/z), Ac-EsxA (9,957 m/z), and EsxB(10,606 m/z Da), respectively.The dotted line was added for clarity.(C) Hemolytic activity of M. marinum strains.The data shown include seven biological replicates each in technical triplicate.Each data point is a technical replicate.Statistical analysis was performed using an ordinary one-way analysis of variance (ANOVA) (P < 0.0001) followed by a Tukey's multiple comparison test.The significance shown is compared to the WT strain, ****P < 0.0001.Other important comparisons are discussed in the text.

FIG 4
FIG 4 Emp1 is required for cytolytic activity and cell-to-cell spread of M. marinum.(A) CFU analysis of M. marinum strains.MOI = 0.2 plated in triplicate, represents three biological replicates.Significance was determined using a two-way RM ANOVA (P < 0.0001) followed by a Tukey's multiple comparison test.The significance shown is compared to the WT strain, for the 96 hpi.However, the CFUs from the ΔeccCb 1 strain were significantly lower than the WT strain throughout the experiment.P values comparing ΔeccCb 1 vs WT were as follows: 0 hpi **P = 0.0035, 24 hpi, *P = 0.0162, 48 hpi ***P = 0.0001, and 72 hpi and 96 hpi ****P < 0.0001.The WT strain was not significantly different from the additional strains at any time point.(B) Schematic of progression through macrophage infection."Panels" refer to this figure.(C) Cytoplasmic bacteriolysis of M. marinum strains in iIFNAR −/− immortalized murine bone marrow-derived macrophages, at 20 and 24 hpi, MOI = 20.The data includes four biological replicates, each performed in three technical replicates.Outliers were identified and removed using ROUT analysis (Q = 0.5%).Cleaned data were tested for significance using a one-way ordinary ANOVA (P < 0.0001) followed by a Tukey's multiple comparison test.*P = 0.0360, ***P = 0.0002, and ****P < 0.0001.(D) Cytolytic activity of M.

TABLE 1
The top five conserved putative NATs between M. marinum and M.

tuberculosis a M. marinum gene M. tuberculosis gene % identity (protein)
a Sequences were obtained from Mycobrowser.The percent protein identity was determined using Protein BLAST.