EspM Is a Conserved Transcription Factor That Regulates Gene Expression in Response to the ESX-1 System.

Mycobacterial pathogens use the ESX-1 system to transport protein substrates that mediate essential interactions with the host during infection. We previously demonstrated that in addition to transporting proteins, the ESX-1 secretion system regulates gene expression. Here, we identify a conserved transcription factor that regulates gene expression in response to the ESX-1 system. We demonstrate that this transcription factor is functionally conserved in M. marinum, a pathogen of ectothermic animals; M. tuberculosis, the human-pathogenic species that causes tuberculosis; and M. smegmatis, a nonpathogenic mycobacterial species. These findings provide the first mechanistic insight into how the ESX-1 system elicits a transcriptional response, a function of this protein transport system that was previously unknown.


RESULTS
The EspM protein binds upstream of the whiB6 gene. To identify transcription factors that regulate genes in response to the ESX-1 membrane complex, we focused on the regulation of the whiB6 gene. The 1 kb of DNA upstream of the whiB6 gene is sufficient for regulation of whiB6 gene expression by the ESX-1 membrane complex (35). We used a DNA pulldown to enrich proteins from M. marinum lysate that specifically bind the 1 kb of DNA upstream of the whiB6 gene ("whiB6 promoter bait," Fig. 1A; bp 6577326 to 6578305 in the M. marinum genome). Using liquid chromatograph-tandem mass spectrometry (LC-MS/MS)-based quantitative proteomics on the proteins eluted from the DNA, we identified several proteins that were specifically and reproducibly enriched for binding the whiB6 promoter bait relative to binding nonspecific DNA (rpoA bait; see Table S1 in the supplemental material). MMAR_5438 was enriched for binding the whiB6 promoter bait Ն64.0-fold Ϯ 0.4-fold relative to the rpoA bait (Fig. 1B). We propose renaming the MMAR_5438 gene "espM," consistent with current ESX-1 nomenclature (40). We generated an M. marinum strain with an unmarked deletion of the espM gene (ΔespM; Fig. S1) and a complementation strain with an integrated constitutive espM expression plasmid (ΔespM/p msp espM). The EspM protein was not identified in the DNA pulldown performed with lysate from the ΔespM strain and was further enriched for whiB6 promoter bait binding compared to the rpoA bait in lysates from the complemented strain (Fig. 1B). We also identified several M. marinum proteins with known DNA binding activity that were not significantly or reproducibly enriched for binding the whiB6 promoter bait relative to the rpoA bait (Table S1). For example, the M. marinum DNA-binding protein Hu homolog HupB (MMAR_1728) bound the two baits comparably following incubation with any M. marinum lysate (Fig. 1B).
To confirm the interaction of the EspM protein with the whiB6 promoter region, we expressed and purified an N-terminally 6ϫHis-tagged EspM MM fusion protein from Escherichia coli (the MM subscript refers to the protein from M. marinum [40]) (see Fig. S2 in the supplemental material) and performed electrophoretic mobility shift assays (EMSAs). We observed a specific shift in mobility of the whiB6 promoter probe (550 bp) (Fig. 1A, "EMSA probe") and a concomitant loss of free whiB6 promoter probe with increasing concentrations of the 6ϫHis-EspM MM protein (Fig. 1C). We did not observe a mobility shift of the rpoA probe, confirming the specific binding of the EspM MM protein to the whiB6 promoter probe.
The espM gene is conserved in M. tuberculosis. The EspM proteins in M. marinum and M. tuberculosis Erdman (ERDMAN_4236, EspM MT ) are 76.25% identical at the amino acid level (41,42). To test if EspM binds the genomic region upstream of the whiB6 MT gene, we expressed and purified 6ϫHis-tagged EspM MT in E. coli (the MT subscript refers to the EspM protein from M. tuberculosis) (Fig. S2). We amplified the 500 bp upstream of the whiB6 gene from M. tuberculosis Erdman and tested if EspM MT specifically bound the whiB6 MT promoter region using EMSAs. Increasing concentrations of 6ϫHis-EspM MT protein led to a specific mobility shift of the whiB6 MT promoter probe and to a corresponding loss of free probe (Fig. 1D). Although bound rpoA probe was not observed at the highest concentrations of 6ϫHis-EspM MT protein, the free probe was reduced, indicating weak binding at the highest protein concentrations. Together, these data indicate that EspM, from both M. marinum and M. tuberculosis, directly and specifically bound the whiB6-espM intergenic region.
espM is divergently transcribed from the whiB6 gene and is immediately adjacent to the esx-1 locus (Fig. 1A). EspM is a predicted conserved regulatory protein (42), but the corresponding function has not been investigated. The EspM MM protein is predicted to have an N-terminal forkhead-associated (FHA) domain (amino acids [aa] 32 to 89) and a C-terminal helix-turn-helix domain (Fig. 1E). We hypothesized that the C-terminal half of the protein mediated DNA binding. We expressed and purified 6ϫHis-tagged EspM NT (aa 1 to 133) and EspM CT (aa 127 to 363) M. marinum proteins from E. coli (Fig. S2). We tested the ability of each protein to bind the whiB6 promoter probe using EMSA. The 6ϫHis-EspM NT protein did not shift the mobility of the whiB6 promoter probe (Fig. 1F). Incubation with increasing concentrations of the 6ϫHis-EspM CT protein caused a shift in mobility of the whiB6 promoter probe and a loss of free probe. We conclude that the C-terminal half of the EspM protein is required for DNA binding.
EspM is a conserved regulator of whiB6 and esx-1 gene expression. We confirmed that the espM transcript was absent in the ΔespM M. marinum strain using quantitative reverse transcription-PCR (qRT-PCR) ( Fig. 2A). The espM expression level was significantly higher in the ΔespM/p msp espM complemented strain than in the wild-type (WT) strain (P Ͻ 0.0001). These data indicate that the complementation strain is an espM overexpression strain. We did not observe a significant reduction of espM gene expression in the ΔeccCb 1 strain relative to the WT strain. These data confirm that espM expression is not regulated by the ESX-1 system in M. marinum, consistent with our previously published transcriptomic analysis (35).
Because EspM bound the region upstream of the whiB6 gene, we tested if EspM regulates whiB6 gene expression. We measured whiB6 gene expression in M. marinum using qRT-PCR. Consistent with our prior findings (35), whiB6 gene expression was significantly reduced in the ΔeccCb 1 strain compared to the WT strain (Fig. 2B, inset, P Ͻ 0.0001). Deletion of the espM gene resulted in a significant increase in whiB6 expression relative to the WT strain (P Ͻ 0.0001). Overexpression of the espM gene resulted in espM expression that was not significantly different from that seen with the WT strain. We conclude that EspM is a repressor of whiB6 gene expression.
We measured the levels of WhiB6 protein in the presence and absence of the espM gene (Fig. 2C). The parental M. marinum strain for these strains includes a whiB6 gene with a C-terminal FLAG epitope tag (WhiB6Fl [35]). Consistent with our previously published data (35), the WhiB6Fl protein was absent from the lysate generated from the ΔeccCb 1 strain (Fig. 2C, lane 2). Consistent with the expression data (Fig. 2B), deletion of the espM gene resulted in increased WhiB6Fl protein levels relative to those seen with the WT strain (Fig. 2C, compare lane 3 to lane 1). The WhiB6Fl protein levels in the ΔespM/p msp espM MM complemented strain (espM overexpression) were lower than those FIG 2 EspM is a conserved regulator of whiB6 and esx-1 gene expression. (A) qRT-PCR measuring the levels of espM expression relative to sigA expression. A one-way ordinary analysis of variance (ANOVA) (P Ͻ 0.0001), followed by a Dunnett's multiple-comparison test relative to the WT strain, was performed. ****, P Ͻ 0.0001. (B) qRT-PCR measuring the levels of whiB6 expression relative to sigA expression. A one-way ordinary ANOVA (P Ͻ 0.0001), followed by a Sidak's multiple-comparison test relative to the WT strain, was performed. ****, P Ͻ 0.0001. The inset shows just the comparison between the WT and ΔeccCb1 strains. A Student's unpaired, two-tailed t test was used to define the significance of the results of the comparisons between the two strains. For panels A and B, the data represent averages of results from at least three biological replicates, each performed in technical triplicate. (C) Western blot analysis of 10 g of protein per lane. Anti-Mpt32 was used as the loading control. All M. marinum strains indicated in this panel contained a C-terminal FLAG epitope tag on the whiB6 gene. Samples were resolved on an 18% Tris-glycine gel. The Western blot shown is representative of at least three independent biological replicates. All strains indicated in panel D contained a C-terminal epitope tag on the whiB6 gene.  Table S3. (F) Heat map of esx-1 locus genes that are significantly differentially regulated in the ΔespM strain versus the ΔespM/p msp espM complemented strain compared to genes expressed in the ΔespM, ΔwhiB6, or eccCb 1 mutant strains relative to the WT strain. in the WT strain (Fig. 2C, lane 4 versus lane 1). Together, these data strongly support the conclusion that EspM represses whiB6 gene expression in M. marinum.
The espM gene is syntenic in M. marinum, M. tuberculosis, and M. smegmatis (Fig. S3). M. smegmatis is a nonpathogenic, soil-dwelling mycobacterial species that uses the ESX-1 system to mediate conjugation (43,44). The EspM orthologs in all three species are conserved at the protein level ( Fig. 2D; see also We performed RNA-seq transcriptional profiling to determine if EspM regulates other genes in addition to whiB6 in M. marinum. Comparison of the WT strain to the ⌬espM strain (both bearing the whiB6Fl allele) identified 134 genes that were upregulated and 300 genes that were downregulated (Ͼ2-fold; false-discovery rate [q value] of Ͻ0.05) ( Fig. S4A; see also Table S3A). Genes controlled by EspM are also expected to be differentially regulated in the ΔespM strain compared to the complemented strain that overexpresses the repressor. We observed 44 genes that were upregulated and 55 genes that were downregulated in the ΔespM strain compared to the complemented strain (Ͼ2-fold; q value of Ͻ0.05) ( Fig. 2E; see also Table S3B). Consistent with repression of whiB6 expression by EspM, we observed that whiB6 expression was induced 1.6-fold and 7.0-fold in the ΔespM strain compared to the WT strain and the complemented overexpression strain, respectively. Of the 44 genes that were induced in the ΔespM strain compared to the complemented strain, 21 genes from the esx-1 locus were identified (MMAR_5436 to MMAR_5457), including 8 genes that were also induced in the ΔespM strain compared to WT strain ( Fig. 2F; see also Fig. S4B). Most of the other genes in the esx-1 locus were significantly induced in the ΔespM strain relative to the WT strain, but with induction levels below 2-fold.
We also observed induction of unlinked esx-associated loci in the ΔespM strain compared to the complemented strain, including MMAR_0187-188 (esxB_1esxA_1), MMAR_3654 (esxP2), and the ESX-1 substrate locus MMAR_2894 (45) ( Table S3B). Several of these genes were previously shown to be regulated by WhiB6 or ESX-1 ( Of the 55 genes downregulated in the ΔespM strain relative to the complemented strain, 39 were also downregulated in the ΔespM strain relative to WT strain (Table S3). These included 24 strongly downregulated genes between MMAR_0159 and MMAR_ 0182 ( Fig. 2E; see also Table S3B), which includes the mce6 locus (Fig. S4C), and genes for amino acid metabolism and lipid anabolism. Prior studies with the ΔwhiB6 and eccCb 1 mutant strains showed induction of the genes in the mce6 locus (Fig. S4C), supporting the idea of ESX-1-dependent regulation. Curiously, we also detected downregulation of an ESX-1-associated operon, MMAR_4166 to MMAR_4168 (espA, espC, and espD). Together, these data strongly support the conclusion that EspM is a regulator of genes broadly associated with the ESX-1 system in M. marinum.
We reasoned that if EspM repressed whiB6 gene expression in the absence of the ESX-1 membrane complex, then deletion of the espM gene in strains lacking the ESX-1 membrane complex would restore whiB6 gene expression. We generated M. marinum strains bearing deletions of each ecc gene (eccA to eccE 1 ) alone or in combination with deletion of the espM gene. Deletion of any ecc gene resulted in a loss of WhiB6Fl protein relative to the WT strain (Fig. 3B, lanes 2 to 7 versus lane 1). The deletion of the espM gene in combination with the ecc genes (ΔespM ΔeccA, ΔespM ΔeccB 1 , ΔespM ΔeccCa 1 , ΔespM ΔeccCb 1 , ΔespM ΔeccD 1 , and ΔespM ΔeccE 1 mutant strains) resulted in levels of WhiB6Fl similar to those in the ΔespM strain (Fig. 3B, lanes 9 to 14 versus lane 8) and higher than those in the WT strain (Fig. 3B, lane 1). We further demonstrated that complementation with the eccA gene or the espM gene in the ΔespM ΔeccA strain resulted in levels of WhiB6Fl similar to those seen with the ΔespM deletion strain or the ΔeccA deletion strain, respectively (Fig. S5).
We reasoned that overexpression of the espM gene might be sufficient to repress whiB6 expression to levels similar to those seen with the ΔeccCb 1 strain. We generated a strain bearing a whiB6 transcriptional reporter in M. marinum. We replaced the whiB6 gene with the lacZ gene, creating a strain lacking the whiB6 gene and with a reporter fusion to the whiB6 promoter (ΔwhiB6::lacZ ϩ ; Fig. 3C). We generated an isogenic ΔeccCb 1 strain (no ESX-1 membrane complex [30]), as well as an isogenic WT strain overexpressing the espM MM gene. The level of ␤-galactosidase activity was significantly reduced in the ΔeccCb 1 strain compared to the WT strain ( Fig. 3D; P Ͻ 0.0001), confirming that the whiB6::lacZ ϩ reporter fusion was regulated by the ESX-1 membrane complex (35). Overexpression of the espM gene in the WT strain significantly reduced the levels of ␤-galactosidase activity compared to the WT strain levels (P Ͻ 0.0001). The levels of ␤-galactosidase activity in the ΔeccCb 1 and espM overexpression strains were not significantly different from each other (P ϭ 0.9915), demonstrating that overexpression of espM is sufficient to repress whiB6 gene expression in M. marinum. Collectively, our data demonstrate that EspM is required for repression of whiB6 gene expression in the absence of the ESX-1 membrane complex. Moreover, because the reporter strain lacks the whiB6 gene, these data indicate that EspM represses whiB6 expression in a WhiB6-independent manner.
The EspM and WhiB6 regulators coordinately control gene expression. The espM and whiB6 genes are divergently organized in mycobacterial genomes (Fig. 4A). Because the whiB6 and espM genes share an intergenic region which likely controls the expression of both genes (Fig. 4A, pink), we sought to further define the relationship between the EspM and WhiB6 regulators.
We generated espM-lacZ ϩ and whiB6-lacZ ϩ integrating transcriptional reporters (Fig. 4). The espM-lacZ ϩ reporter resulted in significantly increased ␤-galactosidase activity in the ΔespM strain compared to the WT strain (P Ͻ 0.0001; Fig. 4A). Loss of the whiB6 gene did not significantly impact ␤-galactosidase activity relative to the WT strain (P ϭ 0.1195). Deletion of both the espM and whiB6 genes (ΔespM ΔwhiB6 mutant strain) resulted in ␤-galactosidase activity comparable to that seen with the WT M. marinum strain (P ϭ 0.9305). We conclude from these data that espM expression is negatively autoregulated. Moreover, in the absence of EspM, WhiB6 is required for the observed increased espM gene expression. We confirmed that WhiB6 binds the whiB6-espM intergenic region by expressing and purifying a C-terminally 6ϫHis-tagged WhiB6 MM fusion protein from Escherichia coli (Fig. S2) and performing EMSAs (Fig. S6). We observed a specific shift in mobility of the whiB6 promoter probe (Fig. 1A, "EMSA probe") and a concomitant loss of free whiB6 promoter probe with increasing concentrations of the WhiB6 MM -6ϫHis protein (Fig. S6). We did not observe a mobility shift of the rpoA probe, confirming the specific binding of the WhiB6 MM protein to the whiB6 promoter probe.
The presence of the whiB6-lacZ ϩ reporter resulted in significantly increased ␤-galactosidase activity in the ΔespM strain compared to the WT strain (P Ͻ 0.0001; A one-way ordinary ANOVA (P Ͻ 0.0001) followed by a Tukey's multiple-comparison test was performed. Significance data shown are relative to the WT strain. ****, P Ͻ 0.0001; **, P ϭ 0.0078. For both panels, the data represent averages of results from at least three biological replicates, each performed in technical triplicate. All strains indicated in Fig. 4, with the exception of the ΔwhiB6 and ΔwhiB6 ΔespM strains, contained a C-terminal epitope tag on the whiB6 gene. Fig. 4B). Loss of the whiB6 gene caused a significant reduction in ␤-galactosidase activity relative to the WT strain (P ϭ 0.0078). Deletion of the espM and whiB6 genes together (ΔespM ΔwhiB6 mutant strain) resulted in significantly increased ␤-galactosidase activity relative to the WT and ΔespM M. marinum strains (P Ͻ 0.0001 for both comparisons). This further supports the idea that EspM represses whiB6 gene expression and confirms positive autoregulation of WhiB6, consistent with prior findings (38). Moreover, the significant increase in whiB6 expression in the absence of both EspM and WhiB6 suggests there is at least one more transcriptional activator of whiB6 expression in M. marinum.
EspM fine-tunes ESX-1 function in M. marinum. Because EspM regulates whiB6 and esx-1 gene expression, we tested if EspM was required for ESX-1 activity. The WT strain produced the EsxA and EsxB substrates and secreted them into the culture supernatant during growth in vitro (Fig. 5A, lanes 1 and 2). Deletion of the eccCb 1 gene, which is required for ESX-1 secretion (10,12,14), reduced production of EsxA and EsxB, and neither protein was secreted (Fig. 5A, lanes 3 and 4). The ΔespM strain exhibited at least WT levels of production and secretion of EsxA and EsxB (Fig. 5A, lanes 5 and 6). The espM complemented strains showed reduced levels of production of EsxA and EsxB (Fig. 5A, lanes 7 and 8, and Fig. S4D) but exhibited at least wild-type levels of secretion of EsxA and EsxB. To further confirm that the levels of ESX-1 proteins were altered, consistent with the observed EspM-dependent expression changes, we performed global proteomics on wholecell lysates of the M. marinum strains represented in panel A (for the WT, ΔeccCb 1 , ΔespM, and complemented strains) (Table S1E and F; see also Fig. S4E). We identified 1,881 proteins at a 1% false-discovery rate. Protein quantification was performed by using label-free quantification (LFQ). We found that, similarly to the EsxA and EsxB proteins, the levels of several ESX-1 substrates (EspF, EspK, and EspB) and components (EccA) and other associated proteins (EspG, EspH, and EspL) were significantly reduced in the complemented strain, consistent with the expression data (Fig. 2). These data demonstrate that EspM is required for fine-tuning the levels of ESX-1-associated proteins, including the EsxA and EsxB substrates, in the mycobacterial cells but not for the secretory function of the ESX-1 system.
The ESX-1 system promotes phagosomal lysis during macrophage infection (6,9,27). M. marinum lyses red blood cells (RBCs) in an ESX-1-dependent manner (14,17,50). Hemolysis analysis is a common way to measure the membranolytic activity of the ESX-1 system (14,17,50). The WT strain caused significantly increased hemolytic activity compared to the phosphate-buffered saline (PBS) control (no bacteria) (P Ͻ 0.0001; Fig. 5B). The ΔeccCb 1 strain exhibited hemolytic activity that was not significantly different from that seen with the PBS control (P ϭ 0.9996). The ΔespM strain exhibited hemolytic activity not significantly different from that seen with the WT strain (P ϭ 0.9602). The complemented strain, which overexpresses espM relative to the WT strain (Fig. 2B), showed significantly reduced hemolytic activity relative to the WT and ΔespM strains (P Ͻ 0.0001).
ESX-1-deficient M. marinum strains fail to lyse the phagosome and fail to lyse macrophages (7,51). We infected RAW 264.7 cells with M. marinum at a multiplicity of infection (MOI) of 7 and measured macrophage lysis by visualizing and quantifying the uptake of ethidium homodimer by permeabilized macrophages (52). Consistent with previous findings (51,53), the WT strain caused macrophage lysis (Fig. 5C). Infection with the ΔeccCb 1 strain resulted in a significant reduction in macrophage lysis compared to infection with the WT strain (P Ͻ 0.0001). In contrast, infection with the ΔespM strain resulted in significantly increased macrophage lysis compared to infection with the WT strain (P Ͻ 0.0001). Infection with the espM overexpression strain restored macrophage lysis to WT levels (P ϭ 0.2138). These data show that in the absence of the espM gene, the ESX-1 system promoted higher levels of macrophage lysis. Moreover, combined with the hemolysis data, these findings indicate that the levels of EspM fine-tune the activity of the ESX-1 system in M. marinum.

DISCUSSION
Collectively, our findings identify EspM as a conserved transcription factor required for the ESX-1-dependent transcriptional response in pathogenic mycobacteria. Although the espM gene is adjacent to the esx-1 locus, EspM has not been previously characterized. The espM gene may not have been linked to the ESX-1 system previously because deletion of the espM gene in M. marinum had only a subtle impact on ESX-1 activity, likely because several transcription factors regulate whiB6 gene expression (39,(54)(55)(56). Moreover, the M. tuberculosis EspM ortholog Rv3863 was previously reported to Black bars indicate median and quartiles. UI, uninfected. Statistical analysis was performed using a one-way ordinary ANOVA (P Ͻ 0.0001) followed by a Tukey's multiple-comparison test. ****, P Ͻ 0.0001 (compared to the WT strain). Each dot represents the number of EthD-1-stained cells in a single field. A total of 10 fields were counted using ImageJ for each well. Processing of 3 wells was performed for each biological replicate. A total of 90 fields were counted for each strain.
be essential for growth in vitro in some genome-wide studies (57,58) and nonessential in others (59), which may complicate study in M. tuberculosis.
The identification of EspM further expands our understanding of the feedback control mechanism that links the levels of ESX-1 substrates, and other genes, to the assembly of the secretory apparatus (35). We found that deletion of the espM gene resulted in levels of whiB6 expression that were higher than those seen with the WT strain ( Fig. 2B and C). We propose that whiB6 expression is repressed by EspM in the WT strain and that whiB6 gene expression is further repressed by EspM in the absence of the ESX-1 system (Fig. 6). Therefore, regulation by EspM is relevant in WT bacteria and not simply when the ESX-1 system is absent, which may or may not be physiologically relevant.
We do not yet understand how the ESX-1 membrane complex controls the magnitude of whiB6 repression by EspM. We do not think that the ESX-1 system transcriptionally regulates the espM gene. We observed no ESX-1-dependent change in espM transcript levels either here ( Fig. 2A) or in our prior work (35). These findings contrast those of Abdallah et al. (36), which indicated that the Rv3863 transcript (espM MT ) is regulated by the ESX-1 system in M. tuberculosis, similarly to the whiB6 gene. This may be an example of differential regulation between M. marinum and M. tuberculosis. The presence or absence of an assembled ESX-1 membrane complex likely posttranscriptionally controls the levels of EspM in M. marinum. We recapitulated the levels of whiB6 gene expression in the ΔeccCb 1 strain by overexpressing espM in the WT strain (Fig. 3D). However, our published proteomic data indicate that EspM MM protein levels are reduced 2-fold in the absence of EccCb 1 , when repression of whiB6 expression is strongest (35). Regulation of the EspM transcription factor may be similar to the control of gene expression by type III secretion systems (T3SS) in Gram-negative bacteria (60)(61)(62)(63). The injectisome T3SS uses cytoplasmic substrates and/or chaperones to posttranscriptionally modulate the levels or activity of transcription factors that regulate secretion-associated genes (62,(64)(65)(66)(67)(68)(69). ESX-1 substrates or chaperones may posttranscriptionally regulate the activity of EspM in response to the presence or absence of the ESX-1 membrane complex.
Posttranscriptional regulation of EspM could occur through the predicted N-terminal forkhead-associated (FHA) domain. FHA domain-containing proteins posttranscriptionally regulate Gram-negative type III and type VI protein secretion systems (70)(71)(72). Staphylococcus aureus has an Ess-type VII secretion system similar to the ESX-1 Regulation of Gene Expression by the ESX-1 System ® system (73,74). The EccC-related protein EssC SA (75,76) includes a twin-FHA domain that is essential for secretion (77). FHA domains also mediate oligomerization (78)(79)(80). We observed a second shift in mobility of the whiB6-espM probe by EMSA with increasing concentrations of EspM MM protein ( Fig. 1C and F). We did not observe this supershifted product when using the C-terminal half of EspM MM alone (Fig. 1F), suggesting that the N-terminal half of the protein is important for this observation. The FHA domain may directly or indirectly control oligomerization of EspM in response to the ESX-1 membrane complex.
Although WhiB6 directly binds the whiB6-espM promoter region, we did not identify WhiB6 in the DNA pulldown ( Fig. 1; see also Table S1 in the supplemental material). We have not routinely identified WhiB6 from M. marinum lysates using mass spectrometry. We also did not identify the PhoP response regulator, which regulates whiB6 expression in M. tuberculosis. Under the conditions of our experiments, EspM may bind the intergenic region preferentially to other regulators, including WhiB6 and PhoP. This idea is supported by the finding that WhiB6 activates espM gene expression only in the absence of EspM (Fig. 4A). Also, it is possible that no single technique can identify all proteins that bind and regulate a specific region. For example, chromatin immunoprecipitation sequencing (ChIP-seq) experiments in strains overexpressing WhiB6 in M. tuberculosis did not identify direct binding of WhiB6 to the whiB6-espM intergenic region. And yet, overexpression of WhiB6 resulted in a significant upregulation of whiB6 gene expression in the same study (55,56). Likewise, although PhoP bound the WhiB6 promoter directly in M. tuberculosis, overexpression of PhoP failed to significantly impact whiB6 gene expression (55,56). Therefore, the absence of enrichment of regulators in our study does not preclude the possibility of a role for them in the regulation of whiB6 and espM expression. Finally, regulation of the whiB6 and espM genes may not be conserved between M. marinum and M. tuberculosis. In the case of whiB6 expression, it has already been established that there is variability in how PhoP regulates whiB6 in M. tuberculosis strains (39). It has not yet been established if whiB6 regulates esx-1 in M. marinum as part of the PhoPR regulon.
The back-to-back divergent arrangement of two regulators is a common theme in microorganisms (81), the best described of which are the cI and Cro regulators of bacteriophage (82). Divergence in organization allows tight coordination of the expression of both transcription factors and of their regulons from a single genetic locus. The intergenic region between the espM and whiB6 genes likely contains binding sites for both WhiB6 and EspM. Indeed, we demonstrated using EMSAs that both EspM and WhiB6 bind this region in vitro ( Fig. 1; see also Fig. S6 in the supplemental material) and that both contribute to regulating the whiB6 and espM genes [ Fig. 4]) (38,39). Therefore, the genes regulated by WhiB6, the ESX-1 system, and EspM may be coordinated to fine-tune the ESX-1 secretion and for additional biological purposes. Moreover, whereas approximately half of the genes induced or repressed in the ΔespM strain versus the complemented strain are associated with the ESX-1 system, other EspMregulated genes may have a currently unrecognized role in ESX-1-associated functions.
Our data clearly demonstrate that EspM impacts the expression of esx-1-associated genes and is associated with corresponding changes in ESX-1 protein levels, supporting the idea that EspM functions to fine-tune ESX-1 function. Consistent with these findings, we observed reduced hemolytic activity upon overexpression of espM and increased cytolytic activity in the absence of EspM. Although hemolysis and macrophage cytolysis are both measures of ESX-1 function, our prior work indicated that the results of the two assays do not always align, especially when using strains with intermediate ESX-1 production or secretion levels (45). It is possible that there are additional roles for EspM in ex vivo infection that differ from those seen in our studies in vitro. Alternatively, EspM could impact the expression of additional genes required for phagosomal lysis or macrophage cytolysis. For example, phthiocerol dimycocerosate (PDIM) has been implicated in both phagosomal lysis and macrophage cytolysis (83,84). However, we did not see changes in the expression of genes required for PDIM synthesis and transport in our RNA sequencing analysis (Table S3) or in the production of PDIM (Fig. S4F).
Unlike most examples of T3SS-dependent gene expression, the genes regulated by EspM and the ESX-1 membrane complex are not restricted to the ESX-1-associated genes (35,36). The C. trachomatis T3SS, which impacts global gene expression, may represent a temporal cue for regulating gene expression during infection (85). Likewise, the assembly of the ESX-1 system may serve as a temporal cue to regulate mycobacterial gene expression. While pathogenic mycobacteria elicit a transcriptional response essential for survival in the phagosome (1,4), there has been no report of a transcriptional response to interaction with the macrophage cytosol. The cytoplasm is considered restrictive for bacterial survival and growth unless the pathogen adapts (86). Listeria monocytogenes, a pathogen that lyses the phagosomal membrane and accesses the cytoplasm (87), adapts by altering metabolism and inducing stress response pathways (86). We propose that the assembly of the ESX-1 membrane complex elicits gene expression pathways to link ESX-1-mediated phagosomal lysis and cytoplasmic adaptation. Indeed, several of the genes regulated by EspM and by the ESX-1 system are predicted to be associated with metabolism (Table S3). This is most notable in genes that are downregulated in the ΔespM strain or upregulated in the ΔwhiB6 and eccCb 1 mutant strains. For example, the genes in the mce6 locus and surrounding genes were significantly downregulated in the ΔespM strain but were upregulated in the ΔwhiB6 and eccCb 1 mutant strains (Fig. S4C), although, due to variability in the data, the results representing the gene induction in the ΔwhiB6 and eccCb 1 mutant strains were not statistically significant. These data are supportive of the conclusion that the mce6 genes are repressed in a whiB6-dependent manner, although further characterization studies will be required to support this hypothesis. mce genes have been associated with carbon nutrient uptake, including mce1, promoting uptake of fatty acids (88), and mce4, promoting uptake of cholesterol (89). mce6 is absent in the M. tuberculosis genome but is present in the genomes of many nontuberculous mycobacterial species (90) and could play a role in controlling metabolite import to promote survival in the phagosome or cytosol. The mce6 locus may be important for the cytosolic lifestyle of M. marinum, which polymerizes host actin and exhibits cytosolic motility (5), which is not conserved in M. tuberculosis.
Finally, because EspM regulates a subset of genes controlled by the ESX-1 system, there are likely additional transcription factors that make up an ESX-1-dependent transcriptional network. We focused on proteins that specifically bound the whiB6/espM intergenic region. Studies aimed at identifying proteins that bind additional ESX-1responsive promoters would identify additional transcription factors in the ESX-1responsive network.
In conclusion, we have identified a conserved transcription factor, EspM, which is encoded by a gene adjacent to the esx-1 locus that is required for the repression of whiB6 gene expression in the absence of the ESX-1 system. Our study results begin to define a transcriptional network that links the assembly of the ESX-1 system to widespread changes in gene expression, including the regulation of the ESX-1 apparatus and substrates.

MATERIALS AND METHODS
A fully detailed explanation of the methods used in this study can be found in Text S1 in the supplemental material. All M. marinum strains were derived from M. marinum strain M (BAA-535). Where indicated, the parental strain included a FLAG epitope tag at the C terminus of the whiB6 gene (35). Maintenance of the M. marinum strains and E. coli strains is described in Text S1. Enriched proteins were analyzed using quantitative nano-high-performance liquid chromatography-tandem mass spectrometry (nano-UHPLC-MS/MS) proteomics. All mycobacterial strains were generated using the allelic exchange protocol developed by Parish and Stoker (91) as described previously (35,45,52,92). All strains, constructs, and primers (IDT, Coralville, IA) used in this study are listed in Table S2 in the supplemental material. All plasmids and genetic deletions were confirmed by targeted DNA sequencing performed at the Notre Dame Genomics and Bioinformatics Facility. All proteins were expressed in E. coli with 6ϫHis affinity tags and purified using metal chelation affinity chromatography as described in Text S1. EMSAs were performed as reported previously (93)(94)(95), with modifications listed in Text S1. ␤-Galactosidase assays on M. marinum strains bearing the whiB6::lacZ ϩ , attB::pwhiB6-lacZ ϩ , or attB::pespM-lacZ ϩ reporter were performed as described previously (52). Hemolysis assays were performed as described previously (35). ESX-1 secretion assays were performed as described previously (35), except that 10 g of protein was analyzed for all protein fractions. Western blot analysis was performed as described previously (35). Macrophage (RAW 264.7 cells) infections were performed as described previously (45) at an estimated multiplicity of infection (MOI) of 7 (2.5 ϫ 10 6 cells/ml). Cells were imaged and ethidium-homodimer uptake by perforated cells was quantified using ImageJ (35,52). For transcriptional profiling, M. marinum strains were grown and RNA was extracted exactly as described previously (35). RNA sequencing was conducted as described previously (96), and the results were analyzed using SPARTA software (97). For analysis of differentially expressed genes (Ͼ2-fold; q value of Ͻ0.05), lists were filtered for genes with average counts greater than 4 (log 2 CPM), with full unfiltered data sets available in Table S3.
Data availability. The transcriptional profiling data are available at the NCBI GEO database (accession number GSE135072). All statistical analysis was performed as described in each figure legend, using PRISM v8.1.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, PDF file, 0.2 MB.