Functional Analysis of EspM, an ESX-1-Associated Transcription Factor in Mycobacterium marinum

ABSTRACT

Pathogenic mycobacteria use the ESX-1 secretion system to escape the macrophage phagosome and survive infection. We demonstrated that the ESX-1 system is regulated by feedback control in Mycobacterium marinum, a nontuberculous pathogen and model for the human pathogen Mycobacterium tuberculosis. In the presence of a functional ESX-1 system, the WhiB6 transcription factor upregulates expression of ESX-1 substrate genes. In the absence of an assembled ESX-1 system, the conserved transcription factor, EspM, represses whiB6 expression by specifically binding the whiB6 promoter. Together, WhiB6 and EspM fine-tune the levels of ESX-1 substrates in response to the secretion system. The mechanisms underlying control of the ESX-1 system by EspM are unknown. Here, we conduct a structure and function analysis to investigate how EspM is regulated. Using biochemical approaches, we measured the formation of higher-order oligomers of EspM in vitro. We demonstrate that multimerization in vitro can be mediated through multiple domains of the EspM protein. Using a bacterial monohybrid system, we showed that EspM self-associates through multiple domains in Escherichia coli. Using this system, we performed a genetic screen to identify EspM variants that failed to self-associate. The screen yielded four EspM variants of interest, which we tested for activity in M. marinum. Our study revealed that the two helix-turn-helix domains are functionally distinct. Moreover, the helix bundle domain is required for wild-type multimerization in vitro. Our data support models where EspM monomers or hexamers contribute to the regulation of whiB6 expression.

IMPORTANCE Pathogenic mycobacteria are bacteria that pose a large burden to human health globally. The ESX-1 secretion system is required for pathogenic mycobacteria to survive within and interact with the host. Proper function of the ESX-1 secretion system is achieved by tightly controlling the expression of secreted virulence factors, in part through transcriptional regulation. Here, we characterize the conserved transcription factor EspM, which regulates the expression of ESX-1 virulence factors. We define domains required for EspM to form multimers and bind DNA. These findings provide an initial characterization an ESX-1 transcription factor and provide insights into its mechanism of action.

INTRODUCTION

Pathogenic mycobacteria cause chronic and acute disease in humans and in animals. Mycobacterium tuberculosis causes human tuberculosis (1). Environmental nontubercular mycobacterial (NTM) species cause disease in humans and animals (2–4). Mycobacterium marinum is an NTM that causes tuberculosis-like infections in poikilothermic fish, and occasionally causes infections in humans (5–7). M. tuberculosis and M. marinum use conserved molecular mechanisms to interact with their hosts (8). Both pathogens are taken up by host phagocytic cells and are initially retained in the phagosome (8–11). M. marinum and M. tuberculosis alter their gene expression and secrete virulence factors to survive within the phagosome (12, 13). Both species then damage the phagosomal membrane and interact with the macrophage cytosol before killing the phagocyte and spreading to another cell (14–19).

Protein secretion systems transport bacterial virulence factors that mediate interaction with the host and are key virulence determinants of bacterial pathogenesis. ESX-1 (ESAT-6-system-1) is a conserved type VII secretion system that is required for the survival of both M. marinum and M. tuberculosis during early infection (20–23). The ESX-1 system is required for perforation of the phagosome (14, 18). Strains of M. marinum and M. tuberculosis that lack the ESX-1 system are retained in the phagosome and attenuated (14–19, 21, 23).

In Gram-negative bacterial pathogens, protein secretion is linked to gene expression (24, 25). By linking gene expression and protein secretion, the bacteria can avoid depletion or accumulation of protein substrates in the bacterial cell (26). Although mycobacteria are unrelated to Gram-negative pathogens, we discovered that the ESX-1 secretion system controls gene expression in M. marinum (27–29). We identified ~160 genes whose expression was significantly altered in response to the presence or absence of the ESX-1 system (29). The ESX-1 system in M. tuberculosis has since been similarly linked to changes in gene expression (30, 31).

ESX-1 controls gene expression in part through feedback control. The ESX-1 system positively controls the expression of ESX-1 substrate genes and the WhiB6 transcription factor in M. marinum (29). WhiB6 positively regulates the expression of the ESX-1 substrate genes (29, 32, 33), likely ensuring substrate production in the presence of the ESX-1 system. In the absence of the ESX-1 system, whiB6 gene expression is repressed, resulting in reduced levels of ESX-1 substrates, likely ensuring that substrates do not accumulate in the absence of protein secretion.

To identify the transcription factors responsible for regulating gene expression in response to the ESX-1 system, we performed a promoter pulldown assay with the whiB6 promoter. This approach led to the identification of EspM as a new conserved transcription factor in pathogenic mycobacteria. We demonstrated that EspM was bound to the whiB6 promoter in M. marinum and in vitro. The espM gene is divergently transcribed from the whiB6 gene and is adjacent to the ESX-1 locus in both M. marinum and M. tuberculosis and in the nonpathogenic species Mycobacterium smegmatis. Deletion of the espM gene increased expression of the whiB6 gene and altered expression of 99 additional genes, the majority of which are not associated with ESX-1 (28). The changes in gene expression were complemented by expression of the espM gene from M. marinum. Importantly, expression of the espM gene from M. tuberculosis in the ΔespM M. marinum strain restored repression of the whiB6 gene. Finally, we demonstrated that EspM was required for the repression of whiB6 gene expression in the absence of the ESX-1 system (28).

Although EspM clearly and directly regulates gene expression, EspM has not been functionally characterized. Importantly, it is not known how EspM is regulated to repress whiB6 gene expression in the absence of the ESX-1 system. We reasoned that we could gain insight into the regulation and activity of EspM by performing a structure-function analysis of the EspM protein. Our findings provide initial insight into the DNA binding ability and multimerization of EspM.

RESULTS

The EspM C terminus is sufficient for whiB6 repression in vivo.

To further understand the functional domains of the EspM protein, especially in the C-terminal half, we generated predicted structural models of EspM using Robetta, LOMETS, and AlphaFold (Fig. 1A) (34–36). All three algorithms predicted that EspM contains four defined domains: an N-terminal forkhead-associated (FHA) domain and two helix-turn-helix [HTH(A) and HTH(B)] DNA binding domains separated by a helix bundle of unknown function (Fig. 1B). The Robetta-generated structure has confidence scoring of 73%, with the largest error contributions from the two predicted linker regions between the FHA and HTH(A) domains and between HTH(A) and the helix bundle. The LOMETS-generated structure was largely equivalent to the Robetta generated structure. The AlphaFold-generated structure was very similar to the other two structures, but with less predicted secondary structure outside of the conserved domains.

FIG 1.

The EspM_CT domain is sufficient to repress whiB6 expression in M. marinum. (A) Comparison of models of the EspM protein from the Robetta protein structure prediction server with a confidence scoring of 73% (34), LOMETS (35), and AlphaFold2 (36). The FHA domain is in cyan, HTH(A) is in green, HTH(B) is in magenta, and the helical bundle is in orange. The structure was visualized using the PyMOL molecular graphics system, version 2.0 (Schrödinger, LLC) (43). (B) Schematic of the predicted domains of the EspM protein. FHA, forkhead-associated domain; HTH, helix-turn-helix; NT, N terminus; CT, C terminus. (C) Western blot analysis of M. marinum cell-associated proteins. Protein (10 μg) was loaded in each lane of a 4 to 20% Tris-glycine gel. Anti-RpoB antibody was used as the loading control. All M. marinum strains in this panel contain a C-terminal 3×FLAG epitope tag on the whiB6 gene. α-FLAG antibody was used to measure WhiB6. α-V5 antibody was used to measure EspM. The Western blot shown is representative of at least three independent biological replicates.

We previously demonstrated that amino acids 154 to 363 of the EspM protein (EspM_CT) were necessary and sufficient for binding DNA in vitro (28). We sought to test if the C terminus of EspM was sufficient to repress whiB6 expression in M. marinum. We generated plasmids constitutively expressing either the espM or the espM_CT (154 to 363 amino acids [AA]) genes tagged with a V5 epitope to verify the expression and stability of each protein in M. marinum. The resulting plasmids were introduced in the ΔespM M. marinum strain with a whiB6 allele including a FLAG epitope tag integrated into the whiB6 locus (29). The EspM-V5 and EspM_CT-V5 proteins were both expressed in M. marinum (Fig. 1C, lanes 4 and 5).

Both the EspM-V5 and the EspM_CT-V5 proteins resolved as multiple bands specifically recognized by the V5 antibody. The lower EspM-V5 (expected, 42.1 kDa) band was between the 25 kDa and 37 kDa markers, with a second band between the 37 and 50 kDa markers, respectively (lane 3). The lower EspM_CT-V5 (expected 24.6 kDa) band ran approximately at the 20 kDa marker, with a second band between the 25 kDa and 37 kDa markers (lane 4).

As we demonstrated previously, deletion of the espM gene resulted in increased levels of WhiB6 protein due to the loss of the EspM repressor (Fig. 1C, lane 3 versus lanes 1 and 2). Expression of either the EspM-V5 or EspM_CT-V5 protein reduced the levels of WhiB6-Fl protein in the espM strain. Expression of the EspM_CT-V5 protein resulted in levels of WhiB6-Fl protein lower than those in the wild-type (WT) strain, suggesting that the N terminus regulates the repressor function of EspM. From these data, we conclude that the EspM_CT is sufficient to repress whiB6 expression under the conditions tested, consistent with our previous in vitro findings.

Identification of domains and residues required for EspM multimerization.

We previously used electrophoretic mobility shift assays (EMSAs) to demonstrate direct DNA binding by a purified, epitope-tagged version of EspM, 6×His-EspM (28). We observed two species of shift products, suggesting that EspM may bind DNA as a multimer (28). To test this, we analyzed a prediction of EspM multimerization using AlphaFold2 as implemented in ColabFold (36, 37). AlphaFold2 predicted that EspM forms a dimer, primarily through association of the helical bundle and the HTH(B) domain, with high confidence scores for the interacting domains, with a predicted local distance difference test (pLDDT, >90) and low predicted aligned error (PAE) (Fig. 2A, pLDDT and PAE scores are shown in Fig. S1 in the supplemental material). We used this model to guide our investigation of EspM function.

FIG 2.

Identification of residues required for EspM multimerization. (A) A dimer of EspM predicted by AlphaFold2 is shown in cartoon representation. One protomer is colored by domains as in Fig. 1A; another protomer is colored blue. The dimerization is mediated by the helical bundles and the HTH(B) domains. (B) Schematic of the LexA-EspM proteins used in this study. DBD, DNA binding domain; FHA, forkhead-associated domain; HTH, helix-turn-helix; NT, N terminus; CT, C terminus. (C to F) Bacterial mono-hybrid assays measuring multimerization capacity of and the following: (C) the N- and C-terminal domains. Significance was determined using an ordinary one-way ANOVA (P < 0.0001) followed by a Tukey’s multiple-comparison test. The significance shown is relative to the bacteria-alone control; ****, P < 0.0001. LexA-EccCb₁ and LexA-EsxB are controls for intermediate interactions. EspM, EspM_NT, and EspM_CT interactions were measured in SU101 in a one-hybrid system. (D) Multimerization of EspM_CT variants on strains isolated from the genetic screen based on the blue colony phenotype on agar with X-Gal and IPTG. Significance was determined using a one-way ordinary ANOVA (P < 0.0001) followed by Dunnett’s multiple-comparison test, relative to LexA-EspM_CT; ****, P < 0.0001; **, P = 0.0035. (E) EspM variants with individual mutations generated by site-directed mutagenesis. Significance was determined using a one-way ordinary ANOVA (P < 0.0001) followed by Dunnett’s multiple-comparison test relative to LexA-EspM; ****, P < 0.0001; **, (R145C) P = 0.0056; **, (G193R) P = 0.0021; *, P = 0.0265. (F) Domains within the predicted EspM C terminus. HB, helix bundle. Significance was determined using a one-way ordinary ANOVA (P < 0.0001) followed by Tukey’s multiple-comparison test. The comparisons shown are compared to LexA-EspM_CT; ****, P < 0.0001. The data represent averages of results from at least three biological replicates, each performed in technical triplicate. Each data point represents a technical replicate.

To test if EspM multimerizes in vivo, we used the LexA bacterial mono-hybrid system (38). This system exploits the dimeric LexA transcriptional repressor, a LexA mutant (LexA₄₀₈) with altered DNA binding activity and an Escherichia coli reporter strain with a LexA-repressible lacZ gene. In the SU202 E. coli strain, the formation of heterodimers of the WT and mutant LexA fusion proteins allow binding of the lexA operators, leading to repression of lacZ gene expression. In the SU101 E. coli strain, self-association of LexA fusion proteins represses lacZ (39).

We generated a gene fusion between the lexA DNA binding domain and espM and expressed them in the SU101 E. coli strain (Fig. 2B). If the LexA-EspM proteins dimerize, then lacZ expression will be repressed due to LexA binding to the DNA. We used the interaction between the Fos and Jun eukaryotic transcription factors as a positive control (38). As shown in Fig. 2C, expression of LexA-Fos and LexA₄₀₈-Jun proteins significantly repressed lacZ gene expression compared to that of the bacteria alone (P < 0.0001). Expression of LexA-EccCb₁ and LexA₄₀₈-EsxB, two ESX-1-associated mycobacterial proteins (21), served as controls for an intermediate interaction (40), which likewise resulted in significant repression of the lacZ reporter compared to bacteria alone (P < 0.0001). Expression of the LexA-EspM protein in the SU101 strain resulted in β-galactosidase activity that was significantly reduced from that of the bacteria alone (P < 0.0001) and not significantly different from that of the LexA-Fos/LexA₄₀₈-Jun control. The LexA-EspM proteins were expressed in E. coli as measured by Western blot analysis (Fig. S2). Together, these data support that EspM multimerizes.

In addition, we generated gene fusions between the lexA DNA binding domain and espM_NT or espM_CT as shown in Fig. 2B and expressed them in the SU101 E. coli strain (Fig. 2C). Expression of either the LexA-EspM_NT or LexA-EspM_CT fusion proteins in SU101 resulted in significant repression of lacZ gene expression compared to the SU101 cells alone (P < 0.0001). Both the LexA-EspM_NT or LexA-EspM_CT fusion proteins were expressed in E. coli as measured by Western blot analysis (Fig. S2A). Expression of LexA-EspM_NT resulted in β-galactosidase activity that was significantly higher than that of those mediated by the LexA-EspM protein (P < 0.0461). Expression of LexA-EspM_CT resulted in β-galactosidase activity that was similar to that of those mediated by the LexA-EspM protein. These data support the conclusion that both the EspM_NT and EspM_CT proteins self-associate in vivo. Notably, coexpression of LexA-Fos and LexA₄₀₈-EspM_CT, or the LexA-EspM_CT and LexA₄₀₈-Jun proteins in SU202, did not repress lacZ expression, indicating that the self-association of the LexA-EspM_CT was specific (Fig. S3).

Because the AlphaFold2 model predicted that the dimerization of EspM was mediated through the EspM C-terminal half (Fig. 2A), and the C terminus of EspM was sufficient to repress whiB6 in vivo (Fig. 1C), we sought to identify specific domains in the EspM_CT required for multimerization. Therefore, we used the EspM_CT, which had the most robust phenotype in the bacterial two-hybrid assay, to screen for EspM variants with altered multimerization properties. We generated a library of LexA-EspM_CT variants with ~4,300 colonies expressing lexA-espM_CT genes mutagenized using error-prone PCR. Using targeted DNA sequencing, we found that ~50% of the lexA-espM_CT plasmids contained at least one mutation in the espM gene. We screened ~4,000 colonies by plating the reporter strains with the mutagenized plasmids on agar with X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) and IPTG (isopropyl-β-d-thiogalactopyranoside) to induce LexA fusion protein expression. Colonies with a blue phenotype when grown on agar with X-Gal and IPTG reflected reporter derepression, signifying a potential loss of EspM_CT multimerization.

Using this approach, we identified 130 colonies with blue phenotypes on agar with X-Gal compared to the SU101 strain expressing the lexA-espM_CT parental plasmid. We selected 46 blue colonies to characterize further. Using targeted DNA sequencing, we defined specific mutations in the lexA-espM_CT gene (Table S1). Based on the DNA sequencing results for the 46 plasmids, we estimate that the mutation rate of the library was 1 per 375.5 bp. Notably, many of the strains contained plasmids with the same mutations.

We measured the levels of β-galactosidase activity for each of the 46 strains exhibiting blue colony phenotypes on agar with X-Gal. We pooled the data from the strains with identical mutations in the lexA-espM_CT gene, as shown in Fig. 2D. We observed various levels of derepression of lacZ gene expression, some of which were significantly different from the levels of β-galactosidase activity measured in the strain expressing the parental LexA-EspM_CT protein.

The majority of the LexA-EspM_CT proteins identified in our screen contained multiple amino acid changes. To determine which mutations impacted LexA-EspM multimerization, we used site-directed mutagenesis to create mutations that would result in single amino acid changes in the LexA-EspM protein. All of the LexA-EspM fusion proteins were expressed to various levels in E. coli (Fig. S2B). As shown in Fig. 2E, three variants resulted in a significant derepression of the lacZ reporter compared to the LexA-EspM protein. The LexA-EspMR145C and LexA-EspMG193R variants resulted in an intermediate level of lacZ derepression (Fig. 2E), with β-galactosidase activity that was significantly higher than that of the LexA-EspM expression strain (R145C P = 0.0056, G193R P = 0.0021). Both mapped to the predicted HTH(A) domain of EspM (Fig. 1B).

A total of 72% of the colonies isolated in our screen had mutations in two base pairs (C722T/T853G) resulting in a double amino acid change in the EspM_CT protein (S219F/L263V, Fig. 2D). The LexA-EspMS219F variant repressed lacZ expression to levels similar to that of the WT LexA-EspM protein. In contrast, the LexA-EspML236V variant resulted in significant derepression of lacZ expression relative to the LexA-EspM protein (P < 0.0001). The levels of derepression could reflect the reduced levels of LexA-EspML236V in E. coli (Fig. S2B). The L236V change mapped to the helix bundle. Together, these data suggested that two domains, HTH(A) and the helix bundle, may contribute to EspM_CT multimerization.

To test if HTH(A) and the helix bundle were sufficient to mediate multimerization, we generated the LexA-HTH(A)-helix bundle and LexA-helix bundle-HTH(B) fusion proteins (Fig. 2F). Both fusion proteins were expressed and stable in E. coli as measured by Western blot analysis (Fig. S2C). As shown in Fig. 2F, the LexA-HTH(A)-HB protein was unable to repress lacZ, similar to the SU101 negative control, and significantly different from the LexA-EspM_CT fusion protein (P < 0.0001). The LexA-HB-HTH(B) protein moderately repressed lacZ expression, resulting in intermediate levels of repression that were significantly different from both the SU101 negative control (P < 0.0001) and the LexA-EspM_CT fusion protein (P < 0.0001). From these data, we conclude that all three domains are necessary for mediating multimerization of the EspM C-terminal half under these conditions.

The EspM variants have altered function in M. marinum.

We tested if the EspM variants we isolated in our screen impacted regulation of whiB6 gene expression in M. marinum. The mutations were introduced in the espM gene behind a constitutive promoter on an integrating plasmid. Each espM mutant gene was tagged with a V5 epitope (espM-V5) to verify expression and stability of each protein in M. marinum. The resulting plasmids were introduced in the ΔespM M. marinum strain with a whiB6 allele including a FLAG epitope tag integrated into the whiB6 locus (29).

We measured the impact of each EspM variant on the levels of whiB6 expression and WhiB6 protein in M. marinum. As shown in Fig. 3 and consistent with our prior findings, whiB6 was expressed in the WT M. marinum strain (Fig. 3A), and we detected WhiB6-Fl protein in the WT M. marinum strain (Fig. 3B, lane 1). Deletion of the eccCb₁ gene, which causes a loss of the ESX-1 secretory apparatus, resulted in a significant reduction of whiB6 expression (Fig. 3A, inset; P < 0.0001 relative to the WT strain) and a corresponding loss of WhiB6-Fl protein (Fig. 3B, lane 2) (29). The ΔespM strain had significantly increased whiB6 expression (Fig. 3A; P = 0.0003) and WhiB6-Fl protein levels (Fig. 3B, lane 3), reflecting the derepression of whiB6 gene expression (28). Expression of the espM-V5 genes in the ΔespM strain led to detectable levels of the WT and mutant EspM-V5 proteins (Fig. 3B, lanes 4 to 8). These bands are specific because they are only present in strains expressing espM-V5 and are absent from lanes 1 to 3. Densitometry analyses of the levels of the EspM-V5 variants did not reveal significant differences in the steady-state levels of these proteins in M. marinum (Fig. S4).

FIG 3.

EspM multimerization is uncoupled from whiB6 regulation. (A) qRT-PCR measuring the levels of whiB6 expression relative to sigA expression. A one-way ordinary ANOVA, followed by Dunnett’s multiple-comparison test relative to the WT strain, was performed; ***, P < 0.0003. The inset shows just the comparison between the WT and ΔeccCb₁ strains. Student’s unpaired, two-tailed t test was used to define the significance of the results of the comparisons between the two strains; P < 0.0001. The data represent averages of results from at least three biological replicates, each performed in technical triplicate. (B) Western blot analysis of 10 μg of protein per lane. Anti-RpoB antibody was used as the loading control. Anti-FLAG antibody was used to measure the expression of WhiB6. Anti-V5 antibody was used to measure the stability and expression of EspM and EspM mutants. All M. marinum strains indicated in this panel contained a C-terminal FLAG epitope tag on the whiB6 gene. The ΔespM M. marinum strains were complemented with a C-terminal V5 tag on the espM gene. Samples were resolved on a 4 to 20% Tris-glycine gel. The Western blot shown is representative of at least three independent biological replicates.

Expression of EspM-V5 in the ΔespM strain reduced whiB6 expression to WT levels and WhiB6-Fl protein below WT levels, consistent with complementation by the espM gene observed previously (28). Similarly, expression of the EspMG193R-V5, EspMS219F-V5, and EspML263V-V5 proteins resulted in whiB6 expression (Fig. 3A) and WhiB6-Fl protein levels similar to those of the WT strain (Fig. 3B, lanes 6 to 8). The expression of these three mutant genes did not complement as well as the WT espM gene, despite maintaining their ability to regulate whiB6 gene expression in M. marinum.

The EspMR145C-V5 variant resulted in phenotypes distinct from the parental EspM protein. The EspMR145C-V5 protein was unable to repress whiB6 gene expression (Fig. 3A), resulting in whiB6 expression that was significantly different from that of the WT strain (P < 0.0003) and similar to that of the ΔespM strain. The WhiB6-Fl protein levels were similar to those of the ΔespM strain (Fig. 3B, lane 5 versus 3). From these data, we conclude that EspMR145C was unable to repress whiB6 expression in M. marinum.

The HTH(A) domain, but not the HTH(B) domain, is essential for binding the whiB6 promoter.

To better understand why the EspMR145C protein failed to repress whiB6 expression in M. marinum, we expressed and purified espM, espMR145C, espMS219F, and espML263V fused to maltose binding protein (MBP)-6×His from E. coli (Fig. S5A and B). The resulting purified MBP-6×His-EspM proteins were cleaved with Tobacco Etch Virus (TEV) protease, which removed both the MBP and 6×His epitope tag from the EspM proteins. We tested the ability of the resulting EspM variant proteins to bind DNA in vitro. As shown in Fig. 4A, EspM specifically bound the whiB6 promoter probe in a concentration-dependent manner. Increasing concentrations of EspM protein shifted the mobility of the whiB6 promoter probe (bound probe) and resulted in a corresponding loss of the unbound whiB6 probe. Addition of EspM did not shift the control rpoA probe, consistent with our previous findings (28).

FIG 4.

The HTH(A) and HTH(B) domains are functionally distinct. (A and B) EMSA measuring binding of heterologously expressed and purified (A) EspM and (B) EspMR145C. (C) Alignment of the two predicted HTH domains from EspM; green, HTH(A); magenta, HTH(B). The triple R residues are annotated for each domain. R317C is the equivalent change to R145C. (D) Western blot analysis of 10 μg of protein per lane. Anti-RpoB antibody was used as the loading control. Anti-FLAG antibody was used to measure expression of WhiB6. Anti-V5 antibody was used to measure stability and expression of EspM and EspM mutants. All M. marinum strains indicated in this panel contained a C-terminal FLAG epitope tag on the whiB6 gene. All M. marinum strains were complemented with a C-terminal V5 tag on the espM gene. Samples were resolved on a 4 to 20% Tris-glycine gel. The Western blot shown is representative of at least three independent biological replicates. (E) EMSA measuring binding of heterologously expressed and purified EspM and EspMR317C with the whiB6 probe The EspMR145C protein is used as a control for loss of DNA binding. The figure is representative of three biological replicates.

Unlike the wild-type protein, EspMR145C did not bind DNA. As shown in Fig. 4B, increasing amounts of EspMR145C did not shift the whiB6 probe as determined by EMSA. Higher concentrations of EspM protein were used to rule out reduced binding affinity of the mutant protein compared to that of the WT protein. Because EspMR145C did not bind DNA, but phenocopied the ΔespM deletion strain in M. marinum, we performed circular dichroism (CD) analysis to define the secondary structure of the EspMR145C protein. As shown in Fig. S6A, both the WT and R145C mutant show distinct negative peaks at 208 nm and 222 nm, characteristic of proteins containing alpha helices (41). The beta-sheets that make up the FHA domain may contribute to the flattening of the spectra in the 210- to 220-nm region, as beta-sheets are known to show up as a broad negative peak at approximately 218 nm (41). The amplitude, which correlates with protein concentration rather than structure, of the EspMR415C trace is much shallower than that of the WT EspM trace. However, the general shape and peak locations remain consistent (Fig. S6), indicating that EspM WT and EspMR145C have similar secondary structures. These data were further confirmed via CAPITO (CD Analysis and Plotting Tool) analysis of the CD data (42). The CAPITO analysis includes data comparison of the predicted secondary structure between inputted CD data and that from a database of known structures. The WT and R145C EspM proteins clustered together in this plot, supporting the finding that the two proteins are similarly folded (Fig. S6B). Both the EspMS219F and EspML263V proteins bound the whiB6 promoter probe at least as well as the WT EspM protein (Fig. S7), consistent with the ability of each protein to effectively complement the ΔespM strain. These data are consistent with the data in Fig. 3A, suggesting that each protein repressed whiB6 expression because they bound the whiB6 promoter. Moreover, these data suggest that mutations in specific domains of EspM protein disrupt DNA binding.

Together, these data led us to conclude that the HTH(A) domain is required for DNA binding of EspM to the whiB6 promoter. These data suggest that the EspMR145C protein was unable to repress whiB6 expression in M. marinum (Fig. 3A) because it fails to bind the whiB6 promoter.

EspM is predicted to have a second helix-turn-helix domain [HTH(B); Fig. 1 and 2]. To test if the HTH(B) domain was also required for the regulation of whiB6 expression, we sought to identify residues within HTH(B) that would mediate DNA binding. To this end, we aligned the HTH(A) and HTH(B) domains using PyMOL to perform a structure-based sequence alignment (Fig. 4C) (43). We identified three arginine residues [HTH(A), R144-146, HTH(B), R317-R319] that are conserved between both HTH domains. Because the R145C mutation in HTH(A) disrupted DNA binding at the whiB6 promoter, we created an analogous mutation in HTH(B), R317C. This mutation was introduced into the espM-V5 gene behind a constitutive promoter on an integrating plasmid. The resulting plasmids were transformed into the ΔespM M. marinum strain with a whiB6 allele including a FLAG epitope tag integrated into the whiB6 locus (29).

As shown in Fig. 4D, and consistent with our prior findings, the ΔespM strain had increased WhiB6-Fl protein levels compared to the WT strain (Fig. 4D, lane 2 versus 1), reflecting the derepression of whiB6 gene expression (28). Expression of the espM-V5 genes in the ΔespM strain led to detectable levels of the WT and mutant EspM-V5 proteins (Fig. 4D, lanes 3 to 5). However, over multiple biological replicates, the levels of the EspMR317C protein were significantly reduced compared to those of the EspM-V5 protein, based on densitometry analyses (Fig. S8A). The EspM-V5 and EspMR317C-V5 proteins complemented the WhiB6-Fl levels in the ΔespM strain (lanes 3 to 5 versus lane 2). In all additional replicates (Fig. S8), expression of the EspMR317C protein reduced WhiB6 levels, but not consistently, to the level of reduction by the WT protein. Based on densitometry analysis over five biological replicates, the EspMR317C protein did not result in significantly higher levels of WhiB6-Fl protein compared to the WT EspM protein (Fig. S8B and C). From these data, we conclude that the EspMR317C protein regulates whiB6 expression in M. marinum. Because the EspMR317C protein retained function and the EspMR145C protein lost regulatory function, we conclude that the two domains are functionally distinct.

The EspMR317C mutant was expressed in and purified from E. coli to determine its ability to bind DNA and multimerize (Fig. S5). To determine if a mutation analogous to HTH(B) would impact EspM binding to the whiB6 promoter, we performed EMSAs. As shown in Fig. 4E, the EspMR317C protein retained DNA binding activity, as evidenced by the concentration-dependent shift in mobility of the whiB6 probe (Fig. 4E, left and right). In addition, the secondary structure of EspMR317C compared to that of the WT protein was measured via circular dichroism (Fig. S6). The amplitude and shape of the observed peaks closely match that of both the WT protein and of the EspMR145C mutant. CAPITO analysis confirmed that the mutant clusters with the WT protein (Fig. S6) (42). Together, these data suggest that the EspMR317C protein may have a higher DNA binding affinity for the whiB6 promoter than the EspM protein but does not have a significant change in secondary structure. Moreover, the HTH(B) domain, and the three conserved arginine residues therein, is functionally distinct from the HTH(A) domain.

To look more closely at how the amino acid change might impact function, we modeled each mutant using the Robetta protein structure prediction server. Each of the mutant proteins were predicted to have secondary structure almost identical to that of the WT protein and had similar confidence scores (72 to 73%), with the largest error contribution coming from the unstructured linker regions between the FHA and HTH(A) domains and between HTH(A) and the helix bundle. The mutant structures were aligned against the WT structure to more directly compare side chain conformation and predicted hydrogen bonding activity (Fig. S9). In the WT EspM protein, R145 is predicted to form a salt bridge with E170 within the HTH(A) domain. This salt bridge cannot form in the R145C mutant as shown in Fig. S9A. This missing salt bridge may destabilize the helix-turn-helix domain’s structural integrity, preventing the protein from binding the whiB6 promoter. Interestingly, this glutamic acid is conserved in HTH(B) (E342). However, the helix containing R317 is rotated relative to the analogous helix in HTH(A), so no salt bridge is predicted to form in this domain.

The helical bundle is required for WT multimerization in vitro.

We next sought to determine if the EspM variants had altered multimerization in vitro. As shown in Fig. 5A, size exclusion chromatography (SEC)-high-performance liquid chromatography (HPLC) analysis of the EspM protein, which is predicted to be 38.3 kDa, resulted in three distinct peaks. The broad peak with the largest mass eluted to an overall average mass of 196 ± 22 kDa (Fig. 5A), which correlated with pentamers and/or hexamers of EspM (Fig. 5A, Table S2). The second-largest peak eluted to an overall mass of 83 ± 14 kDa, which corresponds to dimers of EspM (Fig. 5A and B, Table S2). The final peak eluted at 40 ± 5 kDA, which corresponded to monomers of EspM. SDS-PAGE demonstrated that the EspM protein was found in the fractions corresponding to each of these peaks (Fig. 5A and B, Table S2, Fig. S10). From these data, we conclude that the EspM protein forms higher-order multimers in vitro.

FIG 5.

Multimerization of EspM proteins in vitro. (A, C, and E) HPLC chromatograms of (A) EspM, (C) EspMS219F, and (E) EspML263V proteins. HPLC was performed with a HiLoad 16/60 Superdex XK column and a flow-rate of 0.5 mL/min. Peaks represent enriched multimeric forms of EspM. Each figure represents three biological replicates. (B, D, and F) Quantification of the size of primary peaks from HPLC chromatograms for (B) EspM, (D) EspMS219F, and (F) EspML263V proteins. The standard curve for sizing was generated using a size-standard kit using β-amylase (200.0 kDa), alcohol dehydrogenase (150.0 kDa), albumin (66.0 kDa), carbonic anhydrase (29.0 kDa), and cytochrome c (12.4 kDa). V₀ represents the void volume of the column, and V_e represents average elution volume for peaks across three replicates. Line graph curves were generated using standard linear regression.

Both EspMS219F and EspML263V multimerized in a way that was distinct from the WT protein. SEC-HPLC of the EspMS219F and EspML263V proteins revealed two distinct peaks, with the largest peak eluting to 235.8 ± 0.9 and 238 ± 4 kDa, respectively, corresponding to the hexameric form (Fig. 5C to F, Table S2). The second peak eluted to 45.2 ± 0.6 kDa and 44 ± 1 kDa, respectively, which was consistent with the monomer (Fig. 5C to F, Table S2). Neither the EspMS219F nor the EspML263V protein formed distinct dimers or pentamers.

Using SEC-HPLC, we found that both the EspMR145C and EspMR317C proteins showed peak patterns similar to those of the WT protein (Fig. S11). The only notable difference was that the EspMR317C SEC-HPLC chromatogram included a peak at 695 ± 99 kDa, outside the range of the standards used to determine size in this protocol. This peak may correspond to the multimerization of MBP-His-TEV-EspMR317C, which is also visible on the SDS-PAGE gel following protein purification (Fig. S5). The multimerization species of the R317C mutant does not appear to be significantly changed from that of the WT protein. However, the monomer peak did appear larger than that of the WT in the R317C chromatogram. The largest peak in the WT EspM data was very broad and had variable elution time, suggesting that the peak encompassed multiple large multimeric states. In the case of the R145C mutant protein, the largest peak was split into heptameric and pentameric peaks, each of which had a consistent elution time, suggesting that the R145C mutant protein primarily formed heptamers and pentamers. From these data, we conclude that the mutations in the helical bundle specifically altered multimerization of EspM in vitro.

We modeled each mutant using the Robetta protein structure prediction server (34). In the WT EspM protein, S219, is predicted to form a network of hydrogen bonds with T218, D220, and S224, which stabilizes the turn at the end of the helix bundle leading to HTH(B) (Fig. S9B). The S219F mutation results in a destabilization of this hydrogen bond network, creating a bulge within the turn. Although this bulge is not predicted to change the overall secondary structure of the protein, it could lead to subtle changes in the tertiary structure, in turn affecting multimerization. In the WT EspM protein, L263 is an interior residue within the hydrophobic core of the helix bundle. The mutation to V263 does not cause any predicted secondary structure changes or any large rotamer changes in nearby residues (Fig. S9C). However, the valine does open the interior of the helix bundle slightly, allowing H294 to move into the space. This subtle change could be enough to disrupt the tertiary structure, or the predicted model may not accurately display the mutation’s effect on the protein.

DISCUSSION

We previously demonstrated that the ESX-1 system has physiological roles regulating gene expression in the mycobacterial cell (27–29). We discovered a new, conserved transcription factor that ensures reduced levels of ESX-1 substrates in the absence of a functional ESX-1 transporter (28). In this study, we found that the HTH(A) domain is specifically required for DNA binding at the whiB6 promoter. We identified an EspM variant that was unable to bind DNA in vivo or in vitro. The equivalent mutation in the predicted HTH(B) domain resulted in reduced levels of EspM protein, which retained the ability to regulate whiB6 expression. We do not understand the mechanism leading to reduced levels of EspMR317C protein in M. marinum. We found that the full-length EspM protein, as well as its N- and C-terminal halves have multimeric capacity. Our findings suggest that the helical bundle in the EspM C terminus promotes dimer and trimer formation but is dispensable for higher order oligomerization. We isolated two EspM variants which formed high-order multimers, but not dimers or trimers, supporting the idea that formation of higher-order multimers does not require the prior formation of dimers or trimers. Finally, our data indicate that either the hexameric or monomeric form of EspM mediates DNA binding to the whiB6 promoter. This study represents the initial structure-function analysis of a conserved ESX-1-associated transcription factor and supports the overall idea that EspM is a multifaceted transcription factor that performs a number of different roles in pathogenic mycobacteria.

HTH domains mediate DNA binding of transcription factors (44). Here, we described the isolation and characterization of the EspMR145C variant, which was unable to bind DNA in vitro, was unable to repress whiB6 expression in M. marinum, and did not form hexamers like the WT EspM. The R145C mutation is located within a highly conserved amino acid sequence in pathogenic mycobacteria. Sequence alignments of M. marinum, M. tuberculosis (H37Rv), and M. smegmatis orthologs of EspM show a 47-residue domain that is 100% conserved across the three species (Fig. S12). Since this domain is highly conserved, and mutation of this domain leads to loss of whiB6 regulation, this indicates that the HTH(A) domain is essential for repression of whiB6 expression. EspM has multiple HTH domains, an uncommon characteristic in transcription factors. Prior reports demonstrate that transcription factors can use the multiple HTH domains to bind distinct binding sites and differentially regulate gene expression (45, 46). Our previous RNA sequencing data revealed that EspM may directly or indirectly regulate genes both positively and negatively (28). Given that the EspMR145C variant, with a single change in the HTH(A) domain, did not bind the whiB6 promoter and was unable to repress whiB6 expression in M. marinum, the second HTH domain [HTH(B); Fig. 1] may target other genes in the EspM regulon. In support of distinct functions for the two HTH domains, the equivalent change in the HTH(B) domain, EspMR317C, did not abolish DNA binding or multimerization of the EspM protein.

The EspML263V and EspMS219F proteins, which we showed formed both hexamers and monomers, bound the whiB6 promoter in vitro and repressed whiB6 expression in M. marinum. Although our studies did not directly test which form is relevant in vivo, our data are consistent with EspM binding the whiB6 promoter as a hexamer or a monomer. It could be that either form (hexamer or monomer) is capable of binding DNA. DNA-binding proteins that form hexamers have been reported previously, such as DNA helicases and proteins involved with DNA replication (47–49). Additionally, there are examples of hexameric transcription factors that wrap promoter DNA to regulate gene expression (50, 51). An example of this is RovC, a conserved hexameric DNA-binding protein that promotes expression of the type VI secretion system in Yersinia pestis (50). EspM may similarly bind to several sites on the whiB6 promoter and repress whiB6 expression by wrapping the DNA around itself. In support of the DNA binding capacity of EspM in its monomeric form, the C-terminal half of the EspM protein is sufficient to bind the whiB6 promoter (28) and, based on this study, forms dimeric or monomeric multimers. Likewise, in this study (Fig. 4) and our previous study, multiple shifted forms of EspM-whiB6 promoter DNA are visible by EMSA analyses.

Alternatively, it is possible that one form or the other binds DNA. For example, the hexameric nature of EspM could reflect the hexameric nature of ESX-systems. The conserved components of the ESX-3 and ESX-5 secretion systems that are associated with the cytoplasmic membranes in pathogenic mycobacteria form hexameric cores (52–54). The EccCb₁ and EccA conserved components of the ESX-1 system are also hexameric (7, 54, 55). We previously showed that EspM represses whiB6 gene expression specifically in the absence of a functional ESX-1 system (28, 29). One possibility is that EspM interacts with a hexameric ESX-1 component, connecting the ESX-1 membrane complex to changes in gene expression. For example, the EccCb₁ hexamer could bind the EspM hexamer, preventing EspM from repressing whiB6 expression when ESX-1 is functional. In the absence of ESX-1, EspM monomers could then bind DNA and repress whiB6 expression, leading to reduced substrate levels. This model is consistent with the well-established idea that the activity of a transcription factor can be altered by multimerization status (56, 57). The isolation of an EspM variant that fails to form higher-order multimers would assist in testing this model. If such a model is correct, EspM could switch between forming hexamers and monomers. Examples of regulation of the multimeric state by posttranslational modification are found in both type III and ESX/type VII secretion systems (58–62). The function of FHA domains can be regulated by phosphorylation (58–60). Understanding the posttranslational modification of EspM at the N terminus may provide insight into the drivers of EspM multimerization.

Our data showed that the N and C termini multimerized differently from the full-length EspM protein. The N terminus formed trimeric species, while the C terminus formed dimeric species. The formation of dimeric and trimeric species is dispensable for the formation of higher-order multimers of EspM, as evidenced by the variants isolated here. It has been previously demonstrated in E. coli that distinct multimerization domains in proteins can have inhibitory effects on each other (63). If this is the case with the C terminus and the N terminus of EspM, loss of multimerization of one could lead to enhanced multimerization of the other, perhaps explaining the ubiquitous presence of the hexameric species in the nondimerizing mutant proteins.

An outstanding question remains as to how the screen conducted using a bacterial mono-hybrid system connects to EspM activity in vitro and in vivo. The bacterial two-hybrid system used in this study has been used previously to study bacterial transcription factors (64–66). We were surprised that the mutations resulting in blue colony phenotypes were not all functionally related to multimerization of the C-terminal half of EspM. We suspect we isolated the EspMR145C variant, which does not bind DNA, because the DNA binding activity of EspM was contributing the lacZ repression mediated by the LexA-EspM_CT protein (Fig. 2). The EspMS219F variant did not have a phenotype in the bacterial mono-hybrid assay, yet this variant was unable to form dimers and trimers in vitro. Although the EspML263V variant seemed unable to repress lacZ expression in the bacterial mono-hybrid assay, the protein formed high-order multimers and repressed whiB6 expression in M. marinum. It is possible that the bacterial mono-hybrid system measures only dimerization and that the formation of higher-order multimers in this system is reflected by various levels of derepression of the LexA-regulated reporter.

In conclusion, we have interrogated the mechanisms of action for an ESX-1 transcription factor. Our initial characterization revealed that EspM multimerizes and binds DNA through at least one of two HTH domains. Our findings provide a foundation for future investigation into the connection between the ESX-1 system and the control of mycobacterial gene expression.

MATERIALS AND METHODS

Growth of bacterial strains.

All strains and plasmids used in this study are listed in Table S3. M. marinum strains are all isogenic and based upon the M strain (ATCC BAA-535). M. marinum was grown in Middlebrook 7H9 broth (Thermo Fisher) supplemented with 0.5% glycerol and 0.1% Tween 80 (VWR) or on Middlebrook 7H11 agar (Thermo Fisher) supplemented with 0.5% glycerol and 0.5% glucose. M. marinum strains were maintained at 30°C. M. marinum strains were supplemented with the appropriate amount of antibiotic as necessary (20 μg/mL kanamycin or 50 μg/mL hygromycin).

E. coli were grown in Luria-Bertani (LB) broth or on LB agar. Plasmids were maintained in DH5α cells (Thermo Fisher). Protein expression was performed in BL21 DE3 (Star) cells (Invitrogen). For the LexA bacterial two-hybrid system, all homodimer interactions were tested in SU101 reporter cells. All heterodimer interactions were tested in SU202 reporter cells (38). E. coli strains were maintained at 37°C, except for strains with the pMS604, pDP804, pSR658, and pSR659 plasmids (38), which were maintained at 30°C. E. coli cultures were supplemented with antibiotics when appropriate (50 μg/mL kanamycin, 200 μg/mL hygromycin, 200 μg/mL ampicillin, or 12 μg/mL tetracycline). Strains bearing the pSR659 and pDP804 plasmids were maintained on 100 μg/mL ampicillin.

Generation of DNA constructs.

All DNA constructs were generated using oligonucleotide primers purchased from Integrated DNA Technologies (IDT, Coralville, IA). All resulting constructs were verified using targeted Sanger DNA sequencing from GeneWiz or the Genomics and Bioinformatics Facility at the University of Notre Dame. All strains and constructs are listed in Table S3. All oligonucleotides used in this study are listed in Table S4.

(i) Bacterial mono-hybrid constructs. The lexA-fwd and lexA-rev primers were used to amplify the pSR658 and pSR659 plasmids by PCR (39). The inserts for the pSR658 and pSR869 plasmids included the espM, espM_CT, espM_NT, espM_{HTH(A)_HB}, espM_{HB_HTH(B)}, eccCb₁, and esxB genes, which were amplified using the primers listed in Table S4. The inserts were introduced into the pSR658 or pSR659 plasmid into chemically competent DH5α cells using FastCloning (67). The resulting clones were selected exactly as described in Daines et al. (39) and extracted using the AccuPrep plasmid mini extraction kit (Bioneer). Plasmids were confirmed by PCR using SR-MCS-Fwd and SR-MCS-Rev primers followed by targeted DNA sequencing. The confirmed constructs were introduced into the SU101 and SU202 reporter strains previously described (38, 68, 69).

(ii) Complementation plasmids. All espM complementation plasmids were generated using site-directed mutagenesis (described below) or restriction cloning. The espM genes were excised from the pET15b-espM plasmid using NdeI and SpeI (23) and introduced into the pMH406 plasmid using T4 ligase (New England Biolabs [NEB]). The resulting ligations were introduced into chemically competent DH5α cells, plated on LB agar with 200 μg/mL hygromycin, and incubated at 37°C. Colonies were grown in LB broth supplemented with hygromycin before being extracted and confirmed as described.

(iii) Expression plasmids. We expressed mycobacterial proteins in E. coli using plasmids pET15b and pMF666. espM genes were introduced into the pET15b plasmid using FastCloning (23). The pET15b vector (Novagen) was amplified using the OMF512 and OMF513 primers. The espM_NT or espM_CT genes were amplified for insertion into pET15b from M. marinum M genomic DNA (gDNA) using the OMF624/OMF626 and OMF625/OMF627 primer pairs, respectively. The presence of the espM inserts was confirmed using targeted DNA sequencing. pMF666 is a modified version of pET28, with an N-terminal maltose-binding-protein (MBP) and 6×His-tag epitopes followed by a Tobacco Etch Virus (TEV) cleavage site. The vector was amplified from a modified version of pET28 (70) using the OMF33 and OMF34 primers. The espM gene was amplified from M. marinum gDNA, and the espMR145C, espMS219F, and espML263V genes were amplified from the appropriate complementation plasmids using the OMF628 and OMF629 primers. The espMR317C mutant gene was created via site-directed mutagenesis (described below) of the pMF666-espM plasmid using the ORP85 and ORP86 primers. The resulting constructs were introduced into chemically competent DH5α cells and selected on LB agar with 50 μg/mL kanamycin. Colonies were grown in LB broth supplemented with kanamycin before being extracted and confirmed using PCR and targeted DNA sequencing.

EspM_CT library construction.

Mutagenesis of the espM gene was performed using the GeneMorph II EZClone domain mutagenesis kit (Agilent) according to the manufacturer’s instructions. The espMct-fwd/espMfull-rev primer pair was used to amplify 100 ng of the target gene from the pSR658-espM_CT plasmid using Mutazyme II polymerase to generate the megaprimers as directed. Then, 50 ng of the pSR658-espM_CT vector was incubated with 250 ng of the megaprimers, amplified using EZClone high-fidelity polymerase, treated with DpnI, and introduced into chemically competent DH5α. The presence of mutations in the espM_CT gene was established by targeted DNA sequencing of the plasmids from 10 pilot colonies. The bacterial library of 3,936 colonies was collected, and the DNA was isolated using the AccuPrep plasmid mini DNA extraction kit (Bioneer). The resulting plasmid library was introduced into SU101 reporter cells via electroporation and grown on LB agar plates with 12 μg/mL tetracycline, 1 mM IPTG, and 60 mM 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-Gal) at 30°C. The resulting colony phenotypes were compared to the reporter strains bearing the pSR658espM_CT plasmid. We selected colonies which appeared blue on agar with X-Gal. Colonies were grown in LB broth supplemented with 12 μg/mL tetracycline. Using directed DNA sequencing with Genewiz, the mutation rate of the library was determined to be 1 in 357.5 bp using the 46 colonies that had the highest activity of X-Gal relative to the control.

Generation of point mutations using site-directed mutagenesis (SDM).

To generate individual point mutations identified in the espM_CT gene in the pMOPS-espM-V5 and pSR658-espM_Full plasmids, we designed primers (IDT) with the desired mutation flanked by 12 bp on each side as suggested by the QuikChange II site-directed mutagenesis kit (Agilent). The primers used to introduce each mutation into the espM gene are listed in Table S4. Mutagenesis was performed using Pfu Turbo polymerase with the pMOPS-espM-V5 and pSR658-espM-full plasmids as the templates. The PCR cycle included an extension time of 2 min per kb (Agilent) for 16 cycles. The PCR product was treated with DpnI (NEB) for 2 h, heat-killed at 65°C for 10 min, and introduced into DH5α E. coli cells as described above. The resulting plasmids were subjected to targeted DNA sequencing.

Protein expression and purification.

To generate V5-tagged versions of espM, espM_NT, or espM_CT in pET15b, we added C-terminal V5 tags (NPLLGLDST) to each of the pMOPS-espM constructs using primers OMF635 and OMF636 using FastCloning. The presence of V5 tags was confirmed by targeted Sanger sequencing. The V5-tagged versions of EspM were shuttled using restriction digest into the pET15b plasmid. The resulting ligations of 6×His-EspM-V5 (6,577,820 to 6,578,911 bp in the M. marinum genome, based on MycoBrowser annotation [71]), 6×His EspM_NT-V5 (6,577,820 to 6,578,219 bp), and 6×His-EspM_CT-V5 (6,578,198 to 6,578,911 bp) with the pET15b expression plasmid were introduced into chemically competent DH5 alpha E. coli cells as previously described and plated on LB agar with 200 μg/mL ampicillin before incubation at 37°C. Colonies were grown overnight at 30°C or 37°C in LB broth supplemented with ampicillin. Plasmids were extracted using the AccuPrep plasmid mini extraction kit (Bioneer). All constructs were confirmed by restriction digest and Sanger sequencing. All proteins were expressed and purified as previously described using auto-induction medium and Ni-NTA column purification (23).

To remove the MBP and 6×His-tag in the pMF666 plasmid, elution fractions from the primary purifications were combined and mixed with 125 μg of purified TEV and loaded into a 7,000-molecular weight cutoff (MWCO) dialysis tubing. The mixture was dialyzed at 4°C overnight against cleavage buffer (20 mM HEPES, pH 7.0, 300 mM KCl, 5% glycerol, 5 mM β-mercaptoethanol). The dialysis tubing was transferred to reductant-free buffer (20 mM HEPES, pH 7.0, 300 mM KCl, 5% glycerol) and dialyzed at 4°C for 5 h. Cleaved proteins were removed from the dialysis tube, mixed with buffer-equilibrated nickel-nitrilotriacetic acid (Ni-NTA) resin, and batch incubated at 4°C for overnight. Resin was then poured over a p10 column, and the flowthrough was collected for further analysis. Flowthrough containing a protein of interest was concentrated using Vivaspin spin columns (Sartorius) at 15°C at 2,990 rpm.

Electrophoretic mobility shift assays (EMSAs).

The EMSAs were performed as described previously (28) with the following changes. Briefly, a probe containing the intergenic region of espM and whiB6 was generated using the OKS5 and OKS18u primers before being purified on a silicone column (Bioneer). Then, ~35 ng of the resulting DNA probe was incubated with increasing amounts of each of the purified EspM proteins. These were incubated in 0.5× binding buffer (the 2× stock is 10 mM Tris-HCl [pH 8.0, room temperature (RT)], 40 mM KCl, 2.5 mM dithiothreitol [DTT], 0.25% Tween 20, and 0.02 mg/mL salmon sperm DNA) for 15 to 25 min at 4°C before being separated on a nondenaturing, Tris-acetate-EDTA [TAE]-based acrylamide gel that was precleared as previously described (28). Visualization of DNA was performed by incubating the gel for 20 min with SYBR green dye (Thermo Fisher), followed by imaging on a Bio-Rad Gel Doc EZ imager.

LexA bacterial two-hybrid system.

The bacterial two-hybrid system, including plasmids pMS604, pDP804, pSR658, and pSR659 (69), was the generous gift of Timothy Yahr at the University of Iowa. The multimerization of the proteins of interest was measured using a β-galactosidase assay as previously described (23) but with E. coli lysate instead of M. marinum lysate. Briefly, E. coli cultures were grown overnight at 30°C, before being diluted to an optical density at 600 nm (OD₆₀₀) of 0.01. Cultures were grown to an OD₆₀₀ of 0.7 to 1.3. Then, 500 μL of culture was collected and lysed by adding 5 μL of chloroform and vortexing the mixture for 10 to 20 s before incubating it at RT for 10 to 20 min. Next, 1× Z buffer (60 mM Na₂HPO₄ [Millipore Sigma], 40 mM NaH₂PO₄ [Millipore Sigma], 10 mM KCl, 1 mM MgSO₄) was freshly treated with 50 mM β-mercaptoethanol (βME). Ortho-nitrophenyl-β-galactoside (ONPG) (10 mg/mL) reporter solution was made by dissolving ONPG into the 1× Z+βME buffer using a sonicator at 25% amplitude for 20 s. The lysates were added in technical triplicate to a 96-well plate. Each well contained 192.5 μL of 1× Z+βME buffer, 7.5 μL of sample, and 50 μL of ONPG reporter solution. After 12 to 20 min, the plate was read at OD₄₂₀ on a plate reader (Molecular Devices). To calculate the Miller Units, we used the following equation:

$$\text{1\hspace{0.17em}Miller\hspace{0.17em}unit\hspace{0.17em}}=\text{\hspace{0.17em}1,}000\times ({\text{OD}}_{\text{42}0})/({\text{OD}}_{\text{6}00}\cdot \text{volume\hspace{0.17em}}(\text{mL})\cdot \text{time}(\text{minutes}))$$

Size exclusion chromatography.

Multimerization analysis of EspM proteins was performed in the same buffer composition as the protein purification but without imidazole (20 mM HEPES, 300 mM KCl, 2 mM MgCl₂, and 5% glycerol). Analysis occurred using dextran covalently linked to cross-linked agarose on a HiLoad 16/60 Superdex XK column (GE Healthcare), which has a column volume of 120 to 124 mL. EspM was added to the column through a 0.22-μm cellulose acetate filter (VWR) and was added using a 500-μL loading loop. EspM and its variants were analyzed at a flow rate of 0.5 mL/min at a concentration of 1.2 mg/mL for EspM, 1.3 mg/mL for EspMR145C, 0.2 mg/mL for EspMS219F, 0.34 mg/mL for EspML263V, and 1.8 mg/mL for EspMR317C. Primary peaks were visualized and analyzed using Unicorn software (Cytiva Life Sciences). Fractions were eluted into a 2-mL 96-well plate (PlateOne), and fractions corresponding to peaks of interest found on the chromatogram were further analyzed using SDS-PAGE followed by staining with Coomassie as described below. Fractions were taken from each of the major peaks as well as several fractions that did not correspond to a primary peak as a negative control.

Sizing of the protein fractions was done by generating a standard curve using a size-standard kit containing β-amylase (200.0 kDa), alcohol dehydrogenase (150.0 kDa), albumin (66.0 kDa), carbonic anhydrase (29.0 kDa), and cytochrome c (12.4 kDa) (Sigma). The standard curve generated an equation that was used to calculate the size of the eluted proteins. The equation is kDa = 24396e^{−3.487 Ve/Vo}. In this equation V_o is the void volume of the column, and V_e is the elution volume of the fraction of interest. When the spectra were analyzed, the volumes at which the primary peaks from SEC spectra were eluted were identified using the peak imaging feature in Unicorn software, using a baseline of zero. The elution volumes corresponding to these peaks were input into the equation above to identify the size of the eluted proteins in kDa. The sizes obtained from the primary peaks from three biological replicates were averaged to obtain a predicted size for each distinct peak. This was divided by the size of EspM (38.3 kDa) in order to obtain the predicted multimeric state of the protein occupying that peak. Line graphs to visualize the size of the protein multimers were generated by plotting the average log kDa of the eluted protein against the V_e/V_o. Line graph curves were generated using standard linear regression.

Circular dichroism.

The WT EspM EspMR317C, and EspMR145C mutant proteins were dialyzed into 300 mM KCl and 5% glycerol before being diluted to 14 μM (WT and R317C) and 12 μM (R145C), respectively. CD was performed using a Jasco J-815 CD spectrometer. CD was performed at 20°C with a path length of 0.1 cm, a digital integration time (DIT) of 2 s, and a scanning range of 190 to 250 nm at 50 nm/min. Samples were analyzed in a 0.1-cm cuvette (StarnaCells). The data are the result from three accumulated runs and a blank run. For analysis, the blank was subtracted from the accumulated runs. The mean residue ellipticity (θ_MRE) was then calculated using θ_MRE = (M/(c · l · n_r)) · θ_deg, where M is the molecular weight of the protein in g/dmol, c is the concentration in g/mL, l is pathlength in cm, n_r is the number residues, and θ_deg is the readout from the spectrometer in degrees. This calculated MRE with associated wavelength was then entered into the CAPITO online server for further data analysis (42).

Western blot analysis.

Western blot analysis was performed as previously described (28) to detect the V5-tagged EspM and FLAG-tagged WhiB6. Briefly, membranes were blocked in 5% milk in phosphate-buffered saline (PBS) supplemented with 0.1% Tween 20. Monoclonal anti-V5 (1:5,000, Sigma-Aldrich), monoclonal anti-FLAG (1:5,000, Sigma-Aldrich), LexA DNA binding domain (1:20,000, ActiveMotif), and RNA polymerase (RNAP) subunit β (1:10,000, Abcam) antibodies were used in this study. Horseradish peroxidase-conjugated goat and mouse antibodies were used for secondary blotting (1:5,000, Bio-Rad). Protein was visualized using a LumiGLO peroxidase chemiluminescent substrate kit (SeraCare) with detection on X-ray film (RPI). The densitometry to relatively quantitate the V5 antibody signal was performed using ImageJ as follows. Measurements were set to mean gray value, and the same frame was used across all protein bands and backgrounds of the same type (V5 and RNAP). Pixel density was inverted by the formula (255−measurement). Net values were expressed by subtracting the background measurement from the inverted band value. Relative quantitation of protein expression was determined by dividing the V5 net value by the RNAP net value.

qRT-PCR.

Reverse transcriptase quantitative PCR (qRT-PCR) was performed as previously described, but in 7H9 medium instead of Sauton’s medium, as described above (28). Briefly, 500 ng of RNA was treated according to the manufacturer’s instructions (Promega) with the addition of 100 nM CaCl₂ and 50 nM MgCl₂. Then, 1 μL of DNase-treated RNA was converted into cDNA using Superscript II reverse transcriptase according to the manufacturer’s instructions (Invitrogen). cDNA was quantified using a Nanodrop device (Thermo Fisher) and diluted to 250 ng/μL. qRT-PCR was performed using 1 μL of the synthesized cDNA and PowerUP SYBR green master mix according to the manufacturer’s instructions (Applied Biosystems). The ORS225 and ORS226 primers were used to measure the whiB6 transcript at a final concentration of 30 nM. The sigA-F and sigA-R primers were used to measure the sigA transcript at a final concentration of 300 nM. The QuantStudio 3 qRT-PCR machine (Thermo Fisher) was used to run the reaction using standard cycling conditions. For analysis, a ΔΔC_T method was used to compare the levels of each transcript to the M. marinum parental whiB6-Fl strain.

Statistical analysis.

All statistical analysis was performed in using GraphPad Prism 7 and 8. In each case the significance was determined using a one-way ordinary analysis of variance (ANOVA) test followed by either Dunnett’s or Tukey’s multiple-comparison test. The P value for each experiment is specified in the figure legends and the text.