The antagonistic transcription factors, EspM and EspN, regulate the ESX-1 secretion system in M. marinum

ABSTRACT Bacterial pathogens use protein secretion systems to transport virulence factors and regulate gene expression. Among pathogenic mycobacteria, including Mycobacterium tuberculosis and Mycobacterium marinum, the ESAT-6 system 1 (ESX-1) secretion is crucial for host interaction. Secretion of protein substrates by the ESX-1 secretion system disrupts phagosomes, allowing mycobacteria cytoplasmic access during macrophage infections. Deletion or mutation of the ESX-1 system attenuates mycobacterial pathogens. Pathogenic mycobacteria respond to the presence or absence of the ESX-1 system in the cytoplasmic membrane by altering transcription. Under laboratory conditions, the EspM repressor and WhiB6 activator control transcription of specific ESX-1-responsive genes, including the ESX-1 substrate genes. However, deleting the espM or whiB6 gene does not phenocopy the deletion of the ESX-1 substrate genes during macrophage infection by M. marinum. In this study, we identified EspN, a critical transcription factor whose activity is masked by the EspM repressor under laboratory conditions. In the absence of EspM, EspN activates transcription of whiB6 and ESX-1 genes during both laboratory growth and macrophage infection. EspN is also independently required for M. marinum growth within and cytolysis of macrophages, similar to the ESX-1 genes, and for disease burden in a zebrafish larval model of infection. These findings suggest that EspN and EspM coordinate to counterbalance the regulation of the ESX-1 system and support mycobacterial pathogenesis. IMPORTANCE Pathogenic mycobacteria, which are responsible for tuberculosis and other long-term diseases, use the ESX-1 system to transport proteins that control the host response to infection and promote bacterial survival. In this study, we identify an undescribed transcription factor that controls the expression of ESX-1 genes and is required for both macrophage and animal infection. However, this transcription factor is not the primary regulator of ESX-1 genes under standard laboratory conditions. These findings identify a critical transcription factor that likely controls expression of a major virulence pathway during infection, but whose effect is not detectable with standard laboratory strains and growth conditions.

In several Gram-negative pathogens, transcription is responsive to protein secretion systems (2,18,19).We and others have shown that the loss of the membrane-associ ated ESX-1 secretion system in M. marinum and M. tuberculosis resulted in widespread transcriptional changes (20)(21)(22).ESX-1 genes are a subset of the ESX-1 responsive genes.The gene most responsive to ESX-1 was whiB6.WhiB6 activates the transcription of ESX-1 substrate genes (20,23,24).EspM directly represses whiB6 gene transcription in the absence of the ESX-1 components in the CM (20,22).Reduced WhiB6 levels led to reduced ESX-1 substrate gene transcription, preventing accumulation of ESX-1 substrates in the absence of the transport machinery (20).Genetic deletion of the espM gene, which is divergently encoded from the whiB6 gene, derepressed whiB6 transcription and increased levels of ESX-1 substrates in M. marinum during standard laboratory growth conditions (22).
We previously used lacZ+ transcriptional fusions to measure EspM-and WhiB6dependent transcription from the M. marinum whiB6 promoter.β-Galactosidase activity was ~2-fold higher in the absence of both EspM and WhiB6 than in the absence of EspM alone (22), suggesting that whiB6 transcription was activated in the absence of both EspM and WhiB6.We exploited our knowledge of the EspM repressor as a tool to identify an additional regulator of ESX-1 transcription.

EspN activates whiB6 transcription in the absence of EspM
To identify regulators activating whiB6 transcription in the absence of EspM, we previously used the whiB6-espM intergenic region to enrich proteins from lysates from the ΔespM M. marinum strain (22).Mass spectrometry identified MMAR_1626 as the only protein enriched for binding the whiB6-espM intergenic region in the absence of espM [Fig.1A; original data in Table S1 D of reference (22), analyzed again in SI Data set SIA through C].
We hypothesized that EspN activates transcription of the whiB6 gene.To test this hypothesis, we generated an unmarked espN deletion strain using allelic exchange (Fig. S1A and B) (35).The parental M. marinum M strain [wild type (WT)] has a whiB6 allele with a C-terminal 3× FLAG epitope integrated at the whiB6 locus (20).We generated the complementation strain by integrating a copy of the espN gene behind the constitutive mycobacterial optimal promoter at the attL site (Fig. S1C).We measured espN transcrip tion using quantitative reverse transcription PCR (qRT-PCR) (Fig. S2A).The espN transcript was present in the WT strain but absent from the ΔespN strain (P = 0.0002, compared to the WT strain, inset).espN transcription was significantly increased (≥40-fold) in the ΔespN/pespN strain compared to the WT strain (P < 0.0114, Fig S2A).
To evaluate the role of EspN on whiB6 transcription, we measured the whiB6 transcript and protein levels in the ΔespN and ΔespN/pespN strains relative to the WT strain and ΔeccCb 1 strains using qRT-PCR (Fig. 1C).We detected the whiB6 transcript and protein in Significance was determined using ordinary one-way ANOVA (P = 0.0001), followed by Tukey's multiple comparisons test.Significance shown is relative to the ΔespM strain, with additional statistics of interest discussed in the text.****P < 0.0001, ***P = 0.0002 for ΔeccCb 1 , **P = 0.0010, ***P = 0.0002 for ΔespM/pespN, ***P = 0.0009 for ΔwhiB6.Western blots are representative of three independent biological replicates.All qRT-PCRs include at least three independent biological replicates, each in technical triplicate.ANOVA, analysis of variance; au, arbitrary units; qRT-PCR, quantitative reverse transcription PCR; SCP2, sterol carrier protein 2; wHTH, winged helix-turn-helix; WT, wild type.the WT strain (Fig. 1C, lane 1).The ΔeccCb 1 strain does not produce the EccCb 1 protein, a cytoplasmic component of the ESX-1 secretion system (4).The loss of EccCb 1 destabilizes the ESX-1 membrane complex (7,20,36), resulting in transcriptional repression of whiB6 (20,22).Consequently, the WhiB6-Fl protein was not detected in the ΔeccCb 1 strain (Fig. 1C, lane 2) (20).In contrast, whiB6 transcription in the ΔespN and ΔespN/pespN strains was stochastic, with the mean from three experiments not significantly different from the WT strain.The levels of WhiB6-Fl protein in the ΔespN strain and the ΔespN/pespN strain were similar to that in the WT strain (Fig. 1C, lanes 3 and 4).We also tested the levels of espF and esxA transcription in the ΔespN strain and the ΔespN/pespN strain.EspF and EsxA are ESX-1 substrates that are both transcriptionally regulated by WhiB6 (20,24).Similar to the whiB6 transcript, the espF (Fig. S2B) and esxA (Fig. S2C) transcripts were stochastic.From these data, we conclude that whiB6 transcription, and the transcription of espF and esxA was not dependent on EspN under the conditions tested.
We identified both EspM and EspN using DNA affinity chromatography with the whiB6 promoter DNA (22).EspM was specifically enriched from the WT M. marinum lysate, while EspN was only specifically enriched from the ΔespM lysate [Fig.1A; Table D in reference (22), and Data set SIA through C].This led us to hypothesize that EspN regulated whiB6 transcription in the absence of and in opposition to the EspM repressor.To test this, we deleted espN from the ΔespM M. marinum strain.We also introduced an integrating plasmid constitutively expressing espN into the ΔespM strain.Using qRT-PCR (Fig. S2D), we found that the espN transcript was absent from the ΔespMΔespN strain and significantly increased in the ΔespM/pespN strain (~60-to 80-fold, P < 0.0001).We measured changes in the whiB6 transcript and protein levels in the ΔespMΔespN and ΔespM/pespN strains compared to the ΔespM strain (Fig. 1D).Consistent with our prior work, whiB6 transcript was significantly reduced in the ΔeccCb 1 strain (P < 0.0001) and significantly increased in the ΔespM strain relative to the WT strain (P < 0.0001) (20,22).The WhiB6-Fl protein reflected the differences in transcript levels (Fig. 1D, lanes 1 through 3) (22).Deletion and overexpression of the espN gene in the ΔespM strain significantly reduced the whiB6 transcript and WhiB6 protein levels (Fig. 1D, lanes 4 and 5) compared to the ΔespM (P = 0.0002 and P = 0.0009) and WT (P < 0.0001) strains.We conclude that espN overexpression in the ΔespM strain causes a loss of EspN function, similar to deletion of espN (37).Our data support that EspN is required for the elevated whiB6 transcription in the ΔespM strain.Together, our findings suggest that EspN activates whiB6 transcription in the absence of EspM, working in opposition to EspM at the whiB6 promoter.

EspN activates the transcription of ESX-1 component and substrate genes
WhiB6 positively regulates the transcription of several ESX-1 substrate genes (20,24).We therefore tested how regulation by EspM and EspN impacts ESX-1 function during laboratory growth.M. marinum exhibits contact-dependent, ESX-1-dependent hemolytic activity (5,38).We measured sheep red blood cell (sRBC) lysis to define how EspM and EspN impact ESX-1 activity (Fig. 2A).Water and phosphate-buffered saline were "no bacteria" controls, causing total and baseline sRBC lysis as measured by OD 405 , respectively (Fig. 2A).M. marinum lysed sRBCs in an ESX-1-dependent manner (ΔeccCb 1 vs WT, P < 0.0001).Deletion or overexpression of espN did not significantly impact hemolytic activity.The ΔespM strain lysed sRBCs slightly less than the WT strain (P = 0.0122).Although the loss of EspM or EspN alone did not greatly impact hemolytic activity, deletion or overexpression of espN in the ΔespM strain abolished hemolysis (P < 0.0001), similar to the ΔeccCb 1 strain.These findings suggest that EspM and EspN are collectively required for ESX-1-dependent lytic activity.Moreover, EspN is essential for lytic activity only in the absence of EspM.
We sought to determine why EspM and EspN were essential for ESX-1 hemolytic activity.Considering that WhiB6 regulates ESX-1 transcription (20,23,24), if EspM and EspN were essential for lytic activity solely because they regulate whiB6 transcription, then we would expect that the hemolytic activity of the ΔespMΔespN strain would phenocopy the ΔwhiB6 strain.Consistent with our previous findings, the ΔwhiB6 strain retained hemolytic activity [Fig.2A and reference (20)].These data indicate that the loss of WhiB6 alone in the ΔespMΔespN strain is insufficient to explain the loss of hemolytic activity.
To determine what caused the loss of hemolytic activity of the ΔespMΔespN and ΔespM/pespN strains, we measured the production and secretion of two ESX-1 sub strates, EsxB and EspE, using Western blot analysis.EsxB is an early substrate and likely secreted component of the ESX-1 system (4, 9).The loss of EsxB abolishes ESX-1 substrate secretion and hemolytic activity (5,28).EspE, a late substrate, abolishes hemolytic activity but does not affect ESX-1 substrate secretion, except EspF (9,28).EsxB and EspE were produced in (Fig. 2B, lane 1, α-EsxB and α-EspE) and secreted from (lane 9) the WT strain during in vitro growth.The ΔeccCb 1 strain had reduced EsxB and EspE levels (lane 2) due to EspM-dependent repression of whiB6 transcription (22).Neither protein was secreted (lane 10) due to a loss of the ESX-1 membrane components.EspE and EsxB were produced (lane 3) and secreted from the ΔespM strain (lane 11), consistent with our prior findings (22).Deletion or overexpression of the espN gene in the ΔespM strain abolished EspE production (lanes 4 and 5) and therefore secretion (lanes 12 and 13).EsxB was produced in both strains, but EsxB secretion was reduced compared to the WT and ΔespM strains (lanes 12 and 13).
Reduced EsxB secretion could be explained by reduced ESX-1 components in the CM.EccCb 1 is required for the stability of the ESX-1 complex (20).As measured by Western blot analysis (Fig. 2B), the EccCb 1 protein was detected in the cell-associated proteins from the WT strain and absent from the ΔeccCb 1 strain (lanes 1 and 2, lower band).The ΔespM strain produced EccCb 1 protein (lane 3,).In the absence of EspM and EspN, EccCb 1 protein levels were lower than the WT strain (lanes 4 and 5).
The loss of EspE protein and reduced EccCb 1 protein could be explained by reduced transcription.We tested if EspN regulated the transcription of espE or the ESX-1 component genes when EspM was absent.As measured by qRT-PCR (Fig. 2C), eccCb 1 deletion significantly reduced (P < 0.0001) and espM deletion significantly increased (P = 0.0010) espE transcription relative to the WT strain, consistent with Fig. 2B and our previous findings (22,39).Deletion or overexpression of espN in the ΔespM strain abolished espE transcription, significantly different from the ΔespM (P < 0.0001) and the WT strains (P < 0.0001).Importantly, whiB6 deletion significantly reduced (P = 0.0003) but did not abolish the espE transcript.Therefore, the loss of WhiB6 was insufficient to cause the loss of espE transcription in the ΔespMΔespN strain.We conclude that EspN is required for transcription of espE in the ΔespM strain (Fig. 2C).
The eccCb 1 gene is downstream of three other ESX-1 component genes (Fig. 2D; eccA, eccB, and eccCa 1 ).It is not known if the ecc genes are operonic.To test if reduced EccCb 1 protein was due to reduced ecc transcription, we measured eccA transcription using qRT-PCR (Fig. 2D).The eccA transcript was significantly reduced (P < 0.0001) in the ΔeccCb 1 strain and significantly increased (P = 0.0012) in the ΔespM strain as compared to the WT strain.Together, these data suggest that EspM represses eccA transcription.include a whiB6-Fl allele.RpoB is a control for lysis.MPT-32 is a loading control for the secreted fractions.Blot is representative of three independent biological replicates.(C) Relative qRT analysis of the espE transcript compared to sigA transcript levels in M. marinum.Statistical analysis was performed using ordinary one-way ANOVA (P < 0.0001) followed by Tukey's multiple comparisons test.Significance is shown relative to the ΔespM strain.***P = 0.0010 (WT), ****P < 0.0001, ***P = 0.0003 (ΔwhiB6).(D) Relative qRT analysis of the eccA transcript compared to sigA transcript levels in M. marinum.Statistical analysis was performed using ordinary one-way ANOVA (P < 0.0001) followed by Tukey's multiple comparisons test.Significance shown relative to the ΔespM strain.**P = 0.0012 (WT), P = 0.0026 (ΔwhiB6); ****P < 0.0001.(E) Relative qRT-PCR of the espN, espM, whiB6, espE, and eccA transcripts during macrophage infection.RAW 264.7 cells were infected with a multiplicity of infection of 20, and M. marinum strains were isolated at 4 hours post-infection.Outliers were identified using Robust regression and Outlier(ROUT) analysis, Q = 0.05%.Statistical analysis was performed using ordinary one-way ANOVA (P = 0.0004 for ΔespN, P < 0.0001 for ΔespM and ΔespMΔespN) followed by Dunnett's multiple comparisons test relative to the WT strain (dotted line) in each strain.For ΔespN, *P = 0.0352, ΔespM, *P = 0.0320, ****P < 0.0001; for ΔespMΔespN, ****P < 0.0001.For all qRT-PCR, data include three independent biological replicates each in technical triplicate.
Deleting or overexpressing espN in the ΔespM strain significantly reduced eccA transcrip tion compared to the ΔespM (P < 0.0001) and WT strains (P < 0.0001).Interestingly, while the eccA transcript in the ΔwhiB6 strain was significantly reduced compared to the ΔespM strain (P = 0.0026), it was not significantly different from the WT strain.These findings suggest that EspN activates and EspM represses eccA transcription independently of WhiB6.
To test how EspN and EspM impact transcription during infection, we infected RAW 264.7 macrophages with M. marinum strains lacking espM, espN, or both espM and espN.M. marinum can escape the phagosome between 2 and 4 hours post-infection [hpi (40)].Since ESX-1 functions in the phagosome, we harvested the bacteria at 4 hpi and used qRT-PCR to measure whiB6, espN, espM, espE, and eccA transcripts in M. marinum.Transcription of whiB6, espM, espE, and eccA was not significantly different between the WT and ΔespN strains (Fig. 2E), in agreement with Fig. 1C.Similar to our findings during laboratory growth, deletion of espM significantly increased whiB6 (P < 0.0001) and eccA (P < 0.0001) transcript levels relative to the WT strain.The espN and espE transcripts were also higher than the WT strain but did not reach significance.Strikingly, all five transcripts were significantly reduced in the ΔespMΔespN strain compared to the WT strain.These data suggest that EspM and EspN jointly regulate whiB6, espE, and eccA transcription during macrophage infection under the conditions tested.

EspN is required during M. marinum infection of macrophages and zebrafish
Our data propose a model where EspM and EspN counterbalance to regulate ESX-1 gene transcription both in the laboratory and during infection.However, EspM obscures the regulatory role of EspN under laboratory conditions and during macrophage infection.M. marinum spp.cause macrophage cytolysis in an ESX-1-dependent manner.Strains lacking the ESX-1 system remain in the phagosome and are non-cytolytic (10).We infected RAW 264.7 cells with M. marinum (multiplicity of infection of 4) and imaged these 24 hpi using ethidium homodimer (EthD-1) staining to measure macrophage cytotoxicity (Fig. 3A).EthD-1 selectively stains DNA in cells with permeabilized cell membranes, reflecting the cytolytic activity of M. marinum (5,41).Infection with WT M. marinum led to macrophage cytotoxicity (Fig. 3A).Uninfected and ΔeccCb 1 -infected macrophages exhibited significantly less cytotoxicity than WT-infected macrophages (P < 0.0001).Infection with the ΔespN strain significantly decreased macrophage cytotoxicity, similar to the ΔeccCb 1 strain (P < 0.0001).Overexpressing espN in the ΔespN strain (P < 0.0001) partially complemented cytolytic activity.The ΔespMΔespN and ΔespM/pespN strains were non-cytolytic, similar to the ΔespN and ΔeccCb 1 strains.From these data, we conclude that EspN is independently essential for M. marinum to cause macrophage cytotoxicity.
To test if espN was required for mycobacterial growth in macrophages, we performed CFU analysis (Fig. 3B).WT M. marinum grew within macrophages during infection.In contrast, the ΔeccCb 1 strain was attenuated for growth in macrophages, resulting in CFUs significantly different from the WT strain at 72 and 96 hpi.Consistent with the cytotoxic ity data, the ΔespN strain was attenuated for growth in the macrophages, similar to the ΔeccCb 1 strain.Overexpression of the espN gene restored growth of M. marinum similar to the WT strain.From these data, we conclude that EspN is required for M. marinum growth and during macrophage infection.
To assess EspN's importance during animal infection, we made constitutively fluorescent versions of the WT, ΔespN, and complemented M. marinum strains.Using the zebrafish larval model of mycobacterial infection, we monitored burden over 5 days of infection as measured by a validated fluorescence pixel count assay [Fig.3C and D (42,43)].Bacterial burden was substantially reduced at 5 days post-infection.In vivo, we found that ΔespN had 3.4-fold reduced burden compared to WT at 5 days post-infection (P = 0.00012) and was complemented by constitutive expression of the espN gene (Fig. 3C and D).From these data, we conclude that EspN is required during zebrafish infection Significance was determined using ordinary two-way ANOVA (P < 0.0001) followed by Tukey's multiple comparisons test compared to the WT strain.The significance shown is compared to the WT strain at 96 hours post-infection (P < 0.0001).The ΔeccCb 1 and ΔespN strains were also significantly different from the WT strain at 72 hpi (P = 0.0021 and P = 0.0024, respectively).(B) M. marinum burden in zebrafish infection measured using bacterial mCerulean fluorescence.Data are composed of two biological replicates with 20-30 independent infections per replicate.Statistical analyses were performed using one-way ANOVA followed by Tukey's multiple comparisons of each group to the WT strain (***P = 0.00012, *P = 0.021).(C) Representative images of zebrafish infected with an initial dose of 150-200 fluorescent bacilli for (1.) WT, (2.) ΔespN, or (3.) ΔespN/pespN at 5 days post-infection.Scale bar is 500 µm.(D) Macrophage cytolysis as measured by EthD-1 staining 24 hours post-infection with M. marinum at an MOI of 4. Statistical analysis was performed using one-way ANOVA followed by Dunnett's multiple comparisons test relative to the WT strain (****P < 0.0001).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 three wells was performed for each biological replicate.A total of 90 fields were counted for each strain.CFU, colony-forming unit; ns, not significant.by M. marinum.Together, our data support that EspN is essential for mycobacterial infection.

Overexpression of EspN causes loss of ESX-1 function
We found it curious that espN overexpression in the ΔespM strain phenocopied the ΔespMΔespN strain during laboratory growth and macrophage infection.Our data argue against an unidentified mutation as the reason for the shared phenotypes.espM expression in the ΔespM strain complements all phenotypes associated with EspM loss.We performed whole-genome DNA sequencing on the WT, ΔespM, ΔespN, ΔespMΔespN, and the ΔespM/pespN strains (Data set S2).No consistent mutations were identified between the ΔespMΔespN and ΔespM/pespN strains that could account for the observed phenotypes.
We hypothesized that espN overexpression was affecting endogenous EspN activity.We designed mutations in the mycobacterial optimal promoter focusing on −7 and −12 positions (Fig. 4A).We confirmed the mutations by DNA sequencing, introduced the pespN plasmids into the ΔespM or ΔespMΔespN M. marinum strains, and measured espN expression using qRT-PCR.The mutated plasmids resulted in significantly reduced espN expression compared to the parental plasmid (Fig. 4B; WT, P < 0.0001).Although overexpression of espN abolished the hemolytic activity of the ΔespM strain (P < 0.0001), reduced espN expression did not significantly alter hemolysis of the ΔespM strain (Fig. 4C).From these data, we conclude that the espN overexpression caused the loss of hemolysis in the ΔespM strain.Notably, reduced espN expression from the mutated promoter did not restore hemolysis of the ΔespMΔespN strain (Fig. 4C).

EspE and the N-terminus of EspM are linked to EspN function
To understand the transcriptional network (Fig. 5A), we examined regulatory interactions between EspM and EspN.Assessing espN transcription in the ΔespM strain and espM transcription in the ΔespN strain (Fig. S4A and B) revealed no significant changes.We conclude that EspM and EspN do not regulate each other transcriptionally under the conditions tested.
Because EspN overexpression abolished hemolytic activity and ESX-1 gene expression in the ΔespM strain, we asked if overexpressing other regulators impacted ESX-1 function in the absence of EspM.EspM has an N-terminal forkhead-associated (FHA) domain (EspM NT , Fig. 5B) and two HTH domains between a helical bundle (EspM CT in Fig. 5B) at the C-terminus.We constitutively expressed the espM, espM NT , and espM CT genes including a C-terminal V5 epitope tag in the ΔespM strain.All EspM-V5 proteins were expressed in M. marinum (Fig. S5A, lanes 4-6).espM-V5 expression in the ΔespM strain did not significantly impact hemolytic activity (Fig. 5C).espM NT -V5 expression abrogated hemolytic activity of the ΔespM strain (P < 0.0001).espM CT -V5 expression in the ΔespM strain significantly impacted hemolytic activity (P = 0.0034), causing stochasticity relative to the other strains.These data suggested that the ΔespM/pespM NT -V5 strain effectively phenocopied the ΔespMΔespN strain regarding ESX-1-dependent hemolytic activity (Fig. 3B).
EspE and EspF are dual-functioning ESX-1 substrates.Both proteins negatively regulate WhiB6 activity in M. marinum, and their secretion is required for hemolytic activity and for virulence (28).Overexpressing espE or espF did not impact hemolytic activity (Fig. 4E) but significantly reduced whiB6 transcription in the ΔespM strain (Fig. 4F).However, while espE overexpression reduced whiB6 transcription to levels below the WT strain, overexpression of espF resulted in transcription levels ~20× higher than the WT strain.From these data, we conclude that overexpression of EspE and EspM NT , but not EspF, results in phenotypes consistent with EspN loss of function in the absence of EspM.

DISCUSSION
Our prior studies hinted at the existence of an additional ESX-1 activator (22).Here, we discovered and characterized EspN, a transcriptional regulator of the ESX-1 system that is essential for infection.Our data suggest that EspN and EspM function as a switch that regulates the transcription of ESX-1 genes.When the ESX-1 system is absent, EspM represses whiB6, ESX-1 component (EccA and others), and substrate genes, preventing substrate accumulation in the absence of secretion (20,22).EspM also regulates the transcription of additional genes in M. marinum (22).In the presence of the ESX-1 system, WhiB6 activates ESX-1 substrate gene transcription and other genes, allowing production of substrates during active secretion (20,23,24).The presence of EspM masked a role for EspN in regulating ESX-1 gene transcription.However, in the absence of EspM, EspN is essential for ESX-1 activity likely because it regulates espE transcription and contributes to eccA transcription, optimizing ESX-1 component production and substrate secretion.Deleting espN attenuated the cytolytic activity and growth of M. marinum during macrophage infection, similar to a loss of ESX-1.EspN was likewise necessary for robust infection in zebrafish larvae.Our findings suggest that EspN is required for ESX-1 function during infection possibly due to transcriptional regulation of ESX-1 genes.Deleting both espM and espN from M. marinum abolished cytolytic activity and transcription of whiB6, espE, and eccA during macrophage infection, which differed from deleting espN alone.Finally, overexpression of espN, espM NT , and espE specifically disrupted ESX-1 regulation, demonstrating connections in the transcriptional network.Together, these data demonstrate a critical role for EspN and the ESX-1 transcriptional network during infection.
Our data suggest that EspN may function both independently and together with EspM to regulate transcription and to promote pathogenesis.espN deletion did not have regulatory phenotypes under laboratory conditions or during macrophage infection under the conditions we tested.However, the ΔespN strain was attenuated both in macrophage and in zebrafish larvae.EspN-dependent regulation of ESX-1 or additional genes may be essential for virulence.The deletion of both EspN and EspM abolished ESX-1 transcription under laboratory conditions and during macrophage infection, which differed from deleting either regulator alone.Defining the regulatory targets of EspN, as well as how the EspM and EspN regulons compare, will likely require measuring global gene expression during specific stages of infection.Some of our strains exhibited variable transcript levels (the ΔespN and ΔespM strains; Fig. 1C, D, 2D, E, 5C and D) or hemolytic activity (ΔespM, Fig. 2A; ΔespM/pespM CT , Fig. 5F) across multiple experiments.Our studies relied on population averages for each assay.In regulatory systems such as Type III secretion-dependent transcriptional regulation in P. aeruginosa, persister cells in Escherichia coli, and cannibalism and competence in B. subtilis, stochasticity reflects that individual cells within a population have different gene expression patterns (45,46).This "bistability" indicates a switch between two states rather than intermediate ones (45).The stochasticity of our results may suggest that ESX-1 regulatory network includes a bistable switch that operates early during infection, possibly through pairs of mutually exclusive regulators or positive autoregulation (45).analysis was performed using one-way ANOVA (P < 0.0001) followed by Dunnett's multiple comparisons test (**P = 0.0034, ****P < 0.0001).The data include at least three independent biological replicates each in technical triplicate.(D) Relative qRT-PCR analysis of whiB6 compared to sigA transcript levels.Significance was determined using ordinary one-way ANOVA (P < 0.0001), followed by Dunnett's multiple comparisons test (****P < 0.0001) relative to the ΔespM strain.
The qRT-PCR data include at least three independent biological replicates each in technical triplicate.(E) sRBC lysis measuring hemolytic activity of M. marinum.
Outliers were identified using ROUT analysis, Q = 0.05%.Statistical analysis was performed using ordinary one-way ANOVA (P = 0.9639), which did not indicate significant differences.The data include at least three independent biological replicates each in technical triplicate.(F) Relative qRT-PCR analysis of whiB6 compared to sigA transcript levels.Outliers were identified using ROUT analysis (Q = 0.5%).Significance was determined using ordinary one-way ANOVA (P < 0.0001), followed by Dunnett's multiple comparisons test (**P = 0.0070, ****P < 0.0001) relative to the ΔespM strain.The qRT-PCR data include at least three independent biological replicates each in technical triplicate.FHA, forkhead-associated domain; HTH, helix-turn-helix; NT, N-terminus.
While the molecular nature of the switch is unknown, EspM and EspN likely cannot simultaneously occupy the whiB6, eccA, or espE promoters.Host-specific signals may regulate the switch, including oxidative stress, pH, and other phagosomal cues.espN is divergently transcribed from a putative methyltransferase gene (MMAR_1627) (32).We do not know yet if the genes surrounding espN are important for regulation of ESX-1; methylation has been shown to control bacterial genetic switches (47,48).
ESX-1 expression varies in different strains under laboratory conditions (24).Consistent with our model, ESX-1 gene expression is upregulated in the host (49), and transcriptomic studies in M. marinum suggested differential regulation of virulence genes in a variety of environments (50).The differential regulation of protein secretion systems under laboratory conditions and in the host is a common theme in protein secretion in bacterial pathogens.For example, some clinical Vibrio cholerae strains have active Type VI systems under laboratory conditions, while the pandemic strains only activate their Type VI systems in the host (51).The specialized Type III secretion systems in Vibrio species are regulated by bile salts (52) and by Ca 2+ and host cell contact in Pseudomonas (53)(54)(55).
Our findings suggest that EspN activity or expression might respond to host-specific signals.Notably, transcriptional profiling of M. tuberculosis from infected macrophages isolated from mouse infections revealed significant upregulation of the espN (Rv1725c) transcript compared to broth-grown culture (56).At 1 week post-infection in a mouse intravenous model, a TnSeq screen identified a ~2-fold decrease in representation of transposon insertions in M. tuberculosis Rv1725c [P = 0.0039 (57)].However, at 4 and 8 weeks, this effect was no longer present (57,58).Together, these data suggest that complex regulatory circuits and genetic requirements operate at discrete stages of infection.EspN's C-terminal SCP2 domain (Fig. 1B) could differentially localize the protein in the mycobacterial cell in response to the host environment (59,60) to mediate interaction with the mycobacterial cell membrane under specific conditions.SCP2 domains also mediate lipid transfer of sterols such as cholesterol and fatty acids (61,62), which are an energy source for mycobacterial pathogens during infection (63,64).Focusing on EspN's SCP2 domain may reveal how EspN senses and responds to the host environment.
Gene dosage is important for biological function across organisms.Increases in copy number or expression of wild-type genes can cause mutant phenotypes, such as aggregation or mislocalization (37).We were surprised that the overexpression of espN, espM NT , and espE in the absence of espM resulted in the same regulatory phenotypes in M. marinum as deleting espN from the ΔespM strain.espN, espM NT , or espE overexpression could result in EspN aggregation or mislocalization.Transcription factor multimerization is a regulatory mechanism in bacterial gene expression (65,66).In higher organisms, transcription factors can form aggregate-like bodies that serve as functional regulatory mechanisms which mimic loss of function (67).EspN may form multimeric aggregates that prevent DNA interaction.Alternatively, EspN may be mislocalized under overexpres sion conditions.The EspM NT is a predicted FHA domain.Proteins with FHA domains regulate other Gram-negative secretion systems post-transcriptionally (68)(69)(70).Our data also suggest that EspM may be processed in vitro.EspM cleavage might remove it from the promoter, similar to the cI repressor from the λ phage (65).Alternatively, cleavage may liberate the N-terminus to regulate EspN activity, allowing the EspM CT to bind DNA and repress gene expression.
Our studies were limited by an inability to complement espN expression in the ΔespMΔespN strain.We tested overexpression, endogenous, and inducible promoters to restore espN transcription in the ΔespMΔespN strain.We were unable to restore transcription to the levels of the ΔespM strain.We suspect that architecture or chromoso mal location of espN may be important for proper regulation and expression.
Overall, this study further defined the regulatory network underlying control of the ESX-1 secretion system and demonstrated its importance during infection.For the first time, we have identified an infection-dependent transcriptional activator responsible for regulating both the ESX-1 components and substrates.Our study will serve as a foundation for understanding the molecular complexities of ESX-1 regulation in the host.We are now poised to define the molecular mechanisms underlying how the ESX-1 system senses and responds to a changing host environment.

MATERIALS AND METHODS
Bacterial strains were derived from the M. marinum M parental strain (ATCC BAA-535) and were maintained as previously described (20,22,28,84).Nomenclature follows the conventions proposed by Bitter et al. (34).Genetic deletions were performed using allelic exchange as previously described (20,22,28,84,85).Hemolytic activity was measured against sRBCs as described previously (20,22,28,84).Cell-associated and secreted mycobacterial proteins were isolated and analyzed as described in references (28,84).Protein levels were measured using Western blot analysis.RNA extraction was performed from M. marinum using the Qiagen RNeasy kit, followed by qRT-PCR relative to the levels of sigA as described previously (22,44).Macrophage (RAW 264.7) cytotoxicity was measured using ethidium homodimer uptake following infection by M. marinum as described in references (39,84).Mycobacterial CFUs were obtained similarly to reference (28).Protein modeling was performed using Robetta and Pfam as indicated.Zebrafish larvae infections with M. marinum were performed, and bacterial burden was measured using fluorescent pixel counts as reference (42).Statistical analysis was performed using GraphPad Prism v.9 or R within the latest version of R Studio IDE.Detailed methods are in the supplemental material.

FIG 1
FIG 1 EspN binds the whiB6 promoter and activates whiB6 expression in the absence of EspM.(A) Mass spectrometry analysis of the DNA affinity chromatogra phy showing enrichment of the MMAR_1626 and HupB proteins.The HupB protein binds non-specifically to both DNA probes.The scale represents the log2 intensity of Mass Spectral peak area (MS peak areas).The data were published in Sanchez et al. (22) and adapted in Data set S1. (B) The predicted domain structure of MMAR_1626, which we renamed EspN.Modeled using RoseTTAFold from Robetta (29).Model confidence: 0.80.(C) (Top) Western blot analysis of M. marinum cell-associated proteins.RpoB is a loading control.All strains include a whiB6-3xFl allele at the whiB6 locus (20).(Bottom) Relative qRT-PCR analysis of M. marinum strains compared to sigA transcript levels.Statistical analysis was performed using one-way ANOVA (P = 0.0359), followed by a Dunnett's multiple comparison test, which revealed no significant differences relative to the WT strain.(D) (Top) Western blot analysis of 10 µg of M. marinum whole-cell lysates.RpoB serves as a loading control.All strains include a whiB6-3xFl allele at the whiB6 locus (20); (bottom) qRT-PCR of the whiB6 transcript relative to sigA.

FIG 2
FIG 2 EspN and EspM control transcription of ESX-1 components and substrates.(A) Sheep red blood cell lysis measuring hemolytic activity of M. marinum.Statistical analysis was performed using ordinary one-way ANOVA followed by Dunnett's multiple comparisons test relative to the WT strain.****P < 0.0001, (Continued on next page)

FIG 3
FIG 3 EspN is required for pathogenesis.(A) CFU of M. marinum strains (MOI = 0.2).The data points are the average of three independent biological replicates.

FIG 4
FIG 4 High levels of EspN transcription are required for dominance in the ΔespM strain.(A) Schematic of the −10 region from the mycobacterial optimal promoter driving espN transcription.Pink residues are mutations in the −7 and −12 positions.(B) Relative qRT analysis of the espN transcript compared to sigA transcript levels in M. marinum.Outliers were identified using ROUT analysis, Q = 0.5%.Statistical analysis was performed using ordinary one-way ANOVA (P < 0.0001) followed by Dunnett's multiple comparisons test.Significance is shown relative to the ΔespM/pespN or ΔespMΔespN/pespN strain.****P < 0.0001.Data include three biological replicates each in technical triplicate.(C) sRBC lysis measuring hemolytic activity of M. marinum.Statistical analysis was performed using ordinary one-way ANOVA followed by Dunnett's multiple comparisons test relative to the ΔespM or ΔespMΔespN strain.****P < 0.0001.Data include three biological replicates each in technical triplicate.