Integrated transcriptomic and proteomic analysis of pathogenic mycobacteria and their esx-1 mutants reveal secretion-dependent regulation of ESX-1 substrates and WhiB6 as a transcriptional regulator

The mycobacterial type VII secretion system ESX-1 is responsible for the secretion of a number of proteins that play important roles during host infection. The regulation of the expression of secreted proteins is often essential to establish successful infection. Using transcriptome sequencing, we found that the abrogation of ESX-1 function in Mycobacterium marinum leads to a pronounced increase in gene expression levels of the espA operon during the infection of macrophages. In addition, the disruption of ESX-1-mediated protein secretion also leads to a specific down-regulation of the ESX-1 substrates, but not of the structural components of this system, during growth in culture medium. This effect is observed in both M. marinum and M. tuberculosis. We established that down-regulation of ESX-1 substrates is the result of a regulatory process that is influenced by the putative transcriptional regulator whib6, which is located adjacent to the esx-1 locus. In addition, the overexpression of the ESX-1-associated PE35/PPE68 protein pair resulted in a significantly increased secretion of the ESX-1 substrate EsxA, demonstrating a functional link between these proteins. Taken together, these data show that WhiB6 is required for the secretion-dependent regulation of ESX-1 substrates and that ESX-1 substrates are regulated independently from the structural components, both during infection and as a result of active secretion.

Introduction Mycobacteria use several different type VII secretion systems (T7S) to transport proteins across their thick and waxy cell envelopes. One of these T7S systems, ESX-1, is responsible for the transport of a number of important virulence factors. Disruption of the esx-1 gene cluster severely reduces the virulence of Mycobacterium tuberculosis [1], whereas restoration of esx-1 in the Mycobacterium bovis-derived vaccine strain BCG, which lacks part of the esx-1 region due to continuous passaging, leads to increased virulence [2]. Many studies have attempted to elucidate the function of ESX-1 substrates in virulence. In the case of pathogenic mycobacteria, such as M. tuberculosis and the fish pathogen Mycobacterium marinum, ESX-1 is responsible for the translocation of the bacteria from the phagolysosomal compartments to the cytosols of macrophages [3][4][5]. This translocation activity has been attributed to the ESX-1 substrate EsxA (previously also known as ESAT-6) [6,7]. Interestingly, a closely related homologue of this protein is also secreted by non-pathogenic and non-translocating mycobacteria, such as Mycobacterium smegmatis. A report indicated that, although the EsxA proteins of M. smegmatis and M. tuberculosis are highly homologous, the membrane lysis potentials of these proteins are different [8]. In M. smegmatis, ESX-1 is involved in a completely different process, i.e., conjugative DNA transfer [9]. The direct involvement of EsxA in phagosomal rupture is not undisputed, a recent report indicated that ESX-1-mediated cell lysis occurs through gross disruptions at points of bacterial contact and not through pore formation by EsxA (5). Interestingly, there are more differences between pathogenic and non-pathogenetic mycobacteria, one of which is the presence/absence of the espACD operon, which is exclusively present in mycobacterial species that were shown to be able to induce phagosomal rupture in host-phagocytes such as M. kansasii Type I, M. bovis, M. tuberculosis [10]. Possibly also other ESX-1 substrates are involved in membrane disruption. However, these other substrates could also be involved in other proposed functions of ESX-1 in pathogenic mycobacterial species, including host cell entry and intercellular spread [11][12][13].
The esx-1 locus contains both genes that code for the structural components of the ESX-1 secretion system [14], genes that code for accessory protein [15]and genes that code for ESX-1 substrates, including EsxA, EsxB (also known as CFP-10), PPE68, EspE, EspF, EspK and EspB (reviewed in [16,17]). Two of the exceptions are EspA and EspC [18,19], which are both part of the espA operon, which is located elsewhere in the genome. However, these genes are homologous to the espE and espF genes, respectively, which belong to the esx-1 locus. A peculiar characteristic of ESX-1 substrates is that these substrates are mutually dependent, i.e., the secretion of each of these substrates is dependent on the secretion of the other substrates [18]. The secreted ESX proteins contain a conserved WxG amino acid motif located between two αhelices [20]. In addition, these substrates also contain a conserved secretion signal, present in all secreted protein pairs [21]. This C-terminal YxxxD/E motif that is immediately following the helix-turn-helix domain is targeting these proteins for secretion, but does not determine the specificity for a particular type VII system [22]. Therefore, it remains difficult to bioinformatically predict ESX-1 substrates.
To establish successful infection, mycobacteria need regulatory mechanisms to express the right proteins at the right time. In different environments, mycobacteria require specific transcriptional responses to successfully respond to the stress conditions encountered. During the first stages of infection, ESX-1-mediated protein secretion is one of the most important virulence mechanisms of pathogenic mycobacteria [16,17]. Consequently, tight transcriptional regulation of esx-1 and the associated genes is required. The transcriptional regulator PhoP of the two-component system PhoPR positively regulates the transcription of many esx-1-associated genes, including genes in the espA operon [23][24][25]. It has been proposed that PhoP regulation is dependent on environmental pH [26], which could indicate that the acidic environment of the phagosome induces esx-1 gene transcription via PhoP, leading to bacterial escape from this compartment. Other studies have shown that the espA operon is, in addition to PhoP, also regulated by the transcription factors EspR and MprAB and the repressors CRP and Lsr2, indicating that tight regulation of this operon is essential and, furthermore, suggesting that the espA operon may be regulated in a manner distinct from the regulation of other ESX-1 substrates [27][28][29][30].
Here, we apply RNA-seq and quantitative proteomics to determine the gene expression and proteomic profiles of the pathogenic mycobacteria M. marinum and M. tuberculosis in the absence of a functional ESX-1 secretion system. During short-term infection of macrophages, we observed highly increased transcript levels of the espA operon. In contrast, during in vitro growth in culture medium, transcription of most ESX-1 substrates and some putative new substrates was seen to be decreased. Among the putative regulatory protein that was down-regulated in esx-1 mutant strains of both M. marinum and M. tuberculosis, we confirmed a regulatory role for the putative transcriptional regulator WhiB6 in the gene expression of ESX-1 substrates.

Bacterial strains and growth conditions
We used strain M. marinum E11 for our experiments [31,32]. The esx-1 mutants of the M. marinum E11 wild-type strain used in this study contain transposon insertions in eccB 1 , eccCa 1 , eccCb 1 , eccD 1 and eccE 1 and have been described previously ([33]; Stoop et al, doctoral thesis 2013, "Mycobacterium marinum zebrafish embryo screen identifies polyphosphate kinase 1 (Ppk 1) as an important factor for virulence and granuloma formation"). The whiB6 mutant was created in the M. marinum M strain, which is more amenable to genetic manipulation. As parent strain we used the WT MUSA strain and the spontaneous eccCb1 mutant Mvu [34,35]. The whiB6 mutant was created using the phage method described by Bardarov S [36] and the primers WhiB6 KO listed in S8 Table to produce the whiB6 flanking regions. Mutants were checked using PCR and sequencing of the amplified fragments. For M. tuberculosis, the attenuated double-deletion strains mc 2 6020 and mc 2 6030 of H37Rv were used, with deletions of lysA and panCD and of RD1 and panCD, respectively [37,38]. Bacterial strains were grown with shaking at 30˚C (M. marinum) or 37˚C (M. tuberculosis) in Middlebrook 7H9 culture medium supplemented with 10% ADC (albumin-dextrose-catalase, BD Biosciences) and 0.05% Tween-80. Culture medium containing the auxotrophic M. tuberculosis deletion strains was supplemented with 50 μg/ml pantothenic acid and, for mc 2 6020, 100 μg/ml L-lysine.

Infection of human macrophages
THP-1 monocytes were cultured at 37˚C in 5% CO 2 in RPMI-1640 with GlutaMAX-1 (Gibco) supplemented with 10% FBS, 100 μg/ml streptomycin and 100 U/ml penicillin. Cells were seeded at a density of 3 × 10 7 cells per T175 flask and differentiated into macrophages by 48 hours of incubation with 25 ng/ml PMA (Sigma-Aldrich). Then, 1.8 × 10 8 THP-1 cells were infected with M. marinum at a multiplicity of infection (MOI) of 20 for 2 hours, after which the cells were washed with PBS to remove extracellular bacteria. After 4 additional hours of infection at 33˚C, the THP-1 cells were lysed with 1% Triton X-100. After a low-speed centrifugation step to remove cellular debris, mycobacteria were pelleted, after which RNA was extracted as described in the following section. and the plasmid) with HGAP assembler [40] using the default parameters. The sequence was improved with iCORN2 [41] with three iterations, correcting 20 single base pair errors and 61 insertions and deletions. To transfer the annotation from the current reference, we used RATT [42] with the PacBio parameter. Gene models around gaps were manually improved on the new sequence. The updated genome annotation was resubmitted under the same accession numbers (HG917972 for the M. marinum E11 main chromosome genome and HG917973 for the M. marinum E11 pRAW plasmid; complete sequences).

RNA extraction and qRT-PCR
M. marinum and M. tuberculosis cultures were pelleted and bead beated in 1 ml of TRIzol (Invitrogen) with 0.1-mm zirconia/silica beads (BioSpec Products). After centrifugation, supernatants were extracted with chloroform, and RNA was precipitated with isopropanol. RNA pellets were washed with 80% ethanol and dissolved in RNAse-free water. Contaminant DNA was removed by incubation with DNAse I (Fermentas). For RT-PCR, cDNA was generated using a SuperScript VILO cDNA Synthesis Kit (Invitrogen). An equivalent of 5 ng of RNA was used in the quantitative PCRs. qRT-PCR was performed using SYBR GreenER (Invitrogen) and a LightCycler 480 (Roche) instrument. Transcript levels were normalized to the levels of the housekeeping gene sigA [43] using ΔΔCt analysis. All primer sequences used for qRT-PCR are listed in S8 Table. RNA preparation for Illumina sequencing Total RNA was extracted with TRIzol (Invitrogen) and then purified on RNeasy spin columns (Qiagen) according to the manufacturer's instructions. RNA integrity (RNA integrity score�6.8) and quantity were determined on an Agilent 2100 Bioanalyzer (Agilent; Palo Alto, CA, USA). As ribosomal RNA constitutes a vast majority of the extracted RNA population, depletion of these molecules via RiboMinus-based rRNA depletion was conducted. For mRNA enrichment, Invitrogen's RiboMinus Transcriptome Isolation Kit, bacteria was used according to manufacturer's instructions. Briefly, 2μg of total RNA samples was hybridized with prokaryotic rRNA-sequence-specific 5 0 -biotin-labelled oligonucleotide probes to selectively deplete large rRNA molecules from total RNA. Then, these rRNA-hybridized, biotinylated probes were removed from the sample with streptavidin-coated magnetic beads. The resulting RNA sample was concentrated using the RiboMinus concentrate module according to the manufacturer's protocol. The final RiboMinus RNA sample was subjected to thermal mRNA fragmentation using the Elute, Prime, Fragment Mix from the Illumina TruSeq RNA Sample Preparation Kit v2 (Low-Throughput protocol). The fragmented mRNA samples were subjected to cDNA synthesis using the Illumina TruSeq RNA Sample Preparation Kit (lowthroughput protocol) according to the manufacturer's protocol. Briefly, cDNA was synthesized from enriched and fragmented RNA using SuperScript III reverse transcriptase (Invitrogen) and the SRA RT primer (Illumina). The cDNA was further converted into double-stranded DNA using the reagents supplied in the kit, and the resulting dsDNA was used for library preparation. To this end, cDNA fragments were end-repaired and phosphorylated, followed by adenylation of 3 0 ends and adapter ligation. Twelve cycles of PCR amplification were then performed, and the library was finally purified with AMPure beads (Beckman Coulter) as per the manufacturer's instructions. A small aliquot (1 μl) was analysed on an Invitrogen Qubit and an Agilent Bioanalyzer. The bar-coded cDNA libraries were pooled at equal concentrations before sequencing on an Illumina HiSeq2000 using the TruSeq SR Cluster Generation Kit v3 and TruSeq SBS Kit v3. Data were processed with Illumina Pipeline software v1.82.

RNA-seq analysis
The Illumina reads were mapped with SMALT (http://www.sanger.ac.uk/science/tools/smalt-0) (default parameters) against the new PacBio reference. From the read count, which was obtained with bedtools ( [44], parameter multicov, with -D to include duplicates and -q 5 to exclude repetitive mapping reads), we performed a differential expression analysis with DESeq [45] using default parameters.

Plasmid construction
The E. coli-mycobacterial shuttle vector pSMT3 was used for the construction of all plasmids. To overexpress PE35-PPE68_1 (MMARE11_01740-MMARE11_01750), we used a previously described plasmid [22]. For construction of the plasmid containing espG 1 , this gene was amplified from the M. marinum E11 genome by PCR using primers containing NheI and EcoRV restriction sites and a 3' HA epitope. The resulting PCR product and empty pSMT3 were digested with NheI and EcoRV followed by ligation of espG 1 into the vector by T4 ligase (Fermentas). For construction of the plasmid containing whib6, this gene was amplified from the M. marinum E11 genome by PCR using primers containing NheI and BamHI restriction sites. For the other construct, espI was amplified from the M. marinum E11 genome by PCR using primers containing NheI and BglII restriction sites. The PCR product was digested with NheI and BamHI. Empty pSMT3 was digested with NheI and BamHI, after which the PCR product was ligated into the vector. All plasmids were introduced into the M. marinum wild-type E11 and isogenic eccCb 1 mutant strains by electroporation. All primer sequences are listed in S8 Table.

Analysis of protein expression and secretion
M. marinum cultures were grown to mid-logarithmic phase in 7H9 culture medium supplemented with 0.2% glycerol and 0.2% dextrose. Bacteria were pelleted, washed in PBS and incubated in 0.5% Genapol X-080 (Sigma-Aldrich) for 30 minutes to extract cell wall proteins. Genapol X-080-treated M. marinum cells were disrupted by sonication. Secreted proteins were precipitated from the culture supernatant by 10% trichloroacetic acid (TCA, Sigma-Aldrich). Proteins were separated according to molecular weight on 15% SDS-PAGE gels and subsequently transferred to nitrocellulose membranes (Amersham Hybond ECL, GE Healthcare Life Sciences). Immunostaining was performed with mouse monoclonal antibodies directed against the HA epitope (HA.11, Covance), EsxA (Hyb76-8), or rabbit polyclonal sera recognizing EspE [46].

LC-MS analysis
Peptide preparation from the M. marinum E11 and isogenic eccCb1 mutant strains was performed as previously described [47]. Approximately 100-μg protein digests of each sample were labelled with 4plex iTRAQ reagents (Applied Biosystems). The combined iTRAQlabelled samples were fractionated using strong cation exchange chromatography. The eluted fractions were dried and desalted using a Sep-Pak C-18 SPE cartridge (Waters, Milford, MA, USA). LC-MS analysis as well as MS data processing was carried out following our published procedure [48]. Briefly, each fraction was analysed three times using an LTQ-Orbitrap Velos (Thermo Scientific). The MS spectra were recorded in the Orbitrap, whereas the MS2 spectra were recorded in the c-TRAP for HCD fragmentation and in the LTQ for the CID fragmentation. Both HCD and CID spectra were extracted separately using Proteome Discoverer software and processed by an in-house script before a Mascot search against the M. marinum E11 proteome. The Mascot results (.dat file) were processed by Scaffold software for validation of protein identification and quantitative assessment. For protein identification, local false positive rates (FDR) were maintained below 1% for both protein and peptide identification (0.91% and 0.9% for peptides and proteins, respectively, for this dataset). Protein quantitation was processed using Scaffold Q+, which is based on the i-Tracker algorithm [49]. The iTRAQ quantitation using HCD is highly accurate, and a change of more than 2-fold was considered significant differential expression in this study.

Global features of the M. marinum eccCb1 mutant transcriptome and proteome
To investigate the effect of ESX-1 disruption on gene expression and protein production, RNA and protein were extracted from three independent exponential phase cultures of the M. marinum E11 strain and the isogenic esx-1-mutant during growth in 7H9 culture medium to characterize the transcriptome and proteome. Using transcriptomics (RNA-seq) and mass spectrometry (MS)-based proteomics with isobaric labelling for quantification, we captured the expression dynamics of the transcripts and proteomes of the eccCb1 mutant. Data quality was assessed using Euclidean distance matrices for RNA (S1 Fig Table). To determine parallel changes in protein levels, 1,657 proteins were identified by the presence of 2 or more peptides, of which 576 proteins passed our filter and we classified them as DE. Of these, 412 proteins were found to be down-regulated and 164 were up-regulated (S2 Table), and 482 protein-coding genes were shared and were identified in both the RNA-seq and quantitative proteomic datasets ( Fig 1C).
The degree of global correlation between the gene expression and protein abundance scores among the shared genes was relatively low (S3A Fig), which has also been noted in other bacterial studies [50]. However, within certain classes of M. marinum functional categories (http:// mycobrowser.epfl.ch/marinolist.html), the degree of correlation was much higher than that in other classes, with R 2 exceeding 0.8 for the categories such as lipid metabolism (Fig 1D), regulation ( Fig 1E) and cell wall and cell processes ( Fig 1F). Of the DE genes at the RNA and protein levels, 28% were in the intermediary metabolism and respiration category, 18% were in the cell wall and cell process category, 15% were in the information pathways category and 14% were in the lipid metabolism category (S4 Fig).
Transcriptional profiling analysis of the double auxotrophic M. tuberculosis mc 2 6020 mutant strains [38] and their isogenic esx-1 mutants during growth was carried out to identify genes for which expression was dependent on ESX-1 disruption (Fig 1B and S3 Table). For this species, the same trends could be identified as for M. marinum.

Major effects of esx-1 mutation on genes encoding ESX-1 substrates and biosynthetic pathways
Analysis of differential expression in M. marinum identified changes in genes involved in a variety of cellular processes (Fig 2), although a majority of the most differentially regulated genes were associated with cell wall and cell processes and lipid metabolism. We noted that a substantial number of esx-1-associated genes were down-regulated in the mutant strains during growth in culture medium, including 11 genes that were located within or directly adjacent to the esx-1 gene cluster. Among these down-regulated genes were those coding for known ESX-1 substrates, such as EsxA, EsxB, EspE and EspB. Remarkably, mRNA levels of core components of the ESX-1 secretion system, i.e., those encoding members of the type VII secretion complex, such as EccB 1 , EccD 1 , EccE 1 and MycP 1 , remained unchanged, even though the corresponding genes are interspersed with genes encoding ESX-1 substrates. In contrast to the mRNA levels, we noted a strong increase in the protein levels of EsxA and EsxB, probably reflecting the accumulation of these proteins in the cell due to the secretion defect (Fig 2). Our data also indicate a significant effect of esx-1 disruption on genes associated with lipid metabolism (Fig 2), including genes associated with the synthesis of mycolic acids. Strong down-regulation was observed at the mRNA and protein levels for several polyketide synthases, including genes involved in mycolic acid biosynthesis, such as umaA, mmaA3, accD5, accD6, and pks15/1, which encode components of the lipid biosynthesis pathway (Fig 2  and S1 and S2 Tables). The changes observed in esx-1 and lipid-metabolism-associated genes at the mRNA and protein levels were not unexpected; it has been reported previously that ESX-1-dependent protein secretion and mycolic acid synthesis are critically linked [51]. However, we also noted a surprisingly broad impact of ESX-1 mutation on major biosynthetic pathways, including ribosomal protein synthesis and DNA biosynthesis (S1 and S2 Tables). Downregulation was observed at the mRNA and protein levels for several genes encoding ribosomal proteins and DNA gyrase and a ribonucleotide-diphosphate reductase, which are components of protein and DNA biosynthesis, respectively. We also identified changes at both the mRNA and protein levels in genes involved in general stress response (grpE, dnaK, groES, groEL1), genes involved in stress response regulation (sigA, sigB, devS), members of the WhiB family (whiB2, whiB4, whiB6) and several PE_PGRS genes (Fig 2). The M. tuberculosis esx-1 mutation did not seem to have a significant effect on the expression of genes involved in lipid metabolism compared to the effect seen in M. marinum (Fig 2 and S1 and S3 Tables). Finally, a significant number of genes that are associated with information pathways, including genes encoding ribosomal proteins, were up-regulated at the mRNA level in the eccCb mutant ( Fig  2). Taken together, the observed changes in the transcriptome and proteome of mutants defective in ESX-1 secretion reflect the role this cluster employs for major biochemical pathways in M. marinum and M. tuberculosis.

Global transcriptional profiling of intraphagosomal M. marinum and the eccCb1 mutant
We next determined the effect of ESX-1 abrogation in M. marinum on gene transcription during infection of primary macrophages. Using a PMA-differentiated THP-1 cell line as a model of primary macrophages, we analysed the global gene expression of the wild-type and eccCb1 mutant strains of M. marinum after 6 hours of infection. Wild-type mycobacteria can escape the phagosome within two hours after infection [52], whereas ESX-1 secretion mutants of both M. marinum and M. tuberculosis are known to be limited to the phagosomal compartment [53]. The intraphagosomal transcriptome of the eccCb1 mutant was compared with the intracellular transcriptome of wild-type M. marinum. Furthermore, these intracellular the expression value of a gene. Red dots indicate statistical significance (q < 0.05), and black dots indicate a lack of statistical significance. Selected genes that are most down-or up-regulated in the esx-1 mutant strains are highlighted. (C) Venn diagram of the number of differentially expressed transcripts and proteins quantified of M. marinum eccCb 1 mutant using RNA-seq and quantitative proteomics, respectively. Scatterplots of the relationship between differentially expressed genes of M. marinum eccCb 1 transposon mutant and those of the isogenic wild-type strain E11, quantified in both data sets and classified into the following categories: (D) lipid metabolism, (E) regulatory proteins and (F) cell wall and cell process. Scatterplots and bar chart show the rectilinear equation and the Pearson correlation coefficient (R 2 ). transcriptomes were also compared with the transcriptome of wild-type M. marinum grown in standard broth culture. We identified 720 (p<0.05) genes in the eccCb1 mutant that exhibited significant changes in expression after THP-1 infection compared to the expression levels in the wild-type strain. Of these genes, 465 were down-regulated and 255 were up-regulated (S4 Table and S5 Fig). Remarkably, none of the genes within the esx-1 region were significantly differentially expressed in the esx-1-mutant compared to the wild-type strain. However, we found a specific and pronounced increase in the transcript levels of the espA operon in the WhiB6 is required for the secretion-dependent regulation of ESX-1 substrates in pathogenic mycobacteria intraphagosomal transcriptome of the eccCb1 mutant as compared with the levels in the in vitro transcriptomes (Fig 3A). During growth in culture medium, the mRNA levels of espA did not differ between the wild-type and esx-1-deficient M. marinum strains, which was confirmed by quantitative RT-PCR (qRT-PCR) (Fig 3B). Therefore, these data suggest that proteins encoded by the espA operon, i.e., EspA, EspC and EspD, may play an important role in ESX-1-specific processes during the first stages of macrophage infection. The espA operon was also induced in the wild-type bacteria inside macrophages, albeit at a lower level. Perhaps this difference exists because the wild-type bacteria are able to escape from the phagosome, whereas the eccCb1 mutants are not.
Further analysis showed that a significant number of genes that code for proteins involved in cell wall and cell processes were differentially regulated by intracellular wild-type M. marinum and the ESX-1-deficient strain in comparison with their counterparts grown in culture medium (S5 and S6 Tables). M. marinum genes involved in mycolic acid synthesis, phthiocerol dimycocerosate (PDIM) synthesis and transport to the cell surface, such as fabG1, accDs, ppsC, ppsD, pks11_1, pks13, as well as genes coding for polyketide synthases and the mycolic acid methyltransferase umaA were differentially expressed during infection of THP-1 cells (Fig 3C  and 3D). Furthermore, cpsY, a gene that encodes UDP-glucose 4-epimerases and is essential for linking peptidoglycans and mycolic acid [54], exhibited a pronounced increase in mRNA level in the intracellular eccCb1 mutant (S4-S7 Tables). We also found that many genes associated with cell division and peptidoglycan assembly, such as ftsE, ftsH, ftsW, murC, and murG [55,56], were down-regulated by intracellular bacteria (S4-S7 Tables).
A significant number of genes that code for proteins associated with lipid metabolism and metabolic adaptation were differentially regulated in macrophages (S6A Fig). This subset includes genes involved in fatty acid metabolism such as isocitrate lyase (icl), an enzyme necessary for the glyoxylate cycle and required for intracellular survival [57,58], and pckA, which encodes the phosphoenolpyruvate carboxykinase and is essential for mycobacterial survival in both macrophages and mice [59,60] and is involved in energy metabolism (S6B Fig), and the KstR-dependent cholesterol regulon (S6C Fig), which is involved in lipid degradation and carbon metabolism [61]. We also observed effects of a number of genes involved in general stress response (groES, groEL1, hsp, ahp, dnaK), genes involved in stress response regulation (sigB, devR, devS, hspR, kstR), members of the WhiB family (whiB2, whiB3, whiB4, whiB, whiB6, whiB7) and alternative sigma factors (sigE, sigL, sigM) in the eccCb1 mutant during infection to macrophages. This pattern is illustrated in S6D Fig and is probably associated with stressful intraphagosomal conditions.

Different M. marinum esx-1 transposon mutants have similar gene transcription profiles
The ESX-1-deficient strain of M. marinum used for RNA sequencing contains a transposon in the eccCb 1 gene. To confirm that the observed gene transcription effects were due to a defective ESX-1 system and not due to a side effect of this particular mutation, we analysed several mutants containing transposon insertions in different genes from the esx-1 gene cluster and compared the mRNA levels of the selected genes by qRT-PCR. Our results showed decreased transcript levels of the known ESX-1 substrate esxA and other esx-1 secretion-associated (esp) genes, namely, espL, espK and espJ, for all tested eccCb1 mutants, whereas the transcript levels of eccD 1 , which encodes a structural component of the ESX-1 system [14], did not differ from the transcript levels in wild-type M. marinum (Fig 4). These gene expression patterns in the eccB 1 , eccCa 1 , eccD 1 and eccE 1 transposon mutants were similar to the RNA sequencing results obtained for the eccCb 1 mutant. The only exception was that for the mutant containing a transposon insertion in eccD 1 , we observed an increase of eccD 1 transcription itself and, to a lesser extent, an increase of the adjacent gene espJ (Fig 4). However, this increase was most likely due to the presence of a strong promoter on the transposon, driving the transcription of the kanamycin resistance cassette, as the measured mRNA is transcribed from sequences directly downstream of this promoter. Altogether, our results demonstrate that inactivation of the ESX-1 secretion system leads to down-regulation of the transcription of ESX-1 substrates and associated proteins.

ESX-1 substrate gene transcription is reduced by a regulatory mechanism
We next sought to determine the molecular mechanism underlying the down-regulation of specific transcripts in eccCb1 mutant strains of M. marinum. It is possible that the decrease in mRNA levels is due to a regulatory effect at the transcriptional level. Alternatively, the downregulated mRNA may be degraded via a post-transcriptional mechanism. To investigate these possibilities, we expressed an extra copy of the espL gene under the control of a constitutively active promoter in the wild-type and eccCb 1 mutant strains of M. marinum and determined the espL gene transcript levels. We found a similar increase in espL transcripts in both the wild-type and eccCb 1 mutant strains, indicating that degradation of specific mRNA is probably not the cause of the decreased mRNA levels in the mutant strain ( Fig 5A). Expression levels of the downstream gene espK were not affected by the introduction of espL. These results indicate that there is a regulatory mechanism that prevents the transcription of genes encoding ESX-1 substrates and associated proteins in the absence of a functionally active ESX-1.

PE35 and PPE68 play an important role in ESX-1 secretion but not in gene regulation
Previously, PE35, which is located within the esx-1 gene cluster, has been implicated to be essential for EsxA and EsxB secretion in M. tuberculosis [62]. In contrast to this proposed function, the PE35/PPE68_1 protein pair in M. marinum is secreted via ESX-1 [63,64]. To determine whether PE35 plays a role in the regulation of ESX-1 substrates, we overexpressed the pe35/ppe68_1 operon in M. marinum. Interestingly, although there was no effect on gene transcription (Fig 5C), we noticed a substantial increase in EsxA secretion in the wild-type strain (Fig 5B). This increased EsxA secretion does not seem to represent a general increase in ESX-1 secretion, as protein levels of the cell-surface-localized EspE remained similar (Fig 5B). To study this effect in more detail, we introduced PE35 with a truncated version of PPE68_1 that contained only the PPE domain and was devoid of the C-terminal portion. Although the introduced PE35 protein was expressed and secreted efficiently by ESX-1 (Fig 5B), the levels of secreted EsxA were not increased, indicating that the C-terminal portion of PPE68_1 plays a role in EsxA secretion. To determine whether secretion of the PE35/PPE68_1 protein pair itself was important for this process, we also determined the effect of removal of the last 15 amino acids of the PE protein, which contained the general secretion signal. This small deletion not only abolished the secretion of the introduced PE35 protein but also abolished EsxA secretion completely, despite the presence of an intact chromosomal copy of the pe35/ppe68_1 operon (Fig 5B). This result suggests that the truncated form of PE35 somehow interferes with EsxA secretion. Together these data show that, although PE35 and PPE68_1 do not seem to regulate the transcription of genes encoding ESX-1 substrates, these proteins have a strong effect on EsxA, as previously observed [62]. WhiB6 is required for the secretion-dependent regulation of ESX-1 substrates in pathogenic mycobacteria

Increasing EspI and EspG 1 levels does not affect esx-1 gene expression
A second candidate protein that might regulate gene expression levels of ESX-1 substrates is EspI. The gene encoding this esx-1-secretion-associated protein of unknown function is located within the esx-1 region and is down-regulated in esx-1 mutants of both M. marinum and M. tuberculosis (Fig 2). In contrast to the other Esp proteins, EspI contains a putative nucleotide-binding domain. Previous study has shown that in M. tuberculosis EspI is involved in reduction of ESX-1 secretion in response to low cellular ATP levels [65]. However, when we overexpressed this protein, we did not observe a change in the down-regulation of esx-1-associated gene transcription in the M. marinum eccCb 1 transposon mutant, suggesting that EspI does not regulate this process in our strain (Fig 5C). We next focused on EspG 1 as a candidate esx-1 gene regulator. EspG 1 , interacts specifically with PE35/PPE68_1 in M. marinum [63] and might function as a sensor that measures protein levels of intracellular ESX-1 substrates. To investigate the effect of EspG 1 on esx-1-associated gene expression and protein levels, we increased EspG 1 levels by overexpressing the espG 1 gene in wild-type and ESX-1-deficient M. marinum. This overexpression did not result in altered gene transcription (Fig 5C) or ESX-1 protein secretion. Together, our data show that EspI and EspG 1 do not appear to play key roles in esx-1-associated gene regulation.

WhiB6 plays a role in the transcription of ESX-1 substrates
In addition to espI, another gene encoding a putative regulatory protein was down-regulated in esx-1 mutant strains of both M. marinum and M. tuberculosis, namely, whiB6 (Fig 2). WhiB proteins are actinobacteria-specific regulators that contain iron-sulfur clusters and are thought to act as redox-sensing transcription factors that can cause both gene activation and repression [66]. WhiB6 was suggested to be involved in the regulation of EsxA secretion [67], and subsequent studies have confirmed this suggestion [68][69][70]. To determine whether Whib6 has an effect on the expression levels of esx-1 associated genes, we overexpressed this protein in the ESX-1-deficient M. marinum eccCb 1 transposon mutant strain. We found that particularly those genes that were down-regulated in the mutant strain, such as esxA and espK, showed an increased secretion when whib6 levels were increased (Fig 5D). Furthermore, expression of eccD1 was unaltered by whib6 overexpression, indicating that whib6 is involved in the transcription of ESX-1 substrates and associated genes but not of the components of the ESX-1 system. Surprisingly, whiB6 itself is also one of the genes that is down-regulated upon abrogation of ESX-1-mediated protein secretion. Our results indicate that the presence of WhiB6 in nonsecreting strains has a positive effect on transcription of genes on coding ESX-1 substrates.

WhiB6 is required for the regulation of the ESX-1 system
To further study the involvement of WhiB6 in ESX-1 regulation, we constructed a deletion mutant of whiB6, both in M. marinum WT and in the eccCb1 mutant background (M. marinum M USA -⊿whiB6 and M. marinum M VU -⊿whiB6). First, we analyzed the effect of this mutation on all genes, except for the genes of the esx-1 locus. This analysis identified 34 genes (p<0.05) that were downregulated in the eccCb1 mutant strains (Fig 6A) and therefore putatively influenced by WhiB6. Complementation of both mutants with the whiB6 gene on a mycobacterial shuttle plasmid reversed the upregulation of these genes and generally resulted in decreased expression levels (Fig 6A and 6B). As expected, several genes that are associated with oxidative stress (ahpC, ahpD, rebU) were found in the differently expressed gene pool. Also, the enrichment analysis of the associated Gene Ontology terms for the differently expressed genes (dnaB, dinP) reveal that WhiB6 may also regulate DNA replication or repair through regulating DNA-directed DNA polymerase and DNA helicase (S7 Fig). Another noteworthy gene affected by whiB6 deletion is iniA, which is associated with cell wall stress induced by specific antibiotics. Interestingly, whiB7 is within the whiB6-active gene set, which implies that WhiB7 is active by, or works with WhiB6. Other than the whiB6-active gene set, 13 genes, which are involved in iron-sulfur cluster binding, cellular lipid metabolic processes are downregulated.
Remarkably, these genes are almost exclusively downregulated in the whiB6 esx-1 double mutant, reinforcing a functional link between WhiB6 and the ESX-1 system. However, many of the downregulated genes are encoding hypothetical proteins and hence needed to be further characterized.
Separately, we analyzed the effect of the whiB6 deletion on all esx-1 genes. In line with our previous results, overexpression of whiB6 in the eccCb1 mutant resulted in downregulation of many esx-1 genes (Fig 6C), whereas deletion of whiB6 did not have a strong affect in the eccCb1 mutant background. The effect was the opposite for the mutant with a functional ESX-1 system, there the whiB6 deletion had a strong positive effect on transcription of esx-1 genes. (Fig 6C) and also this effect could be complemented. Only the structural eccE1 and mycP1 genes behaved differently. There are six genes that show the same pattern of up-or down-regulation as most of the esx-1 genes in the two different whiB6 mutants and the complemented strains. Of these six genes, 4 encode proteins that are homologous to secreted substrates and therefore are putative ESX-1 substrates or ESX-1 chaperones, i.e. MMAR_2894 (PE34-like protein), MMAR0299 (PE_PGRS1), MMAR5414 ((EspA-like) and MMAR5432 (EspD-like). Recently, Phan et al. (15) showed using proteomics that MMAR_2894 is highly reduced in the WhiB6 is required for the secretion-dependent regulation of ESX-1 substrates in pathogenic mycobacteria cell surface fractions of M. marinum esx-1 mutants. Together, these experiments show a strong linkage between WhiB6 and the regulation of different esx-1 genes in response to secretion activity.

Discussion
In this study, we determined the transcriptomes of the M. marinum E11 wild-type and the double-auxotrophic M. tuberculosis mc 2 6020 mutant strains and compared these transcriptomes with those of the corresponding isogenic esx-1 mutants. We found that during the growth of M. marinum in 7H9 culture medium, genes encoding ESX-1 substrates, such as EsxA and other ESX-1-associated proteins, were down-regulated in the mutant strains, whereas the transcription of genes encoding several structural components of the ESX-1 system remained unaffected. This specific decrease in transcription might function as a mechanism to avoid toxic accumulation of ESX-1 substrates. Interestingly, a similar decrease in substrate production has been shown for the ESX-5 secretion system, where the PE_PGRS substrates do not accumulate intracellularly when secretion is blocked [71,72]. However, for these PE_PGRS substrates, regulation was shown to occur post-transcriptionally [72], implying that a different mechanism is involved.
The most prominent change in gene expression that was observed upon host cell infection by the M. marinum eccCb1 mutant strain was the increase in transcription of the espA operon. The specific and pronounced transcriptional increase in the expression of this operon, and not of any other esx-1 associated gene, indicates that transcription of the espA operon is regulated independently of the other substrates during infection. Previously, it has been shown that the espA operon is regulated by different transcription and regulation factors, including EspR, MprAB and PhoPR [28, 73,74]. Our new finding also suggests that EspA, EspC and EspD are vital for the bacteria during the early phase of infection. Since ESX-1 has been shown to be responsible for mycobacterial escape from the phagosome, which occurs within the first few hours of infection with M. marinum [53], the proteins produced by the espA operon may play an important role in this process. Consequently, the avirulent phenotype of ESX-1-deficient mycobacteria might be partly attributable to the inability to secrete EspA and/or EspC early in infection.
To determine the mechanism via which ESX-1 substrate regulation is mediated, we overexpressed proteins that may have a regulatory function. Overexpression of the esx-1-encoded EspI and EspG 1 proteins did not have an effect on the reduced transcription of ESX-1 substrates in ESX-1-deficient M. marinum. The putative regulatory protein WhiB6, however, did affect the transcription of these genes. While the transcript levels of whib6 itself were decreased in esx-1 mutants of M. marinum and M. tuberculosis, increasing WhiB6 levels by overexpression resulted in a further decrease in transcription of the ESX-1 substrate in ESX-1-deficient M. marinum. This result clearly indicates that WhiB6 is involved in ESX-1-associated gene regulation, as previously suggested [70]. WhiB6 response to the block in ESX-1 function to repress the genes encode substrates while whiB6 is downregulated in esx-1 mutant which may agrees with Rachel E. Bosserman's results that whiB6 regulation of ESX-1 gene expression is controlled by a negative feedback loop. Indeed, there is accumulating evidence that WhiB proteins function as transcription factors that may play a role in survival within the host (reviewed in [75]). Recently, other groups have also presented evidence supporting a role for WhiB6 in the regulation of the transcription of esx-1 genes [68][69][70].
A remarkable finding in this study was that overproduction of PE35/PPE68_1 resulted in a large increase in EsxA secretion. Previously, deletion of M. tuberculosis PE35 had already been shown to abolish esxA transcription and secretion of the corresponding gene product [62].
Here, we found that EsxA and PE35 secretion are linked, as an increase in PE35 secretion resulted in a concomitant increase in EsxA secretion. The C terminus of PPE68_1 is required for this effect, indicating that this is a specific process, which is consistent with the fact that cell-surface localization of another ESX-1 substrate, namely, EspE, is unaffected by overproduction of PE35/PPE68_1. It is possible that the PPE68 proteins serve as chaperones to escort EsxA outside the bacterium, or these proteins, may be part of the secretion apparatus, making the secretion of specific substrates highly efficient.
During M. marinum infection of human macrophages, we found that transcription of many pe_pgrs and ppe family genes was strongly down-regulated when ESX-1 function was abrogated. As many as 50% of all genes with decreased transcript levels in the eccCb1 mutant strain belongs to one of these gene families (S4 Table). Notably, in the wild-type strain, transcription of the pe_pgrs and ppe genes was decreased during infection in comparison to the levels observed during growth in 7H9 medium (S5 Table). As part of an adaptive response to the macrophage environment, expression of these cell-wall-localized proteins may be down-regulated in order to evade immune recognition or to reduce cell permeability [76]. The fact that in the absence of a functional ESX-1 secretion system these genes are even further down-regulated suggests that there are functional links or shared transcriptional pathways between ESX-1 and (some of the) PE_PGRS and PPE proteins, which are generally ESX-5 substrates [34].
Taken together, our results show that transcription of the espA locus plays an important role in ESX-1 mediated processes during the first hours of infection. Furthermore, we established a functional link between PE35 and EsxA secretion and provided evidence of a regulatory role of WhiB6 in the transcription of ESX-1 substrates and associated genes.