Escherichia coli DH5α (Hanahan, 1985) was cultivated aerobically in 500-ml flasks containing 70–100 ml of 2xYT medium (Green and Sambrook, 2012) at 37°C. Wild-type (WT) C. glutamicum ATCC 13032, its deletion derivative C. glutamicum ΔsigD (Taniguchi et al., 2017) and C. glutamicum sigD-overexpressing strain (Taniguchi et al., 2017) were used for DNA and RNA isolations, and as hosts for testing the activities of promoters cloned in the promoter-test vector pEPR1. C. glutamicum was cultivated aerobically in 500-ml flasks with 70–100 ml of complete 2xYT medium or in minimal CGXII medium (Keilhauer et al., 1993) at 30°C. Media were supplemented with antibiotics, when necessary: kanamycin (Km; 30 μg/ml), tetracycline (Tc; 10 μg/ml) or ampicillin (Ap; 100 μg/ml). The plasmid vectors used are listed in Table 1. The oligonucleotides used are listed in Supplementary Table S1.

DNA isolation, PCR, cutting by restriction enzymes, ligation and transformation of E. coli were done using the standard techniques (Green and Sambrook, 2012). C. glutamicum cells were transformed by electroporation (van der Rest et al., 1999). DNA fragments for cloning promoters in pRLG770 and pEPR1 were assembled from the synthetized complementary oligonucleotides of around 75 nt in length, with overhangs ready for ligation. Their sequences are shown in Supplementary Table S1. Mutations in sigH were constructed using a Q5^® Site-Directed Mutagenesis Kit (New England BioLabs^® Inc.) according to the recommendations of the manufacturer. The specific oligonucleotide primers for mutagenesis were designed by NEBaseChanger^TM v1.2.8 (New England BioLabs^® Inc.).

Corynebacterium glutamicum ATCC 13032 was cultivated in minimal medium CGXII with glucose (2%) or phenol (3.4 mM). The sigD-overexpressing C. glutamicum strain was used as described previously (Taniguchi et al., 2017). Total RNA was isolated from 3 biological replicates of C. glutamicum cells grown to the exponential phase using a Quick-RNA Miniprep Plus kit according to the manufacturer’s instructions (Zymo Research). After additional DNase treatment, RNA samples were purified with an RNA Clean&Concentrator-5 kit (Zymo Research) and quantified with a DropSense 16 (Trinean). The quality of total RNA was controlled with an RNA 6000 Nano kit in an Agilent 2100 Bioanalyzer (Agilent Technologies). To construct the whole transcriptome cDNA library, 2.5 μg total RNA (RIN > 9) was used for the depletion of rRNA with a Ribo-Zero rRNA Removal Kit (Bacteria) according to manufacturer’s instructions (Illumina). The rRNA removal was checked with an Agilent RNA Pico 6000 kit and the Agilent 2100 Bioanalyzer (Agilent Technologies). The mRNA obtained was converted to a cDNA library according to the TruSeq Stranded mRNA Sample Preparation guide (Illumina). The quality and quantity of the cDNA library was checked with an Agilent High Sensitivity DNA kit and the Agilent 2100 Bioanalyzer (Agilent Technologies). Sequencing was performed on an Illumina HiSeq 1500 using 70 bases read length (Illumina).

A primary 5′-end specific cDNA library was synthesized using 2 × 2.5 μg total RNA (RIN > 9) as described previously (Kranz et al., 2018). Briefly, after rRNA depletion and quality control, as described for the whole transcriptome cDNA libraries, a terminator 5′-phosphate-dependent exonuclease treatment (TEN, Illumina) was carried out to digest non-primary transcripts. The remaining non-digested non-primary transcripts were tagged by the ligation of the RNA 5′-index adapter 5′-CCCUACACGACGCUCUUCCGAUCGAG-UACCCUAG (index in bold) to the 5′-monophosphorylated ends. The primary 5′-triphosphate ends were converted to 5′-mono-phosphate by RNA 5′-polyphosphatase (RPP) treatment (Epicenter) to ligate the 5′-adapter to the 5′-ends of primary transcripts. Finally, reverse transcription with a stem-loop DNA adapter and library amplification was performed. The quality and quantity of the cDNA library was checked in the same way as for the whole transcriptome cDNA library. Prior to sequencing, primary 5′-end cDNA libraries were purified and size-selected for fragments approximately 100–1000 bases long via gel electrophoresis, and quantified again. Paired end sequencing was performed in an Illumina MiSeq using MiSeq Reagent Kits v3 with a read length of 2 × 75 bases (Illumina).

Paired-end reads were mapped to the C. glutamicum ATCC 13032 reference genome sequence accession number BX927147 (Kalinowski et al., 2003) with bowtie2 v2.2.7 (Langmead and Salzberg, 2012) using the default settings for paired-end read mapping.

False-positive primary transcriptome cDNA reads containing the barcode sequence TACCCTAG at their 5′-ends were discarded. The remaining R1 cDNA reads were mapped to the C. glutamicum ATCC 13032 reference genome sequence accession number BX927147 (Kalinowski et al., 2003) with bowtie2 v2.2.7 (Langmead and Salzberg, 2012) using the default settings for single-end read mapping. Read Explorer v.2.2 (Hilker et al., 2014) was used for the visualization of short read alignments, transcription start site (TSS) detection and differential gene expression analysis.

Transcription start sites (TSS) were detected essentially as described by Wittchen et al. (2018). Primary 5′-end cDNA library data were analyzed with the software ReadXplorer v2.2 (Hilker et al., 2016) using the Transcription analysis function. The parameters for TSS detection were chosen by ReadXplorer using its automatic parameter estimation. TSS were detected with at least 10 read starts and a minimal coverage increase of 100%. The resulting list of predicted TSS was manually checked for false-positives.

Differential gene expression analysis of C. glutamicum grown with and without phenol (3.4 mM), including normalization, was performed using the whole transcriptome data and Bioconductor package DESeq2 (Love et al., 2014) included in the software ReadXplorer v2.2 (Hilker et al., 2016). The signal intensity value (a-value) was calculated by the log2 mean of normalized read counts, and the signal intensity ratio (m-value) by log2 fold-change. The evaluation of the differential RNA-seq data was performed using an adjusted p-value cut-off of P ≤ 0.01 and a signal intensity ratio (m-value) cut-off of ≥1 or ≤-1.

The in vitro transcription assay was carried out essentially as described previously (Holátko et al., 2012). The holo-RNAP was reconstituted from the RNAP core enzyme isolated from C. glutamicum and individual C. glutamicum sigma factors isolated as His-tagged recombinant proteins from E. coli. The RNAP core (100 nM) was mixed with the respective σ factor (σ^C, σ^D, σ^E, σ^H, or σ^M) in a molar ratio of 1:30. The holo-RNAP was assembled for 10 min at 37°C. The transcription mixture was incubated for 15 min at 37°C. The transcripts labeled with [α-³²P]UTP were separated in 5.5% polyacrylamide gel. In vitro transcription assays were done 2 or 3 times for each promoter, with essentially the same results.

Sigma factors were assigned to individual promoters in vivo using the two-plasmid system for C. glutamicum described recently (Dostálová et al., 2017) and similar that developed for the identification of σ^E-dependent promoters in E. coli (Rezuchova and Kormanec, 2001). In principle, the sigma factors overproduced using the expression vector pEC-XT99A drove transcription from the individual promoters transcriptionally fused to the gfpuv reporter gene in the other plasmid present in the cell, the promoter-test vector pEPR1. Promoters were cloned to pEPR1 as approximately 75-nt fragments using the restriction sites PstI and BamHI. The same pEC-XT99A constructs as described previously (Dostálová et al., 2017) were used to overexpress the sig genes after the addition of isopropyl-β-thiogalactopyranoside (IPTG). Two-plasmid strains were cultivated in 2xYT medium (containing Km and Tc) for 24 h and the cell samples were collected. The cells were disrupted with a FastPrep homogenizer (MP Biomedicals). The fluorescence of the cell-free extracts was determined with a Saphire2 microplate spectrophotometer (Tecan; excitation wavelength 397 nm; emission wavelength 509 nm). Cells harboring the pEPR1 construct and the expression vector without a sig gene were used as a control. To determine the background fluorescence of C. glutamicum cells, the strain only carrying the promoter-less vector pEPR1 was used. Protein concentration in the extract was determined by Bradford assay, and promoter activity was expressed in arbitrary units/mg protein.

The homology models of the σ^D and σ^H domains which recognize the -10 and -35 sequences of the respective promoters were produced by using the Swiss-Model server (Waterhouse et al., 2018). The crystal structures of E. coli σ^E, PDBid: 4LUP (for -10 element GTC) (Campagne et al., 2014) and PDBid: 2H27 (for -35 element GGAAC) (Lane and Darst, 2006) were used as templates. The nucleotides within the E. coli σ^E consensus were replaced to match the consensus for C. glutamicum σ^H or σ^D, where necessary. Molecular dynamics simulations were done using the software package AMBER (Salomon-Ferrer et al., 2013) and Linux computer nodes with powerful NVIDIA GPUs that enable the accumulation of 50-ns MD trajectories at 280 K.

RNA-seq with a C. glutamicum WT strain, a strain overexpressing sigD and a sigD deletion (ΔsigD) strain was performed previously (Taniguchi et al., 2017). Using this approach, the 29 genes were identified, which showed the increased transcription in C. glutamicum overexpressing sigD and decreased transcription in the ΔsigD strain. The genes selected may be those which are directly under the control of σ^D, or the genes whose transcription is influenced by an indirect effect of sigD overexpression or deletion.

To further investigate these σ^D-dependent genes and to map their transcriptional start sites (TSS), RNA-seq of 5′-enriched primary transcripts of C. glutamicum overexpressing sigD was used. The sequencing of primary 5′-end specific cDNA library allows the exact mapping of TSS. To detect these TSS, the data obtained from the primary 5′-end cDNA sequencing were mapped to C. glutamicum ATCC 13032 reference genome sequence, accession number BX927147 (Kalinowski et al., 2003) and analyzed via ReadXplorer (Hilker et al., 2016). To distinguish between the background and actual TSS signal, we considered a position to be +1 (TSS) if the number of read starts here was 10 times higher than at position -1. In this way the 5′- ends of the transcripts of the above mentioned 29 putative σ^D-dependent genes were detected. Since several genes were cotranscribed in operons, the total number of TSS was 23 (Supplementary Table S2). The 50-nt sequences (-49 to +1) were aligned at the TSS, and the consensus motifs within the region of the supposed promoters were searched using the software Improbizer (Ao et al., 2004) as described previously (Albersmeier et al., 2017). The consensus sequence or closely similar motifs within the -35 and -10 regions were found in 11 sequences (Table 2). The products of the respective genes are involved in the synthesis of corynomycolic acid (fadD2, cmt1, cmt2, cmt3), peptidoglycan formation (lppS), inhibition of σ^D activity (rsdA) or have hypothetical functions (Table 3). The σ^D is thus involved in the control of cell wall integrity and cell envelope stress response, as described previously in the strains C. glutamicum ATCC 13032 (Taniguchi et al., 2017) and C. glutamicum R (Toyoda and Inui, 2018).

The sequences were used to generate a sequence logo with Weblogo3 (Figure 1). The consensus sequences GTAAC^A/_G as the -35 motif and GAT as the -10 motif are apparent.

The activity of the 11 proposed σ^D-dependent promoters, which were deduced from the TSS determination by RNA-seq, was tested using the in vitro transcription system (Holátko et al., 2012). The promoter sequence elements of the proposed σ^D-dependent promoters were apparently different from the consensus sequence of σ^A/σ^B-driven promoters (Pátek and Nešvera, 2011; Albersmeier et al., 2017) and therefore ECF σ factors (σ^C, σ^D, σ^E, σ^H, and σ^M) were only used. This assay may directly prove whether a promoter is specifically recognized by RNAP associated with the chosen σ factor, and thus distinguish between direct and indirect control by the tested σ. As shown in Figure 2, all but one promoter were found to be exclusively σ^D-specific. The only exception was the promoter of cg0607 (Pcg0607), which was active with both σ^D and σ^H. The in vitro transcript level with σ^H was even higher than with σ^D (Figure 2). It is worth mentioning that Pcg0607 is the only one out of the 11 analyzed promoters that has the -35 sequence element GGAAC which is a consensus for σ^H-dependent promoters (Busche et al., 2012). The ECF σ factors σ^C, σ^E, and σ^M did not provide any signal in the assays (data not shown). The σ^H-specific clpP1 promoter (Busche et al., 2012) was used as a control.

To confirm the results of in vitro transcription, the activity of σ^D-dependent promoters was determined in vivo using the two-plasmid C. glutamicum cells carrying a sig gene in the expression vector pEC-XT99A and the analyzed promoter in the promoter-test vector pEPR1 (Dostálová et al., 2017). The promoter activity was measured as the fluorescence intensity of the Gfp reporter protein. The results of the assays with 7 most felicitous example promoters which confirmed the activity of σ^H in transcription are presented in Figure 3. As shown in the left part of Figure 3, all promoters exhibited high activity with the overexpressed sigD. Results with Pcg1056 were included to document that not all σ^D-controlled promoters are also recognized by σ^H, although there is no apparent reason found by inspection of the promoter sequence only. Activity of Pcg2047 with σ^H was approximately twice higher after induction at time point 3 (T3) and time point 6 (T6) although it was at the same level of the control. It is an unusual feature of this two-plasmid system in some cases, which we described in our previous paper (Dostálová et al., 2017). We concluded, therefore, that there was an activity of Pcg2047 initiated by σ^H.

The level of Gfp reporter fluorescence was lowest 3 h (T3) after the addition of the inducer (IPTG) in all cases and highest after 24 h (T24) in 10 of the 11 tested promoters. The promoters Pcmt2, PfadD2, PrsdA, Pcg2047, PlppS, Plpd, and Pcg0607 were also found to be active with σ^H, as shown in the right part of Figure 3. Although the maximal σ^H-induced activities were 1.5- to 64-fold lower than those with σ^D under the conditions used, the effect was apparent (Figure 3, right part). The promoter Pcg1056 is an example of a promoter which did not demonstrate any increase in activity with sigH overexpression. Neither of the other overproduced σ factors supported a promoter activity increase with any of the promoters tested (data not shown). Interestingly, the trend of changes in the σ^H-directed promoter activity was mostly the opposite of that with σ^D: the activity with σ^H was highest 3 h after the addition of IPTG, then mostly decreased at T6, and the difference in Gfp level between the cells with IPTG-induced and non-induced σ^H-overproduction nearly vanished at T24 in most cases (Figure 3, right part). Due to this trend, σ^D- and σ^H-incited promoter activities were nearly comparable at T3. This effect was most apparent in Pcg0607, which was also found to be active with σ^H by in vitro assay (Figure 2): this ratio between promoter activity with σ^D/σ^H was 1.5 at T3 but 20.5 at T24 (Figure 3). The overall conclusion is that most of the σ^D-controlled promoters are also to a lesser extent σ^H-activated.

To prove that σ^H is directly responsible for the minor activity of σ^D-controlled promoters in vivo, we tried to overexpress sigH in the ΔsigD strain also harboring pEPR1 with the target promoters. The ΔsigD strain grew slightly slower than the wild type strain (Taniguchi et al., 2017). However, the ΔsigD clones carrying two vector constructs grew extremely poorly. Therefore, we only succeeded in testing the clones with Pcmt2, PlppS, and PrsdA (cloned in pEPR1), moreover, only on a solid agar medium. The increase in promoter activity, (induction ratio of WT and ΔsigD strains with sigH overexpressed), was 2.1 and 1.8, respectively, for Pcmt2, 2.2 and 3.5, respectively, for PlppS and 1.1 and 6.1, respectively, for PrsdA (Figure 4). Thus, RNAP+ σ^H was apparently directly active in transcription from all three tested σ^D-dependent promoters.

To explain the observed function of σ^H in transcription from some σ^D-dependent promoters, the similarity of the amino acid sequences of C. glutamicum σ^D and σ^H proteins was analyzed. The level of overall similarity of the σ^D and σ^H sequences is very low (<25% identical AA), in contrast to C. glutamicum ECF sigma factors σ^H, σ^E, and σ^M, which exhibit a 25–38% mutual identity of AAs. However, when the regions which are thought to interact with the -35 and -10 promoter sequences (Lane and Darst, 2006; Campagne et al., 2014) were compared, some AAs were found to be identical or similar in σ^D and σ^H (shown in green and yellow, respectively, in Figure 5). As shown in Figure 5, majority of these AAs were also found to be conserved in the model ECF sigma factor, σ^E from E. coli, whose crystal structure was previously used for studies on sigma-promoter DNA interactions (Lane and Darst, 2006; Campagne et al., 2014).

We have shown that the consensus sequences of the C. glutamicum -10 and -35 elements recognized by σ^D (this study) and σ^H (Busche et al., 2012) differ (i.e., GAT vs. GTT at -10 and GTAAC^A/_G vs. GGAA^T/_C at -35). To identify the key AAs responsible for recognizing different nucleotides at the second position of these consensus sequences and to understand why σ^H is, to a certain extent, able to initiate transcription from the σ^D-dependent promoters, we used computer modeling tools. First, we created homology models of those σ^D and σ^H domains, which recognize the -10 and -35 sequences of the respective promoters. The structure of E. coli σ^E based on the crystal protein in complex with promoter DNA was used as a template (Lane and Darst, 2006; Campagne et al., 2014). It should be noted that the DNA promoter sequences recognized by E. coli σ^E are closer to the C. glutamicum consensus of the σ^H-specific -10 promoter element (GTC in E. coli σ^E, i.e., T at the second position in both cases) as well as the -35 element (GGAAC in E. coli σ^E, i.e., G in the second position in both cases). This also corresponds to the similarity of the AA sequences of E. coli σ^E and C. glutamicum σ^H which are supposed to interact with these nucleotides (Lane and Darst, 2006; Campagne et al., 2014) (shown in bold in Figure 5).

As for the -10 element, there is no X-ray or cryo-electron microscopy structure of the RNAP complex with any of the group IV (ECF) sigma factors described, and a model of the complete DNA transcription bubble is not yet available. The key nucleotide at the second position of the C. glutamicum -10 elements (i.e., A in σ^D and T in σ^H) is in close contact with the surface of the σ subunit in our homology models based on the E. coli σ^E crystal structure 4LUP (Figure 6). According to these models, the presence of Ala 60 in σ^D and Lys 53 in σ^H seems to be crucial for the recognition of the nucleotide at the second position of the -10 element. Ala 60 with the minimal side chain creates space for a larger adenine at the second position in the -10 element at σ^D (Figure 6A). In contrast, Lys 53 with a larger side chain fills the free space made by a smaller base, thymine, at the second position of the -10 consensus element at σ^H (Figure 6B).

As for the -35 element, template and non-template DNA strands clearly form a duplex. There is no potential conformational polymorphism that could appear with the -10 element, where the transcription bubble opens. Therefore, our homology models of σ^D and σ^H in complex with the respective -35 elements can be considered highly plausible. To verify the stability and reliability of these homology models, we introduced a water envelope into the models and carried out 50-ns MD simulations. It was found, that the hydrophobic interactions of the methyl group of T at the second position of the σ^D-consensus of the -35 element with the side chains of Val 169 and Ala 166 in σ^D are crucial for its recognition (Figure 6D). Conversely, a hydrogen bond with the side chain of Ser 171 is most important for the recognition of G at the second position of the σ^H-consensus of the -35 element (Figure 6E). The hydrogen bond between the hydroxyl group of Thr 168 and the phosphate group of the sugar-phosphate DNA backbone provides another stabilizing interaction.

We used the established models to investigate why σ^H is, to some degree, capable of recognizing σ^D-controlled promoters.

Most of the σ^D-controlled promoters were to a lesser extent also activated by σ^H, although the highly conserved sequences of the σ^H-specific promoters (-35 GGAA^T/_C, -10 GTT) differ from those of the identified σ^D-specific promoters (-35 GTAAC, -10 GAT). Computer modeling showed that σ^H can bind to the key regions of promoter DNA in a similar manner as σ^D to the analogous elements. Based on the results of the sequence similarity analysis and computer modeling, we designed mutations to make σ^H more σ^D-like and possibly more efficient at transcription from σ^D-dependent promoters. We introduced the mutations into the sigH gene cloned in the expression vector pEC-XT99A, which resulted in the production of the mutant proteins σ^HLys-53-Ala or σ^H168-AlaValArgValAla-172. These mutant genes were used in the two-plasmid assays with the promoters which exhibited the highest activity with σ^H: Pcg0607, Plpps, Pcmt2, and PrsdA. We assumed that the overproduced σ^HLys-53-Ala and σ^H168-AlaValArgValAla-172 would trigger higher promoter activity in two-plasmid assays with the chosen σ^D-dependent promoters than σ^H. This assumption was only confirmed with σ^HLys-53-Ala and Pcg0607 (Figure 7). This effect was particularly apparent at T24, when promoter activity with σ^HLys-53-Ala was 1.7-fold higher than that with σ^H. This higher activity compared to other promoters could be explained by the unparalleled presence of the GGAA sequence in the -35 region of Pcg0607, which is a consensus sequence of -35 elements in σ^H-dependent promoters. This result indicated that Lys 53 in σ^H and Ala 53 in σ^HLys-53-Ala (and Ala 60 in σ^D) most likely interact with the -10 promoter elements.

Contrary to the expectation, σ^H168-AlaValArgValAla-172 recognized Pcg0607 less efficiently. To assess how the mutant σ^H168-AlaValArgValAla-172 could interact with the σ^D-controlled promoters, we analyzed the homology model with this mutant σ factor in more detail. Using a 50-ns MD simulation, we found that the complex of σ^H168-AlaValArgValAla-172 with the -35 element of the σ^D-controlled promoter is probably not stable. This was something of a surprise given that the native σ^H was able to efficiently bind σ^D-controlled promoter DNA in previous MD simulations. A detailed analysis of the MD trajectories showed that this instability appeared to be due to the absence of Arg at the final position of the AA sequence AlaValArgValAla (ThrValMetSerArg in σ^H). The positively charged Arg 172 side chain of σ^H was found to create a significant stabilizing interaction (a salt bridge) with the negatively charged backbone of promoter DNA. This hypothesis, which might explain the low efficiency of the mutant σ^H168-AlaValArgValAla-172, may be tested in the future by creating additional mutations suggested by the in silico analysis.

In addition to the TSS ascribed to the σ^D-controlled promoters (Table 2), start sites of transcription driven from different overlapping or nearby promoters were detected by RNA-seq for 8 of the 11 analyzed genes. We suggest that most of these promoters are activated by σ^A and/or σ^B, and two are activated by σ^H, according to the upstream sequences (Figure 8A, in green and red, respectively). The putative housekeeping promoters are always located upstream of the σ^D-dependent promoters if there is a single additional σ^A- and/or σ^B-dependent promoter (upstream of the cg0420, cg0607, cg2047, and cmt2 genes). Promoters which we consider to be σ^H-dependent according to the sequences of the putative -35 (GGAA) and -10 (GTT) elements were found upstream of lppS and fadD2 genes. The detection of all TSS in the upstream regions of the σ^D-dependent genes was important to avoid effects of transcription started from the additional promoters. The position of the cloned promoter fragments for in vivo and in vitro assays could thus be chosen in such a way, that only the activity of the supposed σ^D-controlled promoters was determined. The organization of the multiple promoters within the upstream regions of the 8 genes is shown in Figure 8B. In the case of fadD2, the studied σ^D-dependent P1fadD2 and the additional σ^H-dependent P2fadD2 partially overlap and the activity of the additional σ^H-specific promoter might thus contribute to the overall σ^H-dependent activity of the fragment. However, such activity was not detected by in vitro transcription. In vivo, the σ^H-dependent activity was similarly low as in other cases. The σ^A- and/or σ^B-dependent promoters P3cmt1, P2cg0420, P2cg0607, P2cg2047, P3lppS, and P2cmt2 (Figures 8A,B) were found to partially (mostly by the -10 region) reside on the 70-nt fragments with the analyzed σ^D-dependent promoters. However, their possible activity most probably did not interfere with the in vivo measurements of the effects of σ^D and σ^H overexpression.

We have tested a number of stress conditions (heat shock, SDS, penicillin G, glycine, mitomycin C, phenol and limitation by glucose) to detect a slower growth of the C. glutamicum ΔsigD strain or to induce transcription from PsigD or from the promoters found to be σ^D-dependent (data not shown). Out of the conditions tested, only phenol exhibited a stronger inhibitory effect on the growth of the ΔsigD strain carrying the vector pEC-XT99A than on the growth of the analogous WT strain. We found that growth rate of C. glutamicum WT culture in 2xYT medium after the addition of 7.5 mM phenol decreased from 0.88 to 0.46 h^-1 (1.9-fold), whereas the growth of the C. glutamicum ΔsigD decreased from 0.68 h^-1 to 0.22 (3.1-fold). A difference in the effect of phenol was not significant in plasmidless strains.

To analyze the transcriptional response of C. glutamicum to phenol stress, we performed RNA-seq using the RNA isolated from the C. glutamicum WT cells grown in minimal medium with glucose (2%) or phenol (3.4 mM). Differential gene expression analysis including normalization, was performed using the whole transcriptome data and Bioconductor package DESeq2 (Love et al., 2014) included in the software ReadXplorer v2.2 (Hilker et al., 2014). The evaluation of the differential RNA-seq data was performed using an adjusted p-value cut-off of P ≤ 0.01 and a signal intensity ratio (m-value) cut-off of ≥1 or ≤-1. According to the results (Supplementary Tables S3, S4), five genes previously found to be σ^D-dependent were more highly expressed on phenol than on glucose: cg0607, lppS, rsdA, lpd, and cg2047 (Supplementary Table S4). The respective TSS identified by RNA-seq of 5′-enriched primary transcripts using the cultures grown on phenol were identical with those found using the sigD overexpressing strain.

The effect of phenol on the activity of the promoters Pcg0607, PlppS, PrsdA, Plpd, and Pcg2047 was tested in vivo with the two-plasmid system. The strains carrying sigD or sigH inserted in pEC-XT99A and the promoters in pEPR1 were cultivated in minimal medium with glucose (2%) or phenol (3.4 mM). The promoter activity was measured in the extract of the cells before the addition of IPTG and at the end of exponential growth phase.

The activity of four promoters (Pcg0607, PlppS, PrsdA, and Plpd) in the strains with overexpressed sigD was clearly higher after cultivation with phenol (Figure 9). The activity of the σ^C-dependent Pcg2556 used as a control was much lower in the medium with phenol under sigC overexpression conditions. The overexpression of sigH in the presence of phenol only resulted in a small increase in the Plpd activity. These results confirmed the data from RNA-seq (with the exception of Pcg2047 which was induced on phenol according to RNA-seq, but not in the two-plasmid assay). This is one of a few examples of the stress response to the action of phenolic compounds regulated by the ECF sigma factors in C. glutamicum (Chen et al., 2016, 2017). The phenol stress response mediated by ECF sigma factors will be further studied in C. glutamicum on the basis of the large amount of data produced by RNA-seq.