All of the B. subtilis strains used in this study are listed in Supplementary Table S1. One-step competence medium (Kunst et al., 1994) [MC], Schaeffer’s sporulation medium (Schaeffer et al., 1965), and Luria-Bertani (LB) medium (Difco, Lennox) were used. Antibiotic concentrations were described previously (Ogura and Tanaka, 1996; Ogura et al., 1997). Synthetic oligonucleotides were commercially prepared by Tsukuba Oligo Service (Ibaraki, Japan) and are listed in Supplementary Table S2.

Strains were grown on a LB agar plate (1.5%) containing appropriate antibiotics at 37°C overnight. The cells were scraped from the plate and suspended in the sporulation medium. The suspension was inoculated into 50 ml sporulation medium (with or without glucose) without antibiotics in a 200-ml flask. Klett value was adjusted around 10 units. The flask was gently shaken (110 reciprocation/min) at 37°C. Cell growth was monitored using Klett colorimeter (Klett Mfg., Co., Inc., New York, NY, USA).

The plasmids used in this study are listed in Supplementary Table S1. pIS284-sigM-del1 and pDG1663-sigX-del2 were constructed by cloning of the double-stranded oligonucleotides PsigM-F/PsigM-R and PsigX-F/PsigX-R into pIS284 and pDG1663 which were treated with EcoRI/BamHI, respectively (Guérout-Fleury et al., 1996; Tsukahara and Ogura, 2008). To construct pDG1663-sigX-Wt and pDL2-sigX-del1, PCR products amplified by using the oligonucleotide pair SigX-F/SigX-R and SigX-F/SigX-R2, respectively, were digested with EcoRI/BamHI and cloned into pDG1663 and pDL2 treated with the same enzymes (Yuan and Wong, 1995). To construct pDG1729-PcshA, PCR products amplified by using the oligonucleotide pair, pDG1729-PcshA-E/pDG1729-PcshA-H, were digested with EcoRI/HindIII and cloned into pDG1729 treated with the same enzymes (Guérout-Fleury et al., 1996). To construct pMUTIN3DZ, PCR products amplified by using the oligonucleotide pair, ypuN-F/ypuN-R, were digested with HindIII/BamHI and cloned into pMUTIN3DZ treated with the same enzymes (Yoshimura et al., 2004).

The bkdB::Pxyl-cshA unit was constructed using PCR (Supplementary Figure S1). The bkdB-up linked to Km^r and xylR-Pxyl units were amplified from the total DNA of the strain containing bkdB::sinR and the pX plasmid, respectively (Hori et al., 2002; Ogura, 2016). The other fragments were amplified from the wild type derived total DNA. PrimeSTAR MAX DNA polymerase was used (TaKaRa). Four DNA fragments were assembled by using Gibson Assembly Master Mix (New England Biolabs). The sequences of the cshA ORF in the resultant strains were confirmed by sequencing using the primers pX-cshA-seq-F and pX-cshA-seq-R. The cshA::Tc^r unit was constructed using PCR. Briefly the upstream and downstream regions and Tc^r from pBEST304 (Itaya, 1992) were amplified using the indicated primers (Supplementary Table S2) and then combined by PCR.

Spot-on-lawn assays were performed as previously reported (Luo and Helmann, 2009). To prepare the lawn, overnight culture of the Sublancin-sensitive Y13 cells in liquid LB medium was inoculated into 4 ml of melted 0.7% LB agar (2%), and poured onto the solid 1.5% LB agar plate. Both media were supplemented with or without 1% glucose. Plates were dried for 30 min in an incubator (37°C), and 3 μl of each overnight culture in liquid LB medium was spotted on plates and incubated at 37°C.

The transposon delivery vector pMarA (Le Breton et al., 2006) was introduced into the strain OAM709. The resultant strain was grown in liquid LB medium containing kanamycin at 30°C overnight. The cells were diluted and plated onto sporulation medium with 1.5% agar plates containing 2% glucose, X-gal (100 μg/ml), kanamycin, and erythromycin. The plates were incubated at 42°C and the pale colonies were chosen. Otherwise, the diluted cells were plated onto LB medium with 1.5% agar plates containing kanamycin and erythromycin, and incubated at 42°C. Colonies were then transferred onto sporulation medium with 1.5% agar plates containing 2% glucose, X-gal (100 μg/ml), kanamycin, and erythromycin. The plates were incubated at 37°C to choose pale colonies. The insertion mutations were backcrossed into the parental strain and the resultant strains were used for the Lac-assay. Total DNA was taken from the candidate strain, SauIIIA1-digested, self-ligated, and subjected to inverse PCR using oligonucleotides 695 and 696 as described previously (Chan et al., 2014, Supplementary Table S2). The PCR products were sequenced using the oligonucleotide 696.

CshA-His was purified using a Ni-affinity column from B. subtilis OAM730 (cshA bkdB::Pxyl-cshA-His) cells as described previously (Ogura et al., 2014).

Western blot analysis was performed by a method described previously (Hata et al., 2001). Monoclonal anti-acetylated lysine rabbit antibody was purchased from Cell Signaling Technology (Danvers, MA, USA). This antibody was diluted (1/1000) in 1x TBS with 5% BSA and 0.1% Tween20, and the solution was incubated with a protein-blotted nitrocellulose filter overnight at 4°C. To enhance signal, Can Get Signal solution 2 (ToYoBo) was used for secondary antibody.

We observed GI of the spx gene in sporulation medium (Shiwa et al., 2015). The transcription of the yjbC-spx operon is complex. Five promoters and five σ factors have been reported to play roles in transcription of the operon (Antelmann et al., 2000; Thackray and Moir, 2003; Cao and Helmann, 2004; Jervis et al., 2007, Figure 1). To clarify which σ factors other than σ^A are responsible for GI, we introduced several mutations in the genes encoding σ factors and examined expression of the promoter for yjbC fused to lacZ. The typical cell growth in the sporulation medium with or without 2% glucose is shown (Supplementary Figure S2A). In the sigX and sigM mutants GI was reduced (Figure 1). In a double mutant of sigX and sigM, GI was completely lost, while in the sigB and sigW mutants, GI was still observed (Figure 1). Similar results were obtained using an spx-lacZ fusion (Supplementary Figure S3). This raised the possibility that sigX and sigM themselves might be induced by glucose.

To examine this possibility, we tested the expression of two genes in glucose-containing medium. As expected, sigX and sigM were induced by glucose in sporulation medium (Figure 2). This induction was very sensitive to low concentrations of glucose, 0.05% for sigX and 0.1% for sigM. In B. subtilis the ccpA gene has been known to play a major role in glucose-dependent gene induction and repression (Fujita, 2009). Thus, to test whether CcpA is involved in GI of sig genes, we examined the expression of sigX and sigM in a ccpA disruptant in the glucose-containing medium and observed normal GI (middle panels in Figure 2). This demonstrated that CcpA is not involved in GI.

The expression of sigX and sigM has been shown to be downregulated by their respective anti-σ factors (Brutsche and Braun, 1997; Horsburgh and Moir, 1999). To investigate the involvement of the anti-σ factors in the GI phenomenon regulating sigX and sigM, mutations in anti-σ factor genes were introduced into each strain with the lacZ fusion. It should be noted that a strain with an IPTG-dependent anti-σ^M factor gene was constructed and the expression was tested in medium containing 0.05 mM IPTG, since the anti-σ^M factor gene is essential (Horsburgh and Moir, 1999). We observed enhanced expression of both genes as expected, because the mutations increased the free forms of the σ factors. The increase was 3.5-fold for sigX and threefold for sigM (right panels in Figure 2). As shown in these panels, GI was observed in both mutants, indicating no involvement of anti-σ factors in GI of two σ-regulated genes. Among the seven ECF σ genes, only these two genes were subject to GI (Supplementary Figure S4). It should be noted that GI of sigX was observed in cells grown in LB medium (Supplementary Figure S5).

Bacillus subtilis has been shown to produce Sublancin, an antimicrobial glycopeptide (Luo and Helmann, 2009). The precursor peptide-encoding gene and the self-resistance gene for Sublancin (sunA and sunI, respectively) are located on the SP-β prophage region in the chromosome (Luo and Helmann, 2009). The sunA, sunT (Sublancin transporter) and sunS (Sublancin glycosyltransferase) genes are activated by the transition regulator Abh, whose transcription requires σ^X in addition to σ^M, although σ^M has a minor role (Luo and Helmann, 2009). Thus, GI of sigX might enhance Sublancin production through the σ^X/Abh/SunA pathway. To test this possibility, we spotted Sublancin-producing wild type 168 cells on the lawn of Y13 cells lacking SP-β, which have defects in the production and immunity for Sublancin. On the 1% glucose-containing LB agar media, a larger inhibition zone of Y13 growth was observed as compared to that observed in LB medium (Figure 3). This indicates that glucose addition induces sigX, leading to enhanced transcription of abh and the operon containing sunA. Recently, glucose addition was shown to enhance Sublancin resistance in LB medium (Garcia De Gonzalo et al., 2015). Thus, the resistance levels of the lawn would be higher, and the apparent enhancement of inhibition zone would also be underestimated. As reported previously, we also observed a smaller inhibition zone in the sigX mutant (Luo and Helmann, 2009).

We determined a minimum promoter region responsible for GI in sigX and sigM by testing GI in lacZ fusions carrying various promoter regions. GI is expected to be observed in all constructs because the strains bear endogenous sigX or sigM genes. In both genes, similar levels of GI were observed in all fusions including the core promoter region fusions (-40 to +2 relative to the transcription start site, Figure 4). Moreover, between the core promoter regions of sigX and sigM there is no common nucleotide sequence except for the -35 and -10 elements for σ^X-RNAP and σ^M-RNAP. Thus, the similar levels of GI in all constructs suggested no involvement of some unknown trans-acting factors in GI.

We were interested in which carbon sources could induce sigX and sigM expression. Thus, three carbon sources were added, and we examined the expression levels of both sig genes. Glycerol or pyruvate addition induced gene expression to some extent, while succinate did not (Figure 5A). We note that the effects of pyruvate were modest. Glycerol and pyruvate have been shown to be incorporated into the glycolysis pathway and succinate is not (Wang et al., 2010). Thus, this pattern is reminiscent of acetyl-CoA production. Recently, protein-lysine acetylation has been studied in B. subtilis (Kim et al., 2013; Kosono et al., 2015; Carabetta et al., 2016). This modification may change protein function. In Escherichia coli, glucose addition changed and enhanced global acetylation profiles (Schilling et al., 2015). Thus, we hypothesized that glucose addition might result in a higher cellular pool of acetyl-CoA, leading to enhancement of acetylation of proteins regulating sigX and sigM. To examine the involvement of acetyl-CoA in GI of sigX and sigM, a disruption mutation of pdhC, which encodes the E2 subunit of pyruvate dehydrogenase and whose reaction product is acetyl-CoA (Gao et al., 2002), was introduced into the lacZ fusion strain. The rationale for the experiment is that in the pdhC mutant an acetyl-CoA pool would be reduced, although the cells still have other acetyl-CoA synthesis pathway involving acetyl-CoA synthetase. It should be noted that the gene encoding acetyl-CoA synthetase has been known to be repressed by glucose (Grundy et al., 1994). In the resultant strains, GI of sigX and sigM was completely abolished (Figure 5B). We observed slow growth and poor cell mass of the mutants as previously reported (Gao et al., 2002, Supplementary Figure S2B). These results strongly suggested acetyl-CoA dependent GI of sigX and sigM.

To further gain insights into the mechanism of GI, we performed transposon mutagenesis for GI-deficient mutants by using a transposon delivery vector, pMarA (Le Breton et al., 2006). Mutants with sigX-lacZ showing pale blue on LB or sporulation agar plate containing X-Gal and 2% glucose were chosen among about 12,000 colonies. Several mutants were obtained and one was a disruptant of cshA, which encodes DEAD-box RNA helicase (Lehnik-Habrink et al., 2010). Transposon was inserted into the 342nd codon of the cshA ORF. In this mutant, sigX and sigM expression was not induced by glucose and basal expression was reduced but not abolished (Supplementary Figure S6). The decrease in sigX expression resulted in a decrease in Sublancin production in the cshA mutant (Figure 3). Among many candidates of the Tn mutagenesis, only additional four mutations negatively affected GI of both sigX and sigM genes (data not shown). Two carried a disruption of a gene of which function was not yet identified (ylxR and yqfO) and the rests were ptsH and gcp mutants. Probably in the ptsH mutant efficient glucose transport would be inhibited, leading to abolishment of GI. The detailed analysis of these mutants will be reported elsewhere.

The chromosomal region surrounding cshA is shown in Supplementary Figure S7A, cshA seems to be a monocistronic operon with its own promoter and terminator (Nicolas et al., 2012). Thus, we created a promoter-lacZ fusion of cshA and examined its expression. The expression was observed and found to be unaffected by glucose addition (Supplementary Figure S7B, left). The σ^X/M/V-driven promoter within the ydbO ORF has been reported to co-transcribe ydbO-(ydbP)-ddlA-murF (Eiamphungporn and Helmann, 2008). Between murF and cshA there is no terminator, and thus this promoter may transcribe cshA. In addition, there is a promoter upstream of ddlA (Nicolas et al., 2012). These suggested a read-through mechanism from the upstream promoters. To confirm this, we examined activities of a lacZ-fusion located at the original locus in the chromosome. This lacZ fusion showed higher activity as compared to its own promoter fused to lacZ. Glucose addition also did not affect the activity (Supplementary Figure S7B, right). It should be noted that the strain with this fusion at the original locus carries the cshA disruption. These observations showed that read-through transcription contributed to cshA expression. It was previously reported that the cshA expression is constitutive at the protein level (Lehnik-Habrink et al., 2010).

We created a xylose-inducible cshA cassette at the bkdB locus in the cshA mutant with sigX-lacZ or sigM-lacZ (Supplementary Figure S1A). Xylose addition to the culture of these strains without glucose recovered the basal expression levels of sigX-lacZ and sigM-lacZ (left panels in Figure 6). Xylose addition to the culture with glucose also rescued GI of these strains, demonstrating that the cshA gene is responsible for GI. We noted that xylose addition did not have any effect on GI in the control strain with sigX-lacZ nor sigM-lacZ (data not shown).

We added His-tag to the CshA protein encoded by the cassette at bkdB using PCR (Supplementary Figure S1B). The His-CshA protein was confirmed to be functional by a complementation test (Figure 7A). Xylose addition to the culture with or without glucose caused expression of His-CshA, and the two protein fractions containing His-CshA from the two time points (T0 and T2), were purified by Ni-affinity column (Figure 7B). A global analysis of the acetylome in B. subtilis revealed that CshA was acetylated at K244 and K296 (Kosono et al., 2015). Thus, we examined the acetylation status of purified His-CshA by Western blot analysis using an anti-acetyl lysine antibody. The acetyl lysine signals were enhanced in His-CshA purified from the cells with glucose, indicating that glucose addition stimulates CshA acetylation (Figure 7B). A similar result was reproducibly observed from independent culture (data not shown).

CshA has been known to be involved in RNA degradosome containing endoribonuclease RNase Y (Lehnik-Habrink et al., 2010). RNase Y has global impact on the gene expression through mRNA metabolism (Lehnik-Habrink et al., 2011; Durand et al., 2012; Laalami et al., 2013). Thus, CshA may cause GI of sigX and sigM due to the inhibition of RNase Y activity through acetylated CshA-dependent mechanism, leading to the longer half-life of mRNAs of sigX and sigM. Otherwise RNase Y may regulate sigX and sigM indirectly. Indeed, sigM expression was enhanced in an RNase Y-depleted mutant, rny, according to the three analyses (Lehnik-Habrink et al., 2011; Durand et al., 2012; Laalami et al., 2013). To examine this, we introduced rny depletion into the strains bearing sigX-lacZ or sigM-lacZ and tested whether GI was observed. GI of sigX was clearly observed in the rny mutant (Figure 8, left). With respect to sigM, we observed enhancement of sigM-lacZ expression especially in the log phase as expected (Figure 8, right). Moreover, GI of sigM was also clearly observed. These results demonstrated that CshA would act on expression of sigX and sigM independent of influence to RNase Y due to interaction of CshA with RNA degradosome.

K–R and K–Q mutations have been shown to mimic non-acetylation and acetylation states, respectively (Kamieniarz and Schneider, 2009). Hence, we also introduced these mutations into the xylose-inducible cshA cassette system (Supplementary Figure S1A). Introduction of these mutations resulted in similar effects on the expression of sigX-lacZ and sigM-lacZ (Figure 6). In the strain with CshA bearing two K–R substitutions, basal expression was recovered to that of wild type CshA, whereas GI was completely abolished. In the strain with CshA bearing two K–Q substitutions, basal expression was not recovered, whereas GI was observed. These results strongly suggested that acetylation of two lysine residues in CshA is required for GI of sigX and sigM. Moreover, the results also indicated that basal expression and GI of sigX and sigM were mutually distinguishable in a genetic context. We note that the strains bearing wild type and mutant CshA proteins did not grow in culture with 2% glucose in the presence of xylose, and the extent of the resistance against glucose changed in the strains bearing sigX-lacZ and sigM-lacZ due to unknown reasons (data not shown). Thus, different glucose concentrations were used for the experiments for sigX-lacZ and sigM-lacZ.

Recent analysis using mass spectrometry of B. subtilis RNAP reported that half of the RNAP pool was associated with CshA (Delumeau et al., 2011). Thus, we present a model for the mechanism by which GI of sigX and sigM occurs as shown in Figure 9.