For gene cloning, E. coli NM522 was used. The strain E. coli M15, carrying pREP4-GroESL (Amrein et al., 1995), was used for biosynthesis of tagged proteins. For in vivo mutant construction, B. subtilis BS514 (B. subtilis 168 trp⁺ Pr:: neoR) was used. All mutants used in this work are listed in the Supplementary Table S1. B. subtilis and E. coli strains were grown in Luria-Bertani (LB) medium with shaking, at 37°C. When relevant, ampicillin (100 μg/ml) and kanamycin (25 μg/ml) for E. coli and erythromycin (1 μg/ml), neomycin (5 μg/ml), and phleomycin (2 μg/ml) for B. subtilis were added to the medium.

All PCR primers with restriction enzymes are listed in Supplementary Table S2. The deletion of tkmA gene was performed using the modified mutation delivery method of Fabret et al. (2002). The deletion of tkmA gene was made using the pairs of primers ΔtkmA-ext fwd/ΔtkmA-ext rev were used. To introduce an in-frame deletion between the codons 10 and 240 of the gene tkmA, partially complementary primers ΔtkmA fwd/ΔtkmA rev were used. Gene fatR was amplified by PCR from B. subtilis 168 genomic DNA and inserted into the vector pSG1729 (Lewis and Marston, 1999) between the restriction enzymes site KpnI and XhoI, which results in the replacement of the gfp gene by Strep-fatR. B. subtilis wild type (WT). ΔsalA (Derouiche et al., 2015), ΔptkA (Jers et al., 2010), and ΔtkmA strains were transformed with the pSG1729-fatR construct and selected for erythromycin resistance. All constructs were sequenced to check for absence of unsolicited mutations.

SalA, Ugd, Asd, and FatR proteins were produced in E. coli M15 as 6x His N-terminal fusions. The cultures were grown with shaking (200 rpm) at 37°C until OD₆₀₀ of 0.5. To induce the expression, 1 mM of isopropyl β-D-1 thiogalactopyranoside (IPTG) was added to the medium, and the cultures were kept in the same growth conditions for three more hours. To purify the 6x His-tagged proteins, Ni-NTA columns were used (Qiagen) as described previously (Mijakovic et al., 2003). The Step- tagged FatR protein was purified from B. subtilis BS514 using the Strep Tactin affinity chromatography (Novagen) as described in our previous study (Shi et al., 2014b). The purity of proteins was checked by SDS-PAGE and aliquots of all purified proteins were kept at -80°C in buffer composed of 50 mM Tris-Cl pH 7.5, 100 mM NaCl and 10% glycerol.

All in vitro phosphorylation assays were performed in the presence of 50 μM ATP [γ-³²P], which corresponds to 20 μCi/mmol of labeled ATP. A standard 40 μl reaction mix contained 1 μM BY-kinase PtkA and 1 μM activator proteins (TkmA-NCter or SalA). Concentration of protein substrates was 5 μM, and the reaction medium contained 1 mM MgCl₂ and 100 mM Tris-HCl (pH 7.5). After 1 h of incubation at 37°C, the reactions were stopped by adding SDS-PAGE loading dye and heating for 5 min at 100°C. The proteins were separated by electrophoresis using SDS-PAGE. The gels were washed and then boiled in 0.5 M HCl for 10 min, and transient staining with Coomassie Blue was used to identify protein bands. The gels were dried overnight and the autoradiography signals of phosphorylated proteins were revealed by a PhosphoImager (FUJI). The experiments were performed in triplicates, each time with protein aliquots purified independently. One representative assay is shown for each reaction.

For enzymatic assays, B. subtilis strains were grown in LB with vigorous shaking in flasks at 37°C. Samples were harvested at the mid-exponential phase (OD₆₀₀ = 0.4). Cells were lysed by sonication. The lysis buffers (composition adapted for each assay) contained 1% tyrosine phosphatase inhibitor cocktail (Sigma) and 1 mg/ml lysozyme (Sigma). For the Ugd assay, the buffer contained 50 mm Tris-HCl pH 7.5, 50 mm NaCl and 10% glycerol. For the Asd assay, the buffer contained 50 mM 3-(N-morpholino) propanesulfonic acid (pH 7.0), 200 mM KCl, 0.1 mM EDTA, and 10 mM 2-mercaptoethanol. After centrifugation, the supernatant was desalted on PD-10 columns (GE healthcare), to remove all metabolites and cofactors. Total protein concentration was standardized using the Bradford assay. The Asd assay was performed with 100 μg of total protein, as described by Jers et al. (2010). The only difference was that instead of adding commercially available aspartate kinase, the aspartate kinases present in the crude extract of B. subtilis (Graves and Switzer, 1990; Roten et al., 1991; Kobashi et al., 2001) were used to convert ATP and aspartate to aspartyl phosphate. The Ugd assay was performed with 100 μg of total protein, as described by Pagni et al. (1999).

To examine competition between SalA and TkmA for PtkA binding, purified TkmA was separated on a 12 % SDS-PAGE gel. Protein content of the gel was transferred to a PVDF membrane using a Trans-blot cell from Bio-Rad. The transfer buffer (pH 8.3) was composed of 25 mM Tris, 125 mM Glycine, and 10% ethanol. The membrane was washed three times with 50 ml MQ water and then blocked overnight (at 4°C) with 5% BSA in a buffer containing 25 mM Tris pH 8, 125 mM NaCl, and 1% Tween 20. The assay was carried out with 20 nM Strep-tagged PtkA alone, or mixed with SalA in different ratios (indicated in the figure legend). PtkA (and SalA where indicated) were added to the membrane, and the mix was incubated for 1 h at room temperature. To detect the interaction between PtkA and TkmA, the membrane was incubated with 1:1000 of conjugate Srep-tactin HRP (IBA Biotechnology) in 1% BSA for 1 h in TBS-T buffer as described previously (Derouiche et al., 2015). The chemiluminescence signals were visualized using an AEC chromagen kit (SIGMA). Three replicates were performed with independently purified proteins, and one representative experiment is shown.

Strep-tagged FatR was purified from B. subtilis in different backgrounds: WT, ΔsalA, and ΔtkmA. The DNA binding assay was performed with a 24 bp double-stranded DNA containing the fatR operator sequence, as described by Derouiche et al. (2013). The molar ratio of the DNA probe to strep-tagged FatR proteins is indicated in the figure legend. The reaction mixtures were incubated for 1 h at the room temperature. The migration was performed for 2 h at 2 V/cm in 0.5 Tris-acetate-EDTA using non-denaturing gels (12% polyacrylamide). The experiment was repeated three times with independently purified proteins. One representative experiment is shown.

Recently, we have established that the C-terminus of SalA can interact directly with the BY-kinase PtkA, which results in phosphorylation of the SalA residue tyrosine 327. This phosphorylation enhances its function as a repressor of scoC (Derouiche et al., 2015). Interestingly, the in vitro phosphorylation of SalA took place in absence of the canonical activator TkmA (Derouiche et al., 2015). This finding is difficult to reconcile with the available structural data. The BY-kinases from Firmicutes require the interaction with TkmA-type activators to stabilize the ATP binding pocket in their active site in order to autophosphorylate or phosphorylate substrates (Olivares-Illana et al., 2008). The presence of TkmA is a strict requirement for phosphorylation of all previously characterized PtkA substrates (Mijakovic et al., 2003, 2006; Jers et al., 2010; Derouiche et al., 2013, 2015; Shi et al., 2016). The first clue in explaining this observation comes from the structural homology of SalA with BY-kinases and MinD proteins (Derouiche et al., 2016). A part of the C-terminal region of SalA which interacts with PtkA shows sequence homology (by circular permutation) with the activating fragment of the Staphylococcus aureus BY-kinase activator CapA and the N-terminus of its cognate BY-kinase CapB (Olivares-Illana et al., 2008). This suggested that the C-terminus of SalA could be expected to activate PtkA. To investigate this, we performed a series of PtkA autophosphorylation reactions, keeping the kinase concentration constant and varying the concentration of SalA (Figure 1A, lanes 2–5). The in vitro phosphorylation assay demonstrated that the presence of SalA strongly stimulates autophosphorylation of PtkA in the absence of its canonical activator TkmA.

We recently proposed a model for PtkA-SalA kinase-substrate interaction that may take the form of a hetero-octamer in which some PtkA subunits have been replaced with SalA (Derouiche et al., 2015). In this PtkA-SalA interaction model, both the SalA phosphorylated residue Y327, and the SalA sequence homologous to the CapA activator domain, face the PtkA active site. Therefore, the interaction model is consistent with the notion that SalA may activate PtkA. Examination of all available structures and structural models indicates that the interactions of TkmA-type activators and SalA with the active site of the BY-kinase are similar and thus structurally overlapping. Thus, TkmA and SalA should not be able to interact with PtkA simultaneously. To verify this, we performed a Far Western experiment, in which we detected the interaction of TkmA and PtkA. The concentrations of TkmA and PtkA in the assay were fixed, and we varied the concentration of SalA. A clear inhibition of the TkmA-PtkA interaction was observed with higher concentrations of SalA, indicating that SalA competes with TkmA for PtkA binding (Figure 1B).

The results presented in the previous section establish SalA as a bona fide candidate for an alternative activator of PtkA. If PtkA can interact with either of the two activators in vivo, and the activation effect is similar, what is the reason for the existence of two activators? To address this question, an in vitro phosphorylation assay with purified proteins was performed, where we compared the ability of PtkA to phosphorylate several characterized protein substrates in the respective presence of either TkmA or SalA (Figure 2). The first characterized PtkA substrate, Ugd (Mijakovic et al., 2003; Petranovic et al., 2009), was phosphorylated in the presence of either of the activators. By contrast, SalA promotes the phosphorylation of the PtkA substrate Asd more efficiently (Jers et al., 2010), while TkmA promoted the phosphorylation of the PtkA substrate FatR more efficiently (Derouiche et al., 2013). When SalA served as an activator in the in vitro reaction, for example in the reaction with Asd and Ugd, phosphorylation of SalA was abolished (Figure 2B, lanes “Asd” and “Ugd”). This could mean that the PtkA-SalA interaction exists in two different and mutually exclusive modes: kinase-substrate and kinase-activator. It would be very interesting from the structural and functional perspective to explore this possibility in the future.

The hypothesis that SalA and TkmA may determine the choice of substrate preferentially phosphorylated by PtkA was next assessed in vivo. We used the same three PtkA substrates: Asd (preferentially phosphorylated in the presence of SalA), FatR (preferentially phosphorylated in the presence of TkmA), and Ugd (equally phosphorylated in the presence of either TkmA or SalA). PtkA-dependent phosphorylation is known to lead to a measurable modification of activity of all these substrates: increase of the Asd enzymatic activity (Jers et al., 2010), decrease in the DNA binding affinity of FatR (Derouiche et al., 2013) and increase of the Ugd enzymatic activity (Mijakovic et al., 2003; Petranovic et al., 2009). To test the impact of SalA and TkmA on the phosphorylation of PtkA substrates in vivo, we constructed strains with seamless deletions of salA and tkmA.

The Asd activity was measured directly in desalted crude extracts with phosphatase inhibitors (to preserve the Asd phosphorylation state). Compared to the WT, the activity was significantly reduced in the ΔptkA and ΔsalA strains, but not in ΔtkmA (Figure 3A). This indicated that Asd in vivo activation via phosphorylation depends on PtkA and SalA, but not on TkmA. For assessing FatR activity, we prepared B. subtilis strains expressing Strep-tagged FatR in the same relevant genetic backgrounds: WT, ΔsalA, and ΔtkmA. FatR was purified in the presence of phosphatase inhibitors and assayed for its ability to bind its target DNA (Figure 3B). The binding was not detectable with FatR purified from the WT and ΔsalA strains, but was detectable with the equal amount of FatR purified from the ΔtkmA strain, indicating that the absence of TkmA coincided with less efficient phosphorylation in vivo. Finally, for Ugd we also used the enzyme activity essay with desalted crude extracts (with phosphatase inhibitors) (Figure 3C). In this case the activity was reduced in ΔptkA compared to the WT, but was not significantly affected in either ΔsalA or ΔtkmA, suggesting that in this case the two activators can complement each other’s absence. Thus all the in vivo and in vitro findings were in agreement, suggesting that SalA and TkmA can indeed play the role of alternative PtkA activators, with a potential to preferentially direct the kinase activity to certain substrates. Interestingly, our two-hybrid interaction assay had previously revealed that TkmA interacts with PtkB, a second BY-kinase from B. subtilis (Shi et al., 2014c). It is therefore tempting to speculate that PtkB might also have two alternative activators: TkmA and TkmB.

Our findings indicate that there exists a mutual regulation between B. subtilis PtkA and SalA, which critically depends on the physical interaction between the C-terminal region of SalA and the catalytic domain of PtkA (Figure 4). This interaction has regulatory consequences for both proteins. In contact with PtkA substrates, SalA activates the kinase but does not get phosphorylated itself. In the absence of other substrates, as described previously (Derouiche et al., 2015), SalA gets phosphorylated by PtkA, this promotes its ATP binding, and in turn stimulates binding to its target DNA sequence, upstream of scoC. Accounts of such mutual regulatory loops are not very common in the literature. One example of negative feedback involves the human ubiquitin E3 ligase SIAH2 which is phosphorylated and activated by the serine/threonine kinase DYRK2. In turn, SIAH2 destabilizes and promotes degradation of the kinase (Pérez et al., 2012). Forward-feeding activation loops are exemplified by the murine Src family kinases, which activate the neurotrophin receptor tyrosine kinase TrkB, which in turn activates the kinases (Huang and McNamara, 2010). To the best of our knowledge, this is the first report of a mutual regulation loop based on a direct protein–protein interaction in bacteria. The case is particularly interesting since the interactants share a common evolutionary origin (Derouiche et al., 2016), suggesting that the interaction of SalA and PtkA with mutual regulatory consequences was maintained in the process of divergent evolution. One may expect that such cases of mutual regulation will turn out not to be uncommon in bacteria, since the reports on “moonlighting” proteins which have more than one known function are steadily accumulating (Jeffery, 1999; Wang et al., 2013; Wang and Jeffery, 2016).