Bacterial strains used in this study are listed in Table 1, plasmids are listed in Table 2.

Corynebacterium glutamicum cells were routinely grown in CASO broth at 30°C. The defined CGXII mineral salt medium (Keilhauer et al., 1993) with a glucose concentration of 10 g L⁻¹ was used in addition to the complex CASO medium for growth experiments, which were carried out in a 2 inch deviation Innova 44R shaker (New Brunswick), using 250 mL shaking flasks and a shaking frequency of 300 rpm. All cultures were inoculated from precultures in complex CASO-broth. An initial OD₆₀₀ of 0.2 was used for cultivation in CASO-broth. For growth analysis in CGXII minimal medium, precultures were washed in main culture CGXII medium and then diluted to an initial OD₆₀₀ of 0.5. Unless stated otherwise, E. coli cells were cultured in LB broth at 37°C. Kanamycin (50 μg mL⁻¹) and ampicillin (100 μg mL⁻¹) were used as selection markers.

All primers used in this study were designed with Clone Manager Professional 9 (Scientific & Educational Software) and are listed in Table S1 with their respective sequences and purposes. For the construction of pK18mobsacB-based deletion constructs, approximately 500 bp long regions up- and downstream of the target gene were amplified by PCR and cloned into the vector backbone by Gibson Isothermal Assembly (Gibson et al., 2009). All constructs were introduced into C. glutamicum CR099 cells by means of electroporation. Preparation of electrocompetent C. glutamicum cells and electrotransformation was performed as described by Tauch et al. for Corynebacterium diphtheriae (Tauch et al., 2002), screening for the first and second recombination events were performed as described previously (Schäfer et al., 1994). Both kanamycin-sensitive and sucrose-resistant clones were examined by colony PCR analysis, using primers that bind 250 bases up- and downstream of the flanks used. The construction of C. glutamicum CR099 double and triple deletion mutants was based on the previously generated single or double deletion mutants using the identical procedure.

The construction of E. coli deletion mutants was performed using the Quick & Easy E. coli gene deletion kit (Gene Bridges). The procedure was carried out according to the manufacturer's instructions using the FRT-PGK-gb2-neo-FRT template DNA. By choosing appropriate short homologous sequences, which were attached to the functional cassette by PCR, the genes relA and spoT could be removed consecutively in the form of start-stop deletions.

To generate expression constructs for the different stringent response-associated genes from C. glutamicum CR099 and C. glutamicum AS1.542, the corresponding genes relS_Cg, relP^*_Cg, rel_Cg, and relP_Cg were cloned into the vector pBAD24 by Gibson Isothermal Assembly (Gibson et al., 2009). The resulting plasmids pBAD::relS_Cg, pBAD::relP^*_Cg, pBAD::rel_Cg and pBAD::relP_Cg were subsequently transformed into the E. coli ΔrelAΔspoT strain.

Since the addition of low arabinose concentrations had no effect on the growth of the investigated strains in complex medium, all cultures were spiked with a low arabinose concentration of 0.001% (m/v) during cultivation in LB medium to ensure basal expression of the target genes. All strains were inoculated with an OD₆₀₀ of 0.05 and incubated for 3 h at 37°C. After cultivation, the cells were washed and diluted in PBS buffer (pH 7.4) to identical cell titers. To investigate the growth characteristics of the different strains, 5 μL aliquots of various dilution stages were spotted onto minimal medium plates (Kasai, 2002), containing 4 g L⁻¹ glycerol and different arabinose concentrations. The plates were incubated at 37°C for 48 h.

The IMPACT kit (New England Biolabs) was used for the tag-free purification of the proteins RelS_Cg, RelP^*_Cg, and Rel_Cg. The corresponding genes relS_Cg, relP^*_Cg, and rel_Cg were introduced into the vector pTXB1 by Gibson Isothermal Assembly and the obtained plasmids were transformed into the expression strain E. coli ER2566. The cultivation of the resulting strains was carried out in shaking flasks, using LB broth and an initial OD₆₀₀ of 0.1. After a 2-h cultivation period at 37°C, protein expression was induced by the addition of IPTG (400 μM final concentration). To avoid inclusion bodies, the cultivation temperature was reduced to 19°C 30 min after induction. Following a cultivation period of 16 h, the cells were centrifuged and resuspended in lysis buffer [25 mM Tris-HCl (pH 8.5), 500 mM NaCl, 1 mM EDTA, 0.15% (w/v) Triton X-100, 20 μM PMSF, and 1 mM TCEP]. A French press was used for cell disruption in two runs at about 1,200 psi. The chitin resin based column purification of the heterologously expressed proteins was carried out according to the stated manufacturer data, using a column buffer composed of 25 mM Tris-HCl (pH 8.5), 500 mM NaCl, and 1 mM EDTA. To wash the column, a high salt column wash buffer with a NaCl concentration of 1,500 mM was used. The eluate was concentrated using a 10 kDa cut-off centrifugal filter unit and then dialyzed twice against a glycerin-containing storage buffer [300 mM NaCl, 50 mM Tris-HCl, pH 8.5, 50% (m/v) glycerol] in a 10 kDA cut-off dialysis cassette. Roti-Nanoquant (Roth) was used to determine the protein concentration and storage was carried out at −20°C.

In order to analyze the biochemical properties of the purified proteins, they were incubated with various substrates under variation of relevant parameters. All assay reactions were carried out in a total volume of 50 μl and the enzyme activity was inactivated after the reaction by the addition of 10 μL of glacial acetic acid. Enzyme free controls and nucleotide standards measured for concentration determination were treated analogously.

The reaction buffer for determining the activity kinetics with respect to various substrate concentrations contained 50 mM HEPES-Na (pH 7.0), 200 mM NaCl, and 5 mM MgCl₂. 25 mM ATP and varying guanosine derivate concentrations were applied to determine substrate concentration dependency. For the analysis of the kinetics with respect to the substrate ATP, (p)ppGpp synthetase assays with a GTP concentration of 25 mM and various ATP concentrations were performed. 250 nM enzyme was used and the reaction was carried out at 30°C for 60 min.

The general activity analysis of the three purified enzymes as well as the characterization of RelS_Cg and Rel_Cg with respect to relevant activity parameters was based on pH optimum values of bifunctional RSH enzymes at pH 8 (Sajish et al., 2007), a physiological Mg²⁺ concentration of 5 mM (Nierhaus, 2014) and the standard C. glutamicum growth temperature of 30°C. The reaction mixture contained 50 mM Tris-HCl (pH 8.0), 5 mM MgCl₂, 4 mM ATP, and 4 mM GTP/GDP/GMP as well as 500 ng of the respective enzyme and was incubated for 4 h at 30°C. For the further analysis of RelS_Cg and Rel_Cg the individual influence factors temperature, MgCl₂ concentration and pH were varied using GTP as guanosine substrate. In order to achieve a broad spectrum of different pH values, citrate, phosphate, Tris and ammonia buffer systems with a concentration of 50 mM each were used.

HPLC analysis of all assay reactions was performed using a LaChrome ULTRA system (HITACHI) and a SeQuant ZIC-pHILIC column (Merck Millipore). A MicrOTOFQ (Bruker Daltonics) was used for the mass spectroscopic identification of the different nucleotide species. The individual components were separated by means of an isocratic elution profile with 38% of 10 mM ammonium bicarbonate buffer (pH 9.3) and 62% of acetonitrile over a period of 30 min at a flow rate of 0.2 mL min⁻¹. The injection volume was 1 μL and the detection was carried out at a wavelength of 252 nm. For the determination of pGpp, ppGpp and pppGpp concentrations, the equimolar produced by-product AMP was quantified since more exact and reproducible UV peaks were obtained for this component and no suitable standards were available for the three alarmone species. In order to analyze the activity spectra of the investigated enzymes, the peaks of the UV chromatograms were identified using the MS data. The respective peak areas were first corrected by the results of an enzyme-free negative control and then normalized to the protein amounts and reaction times.

Data analysis was performed using the program OriginPro 2017 (OriginLab). For the determination of the enzyme kinetic parameters v_max and K_m, polynomial fits for the relationships between specific activity and the substrate concentrations were carried out due to product excess inhibition.

In addition to the already known rel_Cg gene (Wehmeier et al., 1998), three further putative stringent response-associated open reading frames were identified in the genome of C. glutamicum by Atkinson et al. in the course of a high-throughput sensitive sequence analysis of the RSH superfamily (Atkinson et al., 2011). Two of the putative ORFs (cg1330 and cg2324) were assigned to different SAS classes and one (cg1485) to the SAHs. In this study, we focused on the investigation of the two putative monofunctional pyrophosphokinases with the aim of identifying novel (p)ppGpp sources in C. glutamicum.

Based on a phylogenetic analysis of the RSH superfamily, carried out by Atkinson et al., the gene cg1330 of C. glutamicum ATCC 13032 was classified as a member of the RelQ subgroup (Atkinson et al., 2011). However, an in-depth investigation of the corresponding protein sequence indicated that Cg1330 has similarities with SAS enzymes from S. mutans and B. subtilis, which are referred to as RelP/SAS2. Since the corresponding proteins were also classified as representatives of the SAS RelQ subfamily by Atkinson et al. (2011), a mix-up of the original nomenclature seems to exist at this point. The same holds true the other way around, proteins originally labeled as RelQ were classified as representatives of the SAS subgroup RelP, so that the names RelP and RelQ were evidently interchanged in the course of the classification of the RSH superfamily by Atkinson et al. (2011). In order to avoid inconsistency with the already performed analyzes of the SAS proteins, the gene product of cg1330 is hereinafter referred to as RelP_Cg. A BLASTP analysis of the corresponding protein sequence illustrated, that this subgroup occurs almost exclusively in Firmicutes and that C. glutamicum is the sole member of the Corynebacteriaceae possessing a RelP variant (Camacho et al., 2009). In analogy to the bifunctional Rel_Cg enzyme of C. glutamicum, RelP_Cg exhibits high similarities to five synthetase sequence motifs, which are strongly conserved for RSH proteins in other Actinobacteria, such as Mycobacteria, in Firmicutes, in Proteobacteria, and in plants (Bag et al., 2014; Steinchen and Bange, 2016; Figure 1A).

In the course of our protein sequence analysis, it was found that the known C. glutamicum genomes show differences with respect to the cg1330 (relP) gene. In comparison to other strains like C. glutamicum R (Yukawa et al., 2007), C. glutamicum ATCC 13032 features a RelP version that is shortened by 41 aa at its C-terminus and has an alternative 11 aa C-terminal end sequence (Figure 1B). An alignment with the ppGpp synthetase catalytic domain YjbM consensus sequence illustrates the special position of the ATCC 13032 RelP variant. The shorter version of RelP_Cg that is present exclusively in ATCC 13032 and its derivatives shows all SAS synthetase protein motifs but lacks some highly conserved amino acids located in its C-terminal region. These highly conserved amino acid residues are present in the genomes of other C. glutamicum isolates (Figure 1B). A multiple alignment of the amino acid sequence of potential homologs of RelP_Cg selected based on amino acid sequence similarity emphasizes this feature since the RelP protein from C. glutamicum ATCC 13032 has the shortest of all found RelP variants (Figure S1). The difference was further investigated by an alignment of the corresponding nucleotide sequences. The analysis identified a 32 bp deletion for the ORF cg1330 from C. glutamicum ATCC 13032 compared to the otherwise largely identical sequence of the strain C. glutamicum R (Figure 1C). This deletion leads to a frameshift, starting from amino acid position 199, and the formation of an alternative stop codon and explains the differences between the protein sequences within the different C. glutamicum strains. Based on this result, the cg1330 gene of the strain C. glutamicum ATCC 13032 is henceforth referred to as relP^*_Cg, while the other C. glutamicum variant is termed relP_Cg.

The second putative SAS gene, cg2324, was assigned to the actRel subgroup of the small SASs (Atkinson et al., 2011). Since there is no appropriate gene name for members of this group, we here propose the term relS, with the C. glutamicum actRel protein being encoded by relS_Cg. The amino acid sequence of this protein also shows a high similarity to the ppGpp synthetase catalytic domain YjbM (COG2357) over a range of 231 aa, and contains all conserved synthetase motifs (Steinchen and Bange, 2016; Figure 1A). Interestingly, both RelS_Cg, and Rel_Cg show slight deviations from the highly conserved Syn4 amino acid sequence motive ExQIRT (Bag et al., 2014; Steinchen and Bange, 2016). To date, only one representative of the actRel subgroup has been studied in more detail, the enzyme MS_RHII-RSD from Mycobacterium smegmatis (Murdeshwar and Chatterji, 2012). However, due to an additional RNase HII domain, this 64.5 kDa enzyme is significantly larger than RelS_Cg, which has only a molecular weight of 38.9 kDa. The additional RNase domain was also found in further representatives of the subgroup (Atkinson et al., 2011). To our knowledge, RelS_Cg therefore is the first analyzed representative of the actRel subgroup without an RNase domain.

In order to analyze the phylogenetic affiliation of the protein RelS_Cg more precisely, a phylogenetic tree of RelS homologs was created (Figure S2). In addition to proteins with sequence similarity to RelS_Cg, proteins allocated to the actRel subgroup by Atkinson et al. were taken into account during the analysis (Atkinson et al., 2011). In general it was found that a RelS homolog is only present in species belonging to the phylum Actinobacteria. In contrast to RelP, the actRel subgroup is abundant in the family Corynebacteriaceae and the corresponding proteins can be distinguished phylogenetically from further homologous representatives. As already demonstrated by Murdeshwar and Chatterji, some species of the closely related genus Mycobacterium and the more remote genus Marmoricola possess RelS variants for which a second domain, namely the RNase H-associated protein domain COG4328, was identified. It is interesting to note that all mycobacteria which contain this fusion protein (N-terminal RNase H domain and a C-terminal (p)ppGpp synthetase domain) represent exclusively environmental species while it appears to be absent in the representatives of the human pathogenic Mycobacterium tuberculosis (Mtb)-complex. In addition to the two large groups already described, many other actinobacterial genomes also contain a relS homolog. However, the performed phylogenetic analysis does probably not represent the complete picture as all candidates were identified only by similarity to RelS_Cg. This is particularly clear from the fact that some representatives of the actRel subgroup according to Atkinson et al. are located far outside the phylogenetic RelS tree. For that reason, this approach should be extended and ideally tested experimentally.

Since the deletion of the rel gene (cg1861) was known to have a negative effect on the growth of C. glutamicum in a chemically defined minimal medium (Tauch et al., 2001), deletion mutants missing the putative monofunctional (p)ppGpp synthetases RelP^*_Cg or RelS_Cg were analyzed for growth defects. The corresponding genes as well as rel_Cg were removed individually from the genome of the strain C. glutamicum CR099 (Unthan et al., 2015) by homologous recombination, using the vector pK18mobsacB (Schäfer et al., 1994). The strain CR099 is a direct descendant of ATCC 13032 but devoid of three prophages (ΔCGP1/2/3) and all members of two IS-element families (ISCg1 and ISCg2). In order to prevent enrichment of suppressor mutations in precultures of the growth experiment, as observed for E. coli ppGpp⁰ strains if cultivated in minimal medium (Murphy and Cashel, 2003), all precultures were grown in complex CASO medium. To analyze the growth characteristics in minimal medium, the precultures were then washed in the main culture CGXII medium and adjusted to the desired inoculum density.

No significant growth differences were observed for the three analyzed single deletion mutants in complex CASO medium (Figure 2A). Although the strain CR099Δrel showed slightly delayed growth, the growth rate during exponential growth phase was equivalent to that of the parent strain CR099 and to the deletion mutants CR099 ΔrelP^*_Cg and CR099 ΔrelS_Cg.

Since the deletions of relP^*_Cg and relS_Cg also did not lead to a change in growth behavior during the cultivation in minimal medium (Figure 2B), both putative monofunctional SAS enzymes seem to be dispensable individually under standard growth conditions.

While in a previous study (Tauch et al., 2001), a partial rel-deletion mutant based on the strain C. glutamicum RES167 showed a strong growth defect and requirements for L-histidine and L-serine, the complete deletion of rel in C. glutamicum CR099 caused a different growth phenotype in CGXII minimal medium (Figure 2B). In the first phase of the cultivation the strain showed almost no growth difference to the parental strain. In the course of the growth experiment, however, a continuous reduction of the growth rate occurred in contrast to the parental strain. After a cultivation period of about 12 h, strain CR099 Δrel almost reached a stationary state with an OD₆₀₀ of 4. Although the OD₆₀₀ increased only slightly to a value of 5 after a cultivation period of 26 h, the strain reached the same end OD₆₀₀ as the parental strain after 48 h.

In order to analyze the in vivo role of RelS_Cg and RelP^*_Cg or possible interrelations between Rel_Cg and the putative SAS enzymes, the growth experiment in minimal medium was extended by the strains CR099 ΔrelΔrelP^* and CR099ΔrelΔrelS as well as the triple deletion mutant CR099 ΔrelΔrelP^*ΔrelS (Figure 2B). For multiple deletions involving the gene relP^*_Cg, no change in the growth behavior was observed. The strain CR099 ΔrelΔrelP^* behaved identically to the strain CR099 Δrel and the triple deletion mutant grew with the same characteristic as the strain CR099 ΔrelΔrelS. Thus, a deletion of gene relP^*_Cg does not seem to have any physiological effects under the analyzed growth conditions.

In contrast to this, the simultaneous absence of rel_Cg and relS_Cg in the strains CR099 ΔrelΔrelS and CR099 ΔrelΔrelP^*ΔrelS caused a clear growth effect (Figure 2B). In the first hours of cultivation, both strains showed a short lag phase in contrast to all other strains examined and started growing after 1.5 h with a growth rate corresponding to that of the wild type. In the further course, both mutants showed uniform growth and did not show the growth rate reduction as observed for strain CR099 Δrel. The comparison of the corresponding growth characteristics clearly shows an in vivo role of RelS_Cg.

Since the growth rate reduction of the strain CR099 Δrel can be compensated by an additional deletion of relS_Cg, the expression of RelS_Cg appears to be a problem for the cells in the strain without the bifunctional enzyme Rel_Cg.

In order to analyze the in vivo functionality of the identified putative SAS enzymes RelP^*_Cg and RelS_Cg, a complementation experiment of an E. coli ΔrelAΔspoT strain derived from the K12 wild-type MG1655 strain was attempted. The relA spoT double deletion mutant is a (p)ppGpp⁰ strain, which shows multiple amino acid auxotrophies and therefore can no longer grow in minimal medium (Xiao et al., 1991). In order to realize different expression levels of the proteins of interest in this strain, the corresponding genes were cloned into the arabinose-inducible vector pBAD24. This allows a precise control of the strength of expression by the variation of inducer concentration as well as the used carbon source (Guzman et al., 1995). In addition to the two genes relS_Cg and relP^*_Cg, the gene rel_Cg, coding for the C. glutamicum ATCC 13032 RSH enzyme and the non-truncated relP_Cg variant from the strain C. glutamicum AS1.542 were cloned into the vector pBAD24.

As expected, the E. coli ΔrelAΔspoT strain containing the empty plasmid control pBAD showed only single colonies in the lowest dilution stage of 10⁻², whereas the parental E. coli strain MG1655 with the same plasmid formed numerous colonies even in the highest dilution stage of 10⁻⁵ (Figure 3). The non-auxotrophic colonies of the double deletion strain are all likely to be (p)ppGpp⁰ suppressor phenotypes, also called E. coli M+ mutants, which are pseudo-revertants generated by spontaneous mutations (Murphy and Cashel, 2003). A corresponding picture of individual revertants was also shown for both RelP variants RelP^*_Cg and RelP_Cg. Therefore, no complementation effect could be found for these proteins. In contrast, the expression of the gene relS_Cg resulted in a complementation of the auxotrophic test strain even at very low arabinose concentrations of 0.00001%. Since the auxotrophy of the complemented strain is due to the absence of a basal (p)ppGpp level, RelS_Cg thus most probably provides a (p)ppGpp synthetase activity. While significant growth was still evident at this low inducer concentration up to the dilution factor 10⁻⁴, an increased inducer concentration of 0.1% arabinose led to a clearly weaker complementation with significantly smaller colonies up to dilution stage 10⁻³. The expression of the RSH protein-encoding gene rel_Cg, on the other hand, led to a complete and inducer concentration-independent complementation of the amino acid auxotrophy, since the corresponding strain showed no difference to the non-auxotrophic parental strain E. coli MG1655 for all tested inducer concentrations.

In order to investigate the catalytic activities of the putative monofunctional small SASs, RelS_Cg, and RelP^*_Cg, both enzymes as well as the RSH-protein Rel_Cg were analyzed biochemically by means of in vitro synthetase assays. The NEB IMPACT™ system was used for the tag-free preparation of all enzymes, which were successfully checked for purity and identity by SDS-PAGE and MALDI-TOF analysis (data not shown). For a precise evaluation of (p)ppGpp synthetase assays, which were based on established reaction conditions, a LC-MS approach was developed. Compared with the routinely used TLC-based analysis of radioactively labeled nucleotides, this method allows a clear identification and more accurate quantification of the reaction products.

In the reaction mixture containing RelP^*_Cg as well as ATP and GTP as substrates, significant amounts of ADP and GDP were detected (Figure 4A). This result implies a phosphohydrolase activity of RelP^*_Cg. Furthermore, RelP^*_Cg possesses a kinase activity, since ADP and GTP were produced in the assay including ATP and GDP (Figure 4B). In addition, very small amounts of AMP were detected in all assay reactions containing RelP^*_Cg (Figure 4). Despite the very sensitive MS signal, however, no evidence for the production of an alarmone species was found. For this reason, the reaction product AMP apparently is not due to a pyrophosphokinase reaction but results from enzymatic degradation of ADP. To exclude inappropriate reaction conditions as reason for the missing pyrophosphokinase activity, different buffer systems and salt concentration were tested in compliance with positive characterizations of RelP homologs and further SAS enzymes from other organisms (data not shown). However, no alarmones were synthesized in any of the different conditions, indicating that RelP^*_Cg lacks the suggested activity. The result of the in vitro analysis thus complies with the previously performed complementation experiments in which no in vivo (p)ppGpp synthetase activity could be detected for the corresponding enzyme RelP^*_Cg, as well as the non-shortened version RelP_Cg.

For the second putative SAS RelS_Cg, the proposed pyrophosphokinase activity was clearly identified. In the presence of this enzyme, a high amount of pppGpp was detected in the reaction mixture containing ATP and GTP as substrates (Figure 4A). In addition to AMP, which is also produced during the pyrophosphokinase reaction, ADP and GDP were detected. This suggests that RelS_Cg has a phosphohydrolase activity similar to RelP^*_Cg. Using ATP and GDP as substrates, ppGpp was synthesized in the reaction mixture containing RelS_Cg(Figure 4B). Interestingly, for the synthetase assay reaction with ATP and GMP as substrates, a mass-to-charge ratio of 521.99 was found as a product which exactly complies with the corresponding value of GTP. However, since the retention time slightly differed from the one found for GTP and the reaction mechanism of the enzyme seems to be selective for the pyrophosphorylation in the 3′-position, the detected product is thought to be GMP 3′-diphosphate (pGpp) (Figure 4C).

According to expectations, the pyrophosphorylation of GDP and GTP to ppGpp and pppGpp was detected for the bifunctional RSH enzyme Rel_Cg (Figures 4A,B). Interestingly, very high ADP and GTP concentrations were observed using ATP and GDP as substrates (Figure 4B). The enzyme Rel_Cg thus apparently has a kinase activity since this product spectrum can be explained by the transfer of the terminal phosphate from ATP to GDP. Furthermore, this assay reaction contained a mixture of ppGpp and pppGpp. With the help of the sensitive MS signal and an exact evaluation of the retention time, pGpp could be detected for the substrate combination ATP/GMP (Figure 4C). However, due to the very low amount of pGpp, which was below the UV detection limit, the corresponding enzyme activity of Rel_Cg is very low.

Since a (p)ppGpp synthetase activity could not be detected for RelP variants from C. glutamicum, neither in the in vivo complementation experiments nor in the in vitro assays, we focused on the more detailed analysis of the enzyme RelS_Cg.

In order to characterize RelS_Cg in more detail, a comparison of the enzyme activity for various substrates was done. It was recently shown that the reaction products pGpp, ppGpp and pppGpp exhibit slightly different biological functionalities (Mechold et al., 2013). Different enzyme activities with respect to the possible substrates could therefore have a great influence on the biological significance of the enzyme in the context of the stringent response.

To investigate the enzyme kinetics, the specific pyrophosphokinase activity of RelS_Cg was determined for different substrate concentrations. Physiological reaction conditions with a pH of 7.0 as well as a Mg²⁺ concentration of 5 mM, which was found for E. coli (Nierhaus, 2014), were applied. To be able to compare the generated data with known kinetic parameters of other SAS enzymes, further assay components as well as the buffer system used were adapted to the corresponding studies (Murdeshwar and Chatterji, 2012; Steinchen et al., 2015). Using standard substrate concentrations of up to 5 mM, no substrate saturation of the specific pyrophosphokinase activity was achieved for all substrates. Nevertheless, in order to determine parameters of the RelS kinetics, higher GMP/GDP/GTP concentrations were used and the ATP concentration was adjusted to 25 mM to exclude a limitation of the second substrate. To investigate the influence of ATP, kinetics for the conversion of ATP and GTP to pppGpp were recorded with different ATP concentrations. Since the half-maximal specific activity has already been reached for approximately 0.8 mM ATP, this substrate seems not to be rate-determining in comparison with guanosine substrates (Figure S3).

Similar kinetics with nearly identical maximum specific pyrophosphokinase activities (v_max) of approximately 2.8 μmol min⁻¹ mg⁻¹ were determined for either GTP or GDP (Figure 5). However, using GDP v_max is reached at significantly higher concentrations of 20 mM, compared to GTP for which a value of 14 mM was determined. Furthermore, RelS achieves its half-maximal pyrophosphokinase activity for the conversion of GTP to pppGpp at a substrate concentration of 4.1 mM, whereas a higher K_m value of 6.1 mM was determined for the substrate GDP. At very high substrate concentrations, a reduced pyrophosphokinase activity with respect to GDP and GTP was observed.

RelS shows a very untypical kinetics for GMP. At low substrate concentrations of up to about 7 mM, a nearly linear increase in activity takes place, whereby the v_max values are significantly lower than for the other substrates. At higher substrate concentrations of 10–25 mM, a saturation effect occurs analogous to the kinetics for GDP and GTP. In the case of even higher substrate concentrations, however, the specific pyrophosphokinase activity increases significantly again to values around 2.8 μmol min⁻¹ mg⁻¹. Using a GMP concentration of 50 mM the addition of acetic acid to stop the reaction resulted in an increased viscosity, so that the sample could no longer be properly injected. Since the gelatinous consistency also occurred in the negative control without ATP, as well as without addition of enzyme it seems to be a non-pGpp or RelS associated effect.

Next, it was interesting to study the dependence of RelS_Cg with regard to different reaction parameters. In addition to RelS_Cg, the enzyme Rel_Cg was also characterized to compare the characteristics of monofunctional SAS enzymes and the bifunctional multi-domain RSH proteins. In order to obtain results with high information value on the physiological significance of the enzymes analyzed, the corresponding starting parameters were adapted to natural values as far as possible. Therefore, physiological Mg²⁺ concentrations of 5 mM (Nierhaus, 2014) and the standard C. glutamicum growth temperature of 30°C were used. However, preliminary experiments showed low activities for Rel_Cg using the optimal C. glutamicum cultivation pH of 7. For this reason and due to a lack of data regarding SAS proteins, the characterization was carried out starting from pH optimum values of bifunctional RSH enzymes at pH 8 (Sajish et al., 2007), to increase comparability. Since the values chosen do not correspond to the RelS optimum conditions retrospectively, suboptimal specific activities were achieved in the individual measurements. For RelS_Cg, a typical bell-shaped curve with a maximum activity in the neutral range of approximately pH 6.5 was found in the analysis of its pH dependency (Figure 6A). The activity extended to a pH range of approximately pH 5.5–pH 9.25. Since only slight differences of the activity values were determined in the various buffer systems used for pH characterization, the activity appears to be largely independent of the buffer-active components phosphate, citrate, or ammonium. In contrast, a distinctly different pH dependency was found for the bifunctional RSH enzyme Rel_Cg. The enzyme had no measurable pyrophosphokinase activity up to a pH value of 6.25. At higher pH values, an increase in the pppGpp synthesis activity occurred, which reached a maximum value from pH 9 on and still had its full activity at pH 10.

The temperature optimum of RelS was 15°C and at the standard cultivation temperature of 30°C, its activity was only about 70% of the maximum value (Figure 6B). In contrast, a typical exponential increase in activity up to a maximum value at a temperature of 40°C was observed for Rel_Cg. Higher temperatures led to a drastic reduction in pppGpp production. Both enzymes had very low activities at temperatures above 50°C.

The presence of Mg²⁺ ions turned out to be essential for the pyrophosphokinase activity of both RelS_Cg and Rel_Cg, since no activity could be detected without this divalent cation. Under the conditions applied, almost identical pyrophosphokinase activities in the form of saturation curves were observed for both enzymes up to a MgCl₂ concentration of 10 mM (Figure 6C). At higher MgCl₂ concentrations, however, significant differences in activity were observed. For RelS_Cg, an extreme increase in pppGpp synthetase activity was found in the form of a further saturation curve. At a MgCl₂ concentration of 50 mM, a maximum pyrophosphokinase activity of 1.375 μmol min⁻¹ mg⁻¹ was achieved, which was more than 10 times as high as the value reached for 10 mM MgCl₂. By contrast, the activity of Rel_Cg decreased continuously using MgCl₂ concentrations above 10 mM and was only 0.047 μmol min⁻¹ mg⁻¹ at 50 mM MgCl₂.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.