D-Glucose, L-glutamic acid, NH₄Cl, K₂HPO₄, MgSO₄ × 7H₂O, CaCl₂ × 2H₂O, MnSO₄ × H₂O, FeCl₃ × 6H₂O, ZnSO₄ × 7H₂O, Na₂-EDTA, CuSO₄ × 5H₂O and CoCl₂ × 6H₂O were purchased from Carl Roth GmbH + Co., KG (Karlsruhe, Germany). LC/MS-grade ultra-pure water, HPLC-grade chloroform, acetic acid, H₂SO₄ and NH₄HCO₃ were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). 10-Camphorsulfonic acid and tributylamine were purchased from SigmaAldrich (St. Louis, MO, United States).

The bacterial strains and plasmids used and created in this study are listed in Table 1. All cloning steps were carried out in Escherichia coli DH5α. The recombinase positive E. coli strain JM101 was used to obtain plasmids for the transformation of B. subtilis. The strain B. subtilis Δspo (Halmschlag et al., 2019) was used for the development of the promoter library and as production host for γ-PGA production.

For plasmid construction and counter selection, all strains were cultivated at 37°C in LB medium containing 100 μg/mL spectinomycin or 0.5% (w/v) mannose when required. For γ-PGA production and metabolome analysis, the B. subtilis strains were grown in glucose minimal medium.

For the use of xylose as inducing agent, the xylAB genes encoding the xylose isomerase and xylulokinase were deleted in the background of strain B. subtilis Δspo. The thereby obtained strain is not capable of metabolizing xylose. Subsequently, the native P_xyl promoter was integrated into the genome of B. subtilis Δspo ΔxylAB upstream of the pgs operon. For that purpose the plasmid pBs-02 was generated. The plasmid contains the backbone of pJOE8739 linearized with primers BS-25/26 (Supplementary Table S1), P_xyl amplified with BS-09/10 and two integration sites up- and downstream of the native promoter P_pgs that were amplified with the primer pairs TS1_fwd/TS1_rev and TS2_fwd/TS2_rev. The created strain expressing pgs under control of the xylose-inducible promoter Pxyl was designated as B. subtilis PG10.

The reporter gene gfpmut3 was amplified from plasmid pBSMul1-gfpmut3 with primer pair BS-111/BS-112. A part of the amyE locus was amplified with BS-113/-114 to be used as homologous sequence for integration. For the assembly of plasmid pBS-21, the vector backbone pBs-20 was linearized with BS-25/-26. The backbone includes a spectinomycin resistance gene (SpcR), an origin of replication for E. coli, the rop gene with regulator function for replication and the second part of the amyE locus for homologous recombination into the B. subtilis genome. The native promoter P_veg was amplified from genomic B. subtilis DNA using the primers BS-109 and BS-110. The promoter fragment was fused with the gfpmut3 gene, the amyE upstream, and the pBs-20 based backbone by DNA HiFi Assembly (NEB, Frankfurt am Main, Germany). The thereby constructed plasmid pBs-21 was used as template to amplify two fragments, one containing the first homologous part (amyE-front) and the spectinomycin resistance gene (SpcR) and the other one containing the promoter, gfp gene and the second homologous part (amyE-back). The two fragments were amplified using the primer pairs BS-115/-116 and BS-117b/118, respectively. The primer BS-117b was a degenerated primer containing the NNN sequence to vary the −35 box sequence of the promoter. After fusion of the fragments by PCR and cloning into the pJET vector, Bacillus was transformed with the plasmids including promoter variants.

The B. subtilis colonies with varying promoter sequences were analyzed with respect to the promoter activity using the BioLector (m2p-labs, Baesweiler, Germany). The biomass was measured at 620 nm extinction/emission at gain 50. The fluorescence of GFP was detected with 488 nm extinction and 520 nm emission at gain 70. All colonies were analyzed in triplicates and the promoter activity was calculated as mean of the slope calculated as quotient of increase in fluorescence and increase in biomass during the exponential phase.

The promoter sequence information was obtained by sequencing (Eurofins, Ebersberg, Germany). The observed sequences were introduced into pRJ-8 variants by amplification of the complete plasmid using specific primers (BS-180, BS-193, and BS-182 for variant PV35.1, PV35.3, and PV35.26, respectively) with mismatches in the promoter region. The thereby created linear plasmids were phosphorylated with T4 polynucleotide kinase (NEB, Frankfurt am Main, Germany) for 30 min at 37°C and subsequently ligated with T4 ligase (NEB, Frankfurt am Main, Germany). First, E. coli DH5α was transformed with these plasmids. The plasmids were purified and the promoter sequences were determined. Second, B. subtilis was transformed with those three plasmids. Successfully transformed colonies were selected on LB agar plates containing spectinomycin. The correct integration was checked by PCR and sequencing.

For the cultivation of B. subtilis glucose was used as sole carbon source in a minimal salt medium for the main cultures. The minimal medium contained per liter: 20 g glucose, 7 g NH₄Cl, 0.5 g MgSO₄, 0.15 g CaCl₂, 0.1 g MnSO₄, 0.04 g FeCl₃, and 1 mL of a trace element solution. The trace element solution [modified trace element solution after: (Wenzel et al., 2011)] contained per liter: 0.54 g ZnSO₄^∗7 H₂O, 30.15 g Na₂-EDTA, 0.48 g CuSO₄^∗5 H₂O and 0.54 g CoCl₂^∗6 H₂O. The medium was buffered using 0.1 M potassium phosphate buffer at pH 7. All media components were sterilized by autoclaving for 20 min at 121°C except for the trace element solution, which was sterile filtered. For xylose induction of pgs expression with B. subtilis PG10, xylose was added in concentrations of 0, 1, 5, 10, 15, 25, 40, 50, 80, 100, and 200 mM. The first pre-culture in LB medium was inoculated from a glycerol stock. The LB medium contained 10 g L^–1 tryptone, 5 g L^–1 yeast extract, and 10 g L^–1 NaCl at pH 7.4. A second pre-culture in glucose minimal medium was inoculated with cells from the LB medium and grown overnight. The main cultures were inoculated to an OD₆₀₀ of 0.05 from pre-culture. The cultivations were carried out in 250 mL Erlenmeyer flasks filled with 25 mL medium. The cultures were incubated on a rotary shaker (Bio-Shaker BR-3000LF, Taitec, Saitama, Japan) at 37°C, 200 rpm, 25 mm shaking diameter. All cultivations were performed in triplicates whereas the reported data represents the mean. The cell growth was monitored by OD₆₀₀ measurements of samples taken from the shake flasks using the GeneQuant 100 spectrophotometer (GE Healthcare United Kingdom, Ltd., Buckinghamshire, United Kingdom). Additionally, the cell growth was monitored online with the Cell Growth Quantifier (CGQ; Aquila Biolabs, Baesweiler, Germany). The biomass concentration is given as cell dry weight (CDW) calculated from the OD₆₀₀ using the correlation equation CDW [g L^–1] = 0.5381^∗OD₆₀₀ + 0.0074.

The production of γ-PGA was analyzed with a cetyltrimethylammonium bromide (CTAB) assay. To purify γ-PGA, the culture broth was centrifuged for 30 min (10,000g, 4°C) to separate the cells. The double amount of ethanol was added to 15 mL of supernatant containing γ-PGA to precipitate the γ-PGA (1:2 ratio, sample:ethanol). The solution was incubated at 4°C overnight. By centrifugation for 10 min at 4°C and 10,000g, the precipitated γ-PGA was recovered as a pellet. The pellet was resuspended in 1 mL distilled water. The addition of 100 μL 0.07 M CTAB in 2% (w/v) NaOH to 100 μL of γ-PGA sample led to the precipitation of the γ-PGA/CTAB complex resulting in an increased turbidity that was measured in a plate reader (Synergy Mx plate reader, BioTek Instruments, Inc., Winooski, United States) at 400 nm. A calibration curve using a 1 MDa γ-PGA standard [(Henkel AG & Co., KGaA, Düsseldorf, Germany] was used to calculated the γ-PGA concentration in the culture broth. γ-PGA samples were diluted for the CTAB assay to reach the linear range of the assay up to 0.1 g/L γ-PGA (equaling a turbidity of <0.5).

A sample volume which satisfies the equation sampling volume (mL) ^∗OD₆₀₀ = 5 was filtered with a PVDF filter (pore size 0.45 mm; Merck Millipore, Burlington, NJ, United States) using vacuum. The filter was washed with the doubled amount of 300 mM NH₄HCO₃ (1:2 ratio, sample: NH₄HCO₃). Subsequently, the metabolism was quenched by soaking the filter in liquid N₂. The filter was transferred to a 2 mL sample tube and was stored at −80°C until metabolite extraction. To extract the metabolites, 1.875 mL of extraction solvent (1:2:2 ratio, H₂O:MeOH:chloroform, including 7 nM 10-camphorsulfonic acid as internal standard) was added to the sample tube including the filter. After vortexing, the sample tube was centrifuged at 16,000 g for 3 min at 4°C. 350 μL of the supernatant were transferred to a 1.5 mL sampling tube, concentrated by vacuum centrifugation for 2 h and freeze dried overnight. The pellet was stored at −80°C until analyzed by LC-MS.

The dried sample is resuspended in 50 μL ultra-pure water and is transferred into a conical glass vial. The Ion-pair-liquid chromatography coupled with tandem mass spectrometry (LC/MS/MS) analysis was performed using a Shimadzu Nexera UHPLC system coupled with an LCMS 8030 Plus device (Shimadzu Co., Kyoto, Japan). The system was equipped with a PE capped CERI L-column 2ODS column (2.1 mm × 150 mm, particle size 3 mm, Chemicals Evaluation and Research Institute, Tokyo, Japan). As mobile phase a gradient of a mixture of solvent A and B is used, where solvent A is 10 mM tributylamine and 15 mM acetate in ultra-pure water and solvent B is pure methanol. The flow rate was set to 0.2 mL min^–1. For gradient elution of metabolites, starting from 0% concentration of solvent B, the concentration of B was increased to 15% after 1 min with a gradient of 30% min^–1, hold for 1.5 min, increased to 50% within 5 min and subsequently increased to 100% within 2 min. The 100% solvent B concentration was held for 1.5 min, decreased to 0% from 11.5 min on and held at this concentration for 8.5 min. The column oven temperature was set to 45°C. The MS parameters were as follows: probe position, þ1.5 mm; desolvation line temperature, 250°C; drying gas flow, 15 L/min; heat block temperature, 400°C; and nebulized gas flow, 2 L/min. As a Quality Control (QC) sample, 2 μL of each analyzed sample of a batch were pooled into a vial. For each sample (including the QC sample), 3 μL were injected to the ion-pair-LC/MS/MS for metabolite analysis.

The calculation of the peak area was carried out using MRMPROBS ver. 2.38 and manual inspection of the chromatogram was conducted. The data was normalized according to the peak area of the internal standard, 10-camphorsulfonic acid. SIMCA 13 (Umetrics, Umeå, Sweden) was used for principal component analysis (PCA) and partial least squares projection to latent structures (PLS) analysis.

For the promoter library approach, three newly identified promoter variants based on the P_veg promoter were successfully integrated upstream of the pgs genes to control the expression of γ-PGA synthetase and subsequently to vary the γ-PGA production rate. The variants were constructed by changing the -35 box sequence of the P_veg promoter (TTGACA). The promoter variants differ from the original P_veg promoter in one or two positions. For variant PV35.01 (AAGACA), two positions were varied. The large deviation from the consensus sequence resulted in only 10% of the original activity. The change of G to T in the third position decreased the activity by 50 percent in variant PV35.3. Variant PV35.26 (GTGACA) exhibits an increased activity of 120% compared to the original promoter.

Growth of the created strains was tested by online growth measurements (CGQ) to observe possible influence of higher γ-PGA synthetase expression rates. As shown in Figure 1, the growth curves did not differ significantly. Independent from the promoter activity for pgs, all strains grew to comparably high optical densities.

The titer of γ-PGA production as analyzed by the CTAB assay (Figure 2B) increased with an increasing promoter activity, determined in the BioLector assay (Figure 2A). Hence, the γ-PGA synthesis rate increases with the activity of the promoter controlling the γ-PGA synthetase expression.

The created strains harboring newly generated promoter sequences with different strength were subjected to metabolome analysis to reveal key metabolites that are most correlated with γ-PGA synthesis. For metabolome analysis according to previous studies (Mitsunaga et al., 2016; Fathima et al., 2018), 93 metabolites were detected that are part of the central carbon metabolism including pentose phosphate pathway (PPP), tricarboxylic acid cycle (TCA), and glycolysis intermediates as well as amino acids and nucleotides. The samples were taken after 8 h during exponential growth. The differing γ-PGA concentrations for the tested strains at this time point demonstrate varying γ-PGA synthesis rates. The metabolome data were subjected to partial least square (PLS) analysis (Figure 3). The horizontal separation corresponds to the increasing γ-PGA production from left (low value for PC1) to right (high value for PC1) with R² and Q² of 0.98 and 0.87, respectively. Metabolites that exhibit loading values higher than 0.1, are shown in Figure 3D. Here, negative values correspond to metabolites decreasing with increasing γ-PGA production and metabolites with positive values are increased for higher γ-PGA synthesis rates.

The metabolites with high loading value originate from various metabolic pathways. The most striking loading data for metabolites from the central carbon metabolism is the data for ribose-5-phosphate (R5P), pyruvate, and dihydroxyacetone phosphate (DHAP) with negative values as well as acetyl-CoA with a positive value. Notably, the direct precursors for γ-PGA, glutamate and 2-oxoglutarate hardly change in concentration (loading values of 0.098 and 0.032, respectively). Homeostasis of these metabolites might be explained with their prominent functions in carbon and nitrogen metabolism. While other amino acids such as phenylalanine are increased other amino acids such as serine are lower concentrated during increased γ-PGA production. Serine and R5P are two metabolites which are directly derived from glycolysis intermediates, 3-phosphoglycerate and glucose-6-phosphate (G6P), respectively. The decrease in the metabolite concentrations of serine and R5P can therefore be explained by the higher demand for carbon flux through glycolysis toward the TCA cycle and glutamate. The higher flux toward glutamate is also displayed in the decreased synthesis of precursors for secondary metabolites such as xanthosine 5′-phosphate (XMP). Contrary to the XMP concentration, the acetyl-CoA concentration increases with higher promoter activity. As it was shown for the comparison of a non-γ-PGA producer and a γ-PGA-producer B. subtilis PG1, the TCA activity is limiting the precursor supply for γ-PGA production. Meyer et al., demonstrated that citrate synthase, aconitase, and malate dehydrogenase form a protein complex catalyzing sequential reactions of the TCA cycle (Meyer et al., 2011). Additionally, the 2-oxoglutarate dehydrogenase complex and the glutamate synthase are affected by these protein-protein interactions (Meyer et al., 2011). The concentration of succinyl-CoA increases for increasing γ-PGA synthesis rates. With regard to the interaction of TCA cycle enzymes, the higher flux toward 2-oxoglutarate and glutamate likely additionally results in increased succinyl-CoA concentrations. Hence, the metabolome data suggests an increased TCA cycle activity yet being a limiting step for higher γ-PGA production from glucose as sole carbon source. The decreased concentration of R5P for higher promoter activity is surprising with respect to the glucose-6-phosphate dehydrogenase reaction, which is known as main reaction for the supply of NADPH. NADPH is used as cofactor for glutamate synthesis. Therefore, a higher demand for glutamate under γ-PGA producing conditions also results in a higher demand for NADPH. Here, the high positive loading values for both NAD⁺ and NADP⁺ are remarkable, too. Since the applied metabolite extraction method is not suitable for investigating the redox status of the cells, the increasing NAD⁺ concentrations are likely due to a higher demand for NAD(P)H for increasing glutamate de novo synthesis.

Besides controlling the synthetase expression using a promoter library, the expression was also altered using an inducible promoter. The xylose-inducible promoter P_xyl controlling the expression of the xylAB operon in B. subtilis was cloned to control pgs expression in B. subtilis PG10. The xylose promoter in B. subtilis PG10 contained the binding site for the xylose regulator XylR, but not the binding site for catabolite repression via CcpA. These features enabled the induction of the promoter by xylose addition and prevented the repression during growth on glucose. Moreover, the xylAB operon was deleted to ensure that xylose is only used as inducer and not as substrate for cell growth. In B. subtilis, xylose is imported via the L-arabinose permease encoded by araE. The araE expression is reported to be arabinose induced and glucose repressed (Sa Nogueira and Ramos, 1997). However, B. subtilis Δspo that was used to create the strains in this study grows on xylose without addition of arabinose. Hence, the xylose uptake of the strains used in this study was shown not to depend on arabinose availability. As expected, γ-PGA production increased with increasing xylose concentration (Figure 4). Comparable to the promoter library, a higher synthetase concentration leads to a higher γ-PGA production rate. Controls were the non-γ-PGA producing B. subtilis Δspo as well as the B. subtilis ΔxylAB used to create PG10, which cannot grow on xylose. In comparison to the promoter library approach, B. subtilis PG10 exhibits a lower growth rate (0.43 h^–1 compared to 0.59 h^–1). Moreover, the γ-PGA production rate at full induction is higher for the xylose inducible promoter compared to the production rate of the strongest promoter from the library. The determined γ-PGA titers were used to correlate the xylose concentration with γ-PGA production (Figure 4). A logarithmic dependency between the γ-PGA production as represented by the turbidity and the inducer concentration was determined. Since saturation of the production rate was observed for xylose concentrations higher than approximately 100 mM, further concentrations ranging from 0 mM to 50 mM were used for the metabolome analysis to create more variation in the production rate.

For metabolome analysis, intracellular metabolites for 13 conditions (B. subtilis ΔxylAB with 0 and 200 mM xylose as controls, B. subtilis PG10 with 0, 1, 5, 10, 15, 25, 40, 50, 80, 100, and 200 mM xylose) were analyzed. For each strain and condition, the intracellular concentration of 77 metabolites was determined. The γ-PGA production was calculated based on the correlation between turbidity and xylose concentration obtained from Figure 4. Based on the obtained values for γ-PGA production under the investigated cultivation conditions, a PLS analysis was carried out (Figure 5A). The PLS analysis resulted in a model with R² of 0.92 and Q² of 0.87 (see Figure 5B). The samples were separated in t[1] according to the γ-PGA production rate.

Analogously to the promoter library, metabolites with varying concentrations for differing γ-PGA production rates could be identified (Figure 5). The metabolite concentrations with loading values higher than 0 increase with an increasing production rate. Comparing the two approaches for controlling the γ-PGA production rate, the promoter library and the xylose induction, the identified metabolites with high loading values partly differ. This fact is most likely caused by differences in growth rates and γ-PGA production rates for both approaches. Differences in metabolic demand (Koebmann et al., 2005) result in changes of metabolic fluxes and, hence, in metabolite concentrations. Changes in flux distribution have also been observed due to differing pH of the growth media and a negative correlation between acetoin and 2,3-butandiol and γ-PGA production was reported (Zhu et al., 2013). Regardless of the differences between the two approaches, several amino acids exhibit increased concentrations in both approaches. Especially phenylalanine increased with increasing promoter strength as well as increasing inducer concentration.

Further, for the xylose induction approach mostly metabolites that accumulate for higher γ-PGA production rates were identified. In this approach, the increase in γ-PGA production rate is interconnected with a slight increase in growth rate (Figure 6). Here, phosphoenolpyruvate (PEP) is the metabolite exhibiting the highest correlation with γ-PGA production. The PEP concentration follows the increasing trend of both γ-PGA production rate and growth rate for increasing xylose induction strength. Besides PEP, further intermediates of the central carbon metabolism increase simultaneously with the γ-PGA production rate. Several metabolites originating from glycolysis, pentose phosphate pathway and TCA cycle like G6P, R5P and 2-oxoglutarate belong to the group of metabolites with high loading values. Hence, the increased metabolite concentrations are likely due to an increased overall carbon flux and not directly due to the γ-PGA production rate. However, the higher γ-PGA production rate with increasing xylose concentration increases the demand for a higher flux through the metabolism resulting in higher glucose uptake rates as indicated by the increasing G6P concentrations. Among the metabolites that are negatively related to the higher γ-PGA production rate, glutamine is the most striking one. Since glutamine is a substrate for de novo synthesis of glutamate from glucose, its concentration is strongly connected to the demand for glutamate. The increasing demand for glutamate with higher γ-PGA production also influences the proline and NADP concentrations as indicated by the loading values of these metabolites. Succinate is a further metabolite indicating the changes in glutamate demand since the carbon flux at the 2-oxoglutarate branch point can either be directed toward glutamate or succinate. The obtained positive loading value for succinate indicates an increase of the concentration with increasing xylose induction. However, the absolute loading value is low emphasizing a weaker connection of the succinate concentration to the γ-PGA production rate. The comparison of concentrations for 2-oxoglutarate, succinate and glutamate (Figure 7) demonstrates the connection. First, the effect of higher overall carbon fluxes can also be seen for these metabolite concentrations. The concentrations of these metabolites increase to a certain extent with higher xylose concentration. While 2-oxoglutarate is increasing for xylose concentrations up to 50 mM, the glutamate and succinate concentrations are already decreasing from 25 and 10 mM on. Therefore, the positive loading value for succinate is mostly reasoned in the small concentration for the reference and increasing concentration for low xylose concentrations up to 10 mM. Additionally, the concentrations of the three metabolites at the 2-oxoglutarate branch point reflect the increasing carbon flux to glutamate for stronger induction. This effect is emphasized by the fact that succinate starts decreasing for weaker induction than glutamate. The metabolite measurement results match the increasing growth rate and γ-PGA production rate for the xylose induction approach suggesting higher metabolic fluxes caused by the demand for glutamate for γ-PGA production.