Poly-γ-glutamic acid (γ-PGA) is an industrially relevant biopolymer produced by several Bacillus species. γ-PGA consists of D- and/or L-glutamic acid monomers that are linked by gamma-amide bonds. As a water-soluble, non-toxic, edible, and hygroscopic polymer, γ-PGA is used in various applications including medicine (Park et al., 2021), food (Yu et al., 2018), wastewater treatment (Li et al., 2020), and agriculture (Pereira et al., 2017). γ-PGA producing strains are grouped into glutamate-dependent and glutamate-independent producers based on their ability to produce γ-PGA only with or even without the addition of exogenous glutamate. As the glutamate for glutamate-dependent producers accounts for up to 50% of the production costs (Li et al., 2021), γ-PGA synthesis with a glutamate-independent strain is favorable for cost-effective γ-PGA production. However, to date the γ-PGA titers obtained with glutamate-independent strains are insufficient for industrial production (Sirisansaneeyakul et al., 2017).

Recently, metabolic engineering approaches aimed for increasing the γ-PGA titers with decreased production costs. To increase γ-PGA titer especially with glutamate independent strains, the main metabolic engineering targets are an increased glutamate precursor supply, the prevention of γ-PGA degradation, and the deletion of by-product synthesis routes. Approaches to enhance the precursor supply included the use of silencing RNAs to repress the expression of rocG encoding a glutamate dehydrogenase. The repression of rocG successfully increased the intracellular glutamate supply and thereby enhanced the γ-PGA production (Feng et al., 2014). Inspired by glutamate production in Corynebacterium glutamicum. a more energy-efficient glutamate synthesis without ATP consumption was enabled by introduction of the C. glutamicum glutamate synthesis machinery. Moreover, a metabolic toggle switch controlling the expression of the ɑ-oxoglutarate dehydrogenase complex was investigated, as this operon was shown to be completely repressed during glutamate synthesis in C. glutamicum (Feng et al., 2017). To increase the flux from 2-oxoglutarate to glutamate, a gene deletion of odhAB or sucCD, encoding the 2-oxoglutarate dehydrogenase and the succinyl-CoA synthetase, respectively, has been studied. The deletions successfully channeled the metabolic flux towards glutamate, thereby increasing the γ-PGA titer (Massaiu et al., 2019).

Besides the increase in precursor supply, the deletion of γ-PGA depolymerases was reported to optimize the γ-PGA synthesis. Mainly, the three genes cwlO, ggt, and pgdS encoding the D, L-endopeptidase CwlO, gamma-glutamyl transferase Ggt, and the gamma-D, L-glutamyl hydrolase PgdS, respectively, were studied in various Bacillus strains, highlighting strain-specific differences in the effect of the gene deletions. While the double deletion of pgdS and ggt improved the γ-PGA synthesis in a B. subtilis strain (Scoffone et al., 2013), the same deletions had no effect in B. amyloliquefaciens. However, the double deletion of pgdS and cwlO increased the production in B. amyloliquefaciens (Feng et al., 2014). The detailed analysis of single, double, and the triple deletion mutant of the genes cwlO, ggt, and pgdS in B. subtilis using on-line measurement of oxygen transfer rate (OTR) and viscosity recently identified the triple gene deletion as a beneficial modification for γ-PGA synthesis in glucose minimal medium (Hoffmann et al., 2022).

The main by-products during γ-PGA synthesis with B. subtilis in a medium containing glucose, yeast extract, and glutamic acid are 2,3-butanediol and acetoin. Furthermore, minor amounts of acetate and lactate are formed (Zhu et al., 2013). To prevent by-product synthesis and thereby increase γ-PGA production, the deletion of the genes involved in 2,3-butanediol and acetoin production, bdhA and alsSD, respectively, are major engineering targets. The deletion of pta, encoding the phosphotransacetylase that is involved in the synthesis of the by-product acetate, was reported to slightly increase γ-PGA production in B. amyloliquefaciens (Feng et al., 2015). Feng et al. (2015) further reported a beneficial effect for the deletion of the eps and lps operons that are responsible for the synthesis of extracellular polysaccharides and lipopolysaccharides, respectively.

In recent years, rational approaches enabled great progress in the design of whole-cell biocatalysts (Lin and Tao, 2017). These rational approaches were greatly accelerated through the integration of omics data (Amer and Baidoo, 2021). Among these omics technologies, the field of metabolomics is particularly useful to guide metabolic engineering approaches (Putri et al., 2013). Metabolomics studies were successfully applied over a wide range of organisms and products, covering, e.g., amino acid production in yeast (Gold et al., 2015), flavonoid production in tomatoes (Bovy et al., 2007), and biofuel production in microalgae and cyanobacteria (Kato et al., 2022). These examples highlight metabolome analysis as an important tool to guide rational approaches towards metabolic bottlenecks and promising engineering targets. This can also be applied to improve the γ-PGA synthesis with B. subtilis.

This study focused on further improving γ-PGA synthesis of a glutamate-independent B. subtilis 168 derivative (Halmschlag et al., 2019) using glucose as sole carbon source. Minimizing by-product formation was here chosen as Metabolic Engineering target, resulting in strain B. subtilis ΔBP (delta by-products) with deletions of genes bdhA, alsSD, pta, yvmC, and cypX. An in-depth analysis of B. subtilis ΔBP variant PG3 using on-line viscosity measurements and metabolome analysis was carried out to identify strain characteristics and potentially further metabolic engineering targets. Efficient γ-PGA synthesis was demonstrated using modified Medium E that is frequently used for γ-PGA production. The study highlights the value of combining metabolic engineering and in-depth strain characterization for resource- and cost efficient γ-PGA production.

Reagents. 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 (MO, United States).

Strains, plasmids and growth conditions. The bacterial strains and plasmids used and constructed in this study are listed in Table 1. All cloning steps were carried out in E. coli DH5α. The recombinase positive E. coli strain JM101 was used for plasmid propagation prior to the transformation of B. subtilis. The previously reported strain B. subtilis Δspo (Halmschlag et al., 2019) that is a derivative of the manPA-negative strain B. subtilis IIG-Bs2 (Wenzel and Altenbuchner, 2015) was used as starting strain for metabolic engineering.

For plasmid construction and counter selection, 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. The bacterial strains and plasmids used and created in this study are listed in Table 1.

Deletion plasmid construction. The plasmids used for the deletion of genes and integration of the promoter upstream of the pgs operon were assembled using the NEB HiFi DNA Assembly Kit (NEB, Frankfurt am Main, Germany). For the integration of the P_pst or P_xyl promoter, the promoter sequences were amplified from B. subtilis 168 genomic DNA with the primer pairs BS-019/BS-020 or BS-009/BS-010 (see Supplementary Table S1). The two target sequences (TS) for integration and subsequent removal of the plasmid were amplified from B. subtilis 168 genomic DNA, using the primer pairs BS-068/BS-069 and BS-070/BS-071 for P_pst integration or TS1_fwd/TS1_rev and TS2_fwd/TS2_rev for P_xyl integration. The TS1 and TS2 regions for the deletion of the genes alsSD, bdhA, pta, yvmC and cypX, xylAB, and ldh were amplified similarly by using the primer pairs BS-029/BS-030 and BS-031/BS-032, BS-033/BS-034 and BS-035/BS-036, BS-037/BS-038 and BS-039/BS-040, BS-053/BS-054 and BS-055/BS-056, BS-085/BS-086 and BS-087/BS-088 as well as BS-325/BS-326 and BS-327 and BS-328, respectively. The fragments were assembled with the vector backbone pJOE-8739 linearized by BS-020/BS-021. Vector assembly with NEB HiFi DNA assembly was performed as described by the manufacturer. The assembled plasmids were transformed into chemically competent E. coli DH5α cells. Correctly constructed plasmids as confirmed by PCR were sequenced and retransformed to E. coli JM101 to obtain the plasmids for B. subtilis transformation.

Strain development. The strain B. subtilis Δspo was transformed with the constructed deletion plasmids to delete the genes and integrate the promoter. The transformation was performed according to the Paris Method (Harwood and Cutting, 1990). Deletions were carried out by the markerless gene deletion system from Wenzel and Altenbuchner (Wenzel and Altenbuchner, 2015). Briefly, transformed cells were selected on LB agar plates containing spectinomycin. Afterwards, the counterselection with mannose containing agar plates was used to select for cells that lost the plasmid by homologous recombination during incubation in LB without selection pressure. The plasmid loss was confirmed by both spectinomycin sensitivity of the cells and by colony PCR, while the successful deletion of genes or integration of the promoter upstream of pgs was confirmed by sequencing.

Cultivation for metabolite analysis. The main cultures of B. subtilis for the analysis of growth and metabolites were carried out in a glucose minimal medium as previously described (Halmschlag et al., 2020b). The minimal medium contained per liter: 20 g glucose, 41.9 g MOPS buffer, 7 g NH₄Cl, 0.5 g KH₂PO₄, 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. (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 pH of the medium was set to 6.8 with NaOH. The two step precultures were cultivated first on LB medium overnight and second in medium identical to the main culture medium. The LB medium contained per liter: 10 g tryptone, 5 g yeast extract and 10 g NaCl at pH 7.4. The main cultures were inoculated to an OD₆₀₀ of 0.1 from the preculture. The cultures were incubated on a rotary shaker at 37°C with shaking at 200 rpm (25 mm shaking diameter) for experiments in glucose minimal medium (Infors Multitron, Bottmingen, Switzerland). If not stated differently, all cultivations were performed in triplicate, and the reported data represent the mean values.

Cultivation in Medium E. Moreover, γ-PGA production in modified Medium E was investigated, which was modified according to Leonard et al. (Leonard et al., 1958) and ultimately composed of per liter: 30 g glycerol, 15 g citrate, 20 g L-glutamate, 4 g NH₄Cl, 0.5 g K₂HPO₄, 0.5 g MgSO₄·7H₂O, 0.04 g FeCl₃·6H₂O, 0.15 g CaCl₂·2H₂O, 0.104 g MnSO₄·H₂O, 41.9 g MOPS buffer (Mitsunaga et al., 2016). 30 g L^-1 xylose was added to experiments with B. subtilis PG44 and PG55 at the beginning of the cultivation to induce expression of pgs based on the xylose-inducible promoter P_xyl (Halmschlag et al., 2020b). The pH of Medium E was adjusted to 7.2 using 10 M NaOH. Precultures were performed as indicated above. The cultures were incubated on a rotary shaker at 37°C with shaking at 300 rpm (50 mm shaking diameter) for experiments in Medium E (Infors Multitron, Bottmingen, Switzerland). All cultivations were performed in triplicate, and the reported data represent the mean values.

Biomass monitoring. The biomass concentration was determined by OD₆₀₀ measurements of samples taken from the shake flasks using a cell density meter (Amersham Bioscience, Little Chalfont, UK). Additionally, cell growth was monitored online with a Cell Growth Quantifier (CGQ; Scientific Bioprocessing (SBI), Baesweiler, Germany). The biomass concentration is given as cell dry weight (CDW) calculated from the OD₆₀₀ using the following correlation equation that was determined empirically: CDW [g L^-1] = 0.5381*OD₆₀₀ + 0.0074. The correlation of optical density and cell dry weight was determined for the reference strain B. subtilis Δspo. Various dilutions of fermentation broth were prepared in triplicates and the OD₆₀₀ was determined for each dilution. A certain volume of cell suspension that was adjusted to the biomass concentration, was centrifuged to obtain a cell pellet. The pellet was resuspended in 100 µL of 0.9% NaCl solution. The concentrated cell suspension was applied to a cellulose filter of known weight that was dried at 60°C for more than 24 h. The biomass dry weight was determined after drying the filter for 24 h to determine the correlation of CDW and OD₆₀₀.

Cultivation for on-line monitoring (OTR and viscosity measurements). On-line monitoring of the respiration activity determining the oxygen transfer rate (OTR) with the Respiration Activity Monitoring System (RAMOS (Anderlei and Büchs, 2001; Anderlei et al., 2004)) was carried out for the main cultures as previously described (Halmschlag et al., 2020a). Commercial versions of this system are available from Kuhner AG, Birsfelden, Switzerland or HiTech Zang, Herzogenrath, Germany.

The first pre-culture in LB medium was inoculated from 200 µL of cryo stocks. After 5 h cultures were harvested, centrifuged at 11,000 rpm, 4°C for 5°min, resuspended in minimal medium, set to an OD₆₀₀ of 0.1 and used for the second pre-culture. The second pre-culture was harvested after 24 h of cultivation, centrifuged at 11,000 rpm, 4°C for 5°min, resuspended in minimal medium, set to an OD₆₀₀ of 0.1 and used to start the main culture. The main culture was transferred to RAMOS flasks for on-line measurement and to additional Erlenmeyer flasks for off-line samples. The culture conditions for pre-cultures and the main culture were as follows: 250 mL shake flasks, filling volume 20 mL, shaking frequency 350 rpm, shaking diameter 50 mm, temperature 37°C, initial OD₆₀₀ = 0.1.

The on-line viscosity measurement was carried out using a device developed by Sieben et al. (2019) as previously described (Halmschlag et al., 2020a; Hoffmann et al., 2022).

Batch fermentation. Batch fermentations were carried out in stirred tank bioreactors (BioFlo120; Eppendorf, Hamburg, Germany) as previously described (Halmschlag et al., 2019). Cell growth was monitored on-line using the cell growth quantifier for bioreactors (CGQ BioR; Scientific Bioprocessing (SBI)), where indicated and wherever, the culture conditions were kept constant with the shake flask experiments: a temperature of 37°C and an initial OD₆₀₀ of 0.1. The pH was controlled at pH seven by the addition of 2 M HCl or 2 M NaOH. The filling volume was 0.5 L, the aeration rate was set to one vvm, and the stirrer speed was at 800 rpm.

Analysis of γ-PGA with CTAB. The γ-PGA production was analyzed using the cetyltrimethylammonium bromide (CTAB) assay as previously described (Halmschlag et al., 2019). Briefly, cells were separated by centrifugation (20 min, 5,000 g, 4°C). Subsequently, three-times the sample volume of pure ethanol was used to precipitate the γ-PGA. The precipitated γ-PGA was resuspended in distilled water. For measurement, the γ-PGA sample was mixed with CTAB solution, and the turbidity was measured at 400 nm with a Synergy MX microplate reader (BioTek Instruments, Winooski, United States). The γ-PGA concentration was calculated based on a calibration curve using standards of 1 MDa γ-PGA (Henkel AG and Co. KGaA, Düsseldorf) in the range of 0.1–10 μg mL^-1.

Analysis of γ-PGA with GPC. The concentration and molecular weight of γ-PGA was determined by gel permeation chromatography (GPC) as previously described (Halmschlag et al., 2020b). The GPC system (HLC-8320GPC) was equipped with a TSKgel GMPWxl column (300 × 7.8°mm, Tosoh Bioscience GmbH, Stuttgart, Germany) and a TSKgel PWxl guard column (40 × 6 mm, Tosoh Bioscience). The mobile phase used was 30 mM KNO₃ at a flow rate of 1 mL min^-1. The temperature of the GPC system was maintained at 40°C, and 30 µL of each purified, filtered γ-PGA sample was injected. To determine the concentration of γ-PGA in the samples, γ-PGA sodium salt was used as standard. The calibration curve for the molecular weight determination was established using poly(styrene sulfonate) sodium salt standards with a peak molecular weight (Mp) in the range of 4.21–976 kDa (PSS Polymer Standards Service GmbH, Mainz, Germany).

Analysis of glucose, acetoin, 2,3-butanediol and acetate. The consumption of glucose and production of by-products such as acetoin, 2,3-butanediol and acetate was analysed by HPLC-RI using the GL-7400 device coupled with a column oven GL-7432 (GL Science, Tokyo, Japan). The HPLC system was equipped with an Aminex HPX-87H column (7.8 × 300 mm, Bio-Rad, CA, United States). 5 mM H₂SO₄ was used as a mobile phase at a flow rate of 0.6 mL min^-1 for 25 min. The column oven temperature was kept at 65°C. Cell-free culture supernatant was filtrated with a Mini-Uni prep PTFE filter (pore size 0.45 mm, GE Healthcare UK). 10 μL of the filtrate were injected to the HPLC system. The concentration of extracellular pyruvate was quantified using the Amplite® Colorimetric Pyruvate Assay Kit (Cat# 13821, AAT Bioquest, California, United States) according to the manufacturer’s instructions. The concentration of analytes was calculated by calibrations curves established for all standards.

Sample Preparation for metabolome analysis. For metabolome analysis, culture samples were harvested during the exponential phase. The samples were prepared as previously described (Halmschlag et al., 2020a). Briefly, a constant biomass concentration was filtered with a PVDF filter (pore size, 0.45 mm) using a vacuum. After washing the sample with 300 mM NH₄HCO₃, the cellular metabolism was quenched by soaking the filter in liquid N₂. The samples were stored at −80°C until metabolite extraction. The metabolites were extracted with 1.875 mL of extraction solvent (1:2:2, H₂O:MeOH:chloroform, including 7 nM 10-camphorsulfonic acid as internal standard) per sample. The extracted metabolites were obtained by centrifugation and supernatant was transferred to a clean tube, vacuum concentrated, and freeze-dried overnight. The obtained pellet was stored at −80°C until analyzed by LC-MS.

Ion-pair-LC/MS/MS Analysis. The extracted metabolite samples were analyzed by 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). As previously described (Halmschlag et al., 2020a), 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). For the mobile phase, a gradient of a mixture of solvent A and B was used, where solvent A was 10 mM tributylamine and 15 mM acetate in ultra-pure water and solvent B was pure methanol. The flow rate was set to 0.2 mLmin^-1. For gradient elution of metabolites, the concentration of B was increased from 0 to 15% after 1 min with a gradient of 30% min^-1, held for 1.5 min, increased to 50% over 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 then 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^-1; heat block temperature, 400°C; and nebulized gas flow, 2 L min^-1. Per sample, 3 µL was injected into the ion-pair-LC/MS/MS for metabolite analysis. The samples for analysis were obtained by resuspending dried metabolite extraction samples in 50 µL of ultra-pure water.

Metabolite data processing and analysis. For metabolite data analysis, the peak area was calculated using MRMPROBS ver. 2.38 and was checked manually. The data were 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 orthogonal partial least square discriminant analysis (OPLS-DA). The t-tests were performed using MATLAB (MathWorks, MA, United States).

To improve γ-PGA synthesis in B. subtilis with glucose as sole carbon source, the carbon flux must be directed towards the TCA cycle and ultimately glutamate to ensure sufficient precursor supply for the γ-PGA synthetase. The resulting strain with impaired by-product synthesis was analysed in detail using a metabolomics approach and online viscosity measurements. Increased γ-PGA production with either glucose as sole carbon source or with the γ-PGA production Medium E and additional metabolic engineering targets are presented.

The cost-efficient γ-PGA production without addition of γ-PGA precursors such as glutamate or citrate requires the rerouting of the metabolic flux towards glutamate. The de novo glutamate synthesis was reported as bottleneck for γ-PGA production without addition of exogenous glutamate as several studies reported significantly increased γ-PGA titers by addition of exogenous glutamate (Lin et al., 2016; Mitsunaga et al., 2016). Yao et al. investigated the origin of carbon in γ-PGA using 13C-labeled glucose and L-glutamate as carbon sources for a B. subtilis producer, showing that only up to 9% of the γ-PGA was derived from glucose (Yao et al., 2010). Here, we engineered γ-PGA producing B. subtilis 168 derivative strains with or without deletions of by-product synthesis pathways to efficiently direct the carbon flux from glucose towards γ-PGA. The presented results demonstrate the high potential of metabolic flux rerouting with a 3.7-fold or 1.5-fold increase in γ-PGA production for the deletion mutant in glucose minimal medium or modified Medium E, respectively. Integration of the engineered strains into optimized PGA production processes has the potential to decrease the substrate cost for PGA synthesis. Feng et al. (2015) previously reported improved γ-PGA synthesis with an engineered B. amyloliquefaciens strain that lacked the genes eps, sac, lps, and pta. The γ-PGA production with the reported strain slightly increased from 3.8 to 4.15 g L^-1. The authors observed a larger increase to 9.18 g L^-1 by additional deletion of genes encoding γ-PGA degrading enzymes. Further, the beneficial effects on the deletion of genes odhAB or sucCD, encoding the 2-oxoglutarate dehydrogenase and the succinyl-CoA synthetase, respectively, was reported (Massaiu et al., 2019). The deletions successfully channeled the metabolic flux towards glutamate and increased the γ-PGA titer with citrate and glucose as carbon sources. The metabolome analysis carried out here identified the citrate synthase as major bottleneck for cultivations with glucose as sole carbon source. The variations of the influence of investigated engineering strategies throughout various studies underline that the impact of gene deletion strongly depends on the investigated strain and culture conditions.

The improved γ-PGA synthesis with the by-product deletion mutant was not only demonstrated with glucose as carbon source but also in modified Medium E. In this modified version of the medium, only 30 g L^-1 glycerol was used as compared to commonly used 80 g L^-1, aiming for an increased γ-PGA yield due to the streamlined carbon metabolism and anticipated enhanced de novo glutamate synthesis. Using the xylose-inducible promoter P_xyl, the γ-PGA titer produced in modified Medium E by the by-product deficient strain B. subtilis PG55 was increased by 41% (14.46 g L^-1 vs. 10.24 g L^-1 after 72 h) as compared to the parental strain B. subtilis PG44. Moreover, the product yield was improved from 0.33 g_γ-PGA/g_glycerol to 0.51 g_γ-PGA/g_glycerol provided by the newly generated knockout strain B. subtilis PG55 in this medium highlighting the improved synthesis capability of the deletion mutant even with reduced substrate concentration. Limited metabolization of available glycerol might indicate a metabolic arrest of B. subtilis PG55 induced by the prominent decrease in culture pH, which is also depicted by the reduced maximal optical density of the strain. Accumulation of extracellular pyruvate poses one possible explanation for this observation. Although the acidification of the medium may be addressed by active pH control using a bioreactor setup, this has no effect on the metabolic imbalance present in the deletion mutant. Future research approaches will focus on stabilizing the culture pH by rerouting carbon flux around bottlenecks identified in this work, thereby avoiding accumulation of potential acidic metabolites. Notably, the γ-PGA titers obtained in this study ranging from less than 1 g L^-1 in glucose minimal medium to 14.46 g L^-1 in modified medium E are lower than reported titers of more than 40 g L^-1 (Zhao et al., 2013; Zhang et al., 2019), indicating the potential for further optimization of production strain and cultivation conditions. Despite the lower production, the genetic streamlining of the carbon metabolism through deletion of by-product formation pathways improved both γ-PGA yield and titer in the deletion mutant independent of the promotor of synthetase genes and the used medium for polymer synthesis. Furthermore, the metabolic bottlenecks identified in this work will guide the future strain optimization that will be combined with optimized cultivation conditions.

The detailed metabolome analysis as well as the detection of accumulating pyruvate for deletion mutants in Medium E identify the citrate synthase as important metabolic engineering target to further optimize the γ-PGA production with by-product deletion mutants. The conversion of this overflow metabolite into a TCA cycle intermediate can further channel the metabolic flux towards γ-PGA precursor molecules. Strengthening the flux of pyruvate into the TCA cycle and therefore towards γ-PGA production will be object of future studies. The advances in γ-PGA product titer and yield for the deletion mutants even in an acidic environment highlight the potential of the engineered strains to harness the already increased precursor supply induced by a reduction of overflow metabolism. Interestingly, by-product formation was not completely abolished by deletion of the respective pathways. Lactate, acetoin and acetate continued to be produced by B. subtilis PG55 in modified Medium E due to unknown metabolic pathways. This observation is in unison with the work of Zhu et al., who detected acetate formation in Corynebacterium glutamicum even in the absence of the related metabolic synthesis pathways (Zhu et al., 2013). In Escherichia coli, the two-electron flavoprotein pyruvate oxidase PoxB catalyses the decarboxylation of pyruvate to acetate and CO₂ independent of the classic synthesis route involving Pta and Ack (Li et al., 2007). In case of Streptococcus pneumoniae, the analogous pyruvate oxidase SpxB catalyses the synthesis of acetyl-phosphate from pyruvate, which is then further converted to acetate (Echlin et al., 2016). Based on these examples it can be hypothesized that a similar system is present in engineered B. subtilis 168, although so far only marginally investigated. According to the NCBI gene database, an acetyl-phosphate generating pyruvate oxidase is indeed present in B. subtilis 168 at location NC_000963.3 (BP 488830 to 490,554, gene ID 938239). The genomic location is annotated as putative pyruvate oxidase YdaP in B. subtilis 168 in the SubtiWiki database (Pedreira et al., 2022). YdaP has been identified as an acetate forming pyruvate oxidase in Bacillus licheniformis 9A, highlighting its potential role in acetate or acetyl-phosphate synthesis in the constructed deletion mutant B. subtilis PG55 (Wani Lako et al., 2018). Since B. subtilis PG55 is only deficient in pta which is responsible for the conversion of acetyl-CoA to acetyl-phosphate, acetate formation through AckA might occur based on a Pta-independent source of acetyl-phosphate as provided by a B. subtilis pyruvate oxidase. The deletion of putative pyruvate oxidase genes or the acetate kinase gene ackA with their respective effect on acetate production in B. subtilis PG55 will be object of future research.

The presence of acetoin produced by B. subtilis deficient in alsSD might indicate a novel synthesis route. Previously, a successful prevention of acetoin formation upon deletion of the alsSD operon was reported (Ma et al., 2018). The overflow metabolism of B. subtilis not only offers the possibility to redirect accumulating pyruvate into neutral fermentation products but also enables regeneration of NAD + through the synthesis of lactate and 2,3-butanediol (Nakano et al., 1997). Deletion of these synthesis routes might therefore result in an imbalance in redox metabolism, hindering cell growth and ultimately γ-PGA production. Indeed, redox state alteration in B. subtilis has already been shown to accelerate the production of NAD + -dependent products such as acetoin (Zhang et al., 2014). However, introduction of an NADH oxidase into by-product deficient B. subtilis showed no effect in synthesis of N-acetylglucosamine (Gu et al., 2019). Investigations whether the introduction of an NADH oxidase into the engineered B. subtilis mutants reported in this study influences growth or γ-PGA production will be subject of our future work.

Aiming at a cost-efficient and sustainable γ-PGA production, the presented metabolic engineering results demonstrate the high potential of B. subtilis as production host. The provided information will be used to guide metabolic engineering strategies in further studies to optimize γ-PGA production with either glucose as sole carbon source or in the γ-PGA production Medium E.