CRISPR-assisted rational flux-tuning and arrayed CRISPRi screening of an l-proline exporter for l-proline hyperproduction

Optimization of CRISPR/Cas9-assisted genome editing methods for C. glutamicum

Efficient methods for genetic manipulation are required for strain development. Genome editing tools based on both CRISPR/Cas9^31,32,33,34 and CRISPR/Cas12a^35,36,37,38 have been developed for C. glutamicum. Comparing these two systems for application in GC-rich C. glutamicum (53.8% GC-content for type strain ATCC 13032), the CRISPR/Cas9 system recognizing G-rich protospacer adjacent motif (PAM) sequences should have a wider genome targeting scope than the CRISPR/Cas12a system recognizing T-rich PAMs³⁹. Besides, the CRISPR/Cas9 system can distinguish a single nucleotide change in the seed sequence, which is preferred for precise editing⁴⁰. However, the CRISPR/Cas9 system is cytotoxic to C. glutamicum and previous methods usually suffer from few transformants and low accuracy, especially for manipulation of large DNA fragments^32,34.

The double-stranded DNA break (DSB) generated by the CRISPR/Cas9 can be a two-edged sword, which is lethal to most bacterial cells but can be used for counter-selection of edited cells⁴¹. Therefore, the unwanted and untimely leaky expression of Cas9 proteins should be minimized. To optimize the CRISPR/Cas9 system for C. glutamicum, we first constructed a temperature-sensitive plasmid that expressed Cas9 and gRNA and harbored homologous recombination (HR) arms. With a very low dosage of isopropyl-β-d-thiogalactopyranoside (IPTG) (0.01 mM) to induce Cas9-facilitated counter-selection, a few dozen transformants and a moderate editing efficiency of 33.3% were obtained for deletion of a 1.7 kb DNA fragment (cgl0620–cgl0622) with two 1.0 kb HR arms. Interestingly, with the increase in IPTG dosage, the transformant number and editing efficiency gradually decreased (Fig. 1a and Supplementary Fig. 1), again demonstrating the importance of controllable Cas9 expression in genome editing.

Next, the expression of Cas9 was regulated by using a perfectly symmetric lac operator (LacO) that binds the lac repressor (LacI) very tightly⁴² and a weaker ribosome binding site (RBS). By using a green fluorescent protein (GFP) as a reporter, it was found that these modifications almost eliminated leaky expression. Gradually enhanced gene expression was obtained by increasing the IPTG dosage (Supplementary Fig. 2a). When used for cas9 expression, these modifications on LacO and RBS also reduced the cytotoxicity of cas9 expressing plasmid and produced more transformants (Supplementary Fig. 2b). Then, the optimized system was tested for Cas9-mediated gene deletion. The editing efficiency was largely improved up to 91.3% with 0.05 mM IPTG dosage (Fig. 1b and Supplementary Fig. 1). Hundreds of transformants could be obtained because of the strict control of Cas9 expression. More tests were then conducted to evaluate the optimized system. When shorter HR arms (0.5 kb) were used for deletion of a longer DNA fragment (20 kb fragment in the prophage CGP3), an editing efficiency of 46.4% was obtained, which was over the previously reported efficiency of 26.9% for the same test³⁴. The editing efficiency remained 34.8% for deletion of a 219 kb DNA fragment (complete prophage CGP3), which has not been achieved by previous methods³⁴. For insertion of a 4 kb DNA fragment (an artificial proBAC operon), an efficiency of 43.5% was obtained (Fig. 1c and Supplementary Fig. 3).

With exogenous recombinases, recombination with single-stranded DNA (ssDNA) can be more efficient than recombination with plasmid-borne double-stranded DNA (dsDNA)^32,43. To test genome editing with ssDNAs as templates, the RecT recombinase from the Rac prophage of E. coli was expressed using a low-copy and easily curable plasmid under the control of a strong constitutive promoter P_ddh^*. By providing 90 nt ssDNA templates targeting the rpsL^K43R mutation, triple or single nucleotide changes were efficiently introduced to the target site with 100% efficiencies and over 1000 transformants for each test. The editing efficiency remained 80% when the length of ssDNA was reduced to 60 nt (Fig. 1d and Supplementary Fig. 3). Compared with the CRISPR/Cas9 systems previously developed for C. glutamicum^31,32,33,34, the present system shows overall improved editing efficiency. A routine procedure for genome editing and plasmid curing takes as short as four days (Supplementary Fig. 4). The sizable quantity of transformants also make this system easy to operate.

Screening γ-glutamyl kinase variants to facilitate l-proline production

With the optimized genome editing method, we conducted metabolic engineering of C. glutamicum ATCC 13032 for l-proline production. l-Proline is synthesized from the tricarboxylic acid cycle (TCA cycle) intermediate 2-oxoglutarate by l-glutamate dehydrogenase (Gdh), γ-glutamyl kinase (ProB), γ-glutamyl phosphate reductase (ProA), spontaneous cyclization, and pyrroline-5-carboxylate reductase (ProC) in C. glutamicum. l-Proline biosynthesis is tightly regulated by feedback inhibition of ProB by the end-product l-proline (Fig. 2)¹⁹. Therefore, releasing the feedback inhibition of ProB by l-proline is the initial step for the development of l-proline producers⁴⁴. The ~100% single nucleotide editing efficiency of CRISPR/Cas9-assisted ssDNA recombineering makes it possible to perform codon saturation mutagenesis in C. glutamicum chromosome without introducing adjacent synonymous mutations, which do not alter the encoded protein but can influence gene expression⁴⁵. To identify the target amino acid residues for mutagenesis, 946 ProB sequences from UniProt database were aligned and analyzed (Fig. 3a and Supplementary Data 1). It has been reported that the substrate l-glutamate and the feedback inhibitor l-proline bind at overlapping sites⁴⁶. The predicted binding sites of ProB from C. glutamicum ATCC 13032 (CgProB) are D154 and N155³⁵, which are conserved in ProB enzymes from different sources. Interestingly, the adjacent region is less conserved, especially positions 148 and 150. The sequence alignment suggests that T148, G149, V150, and N151 of CgProB are all low-probability amino acid residues for these four sites (probability of 4.55, 0.95, 3.07, and 1.37%, respectively) (Fig. 3a). The 3D protein structure modelling of CgProB indicates that the less conserved region forms a flexible loop and maps away from the binding sites (Fig. 3b and Supplementary Fig. 5a). Analysis of the receptor-ligand interactions suggests that these amino acid residuals unlikely interact with the substrate l-glutamate directly (Supplementary Fig. 5b). However, previous studies on ProB of E. coli show that this flexible loop modulates l-proline inhibition⁴⁷. Therefore, we hypothesize that mutations in this region may release the feedback inhibition of CgProB by l-proline without affecting the substrate l-glutamate binding. Previous research on G149 mutagenesis provides support for our hypothesis³⁵. Therefore, codon saturation mutagenesis at A146, T147, T148, V150, and N151 were performed for CgProB with CRISPR/Cas9-assisted ssDNA recombineering.

To reduce the library size and avoid rare codons, 8–9 ssDNAs containing degenerate bases were designed and synthesized to cover all the 19 amino acid substitutions (Fig. 3c, d). Sixty transformants were randomly picked and tested for editing at each site. Site-directed mutagenesis targeting A146 and V150 produced several mutants with extracellular l-proline accumulation using glucose as a carbon source. Sequencing of proB in these mutants showed that sixteen different amino acid substitutions were obtained for A146 mutation and fourteen different amino acid substitutions were obtained for V150 mutation. All the sequenced proB variants had one to three nucleotide changes, demonstrating a 100% editing efficiency. Compared with A146 substitutions leading to low-level l-proline production (A146W, A146K, and A146R) (Fig. 3c), several V150 substitutions facilitated up to 3.6 g/L l-proline production (V150N), which was even 24.1% higher than that produced by the strain harboring previously reported G149K substitution (2.9 g/L) (Fig. 3d). The l-proline producing mutants overall showed decreased biomass compared with the wild-type strain, suggesting a redirection of metabolic flux from the production of biomass to the target metabolite. The strain expressing the proB^V150N mutant was designated as strain PRO-1.

Metabolic flux analysis for optimal l-proline biosynthesis

Next, to predict the flux-control reactions leading to the optimal l-proline biosynthesis, flux balance analysis (FBA) was performed using a genome-scale metabolic model of C. glutamicum, iCW773²⁴. FBA is a mathematical approach for analyzing the fluxes of metabolites through a metabolic network and has been widely used to predict the growth rate of an organism or the theoretical yield of producing a metabolite of interest⁴⁸. By comparing the metabolic flux distribution between maximum biomass formation and maximum l-proline production, the flux-control reactions can be predicted. The in silico simulation suggests that when the substrate glucose is used for maximum biomass formation, the glycolysis and TCA cycle are active and few metabolic fluxes are channeled to l-proline biosynthesis. When maximum l-proline biosynthesis is the objective of simulation, 2-oxoglutarate flux is drained from the TCA cycle to enter the l-proline biosynthetic pathway, resulting in a theoretical l-proline yield from glucose of 0.86 mol/mol (0.55 g/g) (Fig. 2 and Supplementary Data 2 and 3). Since the TCA cycle is highly active in C. glutamicum⁴⁹, a strong l-proline biosynthetic pathway (Gdh, ProB, ProA, and ProC) is expected to channel 2-oxoglutarate flux from the TCA cycle²³.

According to the simulation, an efficient C4 anaplerotic pathway is needed for the optimal l-proline biosynthesis. The phosphoenolpyruvate carboxylase (Ppc)-based C4 anaplerotic pathway converts phosphoenolpyruvate (PEP) and CO₂ to oxaloacetate⁵⁰. To save PEP for oxaloacetate biosynthesis, glucose needs to be partially transported into cells via the nonphosphotransferase system (non-PTS) (Fig. 2), which is inefficient in C. glutamicum⁵¹. The pyruvate carboxylase (Pyc)-based C4 anaplerotic pathway offers a promising alternative⁵⁰. First, Pyc converts pyruvate and CO₂ to oxaloacetate. As a result, sufficient PEP can be used as the phosphoryl donor for phosphotransferase system (PTS)-mediated glucose uptake, which allows complete glucose uptake via the efficient PTS without the involvement of inefficient non-PTS. Second, the simulation suggests that using the Pyc-based C4 anaplerotic pathway leads to excess ATP production from glucose (2.04 excess ATP mol/mol glucose), which can support cell growth. In the simulation, the excess ATP is consumed by meaningless cyclic reactions, e.g., ATP-consuming l-cysteine transport reactions shown in Supplementary Data 4.

Since significant amount of NADPH is needed for l-proline biosynthesis, almost half carbon flux is channeled into the oxidative pentose phosphate pathway (PPP) for NADPH regeneration at the cost of CO₂ release, which consequently lowers the l-proline yield. The nonphosphorylating NADP-dependent glyceraldehyde-3-phosphate dehydrogenase (GapN) can be used to replace the native NAD-dependent isoenzyme for NADPH regeneration with improved carbon conservation^52,53,54. However, GapN catalyzes the conversion of glyceraldehyde-3-phosphate (G3P) to glycerate-3-phosphate with an NADPH formation but no ATP formation. Therefore, the GapN-based pathway produces less ATP than the native GapA-Pgk (NAD-dependent glyceraldehyde-3-phosphate dehydrogenase and phosphoglycerate kinase) pathway that produces an NADH and an ATP (Fig. 2). By combining the GapN-based G3P oxidation pathway and the previously described Pyc-based C4 anaplerotic pathway, it is possible to provide sufficient ATP and NADPH for cell growth and l-proline biosynthesis. After adding the GapN-catalyzed reaction to the model, a simulation suggests the l-proline yield from glucose increases from 0.86 mol/mol to 0.98 mol/mol (0.63 g/g) (Fig. 2 and Supplementary Data 5). The metabolic flux through the oxidative PPP decreases by 84.9%. The G3P flux is divided into the GapN and GapA-Pgk pathways by a ratio of 86.2% to 13.8%. Based on these analyses, l-proline biosynthetic genes (gdh, proB, proA, and proC), pyc, and gapN are flux-control genes and need to be finely tuned for enhanced l-proline production.

Chromosomally transcriptional tuning of flux-control genes using tailored promoter libraries

Promoter engineering is widely used to enhance gene expression at the transcription level. A number of homologous and heterologous promoters have been developed for constitutive or inducible gene expression in C. glutamicum⁵⁵. However, the choice of strong and constitutive natural promoters is still limited¹⁷, and P_sod and P_tuf promoters are the mostly used ones for strain development^54,56. Moreover, because the 5′ region of coding sequence (CDS) significantly effects gene expression^57,58, a promoter usually shows different strengths for a reporter gene (e.g., a fluorescent protein-encoding gene) and the target gene, making it difficult to predict the promoter performance for metabolic tuning. To overcome this problem, we fused the first 180 bp of target gene with a flexible linker (GGGGS)₃ and a red fluorescent protein (RFP) gene to construct a tailored reporter system. The native promoter of the target gene was used as a template to build promoter libraries for good adaptability to its naturally controlled gene (Fig. 4a). Regarding the screening method, fluorescence-activated cell sorting (FACS) is a commonly used high-throughput method for the first round of screening⁵⁹. However, exponential reproduction of microbial cells during the cultivation of transformants may lower the proportion of variants that are overburdened by high-level protein expression⁶⁰. To maximize the efficiency of the first round of screening, we used a fluorescence imaging system to directly assay the RFP fluorescence of 1/10 transformants on agar plates (~10⁵ colonies) (Fig. 4a and Supplementary Fig. 6). Cultivation in deep-well plates was used for the second round of screening of ~10² colonies with enhanced RFP fluorescence.

Several libraries for native promoters of gdh, proB, and pyc were constructed by installing degenerate bases via chemical synthesis (Fig. 4b and Supplementary Fig. 7). The P_gdh library produced variants with 2.3- to 46.4-fold increases in the expression level of the tailored fluorescent reporter compared with the wild-type promoter. The P_pyc library covered 1.8- to 16.1-fold fluorescence increases compared with the original P_pyc (Fig. 4c and Supplementary Data 6). However, the construction of P_proB libraries failed to generate variants with stronger transcription initiation capabilities. Therefore, we tested the strongest promoter variants from the P_gdh and P_pyc libraries for the expression of the proB reporter. P_gdh-29 and P_pyc-20 led to 6.7- and 20.8-fold increases in the expression level of the proB reporter, respectively. Although the strengths of P_gdh-29 and P_pyc-20 for the proB reporter were different from those for their corresponding reporters, these two promoters can be used to enhance the expression of proB (Fig. 4c).

Next, the flux-control reactions were tuned using the screened promoter variants. All the genetic modifications were conducted in the C. glutamicum chromosome using the optimized CRISPR/Cas9-assisted genome editing method. First, based on strain PRO-1 expressing a chromosomal copy of proB^V150N, an artificial proB^V150NAC operon controlled by P_pyc-20 was inserted into the putA locus for simultaneous enhancement of l-proline biosynthesis and disruption of l-proline degradation, resulting in strain PRO-2 (Supplementary Fig. 8). putA encodes the multifunctional l-proline utilization A flavoenzyme that catalyzes the oxidation of l-proline to l-glutamate in two reaction steps⁶¹ (Fig. 2). Deletion of putA has shown a beneficial effect for l-proline production²⁴. Insertion of proB^V150NAC operon into the putA locus resulted in a 75.8% increase in l-proline production (from 3.3 g/L to 5.8 g/L) in 24-deep-well plate cultivation (Fig. 5a). Meanwhile, biomass production and glucose consumption largely decreased, leading to a 120% increase in the conversion yield (from 0.05 g/g to 0.11 g/g). Second, gdh responsible for competition with TCA cycle for 2-oxoglutarate flux was overexpressed in strain PRO-2 by in-situ replacing the original P_gdh by its variants with gradually increased strengths (P_gdh-1, P_gdh-16, P_gdh-23, P_gdh-26, and P_gdh-29, with 2.3-, 5.3-, 9.2-, 13.3-, and 46.4-fold strength increases, respectively). Strain PRO-6 in which gdh was overexpressed using P_gdh-26 showed the largest improvement in l-proline production compared with strain PRO-2 (34.5% and 18.5% improvement in titer and yield, respectively) (Fig. 5a). The use of the strongest promoter P_gdh-29 slightly decreased l-proline titer, which was possibly due to the trade-offs between the resources used for maintaining cellular metabolism and synthesizing l-proline. Third, the C4 anaplerotic pathway was enhanced by in-situ overexpressing pyc using P_pyc variants with gradually increased strengths (P_pyc-1, P_pyc-9, P_pyc-13, P_pyc-16, and P_pyc-20, with 1.8-, 6.4-, 8.9-, 12.1-, and 16.1-fold strength increases, respectively). Moderate improvement in l-proline production was observed. Strain PRO-11 overexpressing pyc with the P_pyc-16 variant produced 9.7% more l-proline with 5.8% higher yield compared with its parent strain PRO-6 (Fig. 5b). Finally, the NADPH-generating glycolysis branch was introduced to strain PRO-11 by expressing gapN from Streptococcus mutans using five P_pyc variants (P_pyc-1, P_pyc-9, P_pyc-13, P_pyc-16, and P_pyc-20). By using a tailored reporter consisting of the first 180 bp of gapN and an RFP gene, it was observed that the five P_pyc variants led to gradual expression levels of gapN (Supplementary Fig. 9). The strength of P_pyc variants for expressing gapN was different from that for expressing its native pyc gene, which was reasonable considering the effects of the 5′ region of CDS as discussed previously^57,58. Only expression of gapN with P_pyc-1 (strain PRO-13) improved l-proline production. The titer and yield of strain PRO-13 reached 9.0 g/L and 0.15 g/g, which were 13.9% and 15.9% higher than its parent strain PRO-11, respectively. Use of stronger P_pyc variants caused growth retardation, suggesting very high-level expression of gapN negatively affected cellular metabolism (Fig. 5c).

Identification of an l-proline exporter by arrayed CRISPRi screening

Chromosomally transcriptional tuning of flux-control genes improved l-proline production by nearly 3-fold. Besides metabolic pathways, transport systems are also pivotal for the hyperproduction of biochemicals²⁶. To test whether l-proline excretion is a limiting step for its production, the intracellular l-proline concentration of strain PRO-13 during batch fermentation in shake flasks was measured and found to be 300–400 mM (Supplementary Fig. 10), which is almost 10-fold higher than the wild-type nonproducing strain⁶². Therefore, strengthening l-proline excretion is required for further strain improvement. As a compatible solute essential for protection against hyperosmotic shock, uptake systems of l-proline have been well characterized in microorganisms⁶³, whereas l-proline exporter that transports the intracellularly synthesized l-proline molecules to the culture medium has not been discovered yet²⁵. By using TransportDB 2.0⁶⁴, C. glutamicum was predicted to possess nearly 400 membrane transporters (Supplementary Data 7). For a comprehensive analysis of C. glutamicum transporters and identification of l-proline exporters, we adopted CRISPRi technology for C. glutamicum and constructed an arrayed CRISPRi library consisting of 397 plasmids, each of which was designed to repress the transcription of a membrane transporter gene. The plasmid library was then transformed into an l-proline producing C. glutamicum chassis PRO-CRISPRi (C. glutamicum expressing a deregulated ProB^G149D variant²⁴) to evaluate the effect of gene repression on l-proline biosynthesis. Transformation of five plasmids failed to produce any transformants possibly because repression of these genes seriously inhibited cell growth. Because impaired amino acid excretion could significantly decrease amino acid production³⁰, the transporters whose repression brought significant decrease in l-proline production were considered as potential l-proline exporters and further characterized (Fig. 6a).

Two rounds of screening were conducted for identifying potential l-proline exporters. The first round of screening was performed by cultivating all the 392 strains with transporter-targeting CRISPRi systems and a negative control with a nontargeting CRISPRi system. Repression of twenty-one transporters were found to negatively affect l-proline production (Fig. 6b), which were subjected to a second round of screening by cultivation in 24-deep-well plates. After two rounds of screening, repression of six transporters (Cgl2622, Cgl1436, Cgl1933, Cgl2348, Cgl1360, and Cgl1937) decreased l-proline production (Fig. 6c). Cgl1933 (PtsI), Cgl1360 (PtsG), and Cgl1937 (PtsH) belong to the PTS for sugar uptake⁶⁵, repression of which would hinder glucose uptake and consequently decrease l-proline production. Therefore, these three genes were not further investigated here. cgl2622, cgl1436, and cgl2348 were individually deleted in strain PRO-CRISPRi by CRISPR/Cas9 to verify their functions. While CRISPRi of cgl1436 slightly decreased l-proline production, its gene deletion showed no significant affect. Interestingly, the strain with repression of cgl2348 only produced ~20% l-proline compared with the control without CRISPRi. However, deletion of cgl2348 did not decrease l-proline production significantly. Cgl2348 was annotated as a DNA uptake protein or related DNA-binding protein. According to a previous transcriptomic analysis, cgl2348 belongs to an operon containing proB and proA involved in l-proline biosynthesis⁶⁶. CRISPRi of cgl2348 may also affect the expression of its neighboring genes including proB and proA, leading to decreased l-proline production. Partial deletion of the CDS of cgl2348 should not affect the expression of its neighboring genes, which could explain why CRISPRi and deletion of cgl2348 showed different influences on l-proline production. Only deletion of cgl2622 significantly decreased l-proline production, making Cgl2622 a candidate for l-proline exporter (Fig. 6d).

To verify the function of Cgl2622 as an l-proline exporter, two experiments were conducted. First, cgl2622 was deleted, complemented, and overexpressed in the l-proline producing strain PRO-CRISPRi to test the effects on intracellular and extracellular accumulation of l-proline. The resultant strains were cultivated in shake flasks to obtain enough cells for the measurement. In strain PRO-CRISPRi, the intracellular l-proline accumulated to very high levels of 150–300 mM (Fig. 6e). Deletion of cgl2622 further increased the intracellular l-proline concentration especially in the early phase of fermentation and largely decreased extracellular accumulation of l-proline. By complementing or overexpressing cgl2622 in a plasmid with IPTG-inducible promoter P_trc, intracellular l-proline concentration was maintained at relatively low levels between 50 mM to 100 mM, while extracellular l-proline production was largely increased (Fig. 6e). The similar results for complementation and overexpression of cgl2622 are possibly because of the high-level expression of cgl2622 using multi-copy plasmid and strong IPTG-inducible promoter. The results suggest that Cgl2622 plays an important role in maintaining the intracellular l-proline homeostasis and transporting l-proline outside cells.

Second, peptide uptake and amino acid export assay was conducted. cgl2622 was deleted and complemented in C. glutamicum wild-type strain. Upon the addition of Thr-Pro peptide, the cgl2622-deleted mutant showed higher intracellular l-proline level and dramatically decreased l-proline export rate (Supplementary Fig. 11). Conversely, complementation of cgl2622 in a plasmid with IPTG-inducible promoter P_trc largely decreased intracellular l-proline concentration but accelerated l-proline export (Supplementary Fig. 11). These results further verify the function of Cgl2622 as an l-proline exporter.

Transport engineering to enhance l-proline production

To overcome the limitation of l-proline excretion in engineered producing strains, overexpression of Cgl2622 is required. To quantify the expression level of Cgl2622, we used the developed strategy of building a tailored fluorescent reporter by fusing the first 180 bp of cgl2622 with the flexible linker (GGGGS)₃ and fluorescent protein-encoding gene. Both RFP and GFP were tested, whereas the fusion protein failed to produce fluorescence, which was possibly due to the membrane topology of Cgl2622. Therefore, we directly added a second copy of cgl2622 controlled by the derepressed P_trc promoter in the chromosome of strain PRO-13 for overexpression. The resultant strain PRO-18 produced 15.7 g/L l-proline with 0.25 g/g yield, which are 177.4% and 170.5% of the production level of its parent strain PRO-13 (Fig. 7a). Overexpression of Cgl2622 also led to ~100% decrease in intracellular l-proline concentration (Supplementary Fig. 12), which was considered as a major cause of improved l-proline production.

Fed-batch fermentation in a 5 L bioreactor was then conducted to evaluate the l-proline production by strain PRO-18. After 49 h cultivation, 87.2 g/L l-proline was produced with a productivity of 1.78 g/L/h and a yield of 0.19 g/g (Fig. 7b). The biomass reached a OD_600nm value over 180. It seemed that too much carbon flux was directed to the production of biomass rather than the target molecule l-proline. Since C. glutamicum is a biotin-auxotrophic bacterium⁶⁷, biotin limitation can be used to balance the biosynthesis of biomass and target molecules. Therefore, a biotin-limited fed-batch fermentation was conducted. Although biomass formation was reduced, l-proline production was only moderately improved to 97.8 g/L with a productivity of 2.00 g/L/h and a yield of 0.23 g/g. Product analysis showed the high-level accumulation of extracellular l-glutamate up to 77.5 g/L (Fig. 7c).

Improving l-proline production by eliminating biotin limitation-induced l-glutamate excretion

l-Glutamate production by C. glutamicum can be induced by treatments such as biotin limitation, Tween 40 addition, and penicillin addition. Biotin limitation decreases the activity of 2-oxoglutarate dehydrogenase complex (ODHC) that converts 2-oxoglutarate to succinyl-CoA, causing accumulation of intracellular 2-oxoglutarate. Meanwhile, the cell membrane tension is altered and mechanosensitive ion channels responsible for l-glutamate excretion are activated^30,67. To reduce l-glutamate production, the major l-glutamate exporter MscCG encoded by cgl1270 was deleted in strain PRO-18 by CRISPR/Cas9. The resultant strain PRO-19 was then used for l-proline production in biotin-limited fed-batch fermentation. After 49 h cultivation, 142.4 g/L l-proline was produced with a productivity of 2.90 g/L/h and a yield of 0.31 g/g (Fig. 7d). The titer, productivity, and yield of strain PRO-19 were improved by 44.9, 45.0, and 34.8%, respectively, compared with strain PRO-18 under the same cultivation condition. The byproduct l-glutamate was largely decreased to 12.3 g/L, which was only 15.6% of that produced by strain PRO-18. The amino acid profile of the broth from the end of fermentation was further determined using an amino acid analyzer. Besides l-glutamate, 7.9 g/L of l-alanine and 3.2 g/L of l-valine were also detected as byproducts (Supplementary Fig. 13). Because the l-glutamate exporter MscCG encoding gene cgl1270 was deleted in strain PRO-19, l-glutamate may accumulate inside the cells. l-Glutamate is an important amine donor for the biosynthesis of many amino acids⁶⁸. Its accumulation may enhance the transamination and biosynthesis of other amino acids, which may explain the formation of l-alanine and l-valine as byproducts. Therefore, increasing the expression of genes involved in l-proline biosynthesis (such as gdh and proB) and deleting related aminotransferase encoding genes may further improve l-proline production and decrease byproduct formation.