Methanol-Essential Growth of Corynebacterium glutamicum: Adaptive Laboratory Evolution Overcomes Limitation due to Methanethiol Assimilation Pathway

Methanol is a sustainable substrate for biotechnology. In addition to natural methylotrophs, metabolic engineering has gained attention for transfer of methylotrophy. Here, we engineered Corynebacterium glutamicum for methanol-dependent growth with a sugar co-substrate. Heterologous expression of genes for methanol dehydrogenase from Bacillus methanolicus and of ribulose monophosphate pathway genes for hexulose phosphate synthase and isomerase from Bacillus subtilis enabled methanol-dependent growth of mutants carrying one of two independent metabolic cut-offs, i.e., either lacking ribose-5-phosphate isomerase or ribulose-5-phosphate epimerase. Whole genome sequencing of strains selected by adaptive laboratory evolution (ALE) for faster methanol-dependent growth was performed. Subsequently, three mutations were identified that caused improved methanol-dependent growth by (1) increased plasmid copy numbers, (2) enhanced riboflavin supply and (3) reduced formation of the methionine-analogue O-methyl-homoserine in the methanethiol pathway. Our findings serve as a foundation for the engineering of C. glutamicum to unleash the full potential of methanol as a carbon source in biotechnological processes.

In order to test if a functional RuMP cycle for formaldehyde assimilation can circumvent feeding ribose to the rpi deletion mutant ( Figure 1B), strain MDS1(pVWEx1-mdh-hxlAB) for expression of the genes encoding methanol dehydrogenase from B. methanolicus and 3-hexulose 6-phosphate synthase HxlA and hexulose 6-phosphate isomerase HxlB from B. subtilis was constructed and evaluated for methanol-dependent growth ( Figure 2C). No growth of this strain was observed with xylose as sole carbon source, however, upon addition of methanol the strain grew with a growth rate 0.03 h −1 from an OD 600 of 0.52 ± 0.07 to 1.39 ± 0.37. As an independent test, growth with gluconate as sole carbon source with/without addition of methanol was tested since ribulose 5-phosphate is generated during gluconate catabolism. Indeed, strain MDS1(pVWEx1-mdh-hxlAB) grew with a growth rate of 0.04 h −1 from an initial OD600 of 0.50 ± 0.07 to 1.63 ± 0.10. Although no ALE was conducted with strain MDS1, the specific growth rates with both co-substrates were on par with, but not faster than previously reported growth rates for methanol-dependent growth of evolved C. glutamicum strain MX-11 with xylose (0.03 h −1 ) [25].
To exclude that rpi deletion mutants can only be selected for if compensatory mutations occurred elsewhere in the genome, we re-sequenced the genome of strains MDS0(pEKEx3-xylAB) (pVWEx1-mdh-hxlAB) and MDS1(pVWEx1-mdh-hxlAB) that only differ by the presence/absence of rpi. For strain MDS0(pEKEx3-xylAB)(pVWEx1-mdh-hxlAB), 79.25%, 14.23% and 7.17% of the processed reads could be mapped to the genome, plasmid pEKEx3-xylAB and plasmid pVWEx1-mdh-hxlAB, respectively. For the rpi deletion mutant MDS1(pVWEx1-mdh-hxlAB), 79.81%, 13.43% and 6.33% of the processed reads were mapped to the genome, plasmid pEKEx3-xylAB and plasmid pVWEx1-mdh-hxlAB, respectively. As expected, reads mapping to the deleted genes ald (cg3096) and fadH (cg0387) were not found for both strains. As compared to the reference genome sequence, in both strains psp1 (cg2069) coding for a putative secreted protein of the prophage CGP3 was only partly covered and seven non-silent single-nucleotide polymorphisms (SNPs) were found ( Table 1). Three occurred in the remainders of the deleted genes ald and fadH which were left to ensure in frame-deletions and prevent polar effects. The gene functions of loci cg0822 and cg1245 are yet unknown and additional SNPs were found in wzz and hrtA (cg0414, cg2204) encoding for proteins involved in cell surface polysaccharide biosynthesis, chain length determination and an ABC-type transport system and ATPase component [28]. As expected, none of the SNPs matched the ones generated after ALE of a C. glutamicum strain with a similar background [25]. Importantly, while sequenced reads were mapped to the rpi locus (cg2658) for strain MDS0(pEKEx3-xylAB)(pVWEx1-mdh-hxlAB), with a mean feature coverage of 12, no reads were mapped to this gene in strain MDS1(pVWEx1-mdh-hxlAB). Thus, the presence/absence of rpi was the only difference between the strains. Table 1. Single-nucleotide polymorphisms (SNPs) determined by whole-genome sequencing. Non-silent SNPs and mutations found in the strains ∆ald ∆fadH and ∆ald ∆fadH ∆rpi, as compared to the wildtype C. glutamicum ATCC13032, or strains MDS2T8 and MDS2T14 in comparison to their parental strain MDS2, including the corresponding loci, gene names, products and the resulting amino acid substitutions.

Adaptive Laboratory Evolution for Accelerated Methanol-Dependent Complementation of a Ribulose 5-Phosphate Epimerase Mutant
To reduce the number of plasmids to one, we first constructed a new strain containing only one plasmid. Instead of rpi, the ribulose 5-phosphate epimerase gene rpe was deleted in MDS0 to isolate ribose-5-phosphate dependent anabolism from the rest of cellular metabolism. As a result, only one plasmid (i.e., pEKEx3-mdh-hxlAB) was required since in this metabolic setting C. glutamicum naturally catabolizes ribose. Methanol-dependency of the new C. glutamicum strain MDS2(pEKEx3-mdh-hxlAB) was examined in minimal medium with 20 mM ribose as co-substrate. No growth with ribose as sole carbon source was observed for 72 h. By contrast, with methanol plus ribose, strain MDS2(pEKEx3-mdh-hxlAB) formed biomass (∆OD of 1.32 ± 0.03) with specific growth rate of 0.03 ± 0.01 h −1 ( Figure 4A). Thus, methanol dependent growth of two-vector strain MDS1(pVWEx1-mdh-hxlAB) with methanol plus xylose and one-vector strain MDS2(pEKEx3-mdh-hxlAB) with methanol plus ribose were comparable. C. glutamicum strain MDS2(pEKEx3-mdh-hxlAB) was subjected to ALE by iterative transfers of cells grown in minimal medium with ribose and methanol. After the first eight transfers, strain MDS2T8 was conserved for whole-genome sequencing and used to determine to what extent 13 C-methanol is assimilated via the RuMP pathway.
Furthermore, transfers of C. glutamicum strain MDS2T8 were performed from minimal medium to minimal medium with 0.5 g/L of yeast extract. After six further transfers, C. glutamicum strain MDS2T14 was selected as it reached a final ∆OD of 3.95 ± 0.37 in minimal medium with ribose and methanol with a specific growth rate of 0.10 ± 0.01 h −1 , which is comparable to the wild type (WT) grown with 20 mM ribose (∆OD of 3.90 ± 0.29) ( Figure 4A). Therefore, MDS2T14 was 3-times as fast as strain C. glutamicum MX-11 which was cultivated with xylose as co-substrate [25] and exceeded the growth rate of evolved methanol-essential growth with E. coli co-utilizing gluconate (0.081 ± 0.002 h −1 ) [24]. Notably, strain MDS2T14 did not require yeast extract for methanol-dependent growth. Taken together, ALE allowed the selection of a mutant strain of MDS2 named MDS2T14 that showed full methanol-dependent complementation of the rpe deletion.

Genome Sequencing of ALE Strains Revealed Candidate Mutations that May Accelerate Methanol-Dependent Biomass Formation
The genome sequences of ALE strains MDS2T8 and MDS2T14 were determined and compared to that of the parental strain MDS2 to identify candidate mutations that may be causal for accelerated methanol-dependent growth. As compared to the parental C. glutamicum strain MDS2, strain MDS2T8 possessed three mutations: non-silent SNPs in the genes metK (S-adenosylmethionine synthetase) and res (site-specific recombinase) resulting in amino acid exchanges S288N and R91H, respectively, and a 15 bp deletion disrupting the coding sequence (CDS) of an uncharacterized gene (cg3104) which is annotated as ATPase involved in DNA repair. The latter shares structural motifs typically found in proteins with DNA exo-and endonuclease activity, such as the well-characterized SbcCD protein in E. coli [29] that was revealed to be capable of cleaving secondary hairpin structures during plasmid replication [30]. The unique mutation found in strain MDS2T14 was a 1.4 kb transposon insertion near the deleted rpe gene which shares a leaderless promoter with the riboflavin synthesis genes ribG and ribC, encoding putative bifunctional riboflavin specific deaminase/reductase and putative riboflavin synthase, respectively ( Figure 5A) [31].

Identification of Mutations Causal for Methanol-Essential Growth
First, we analyzed the effect that deletion of rpe has on riboflavin biosynthesis. C. glutamicum MDS2(pEKEx3-mdh-hxlAB) and the evolved strains MDS2T8 and MDS2T14 were cultivated on glucose as sole carbon source with or without riboflavin supplementation to find out whether the transposon integration in MDS2T14 influenced growth without riboflavin. Growth of strains MDS2 and MDS2T8 on glucose minimal medium could be enhanced by riboflavin addition ( Figure 5B) as it was previously reported in studies investigating riboflavin transport in a riboflavin deficient C. glutamicum strain [32]. By contrast, without added riboflavin strain MDS2T14 grew as fast as or faster than MDS2 and MDS2T8 with added riboflavin ( Figure 5B). Thus, the growth advantage of strain MDS2T14 is riboflavin-dependent and likely due to the transposon integration which provided an additional potential -10 promoter region (TATTT) upstream of ribG and an alternative ribosome binding motif (AAGGG) 7 bp upstream of the ribG start-codon [33,34]. Deletion of cg3104 in parental strain MDS2 did not increase biomass formation (∆OD: 1.74 ± 0.08) as compared to MDS2 (∆OD: 2.08 ± 0.41) ( Figure 4A). Based on its supposed function as putative ATPase involved in DNA repair, the plasmid copy number of plasmid pEKEx3-mdh-hxlAB was determined by qPCR ( Figure 4C). Strains MDS2T14 and MDS2 ∆cg3104 showed approximately 40-60% increased relative plasmid copy number (1.85 ± 0.21 and 2.17 ± 0.05, respectively) as compared to that of the parental strain MDS2 (1.33 ± 0.02). Because of the elevated copy numbers, higher enzyme activities of the heterologously expressed RuMP genes were expected. A coupled assay of HxlA and HxlB revealed 60% higher enzyme activities (p < 0.05) of MDS2T14 (268 ± 21 mU/mg) and MDS2 ∆cg3104 (268 ± 26 mU/mg) compared to MDS2 (169 ± 22 mU/mg) ( Figure 4B), which is beneficial for formaldehyde fixation. Improved channeling of formaldehyde into central metabolites has been shown to shift the equilibrium of Mdh towards the methanol-oxidation reaction in E. coli, for an overall increased utilization of methanol [35]. Efficient methanol-oxidation and formaldehyde channeling has been argued to be one of the major bottlenecks for synthetic methylotrophy [24,36], which is further reinforced by our finding that increased HxlAB activity is an essential factor for improved growth of strain MDS2T14.
Another key factor for improved methanol-dependent growth of MDS2T14 is an amino acid exchange from serine to asparagine found at position 288 in MetK (Table 1) which is next to the methionine binding K289. MetK catalyzes the formation of S-adenosylmethionine (SAM) from methionine and ATP ( Figure 5C) [37]. Both E. coli and C. glutamicum only harbor one copy of the essential metK gene [38]. When combining deletion of cg3104 with introduction of MetK amino acid exchange S288N, growth with ribose and methanol was improved as strain C. glutamicum MDS2 ∆cg3104 metK_S288N grew to a ∆OD of 4.20 ± 0.47, reaching comparable levels of biomass formation as the evolved strain MDS2T14 (3.95 ± 0.37) and C. glutamicum WT with 20 mM ribose (3.90 ± 0.29) ( Figure 4A). Introduction of the MetK amino acid exchange S288N alone had no positive effect on biomass formation (∆OD: 2.20 ± 0.03). However, reverting the amino acid exchange in MetK in the evolved strain MDS2T14 from S288N to N288S, i.e., to the WT sequence, reduced growth on ribose and methanol considerably (∆OD of 1.11 ± 0.02) ( Figure 4A). Taken together, the MetK amino acid exchange S288N combined with deletion of cg3104 and supplementation with riboflavin was sufficient to explain improved growth of the evolved strain MDS2T14 in minimal medium with ribose and methanol.
The antimicrobial methionine analogue O-methyl-homoserine [39] was shown to be a naturally occurring side-product of O-acetyl-homoserine sulfhydrolase (MetY) of Corynebacterium acetophilum in the presence of homoserine and methanol since the latter is an analogue of methanethiol, the precursor of methionine ( Figure 5C) [40]. The accumulation of growth-inhibiting O-methyl-homoserine in presence of methanol has been argued to be connected to methanol tolerance before, when MetY was found to be mutated leading to amino acid substitution A165T in a methanol tolerant C. glutamicum strain [41]. Furthermore, metY was among the mutated genes of strain C. glutamicum MX-11 that exhibited improved methanol-dependent growth with xylose [25]. During the writing process of this manuscript, a follow-up study from the same group appeared, further elucidating this mutation (G419D) [42]. It was shown to increase the tolerance for methanol, which was confirmed by knock-down and knock-out experiments. Here, we observed elevated methanol-dependent growth with ribose and riboflavin after metY deletion in strain MDS2 ∆cg3104 ( Figure 5D), which was similar to strain MDS2 ∆cg3104 metK_S288N, indicating that both encoded enzymes, MetK and MetY, presumably act on each other when methanol is present and significantly improve methanol-dependent biomass formation of C. glutamicum.

ALE Experiments
ALE of C. glutamicum strain MDS2 (methanol-dependent strain) was carried out in two phases: In phase 1 [MDS2 -MDS2T8] triplicates were cultivated in CGXII minimal medium with 20 mM ribose and 500 mM methanol with intermediate inoculations of LB overnight cultures and vice versa from the highest grown culture. In phase 2 [MDS2T8-MDS2T14], transfers to LB were omitted and minimal medium was supplemented with 0.5 g/L yeast extract, 20 mM ribose and 500 mM methanol.

Molecular Biology Methods
Genomic DNA of C. glutamicum was isolated as described previously [51]. Classical methods including plasmid isolation, molecular cloning and transformation of E. coli by heat shock and of C. glutamicum by electroporation were performed as described before [45]. DNA sequences were amplified with the ALLin HiFi DNA Polymerase (HighQu, Kraichtal, Germany) using plasmid or genomic DNA as template. The oligonucleotides used in this study are listed in Table S1. The gene rpi was amplified from C. glutamicum and assembled into PstI-and BamHI-digested pVWEx1 by Gibson Assembly using the respective primers. Likewise, the genes mdh, hxlA and hxlB were amplified as a single operon from the plasmid pEKEx3-mdh-hxlAB [20] and cloned into pVWEx1 and pEC-XT99A linearized with BamHI. The constructed plasmids and empty vectors were transferred into C. glutamicum by transformation. For deletion, plasmid pK19mobsacB [44] digested with EcoRI and BamHI was assembled with amplified DNA fragments flanking the genes rpe and cg3104 by Gibson Assembly and transferred into E. coli S17-1 to follow a protocol for gene deletion routinely applied [45].

Coupled in vitro Activity of HxlA and HxlB
C. glutamicum cells were grown overnight in BHI medium containing 1 mM IPTG and appropriate antibiotics, harvested by centrifugation (5 min, 4000 rpm), washed in 50 mM phosphate buffer and subsequently disrupted by sonication with an amplitude of 50% and a duty circle of 0.5 for 9 min. Afterwards, total protein concentrations were determined by the Bradford method with bovine serum albumin as standard. Activities of HxlA and HxlB in crude extracts were determined in a coupled assay, derived from a previously described method [20]. A total volume of 1 mL was used for measurements, containing 25 mM phosphate buffer pH 7.4, 5 mM MgCl 2 , 5 mM ribose-5-phosphate, 0.5 mM NAD + , 2 U phosphoriboisomerase from spinach (Sigma), 2 U phosphoglucoisomerase from yeast (Sigma), 2 U Glucose-6-P dehydrogenase from Leuconostoc mesenteroides (Sigma) and 50 µL crude extract. An addition of 5 mM formaldehyde started the reaction and NADH formation was continuously measured at 340 nm and 30 • C for 6 min.

Whole-Genome Sequencing
Whole-genome sequencing was performed with isolated genomic DNA from C. glutamicum strains (PRJNA603493). DNA library preparation, trimming and mapping of the reads and visualization was performed as described previously [52]. For SNP detection in all CDS's of C. glutamicum, the built-in tool from ReadXplorer was used. Minimal scores for base quality, average base quality and average mapping quality were set to 20 and reads with a coverage between 5 and 10 were kept if the corresponding frequency was 100%. If the coverage was above 10, the minimum frequency was 90% instead. Additionally, genomic DNA of C. glutamicum ∆ald ∆fadH ∆rpe strains MDS2, MDS2T8 and MDS2T14 was sequenced using Nanopore MinION sequencing technology (Oxford Nanopore Technologies Oxford, UK) as described previously [53]. After assembly, 11 contigs for MDS2, 6 contigs for MDS2T8 and 2 contigs for MDS2T14 were revealed. Assembled genomes of MDS2T8 and MDS2T14 were aligned with the MDS2 genome sequence using SnapGene software v.4.3 (GSL Biotech, Chicago, USA) to visualize SNPs and insertions or deletions. Variant calling was performed manually to identify SNPs in the MDS2T8 and MDS2T14 genomes, after alignment to the MDS2 parental strain genome reference.

Quantitative PCR
Quantitative PCR (qPCR) was performed in 96-well plates, sealed with transparent adhesive cover, in a CFX Connect TM Real-Time PCR Detection System (Biorad, Hercules, USA). A total volume of 20 µL contained 10 µL SensiFAST TM SYBR No-ROX Kit from Bioline (Heidelberg, Germany), 100 nmol/L forward/reverse primer and 4 µL of diluted template DNA. Genomic template DNA, also containing plasmid DNA, was isolated as described before. Fragments of approximately 250 bp were amplified, targeting gntK encoding gluconate kinase on the chromosome and oriV Cg on the pEKEx3 plasmid. Serial dilutions of DNA from 100 to 0.1 ng/µL were used to construct relative standard curves for chromosomal and plasmid targets. Plasmid and chromosome-specific amplicons were additionally detected in separate reactions and technical triplicates at 10 ng/µL to calculate plasmid copy numbers.

Quantification of 13 C-Enrichment
Cultures of C. glutamicum MDS1 and MDS2T8 (pEC-XT99A-lysC fbr -ldcC), grown overnight in BHI medium, containing 1 mM IPTG and antibiotics were used to inoculate modified M9 medium with 20 mM gluconate and 500 mM 13 C-methanol to a starting OD 600 of 0.5. Modified M9 medium contained 7.5 g/L Na 2 HPO 4 , 3 g/L KH 2 PO 4 , 0.5 g/L NaCl, 0.25 g/L MgSO 4 7 H 2 O, 0.003 g/L CaCl 2 , 0.5 g/L NH 4 Cl and the trace elements as described for CGXII medium [45]. Samples of 1 mL were taken after 24, 48 and 72 h of cultivation and subsequently centrifuged for 10 min at 10,000 rpm. Supernatants were frozen at −20 • C for analysis of isotopologues as previously described [20].

Conclusions
By engineering two alternative metabolic cut-offs in the central carbon metabolism of C. glutamicum to reinforce methanol utilization through RuMP pathway reactions, we achieved methanol-dependent biomass formation on three different co-substrates (ribose, xylose and gluconate). Incorporation of methanol-carbon and functionality of the introduced RuMP pathway was confirmed by following the fate of 13 C-methanol in cadaverine. Growth rates and methanol-dependent biomass formation were improved through ALE. Analysis of mutations identified by genome sequencing of evolved strains revealed three mutations causing the observed metabolic changes concerning improved methanol utilization through an increased plasmid copy number, resistance against methionine-analogue toxicity due to modified metK sequence, which was also achieved by metY deletion and enhanced riboflavin supply because of alternative promoter and RBS motifs. The findings reported here will serve as a basis for future engineering approaches towards synthetic methylotrophy in C. glutamicum and lead to a better understanding of necessary requirements for synthetic methylotrophy in general to unleash the full potential of methanol as biotechnological carbon source for the sustainable production of valuable chemical compounds in the future.