All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless noted otherwise. ¹³C-methanol (99% ¹³C) was purchased from Cambridge Isotope Laboratories (Tewksbury, MA). E. coli NEB5α, Q5 DNA polymerase and NEBuilder HiFi DNA assembly master mix were purchased from NEB (Ipswich, MA). Restriction endonucleases were purchased from Thermo Fisher Scientific (Waltham, MA).

All strains, plasmids, and primers used in this study are listed in Supplementary Tables S1–S3, respectively. E. coli NEB5a was used for plasmid construction and propagation. E. coli BW25113 DfrmA was obtained from the Keio collection and used for growth characterization (Baba et al., 2006). Deletion of ihfA was performed as previously described via the lambda red recombineering system (Datsenko and Wanner, 2000; Bennett et al., 2020a). Methanol assimilation genes were cloned into pETM6 (Xu et al., 2012) for episomal expression as previously described to produce the pUD9 plasmid (Bennett et al., 2018a). Overexpression of crp was achieved as previously described (Bennett et al., 2020a).

E. coli strains were routinely cultured in LB medium supplemented with the appropriate antibiotics (100 μg/mL ampicillin, 25 μg/mL kanamycin) unless otherwise noted. Growth characterization of E. coli strains for methanol or formaldehyde tolerance was performed in 250 mL baffled flasks containing 30 mL LB medium supplemented with methanol or formaldehyde at the specified concentrations at 37°C and 250 RPM. Cell growth rate was determined every hour as follows: ln(C/C₀)/(t − t₀), where C and C₀ represent the biomass concentration at the current (t) and prior (t₀) times (e.g., 3 and 2 h). The highest value was selected as the maximum growth rate. For methylotrophic growth and ¹³C-labeling assays, an overnight culture of the respective E. coli strain in LB medium was used to inoculate fresh M9 minimal medium containing 1 g/L yeast extract with or without 60 mM ¹³C-methanol to an OD₆₀₀ of approximately 0.05. Samples were collected at 48h for labeling analysis.

Chemical mutagenesis was performed as described previously (Sandoval et al., 2015). Briefly, 70 μL of an overnight culture of E. coli ΔfrmA pUD9 was used to inoculate 7 mL of fresh LB medium. Cells were grown aerobically at 37°C until an OD₆₀₀ of 1. Cells were then harvested via centrifugation (4,000 g, 10 min), washed once with 10 mL of PT buffer (0.1 g/L peptone, 8.5 g/L sodium chloride, 1 g/L sodium thioglycolate), harvested again via centrifugation (4,000 g, 10 min) and finally resuspended in 7 mL of fresh LB medium. 200 μL of 2.5 g/L N-methyl-N’-nitro-N-nitrosoguanidine (NTG) was then added, followed by a 20 min incubation at 37°C. Mutagenized cells were then harvested via centrifugation (4,000 g, 10 min) and washed thrice with 10 mL of PT buffer, followed by resuspension in 10 mL of fresh LB medium and outgrowth overnight at 37°C. The lethality of this chemical mutagenesis was determined to be ca. 99% as CFUs/mL directly prior to NTG treatment were ca. 7.2 × 10⁸ and ca. 6.4 × 10⁴ directly following NTG treatment.

After overnight recovery, mutagenized cells were subjected to directed evolution via passaging in fresh LB medium supplemented with methanol. After the initial recovery, cells were used to inoculate fresh LB medium supplemented with 1 M methanol. Upon growth of this culture, cells were used to inoculate fresh LB medium supplemented with 1.25 M methanol. This procedure was continued for 1.5, 1.75, and 2 M methanol. Upon growth in fresh LB medium supplemented 2 M methanol, a frozen stock in 20% glycerol was made. This frozen stock was streaked on a fresh LB agar plate to isolate individual clones. Six of these clones were analyzed for improved methanol tolerance over the non-evolved parent strain in fresh LB medium supplemented with 2 M methanol. The clone exhibiting the most improved methanol tolerance was used for further analysis.

Three other E. coli strains (ΔfrmA, ΔfrmAΔihfA + pUD9 and ΔfrmA + pUD9 + pCrp) were also subjected to ALE without NTG mutagenesis. Serial batch passaging in LB medium supplemented with increasing methanol concentrations was performed in a similar manner until methanol-tolerant clones could be isolated. Approximately 10 passages were required to achieve improved methanol tolerance when chemical mutagenesis was not used.

Mdh, Hps and Phi in vivo assays were performed as described (Whitaker et al., 2017). Briefly, E. coli ΔfrmA strains expressing B. stearothermophilus Mdh and B. methanolicus Hps and Phi were grown from a colony in LB for 6 h at 37°C with shaking (225 rpm). Cells were then washed twice in M9 minimal medium and adjusted to an OD₆₀₀ of 1.0 in M9 minimal medium. Methanol was added to a final concentration of 1 M while formaldehyde was added to a final concentration of 1 mM. At the indicated time points, samples were collected and 400 μL of culture supernatant was mixed with 800 μL of Nash reagent to assay for formaldehyde concentration (Nash, 1953).

Biomass concentration was determined as previously described (Whitaker et al., 2017; Bennett et al., 2018a). Briefly, OD₆₀₀ was measured on a Beckman-Coulter DU730 spectrophotometer. Methanol boost was calculated as the percentage improvement of biomass yield of a culture in the presence of methanol as compared to the control without methanol (Whitaker et al., 2017; Bennett et al.,2020a,b). Methanol was measured via high performance liquid chromatography (HPLC) (Whitaker et al., 2017). Extraction of metabolites and proteinogenic amino acids was performed as previously described and analyzed for ¹³C-labeling using gas chromatography-mass spectrometry (Whitaker et al., 2017; Bennett et al., 2018a, 2020a, 2020b,c; Long and Antoniewicz, 2019). ¹³C-labeling was determined from the measured mass isotopomer data (Whitaker et al., 2017; Long and Antoniewicz, 2019). Statistics were calculated using a two-tailed unpaired t-test with a 95% confidence interval.

Whole genome sequencing of ΔfrmA parental and evolved strains, ΔihfA parental and evolved strains, and KB201, a ΔfrmA strain that was subjected to directed evolution without the pUD9 plasmid, was performed as previously described (Bennett et al., 2020c). Briefly, genomic DNA was extracted using a Qiagen DNeasy Blood and tissue kit per manufacturer’s protocol (Germantown, MD). Genomic DNA was then sequenced on an RSII sequencer system (Pacific Biosciences, Menlo Park, CA) using single molecule, real time (SMRT) sequencing (University of Delaware DNA Sequencing and Genotyping Center), with average read length of 10 kb generated. Sequencing analysis was performed with the SMRT Link software via the resequencing application (Pacific Biosciences). E. coli BW25113 (GenBank CP009273.1) was used as the reference genome. Mutations unique to each sequenced strain in comparison to the respective parental strain were chosen.

As discussed, methanol and formaldehyde are cytotoxic to cells. In LB medium, the growth rate of a synthetic E. coli methylotroph (E. coli BW25113 ΔfrmA + pUD9) is severely inhibited above methanol concentrations of 1 M (Figure 2 and Supplementary Figure S1A). Specifically, the growth rates in the presence of 0, 1, 2, and 3 M methanol are 1.2 ± 0.02, 1.0 ± 0.01, 0.47 ± 0.02, and 0.07 ± 0.00 h^–1, respectively. Since several studies have reported using high methanol concentrations, specifically ≥ 1 M, for natural and synthetic methylotrophs, specifically B. methanolicus (Bozdag et al., 2015) and E. coli (Muller et al., 2015), improved methanol tolerance at these high concentrations would be beneficial.

To improve the methanol tolerance of this synthetic E. coli methylotroph (E. coli BW25113 ΔfrmA + pUD9), chemical mutagenesis and ALE were performed. Briefly, cells were first mutagenized with N-methyl-N’-nitro-N-nitrosoguanidine (NTG), which mutates DNA by alkylating guanine and thymine, resulting in transition mutations between GC and AT, recovered overnight in LB medium and then subjected to several rounds of passaging in LB medium supplemented with increasing methanol concentrations (Supplementary Figure S1B). After outgrowth in the presence of 2 M methanol, a frozen stock was prepared and subsequently streaked onto an LB agar plate to isolate individual clones. Six clones were examined for improved growth in LB medium supplemented with 2 M methanol (Supplementary Figure S2). One of these clones, “Evolved 3,” simply referred to as “evolved,” exhibited the largest improvement in growth and was selected for further analysis. Indeed, this evolved clone exhibited improved methanol tolerance at high methanol concentrations, i.e., 2–3 M (Figures 2C,D and Supplementary Figure S1A). Specifically, the growth rates in the presence of 0, 1, 2, and 3 M methanol were 1.2 ± 0.02, 1.1 ± 0.00, 0.65 ± 0.01, and 0.11 ± 0.01 h^–1, respectively (Supplementary Table S4). Thus, the evolved clone exhibited growth rate improvements of 10, 38, and 57% over the parental strain in 1, 2, and 3 M methanol, respectively. Additionally, the evolved clone achieved higher final biomass titers over the parent strain in LB medium supplemented with 2 and 3 M methanol (Figures 2C,D). Taken together, these results demonstrate the usefulness of chemical mutagenesis and ALE for improving tolerance to toxic substrates.

To investigate whether the improved methanol tolerance of the evolved clone resulted from improved methanol and/or formaldehyde tolerance, the parental strain and evolved clone were cured of the pUD9 plasmid via serial passaging in the absence of the appropriate antibiotic so that formaldehyde tolerance of both strains could be determined. Plasmid curing was essential for this since both the Mdh and RuMP pathway enzymes readily consume formaldehyde, either via reducing it back to methanol or assimilating it into central carbon metabolism, respectively. Both pathways result in inaccurate growth rate measurements since the formaldehyde concentration is continually decreasing to non-toxic levels over time, thus yielding a dynamic growth rate (Supplementary Figure S3). Furthermore, the evolved, plasmid-containing clone was not observed to overcome formaldehyde toxicity more quickly than the plasmid-containing parental strain (Supplementary Figure S3), suggesting that the observed methanol tolerance of the evolved clone results from improved tolerance to methanol and not formaldehyde.

The resulting growth rates of both plasmid-cured strains in LB medium supplemented with varying levels of formaldehyde were similar (Supplementary Figures S4, S5), again suggesting that the observed methanol tolerance of the evolved clone results from improved tolerance to methanol and not formaldehyde, at least at the concentrations tested in this study. Furthermore, as discussed, formaldehyde exerts greater cytotoxicity on cells than does methanol, as indicated by the severe reduction in growth rate at low (i.e., mM) formaldehyde concentrations compared to high (i.e., M) methanol concentrations. Specifically, the growth rates of the plasmid-cured parental strain in the presence of 0, 0.25, 0.5, 1, and 1.5 mM formaldehyde were 1.4 ± 0.02, 1.0 ± 0.03, 0.75 ± 0.01, 0.33 ± 0.01, and 0.16 ± 0.01 h^–1, respectively (Supplementary Table S4). Comparatively, the growth rates of the plasmid-cured evolved clone in the presence of 0, 0.25, 0.5, 1, and 1.5 mM formaldehyde were 1.2 ± 0.00, 1.1 ± 0.01, 0.76 ± 0.00, 0.34 ± 0.00, and 0.18 ± 0.00 h^–1, respectively (Supplementary Table S4). The slight growth defect observed in the plasmid-cured evolved clone compared to the plasmid-cured parental strain likely results from genomic mutations developed during chemical mutagenesis and directed evolution. These mutations are discussed in the WGS section below.

To investigate whether the improved methanol tolerance of the evolved clone results from improved rates of methanol and/or formaldehyde consumption via the Mdh and RuMP pathway enzymes, in vivo formaldehyde production and consumption assays were performed. Briefly, resting cells in minimal medium were used to monitor formaldehyde production following the addition of 1 M methanol (Figure 3A) or formaldehyde consumption following the addition of 1 mM formaldehyde (Figure 3B). The rates of formaldehyde production and consumption were similar between the parental strain and evolved clone, suggesting that the in vivo activities of the methylotrophic enzymes (Mdh, Hps, and Phi) are retained following chemical mutagenesis and ALE. Retention of in vivo activities of the methylotrophic enzymes, and thus a functional synthetic methanol utilization pathway, is further confirmed as methanol-derived carbon is still assimilated into intracellular metabolites following ALE, which is discussed in more detail in the methanol assimilation section below. These results are supported by WGS analysis (discussed in the WGS section below), which did not reveal any unique mutations in the pUD9 plasmid following chemical mutagenesis and ALE. Thus, it does not appear that the improved methanol tolerance of the evolved clone results from an increased rate of methanol and/or formaldehyde consumption through the synthetic methanol utilization pathway. Therefore, chromosomal mutations appear responsible for the improved methanol tolerance of the evolved clone, which agrees with the previous studies in C. glutamicum (Lessmeier and Wendisch, 2015; Wang et al., 2020).

We previously examined mutants of several transcriptional regulators and found that deletion of integration host factor subunit α (ihfA), which is known to repress multiple amino acid metabolic pathways (Goosen and van de Putte, 1995; Karp et al., 2018), resulted in an improved methylotrophic phenotype, as indicated by increased ¹³C-labeling of intracellular metabolites from ¹³C-methanol (Bennett et al., 2020a). Overexpression of cAMP-receptor protein (crp), a known activator of multiple amino acid metabolic pathways (Karp et al., 2018), also resulted in an improved methylotrophic phenotype (Bennett et al., 2020a). Given the success of ALE in improving the methanol tolerance of the synthetic E. coli methylotroph described above (E. coli BW25113 ΔfrmA + pUD9), we sought to also improve the methanol tolerance of these transcriptional regulator mutant strains. E. coli ΔfrmAΔihfA + pUD9 (simply referred to as “ΔihfA”) and ΔfrmA + pUD9 + pCrp (simply referred to as “pCrp”) were serially passaged in LB medium supplemented with increasing methanol concentrations in a similar manner as before (Supplementary Figure S6). However, chemical mutagenesis was not used for this ALE. Once tolerance to 2 M methanol was achieved, which required ca. 10–15 passages, six clones of each strain were examined for improved growth in 2 M methanol as compared to the respective parental strains (Supplementary Figure S7). The “Evolved 2” clone from each strain, hereafter referred to as “evolved pCrp” and “evolved ΔihfA,” exhibited the most improved methanol tolerance and were selected for WGS analyses to identify common mutations responsible for methanol tolerance.

To assess whether a non-methylotrophic E. coli strain could achieve improved methanol tolerance, we serially passaged E. coli ΔfrmA (containing no plasmid) in LB medium supplemented with increasing methanol concentrations in a similar manner as before without chemical mutagenesis. After 11 passages, isolates were examined for improved methanol tolerance in 2 M methanol (Supplementary Figure S8). Interestingly, this non-methylotrophic E. coli strain did achieve improved methanol tolerance through ALE, suggesting that methanol tolerance is not specific to methylotrophic strains. The “KB201” clone (Supplementary Figure S8B), an evolved, non-methylotrophic E. coli BW25113 ΔfrmA strain (containing no plasmid), was selected for WGS analyses to identify common mutations responsible for methanol tolerance.

We next sought to determine whether the evolved strains accumulated common genomic mutations that contributed to the methanol tolerance phenotype. We performed WGS analysis of the original evolved strain (ΔfrmA + pUD9), KB201 and the evolved ΔihfA clone. The evolved pCrp clone was excluded from WGS analysis for simplicity. Each evolved strain accumulated multiple unique mutations when compared to the respective parental strains (Supplementary Table S5). Of considerable interest were mutations found in 30S ribosomal subunit proteins, which were common among all of the evolved strains. Both the evolved ΔihfA clone and KB201 had an identical mutation in rpsQ, a 30S ribosomal subunit protein S17, that resulted in a His31Pro change. The original evolved strain (ΔfrmA + pUD9) had a point mutation in another 30S ribosomal subunit protein S12, rpsL, which resulted in a Gly92Ser amino acid change. The original evolved strain (ΔfrmA + pUD9) was also observed to have a larger number of mutations, especially transition mutations, due to chemical mutagenesis prior to ALE. Since the only common mutation occurring in all three evolved strains was specific to a 30S ribosomal subunit protein, we hypothesize that methanol tolerance results from increased translational efficiency (Haft et al., 2014). To support this hypothesis, an identical mutation in rpsQ (H31P) was found in a previous study that examined ethanol tolerance of E. coli (Haft et al., 2014). This mutation was found to protect cells from ethanol toxicity by increasing the accuracy of protein synthesis. This suggests that the mechanisms of methanol and ethanol tolerance in E. coli are similar and not specific to methylotrophic metabolism, which is why methanol tolerance in the non-methylotrophic E. coli strain was readily achieved. Compared to previous methanol tolerance studies, these results provide a novel insight into alternative mechanisms of methanol tolerance.

Improving the methanol tolerance of a synthetic E. coli methylotroph is not beneficial unless the evolved clone retains the methylotrophic phenotype, i.e., growth on methanol, at low, non-toxic methanol concentrations, which are more practical for industrial bioprocesses to minimize substrate loss via evaporation and ensure complete substrate utilization. Previously, we demonstrated, for the first time, that a synthetic E. coli methylotroph is capable of growth on methanol with a small amount of yeast extract supplementation (Whitaker et al., 2017). To ensure that the original evolved strain (ΔfrmA + pUD9) still exhibits the parental methylotrophic growth phenotype, methylotrophic growth assays in minimal medium supplemented with 1 g/L of yeast extract in the absence and presence of 60 mM ¹³C-methanol were performed. Upon yeast extract exhaustion, we previously demonstrated that methylotrophic E. coli is able to grow on methanol for a brief period, resulting in improved biomass production and termed “methanol boost” (Whitaker et al., 2017). Here, we determined the methylotrophic characteristics of the original evolved strain (ΔfrmA + pUD9) and compared them with the respective parental strain. Cultures were inoculated to an OD₆₀₀ of approximately 0.05, and samples were collected at 48h for determination of biomass production, methanol consumption and ¹³C-labeling. Both strains exhibited similar methylotrophic growth phenotypes in terms of biomass production as the total biomass production in the presence of methanol was approximately 25% higher than that in the absence of methanol, demonstrating that both strains exhibit growth on methanol and a methanol boost of ca. 25% (Figure 4A). Although biomass production profiles were similar between the two strains, total methanol consumption was improved in the original evolved strain (ΔfrmA + pUD9), which consumed 11.1 ± 0.7 mM methanol over the course of 48 h, representing a 16% increase in total methanol consumption over the parental strain, which consumed 9.6 ± 0.2 mM methanol over the course of 48 h (Figure 4B).

Although total methanol consumption was improved in the original evolved strain (ΔfrmA + pUD9), we aimed to determine whether more methanol was assimilated into metabolites and biomass components since methanol consumption includes both assimilation into central metabolism and dissimilation to formate and CO₂, even with ΔfrmA. To quantify methanol assimilation, ¹³C-labeling in intracellular metabolites and proteinogenic amino acids, derived from ¹³C-methanol, was determined. Indeed, significantly higher ¹³C-lableing was realized in the original evolved strain (ΔfrmA + pUD9) as compared to the respective parental strain (Figure 5). For example, average carbon labeling in pyruvate (Pyr), a lower glycolytic intermediate, was increased from 26.7 ± 1.9% in the parental strain to 59.4 ± 0.7% in the evolved clone, representing an increase of 120%. Additionally, average carbon labeling in citrate (Cit), a TCA cycle intermediate, was increased from 19.8 ± 2.0% in the parental strain to 55.5 ± 1.7% in the evolved clone, representing an increase of 180%. Finally, average carbon labeling in alanine (Ala), a proteinogenic amino acid derived from Pyr, was increased from 26.2 ± 2.0% in the parental strain to 56.8 ± 0.9% in the evolved clone, representing an increase of 120%. Taken together, these results suggest that the original evolved strain (ΔfrmA + pUD9) not only consumes more methanol, but also assimilates more methanol-derived carbon into intracellular metabolites and proteinogenic amino acids, which is a crucial characteristic for industrial methanol bioprocesses. Though the evolved strain exhibited increased consumption and assimilation of ¹³C-methanol, the absolute amount of improved consumption was low (ca. 2 mM) compared to the parental strain. This low amount is insufficient to generate a substantial improvement in biomass production, but it provides a step forward toward autonomous methylotrophy as indicated by the ¹³C-labeling analysis. Furthermore, the ¹³C metabolite data represent relative, not absolute, labeling, which further supports why consumption and assimilation of methanol, but not biomass production, are improved in the evolved strain.