Engineering E. coli for simultaneous glucose–xylose utilization during methyl ketone production

Strains, plasmids, and reagents

Genetic manipulations and strain development

Studies were conducted to modulate and optimize the expression of the CRP by testing a set of crp ribosomal binding sites (RBS). To mutate the RBS for CRP, the RBS Library Calculator [22] was used to design RBS mutant library sequences and λ-Red recombineering was used to integrate the library mutants into the host genome. For RBS calculations, the λ-Red-generated 81-bp scar sequence (5′-ATTCCGGGGATCCGTCGACCTGCAGTTCGAAGTTCCTATTCTCTAGAAAGTATAGGAACTTCGAAGCAGCTCCAGCCTACA-3′) was used as the pre-sequence. Based on the native 35-bp RBS sequence for crp (5′-CTCTGGAGAAAGCTTATAACAGAGGATAACCGCGC-3′), a degenerate RBS sequence (5′-CTHTGGTGAAAGCTTATAACTGAGGMRAACCGCGT-3′) was generated with a broad range of predicted translation initiation rates (TIR) for a total 12 variant sequences. RBS sequences and their predicted TIR values are listed in Additional file 1: Table S2. For reference, the native crp RBS sequence in E. coli DH1 was predicted to have a TIR value of 2441 au based on the RBS Calculator [23, 24].

Media and culture conditions

Lysogeny broth (LB) was used for routine cell growth and propagation. Kanamycin was added to the growth medium at a final concentration of 50 µg mL⁻¹, when required. M9-MOPS minimal medium with 1% total sugars (5 g L⁻¹ glucose and 5 g L⁻¹ xylose) or 2% total sugars (10 g L⁻¹ glucose and 10 g L⁻¹ xylose) as carbon sources was used for production experiments. The composition of M9-MOPS minimal medium followed the recipe previously described [2]. For production experiments, strains were first adapted in M9-MOPS minimal medium with 1% total sugars for 3 passages before being inoculated into production medium with glucose and xylose.

Analysis of cell growth and sugar metabolism

Cell growth was monitored by measuring optical density at 600 nm (OD₆₀₀). Sugars were measured with an Agilent 1100 Series HPLC system, equipped with an Agilent 1200 Series refractive index detector (RID) (Agilent Technologies, CA) and Aminex HPX-87H ion-exclusion column (300-mm length, 7.8-mm internal diameter; Bio-Rad Laboratories, Inc., Hercules, CA). The column temperature was 50 °C, and 4 mM sulfuric acid was used as the mobile phase with a flow rate of 0.6 mL min⁻¹ for 24 min. The quantification of glucose and xylose was conducted by external standard calibration with authentic standards.

Production and analysis of methyl ketones

For methyl ketone production, strains were inoculated into 50 mL M9-MOPS minimal medium with 1% total sugars, or 25 mL with 2% total sugars, in 250-mL shake flasks and cultured at 37 °C with 200-rpm agitation. The starting OD₆₀₀ during production was ca. 0.01. Gene expression was induced by adding 0.2 mM IPTG and 1 mM arabinose after 6 h of growth. Five mL of decane (Reagent-Plus ≥ 99% purity, Sigma-Aldrich, St. Louis, MO) amended with perdeuterated tetracosane (C₂₄D₅₀) and 3-tetradecanone (Sigma-Aldrich, St. Louis, MO) as internal standards was also added to the cultures during induction. The decane overlay was sampled for the measurement of methyl ketones; analysis by electron ionization gas chromatography–mass spectrometry (GC–MS) was conducted as previously described [3].

Batch fermentation of strains EGS1895 and XW1075 in a 2-L bioreactor for methyl ketone production

Batch fermentation was carried out in a 2-L bioreactor equipped with a Sartorius BIOSTAT B plus control unit for regulating dissolved oxygen (DO), pH, and temperature. Frozen glycerol stocks of M9-MOPS-adapted cells were used to seed a test tube containing 5 mL of M9-MOPS minimal medium with 1% total sugars (5 g L⁻¹ glucose and 5 g L⁻¹ xylose) as previously described. After 30 h of growth, the cultures were diluted 1:250 into a 250-mL flat-bottom shake flask containing 50 mL of M9-MOPS minimal medium supplemented with 1% total sugars. This culture was grown for another 30 h as before and was used to inoculate 1.25 L of medium in the bioreactor. The medium was adapted from Korz et al. [25] and was composed of M9 salts (6.8 g L⁻¹ Na₂HPO₄, 3.0 g L⁻¹ KH₂PO₄, 1.0 g L⁻¹ NH₄Cl, 0.5 g L⁻¹ NaCl) supplemented with 0.5 g L⁻¹ of MgSO₄·7H₂O, 0.18 g L⁻¹ of NH₄Cl, 1.0 mg L⁻¹ thiamine, 10 nM of FeSO₄·7H₂O, 100 µM CaCl₂·2H₂O, micronutrients as described previously [2], 10 g L⁻¹ of glucose, 10 g L⁻¹ of xylose, and 50 µg mL⁻¹ of kanamycin. The temperature of the bioreactor was maintained at 37 °C throughout the fermentation and the culture was maintained at pH 6.5 automatically by the addition of a 10 M potassium hydroxide solution. The initial stir rate and airflow were set at 200 rpm and 1 VVM (volume of air per volume of liquid per minute), respectively. Dissolved oxygen was maintained above 40% of saturation via cascade control of adjustment of stirrer speed (up to 1600 rpm), followed by air-flow rate (up to 2.0 VVM). Cultures were induced with 1 mM arabinose and 0.5 mM IPTG at 6 h after initiation of batch phase. In addition, 150 mL of dodecane (Sigma-Aldrich, ReagentPlus ≥ 99% purity) amended with 3 mg mL⁻¹ of 3-tetradecanone (Sigma-Aldrich) as an internal standard, was added into the bioreactors.

At selected times, 10–15-mL samples were removed from the bioreactors via a syringe affixed to the sampling tube while the stirrer was still operating. Approximately 50 µL of cell cultures were filtered through a 0.2-µm syringe membrane filter directly into a vial for HPLC analysis and the rest of the cultures transferred to a 15-mL Falcon tube. After allowing the samples to sit in the 15-mL tube for 1 min, the supernatant dodecane overlay was pipetted out into a 2-mL microcentrifuge tube and centrifuged at 21,130×g for 10 min to obtain a better-resolved aqueous-organic interface. The dodecane overlay was transferred into a glass vial and stored at 4 °C until GC–MS analysis.

Engineering simultaneous glucose–xylose utilization in methyl ketone-overproducing strain EGS1895

We engineered several strains by manipulating key genes in pentose metabolism (XW1014, XW1024, and XW1044; Tables 1, 2) and evaluated their ability to simultaneously utilize glucose and xylose (Fig. 1). The control strain, EGS1895, presented a typical diauxic pattern in which xylose utilization began after glucose was fully depleted. In contrast, newly engineered strains displayed glucose–xylose co-utilization to varying degrees rather than a strict diauxic profile. Among these engineered strains, XW1014 (with constitutive expression of xylA and xylF plus a point mutation in the xylA promoter, xylA ^up ) showed the best performance for simultaneous utilization of glucose and xylose (Fig. 1). This strain had identical consumption rates for glucose and xylose at 1% sugar conditions, while a slight decrease in xylose consumption was observed at higher sugar concentration (2%). The inactivation of araE (XW1024; Tables 1, 2) did not result in better sugar co-utilization than was observed for strain XW1014, nor did the manipulations made for strain XW1044 (alleviating AraC-mediated repression through four collective araC-related manipulations, including araC deletion from both the genome and plasmid as well as replacement of promoters for araB and araF). Although both strains XW1024 and XW1044 showed favorable simultaneous consumption rates of glucose and xylose at 1% sugar conditions, their xylose consumption dramatically decreased at higher sugar concentration (2%).

In addition, because ptsG deficiency is a well-studied mechanism for mitigating CCR in E. coli [10], the glucose transporter EIIBC^Glc encoded by ptsG was deleted from EGS1895 to investigate the effect on sugar co-utilization (strain XW1004; Table 1). Strain XW1004 did not display a better sugar co-utilization profile than strain XW1014 (Fig. 1).

Methyl ketone production was also investigated among these strains engineered for hexose–pentose co-utilization. Compared with the titer of the control strain EGS1895 (~ 690 mg L⁻¹), methyl ketone production was significantly reduced in all four modified strains (Fig. 2). The best performing strain for sugar co-utilization, XW1014, only produced ~ 140 mg L⁻¹ total methyl ketones (1% total sugars), which is approximately fivefold lower than for strain EGS1895. Strains with more genetic manipulations produced even lower methyl ketone titers; for example, strains XW1024 and XW1044 produced < 60 mg L⁻¹ methyl ketones. Although the ΔptsG strain (XW1004) showed the highest methyl ketone titer among these four strains, its diminished glucose utilization was not optimal and it was not pursued further. Despite its relatively low methyl ketone titer, strain XW1014 had the most favorable combination of sugar co-utilization and methyl ketone production of the strains tested.

Optimization of methyl ketone production in strain XW1014 by enhancing NADPH availability

Although strain XW1014 was successfully engineered for simultaneous glucose–xylose consumption, the significantly reduced methyl ketone titer in this strain necessitated further engineering to improve commercial relevance. We hypothesized that enhancing NADPH availability could be a fruitful engineering target because (1) the biosynthesis of fatty acids (methyl ketone precursors) in E. coli is an NADPH-demanding process and (2) xylose metabolism, particularly when simultaneous with glucose metabolism, could disrupt NADPH production in a host cell (e.g., strain XW1014) compared to conditions with glucose as a sole carbon source. Fatty acid biosynthesis results in net consumption of NADPH due to demand from two key reductases—FabG (β-ketoacyl-ACP reductase) and potentially, FabI (enoyl-ACP reductase), which can utilize either NADH or NADPH as a cofactor [1, 26]. To illustrate the substantial NADPH demands of fatty acid/methyl ketone biosynthesis, production of 1 mol of a C₁₃ methyl ketone (2-tridecanone) from glucose using the relevant metabolic pathway [2] would result in net consumption of 6 (or 12) mol of NADPH and net production of 9 (or 15) mol of NADH, depending on FabI cofactor usage.

By virtue of where xylose enters central carbon metabolism in E. coli, xylose metabolism tends to result in less flux than glucose metabolism through the oxidative, NADPH-generating steps of the pentose phosphate pathway (PPP), namely reactions catalyzed by glucose-6-phosphate dehydrogenase (Zwf) and phosphogluconate dehydrogenase (Gnd); however, xylose metabolism can take advantage of other sources of NADPH, such as malic enzyme and transhydrogenase [27]. The situation is likely more complex when considering sugar utilization and NADPH production in strain XW1014 compared to that in control strain EGS1895. Compared with the sequential metabolism from glucose to xylose during diauxic growth (strain EGS1895), simultaneous metabolism of glucose and xylose (strain XW1014) could alter NADPH production by re-distributing flux between glycolysis and the PPP. For example, it is possible that the flux of glucose carbon through the oxidative PPP might be reduced when xylose co-utilization is occurring, because xylose metabolism will satisfy the cell’s needs for downstream PPP metabolites required for anabolism, such as erythrose 4-phosphate (needed for aromatic amino acid biosynthesis) and ribose 5-phosphate (needed for nucleic acid biosynthesis).

We implemented two strategies for increasing NADPH supply in strain XW1014: (1) deleting pgi (glucose-6-phosphate isomerase) from the chromosome to divert flux from glycolysis through the oxidative PPP (Fig. 3) and (2) overexpressing maeB (malic enzyme), which leads to NADPH generation by oxidative decarboxylation of malate to pyruvate (Fig. 3). ¹³C Metabolic flux analysis studies in E. coli have shown that pgi deletion results in substantial production of NADPH by diversion of flux from glycolysis through the oxidative PPP, and that excessive accumulation of NADPH (cofactor imbalance) in Δpgi strains can be at least partially ameliorated by NADPH consumption through transhydrogenase [28, 29]. In our Δpgi strain (XW1054; Table 1), it was anticipated that a portion of the NADPH made available by the pgi deletion might facilitate fatty acid/methyl ketone biosynthesis by better satisfying its high NADPH demands than did central carbon metabolism in strain XW1014.

Production results showed that the Δpgi strain (XW1054) had dramatically improved methyl ketone titer (850 mg L⁻¹) relative to strain XW1014 after 96 h at 1% total sugar conditions (Fig. 4); this methyl ketone titer was comparable to that of the control strain (EGS1895). Under 2% total sugar conditions, the methyl ketone titer of strain XW1054 (~ 1300 mg L⁻¹ after 96 h) was also comparable to that of strain EGS1895 (~ 1600 mg L⁻¹). However, xylose showed a slower consumption rate than glucose after pgi was deleted, and slower cell growth was also observed during production. In contrast to methyl ketone titer improvement for strain XW1054, the overexpression of maeB with or without pgi deletion (strains XW1055 and XW1018; Table 1) did not result in improvement in methyl ketone production (Additional file 1: Figure S1).

Based upon the results for strain XW1054 (Δpgi), it is possible that NADPH is more limiting when xylose is used as a carbon source. Indeed, we observed that the control strain (EGS1895) produced very low methyl ketone titers when xylose was used as the sole carbon source in minimal medium (Additional file 1: Figure S2).

Optimization of methyl ketone production in strain XW1014 by mutating the RBS of crp

While enhancing potential NADPH supply (via pgi deletion) substantially improved methyl ketone production with mixed glucose–xylose medium, several lines of evidence suggested that the engineered strains were experiencing suboptimal sugar utilization (e.g., strain XW1054 in Fig. 4), and potentially, suboptimal methyl ketone production, that had causes beyond NADPH limitation. For example, NADPH limitation alone does not seem to explain the dramatic reduction in methyl ketone titer in both strain XW1004 (ΔptsG) and strain XW1014 (introduced constitutive promoters to xylA and xylF) (Fig. 2), as these genetic modifications are not clearly linked to NADPH supply.

A possible explanation for these results is changes in intracellular distributions of the global regulator CRP. For strain XW1014, promoter replacement for xylA and xylF resulted in removal of a CRP binding site from the intergenic region between xylA and xylF [30]. As a global regulator, CRP not only plays an important role in carbon catabolite repression, but also controls the transcription of more than 100 genes in E. coli, such as key genes in fatty acid metabolism (e.g., fadD, fadH) [31] and in central carbon metabolism (e.g., pgi, zwf, gnd) [30, 32]. Thus, the promoter change in strain XW1014 might have altered the level of free CRP and directly and indirectly affected the transcription of many other genes related to fatty acid metabolism. Similarly, changes to intracellular CRP pools might also explain why methyl ketone production was reduced in the ΔptsG strain (XW1004): the absence of PtsG likely increased cAMP availability [33], and in turn, altered the level of free intracellular CRP, which interacts with cAMP to make the cAMP–CRP complex.

Based on this reasoning, one possible strategy for improving methyl ketone production is to optimize the expression level of CRP in strain XW1014. We attempted to modulate CRP availability by replacing the native crp RBS with synthetic RBSs of varying strengths. We created a mutant crp RBS library with broad range of predicted TIR values (8–7290 au, Additional file 1: Table S2). A total of 7 RBS variants with different TIRs were identified by sequencing from the mutant library. Screening of this library was conducted with 5-mL cultures in M9-MOPS medium (50-mL test tubes), and one mutant (strain XW1064) was selected that showed significant improvement in methyl ketone production (~ 900 mg L⁻¹ after 96 h with 1% total sugars, Additional file 1: Figure S3). Notably, the predicted TIR of strain XW1064 was 13 au, which is approximately 188-fold lower than the predicted native TIR (2441 au) of crp. Scaled up production of strain XW1064 in 250-mL shake flasks resulted in methyl ketone titers up to ~ 450 mg L⁻¹ without compromised cell growth (Fig. 4).

This result supported our hypothesis that optimized expression of CRP is able to improve methyl ketone production in the strains engineered for glucose–xylose co-utilization. However, we also noticed that the consumption rate of xylose in strain XW1064 was slower than that of glucose, especially under 2% total sugar conditions (Fig. 4).

Seeking the best candidate by combining engineering strategies

Given the complementary features of the above strategies (Δpgi and CRP downregulation) on cell growth and methyl ketone production, and the fact that they both effectively improved methyl ketone production in strain XW1014, we decided to combine these two strategies to obtain an additive effect. Overall, combining Δpgi and CRP downregulation (strain XW1074; Table 1) created superior phenotypes in cell growth and methyl ketone production compared to use of either strategy alone (Fig. 4). This strain produced up to 570 mg L⁻¹ methyl ketones at 1% total sugar conditions, but reached a higher titer at 2% total sugars (~ 1600 mg L⁻¹) that was comparable to that of the control strain (EGS1895). Glucose and xylose were simultaneously consumed by strain XW1074 (Fig. 4) after a lag period, but utilization of xylose was still slower than that of glucose. Surprisingly, the added maeB overexpression (strain XW1075) dramatically improved sugar co-utilization (albeit with the same lag period, likely caused by pgi deletion; [34, 35]). As a result, strain XW1075 achieved synchronized consumption rates for glucose and xylose at both 1 and 2% total sugar conditions. Methyl ketone titers in strain XW1075 were up to 700 and 1100 mg L⁻¹ at 1 and 2% total sugars, respectively. Thus, these two strains engineered with combined strategies (XW1074 and XW1075) represented a favorable phenotype displaying simultaneous utilization of glucose and xylose without substantially sacrificing methyl ketone production relative to the control strain (EGS1895) (Figs. 4, 5).

Strain XW1075 performance during batch fermentation

Strain XW1075 also compared favorably to control strain EGS1895 in batch fermentation mode. Glucose and xylose were utilized concurrently in strain XW1075 (albeit at unequal rates), whereas strain EGS1895 displayed a typical diauxic pattern, including sequential sugar utilization (Fig. 6). Correspondingly, strain XW1075 had a more consistent methyl ketone production yield (8.7–9.8%) than the control strain (6.9–10.0%). At 72 h, the methyl ketone titer of strain XW1075 was 2 g L⁻¹, which was ca. 33% higher than that of strain EGS1895 (1.5 g L⁻¹).

Comparison of the results in Fig. 6 with those of strains XW1075 and EGS1895 grown with pure glucose or xylose (Additional file 1: Figure S2) reveals that co-utilization of glucose and xylose in strain XW1075 enabled substantially better methyl ketone production than did utilization of either sugar alone. In fact, methyl ketone production was negligible for strain XW1075 utilizing either pure glucose or pure xylose (Additional file 1: Figure S2). Notably, strain EGS1895 also produced negligible methyl ketones when grown on pure xylose (Additional file 1: Figure S2), but produced substantial methyl ketones while metabolizing xylose after diauxic depletion of glucose (Fig. 6). From Figs. 4 and 6, it appears that glucose metabolism supported both growth and methyl ketone production in strain EGS1895, whereas xylose metabolism supported methyl ketone production but little or no growth.