Awakening the natural capability of psicose production in *Escherichia coli*

The study addresses the need for accessible, palatable, low-calorie sugar substitutes to combat rising obesity and diabetes. D-psicose (allulose), a zero-calorie C3 epimer of D-fructose with favorable organoleptic and functional properties, is limited by current production methods that rely on in vitro epimerases with thermodynamic constraints (~50% equilibrium conversion), resulting in costly purification and limited adoption. The research hypothesizes that Escherichia coli harbors a thermodynamically favorable, native route to produce D-psicose from D-glucose via phosphorylation–epimerization–dephosphorylation, where dephosphorylation provides a strong thermodynamic sink. The purpose is to reveal and engineer this native capability, redirect central carbon flux toward D-psicose, and implement dynamic regulation to balance growth and production, achieving high yield, productivity, and complete substrate utilization under simple conditions.

Current commercial D-psicose production employs D-tagatose-3-epimerase and D-psicose-3-epimerase acting on D-fructose, but the reaction is thermodynamically unfavorable (ΔG ~ +5 kJ mol−1), limiting conversion to ~50% and necessitating expensive purification. Engineering efforts have improved enzyme properties (catalysis, thermostability, pH range), yet equilibrium constraints persist. Alternative in vitro systems couple phosphorylation/dephosphorylation to drive flux but require multiple enzymes, elevated temperatures, purified enzymes, and external phosphoryl donors, and generate side products (e.g., D-fructose) that complicate purification. Prior in vivo attempts in E. coli using heterologous pathways yielded low titers (~1.6 g L−1) with poor growth or required supplementation with ATP and sodium hexametaphosphate, increasing costs and leaving residual glucose and fructose. These limitations motivate a whole-cell catalysis approach leveraging native enzymes, thermodynamic driving via dephosphorylation, and dynamic flux control.

Pathway design and hypothesis: Proposed a native E. coli route converting D-glucose to D-psicose via G6P → F6P → psicose-6-phosphate (P6P) → D-psicose, where dephosphorylation of P6P provides a strong thermodynamic sink (predicted ΔGm −31.1 kJ mol−1 at 1 mM). Strategy: accumulate F6P and exploit native epimerase/phosphatase activities, then engineer flux control and competing pathways.

Strains and genetic engineering: Base strains MG1655 and AL3601 (T7 RNAP) were used. Gene deletions were introduced via CRISPR-Cas9-mediated homologous recombination to redirect flux: ΔpfkA (reduce glycolysis; build F6P), Δzwf (limit PPP), ΔrpiB (prevent P6P reassimilation via allose-6-P isomerase), ΔmanA (reduce mannose side-product), Δpgm (block glycogen biosynthesis), and tested ΔptsG, ΔptsH (modulate PTS). Verification by sequencing. Key production strains assembled as per Table 1 in the paper (Strains 1–7).

Enzyme identification and expression: Identified AlsE as the epimerase (F6P↔P6P) by deletion analysis (ΔalsE abolished production). Screened phosphatases (HxpB, YbiV, YidA, HxpA, YihX, YigL) coexpressed with AlsE from Pr7 or PLlacO1; HxpB gave highest D-psicose. Constructed operons expressing alsE–hxpB under PLlacO1 or stationary-phase promoter PgadB. Supplemented glucose uptake via plasmid expression of galP and glk under PLlacO1.

Promoter and dynamic control: Evaluated stationary-phase promoters (PgadB, P(cbpA2), PihfA4, Pdps) using sfgfp reporters; selected PgadB (strong expression in late log/early stationary phase) to phase-match production with cell physiology. Implemented CRISPRi with dCas9 under aTc-inducible Ptet and constitutive sgRNA targeting the pfkB promoter to dynamically knock down residual glycolysis during production. Compared single- vs dual-plasmid CRISPRi architectures; used single-plasmid system to minimize growth burden.

Structural modeling: Used AlphaFold-predicted structures and Rosetta (Relax and GALigandDock) to dock P6P into candidate phosphatases; assessed hydrogen-bonding patterns and active-site geometry constraints (including Mg2+ coordination) to rationalize observed activities; identified key HxpB residues interacting with P6P hydroxyls and phosphate positioning.

Culturing and production conditions: For regular density: inoculate into M9P (M9 + 5 g L−1 yeast extract) with 10–40 g L−1 glucose, grow at 37 °C to target OD600 (~0–1), then induce (as applicable: IPTG 1 mM; aTc 100 ng mL−1) and shift to 30 °C for 24 h. Explored timing of temperature shift and inductions. For high-density production: grow to OD600 ~1, induce, then pellet and resuspend to OD600 ~10 in M9P with 40 g L−1 glucose, continue at 30 °C; sampled at 0, 4, 8 h.

Analytics: Quantified D-psicose, D-glucose, and D-mannose by HPLC (Rezex RCU-USP sugar alcohol column, RI detection; water mobile phase; 0.5 mL min−1; 83 °C column). Identified unknown side-product via GC-MS (derivatization with methoxyamine/MSTFA; LECO Pegasus IV TOF MS) as D-mannose by RT and mass spectra matching standards.

Comparative expression systems: Compared T7 RNAP system (PT7) versus PLlacO1 driving alsE–hxpB; selected PLlacO1 due to reduced growth burden and tighter repression uninduced. Selected PgadB for stationary-phase-driven expression of alsE–hxpB.

Data analysis: Calculated titers, yields (mol psicose per mol glucose, theoretical max 100%), productivities (g L−1 h−1), and specific titers (g L−1 OD600−1).

• E. coli can natively produce D-psicose from D-glucose when F6P pools are increased: ΔpfkA strains produced 0.15–0.24 g L−1; deletion of alsE abolished production, implicating AlsE as the epimerase.
• Phosphatase screening with coexpressed AlsE identified HxpB as the most effective for converting P6P to D-psicose (0.55 g L−1 in AL3601), outperforming YbiV (0.21 g L−1) and YidA (0.20 g L−1).
• Structural modeling suggested that catalytically competent P6P binding requires at least three hydrogen bonds including one to the terminal hydroxyl; HxpB forms extensive interactions positioning P6P for hydrolysis.
• Removal of competing pathways increased production: ΔpfkA Δzwf ΔrpiB (Strain 1) yielded 2.31 g L−1; eliminating mannose formation via ΔmanA (Strain 2) increased D-psicose vs Strain 1 and reduced D-mannose from 2.49 to 0.69 g L−1.
• Stationary-phase promoter PgadB driving alsE–hxpB (Strain 3) improved titers and growth across glucose levels; at 40 g L−1 glucose and shift at OD600 ~1, produced 9.13 g L−1 (vs 5.42 g L−1 with PLlacO1) with higher ΔOD600.
• Enhancing non-PTS glucose uptake via PLlacO1:galP–glk (Strain 4) increased titer to 13.81 g L−1 (specific titer 3.2 g L−1 OD600−1; 55% yield) with IPTG; no IPTG gave 8.67 g L−1 (49% yield). Overexpression imposed some growth burden.
• Glycogen pathway knockout Δpgm (Strain 5) further raised performance: 14.66 g L−1, specific titer 5.5 g L−1 OD600−1, 58% yield. PTS deletions (ΔptsG) were detrimental or neutral, likely due to broader regulatory effects; ΔptsH neutral.
• Dynamic glycolysis control with CRISPRi targeting pfkB (Strain 7) yielded up to 62% yield (with aTc), producing 11.40–13.65 g L−1 with specific titers ~3.6–3.8 g L−1 OD600−1; growth reduced relative to controls.
• High-cell-density production (OD600 ~10, 40 g L−1 glucose): 15.3 g L−1 in 8 h, productivity 1.9–2.0 g L−1 h−1; yield rose from 35% (0–4 h) to 53% (4–8 h). At 24 h, complete glucose consumption and clean chromatograms facilitated downstream processing.
• Overall best metrics: titer 15.3 g L−1; highest yield 62%; productivity ~2 g L−1 h−1 under test-tube conditions; complete glucose consumption achieved.

The work demonstrates that E. coli contains a latent, thermodynamically favorable pathway to produce D-psicose from D-glucose by routing F6P through AlsE-mediated epimerization to P6P followed by dephosphorylation by HxpB. Placing the strongly exergonic dephosphorylation at the end of the pathway circumvents the unfavorable equilibrium of direct fructose–psicose epimerization and drives flux toward product. By removing competing pathways (PPP via Δzwf, glycolysis via ΔpfkA, P6P reassimilation via ΔrpiB, mannose formation via ΔmanA, glycogen via Δpgm) and supplementing glucose uptake with GalP–Glk, intracellular F6P availability is increased and directed toward D-psicose formation. Dynamic regulation matched production gene expression to stationary phase (PgadB) to protect growth, while CRISPRi of pfkB reduced residual glycolytic drain during production, improving yields up to 62%. The strategy achieved high titers and productivities with complete substrate utilization under simple cultivation, highlighting the advantages of whole-cell catalysis for rare sugar manufacture. These results address the thermodynamic and process limitations of current in vitro epimerase-based methods and indicate industrial feasibility with simplified downstream purification due to full glucose consumption and limited side-products.

This study reveals and engineers a native, thermodynamically driven pathway in E. coli to convert D-glucose to D-psicose via phosphorylation–epimerization–dephosphorylation, eliminating reliance on in vitro epimerases constrained by equilibrium. Combining deletion of competing pathways, expression of native alsE and hxpB, augmentation of glucose uptake (galP–glk), and dynamic flux control (stationary-phase promoters and CRISPRi) delivered complete glucose consumption, up to 15.3 g L−1 D-psicose, ~2 g L−1 h−1 productivity, and a maximum yield of 62% in test tubes. The approach offers a sustainable, cost-effective route to D-psicose and is adaptable to other rare sugars. Future efforts may focus on scale-up in bioreactors, optimization of induction and promoter strategies, further minimization of side-products, and exploration of additional host strains and transport/secretion engineering to enhance productivity and downstream processing.

• Growth–production trade-offs persist: overexpression of galP–glk and CRISPRi knockdown of pfkB reduced growth and, in some cases, absolute titers despite improving specific titers and yields.
• Side-product formation: D-mannose was formed due to reverse flux via ManA from accumulated F6P; although ΔmanA reduced it substantially, residual mannose remained.
• PTS gene deletions (ΔptsG) negatively impacted production, reflecting complex global regulatory roles; fine control of PTS signaling remains challenging.
• Reported results are from unoptimized test-tube conditions (M9P with yeast extract); large-scale bioreactor performance, process control, and technoeconomic analysis were not evaluated.
• Dependence on stationary-phase promoter activity and culture phase may complicate transfer to varying process conditions; promoter tuning may be required.
• No detailed downstream processing or product purification workflow was demonstrated beyond analytical confirmation; impurity profiles beyond mannose were not extensively reported.