Construction and modular implementation of the THETA cycle for synthetic CO₂ fixation

The study addresses the challenge of designing, constructing, and validating new-to-nature CO₂-fixation pathways that can operate efficiently both in vitro and in vivo. While biological CO₂ fixation is central to the global carbon cycle and several natural pathways exist, nature occupies a limited solution space, leaving many theoretically feasible pathways unexplored. Synthetic biology enables assembling novel combinations of enzymes and reactions to create oxygen-insensitive, efficient CO₂-fixation cycles. However, implementing such orthogonal pathways in living cells is difficult due to potential interactions with native metabolism and increased network complexity. The authors propose and realize the THETA cycle, a pathway converting CO₂ directly into acetyl-CoA, and investigate its modular deployment in Escherichia coli to evaluate feasibility, performance, and integration with host metabolism.

The work builds on prior identification and implementation of natural and synthetic CO₂-fixation pathways. Seven natural pathways are known and adapted to diverse environments. Over 30 new-to-nature pathways have been theoretically designed, with two oxygen-insensitive synthetic pathways (CETCH cycle and rGPS-MCG cycle) previously realized in vitro. Natural CO₂-fixation pathways have been transplanted into model organisms, but demonstrating a complex, orthogonal synthetic pathway in vivo remains an open challenge. The authors select highly efficient carboxylases—phosphoenolpyruvate carboxylase (Ppc) and crotonyl-CoA carboxylase/reductase (Ccr)—as the core of their design, informed by kinetic, thermodynamic, and mechanistic advantages over RuBisCO, and previous successes using Ccr in synthetic pathways.

Design: The THETA cycle was designed around two fast, thermodynamically favorable carboxylases (Ppc and Ccr). The pathway was modularized into three modules: (1) pyruvate to succinate via PEP synthase, Ppc, malate dehydrogenase, fumarase, and fumarate reduction; (2) succinate to crotonyl-CoA via succinyl-CoA ligase (SucCD), succinate semialdehyde/formaldehyde interconversions and reactions from DC/4HB and 3HP/4HB cycles; (3) crotonyl-CoA to acetyl-CoA and pyruvate via the ethylmalonyl-CoA and 3HP bicycle reactions. Overall stoichiometry: CO₂ + HCO₃⁻ + 4 ATP + 3 NADPH + 2 NADH + CoA → acetyl-CoA + FADH₂ (ΔrGm = −261 kJ mol⁻¹). MDF analysis validated thermodynamic feasibility in vivo. In vitro construction: 17 enzymes (from 9 organisms) were cloned, expressed and purified (most as His-tag fusions); some commercial enzymes (Ppc, Mdh, Fum) were used. Methylsuccinyl-CoA oxidation was implemented with an engineered oxidase (Mco) using O₂ plus catalase; fumarate reduction used NADH-dependent soluble Frd from Trypanosoma brucei. Reactions were performed in MOPS buffer with NaHCO₃ as CO₂ source, started with pyruvate, and monitored by LC-MS for CoA esters. Rational optimization: Initial cycle (THETA 1.0) produced acetyl-CoA but accumulated malyl-CoA due to SucCD promiscuity; a malyl-CoA thioesterase was added (THETA 2.0), and cofactor regeneration (ATP and NAD(P)H) implemented. Bottlenecks were probed by starting from different intermediates; fumarate reduction was limiting. A dihydroorotate-driven fumarate reductase bypass using DHOD1a (Trypanosoma cruzi) and DHOD1b (Bacillus subtilis) replaced Frd (THETA 3.0). Machine learning-guided optimization: The METIS active-learning workflow with Echo liquid handler explored 34-component combinatorial space (up to 8 levels per component), testing 30 conditions in triplicates over 9 rounds in 10 µl assays. Conditions started with 200 µM pyruvate and ran 3 h before LC-MS quantification. Feature importance suggested increasing Mco and lowering Hbs and coenzyme B12 improved productivity. Optimized variant THETA 3.9.9 was validated at 120 µl scale. Product scope extension: For malonyl-CoA, an engineered propionyl-CoA carboxylase (Pcc*) carboxylated acetyl-CoA, increasing output. For glycolate, the OAA-to-succinate segment was replaced with a glyoxylate cycle module to release glyoxylate, then reduced to glycolate. In vivo implementation strategy: Modular transfer to E. coli using auxotrophic strains and growth rescue or 13C labelling: - Module 1 (pyruvate→succinate): Built in a succinyl-CoA auxotroph (SL1; derived from acetyl-CoA auxotroph JCL301 with additional deletions sucAB, aceA). Introduced NADH-dependent Frd (T. brucei) or overexpressed native NadB to enable fumarate reduction. Growth rescue on minimal medium with glucose and acetate; 13C-labelled succinate patterns verified operation. - Module 2 (succinate→crotonyl-CoA): Used acetyl-CoA auxotroph SL2 (JCL301 with additional deletions kbl, ltaE). A C4-acetyl-CoA shunt (AtoB, Bhbd, Crt) plus crotonyl-CoA ligase (DmdB1) enabled growth on crotonate. Module 2 genes (SucCD, Scr, Ssr, Hbs/Cat2, Hbd) were assembled into operons; enzyme homolog swaps (Scr/Ssr from Porphyromonas gingivalis) and replacing Hbs with 4HB-CoA transferase (Cat2) increased activities. Short-term evolution on selection plates yielded strains 1736 and 1737 that grew on succinate with module 2 + shunt. 13C labelling (with [U-13C]succinate) of amino acids confirmed conversion of succinate to acetyl-CoA (Leu predominantly M+2; Ala mainly unlabelled). - Module 3 (crotonyl-CoA→acetyl-CoA + pyruvate): Initial chromosomal integration of epi, ecm, mct, meh, ccl in SL2 with plasmid expression of ccr, mco and DmdB1 did not rescue growth. Metabolomics showed methylsuccinate accumulation from crotonate, implying hydrolysis of methylsuccinyl-CoA. Deletion of thioesterases tesB and yciA reduced methylsuccinate production 27-fold (OD-normalized), identifying them as responsible thioesterases. Native Sdh oxidized methylsuccinate to mesaconate (validated by sdhA/B knockouts reducing mesaconate 80–90%, and purified SdhAB activity ~70% of that with succinate). An acid-bypass strategy used Yersinia pestis itaconate CoA transferase (Ict) to convert mesaconate to mesaconyl-C1-CoA, feeding into the lower module (Meh, Ccl) to produce pyruvate. Overexpressing Sdh increased mesaconate 7-fold. In strain HH422 (MG1655-derived), with epi/ecm integrated and plasmid-borne ict, ccr, ccl, meh, sdhCDAB, dmdB1, non-labelled pyruvate was produced from crotonate in [U-13C]glucose minimal medium with 2% LB, confirming in vivo module 3 activity. Analytical and computational methods: LC-MS quantified CoA esters and extracellular acids; 13C labelling of amino acids and pyruvate traced flux; whole-genome sequencing (Illumina and Nanopore) characterized evolved strains; MDF analysis used equilibrator_api/pathway. Growth rescue assays used M9 minimal medium protocols; short-term evolution on plates selected succinate growth.

- A new-to-nature CO₂-fixation cycle (THETA) was designed and constructed, comprising 17 enzymes from 9 organisms, with overall stoichiometry CO₂ + HCO₃⁻ + 4 ATP + 3 NADPH + 2 NADH + CoA → acetyl-CoA + FADH₂ and ΔrGm = −261 kJ mol⁻¹. - In vitro performance improvements: • THETA 1.0 produced 8.5 µM acetyl-CoA in 60 min but accumulated malyl-CoA due to SucCD promiscuity. • THETA 2.0, adding a malyl-CoA thioesterase and cofactor regeneration, produced 38 µM acetyl-CoA in 60 min with minimal by-products. • A dihydroorotate-driven Frd bypass (THETA 3.0) boosted acetyl-CoA to ~200 µM in 60 min, independent of starting substrate (pyruvate, fumarate, or succinate). CO₂-fixation rate reached 2.7 nmol min⁻¹ mg⁻¹ of core cycle proteins. • Machine learning-guided optimization (METIS) achieved up to 1,150 µM acetyl-CoA from 200 µM pyruvate substrate in 3 h (10 µl assays). Manual 120 µl validation yielded ~890 µM. Overall, a ~135-fold improvement over THETA 1.0. • Lower starting pyruvate increased turnover: starting with 100 µM and 50 µM pyruvate gave ~680 µM and ~470 µM acetyl-CoA, corresponding to ~7 and ~9 total turns. • Beneficial parameter shifts included increasing Mco and lowering Hbs and coenzyme B12; Mco remained rate-limiting. - Product scope extension: • Adding engineered Pcc* to carboxylate acetyl-CoA increased malonyl-CoA output by ~9% (THETA 3.9.9.mc). • Replacing the OAA→succinate segment with a glyoxylate cycle module led to >530 µM glycolate from 50 µM succinate (turnover >10) (THETA 3.9.9.gc). - In vivo module implementations in E. coli: • Module 1 (pyruvate→succinate) restored growth of a succinyl-CoA auxotroph (SL1) via expression of T. brucei Frd or overexpression of native NadB; 13C-labelling confirmed expected succinate patterns. • Module 2 (succinate→crotonyl-CoA), combined with a C4-acetyl-CoA shunt and DmdB1, enabled an acetyl-CoA auxotroph (SL2) to grow on crotonate; after short-term evolution, strains 1736 and 1737 grew on succinate. 13C-labelling with [U-13C]succinate showed leucine predominantly M+2 and alanine mostly unlabelled, confirming succinate-to–acetyl-CoA conversion. • Module 3 (crotonyl-CoA→acetyl-CoA + pyruvate): Identified native E. coli thioesterases (TesB, YciA) as methylsuccinyl-CoA hydrolases (double knockout reduced methylsuccinate ~27-fold). Discovered native Sdh oxidizes methylsuccinate to mesaconate; sdhA/B knockouts reduced mesaconate by 80–90%, while Sdh overexpression increased it ~7-fold. An acid-bypass using Y. pestis Ict enabled mesaconate activation and downstream conversion to pyruvate; HH422 produced non-labelled pyruvate from crotonate in [U-13C]glucose medium (with 2% LB), confirming module 3 activity in vivo.

The THETA cycle demonstrates that highly orthogonal, oxygen-insensitive CO₂-fixation pathways can be designed, constructed, and substantially optimized in vitro, and their modules can be functionally integrated into E. coli. Combining rational engineering with active-learning optimization rapidly navigated a vast parameter space to improve productivity by nearly two orders of magnitude. The findings address the central question of feasibility and performance of synthetic CO₂ fixation, showing competitive in vitro rates relative to the CBB cycle in extracts and other synthetic cycles. In vivo results highlight differences between cell-free and cellular contexts: bottlenecks and solutions diverged (e.g., a dihydroorotate-based Frd bypass was essential in vitro, whereas native or heterologous fumarate reduction sufficed in vivo). Native metabolic activities both hindered (thioesterase hydrolysis) and enabled (Sdh-mediated methylsuccinate oxidation) pathway function, leading to an effective acid-bypass strategy for module 3. The established modules and labelling/growth validations lay the foundation for assembling and evolving the full cycle in cells, advancing our understanding of constraints and opportunities in cellular carbon fixation.

This work introduces the THETA cycle, a modular, new-to-nature CO₂-fixation pathway converting CO₂ to acetyl-CoA, and demonstrates robust in vitro operation with significant productivity gains through rational and machine learning-guided optimization, along with flexible product outputs (acetyl-CoA, malonyl-CoA, glycolate). Three pathway modules were implemented in E. coli and validated via growth rescue and/or 13C labelling, uncovering both barriers and useful native activities that inform in vivo engineering strategies. Future efforts should focus on increasing flux (notably improving Mco activity), reducing undesired cross-talk (e.g., thioesterase activity), balancing cofactor/energy supplies, and integrating the complete cycle for autotrophic operation, potentially powered by external reducing equivalents (electricity, H₂, formate, methanol) and adaptive laboratory evolution to fine-tune regulatory and metabolic burdens.

Despite substantial in vitro improvements, productivity and stability remain limited compared with natural pathways. The Mco step is rate-limiting. In vivo, module 3 required an acid-bypass due to thioesterase-mediated hydrolysis of methylsuccinyl-CoA, and additional nutritional support (2% LB) was needed to detect pyruvate formation in HH422, indicating metabolic burden and integration challenges. Full complementation of acetyl-CoA auxotrophy by module 3 was not achieved, and complete in vivo closure of the THETA cycle remains future work. Autotrophic operation will require additional energy and reducing power and careful management of interactions with host metabolism.