Seven critical challenges in synthetic one-carbon assimilation and their potential solutions

Abstract

Synthetic C₁ assimilation holds the promise of facilitating carbon capture while mitigating greenhouse gas emissions, yet practical implementation in microbial hosts remains relatively limited. Despite substantial progress in pathway design and prototyping, most efforts stay at the proof-of-concept stage, with frequent failures observed even under in vitro conditions. This review identifies seven major barriers constraining the deployment of synthetic C₁ metabolism in microorganisms and proposes targeted strategies for overcoming these issues. A primary limitation is the low catalytic activity of carbon-fixing enzymes, particularly carboxylases, which restricts the overall pathway performance. In parallel, challenges in expressing multiple heterologous genes—especially those encoding metal-dependent or oxygen-sensitive enzymes—further hinder pathway functionality. At the systems level, synthetic C₁ pathways often exhibit poor flux distribution, limited integration with the host metabolism, accumulation of toxic intermediates, and disruptions in redox and energy balance. These factors collectively reduce biomass formation and compromise product yields in biotechnological setups. Overcoming these interconnected challenges is essential for moving synthetic C₁ assimilation beyond conceptual stages and enabling its application in scalable, efficient bioprocesses towards a circular bioeconomy.

In this review, we summarize the challenges in establishing synthetic pathways for assimilation of C₁ feedstocks in microbes and how to solve these issues towards a circular bioeconomy.

List of key abbreviations

ALE

Adaptive laboratory evolution

AOX

Oxygen-dependent alcohol oxidases

C₁

One-carbon (substrate/moiety)

CBB cycle

Calvin–Benson–Bassham cycle

CCM

Carbon-concentrating mechanism

CETCH cycle

Crotonyl-CoA/ethylmalonyl-CoA/hydroxybutyryl-CoA cycle

CoA

Coenzyme A

ECM

Enzyme cost minimization

EDEMP cycle

Entner-Doudoroff–Embden-Meyerhof-Parnas–pentose phosphate cycle

FalDH

Formaldehyde dehydrogenase

FDH

Formate dehydrogenase

GED cycle

Gnd–Entner–Doudoroff cycle

HTS

High-throughput screening

MDF

Maximum-minimum driving force

MeDH

Methanol dehydrogenase

NCRC

Noncanonical redox cofactor

PQQ

Pyrroloquinoline quinone

rGlyP

Reductive glycine pathway

RuBisCO

Ribulose-1,5-bisphosphate carboxylase/oxygenase

RuMP cycle

Ribulose monophosphate cycle

SACA pathway

Synthetic acetyl-CoA pathway

STC

Serine threonine cycle

TCA cycle

Tricarboxylic acid cycle

TRY

Titers, rates, and yields

TRYE

Titers, rates, yields, and emissions

XuMP cycle

Xylulose monophosphate cycle

Introduction

Climate change is one of the most pressing global challenges, with widespread impacts that span ecological, environmental, socio-political, and socio-economic domains (Abbass et al. 2022, Romanello et al. 2024). The continued dependence on fossil fuels and shifts in land use have driven a sustained rise in greenhouse gas emissions, particularly carbon dioxide (CO₂), methane (CH₄), and nitrogen dioxide (N₂O), which accumulate in the atmosphere and intensify global warming (Friedlingstein et al. 2023, Liu et al. 2024). Among these, CO₂ has played a predominant role, with atmospheric concentrations rising from 278 ppm in 1750 to over 427 ppm in 2025 (Lüthi et al. 2008, Filonchyk et al. 2024). If this trajectory persists, global temperatures are projected to increase by 3°C–5°C by 2100. Beyond its climatic impact, the prevailing linear economic model, which prioritizes resource extraction and waste generation (Polasky et al. 2019), poses a major sustainability challenge (Söderholm 2020, Horodecka 2024). Addressing these issues requires an urgent transition to a circular carbon bioeconomy (García and Galán 2022, Pavan et al. 2022). Capturing and repurposing carbon emissions offers a viable path towards reducing reliance on fossil fuels, while mitigating greenhouse gas emissions (Velenturf and Purnell 2021, Qiao et al. 2023, Awogbemi and Desai 2025, Timmis et al. 2025). A CO₂-based circular carbon economy holds the potential to support the sustainable production of chemicals, fuels, materials, pharmaceuticals, and even food (Gleizer et al. 2020, Lee et al. 2025).

Multiple technologies exist for converting CO₂ into value-added products. Chemical methods, e.g. nonreductive CO₂ valorization, have reached commercial implementation but remain confined to simple compounds, e.g. urea, carbonates, and polycarbonates (Liu et al. 2015, Hepburn et al. 2019, Yusuf et al. 2023). Other approaches, e.g. catalytic hydrogenation, plasma-assisted reactions, and electrochemical reduction, enable the conversion of CO₂ into multicarbon molecules but are hindered by challenges that include restricted product range, low selectivity, sensitivity to gas impurities, and the requirement for extreme reaction conditions (Kumaravel et al. 2020, Park et al. 2024). In contrast, biological systems function under mild conditions and exhibit high product specificity, yet they face distinct limitations, e.g. low current densities in direct electron transfer from electrodes to microorganisms or the energetic constraints of chemolithoautotrophic carbon fixation (Satanowski and Bar-Even 2020). Hybrid systems that integrate chemical and biological approaches offer a promising alternative, where CO₂ is first reduced to one-carbon (C₁) intermediates through chemical processes, which are subsequently converted into structurally complex molecules by microbial metabolism (Zhang et al. 2022, Hu et al. 2024, Liu et al. 2025). This strategy harnesses the strengths of both systems, improving specificity and energy efficiency while operating under ambient conditions (Onyeaka and Ekwebelem 2023, Baumschabl et al. 2024).

Microbial C₁ assimilation for biomanufacturing presents a viable alternative to conventional feedstocks, such as sugars, which are primarily sourced from agriculture and compete with food and feed industries (Keasling 2010, Segers et al. 2024, Orfali et al. 2025). This competition for essential resources can be alleviated by adopting second-generation feedstocks or utilizing CO₂ and reduced C₁ substrates derived from renewable energy sources (Cotton et al. 2020). C₁ compounds, including methanol, formate, and CO, can be synthesized through electrochemical reduction, photocatalysis, and heterogeneous catalysis, providing sustainable inputs for microbial bioproduction (Dürre and Eikmanns 2015, Alper and Yuksel Orhan 2017). However, most industrially relevant microorganisms do not naturally metabolize C₁ molecules as carbon sources. Although native C₁-trophic microbes can utilize these compounds, they are often challenging to engineer and have a narrow product spectrum compared to established microbial hosts, e.g. Escherichia coli and yeast. Advances in synthetic biology and genome engineering now enable the modification of industrially relevant microbes to metabolize C₁ substrates for bioproduction (Bae et al. 2022, O’Keeffe et al. 2025). Significant strides have been made in pathway engineering for CO₂, methanol, or formate assimilation; all the pathways that have been implemented in vivo are illustrated in Fig. 1. Heterotrophic microbial hosts have been engineered to express rather diverse C₁ assimilation routes, including the Calvin–Benson–Bassham (CBB) cycle in E. coli and Pichia pastoris, the ribulose monophosphate (RuMP) cycle in E. coli and Saccharomyces cerevisiae, the reductive glycine pathway (rGlyP) in E. coli, Pseudomonas putida, and Cupriavidus necator, and, more recently, the synthetic serine–threonine cycle (STC) in E. coli. Despite these promising developments, achieving full C₁ assimilation in heterotrophs remains highly challenging. A detailed description of these pathways and the obstacles associated with their implementation is provided in Tables 1 and 2. Only a fraction of engineering efforts has resulted in complete C₁ assimilation, and in most cases, pathway modules have been partially functional, requiring cosubstrates to support microbial growth. Notably, the majority of these studies have focused on E. coli, leaving many promising microbial hosts unexplored (Zhong et al. 2023). Additionally, in vitro research has identified several potential assimilation pathways that have yet to be validated in vivo (Guo et al. 2017, Danchin 2021). Similarly, naturally occurring assimilation routes in autotrophic bacteria, e.g. the 3-hydroxypropionate bicycle and the Wood–Ljungdahl pathway (Sanford and Woolston 2022), remain untested in heterotrophic organisms (Box 1). The metabolic utilization of other C₁ substrates, e.g. methane in nonmethanotrophic microbes, also remains largely unexamined (Bennett et al. 2021a, Jiang et al. 2021, Gregory et al. 2022, Tan et al. 2024, Yu et al. 2024).

Figure 1.

Natural and synthetic C₁ assimilation pathways engineered in vivo in heterotrophic organisms. Key enzymes of each pathway are highlighted in orange, dashed lines represent lumped reactions. Key abbreviations: RuMP cycle, ribulose monophosphate cycle; XuMP cycle, xylulose monophosphate cycle; EuMP, erythrulose monophosphate cycle, CBB cycle, Calvin–Benson–Bassham cycle; GED cycle, Gnd–Entner–Doudoroff cycle; rGly pathway, reductive glycine pathway; SACA pathway, synthetic acetyl-coenzyme A (CoA) pathway; THETA cycle, reductive tricarboxylic acid branch/4-hydroxybutyryl-CoA/ethylmalonyl-CoA/acetyl-CoA cycle; MeDH, methanol dehydrogenase; FalDH, formaldehyde dehydrogenase; FDH, formate dehydrogenase; F6P, fructose-6-phosphate; 6PG, 6-phosphogluconate; G3P, glyceraldehyde-3-phosphate; Ru5P, ribulose-5-phosphate; H6P, 3-hexulose-6-phosphate; THF, tetrahydrofolate; 2PG, 2-phosphoglycerate; 3-phosphoglycerate; 1,3BPG, 1,3-bisphosphoglycerate; PEP, phosphoenolpyruvate; RuBP, ribulose-1,5-bisphosphate; FBP, fructose-1,6-bisphosphate; Xu5P, xylulose-5-phosphate; DHA, dihydroxyacetone; DHAP, dihydroxyacetone phosphate; KDPG, 2-keto-3-deoxy-6-phosphogluconate; l-Eu1P, l-erythrulose 1-phosphate; d-Eu4P, d-erythrulose 4-phosphate; and E4P, d-erythrose 4-phosphate.

Table 1.

C₁ assimilation pathways fully^a engineered in vivo in heterotrophs.

Pathway	C-source	Product	Architecture	Host	Challengesb	Potential solutionsc	References
CBB cycle	CO2	3-Phospho-glycerate	Circular	E. coli	3 (branch node efflux)	(Unintended) hypermutagenesisd, ALE and reverse engineering	Gleizer et al. (2019), Ben-Nissan et al. (2023)
				P. pastoris	7 (low growth rates)	ALE and reverse engineering	Gassler et al. (2020, 2022)
RuMP cycle	Methanol	Glyceraldehyde-3-phosphate	Circular	E. coli	1, 3 (cycle acceptor regeneration; branch node efflux), 4, 7 (low growth rates)	In silico analysis, growth-coupled selection, rational metabolic engineering, (unintended) hypermutagenesisd, ALE, retroengineering, inducible systems, copy number variation	He et al. (2018), Chen et al. (2020), Keller et al. (2020, 2022), Reiter et al. (2024)
				S. cerevisiae	4	In vivo colocalization of formaldehyde-producing and assimilating enzymes in membrane-less organelles	Zhou et al. (2023)
rGlyP	Formate and CO2	Pyruvate	Linear	C. necator	3 (topological incompatibility), 5, 7 (availability of engineering tools)	Transcriptomics, ALE	Claassens et al. (2020), Dronsella et al. (2025)
	Methanol/formate and CO2			E. coli	3, 7 (overflow metabolism: acetate accumulation; low growth parameters especially with methanol)	ALE and reverse engineering, (unintended) hypermutagenesisd (implemented); increased tolerance to substrate, toxicity, use of AOX-MeDH (proposed)	Döring et al. (2018), Yishai et al. (2018), Kim et al. (2020, 2023)
	Formate and CO2			E. coli	6, 7	Expression fine-tuning, decrease in temperature (implemented), ALE and/or metabolic engineering to increase substrate assimilation and energy generation, optimizing cultivation conditions (proposed)	Bang and Lee (2018), Bang et al. (2020)
	Methanol/formate and CO2			P. putida	3 (topological incompatibility), 6 (NADH availability), 7	ALE and reverse engineering	Turlin et al. (2022)
STC	Formate	Acetyl-CoA	Circular	E. coli	3 (topological incompatibility), 5, 6 (NADH availability)	ALE and reverse engineering (partially implemented)	Yishai et al. (2017) Wenk et al. (2024)

^aSynthetic C₁ pathways are considered to be fully implemented when no additional carbon sources or additives (e.g. yeast extract) are included in the culture medium, and more than two doublings, consistent growth by the engineered host have been achieved.

^bChallenges are coded as follows: (1) limited activity of carbon-fixing enzymes, (2) poor heterologous gene expression, (3) pathway complexity and low efficiency, (4) formation of toxic intermediates, (5) complex and unknown regulation mechanisms, (6) cofactor imbalance, and (7) low biomass and target product yield (see Fig. 2). These challenges have been either reported by the authors or derived from the results described in the corresponding article.

^cSolutions that have been implemented unless noted (identified as proposed).

^dIn these examples, unintended hypermutagenesis was a consequence of natural evolution rather than a rational engineering strategy.

Table 2.

Characteristics of C₁ assimilation pathways partially^a engineered in vivo in heterotrophs and challenges and solutions during their implementation.

Pathway	C-source	Product	Architecture	Host	Challengesb	(Potential) Solutionsc	References
CBB cycle	CO2	3-Phospho-glycerate	Circular	Methylobacterium extorquens	3 (branch node efflux)	ALE and rational engineering (proposed)	Schada von Borzyskowski et al. (2018)
rGlyP	Formate and CO2	Pyruvate	Linear	S. cerevisiae	3 (topological incompatibility), 6 (low FDH activity)	ALE and reverse engineering	González de la Cruz et al. (2019) Bysani et al. (2024)
STC	Methanol Formate	Acetyl-CoA	Circular	P. putida	3 (cycle acceptor regeneration; branch node efflux), 5	Understanding regulation of the host (partially implemented), ALE (proposed)	Puiggené et al. (2025)
Modified serine cycle	Methanol	Acetyl-CoA	Circular	E. coli	1, 3 (cycle acceptor regeneration; branch node efflux, topological incompatibility)	None mentioned	Yu and Liao (2018)
				P. putida	3 (cycle acceptor regeneration; branch node efflux, topological incompatibility)	Understanding regulation of the host (partially implemented), ALE (proposed)	Puiggené et al. (2025)
Homoserine cycle	Methanol	Acetyl-CoA	Circular	E. coli	1, 3 (branch node efflux)	Enzyme engineering of assimilative reactions and branch points, ALE (proposed)	He et al. (2020)
				P. putida	1, 3 (branch node efflux)	Understanding regulation of the host (partially implemented), ALE, enzyme engineering of assimilative reactions (proposed)	Puiggené et al. (2025)
Enhanced STC	Methanol	Acetyl-CoA	Circular	P. putida	3 (cycle acceptor regeneration; branch node efflux)	Understanding regulation of the host (partially implemented), ALE (proposed)	Puiggené et al. (2025)
GED cycle	CO2	Pyruvate	Circular	E. coli	1, 3 (branch node efflux)	Enzyme discovery and rational engineering; formate dehydrogenase expression; ALE	Satanowski et al. (2020)
SACA	Formaldehyde	Acetyl-CoA	Linear	E. coli	1, 4	Enzyme discovery or rational engineering; increase formaldehyde tolerance	Lu et al. (2019)
EuMP cycle	Formaldehyde	Glyceraldehyde 3-phosphate	Circular	E. coli	3 (topological incompatibility), 7	Rational strain engineering; long-term ALE; enzyme discovery	Wu et al. (2023)
XuMP cycled (hybrid)	Methanol	Glyceraldehyde 3-phosphate and dihydroxyacetone	Circular	E. coli	1 (Energy inefficiency of O2-dependent alcohol dehydrogenase), 4	Use of prokaryotic NADH-dependent MeDHs	De Simone et al. (2020), Sun et al. (2022)
				S. cerevisiae	1, 4, 6	Optimization of peroxisomal reactions, introduction of a Mdh to provide NADH; introduction of a xylose metabolic pathway to promote formaldehyde assimilation; ALE	Zhan et al. (2023)
THETA	CO2	Acetyl-CoA	Circular	E. coli	3 (topological incompatibility), 5, 6	ALE; use of additional reducing power; metabolic proofreading	Luo et al. (2023)
FORCE pathways	Methanol Formate	2-Hydroxy-aldehydee	Linear	E. coli	1, 2, 3 (branch node efflux, topological incompatibility)	Use of cocultures; enzyme discovery and rational engineering; better activation of the substrate	Chou et al. (2021)

^aSynthetic C₁ pathways are considered to be partially implemented when additional carbon sources (e.g. yeast extract or glucose) are included in the culture medium, only some pathway modules have been introduced, or no more than two doublings by the engineered host have been achieved.

^bChallenges are coded as follows: (1) limited activity of carbon-fixing enzymes, (2) poor heterologous gene expression, (3) pathway complexity and low efficiency, (4) formation of toxic intermediates, (5) complex and unknown regulation mechanisms, (6) cofactor imbalance, and (7) low biomass and target product yield (see Fig. 2). These challenges have been either reported by the authors or derived from the results described in the corresponding article.

^cSolutions that have been implemented unless noted (identified as proposed).

^dBoth engineering efforts for the XuMP cycle are considered incomplete due to the requirement of yeast extract for E. coli and only two doublings reported for S. cerevisiae.

^e2-Hydroxyaldehydes and various elongated aldehydes, diols, carboxylic acids, and alcohols.

BOX 1.

Anaerobic C₁ assimilation: native pathways and engineering efforts

A key aspect to consider while engineering C₁ assimilation is the dependence of metabolic pathways on molecular oxygen (O₂). Most synthetic designs for C₁-trophic growth rely on aerobic conditions to enhance ATP generation through oxidative phosphorylation, using O₂ as the terminal electron acceptor. This configuration supports high energy yields but often leads to net CO₂ emissions. As a result, a delicate balance must be struck between generating sufficient energy for biomass and product formation and minimizing carbon losses as CO₂ during the bioprocess. In contrast, anaerobic conditions offer distinct benefits, particularly in large-scale reactors where eliminating aeration simplifies operation (Takors et al. 2018). During anaerobic fermentation, a portion of the carbon source functions as an electron acceptor to counteract excess reducing equivalents. This inherently links production to growth, as observed in acetate synthesis through acetogenesis (Weusthuis et al. 2011). Notably, acetogenesis—i.e. the anaerobic conversion of C₁ substrates into acetate via the Wood–Ljungdahl pathway—operates near the thermodynamic threshold for life (Frolov et al. 2023), thereby maximizing the energetic efficiency of C₁ compounds (Claassens et al. 2019b). However, the coupling of energy conservation to product formation imposes constraints on the structural diversity of compounds acetogens can produce. Genetic tools tailored for these organisms remain underdeveloped, and their manipulation is still far from streamlined (Bourgade et al. 2021, Flaiz and Sousa 2024). Transferring this metabolic framework into alternative hosts is also challenging due to limited understanding of energy conservation mechanisms, including maturation of ferredoxin-dependent enzymes. Despite the complexities of maintaining anaerobic conditions, industrial-scale processes using obligate anaerobes have been implemented commercially. One prominent case is the LanzaTech platform (https://lanzatech.com/), which captures CO₂ to generate bulk chemicals at scale. At present, this acetogen-driven system represents the most broadly deployed biotechnological solution for CO₂ conversion, underscoring the feasibility and scalability of anaerobic bioprocesses. Taking all these considerations together, acetogens—and, more broadly, anaerobic C₁ assimilation—offer promise for the sustainable production of commodity chemicals in a carbon-based bioeconomy (Poehlein et al. 2025), although their potential remains capped by energetic constraints that limit product scope (Bertsch and Müller 2015, Valgepea et al. 2017).

The ultimate goal of metabolic engineering for C₁ assimilation is to enable bioproduction from C₁ substrates once autotrophic growth is established in a selected microorganism. This strategy holds promise for the sustainable synthesis of biofuels, chemicals, and proteins (Bachleitner et al. 2023)—yet it remains technically demanding. Microbial protein production from methanol has drawn considerable interest, with the methylotrophic yeast P. pastoris serving as a widely studied host capable of naturally converting methanol into biomass (Meng et al. 2023). Despite this well-established proof of concept, the efficiency of microbial protein yields from C₁ compounds remains insufficient for large-scale deployment, primarily due to suboptimal methanol assimilation and slower growth rates relative to glucose-based systems. Similarly, formate-utilizing microbes have been investigated for the biosynthesis of platform chemicals (Kim et al. 2023, Tian et al. 2023), yet their production titers lag behind those achieved with sugar feedstocks (Konzock and Nielsen 2024), in part due to imbalances in redox and energy fluxes (Tables 1 and 2). More broadly, bioprocesses relying on C₁ substrates continue to underperform compared to traditional sugar-based fermentation, with multiple unresolved bottlenecks limiting their efficiency. Current literature on systems-level bottleneck analyses of synthetic C₁ metabolism is largely restricted to isolated case studies. While several reviews underscored recent progress in C₁-based bioproduction, critical challenges persist in optimizing the synthetic assimilation of these feedstocks. Addressing these limitations is imperative as the field advances, opening new avenues for overcoming these constraints.

This review identifies seven critical challenges that have consistently hindered progress in the field (Fig. 2) and discusses potential solutions to address each of these obstacles. Most synthetic C₁ assimilation pathways rely on carbon-fixing enzymes, either native or heterologous; however, these enzymes, particularly carboxylases, inherently exhibit low catalytic efficiency. Additionally, constructing functional synthetic pathways requires the coordinated overexpression of multiple genes in the host organism, often leading to poor heterologous gene expression, particularly for metal-dependent or oxygen-sensitive enzymes. Even when enzyme levels are finely tuned, the complexity of the entire metabolic cascade increases, demanding precise control over flux distribution to enhance pathway efficiency. Beyond enzyme-specific limitations, synthetic pathways frequently suffer from poor integration into the host metabolism, often resulting in toxic intermediate accumulation and impaired metabolic connectivity hampered by the native regulation of the host metabolism. Furthermore, cofactor imbalances disrupt redox and energy homeostasis, constraining biomass formation and limiting yields from these carbon-poor substrates (Fig. 2). This review provides our views into overcoming these bottlenecks, offering a roadmap to drive the field towards more efficient and scalable implementation of synthetic C₁ metabolism.

Figure 2.

Challenges frequently encountered when engineering synthetic C₁ assimilation in heterotrophic microbial hosts. The challenges are numbered according to the same nomenclature used in the tables.

Challenges

C₁-processing enzymes, especially carboxylases, have an inherently low activity

When examining potential barriers to engineering C₁ compound assimilation, one of the earliest and most persistent limitations arises from the thermodynamics of carboxylation and carboxyl reduction reactions. Across existing carbon-fixation pathways, these steps impose substantial energetic burdens. In most cases, ATP consumption is directly or indirectly associated with these transformations (Bar-Even et al. 2012). The endergonic nature of carboxylation stems from the requirement to reduce carbon in CO₂, which exists in its most oxidized state (oxidation state γ = 4), considerable energy input. As a result, the assimilation of more reduced C₁ molecules—e.g. formate, formaldehyde, and methanol—offers improved energetic efficiency under both oxic and anoxic conditions (Amoah et al. 2019, Claassens et al. 2019a, Guo et al. 2022a). Enzymes can circumvent thermodynamic constraints by lowering activation energy or embedding these reactions within exergonic metabolic sequences, thereby favourably altering the equilibrium. Nature employs several biochemical tactics for C₁ assimilation, including high-energy intermediates, metal coordination with C₁ substrates, input from external electron donors, and the synthesis of thermodynamically favourable products (Bierbaumer et al. 2023, Mohr et al. 2025). Although these strategies support carbon fixation, carboxylation often remains the most energy-demanding and rate-limiting step, reinforcing the importance of choosing an effective carboxylase in synthetic CO₂ assimilation routes (Schulz-Mirbach et al. 2024a). Among the enzymes available for this purpose, phosphoenolpyruvate carboxylase stands out due to its independence from soluble redox cofactors [NAD(P)H] or ATP and its oxygen insensitivity (Fig. 3), paired to its ability to irreversibly carboxylate phosphoenolpyruvate at atmospheric CO₂ levels with high catalytic efficiency (Kai et al. 2003). RuBisCO (ribulose-1,5-bisphosphate carboxylase/oxygenase) catalyses the primary carboxylation step in the CBB cycle by incorporating CO₂ into ribulose-1,5-bisphosphate, forming two molecules of 3-phosphoglycerate (Fig. 3). This reaction supports the majority of global carbon assimilation and connects atmospheric CO₂ to core metabolic pathways in plants, algae, and numerous bacteria (Prywes et al. 2023).

Figure 3.

Overview of the reactions catalysed by carboxylases, carboligases, and methanol dehydrogenases. The catalytic efficiency of these transformations often limits synthetic C₁ assimilation efforts. The reactions of selected carboligase enzymes—Hps, hexulose-6-phosphate synthase, and serine hydroxymethyltransferase, key enzymes of the RuMP cycle and the rGlyP, respectively—as well as carboxylase enzymes—Ppc, phosphoenolpyruvate carboxylase and RuBisCO, central to the CBB and serine cycles, respectively—are explicitly shown as examples. The main steps of C₁ dissimilation are also depicted, with emphasis on the different types of methanol dehydrogenases and their associated free-energy values (Whitaker et al. 2015). AOX, oxygen-dependent alcohol dehydrogenase.

The assimilation of reduced C₁ compounds predominantly depends on carboligases rather than carboxylases. In this context, carboligase refers broadly to enzymes that catalyse carbon–carbon (C–C) bond formation by linking two carbon-containing substrates. These transformations typically proceed without net incorporation or release of CO₂, instead enabling condensation reactions (Fig. 3)—often involving aldehyde or keto groups—to build extended carbon backbones (Cheon et al. 2024, Dobiašová et al. 2024). Native C₁-assimilation cycles include carboligases such as 3-hexulose 6-phosphate synthase [a lyase involved in formaldehyde assimilation within the RuMP cycle, converting ribulose 5-phosphate into 3-hexulose-6-phosphate (He et al. 2018, Antoniewicz 2019)] and serine hydroxymethyltransferase (encoded by glyA, which catalyses the reversible interconversion of serine and glycine). The latter reaction, relying on tetrahydrofolate as a one-carbon carrier (Fig. 3), plays a central role in the serine cycle (Florio et al. 2011).

Suboptimal enzyme characteristics are not limited to carboxylases and carboligases. In methanol metabolism, for instance, methanol dehydrogenases (MeDH) catalyse the oxidation of methanol to formaldehyde (Gan et al. 2023). These enzymes fall into three categories based on their preferred electron acceptor: oxygen-dependent alcohol oxidases (AOX), pyrroloquinoline quinone (PQQ)-dependent MeDH, and NAD⁺-dependent MeDH (Fig. 3). Although NAD⁺-dependent MeDHs are relatively easy to express and offer the highest energetic efficiency, their catalytic performance is limited by slow kinetics and unfavourable thermodynamics at ambient temperature (Fig. 3). AOXs require elevated oxygen levels, lack energy conservation, and produce excess heat and toxic hydrogen peroxide. PQQ-dependent MeDHs offer a compromise between energy efficiency and catalytic performance, supporting faster growth. However, functional expression of PQQ-dependent MeDHs requires at least 11 genes, including those for PQQ biosynthesis, which are only naturally present in a limited group of microorganisms, e.g. Acinetobacter and Pseudomonas (Chistoserdova et al. 2003, Yang et al. 2010, Krüsemann et al. 2023).

Further limitations in designing C₁ assimilation routes, especially during in vivo implementation, include off-target enzyme activity, structural instability, imbalanced cofactor environment, enzyme inactivation by reactive substrates or intermediates, and loss of key substrates or cofactors due to competing metabolic routes (Lu et al. 2019, Pardo et al. 2022, Bierbaumer et al. 2023)—challenges that extend to other metabolic engineering endeavours. These issues are often not evident in vitro and typically emerge only in the context of the cellular environment.

Poor heterologous gene expression, metal-dependence, and oxygen sensitivity of key enzymes

Heterologous gene expression remains a major bottleneck for synthetic C₁ metabolism, particularly for genes encoding enzymes that are metal-dependent or oxygen-sensitive. Genes encoding metal-containing enzymes (e.g. ferredoxin-dependent proteins), widespread in anaerobic metabolism (Box 1), are especially challenging to express in standard microbial hosts due to the intricate and often poorly characterized maturation mechanisms and their links to energy conservation. Cupriavidus necator employs efficient, metal-dependent formate dehydrogenases (FDHs), but replicating their functionality in other organisms has proven difficult. To circumvent these constraints, metal-free FDHs have been implemented in E. coli, as they are more readily expressed; however, their catalytic performance is inferior (Calzadiaz-Ramírez and Meyer 2021). The expression of genes encoding metal cluster-containing enzymes involved in synthetic C₁ assimilation, e.g. those in the dicarboxylate/4-hydroxybutyrate cycle (Berg et al. 2007), is further limited by incompatibilities with intracellular environments different to those of native hosts (Huber et al. 2008). In addition, oxygen sensitivity significantly impairs the stability and functionality of several key enzymes, e.g. carbon monoxide dehydrogenase (CODH) from the Wood–Ljungdahl pathway (Box 1) and pyruvate formate lyase (Pfl), posing further complications for their use in heterologous systems (Bar-Even et al. 2013).

Pathway complexity and efficiency

The complexity of C₁-assimilating pathways is shaped by both the number of enzymatic conversions and the overall pathway configuration. Extended reaction sequences obviously pose greater challenges than shorter ones, and linear designs are typically easier to implement than cyclic or autocatalytic variants, which require careful tuning of flux distributions to maintain metabolic balance (Boyle and Silver 2012, Nikel et al. 2015, Schulz-Mirbach et al. 2024b). Synthetic and natural C₁ assimilation routes are commonly evaluated based on four key parameters: thermodynamic feasibility, kinetic performance, energy requirements, and structural integration. Pathway topology refers to the organization and intricacy of the metabolic route, while topological compatibility considers how effectively the introduced pathway integrates with the native metabolism of the selected microbial host (Bar-Even et al. 2010, Pandit et al. 2017, Han and Styczynski 2024). In heterotrophic hosts, a significant pathway overlap with endogenous metabolism presents both opportunities and risks for engineering. While it may limit the burden of heterologous gene expression, this overlap can also trigger deleterious interactions through cross-regulation or metabolic interference (Sánchez-Pascuala et al. 2019). Reducing such conflicts can be achieved by employing orthogonal designs that lead to a minimal overlap with the metabolic network of the host (Sánchez-Pascuala et al. 2017, Chou et al. 2021, Orsi et al. 2022).

Most native and synthetic C₁-assimilation pathways depend on autocatalytic cycles, e.g. RuMP cycle, natural and modified forms of the serine cycle (i.e. modified serine cycle, STC, and homoserine cycle), the CBB cycle, Gnd–Entner–Doudoroff cycle, and the malonyl-CoA–oxaloacetate–glyoxylate pathway (Bar-Even et al. 2010, Claassens 2017, François et al. 2019). Autocatalytic cycles are defined as metabolic loops in which the final product also serves as a substrate for the next iteration of the metabolic sequence. While their topology offers advantages for network efficiency, these cycles are notoriously difficult to engineer in surrogate hosts due to inherent complexity (Antonovsky et al. 2016, Barenholz et al. 2017). Natural C₁-assimilation cycles, such as the CBB, RuMP, and STC, are generally well-suited to integration within native metabolic networks and typically demand expression of only a limited set of heterologous genes. However, because of the high overlap with the extant central metabolism, implementation in surrogate hosts has proven more challenging than with linear or orthogonal designs (e.g. the rGlyP), largely due to uncontrolled diversion of key intermediates (Bang et al. 2021, Claassens et al. 2022). Hence, designing synthetic pathways for a given microbial host demands a detailed understanding of its central carbon flux. Identifying overlaps or mismatches in metabolic directionality is essential to avoid bottlenecks that could impair pathway function or, more critically, reduce host fitness when the cycle is active (Bar-Even 2016, Erb et al. 2017, Ko et al. 2020). Examples include reversal of the TCA cycle components, partly occurring in the serine cycle (Yu and Liao 2018), or the rewiring of the glycine cleavage system to implement the rGlyP in E. coli, C. necator, or P. putida (Bang and Lee 2018, Guo et al. 2018, Claassens et al. 2020, Kim et al. 2020, Turlin et al. 2022, Bruinsma et al. 2023).

Formation of toxic intermediates and dead-end products

Reduced C₁ substrates are toxic for most bacterial species (Lv et al. 2023). Formic acid, a weak acid, disrupts the proton motive force by acidifying the cytoplasm and inhibits cytochrome c oxidases (Kirkpatrick et al. 2001). Methanol toxicity is linked to its oxidation into formaldehyde, which alters membrane fluidity and function (Zhan et al. 2021). Formaldehyde, in turn, is highly reactive with proteins and nucleic acids, interfering with DNA replication, transcription, translation, and protein integrity, ultimately resulting in limited fitness and cell death (Chen et al. 2020, Jia et al. 2024a). These challenges have driven interest in understanding C₁ metabolism in industrially relevant bacterial hosts, e.g. E. coli, P. putida, and C. glutamicum (Zaldivar and Ingram 1999, Wang et al. 2020a, Turlin et al. 2023). Strains equipped with highly active formaldehyde and FDHs, e.g. P. putida, have a comparatively high tolerance to formaldehyde and formate (Roca et al. 2008, Wang et al. 2020a, Turlin et al. 2023). In contrast, elevated methanol oxidation rates have been associated with reduced methanol tolerance, due to intracellular formaldehyde formation (Goldberg and Mateles 1975, Adroer et al. 1990, Turlin et al. 2023). Therefore, harnessing C₁ assimilation as a detoxification mechanism remains a promising direction for future research.

Several C₁ assimilation pathways use methanol as a substrate, e.g. the RuMP and xylulose monophosphate (XuMP) cycles, whereas pathways based on formate or CO₂ can incorporate methanol to generate supplementary reducing power [e.g. the rGlyP (Kim et al. 2020), STC, and potentially the CBB cycle (Federici et al. 2023)]. Maintaining favourable growth conditions requires high formaldehyde production fluxes, which in turn impose significant toxicity on the engineered methylotroph. However, efficient C₁ assimilation or conversion can mitigate this toxicity. Adding to the complexity, the balance between assimilation routes and native detoxification mechanisms is a persistent obstacle for synthetic methylotrophy (Whitaker et al. 2015, Chen et al. 2020). Removing detoxification enzymes to exploit C₁ toxicity as a metabolic driver while avoiding substrate competition is an important strategy for overcoming this limitation (Yishai et al. 2016, Tuyishime et al. 2018, Bennett et al. 2021b, Jia et al. 2024b).

Limited metabolic integration of the synthetic pathway(s) and regulation of the host metabolism

In a broad sense, traditional strain engineering has centered on mapping metabolic reactions and modifying stoichiometry to improve specific outputs (Biz et al. 2019, Choi et al. 2019). However, enabling microbial utilization of unconventional substrates requires a system-level understanding of the physiological responses in the surrogate host that extends beyond the mere insertion of heterologous genes (Gonzalez et al. 2018, Volke and Nikel 2018). While genetic constructs for novel pathways can be successfully introduced, rational engineering often falls short in achieving the fine-scale genetic adjustments needed to balance fluxes across native and synthetic metabolic routes (Wenk et al. 2024). This circumstance highlights the importance of integrating systems-level analysis with precise genetic modifications to effectively tune engineered metabolic functions (Fernández-Cabezón et al. 2022). The inherently low catalytic efficiency of enzymes used in C₁ assimilation imposes a disproportionately large proteome allocation to favour carbon fluxes compatible with sustaining microbial growth. Enzymes such as RuBisCO, along with methane and MeDH, constitute a considerable share of the proteome in both natural and engineered methylotrophs and methanotrophs. This proteome investment is accompanied by extensive metabolic rearrangements, including reversed fluxes, altered metabolite pool sizes, and shifts in redox balance. These systemic changes impose high demands on the host, often requiring specialization in native C₁-assimilating organisms or substantial proteome reallocation in facultative species depending on the carbon source (Jahn et al. 2021, Pavan et al. 2022). For synthetic C₁-trophic strains, adopting a facultative phenotype that permits rapid growth would provide a more flexible platform for subsequent engineering efforts. This feature has been largely overlooked in current C₁-assimilation designs, leading to constructs that perform only under narrowly defined conditions (Chen et al. 2020). Paired to this challenge, a persistent bottleneck in deploying nontraditional feedstocks, e.g. C₁ compounds, is the limited responsiveness of regulatory circuits in platform organisms, which do not process these substrates in the same manner as conventional carbon sources.

Cofactor imbalances (redox and energy) during synthetic C₁ assimilation

Selecting an appropriate source of energy and electrons is fundamental for enabling functional rewired metabolism in synthetic C₁-fixation pathways. Detailed assessments of electron donors for C₁ assimilation identify formate and methanol as leading candidates, given their relatively straightforward derivation from the electrochemical reduction of CO₂ (Abel and Clark 2021). Formate offers advantages in terms of water solubility and lower toxicity compared to formaldehyde, the oxidation product of methanol. However, formate delivers only two electrons per carbon, whereas methanol provides six (Cotton et al. 2020). When assimilating these C₁ compounds, the use of a single substrate for both carbon and energy can simplify the overall metabolic design but it introduces the challenge of balancing carbon incorporation with dissimilation for energy generation. The limited pool of electrons from C₁ substrates must be allocated between cellular growth and, eventually, product synthesis, complicating the distribution of reducing equivalents across multiple NAD(P)⁺/NAD(P)H-dependent reactions.

Hydrogen is a cost-effective, nontoxic, and highly reduced alternative for supplying energy during carbon fixation, as exemplified by acetogens and Knallgas bacteria (e.g. C. necator). Nevertheless, its industrial application is constrained by safety risks, particularly its explosiveness in the presence of oxygen (Claassens et al. 2018). An alternative explored in academic settings is synthetic phototrophy (Tong et al. 2025). Rather than reconstructing complete photosystems to extract electrons from water, rhodopsins—enzymes that generate a proton gradient using light—offer a minimalistic solution to support prototrophic growth (Martinez et al. 2007, Tu et al. 2023). While unlikely to serve as the sole energy source, rhodopsins may enhance ATP production when paired with an external electron donor (Toya et al. 2022). At larger scales, however, the low efficiency of biological light capture and challenges in bioreactor illumination limit the feasibility of this strategy.

Limitations of biomass and product yields on carbon-poor substrates

Genome-scale metabolic models and in silico reaction predictions often indicate promising biomass yields for engineered C₁-trophic organisms; however, experimental outcomes frequently fail to meet these projections (Zhu and Chen 2022). Bioproduction from C₁ substrates poses additional hurdles, as substrate conversion must support both cellular growth and efficient flux distribution toward target products. Managing this balance between biomass formation and product synthesis requires fine-tuned control over resource allocation, particularly given the distinct flux patterns characterizing C₁-based systems in contrast to conventional heterotrophic modes. Engineered assimilation pathways have shown that high biomass densities are achievable through adaptive laboratory evolution (ALE). For instance, implementing the RuMP cycle in engineered E. coli strains has enabled optical densities approaching 100 units in fed-batch bioreactors; these strains also have the capacity for efficient product formation, with lactate titers reaching ca. 3 mM (Kim et al. 2023, Reiter et al. 2024). Despite these surely notable achievements, substantial improvements are still needed in titers, yields, and volumetric productivity.

Approaches, tools, and potential solutions

Established and emerging synthetic biology tools and metabolic engineering approaches and tools to facilitate the synthetic C₁ assimilation efforts hindered by the challenges previously discussed are outlined in the next section (Fig. 4).

Figure 4.

Potential solutions to the challenges listed in Fig. 2, presented as a roadmap originating from the selection of a pathway, carbon, and energy source of choice. The sequence begins with in silico analyses and the implementation of computational tools, followed by in vitro prototyping and characterization of functional parts when necessary, and extends to a suite of in vivo tools, with an emphasis on metabolic engineering strategies. CNV, copy number variation.

In silico analysis of pathway suitability

In silico modelling tools, developed to assess pathway engineering across diverse criteria, serve as an effective starting point to tackle the inherent complexity of designing C₁-assimilating pathways (Antonovsky et al. 2016, Satanowski et al. 2020, Keller et al. 2022). Flux balance analysis, for example, can be used to evaluate the compatibility of a candidate pathway with the host metabolism while probing its energetic plausibility (Orth et al. 2010, Khana et al. 2022). To deepen this assessment, thermodynamic and kinetic modelling can be integrated prior to experimental validation. In silico frameworks such as max–min driving force (MDF) and enzyme cost minimization (ECM) support this stage by characterizing the thermodynamic profile under physiological conditions and estimating protein demands based on pathway kinetics, respectively (Noor et al. 2014, 2016, Hädicke et al. 2018, Schulz-Mirbach et al. 2024a). MDF quantifies the lowest thermodynamic driving force across a pathway, incorporating parameters such as metabolite and cofactor concentrations, pH, and ionic strength to reflect physiological relevance. Higher MDF values suggest a more favourable driving force distribution, which can support elevated fluxes (Noor et al. 2014). In parallel, pathway kinetics inform protein investment by determining the minimal enzyme quantity required to sustain maximum flux. The total cost metric of ECM captures this relationship, as faster catalytic rates reduce the metabolic burden of maintaining flux (Noor et al. 2016). Both MDF and ECM simulations are supported by user-friendly platforms such as eQuilibrator and its API (Beber et al. 2022), even in cases where reaction thermodynamic or kinetic parameters are unknown. Additional bioinformatic tools can be employed to investigate physiological parameters after implementing the pathway of interest, e.g. accurate determination of growth rates (Wirth et al. 2023). This experimental data can be used to iteratively refine and validate in silico models, enabling closer alignment between predicted and observed pathway performance.

Selecting the right carbon and energy source to support C₁-trophy

Several parameters must be evaluated when selecting a CO₂-derived molecule as the primary carbon source for a given bioprocess, including its oxidation state, thermodynamic feasibility associated with its assimilation, and compatibility with cosubstrates. Methanol, being more reduced than formate, demands fewer reducing equivalents for assimilation, influencing the availability of NADH or NADPH for other cellular functions—such as bioproduction (Löwe and Kremling 2021). To overcome thermodynamically unfavourable steps, various strategies are employed, including the use of high-energy substrates, CO₂ activation via metal ion or ATP coordination, supplementation with external electron donors (e.g. NADPH or ferredoxins), and the formation of energetically favourable products (Bierbaumer et al. 2023). Microorganisms that depend solely on a single carbon and energy source (e.g. methanol or formate) often grow slowly, which can be mitigated by supplying auxiliary energy sources (e.g. hydrogen or phosphite). For example, in engineered P. putida, faster growth on formate was achieved by using acetate as an additional energy source while blocking its assimilation through the deletion of aceA (Turlin et al. 2022). Therefore, careful consideration of these factors is essential for optimizing carbon source selection and enhancing overall bioprocess efficiency.

Growth-coupled selection and modular engineering to tackle stepwise C₁-trophy

To establish a C₁-trophic metabolic regime in a surrogate host, stepwise coupling of C₁ assimilation to microbial growth through auxotrophies is a central requirement. This can be implemented via growth-coupled selection (Wenk et al. 2018, Schneider et al. 2021, Claassens et al. 2022, Orsi et al. 2025), which involves constructing mutant auxotrophic strains with targeted disruptions in the intracellular metabolic network that impair growth (Cros et al. 2022). Restoration of essential biomass precursor synthesis in these strains is achieved either by supplying external nutrients or by heterologous expression of specific metabolic modules—ranging from individual enzymes to full pathways—to repair the disrupted network. In cases where the latter strategy is employed, growth rate and yield serve as indirect measures of the flux potential of the introduced modules (Orsi et al. 2021, 2022). The conceptual basis for coupling growth with formate or methanol assimilation has been previously outlined (Claassens et al. 2019b), including modular configurations for pathways such as the rGlyP (Claassens et al. 2019b, 2022). Over the past years, this strategy has been applied to implement the CBB cycle in E. coli (Antonovsky et al. 2016, Gleizer et al. 2019), the rGlyP in E. coli (Yishai et al. 2017, 2018, Kim et al. 2020, Bang et al. 2021), C. necator (Claassens et al. 2020, 2022), P. putida (Turlin et al. 2022, Bruinsma et al. 2023), and S. cerevisiae (González de la Cruz et al. 2019, Bysani et al. 2024), the RuMP cycle in E. coli (Chen et al. 2020, He et al. 2020, Keller et al. 2020, 2022), and the STC in E. coli (Wenk et al. 2024). These examples illustrate the robustness of growth coupling strategies, expected to remain key for implementing synthetic C₁-assimilation routes in surrogate microbial hosts (von Kamp and Klamt 2017, Alter and Ebert 2019).

Enzyme engineering as a framework for establishing new C₁-processing reactions

Enzyme engineering in silico

In many instances, the core enzymes involved in C₁ assimilation, including carboxylases, display limited catalytic efficiency, prompting efforts to identify superior variants—whether naturally occurring or engineered (Pan and Kortemme 2021, Listov et al. 2024). Resources such as BRENDA, an extensive enzyme database detailing functions, substrates, and properties across diverse organisms (Chang et al. 2021), and homology-based tools such as BLAST (Camacho et al. 2023), continue to be instrumental in this type of analysis. EnzymeMiner, a webserver designed to streamline candidate selection by ranking sequences based on catalytic potential and soluble expression in E. coli, further accelerates the process by reducing the time required for data analysis and prioritization (Hon et al. 2020). For enzymes that remain challenging to express, recent developments in protein engineering have enabled substantial improvements in both stability and performance (Qing et al. 2022, Goverde et al. 2024). Among these tools is ProteinMPNN, a deep neural network that facilitates de novo sequence generation from a given structure (Dauparas et al. 2022). In contrast to resource-intensive physics-based platforms, e.g. Rosetta (Wang et al. 2024), ProteinMPNN offers a significantly faster alternative, broadening access to custom protein design. Sequences designed using ProteinMPNN have already shown enhanced expression, stability, and function when compared to their original forms (Sumida et al. 2024). While de novo protein design is not yet broadly adopted for creating entirely new enzymatic activities, recent progress (rooted on deep learning) suggests its strong potential for developing efficient biocatalysts to support C₁ assimilation pathways (Lovelock et al. 2022, Watson et al. 2023, Chen et al. 2024, Kortemme 2024).

Enzyme engineering in vitro

Innovative directions in enzyme engineering, encompassing the systematic exploration of Nature’s catalytic repertoire to design synthetic biocatalysts, are being explored to tackle the multifaceted barriers associated with C₁ assimilation. Efforts to overcome thermodynamic limitations in these systems—e.g. circumventing energetically unfavourable steps through high-energy intermediates, activating CO₂, or introducing external sources of reducing power—are evaluated in model microbial hosts to improve the uptake of C₁ feedstocks (Aleku et al. 2021, Bierbaumer et al. 2023, Gan et al. 2023). Targeting enzymes that catalyse individual reactions provides a window into developing alternative metabolic routes based on C₁ substrates. Whether extensively studied (e.g. RuBisCO and phosphoenolpyruvate carboxylase) or less characterized (e.g. lanthanide-dependent MeDHs), our current understanding of enzyme mechanisms is still shaped by a narrow subset of experimentally validated homologs within large protein families (Erb et al. 2017, Holliday et al. 2020, Ribeiro et al. 2023). The expansive protein sequence landscape offers opportunities to enhance catalytic traits (e.g. kinetics and stability) or even design unprecedented reactivities. For instance, analysis of multisequence alignments across medium-chain reductases, such as propionyl-CoA synthase and acrylyl-CoA reductase, enabled the engineering of novel carboxylases with productivities on par with RuBisCO (Bernhardsgrütter et al. 2019).

Expanding this search into the evolutionary past, paleoenzymes reconstructed from ancient DNA or inferred through ancestral sequence reconstruction offer a complementary route to discover biocatalysts with beneficial properties (Mascotti 2022, Klapper et al. 2023). Such enzymes, shaped by different environmental pressures, can reveal catalytic features absent in modern proteins (Gao et al. 2023, Prakinee et al. 2024). Though challenges persist due to limited functional annotations and the labour intensity of experimental validation, computational tools and bioinformatics now play a central role. Machine learning is emerging as a powerful means to navigate the sequence-function landscape, guiding mutagenesis, improving annotations, and enabling the modular reengineering of enzyme function (Markus et al. 2023, Wang et al. 2025). A notable example includes a machine learning–assisted design of synthetic glycolyl-CoA carboxylase, which resulted in a 2-fold increase in carboxylation activity and a 50% reduction in ATP consumption (Marchal et al. 2023).

The emergence of deep learning has enabled the design of synthetic enzymes that catalyse reactions not found in biology. A single structural framework has been reconfigured to support entirely new catalytic activities, as illustrated by the transformation of benzaldehyde lyase into formolase and glycolaldehyde synthase. These enzymes establish links between previously disconnected metabolic routes—e.g. the formolase and synthetic acetyl-CoA (SACA) pathways—supporting novel configurations for C₁ conversion (Siegel et al. 2015, Lu et al. 2019). While formolase mediates the condensation of formaldehyde into dihydroxyacetone, glycolaldehyde synthase facilitates the formation of glycolaldehyde. Additional strategies involve constructing multienzyme complexes that exploit substrate channeling. By restricting intermediates within confined microenvironments, such assemblies enhance catalytic throughput and suppress unwanted side reactions (Wheeldon et al. 2016). Directed substrate flow ensures selective biotransformations and boosts local reactant concentrations (Volke and Nikel 2018), improving thermodynamic favourability in the desired direction. The physical proximity of active sites further prevents the accumulation of harmful intermediates by enforcing temporal and spatial coupling of the cascade. Natural examples underscore this principle, including CODH/acetyl-CoA synthase and formyl-methanofuran dehydrogenase (Bar-Even et al. 2012). This concept has also been extended to engineered variants of MeDH, highlighting its relevance in synthetic systems (Price et al. 2016, Fan et al. 2018).

Protein engineering continues to address persistent limitations that affect biocatalysis—e.g. side reactions, inhibitory intermediates, and cofactor constraints. As an example, a MeDH from Bacillus stearothermophilus DSM 2334 was engineered to improve catalytic efficiency by 20-fold and switch cofactor specificity from NAD⁺ to NADP⁺ (Yang et al. 2024). To refine pathway output, promiscuous enzymes can be exchanged for more selective counterparts or tailored to enhance substrate discrimination. Paired with high-throughput platforms capable of screening vast variant libraries (Orsi et al. 2024a), these developments are closing the gap between theoretical designs and laboratory realization—paving the way for novel synthetic C₁ metabolism.

Enzyme engineering in vivo

In vivo enzyme engineering can optimize individual enzymatic steps or entire metabolic pathways directly within living organisms. By leveraging in vivo genetic tools to generate enzyme variants, this method accelerates the testing phase of the Design–Build–Test–Learn cycle and facilitates the selection of improved variants by coupling enzymatic activity to growth in a high-throughput manner (i.e. growth-coupled selection; Orsi et al. 2024b). By maintaining selective pressure, microbial strains can evolve to enhance flux capacity through the enzyme(s) of interest. The most efficient enzyme variants are then identified based on their reaction rates, which manifest as differences in the growth rates of the selection strains (Wenk et al. 2018, Orsi et al. 2021). Hence, developing in vivo evolution techniques facilitates the engineering of metabolic pathways, as mutagenesis occurs simultaneously with the selection of desired phenotypic traits. Processes that would naturally take extensive evolutionary timeframes can be achieved within laboratory and human timescales (Packer and Liu 2015) while accounting for specific boundary conditions, e.g. intracellular environment of the surrogate host (Danchin and Nikel 2019).

Hypermutator methods further enhance the mutation rate of target genes. In vivo hypermutators are capable of increasing the natural mutation rate, typically between 10^–10 and 10^–9 per cell per generation to as high as 10^–4 within the target gene(s) (Imhof and Schlotterer 2001, Molina et al. 2022), accelerating the adjustment of metabolic network fluxes to improve growth (Sánchez-Pascuala et al. 2019). Various techniques have been developed to elevate the mutation rate either in specific genes or across the entire genome, each leveraging distinct molecular mechanisms. These include polymerases, recombinases, CRISPR–Cas complexes, oligonucleotides, and base editors, among others (Zimmermann et al. 2024). Such hypermutating techniques have become powerful tools for in vivo evolution, enabling rapid optimization of target traits (Fernández-Cabezón et al. 2021, Molina et al. 2022). Nevertheless, directed in vivo hypermutagenesis has not been tested to engineer metabolic pathways for the assimilation of C₁-compounds. Recent evolution experiments aimed at enabling growth on C₁ compounds have often been prolonged due to factors, e.g. population size, slow growth rates, and the number of generations required to achieve improved phenotypes. Several studies reported the emergence of hypermutator phenotypes, characterized by mutations in endogenous mismatch–repair systems, which may accelerate metabolic rearrangements. Such hypermutator strains have been reported during the implementation of the rGlyP and the CBB and the RuMP cycles in engineered E. coli (Antonovsky et al. 2016, Döring et al. 2018, Nieh et al. 2024). Combining large enzyme libraries with directed in vivo mutagenesis presents a promising strategy in this direction. By applying ALE to these hypermutator strains, it would be possible to further enhance target reaction rates and accelerate synthetic C₁ metabolism. This integrated approach leverages the natural evolutionary processes to optimize metabolic pathways efficiently, facilitating the development of robust microbial strains capable of utilizing C₁ compounds for biotechnological applications.

High-throughput screening methods and biofoundries to accelerate pathway implementation

When growth-coupled selection is not applicable, high-throughput screening (HTS) offers a substantial advantage over conventional low-throughput selection formats, with the possibility of exponentially expanding experimental capabilities in the laboratory (Zeng et al. 2020, Bozkurt et al. 2025). HTS is particularly effective for rapid and cost-efficient exploration of enzyme variant libraries. The use of liquid handling systems in microtiter plate-based assays has accelerated discovery efforts by supporting robust design-of-experiment strategies, while maintaining flexibility for colorimetric, fluorometric, and label-free formats (Gurdo et al. 2023). Technologies based on droplet microfluidics further increased screening capacity, reaching up to 10⁹ variants per run, while lowering time and resource demands (Wang et al. 2021). Nevertheless, HTS platforms rely on measurable signals, e.g. fluorometric or colorimetric outputs, which, in general, limits their applicability to enzymes in C₁ assimilation that often do not generate directly detectable products. These enzymes typically require coupling with indicator reagents to enable detection (Kozaeva et al. 2022, De Maria et al. 2024). Although efforts have been made to pair droplet microfluidics with label-free readouts, the resulting throughput has not yet matched that of traditional droplet systems (Gantz et al. 2023). Enhancing the performance of label-free methods—e.g. mass spectrometry or chromatography—through integration with ultra-HTS workflows (Hernández-Sancho et al. 2024), will soon be a milestone in developing and engineering synthetic C₁ assimilation (Zeng et al. 2020).

Carbon-concentrating mechanisms and compartmentalization to enhance C₁-trophy

A possible solution to the low affinity of C₁-fixing enzymes, especially engineered variants, can be informed by natural systems. Autotrophic organisms have developed carbon-concentrating mechanisms (CCMs) to counteract the low specificity of RuBisCO and reduce its oxygenation activity. These systems raise the local CO₂ concentration and the CO₂/O₂ ratio near the carboxylase. CCMs can be biophysical, e.g. carboxysomes in cyanobacteria and pyrenoids in eukaryotic algae (Raven et al. 2008, Barrett et al. 2021, Borden and Savage 2021), where CO₂ is imported as bicarbonate (HCO₃^–) and reconverted by carbonic anhydrases within a specialized subcellular compartment containing RuBisCO (Yamaoka et al. 2025). Alternatively, CCMs are found in C4 and CAM plants, where carbon is shuttled as organic acids between cellular compartments and released in proximity to RuBisCO (Correa et al. 2025). Synthetic implementation of carboxysomes has been explored in E. coli, where the integration of 20 CCM genes from Halothiobacillus neapolitanus led to a RuBisCO-dependent E. coli strain able to grow at ambient CO₂ partial pressure—growth that would otherwise be unachievable in the absence of the carboxysome (Flamholz et al. 2020). This outcome illustrates broader opportunities for bacterial microcompartments beyond carbon assimilation. For instance, in methylotrophic yeasts, methanol-metabolizing enzymes are localized in peroxisomes to prevent cytosolic formaldehyde accumulation (Gassler et al. 2022). This spatial organization has been successfully implemented in S. cerevisiae by relocating components of the AOX–XuMP cycle to peroxisomes, combined with the overexpression of the peroxisomal membrane protein Pex5 to enhance matrix import capacity (Zhan et al. 2023). A comparable strategy may be relevant when developing formatotrophic and methylotrophic capacities in microbial systems (Peiro et al. 2022, Snyder et al. 2025).

ALE, a general framework for optimizing synthetic C₁ pathways

ALE plays a central role in C₁ assimilation engineering, drawing on the principles of natural selection in controlled environments to enable engineered strains to refine their metabolism around the introduced pathway. The technique spans a spectrum of setups, from basic serial passaging in flasks and tubes to more advanced platforms, e.g. chemostats and turbidostats (Dragosits and Mattanovich 2013, Fernández-Cabezón et al. 2019, Sandberg et al. 2019). In chemostats, growth is governed by the availability of a limiting nutrient, whereas turbidostats maintain constant cell density by adjusting dilution rates. This setup allows for microbial growth to be influenced not by nutrient depletion but by constraints in assimilation or cellular functions (Orsi et al. 2024a). ALE enables the refinement of diverse cellular traits across a wide range of targets, including the optimization of metabolic fluxes, enhancement of stress tolerance, rewiring of regulatory circuits, improved gene expression, and tuning of mutation frequencies. Selective pressure applied under defined conditions drives the emergence of adaptive traits, positioning ALE as a versatile tool for solving key problems in metabolic engineering. Several of these applications are discussed in the sections below.

Balancing metabolic nodes and junctions

As indicated previously, autocatalytic cycles remain difficult to engineer due to their functional overlap with endogenous metabolism, a challenge often addressed through ALE. These studies frequently uncover modifications at key metabolic branchpoints. For instance, independent mutations in the ribose phosphate diphosphokinase gene (prs)—a pivotal metabolic node directing flux towards biomass synthesis—were found in all chemoautotrophic clones during efforts to introduce the CBB cycle in E. coli (Antonovsky et al. 2016). In the same system, full autotrophy revealed another critical node involving phosphoglucoisomerase (Pgi), which contributed to the CO₂-dependent phenotype (Gleizer et al. 2019). Loss-of-function mutations in pgi led to a 100-fold reduction in enzyme activity, curbing fructose-6-phosphate efflux and kinetically stabilizing the cycle (Ben-Nissan et al. 2023). Similarly, during adaptation of the RuMP cycle in E. coli, methylotrophic evolution resulted in both downregulation and mutation of enzymes at branching points, including 6-phosphogluconate dehydrogenase (Gnd), with variants retaining only 5% of their native activity (Keller et al. 2022). Beyond laboratory evolution, natural autotrophs and methylotrophs provide valuable design principles for engineering heterotrophic hosts. For example, tuning branchpoint enzymes to reduce competition—e.g. employing an irreversible sedoheptulose-1,7-bisphosphatase (SBPase) to favour ribulose-5-phosphate regeneration in the RuMP cycle—can improve cycle performance (Bennett et al. 2018, Woolston et al. 2018a, Claassens et al. 2019a). Many branchpoints can be identified before in vivo testing (Claassens et al. 2019a, Kozaeva et al. 2021), and constraint-based modelling helps pinpoint these nodes by resolving stoichiometric flux distributions—particularly relevant in noncanonical microbial hosts (Nieto-Domínguez and Nikel 2020, Nogales et al. 2020, Fernández-Cabezón et al. 2022, Bujdoš et al. 2023). Still, these models tend to fall short in fully capturing regulatory dynamics, thermodynamic feasibility, and kinetic constraints, limiting their predictive utility. Further refinement in the near future will help solve these pitfalls.

Understanding the regulation of the host carbon metabolism

ALE has underscored the role of regulatory changes in enabling utilization of novel substrates (Wirth et al. 2022). For example, improved methanol assimilation in E. coli cometabolizing threonine revealed that the leucine-responsive regulatory protein (Lrp) repressed the expression of genes encoding threonine dehydrogenase and serine hydroxymethyltransferase, limiting flux through the threonine/glycine/serine branch point (Gonzalez et al. 2018). Lrp, a global regulator modulating hundreds of genes, is responsive to native starvation conditions (Cho et al. 2008), and mutations in lrp have been associated with enhanced performance during stationary phase (Tani et al. 2002). Similarly, ALE experiments aimed at improving methanol metabolism in C. glutamicum revealed that approximately one-third of the mutations acquired during evolution occurred in genes encoding transcriptional regulators (Tuyishime et al. 2018). In addition to transcriptional effects, allosteric regulation presents further constraints to flux tuning. A key example is the regulatory architecture of the thrLABC operon, encoding enzymes that control threonine biosynthesis, which hindered complete formatotrophic growth via the STC in engineered E. coli (Wenk et al. 2024). Redox cofactor balance, especially NADPH availability, remains equally critical in engineering C₁-utilizing systems (Ducat and Silver 2012). ALE of E. coli strains optimized for formate assimilation through either the STC or the rGlyP independently led to identical single base-pair changes in the pntAB promoter, resulting in a 13-fold increase in the pnt transcript levels (Ducat and Silver 2012, Kim et al. 2020). PntAB, a membrane-bound transhydrogenase, facilitates NADPH regeneration by catalysing NADP⁺ reduction via NADH oxidation (Sauer et al. 2004, Nikel et al. 2016). In P. putida, evolution under formatotrophic conditions using the rGlyP also triggered elevated expression of transhydrogenase genes (Turlin et al. 2022). While ALE can shed light on global regulatory rewiring relevant for synthetic C₁ metabolism, its implementation is not without drawbacks. The process is time-consuming, often requiring extended propagation to accumulate beneficial mutations, and pinpointing causative mutations among broader genomic changes remains difficult. Still, ALE has exposed regulatory loops that would have been difficult to predict from first principles—including, for instance, the observation that only a limited number of genetic alterations are needed to support autotrophic growth in E. coli via either the CBB cycle or the STC (Ben-Nissan et al. 2023, Wenk et al. 2024).

Solutions to relieve the toxicity of C₁ substrates and metabolic intermediates

Increased methanol tolerance in synthetic methylotrophs is often linked to elevated expression of genes directly involved in alcohol assimilation; however, additional unsuspected targets have also been identified, e.g. O-acetyl-l-homoserine sulfhydrylase (MetY) in engineered C. glutamicum (Wang et al. 2020b). A mutated version of MetY in an evolved synthetic methylotroph blocked the formation of O-methyl-l-homoserine, an l-methionine analogue, thereby preventing the incorporation of nonfunctional amino acids into proteins (Wang et al. 2020b). Although ALE has proven useful in enhancing resistance to toxic C₁ substrates, we argue that it may not fully address the core issue—as the C₁ assimilation pathway itself should act as a detoxification route. Other strategies that have shown potential in alleviating toxicity are discussed below.

Copy number variation

Copy number variation has been employed to mitigate formaldehyde toxicity and DNA–protein cross-linking during the implementation of the RuMP cycle in engineered E. coli (Chen et al. 2020, Nieh et al. 2024). Insertion sequence-driven copy number changes enable stochastic amplification of specific tandem repeat regions in the genome. When essential phenotype-related genes are embedded within these repeats, copy number fluctuations can be harnessed to dynamically adjust gene expression under selective pressure (Li et al. 2022, Kozaeva et al. 2024), similarly to repetitive transposon insertion protocols (Federici et al. 2025). Despite its potential, this strategy has seen limited use outside the context of synthetic methylotrophy and warrants greater exploration in future work.

Inducible systems based on methylotrophic transcriptional regulation

A detailed understanding of native architectures for C₁ substrate-dependent regulation in methylotrophs is key to advance synthetic methylotrophic engineering and to develop strategies to counteract toxicity issues. Several inducible systems based on formaldehyde-responsive elements have been characterized in both native methylotrophs, e.g. EfgA in Methylobacterium extorquens, and surrogate microbial hosts, e.g. FrmR from E. coli (Woolston et al. 2018b, Wang et al. 2020b, Bazurto et al. 2021). The role of EfgA, for instance, became evident while studying the consequences of elevated intracellular formaldehyde levels, known to halt growth in M. extorquens (Bazurto et al. 2021). Additionally, the native E. coli formaldehyde-inducible system FrmR/P_frm has shown promise for metabolic engineering purposes, e.g. enabling dynamic control of genes encoding pentose phosphate pathway enzymes to improve methanol assimilation—although this configuration did not outperform constitutive promoters (Rohlhill et al. 2020). Methanol- and formate-responsive transcriptional regulators have also been explored for constructing tightly controlled gene circuits (Wang and Gunsalus 2003, Rother et al. 2005). For carbon-fixation pathways that bypass methanol, formaldehyde, or formate, transcriptional modulation based on metabolite concentrations may allow conditional regulatory layers through basal-to-high gene expression levels. Inducible regulation, however, remains largely unexplored in synthetic methylotrophy, yet novel or redesigned control elements may help mitigate formaldehyde toxicity and reduce DNA–protein cross-linking (Whitaker et al. 2015, Wang et al. 2020a). Moreover, dynamic control could enable reallocation of carbon and energy resources, easing metabolic stress in hosts engineered with extended synthetic pathways.

Metabolic proofreading

A potential solution for the increased toxicity and carbon loss caused by promiscuous enzymes in synthetic assimilation pathways is metabolic proofreading (Linster et al. 2013, Danchin 2017). This strategy relies on incorporating auxiliary enzymes that eliminate toxic by-products or recover dead-end metabolites (Sun et al. 2017). Although scavenging and proofreading enzymes are widespread in natural systems, they have been somewhat overlooked in synthetic biology and pathway design. Metabolic proofreading, combined with protein engineering, has been applied in vitro to enhance the performance of the CETCH cycle (Schwander et al. 2016), the THETA cycle (Luo et al. 2023), and to recycle erythrose-4-phosphate—a dead-end intermediate—in a reconstituted pathway for poly(3-hydroxybutyrate) synthesis from glucose (Opgenorth et al. 2016).

Omic approaches can uncover and remediate bottlenecks at a systems-level

Efforts to engineer model microorganisms for the utilization of C₁ compounds seem to be limited by our understanding of the complex regulatory systems endogenous to the host. are often constrained by limited insight into the host's intricate regulatory networks. Systems biology, supported by omics technologies, offers a means to dissect and map these networks (Kato et al. 2022, Volke et al. 2023), revealing the dynamic interactions that govern microbial metabolism and aiding in the identification of critical pathways and control points (Amer and Baidoo 2021). The data generated through these high-throughput methods can be incorporated into genome-scale metabolic models (Roy et al. 2021), improving the predictive accuracy of pathway performance and informing rational design strategies (Bierbaumer et al. 2023).

Solutions to improve C₁-based bioproduction

From the examples discussed so far in this review, it becomes evident that the ultimate objective of engineering efficient C₁-assimilating systems is to convert C₁ feedstocks into value-added products. Hence, once robust C₁-trophic growth is established, metabolic engineering efforts shift toward biosynthesis. Target molecules typically include C₃–C₄ compounds, which are suitable for polymerization into materials that retain fixed carbon without reemission (Mezzina et al. 2021, Scown 2022). Both natural C₁-trophs and synthetic configurations, e.g. the malonyl-CoA bypass, have provided critical insights for metabolic rewiring (Orsi et al. 2023, Favoino et al. 2024). This energetically favourable route, which combines acetyl-CoA carboxylase and acetoacetyl-CoA synthase from Streptomyces sp. CL190, was initially demonstrated for photoautotrophic 1-butanol production in cyanobacteria (Lan and Liao 2012). The same approach was later applied for the conversion of formate into crotonate in engineered C. necator (Collas et al. 2023).

C₁ feedstocks may also be used as cocarbon sources to enhance redox cofactor regeneration and boost productivity. A synthetic methylotrophic E. coli relying on the RuMP cycle, for instance, achieved a succinic acid yield of 0.98 g g^–1 with methanol as an auxiliary substrate (Zhang et al. 2018) and produced 0.512 mM d-allulose per 1 mM methanol when the C₁ feedstock was cofed with xylose (Guo et al. 2022b). In another case, C. glutamicum expressing the RuMP cycle genes showed elevated cadaverine production when grown on ribose and methanol (Leßmeier et al. 2015). Beyond growth-coupled designs, production initiated during the stationary phase has shown substantial value in biotechnological setups. As an example, an ATP-dissipating module led to a 24% increase in lactate output under anaerobic conditions (Wichmann et al. 2023). In a different metabolic configuration, E. coli strains were modified to integrate fermentative and respiratory pathways, enabling aerobic production of lactate and isobutanol from glycerol, circumventing redox constraints and extending the potential of C₁-based manufacturing systems (Schulz-Mirbach et al. 2024b).

Solutions for cofactor imbalance and use of surrogate reducing equivalents

Noncanonical redox cofactors (NCRCs) encompass a potential solution to the cofactor-related metabolic constraints in synthetic C₁ fixation. These alternative cofactors retain the reactive features of native redox mediators but exhibit structural variations that render them incompatible with natural enzymes (Weusthuis et al. 2020). While MeDHs have not yet been tailored for NCRC utilization, several studies have explored the use of surrogate cofactors in different alcohol dehydrogenases and other C₁-related enzymes (Campbell et al. 2012, Wang et al. 2022, Aspacio et al. 2024, Orsi et al. 2024b). Nicotinamide-based NCRCs display redox potentials comparable to their natural analogs (Paul et al. 2014), and their orthogonality enables tight regulation of the redox state—addressing thermodynamic bottlenecks typical of most MeDHs while improving the utilization of reducing equivalents from C₁ molecules (Aspacio et al. 2024). These cofactors can support C₁ incorporation into biomass, target products, or both, while bypassing native electron flow through central carbon metabolism (Black et al. 2023). Despite their limited interaction with native redox systems, maintaining a proper redox balance and ATP supply remains essential under both aerobic and anaerobic conditions. In anaerobic fermentations—amenable to industrial scaling (Liew et al. 2022)—NCRCs can limit electron distribution to desired reactions, thereby reducing by-product formation. Engineering MeDHs to function with NCRCs may overcome thermodynamic constraints and support energy-efficient biosynthesis. Moreover, deploying these alternative cofactors could resolve conflicts arising from host dehydrogenase redundancy, as native enzymes often exhibit incompatible cofactor preferences (Volke et al. 2021). Hence, developing NCRC-compatible dehydrogenases continues to be an exciting direction in C₁-based biotechnology (Orsi et al. 2024b).

Selecting an appropriate microbial host

Selecting an appropriate microbial host for a given synthetic C₁ pathway remains a complex decision, requiring careful consideration of compatibility of the route(s) with the native metabolism, substrate range, stress tolerance, compartmentalization needs, availability of genetic tools, regulatory landscape, and production capacity (Calero and Nikel 2019). Although E. coli is frequently used due to its established role in biotechnology, nonconventional bacterial hosts with distinct metabolic features may offer superior performance for synthetic C₁ assimilation, particularly when pathways demand enhanced stress tolerance or metabolic adaptability (Gurdo et al. 2022, de Lorenzo et al. 2024). Minicells are yet another attractive system for C₁ biotechnology (Kim et al. 2022, Kozaeva et al. 2025). Additionally, cocultures systems offer an alternative for optimizing production by harnessing the complementary traits of autotrophic and heterotrophic microbes (Löwe et al. 2017, Fedeson et al. 2020, Kang et al. 2025). This division of labor enables the resolution of issues such as incompatible metabolic demands or oxygen sensitivity that arise in single-species settings (Claassens et al. 2016). As an example, pairing autotrophic acetogens that fix CO₂ and produce acetate (Box 1) with heterotrophs engineered for targeted product synthesis can substantially improve system performance and boost energy conversion efficiency.

Outlook

Engineering synthetic C₁ assimilation has largely centered on a limited set of model microorganisms, a situation favoured by the availability of robust genetic tools and detailed knowledge of their metabolic and regulatory networks (Tables 1 and 2). Nonetheless, current production levels using these microbial systems remain below the threshold for economic feasibility, emphasizing the need to investigate nonmodel microbes and native C₁-trophs to support the emergence of a true circular carbon bioeconomy. Addressing key challenges—including pathway compatibility, metabolic constraints, and enzyme performance—is essential for expanding synthetic C₁ assimilation to a broader spectrum of microbial hosts. Computational methods can guide the a priori selection of pathways that align with target products and meet both metabolic and thermodynamic process demands (Fig. 4). For example, C₃-yielding synthetic routes, e.g. the rGlyP or the RuMP cycle, are optimal for lactate biosynthesis, whereas acetyl-CoA-based systems, e.g. the serine cycle, are better aligned with C₂-based bioproduction. The oxidation state of the substrate is also a determining factor defining the success of the overall strategy; methanol provides greater energy potential than formate, but its assimilation introduces more challenging metabolic limitations. Techniques that include in vitro pathway assembly, enzyme engineering and catalytic refinement, regulatory tuning, and synthetic mutagenesis offer routes to fine-tune novel pathways ahead of in vivo implementation. Recent developments in artificial intelligence and deep learning have lowered the barrier to de novo design of enzymes and synthetic pathways, while ancestral DNA continues to serve as a reservoir of unconventional metabolic functions. The integration of fully automated platforms for engineering synthetic organisms is expected to further accelerate progress, facilitating the construction of tailored C₁-assimilating systems (Fig. 4). We argue that the convergence of these approaches positions synthetic C₁ metabolism as a viable pillar for advancing circular bioeconomy initiatives in the near future.

Funding

The Nikel Lab gratefully acknowledges financial support from The Novo Nordisk Foundation through grants NNF20CC0035580, LiFe (NNF18OC0034818), TARGET (NNF21OC0067996), FM·Pseudomonas (NNF24OC0091501), and NovoF (NNF23OC0083631), and the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement numbers 814418 (SinFonia) and 101082049 (TOLERATE).

Conflict of interest

The authors declare no competing interests.