ARCANE — Closed-Loop Molecular Food Synthesis

Closed-loop human food at habitat scale is not a single problem. It is at least three. The first is whether the underlying biology can be made to work at all — whether single-cell organisms can convert CO₂, electrolytic hydrogen, and waste-derived nitrogen into nutritionally complete human food at industrially relevant rates. The second is whether the integration of those organisms into a multi-compartment, multi-organism habitat-scale loop is robust enough to keep humans alive when a single subsystem fails. The third is whether the resulting food is palatable, varied, and culturally acceptable enough that humans will actually eat it for years at a time. Each of these questions has a different state of the art, a different failure mode, and a different cohort of researchers. The ARCANE program treats them as one program because the failure of any one of them collapses the others, but the analysis here keeps them separate, because the literature has frequently conflated them and the conflation has produced confused arguments on all sides.

The three questions are different

The phrase “closed-loop food” gets used to mean at least three distinct things, and the conflation has done the field genuine harm. The first is the biochemical claim: that the molecular pathways for converting one-carbon substrates (CO₂, methanol, formate) and electrolytic hydrogen into proteins, carbohydrates, and lipids are well-characterized, that several of them have been demonstrated at industrial scale, and that the per-gram cost of such food is converging toward parity with conventional protein. The second is the systems-engineering claim: that those pathways can be integrated into a multi-compartment closed loop — waste in, food and breathable atmosphere out — that survives months or years of operation without external resupply. The third is the human-acceptance claim: that astronauts, settlers, or extreme-environment crews will actually consume microbial-derived foods at the volumes and durations the loop assumes, rather than reverting to cravings, deficiencies, and the pre-packaged-food regression that has sunk every previous closed-ecosystem experiment.

The three claims are independent. A program can succeed at the biochemistry and fail at the systems integration; this is the present state of the field, with chemoautotrophic single-cell protein operational at industrial scale¹ but no demonstrated multi-year, multi-crew closed-loop habitat. A program can succeed at the systems integration and fail at the human-acceptance question; the Soviet BIOS-3 facility in Krasnoyarsk and the Arizona Biosphere 2 closure, taken together, are case studies in technically functional life support whose human-side failures dominated the historical narrative.²³ And a program can succeed at the acceptance question — humans will eat almost anything if the alternatives are worse — and fail at either of the first two, in which case the system kills its inhabitants. The relevant question for ARCANE is not which of the three is hardest. The relevant question is how to engineer all three simultaneously in a way that the failure modes of each do not propagate into the others.

The most-cited starting point for this conversation is the Krasnoyarsk BIOS-3 program, which between 1972 and 1973 sustained a three-person crew for 180 days on a closed-loop architecture combining wheat phytotrons and Chlorella algal cultivators that recycled approximately 100% of water and atmosphere and roughly half of the dietary requirements.² BIOS-3 was the existence proof that closed-loop life support is possible. It was also the existence proof that closed-loop life support is hard: the 180-day duration, the partial-rather-than-full food closure, and the heroic operating conditions are not a template for civilian deployment. Biosphere 2, two decades later and at twenty times the scale, was the second existence proof and the second lesson. The eight Biospherians completed two years inside the structure, but only after oxygen levels dropped from the ambient 21% to roughly 14% — equivalent to living at 4,500 metres of altitude — because organic matter in the soil respired faster than the photosynthetic compartments could replenish, and the CO₂ that resulted was sequestered into the structure’s exposed concrete as calcium carbonate rather than recycled by the plant compartments.³ The biological loop and the civil-engineering loop were not separately characterized. They interacted, in a direction that nobody had modelled, and the system nearly failed.

So the three questions need to be held apart. The biochemical question is whether the molecular pathways exist. The systems-engineering question is whether they integrate. The acceptance question is whether humans will live inside the result. The remainder of this page is structured accordingly. First the biochemistry, where the recent progress is genuinely striking. Then the integration, where the evidence base is much thinner. Then the acceptance question, where the data is almost entirely from short-duration analogues and where the honest answer is that we do not know.

The mass-and-energy economics of off-Earth food

Before getting to the biology, the engineering economics. A six-person Mars crew on a conjunction-class mission requires food, water, and atmospheric revitalization for roughly 900 days of round-trip duration, plus surface activity. The food requirement alone, at a per-person daily metabolic budget of approximately 0.62 kg, is on the order of 3,400 kg of food mass per crew over the mission, before accounting for packaging, redundancy, and the calorie-density premium that pre-packaged food carries.⁴ The launch-mass cost of that quantity, even on a fully reusable architecture, is the largest non-propellant consumable in the mission budget. Stored food also degrades. Vitamin content drops significantly over multi-year shelf lives even under optimal storage conditions, and the menu fatigue documented in long-duration submarine and Antarctic deployments produces measurable caloric under-consumption that compounds nutritional deficiencies. The pre-packaged regression is not a thought experiment. It is what happens by default if no other architecture is deployed.

Hydroponic and aeroponic plant cultivation is the standard alternative, and the standard alternative has been carefully studied. The European EDEN-ISS project, operated by the German Aerospace Center (DLR) at the Antarctic Neumayer III station between 2018 and 2023, produced approximately 270 kg of fresh food across nine months of its first deployment — 26 different crops grown across 12.5 m² of aeroponic cultivation area, supplied by LED lighting and a closed-loop nutrient-and-water delivery system.⁵ This is the most carefully characterized multi-year off-grid plant-cultivation system in existence, and the result is the relevant data point: 30 kg of fresh food per square metre per year, achieved under expert operation in a stable analogue environment. The data from EDEN-ISS, combined with comparable work on the ISS Veggie and Advanced Plant Habitat platforms, establishes that plant cultivation in closed environments is operationally feasible, that crew morale and dietary diversity benefits are real, and that the systems are not light. Every kilogram of plant biomass requires multiple kilograms of water in circulation, several square metres of cultivation area, and tens to hundreds of watts of grow-light power. For a multi-year settlement scaling beyond a half-dozen crew, the area, water, and power budgets exceed what can be reasonably allocated to food production alone.

The chemoautotrophic alternative inverts the area-and-water economics. Hydrogen-oxidizing bacteria growing on H₂ + CO₂ + minerals produce a complete protein from inputs that Mars provides natively (CO₂ from the atmosphere, hydrogen from electrolyzed water, minerals from regolith leaching) and require neither lighting nor large volumes of water. The Solar Foods commercial process, the most-developed implementation, achieves industrial-scale chemoautotrophic protein production at a productivity of approximately 1 g per litre per hour and an energy efficiency (O₂ consumed per CO₂ fixed) of 2.7, with a per-kilogram CO₂ requirement of approximately 1.85 kg and a target operating cost of approximately €5 per kilogram at full scale.¹⁶ The resulting Solein powder is a complete protein at 65–70% protein content, with an amino acid profile suitable for human nutrition. It is not a complete diet. It is the first demonstration that the protein leg of the closed loop is solvable with currently-deployed technology. The remaining legs — carbohydrate, lipid, micronutrients, fibre, and the social-acceptance question of palatability — are where the ARCANE program’s research focus lies.

The energy economics are non-trivial. Producing one kilogram of microbial protein from electrolytic hydrogen requires on the order of 8–12 kWh of electrical energy for the electrolysis step alone, plus additional energy for bioreactor mixing, gas compression, harvesting, and drying. At grid prices in industrial regions today, this puts the energy cost per kilogram in the range of $0.50 to$1.50, well below the cost of equivalent conventional protein. At Mars, where electrical energy must be generated locally from solar or nuclear sources, the energy cost is the binding constraint, and the question of how much surplus power the habitat can spare for food production is a co-design question with the ENERA energy program. ARCANE without ENERA is a research curiosity; ENERA without ARCANE is power without a use. The two programs are coupled by design.

What ARCANE is, technically

Six technology streams converge to make closed-loop molecular food synthesis tractable. None of them is the answer alone. The integration of several into a habitat-scale loop is the open frontier, and it is what the ARCANE research program pursues. The technical state of each stream is summarized below, with citations to the most directly relevant primary literature so that the claims can be checked.

Hydrogen-oxidizing bacteria and chemoautotrophic protein

Microorganisms in the genera Cupriavidus, Xanthobacter, and related chemoautotrophs grow on a feedstock of hydrogen, CO₂, and mineral nutrients via the Calvin–Benson–Bassham cycle and energy-yielding hydrogen oxidation. The biomass is harvested, dried, and processed into a high-protein powder. The intellectual lineage here is much older than the recent commercial activity. Imperial Chemical Industries operated a 50,000–75,000 tonnes-per-year Pruteen single-cell-protein plant at Billingham in the late 1970s, using Methylophilus methylotrophus on methanol as a feedstock for the European pig-feed market.⁷ The Pruteen plant was technically successful and commercially abandoned in the mid-1980s when the price of soybean meal collapsed; the technology was sold and adapted but the platform did not return to commercial scale until the gas-fermentation revival of the 2010s. The relevant lesson is that the underlying biology has been industrially proven for decades. The current generation of work is a return, with better strains, lower-cost electrolytic hydrogen, and a dramatically different competitive landscape.

The leading current implementation is Solar Foods (Pasi Vainikka, Juha-Pekka Pitkänen; Vantaa, Finland), whose Factory 01 was commissioned in April 2024 and which reached its design-capacity target of 160 tonnes per year of Solein production in October 2025, with planned expansion to 230 tonnes per year in 2026 and a Factory 02 of 6,400-tonne nameplate capacity slated for 2028 commissioning.⁶ The organism is Xanthobacter sp. SoF1, isolated from Baltic Sea sediment. The process electrolyzes water to H₂ and O₂, captures CO₂ from air or industrial flue gas, and feeds H₂ + CO₂ + mineral nutrients to the chemoautotroph in an 8-metre-diameter bioreactor. Solein received its first novel-food regulatory approval from the Singapore Food Agency in October 2022 — the first such approval anywhere — and obtained self-affirmed GRAS status in the United States in September 2024.⁸ In NASA’s Phase 3 Deep Space Food Challenge in August 2024, Solar Foods was the international winner among the four-team international finalists, and the only winner whose process inputs (CO₂, electricity, water) directly match the resource availability profile of a Mars surface habitat.⁹ No other food technology in existence today is as directly Mars-applicable. The Solar Foods–OHB HOBI-WAN (Hydrogen Oxidizing Bacteria In Weightlessness As a source of Nutrition) microgravity prototype is the explicit space-development descendent of the Earth-side commercial process. The relevance to ARCANE is direct: the protein leg of the loop is solved, in operating commercial form, today.

The Air Company / Air Protein / NovoNutrients family represents the chemistry-first alternative. Air Company (Greg Constantine, Stafford Sheehan; Brooklyn, NY) hydrogenates CO₂ catalytically into food-grade ethanol, which is then fermented into yeast biomass — a two-step path that is more energy-intensive than direct chemoautotrophy but uses better-characterized industrial chemistry. Air Company won NASA’s CO₂ Conversion Challenge Phase 1 in May 2019 and reached Phase 3 of the Deep Space Food Challenge with a “yeast pasta from astronaut breath” demonstration. NovoNutrients (David Tze) and Air Protein (Lisa Dyson) are HOB-based competitors targeting fish-meal-class protein flour for aquaculture as a near-term commercial market. LanzaTech (Jennifer Holmgren, Sean Simpson) commercialized acetogen-based gas fermentation using Clostridium autoethanogenum at steel-mill scale, producing more than 100 million gallons of ethanol from waste industrial gases, with operating facilities in China, India, Belgium, and the United States.¹⁰ LanzaTech’s commercial focus is fuels and chemicals rather than human food, but the underlying acetogen platform is directly portable to space-relevant one-carbon biology, and the LanzaJet sustainable-aviation-fuel pathway has received commercial-aviation certification — proof that the chemistry has cleared regulatory bars for human-adjacent uses.

Synthetic methylotrophy and formatotrophy

The next stream goes a level deeper. Rather than relying on naturally chemoautotrophic organisms, recent work has re-engineered the most-studied bacterium in biology — Escherichia coli — to grow on methanol or formate as the sole carbon source. The achievement here is genuinely new. Methanol can be synthesized from CO₂ and H₂ by well-known catalytic chemistry; formate can be produced from CO₂ by direct electrochemical reduction at room temperature. Either is a transportable, room-temperature liquid feedstock that can be moved through pipework rather than gas plumbing, which dramatically simplifies bioreactor engineering. Until very recently, however, no engineered organism could grow on either as the sole carbon source at industrially relevant rates.

The breakthrough was achieved in the laboratory of Julia Vorholt at ETH Zurich. The 2024 paper in Nature Catalysis by Reiter, Vorholt, and colleagues describes a synthetic methylotrophic E. coli with a doubling time of 4.3 hours on methanol, comparable to many natural methylotrophs, and capable of producing four distinct value-added products — lactic acid, polyhydroxybutyrate, itaconic acid, and p-aminobenzoic acid — from methanol as the sole carbon source.¹¹ The 2022 Nature Communications paper by Keller, Reiter, and colleagues established the underlying methylotroph by 250 generations of adaptive laboratory evolution in a chemostat, producing the first ribulose-monophosphate-cycle methanol-growing E. coli with no external carbon supplementation.¹² These are engineering feats that did not exist five years ago. They convert E. coli — the most extensively engineered organism in biology, with the deepest synthetic-biology toolkit available — into a chassis for one-carbon bioproduction.

The implications for closed-loop food are direct. Methanol is a Mars-compatible feedstock: it can be synthesized from atmospheric CO₂ and electrolytic hydrogen via the methanol synthesis route used industrially since the 1920s. Formate, the alternative, is producible by electrochemical CO₂ reduction at room temperature with reasonable Faradaic efficiency. Either gives the food-production loop a liquid one-carbon feedstock that is easier to handle than gaseous H₂ + CO₂ and that allows the engineering toolkit of standard E. coli to be deployed against the food-synthesis problem. The Vorholt-lab work provides the engineering proof. The remaining work, which ARCANE focuses on, is the integration of methylotrophic E. coli (and the equivalent yeast platforms — Yarrowia lipolytica, Saccharomyces cerevisiae — that have been engineered for similar capabilities¹³) into a complete-nutrition, palatability-acceptable food output rather than an industrial-chemistry output.

Synthetic CO₂ fixation pathways

Above the level of “find a chemoautotroph or engineer one” sits the deeper question of whether CO₂ fixation can be made faster than nature has evolved it. The Calvin–Benson–Bassham cycle, the dominant carbon-fixation pathway in plants and cyanobacteria, evolved under different atmospheric conditions and is known to be substantially sub-optimal in modern atmospheric CO₂ concentrations. Tobias Erb’s group at the Max Planck Institute for Terrestrial Microbiology in Marburg has spent more than a decade designing and testing synthetic CO₂-fixation pathways that out-perform their natural counterparts.

The 2016 Science paper introducing the CETCH cycle — Crotonyl-CoA / Ethylmalonyl-CoA / Hydroxybutyryl-CoA — describes a 17-enzyme synthetic pathway assembled from enzymes originating in nine different organisms across all three domains of life, with three of the enzymes engineered specifically for the role.¹⁴ The cycle was designed by metabolic retrosynthesis, optimized through several rounds of enzyme engineering and metabolic proofreading, and shown to fix CO₂ at a rate of approximately 5 nanomoles per minute per milligram of protein in vitro — substantially faster than the Calvin cycle under comparable conditions, and at approximately 20% lower energy per CO₂ fixed. The 2020 Nature Biotechnology follow-up reconstituted the CETCH cycle inside microfluidic droplets containing isolated thylakoid membranes from spinach, producing a functional synthetic chloroplast operating outside any living cell. The 2023 Nature Catalysis paper by McLean, Erb, and colleagues describes the THETA cycle — a 17-enzyme synthetic pathway using crotonyl-CoA carboxylase/reductase and phosphoenolpyruvate carboxylase, the two most efficient CO₂-fixing enzymes characterized — that converts two CO₂ molecules into one acetyl-CoA per cycle.¹⁵ The THETA cycle was modularized into three components and successfully implemented in E. coli, with each module’s functionality verified through growth-coupled selection and isotopic labelling. Yield improvements of two orders of magnitude were achieved using a combination of rational and machine-learning-guided optimization. The work is the first demonstration of a synthetic-to-nature CO₂-fixation pathway operating in a living host.

These results are not yet at industrial scale. They are, however, the proof that the rate-limiting step in chemoautotrophic food synthesis — the speed of CO₂ fixation — is not fixed by evolution. If the CETCH/THETA-class pathways can be ported into production-strain hosts at industrially relevant titres, the per-litre productivity of microbial food production rises by roughly an order of magnitude, and the bioreactor footprint required for habitat-scale food shrinks correspondingly. This is a long-arc research bet rather than a near-term deployment, and ARCANE treats it as such; the program funds the basic engineering work but does not depend on it for the deployable architectures of the next decade.

Cyanobacterial cultivation and the Mars-atmosphere case

Cyanobacteria — photosynthetic prokaryotes — occupy a complementary niche in the closed-loop architecture. Unlike chemoautotrophs, which need supplied hydrogen, cyanobacteria fix CO₂ photosynthetically using sunlight as the energy source, and the more interesting strains (Anabaena, Nostoc) also fix atmospheric nitrogen, eliminating the requirement for supplied nitrogen fertilizer. The biomass is consumed directly as food (e.g., the long-established Limnospira / Spirulina nutritional supplement market) or fed into a downstream organism — typically E. coli or yeast — that converts it into a more processed end product.

The Mars case is the most-studied. The Laboratory of Applied Space Microbiology at the University of Bremen ZARM, led by Cyprien Verseux, operates the Atmos photobioreactor for testing cyanobacteria under low-pressure (100 hPa), Mars-analogue atmospheres consisting of 96% N₂ and 4% CO₂ on Mars Global Simulant (MGS-1) regolith — a Rocknest-based open-standard basaltic regolith analogue developed by the Center for Lunar and Asteroid Surface Science at the University of Central Florida.¹⁶ The 2021 Frontiers in Microbiology paper by Verseux and colleagues demonstrated that Anabaena sp. PCC 7938 grows at biomass yields equivalent to ambient-pressure controls under the simulated Mars atmosphere, fixing nitrogen and CO₂ together, on a medium prepared by leaching MGS-1 regolith.¹⁷ The follow-up work demonstrated that anaerobic digestion of Anabaena biomass, fertilized only from MGS-1, yielded a working fertilizer that supported edible duckweed (Lemna sp.) production at a ratio of 27 grams of fresh duckweed per 1 gram of dry cyanobacterial input. This is the most operationally relevant Mars in-situ-resource-utilization food experimental work currently active anywhere in the world.

The ESA ARTHROSPIRA-B experiment, conducted aboard the International Space Station from 2017 onward, demonstrated Limnospira indica PCC 8005 growing in microgravity in a four-bioreactor photobioreactor system within the Biolab incubator, with online measurements of oxygen production rate and growth rate.¹⁸ The result — that microgravity has no first-order effect on Arthrospira growth in a properly mixed photobioreactor — eliminates a long-standing concern about whether cyanobacterial cultivation could be deployed on the ISS, the lunar surface, or transit-class missions. The follow-up ARTHROSPIRA-C, semi-continuous and at multiple light intensities, is in progress. The combination of the Verseux ZARM Mars-atmosphere work and the Poughon–Creuly ARTHROSPIRA microgravity work bounds the space-environmental envelope within which cyanobacterial food production has been experimentally validated.

Precision fermentation and cellular agriculture

The two complementary streams that round out the ARCANE biochemistry are precision fermentation and cultivated cellular agriculture. Precision fermentation — using engineered microbes to produce specific animal-protein-equivalent ingredients (whey, casein, egg-white, heme, collagen) — is now a multi-billion-dollar market, with the precision-fermentation-ingredients sector valued at approximately $5.0 billion in 2025 and projected to reach$36.3 billion by 2030 at a CAGR of 48.6%.¹⁹ Perfect Day, Geltor, The EVERY Company, Motif FoodWorks, and Imagindairy have all reached commercial scale on specific ingredients. Perfect Day’s Gujarat, India production facility is on track for second-half-2026 commissioning with full ramp-up in 2027. The 2024 partnership with Unilever’s Breyers brand, putting Perfect Day’s recombinant whey protein into a national U.S. retail dairy product, is the first mass-market consumer demonstration. The technology is no longer speculative; it is operating at commercial scale and clearing regulatory bars at the rate of several products per year.

Cultivated cellular agriculture — growing animal muscle, fat, and connective tissue directly from cell culture rather than from whole animals — has progressed through a parallel commercial trajectory. Mark Post’s 2013 Maastricht University laboratory demonstration produced the first cultivated hamburger at a per-kilogram cost of roughly $2.3 million, requiring two years of work and over 20,000 individual muscle-fibre strands per 142-gram patty.²⁰ The technology cost has since fallen by approximately five orders of magnitude. Believer Meats received an FDA “no questions” letter and USDA approval for its Wilson, North Carolina production facility in late October / early November 2025 — the first cultivated-meat plant of true industrial scale, with 200,000 square feet of capacity rated to produce approximately 26 million pounds (11,800 tonnes) of cultivated chicken annually, beginning with a 21-million-pound first-year output.²¹ Believer Meats is the fifth company approved for cultivated-meat sale in the United States, joining Upside Foods and GOOD Meat (chicken), Wildtype (salmon), and Mission Barns (pork fat). The technology has crossed the regulatory threshold and is entering the commercial-scale cost-reduction phase.

For ARCANE, neither precision fermentation nor cultivated agriculture is the central pillar of the closed-loop food architecture. The central pillar is chemoautotrophy plus engineered methylotrophy: CO₂ + H₂ + N₂ in, complete protein and other macros out, with the smallest-possible water and area footprint. Precision fermentation and cellular agriculture are the lipid, the flavor, the texture, the cultural-acceptability layer that turns nutritionally-adequate microbial powder into food that humans will eat. The integration of the two is the open frontier.

Definitional bounds

Before moving to the systems-engineering and acceptance questions, four exclusions are worth being explicit about, because the loose use of “molecular food” and “post-scarcity food” in the discourse has produced confusions that make the substantive arguments harder to have.

ARCANE does not mean replication of Earth food. The output of a chemoautotrophic process is not a hamburger, a tomato, or a glass of milk. It is a high-protein powder, a microbial biomass, a precision-fermented ingredient. The texture, flavour, and cultural-context attributes that make Earth food culturally meaningful are not produced by the underlying biology; they are added by downstream processing, formulation, and culinary work. A society that consumes ARCANE-derived food at scale will eat differently from a society that consumes traditionally-farmed food. The closed-loop architecture is not a substitute for traditional farming on Earth, where the inputs (sun, water, soil) are abundant and the outputs (whole plants, whole animals) carry cultural meaning that microbial powders do not. The closed-loop architecture is the food architecture of contexts where traditional farming is infeasible: extended off-Earth missions, deep-Earth habitats, bunker-class survival contexts, and — increasingly — environmental-impact-conscious supplementation of Earth diets.

ARCANE does not mean abandonment of plants. Higher plants remain essential to closed-loop architectures for at least four reasons. They are the most efficient atmosphere-revitalization compartment available, fixing CO₂ and producing O₂ at rates that microbial-only systems struggle to match. They produce dietary fibre, vitamins (particularly the C, A-precursor, and K families), and phytochemicals that are difficult to synthesize microbially. They provide the textural and cultural-context food experiences (a fresh leaf, a tomato, a strawberry) that microbial-only diets cannot. And they buffer the system against single-organism failure modes, which are the most-feared failure mode in closed-loop architectures. The MELiSSA reference architecture explicitly couples a higher-plant compartment to its three microbial compartments precisely for these reasons, and the EDEN-ISS Antarctic deployments validate the operating performance of plant compartments in closed environments.²²⁵ The relationship between microbial food production and plant food production is complementary, not substitutive.

ARCANE does not mean a single organism solves the problem. Every closed-loop architecture that has been seriously studied — BIOS-3, MELiSSA, Biosphere 2, the NASA Lunar–Mars Life Support Test Project — has converged on a multi-compartment, multi-organism design, for reasons that are both biological and engineering. Single-organism systems are catastrophic-failure-prone: a single contamination, mutation, or operating-parameter excursion can collapse the entire food output. Multi-organism systems provide redundancy, niche specialization (one organism is good at protein, another at fibre, another at fat, another at nitrogen fixation), and natural buffering against the operating-condition variations that occur in any real-world deployment. The engineering question is not “find the best single organism”; it is “find the right consortium and the right inter-compartment plumbing.” The relevant model for ARCANE is the MELiSSA five-compartment architecture, scaled up and modernized with the chemoautotrophic and methylotrophic technologies that did not exist when MELiSSA was first specified in 1989.

ARCANE does not mean food synthesis is a finished problem. The biochemistry is operational at industrial scale for the protein leg. The lipid leg is partially solved at laboratory scale and not yet at commercial scale. The fibre, vitamin, and micronutrient legs are unsolved or solved only by supplementation from external sources. The systems-integration leg has been demonstrated at 180-day duration with three crew (BIOS-3) and at 24-month duration with eight crew (Biosphere 2, with the oxygen-decline failure documented above) but never at the multi-year, multi-crew, full-closure scale that off-Earth deployment requires. The acceptance leg has been studied only in short-duration analogues. ARCANE is a research program because the integration is genuinely incomplete; we are explicit about this rather than the alternative.

These exclusions are not throat-clearing. They are the load-bearing definitional choices that determine what the rest of the analysis is about, and they constrain what the ARCANE program is allowed to claim.

The historical precedents

Closed-loop life support is not a new field. The cumulative body of experimental work spans roughly six decades and four programs, and any new architecture has to contend with the lessons that have been paid for by those programs.

BIOS-3 (Krasnoyarsk, 1972–1973 onwards)

The Soviet BIOS-3 facility at the Institute of Biophysics in Krasnoyarsk, Siberia, was the first closed-loop human life-support experiment to sustain a multi-person crew for multi-month duration. Construction began in 1965 under the leadership of Joseph Gitelson; the first crewed experiment occurred in 1972–1973 and sustained three crew members for 180 days inside a 315 cubic-metre underground steel structure divided into four compartments — a crew habitat, an algal cultivator, and two phytotrons growing wheat or vegetables.² The compartments were illuminated by water-cooled 20-kW xenon lamps providing photosynthetically-active light comparable to sunlight. Water and atmosphere were closed at approximately 100% recycling; food was closed at approximately 50% (the wheat compartment provided staple grain; the algal compartment provided supplementary protein; the remainder was supplied from external stocks). The system survived. The crew survived. The 1986 meeting between Gitelson and the Biosphere 2 leadership team (John Allen, Mark Nelson, Oleg Gazenko, and others) explicitly acknowledged the BIOS-3 work as the experimental and intellectual predecessor to Biosphere 2 and to subsequent closed-loop programs.

BIOS-3’s lessons are foundational. Human-scale closed-loop life support is biologically achievable. Atmosphere and water close more easily than food. Higher plants are the most efficient atmosphere-revitalization compartment. The duration limit is set by the systems that haven’t yet been characterized — by what fails at 200 days that didn’t fail at 180. And the human-side challenges (psychological strain, dietary monotony, social-confinement effects) are at least as severe as the engineering challenges. Subsequent programs have inherited each of these.

Biosphere 2 (Arizona, 1991–1994)

Biosphere 2 was the largest and most ambitious closed-ecosystem experiment ever attempted. Constructed in Oracle, Arizona between 1987 and 1991 by a private team led by John Allen and the Space Biospheres Ventures consortium, the facility enclosed approximately 12,700 square metres under glass, with seven distinct biome compartments — rainforest, ocean, savanna, desert, marsh, intensive-agriculture, and human habitat — supporting a crew of eight Biospherians for the first closure mission from September 1991 to September 1993.³ The atmospheric leakage rate was the lowest ever achieved for a closed structure of this scale, less than 10% per year.

The first-mission failure modes have been documented in detail. The most consequential was the oxygen decline. Atmospheric oxygen dropped from 20.9% at closure to 14.2% by the sixteenth month — the equivalent of living at approximately 4,500 metres of altitude — and several of the crew exhibited measurable altitude-sickness symptoms before external oxygen had to be added in January 1993 to prevent further deterioration. The mechanism was not initially obvious. Excess soil organic matter respired CO₂ at rates that the photosynthetic compartments could not match, and the resulting CO₂ was sequestered by reaction with the structure’s exposed concrete to form calcium carbonate rather than recycled by the plant compartments. The civil-engineering loop and the biological loop interacted, in a direction nobody had modelled.

Biosphere 2’s lessons include several that are directly relevant to ARCANE. Closed ecosystems exhibit dynamics that the component characterizations do not predict; the system is not the sum of its parts. The materials of the structure participate in the biogeochemistry. Closed-loop architectures need conservative oxygen-and-CO₂ buffers and active monitoring, not just passive design. And dietary restriction, even when caloric-and-nutritional adequacy is maintained, produces measurable weight loss and physiological stress in the crew. The first-mission Biospherians lost on average 16% of their body weight during the two-year closure. The food output was nutritionally adequate; the human-acceptance leg was not.

MELiSSA (ESA, 1989–present)

The Micro-Ecological Life Support System Alternative is the European Space Agency’s reference architecture for closed-loop life support, initiated in 1989 under the leadership of Christophe Lasseur and operated as a 30-organization, 35-year-and-counting research program centred on the MELiSSA Pilot Plant at Universitat Autònoma de Barcelona, inaugurated in 2009 under principal investigator Francesc Gòdia.²² The architecture comprises five interconnected compartments: a liquefying compartment (thermophilic anaerobic fermenters that convert mission waste into volatile fatty acids, NH₄⁺, and CO₂), a photoheterotrophic anoxygenic compartment (using Rhodospirillum rubrum to consume the volatile fatty acids), a nitrifying compartment (Nitrosomonas + Nitrobacter converting NH₄⁺ to nitrate), a photoautotrophic compartment (the Limnospira indica PCC 8005 strain that is the subject of the ARTHROSPIRA-B and ARTHROSPIRA-C experiments) and a higher-plant compartment with 32 candidate crops. The crew compartment closes the loop.

The MELiSSA Pilot Plant operates the integrated loop in stages. Rats currently serve as the model crew, with the goal of human-crew support targeted for the late 2020s. The cumulative output of the program is the most extensive characterization of closed-loop biology in existence: dozens of doctoral theses, hundreds of peer-reviewed papers, and the engineering documentation behind the ATHROSPIRA-B and ATHROSPIRA-C ISS experiments. ARCANE’s reference architecture is a modernized version of the MELiSSA five-compartment design, with chemoautotrophic and methylotrophic compartments replacing some of the original microbial compartments and with the higher-plant compartment retained for fibre, vitamin, and palatability supplementation.

ICI Pruteen (Billingham, 1976–1985)

The fourth precedent is the only one that operated at full commercial scale, and it is the only one that was not designed for closed-loop life support. Imperial Chemical Industries’ Pruteen plant at the Billingham, UK chemical-works complex commissioned in 1976 and operated through approximately 1985, producing 50,000–75,000 tonnes per year of single-cell protein from Methylophilus methylotrophus on a methanol feedstock for the European pig-feed market.⁷ The plant was technically successful: the bioreactor (the largest continuous fermenter ever built at the time, at 1,500 cubic metres) operated at design capacity; the product met its protein and amino-acid specifications; the regulatory and customer-acceptance work cleared. The plant was commercially abandoned in the mid-1980s when soybean meal prices fell to a level that Pruteen could not match. The technology was sold and adapted, but no comparable commercial single-cell-protein plant operated again until the 2010s revival.

The Pruteen lesson for ARCANE is sobering. The biology works at industrial scale. The economics are sensitive to the commodity price of substitute protein. Off-Earth deployment, where soybean meal is not an option, may give the ARCANE-class architecture its decisive cost advantage; on Earth, the architecture has to compete against the world soybean-meal-and-fishmeal market on price, and the price competition is not yet decisively won. The current generation of work — Solar Foods at €5/kg target, NovoNutrients at fishmeal parity for aquaculture, the precision-fermentation cost curve — is converging on parity, but the convergence is recent.

What has been demonstrated, recent and ongoing

The list of operational and clinical-stage closed-loop food results separates ARCANE from speculative replicator-style food fabricators. Each entry below is concrete, citable, and either operating at commercial scale or in active demonstration.

Solar Foods Factory 01. Industrial-scale chemoautotrophic protein production at 160 t/year capacity (April 2024 commissioning, October 2025 productivity-target hit), planned 230 t/year in 2026 expansion, Factory 02 at 6,400 t/year planned for 2028.¹⁶ Singapore Food Agency novel-food approval (October 2022), U.S. self-affirmed GRAS (September 2024), NASA Deep Space Food Challenge Phase 3 international winner (August 2024).⁹ HOBI-WAN microgravity prototype in development with OHB. The most directly Mars-applicable food technology in operating commercial form.

Vorholt-lab synthetic methylotrophy. E. coli growing on methanol as the sole carbon source at 4.3-hour doubling time (Reiter et al., Nature Catalysis 2024¹¹) and on methanol via the ribulose-monophosphate cycle after 250 generations of adaptive laboratory evolution (Keller et al., Nature Communications 2022¹²). Bioproduction of lactic acid, polyhydroxybutyrate, itaconic acid, and p-aminobenzoic acid demonstrated. The chassis for one-carbon liquid-feedstock food production.

Erb-lab synthetic CO₂ fixation. The CETCH cycle (Schwander et al., Science 2016¹⁴) — 17-enzyme, 9-organism, three-engineered-enzymes synthetic pathway fixing CO₂ at approximately 5 nmol/min/mg in vitro. Synthetic chloroplast in microfluidic droplets (2020). The THETA cycle (McLean et al., Nature Catalysis 2023¹⁵) — modular implementation of synthetic CO₂ fixation in living E. coli, two-orders-of-magnitude yield improvement via machine-learning-guided optimization. The first synthetic-to-nature CO₂-fixation pathway operating in a living host.

Verseux ZARM Mars-atmosphere cyanobacteria. Anabaena sp. PCC 7938 growing under simulated Mars atmosphere (96% N₂ / 4% CO₂ at 100 hPa) on MGS-1 regolith-derived medium, with biomass yields equivalent to Earth-ambient controls (Verseux et al., Frontiers in Microbiology 2021¹⁷). End-to-end demonstration of MGS-1 → cyanobacterial biomass → anaerobic digestion → fertilizer → 27 g fresh duckweed per 1 g cyanobacterial input.

ESA ARTHROSPIRA-B and ARTHROSPIRA-C. Limnospira indica PCC 8005 photobioreactor cultivation in microgravity aboard the ISS Biolab incubator, with online oxygen-production-rate and growth-rate measurement, demonstrating no first-order effect of microgravity on cyanobacterial growth in properly-mixed photobioreactors (Poughon et al.¹⁸). ARTHROSPIRA-C semi-continuous and multi-light-intensity operation in progress.

MELiSSA Pilot Plant (UAB, Barcelona). Five-compartment closed-loop life-support system in integrated multi-compartment operation with rats as model crew, targeting human-crew support in the late 2020s (Lasseur, Gòdia²²).

EDEN-ISS Antarctic greenhouse. 270 kg of fresh food across 26 crops in 9 months, 12.5 m² aeroponic cultivation area, autonomous nutrient-and-water delivery, between 2018 and 2023 at the Neumayer III Antarctic station.⁵

LanzaTech commercial gas fermentation. Clostridium autoethanogenum converting steel-mill, refinery, and biomass-derived waste gases into ethanol at >100 million gallons cumulative production across operating facilities in China, India, Belgium, and the United States.¹⁰ LanzaJet sustainable-aviation-fuel certification for commercial aviation.

Believer Meats Wilson, NC. First true industrial-scale cultivated-meat plant, 200,000 sq ft, 11,800 tonnes/year nameplate capacity, FDA “no questions” + USDA approval received late October / early November 2025, production launch end of 2025.²¹

NASA Deep Space Food Challenge Phase 3 finalists (August 2024). Interstellar Lab NuCLEUS modular bioregenerative system (US grand prize, $750K), Nolux artificial-photosynthesis dark chamber (UC Riverside,$250K), SATED cooking appliance (Boulder, $250K), Solar Foods (international winner).⁹ The follow-on Mars-to-Table challenge concluding September 2026.

The cumulative weight of these results is the relevant data point. Closed-loop molecular food synthesis is not a thought experiment. It is a research program with operating commercial implementations on multiple legs, with active space-environmental validation, and with the regulatory and acceptance bars cleared in multiple jurisdictions. The remaining work is integration, lipid biosynthesis at scale, and palatability — three legs that are tractable but unsolved.

Three risk scenarios

Honest planning for the program requires honest enumeration of the failure modes. We name three. They are not exhaustive. They are the ones the design has to take seriously.

Scenario A — Brittle integration

The first failure mode is the multi-compartment-failure-cascade scenario, the lesson of Biosphere 2 generalized. A closed loop integrating chemoautotrophs, methylotrophs, cyanobacteria, and higher plants is a system whose dynamics are not predictable from the dynamics of the individual compartments. A single contamination event, a single operating-parameter excursion, a single materials-of-construction interaction, can collapse the food output of multiple compartments simultaneously, with effects that propagate to atmosphere and water before the problem is noticed. The Biosphere 2 oxygen decline is the canonical example: nobody had modelled the concrete-to-calcium-carbonate sequestration; the failure was not a failure of any single compartment but of the inter-compartment chemistry that nobody had explicitly designed.

The implication is that ARCANE-class architectures have to be engineered with conservative biological-and-chemical buffers (atmospheric oxygen reserves, water reserves, food reserves), explicit redundancy across compartments (so that any one compartment can fail without the loop collapsing), and active monitoring with aggressive instrumentation. This is the discipline that the Agentic Systems and Physical Intelligence layers are coupled to — the multi-tool, oversight-capable autonomous systems that operate the loop have to do so at a level of biological and chemical literacy that today’s process-control systems do not approach. Failure of the integration is not a recoverable failure for an off-Earth deployment. The design has to take this seriously.

Scenario B — Palatability and adherence collapse

The second failure mode is the human-acceptance scenario, the lesson of the Biosphere 2 dietary deficit and of every long-duration submarine and Antarctic dietary record. Microbial-derived foods have texture, flavour, and aroma profiles that are different from conventional food, and the differences are not always favourable. Solein in its raw form is described in tasting reports as having a powdery, bland, slightly mineral character — adequate as a protein supplement, indifferent as a meal in itself. The processing required to produce a Solein-based food product that humans will eat with enthusiasm for years at a time is non-trivial, and the body of research on this question is small relative to the body of research on the underlying biology.

The risk is that the closed-loop architecture delivers a nutritionally-adequate food output that the crew will not eat at the necessary volume, producing the dietary-deficit pathway that Biosphere 2 documented. The mitigation is multi-pronged: precision-fermentation flavour and texture ingredients, cultivated-meat fat-and-flavour modules, cyanobacterial and higher-plant freshness-and-cultural-context inputs, and culinary-design work that takes the social and aesthetic aspects of food seriously rather than treating them as decoration. The honest summary is that the engineering of palatability is at least as hard as the engineering of nutrition, and the program has to be designed around this rather than around an assumption that nutritionally-adequate food is acceptable food. It is not.

Scenario C — Successful staged deployment

The third scenario, which we treat as the base case if the engineering and human-acceptance work are competent, is staged deployment in which the closed loop is closed in stages, with each stage validated at progressively longer duration and progressively larger crew before the next is deployed. The reference architecture is the MELiSSA pilot-plant trajectory: laboratory-scale single-compartment work, then bench-scale two-compartment integration, then pilot-scale multi-compartment integration with model-organism (rat) crew, then human-crew partial-closure work, then full closure for short duration, then duration extension. Each stage validates the assumptions of the next. The process is multi-decade and not glamorous. It is also the only process that has any track record of working.

The base-case deployment for ARCANE-class architectures, on this view, is not a Mars-direct megaproject. It is a sequence of Earth-side analogue deployments — Antarctic stations, deep-mine habitats, off-grid settlement contexts — that progressively validate the integration before any space deployment is attempted. The same architecture serves the climate-resilience question on Earth (food security under disrupted-supply-chain conditions, urban food production with low water and area footprint, post-collapse-scenario survival infrastructure) as well as the off-Earth question. The Earth-side deployments are not delays. They are the validation pathway.

What technical work bears on this

The reason ARCANE appears on a research-company website at all, rather than in a journal of bioengineering, is that the technical work is coupled to the rest of the civilizational stack in ways that are not always obvious. We pull three threads back from the food question to the broader technical agenda.

The first is that ARCANE depends on ENERA for the kilowatt-to-megawatt power loads that electrolytic hydrogen production, bioreactor mixing and gas compression, and (for plant compartments) grow-light operation require. Without abundant power, no closed-loop food. The energy economics of ARCANE are genuinely sensitive to the underlying electricity cost, and the path to sub-€1/kg microbial food on Earth runs through the same cost-of-electrolytic-hydrogen curve that is the bottleneck for green steel, green ammonia, and grid-scale long-duration storage. ARCANE is one of the larger non-trivial demand-side opportunities for cheap electrolytic hydrogen, which is itself one of the larger demand-side opportunities for cheap clean electricity. The food-energy coupling is not incidental. It is structural.

The second is that the bioreactor monitoring, process control, and multi-compartment loop optimization are exactly the kind of constrained-coordination problem that the Agentic Systems and Economic Orchestration layers are built for. A closed-loop food system at habitat scale has hundreds to thousands of monitored process variables, dozens of intervenable parameters, and a coordinator-of-multiple-tools task structure that is much closer to an autonomous-agent control problem than to a classical PID-control problem. The deployment surface for agentic systems with appropriate oversight discipline is the operating layer of ARCANE-class facilities. The same multi-tool composition, oversight, and verifiable-protocol problems that the AI Safety program addresses in the abstract are addressed concretely by the requirement that the operating system of the food loop be both autonomous (because human reaction time is too slow for many of the relevant control problems) and verifiable (because the cost of a control failure is the loss of the loop and the loss of the crew).

The third is that the microbiome-engineering surface — the engineered probiotic strains delivering specific metabolites to the human gut, the personalized-nutrition pipelines that match microbial-food output to individual physiology, the stress-and-aging-resilience strains that are the long-arc downstream of the INTEGRITISSUE program — is simultaneously food and medicine. The boundary between the food layer and the medical layer is, in closed-loop habitat contexts, much thinner than the regulatory framework on Earth has historically recognized. The convergent technical question — how to engineer microbial outputs that simultaneously meet nutritional, palatability, and health-outcome targets — is the same question on both sides of the boundary, and the program treats it as such.

The summary, drawn back from the technical layers to the food question, is this. A civilization that has solved the energy problem and the cognitive-substrate problem and the planetary-coordination problem but cannot reliably feed its inhabitants in resource-constrained or off-Earth contexts is a civilization that is not actually capable of operating outside the conditions in which it evolved. ARCANE is the food leg of the broader civilization-stack architecture. Without it, none of the rest matters at scale. With it, the rest matters more, because the constraints that have historically pinned humanity to the Earth’s surface — the requirement of an external photosynthetic biosphere providing all calories — are progressively replaced with closed-loop architectures that humans operate themselves. The technical work and the civilizational work are not parallel programs. They are the same program, seen from two angles.

Open questions in the field

The research-program agenda. We name nine. They are the questions the program is funded to address. They are not exhaustive of the field, but they are the ones the program treats as load-bearing.

1. Integrated closed-loop habitat food systems validated at ≥1-year, ≥6-crew scale. This is the MELiSSA Pilot Plant target and the natural extension to fuller closure with the modernized chemoautotrophic and methylotrophic compartments described above.
2. Multi-organism microbial consortia that convert CO₂ + H₂ + N₂ into a complete macronutrient profile (carbohydrate + protein + lipid + micronutrients) without external supplementation. The protein leg is solved; the lipid and micronutrient legs are not.
3. Lipid biosynthesis from one-carbon substrates at gram-per-litre titre, including essential fatty acids. Omega-3 EPA/DHA from formate or methanol at industrially-relevant scale is the most-overlooked nutritional gap. The Yarrowia lipolytica methanol-assimilation work is the relevant lead.¹³
4. Palatability and texture of microbially-derived foods at production scale — the “Solein tastes like sawdust” problem, addressed by precision-fermentation flavour ingredients, cultivated-meat fat modules, and culinary-design work that is currently funded at a tiny fraction of the underlying biology.
5. Mars-atmosphere-direct food synthesis at multi-crew scale, building on the Verseux ZARM cyanobacterial work.¹⁷
6. Plant cultivation under low-pressure, low-light, Mars-soil conditions at productive scale, building on EDEN-ISS Antarctic and ZARM analogue work.
7. Cultivated-meat bioreactor scaling from the current 25,000-litre commercial scale to >100,000 litres while maintaining cost competitiveness. The Believer Meats Wilson facility is the current state-of-the-art and the relevant cost-curve data point.²¹
8. Robust closed-loop water recovery at >99.5% efficiency, including the recovery of trace contaminants that current ECLSS systems either remove or vent.
9. Engineered photosynthesis (CETCH/THETA-class) at industrially-relevant titre in living cells. The Erb-lab work has crossed the in-vivo demonstration threshold; the next step is commercial-scale titre.¹⁴¹⁵

Each of these is a multi-year research effort. None of them is solved. All of them are tractable. The ARCANE program funds work on each, in collaboration with the laboratories named above and with the open scientific community that is the natural home of this work.

Connections

ARCANE depends on ENERA for the electrolytic hydrogen and bioreactor power; without abundant power, no closed-loop food. It feeds into INTEGRITISSUE through the microbiome-engineering and personalized-nutrition surface, where engineered probiotic strains delivering specific metabolites to the gut are simultaneously food and medicine. The autonomous bioreactor monitoring and multi-compartment process control is the deployment surface for Agentic Systems and Physical Intelligence, with the same multi-tool composition and oversight problems that AI Safety research is addressing in the abstract. The mass-balance, water-recovery, and feedback-loop optimization is exactly the kind of constrained-coordination problem Economic Orchestration is built to handle at scale. The food layer is not an isolated program. It is a deployment surface for several of the others, and a binding constraint on what the rest of the stack can plausibly support.

Where to read further

If this framing is useful, the related pieces of the stack complete the picture. ENERA treats the energy infrastructure that ARCANE is downstream of. INTEGRITISSUE treats the biological resilience and microbiome-engineering surface that ARCANE feeds into. The Agentic Systems and Physical Intelligence research pillars treat the autonomous-control and bioreactor-deployment substrate. Economic Orchestration treats the supply-chain and resource-allocation question — who routes the inputs and the outputs, at what cost, with what incentive properties. The manifesto provides the broader architectural framing within which all of these sit.