Protein from renewable resources: mycoprotein production from agricultural residues

Protein fermentation, intended for human consumption, from agricultural residues has been shown to be a potential viable process from a technical, economic and sustainability perspective.


Introduction
The projected 50% increase in global population combined with Global South growth is expected to bring over a 50% rise in food demand by 2050 with global animal-sourced protein demand estimated to nearly double (1). Despite the efforts to tackle malnutrition, one in nine people (821 million) in the world are still undernourished (2) and current trends suggest that global malnourishment is increasing. Furthermore, the positive feedback loops posed by the effects of Climate Change (3) may risk severe disruption to agricultural activities and threaten protein security as an estimated 56% of global protein nutrition is derived from crops in which the protein content is particularly sensitive to elevated CO 2 levels(4). Furthermore, plantsourced and animal-sourced protein contributes significantly to greenhouse gas (GHG) emissions and is highly dependent on arable lands. Notably, the global livestock sector is responsible for substantial GHG emissions and is responsible for 30% of terrestrial land use for grazing and compound feeding (5). Research has emerged in recent years on alternative protein sources with minimal environmental damage, in particular food-grade single cell proteins (SCPs). SCPs refer to the protein derived from microorganisms, such as bacteria, algae and fungi that can be used as a protein supplement for human consumption or animal feeds. Animal-feed SCP research has utilised feed-safe waste such as agro-industrial residues; sugarcane molasses (6), soy molasses (7), wheat bran, (8,9) starch processing wastes(9, 10) and mixed-food industry wastes (11). Whereas, food-grade SCP are mainly sourced from arableland-dependent food crops due to strict regulatory requirements (12) and predominantly used as food-supplements rather than meat-substitutes. Saccharomyces cerevisiae and Pichia jadinii are examples of successful SCPs produced from starch-derived glucose for human consumption. There has been an increasing research interest in alternative non-food feedstocks such as carbon dioxide (13) which were investigated for SCP for human consumption on a technical level (13,14). However, food-grade SCP production from non-crop feedstocks, which are independent of arable lands and offer low-carbon solutions, remains an open area for investigation.
This study presents exploratory research on Mycoprotein derived from food-grade lignocellulosic agricultural residues, which offers potential sustainable solutions for future protein security. Mycoprotein, developed in response to global protein deficiency since 1967, is derived from the fermentation of glucose with Fusarium venenatum A3/5 and has been produced by Marlow Foods Ltd under the trade name Quorn™ since 1985, with current annual sales over £200m (14,15). Containing high protein (45 % mass) and essential amino acids (44% of total protein), Mycoprotein offers a range of positive attributes compared with animalsourced protein such as favourable fatty acid profile and high fibre content. However, its starchderived glucose sourced from food crops is still dependent on arable land. Alternatively, lignocellulosic-sourced sugar offers a promising solution. Cellulose the most abundant organic material on earth, can be derived from agricultural residues and other food-safe lignocellulosic resources such as rice straw. Rice as a staple crop feeds over 45% of the world's population (16). With continuous global growth in rice demand, there is a corresponding increase in rice straw production. Much of this straw is deemed as waste and is burned, leading to severe health complications, adverse impacts to the atmospheric quality and soil pollution (17). Several pretreatment technologies have been investigated to fractionate lignocellulose enabling glucose production, amongst which, the use of ionic liquids (ILs) is one of the most promising technologies due to its high sugar release (18). Furthermore, the use of promising low-cost food-grade ILs may eliminate many hurdles relating to food safety and enabling lignocellulosic-SCP. Despite the advancements in lignocellulosic fractionation and biochemical production, no publicly available research has been found on food-grade lignocellulosic-glucose extraction for SCP production.
For establishing such a process at industrial scale, biorefinery modelling is an important area of research, however, biorefinery models with IL pretreatment are a scarcity. Baral and Shah (19) investigated IL pretreatment for ethanol production and concluded that IL recovery and cost were significant economic factors. SCP production modelling also represents an open research area with very few published studies focusing on feed-grade SCP. Strong,Kalyuzhnaya (20) discussed the potential for a methanotroph-based multiple product biorefinery which considered protein production via a biogas-derived methane-fed fermentation. Their modelling suggested a technically feasible system producing 0.54 tonnes of SCP per tonne of methane feed. Molitor,Mishra (13) presented an economic viability study for a two-stage power-to-protein process which fixed CO 2 in the presence of electrolysisderived H 2 in the first stage, followed by S. cerevisiae fermentation on the acetate effluent from stage A. The biorefinery process proposed by Aggelopoulos, Katsieris (11) for solid state fermentation of food industry wastes to feed SCP revealed economically competitive wastederived SCP (~$460-550/tonne vs $425/tonne (soybean meal))(21). However, a research gap remains on modelling of lignocellulosic-derived food-grade single cell protein.
In this work, through both experimental investigation and modelling, we investigate, for the first time, the technical feasibility and sustainability of lignocellulosic-SCP, in particular, lignocellulosic-mycoprotein production. We employ food-grade ionic liquids for sugar extraction from rice straw to assess the pretreatment performance of the ionic liquid. Our experimental data then forms the basis of the biorefinery model developed to assess the technoeconomic feasibility and environmental sustainability of the process. Finally, the model was probed to highlight the key variables for process optimisation and intensification. The novelty of this exploratory research is based on the particular application of food-grade ionic liquids and enzyme in the treatment of rice straw with the aim of valorisation of the recovered fermentable sugar through producing mycoprotein for human consumption.

Methodology
We experimentally investigated food-grade lignocellulosic sugar production from rice straw using the food-safe ionic liquid, [Ch][HSO 4 ], and cellulase enzyme Celluclast 1.5L where a non-food grade ionic liquid ([TEA][HSO 4 ]) and enzyme (Novozyme CELLIC ® CTEC 2) were used as a benchmark. The methodology of this paper ( Figure 1) consists of laboratory experiments and modelling. The data from the experimental work formed the basis of the process design and modelling. The developed modelling framework consists of three components: process design and simulation, techno-economic analysis, and sustainability performance.
The process was synthesised in the process simulation software Aspen Plus V9. Coupling Aspen Plus with Microsoft Excel and MATLAB, the economic performance of the process including capital and operating costs were evaluated. Subsequentially, the Minimum Selling Price of the protein product was determined as part of the economic evaluation of the process. Finally, a life cycle assessment of lignocellulosic-mycoprotein was conducted in the software SimaPro to evaluate the sustainability performance.

Compositional analyses of rice straw samples
The straw samples from three rice varieties cultivated by International Rice Research Institute (IRRI) Philippines were sent to the UK for analyses. Prior to compositional analyses, samples were ground using a Retsch (M200) cutting mill and sieved to a defined particle size of between 180-850 µm to achieve a homogenous mixture of samples. The moisture content of each ground biomass sample was determined by removal of a subsample and oven-drying at 105°C overnight (Supplementary S.1.1). The chemical composition analyses for cellulose, hemicellulose, lignin and ash contents of the untreated samples were conducted by Celignis using Near Infrared Analysis (NIR) following P10 analytical methods developed by Hayes (22), (23). The compositional analyses results were used as data input for process simulation. Similarly, chemical composition analysis of the pretreated pulp was conducted by Celignis using the same protocol.  Figure S.1). We conducted ionoSolv pre-treatment(24) (Figure S.2) to fractionate rice straw samples into three distinct components -a cellulose-rich pulp, lignin, and hemicellulose liquor. The hemicellulose and lignin were dissolved during the process into the ionic liquid, from which lignin was recovered with the addition of water as antisolvent.

Lignocellulose fractionation to determine ionic liquid performance
To bridge the knowledge gap on the optimum condition for pretreatment of lignocellulosic biomass using [Ch] [HSO 4 ]. we have designed experiments based on our previous research on other ILs, where effective fractionation could be achieved at 170 °C for 30 min (25). However, reports concerning the general thermal stability of choline based ionic liquids have demonstrated that degradation will begin to occur at temperatures above 150 °C, although this is anion dependent. Therefore, due to concerns for the stability of the ionic liquid, a lower temperature was selected (150 °C for 1 hour) than previously published optimal temperature on grass-type lignocellulosic biomass.
IonoSolv fractionation experiments were performed in triplicates, where both ionic liquids ([Ch] [HSO 4 ] and [TEA][HSO 4 ]) were investigated. Each fresh straw sample equivalent to 1g oven dry weight (ODW) with 9 g ionic liquid was placed into a 15 mL pressure tube and thoroughly mixed using a vortex mixer. Pressure tubes were placed in the oven at 150˚C for 1 hour. After cooling for 15-20 mins to room temperature, the sample was transferred to a 50ml centrifuge tube and washed with absolute ethanol 4 times. At each wash cycle, 40 ml ethanol was added to each tube; after mixing with the vortex mixer, they were rested at room temperature for one hour. Next, the samples were centrifuged (VWR Mega Star 3.0 centrifuge) at 3000 rpm for 50 minutes. Supernatants were removed at each wash whereas pellets (i.e. pretreated pulp) remained in the centrifuge tube. After 4 washes the pulp was transferred to a Whatman cellulose thimble for soxhlet extraction with 150 ml of absolute ethanol at 135˚C for 20 hours in a fume hood. After the completion of the extraction the ethanol was removed, and the contents of the thimbles were transferred to a pre-weighed 50 ml centrifuge tube for two water-wash cycles. Water and pulp mixture were centrifuged at 3000 rpm for 50 minutes in each cycle. The water was removed via decanting and the tube containing pulp was weighed to determine the moisture content of the pulp and the cellulose pulp recovery yields calculated.

Saccharification assay to recover lignocellulosic glucose
Saccharification assay was performed in triplicates to determine the yield potential of foodgrade lignocellulosic glucose from rice straw. Ionic liquid pre-treated pulp equivalent to 100mg ODW was incubated in a 15ml sterile plastic vial with 5 ml sodium citrate buffer (0.1M at pH 4.8), 40μl Tetracycline antibiotic (10mg/mL in 70% ethanol), 30μl Cycloheximide antibiotic (10mg/mL in purified water), 3.4ml of purified water and 20μl cellulase enzyme (Novozyme CTEC 2 or Celluclast 1.5L). A reaction blank was prepared containing the aforementioned reagents. Samples were incubated for 7 days at 50°C at 250 RPM in a Stuart Orbital Incubator. The glucose concentrations were determined with the HPLC (Shimadzu Prominence) equipped with a UV detector (SPD-M20A) as well as a RI detector (RID-10A). The installed column was the Aminex HPX-87P produced by Bio-Rad, operated at 85 °C with deionized water as the mobile phase with a flow rate of 0.6 mL /min.

Elemental analyses to determine Fusarium venenatum biomass composition
To bridge the knowledge gap on the Fusarium venenatum A3/5 biomass composition, we performed elemental analysis to provide data for process simulation. Live fermentation broth was collected from fermenter at Quorn production site and was frozen at -30 C. The sample was then freeze-dried for 2 days in a vacuum at less than 0.01 bar at room temperature. Elemental analyses were performed by OEA Laboratories Ltd using an automatic combustion elemental analyser (Thermoquest EA1110 elemental analyser) to determine carbon (C), hydrogen (H), nitrogen (N), oxygen (O), sulphur (S) and phosphorus (P) content alongside other elements (sodium (Na), potassium (K), ash content). The detailed experimental methods of this analysis are given in Supplementary S.1.3.

Process Design and Simulation
Based on our experimental results supplemented by publicly available data, we modelled a lignocellulosic-mycoprotein biorefinery with process simulation software Aspen Plus V9. The production capacity was based on an estimate of the current annual production capacity obtained from literature (13,26) and information regarding the capacity increase due to new investments at the production site (27). Overall, the expected capacity, and thus the production rate used in this work is 40,000 tonnes/year. The process design and simulation aimed to understand the technical feasibility to scale-up Mycoprotein production using foodgrade lignocellulosic sugar derived from rice straw. As shown in Figure 1, the process was decomposed into distinct processing areas, of which, pretreatment, separation and fermentation are a primary focus of this model whilst the other sections are adapted from previous work (28). The design for each processing area is detailed in result section 3.2. The origin of the physical properties for the compounds used in this work are described in (

Economic Evaluation
We used the Minimum Selling Price (MSP) to assess the economic viability of the process to produce lignocellulosic-mycoprotein from rice straw. The capital costs, denoted as Fixed Capital Investment (FCI), were determined based on detailed sizing and costing of each process section described in the process area (Supplementary S.2.1). The input and output streams of reagents, utilities, products and byproducts contribute to the annual operating costs (Supplementary S.2.2). To determine the MSP accurately, we split the process into upstream and downstream segments. For the 'upstream' process, which involves pretreatment of rice straw, enzymatic hydrolysis and subsequent fermentation to mycoprotein paste and the auxiliary facilities, we calculated a MSP of the paste (MSP paste ) by finding the MSP of the paste that makes the upstream Net Present Value (NPV) 0 (Supplementary S.2.4. and Supplementary Figure 1. Simplified schematic of research scope of this article divided into experimental investigations, process design and simulation, techno-economic analysis, and sustainability analysis. Within process design and simulation, the main process areas are shown and the input and outputs to the process are shown with solid arrows. The process was synthesised in the process simulation software Aspen Plus V9. Coupling Aspen Plus with Microsoft Excel, the economic performance of the process including capital inputs and operational inputs were evaluated. Subsequentially, the Minimum Selling Price of the protein product was determined as part of the economic evaluation of the process. Finally, a life cycle assessment of lignocellulosic-mycoprotein was conducted in the software SimaPro to evaluate the sustainability performance. S.2.5). As the Mycoprotein paste is not the final supermarket product, and in the interest of finding the true cost of lignocellulosic-mycoprotein to the consumer, we pursued an estimate of the MSP of the final product (MSP product ) by analysing the published Quorn revenue figures. Furthermore, we made a prediction of the current production price of the mycoprotein paste using the same NPV analysis based on several assumptions relating to the input materials to the process. These are summarised in Supplementary S.2.6.
The process design for the conversion of rice straw to mycoprotein paste excludes the subsequent texturisation steps required to produce the final consumer product. In this study, the lignocellulosic mycoprotein paste was modelled as the crude protein and compared to other crude protein sources, namely, livestock. The crude protein contents of mycoprotein and livestock-sourced protein are given in Supplementary S.2.7.
To investigate the effects of certain process parameters, economic assumptions, a sensitivity analysis was carried out considering an upper and a lower bound of the design space. The bounds of most process parameters were selected based on representative literature data; whereas for the price of cellulase enzyme and rice straw were varied ±20% of the base-case value.

Surrogate-based Retro-Techno-Economic Analysis
In this study, a novel surrogate-based retro-techno-economic analysis (RTEA) was developed to identify the feasible design space for performance optimisation of the lignocellulosic-mycoprotein process. Reversing the TEA process i.e. determination of economic key performance indicators (KPI), RTEA (29-31) allows the declaration of a minimum acceptable KPI, in order to identify the combination and value of operating conditions and process targets (feasible design space) that would result in the desired KPI e.g. MSP. Thus, RTEA offers a powerful tool to provide performance targets enabling process design to effectively screen alternatives and focusing on performance-limiting steps. RTEA has been applied in previous research to equation-orientated simulators (29)(30)(31) in which the economic model is implemented explicitly in equation form, and the model converges only when all equations are satisfied. In this study, the simulation and design were conducted with a sequential-modular programme, where iterative procedures are used to converge the flowsheet. Thus, we employed a novel RTEA approach. Kriging surrogate models(32) were constructed using the Surrogate Modeling (SUMO) Toolbox in Matlab (33). A latin hypercube sampling method with corner points were selected as initial design where the sample space was constructed with a combination of input variables (cellulase dosage, rice straw cellulose and xylan composition, and saccharification yield) and output variable was MSP. In all cases, the feedstock flowrate was fixed at 378,000 tonnes/year.

Life Cycle Assessment (LCA)
To assess the cradle-to-factory gate environmental sustainability of the lignocellulosicmycoprotein process, we conducted a life cycle assessment using SIMAPRO V9. The inputoutput flows derived from process simulation were fed into the LCA as inventory data. The inputs and assumptions for the Life Cycle Inventory (LCI) are detailed in Supplementary S.3 Through contributional analyses, LCA comparison and sensitivity analyses, we aimed to a) identify the main environmental contributors and performance-limiting processes and outline potential improvements, b) compare lignocellulosic-mycoprotein with food-crop derived mycoprotein and traditional protein sources, highlighting benefits and drawbacks, and c) investigate the most sensitive parameters on the environmental performance of the lignocellulosic-mycoprotein system.
The system boundary defined in this 'cradle-to-gate' study is shown by the black dashed line in Figure 1 and includes the biorefinery process, converting rice straw to mycoprotein, as well as cultivation of rice, and ionic liquid synthesis. The functional unit was defined as 1kg Mycoprotein paste at biorefinery gate, with a solids content of 25% based on a production capacity of 40,000 tonnes/year. We applied an economic allocation approach where multipleproducts occur at feedstock cultivation (rice grain and straw) and lignocellulosic-mycoprotein production stages (mycoprotein paste and furfural). A problem oriented (midpoint) approach ReCiPe 2016 Midpoint (hierarchist version) was adopted as the baseline Life Cycle Impact Analysis (LCIA) method. Under this approach, the impact categories of acidification, climate change, depletion of abiotic resources, ecotoxicity, eutrophication, human toxicity, ozone layer depletion, particulate matter, and photochemical oxidation. Land use consumption characterisation methods are detailed in Supplementary S.3.1 to account for arable and pastureland occupation by arable crop cultivation and livestock grazing/feeding to meet protein demand including animal-sourced and plant-based protein.
Sensitivity analyses were conducted to understand the implications of varying process parameters and economic assumptions on LCIA results. The LCA results of lignocellulosicmycoprotein derived from rice straw was compared to potential lignocellulosic-mycoprotein derived from other sources, namely, miscanthus, switchgrass and corn stover to highlight sustainability improvements that could be attained by switching feedstock.

Process design
Based on our experimental work, we modelled a lignocellulosic-mycoprotein biorefinery (40,000 tonnes/year) using Aspen Plus V9. The design overview is presented in Figure 3 and below with more detail in (Supplementary S.5).

Feed Handling
Milled rice straw is delivered via the 'uniform-format feedstock supply system'(34) at 20% moisture content. Therefore, the dry-composition value of rice straw was adjusted to account for the moisture content in the process model. In modelling the biomass, a normalised average ash content of 17.5%(35) (Table S.21) was used. In the elemental analysis, acetyl groups were not explicitly measured, however, hemicellulose acetyl groups account for between 1-2% of cereal straws (36). In our study, 1.7% was assumed. The reception and processing of the milled straw is based on previous models (37) where rice straw is delivered from delivery trucks to receiving bins at the pretreatment reactor. At the same time, recycled ionic liquid is conditioned with fresh water and ionic liquid to achieve a specified water loading of 20% and sufficient quantity of solvent mixture for pretreatment (rice straw =10%, solvent=90%). More details about feedstock handling are detailed in Supplementary S.5.3.

Pretreatment
The rice straw and the ionic liquid solvent mixture (ionic liquid/water: 80%wt/20%wt) are mixed before entering the reactor. Stoichiometric equations define the conversion of rice straw constituents (Table S. 22). Conversions of the feedstock constituents (cellulose, xylan, mannan, galactan, arabinan and lignin) were calculated based on the conversions that would result in the pulp composition of the samples that was determined experimentally (Section 3.1 and Table S.15). Previous experimental results reveal degradation of xylose to furfural in acidic conditions at longer pretreatment times (38). Although the pretreatment time in this work is relatively short, it was assumed that all xylose released from the straw was further converted to furfural, this then provides an optimistic case for furfural recovery for sale as a co-product. The reactor effluent is immediately flashed in a tank at the reactor temperature to utilise the simple heat of the stream to evaporate volatiles (furfural, HMF, acetic acid) and some water, which is sent to furfural recovery. This reduces the volumetric flow through the rest of the pretreatment sequence, reducing energy demand in the evaporation of water from the ionic liquid as discussed later. After the flash, the liquid stream enters a thickener followed by a filter to remove the IL-rich liquid from the solid pulp; subsequentially, the pulp enters a countercurrent belt washer which was modelled to reflect current pulp washing technology(39) producing a washed pulp stream sent to enzymatic hydrolysis, and a stream containing the recovered solutes.

Enzymatic Hydrolysis
Cellulases are used to hydrolyse cellulose to produce a 10% glucose concentration for feed to fermentation (reaction and conversion detailed in Supplementary Table S.23). The design is based on previously published models (28). The washed pulp from pretreatment is heated to 48°C. The pulp is first partially saccharified to less than 20wt% solids so that the slurry is pumpable, followed by saccharification in a batch reactor for 72 hours. Afterwards, the hydrolysate is filtered and passed through a triple-effect evaporator to recover a liquid hydrolysate containing soluble sugars with a specification of maximum 10% glucose. The solid cake is sent to the boiler for energy recovery.

Fermentation
The fermentation unit for mycoprotein production was modelled, in which the definition of the reactor and the required reagents were based on the continuous air-lift fermenter currently in operation at Quorn Foods (40). The diluted glucose stream is heated to 30°C and pumped to the fermenter. Air, ammonia and a nutrient medium are added to the fermenter to assist F. venenatum A3/5 growth. Elemental analysis data described in Moore, Robson (40) combined with our elemental analysis of the macro-elements were used to determine the stoichiometric requirement of reagents (Supplementary S.5.6). The main elements required for biomass growth are carbon, hydrogen, oxygen, nitrogen and phosphorus, which are supplied by glucose, O 2 , ammonia and phosphoric acid. The remaining 4% of the biomass consists of potassium, sodium, calcium, magnesium, iron, copper, manganese and zinc, therefore, this provided the basis to determine the remaining nutrients for the growth medium (Supplementary S.2.2.2). The compounds considered in the nutrient mix were based on study by Moore,Robson (40), with any missing compounds taken from those used in Vogel's medium (41). After the fermenter, steam is added directly to the stream and the temperature is raised to 68°C to simulate RNA reduction. which is necessary to reduce RNA content to avoid health effects due to excessive RNA consumption in the diet. As a consequence, 30% of biomass is lost. Subsequentially, additional steam raises the temperature to 90°C, before centrifugation to produce a mycoprotein paste containing 25% solids (cell mass), and a liquid centrate. The centrate has potential for concentration as a flavouring product (14), however, in our design it is sent to wastewater treatment for energy recovery.

Separation
Three separation steps can be distinguished: lignin precipitation, IL recovery, and furfural recovery. The ionic liquid-rich filtrates from pretreatment are treated with anti-solvent (water) to precipitate lignin to recover solid lignin which is washed with water, and sent to the boiler for energy recovery. The IL rich stream is diverted to IL recovery, and the spent-wash stream is sent to wastewater treatment. Small amounts of glucose and hemicellulose sugars are also in solution with lignin at the precipitation stage. In this design the sugars are also assumed to be precipitated in the same stage as the lignin, therefore, in a similar fashion to Klein-Marcuschamer, Simmons (42), this separation is partially defined. In future work, further investigation, both experimentally and from a design perspective should be carried out.
The IL rich stream is treated through a three-effect evaporator to reduce the water content to 20%. The vapour stream from the first effect is mixed with the vapours from the pretreatment flash. This combined stream is the feed to the furfural recovery section. The vapours from the second and third effects are sent to wastewater treatment.
A minimum-boiling azeotrope, which exists between water and furfural at 97.78°C (1 bar), complicates the recovery of furfural. However, the presence of acetic acid acts as an entrainer itself to facilitate distillation via heteroazeotropic distillation by taking advantage of the vapour-liquid-liquid equilibrium formed between the 3 components. Previous work has investigated this separation problem (43,44), achieving a 95% pure furfural product. In this case 99% purity is desired, therefore an extra column was employed. The sequence used in this work involved a first column of 7 stages with no reboiler where the distillate splits into a fufural rich phase and a water rich phase in the decanter, from which the fufural rich phase is sent to a smaller column (5 stages) where the purity is increased from 95% to 99% and is recovered in the bottoms. The distillate is recycled to the first column. There is an opportunity for recovery of the acetic acid with an alternative distillation configuration, however, Galeotti, Jirasek (44) concluded that it was not economically justified.

Auxiliary Sections
The auxiliary sections of the process design include wastewater treatment (WWT), the combustor, and utilities. These sections are based on previous designs and specifications (28). Wastewater treatment includes an anaerobic digestion step, producing biogas, followed by aerobic treatment, which produces a clean water stream that is recycled back to the process, and a solid sludge that is sent to the combustor (28). In the combustor, the solid waste streams produced from pretreatment (lignin), enzymatic hydrolysis (non-hydrolysed pulp), and the solid sludge produced from wastewater treatment undergo combustion to generate heat for the on-site production of steam and electricity. Fly-ash and stack gas are produced. Finally, the utilities include the process water, cooling tower, chilled water and electricity requirement for the plant. A cooling tower system is used to provide cooling water at 28°C with a maximum temperature change to 37°C, with a tower blowdown of 0.15%. The chiller system required 0.56kw/ton of refrigeration for the compressor electricity demand.

A B
The Fixed Capital Investment for Scenarios 1, 2, 3 and 4 is $756m, $737m, $575m and $575m respectively ( Figure 4A; Table S. 26), and total annual operating costs are $125m, $122m, $92m and $92m respectively ( Figure 4B; Table S.27). Overall, these costs are largely dependent on the pretreatment performance, with significant economic improvements due to the lower pulp yield (thus greater lignin and hemicellulose removal), which facilitates saccharification to glucose. The choice of enzyme also affects costs, with the food-grade CELLIC ® CTEC2 enzyme scenarios being economically inferior to the equivalent non-food grade cellulase scenarios, however, the effect is less pronounced than the degree of pretreatment.
The MSP paste reflects the trend in costs, with the resulting prices of $5.04/kg, $4.95/kg, $3.78/kg and $3.78/kg for scenarios 1-4 respectively. This reveals that significant improvements in the MSP paste can be achieved with selection of a high performance ionic liquid solvent. However, of greatest interest is the present possibility of lignocellulosic-mycoprotein production, in which case, food-grade reagents are a necessity, therefore, scenario 1 is the focus of our analyses.

Sensitivity analysis
The sensitivity analysis is presented in Figure 5 (Table S.28), in which a set of process parameters and economic assumptions were varied to understand their implications on the system performance of food-grade lignocellulosic mycoprotein (expressed as MSP). Sensitivity analysis results revealed that the maximum variation in price was +1.40 $/kg and -0.85 $/kg (-16.9%/+27.8%). The dominant parameter influencing the MSP was the uncertainty in the overall capital investment, which at this level of process design is generally ±25% (45). The other economic assumptions (i.e. discount rate, tax rate and applied operating costs) are sensitive parameters, producing significant effects on the MSP, highlighting the need for detailed economic analysis. The most sensitive process parameter is the saccharification yield of glucose from rice straw. This demonstrates the importance of ionic liquid selection and pretreatment performance for the process viability.
The process variables with the greatest impact are the saccharification yield and the potential utilisation of xylose in tandem with glucose by F. venenatum A3/5. The scenario with greater glucose saccharification yield is modelled to represent the improvement in food grade ionic liquid performance or enhancement in cellulase activity. Xylose utilisation represents a promising future scenario with an increased carbon utilisation efficiency by F. venenatum A3/5 where the strain is capable of growth on a mixed sugar substrate derived from lignocellulosic resources. This alters the aim of the pretreatment process design and ionic liquid selection. For the case in which only glucose can be utilised by F. venenatum A3/5, pretreatment should aim to remove both lignin and hemicellulose using high-performing IL e.g.
[TEA] [HSO 4 ]. However, if xylose can be utilised, pretreatment should target lignin removal and limit hemicellulose loss which can be achieved by the low-performing food grade [Ch] [HSO 4 ]. Our modelling results suggest the potential degree of process improvement and highlight the future research frontier for lignocellulosic mycoprotein technology, i.e. ionic liquid performance and F. venenatum A3/5 growth on a wider variety of sugar substrates.
The impacts of rice straw prices on the MSP product suggest the importance of feedstock screening in particular, considering the dependence of agro-crops on environmental variables (e.g. climate). Therefore, an ideal feedstock would be the lignocellulosic plant species with resilient traits towards environmental change and other extreme events. The ionic liquid price has been factored in with a range of $0.5/kg and $25/kg to represent the general cost of various ionic liquids. The extent to which the MSP is affected by ionic liquid price is reduced with ionic liquid recycling. However, at a high price of $25/kg, solvent costs could hinder economic feasibility significantly. Fortunately, protic ionic liquids, as demonstrated in our study, offer a cost-effective alternative to costly ionic liquids (46)(47)(48).Finally, benefiting from the high recycle rate of ionic liquids, biomass loading of the solvent has a negligible effect on the MSP, Water loading of the solvent has a greater effect, however, this model does not account for any change in pretreatment performance due to the change in water present, as has been found in experimental studies.  Table S.28.

Surrogate-based Retro-techno-economic analysis
A surrogate-based RTEA was performed to highlight the feasible design space of lignocellulosic-mycoprotein processes, where key variables such as biomass composition, ionic liquid selectivity, cellulase dosage, and carbon source utilization by F. venenatum were investigated. Kriging surrogate models were constructed by running the model with 54 samples spanning the design space and the corresponding resulting MSP. The surrogate models were then used to conduct the RTEA. The Training data and model details can be found in Supplementary S.29-S.32. The results of the RTEA are shown in Figure 6.
In Figure 6A, contours representing different MSP prices are shown for combinations of saccharification yield and cellulase dosage. The primary driver for economic improvement in this scenario is the saccharification yield. Trivially, for a single cellulase mixture, as dosage increases one would expect the saccharification yield to increase. However, in this case, Figure   Figure 6. Results of the retro-techno-economic analysis (RTEA). A surrogate model of the process was created for certain input variables to find the response of the Minimum Selling Price (MSP) for a change in the input variables. Figures 6A, 6B and 6C are for a two variable RTEA in which the variables on the axes were varied within their displayed ranges. The contour lines for A, B and C show a number of MSP targets ($/kg) and the arrow represents the direction of improvement of the MSP. The solid lines of Figure 6C are for the scenario where cellulose and xylan composition of rice straw is varied where xylose cannot be utilised by F. venenatum.The dashed lines represent the scenario in which xylose can be utilised by F. venenatum. Finally, Figure 6D is a three variable RTEA where rice straw cellulose and xylan composition were varied as well as the saccharification yield of cellulose and xylan to glucose and xylose respectively. For this scenario, F. venenatum utilisation of xylose was assumed. The solid contour lines show the cellulose and xylose compositions required at selected saccharification yield (80% and 90%) to achieve an MSP than is no more than 100% of the predicted current minimum selling price (CMSPsee Supplementary S.2.6 for derivation). 100% increase was chosen as there is no combination of variables (feasible region) capable of achieving the CMSP. 6A can be used as a tool to cross-reference with existing cellulase formulations to identify alternatives that can be used to reduce MSP.
Significant improvements in the MSP can be achieved by combining a high cellulose composition and saccharification yield ( Figure 6B), which maximises the glucose yield per unit of feedstock. Figure 6B also offers a tool to quickly screen potential alternative feedstocks and ionic liquid pretreatment combinations based on experimental results in literature to highlight their potential for implementation in the lignocellulosic-mycoprotein process to reduce the MSP. Finally, a three variable RTEA was conducted where saccharification yield was also investigated alongside feedstock cellulose and xylan composition ( Figure 6D) as these are the most promising parameters to reduce the MSP. In this scenario xylose utilisation by F. venenatum was also assumed. The (predicted) current minimum selling price (CMSP) of the paste (see Supplementary S.2.6 for derivation) was used as a benchmark to judge how competitive lignocellulosic-mycoprotein could be. No feasible region could be determined for achieving the CMSP . Therefore, the RTEA was run for a 100% increase of the CMSP. This was selected as it was deemed to be potentially acceptable by consumers for a product that could be produced with a reduced environmental impact (to be discussed later). As can be seen, even at a modest 80% saccharification yield, high cellulose feedstocks could be used to achieve that is no greater than twice the CMSP. If 90% yields could be obtained, then feedstocks with a cellulose content greater than 43% would be promising.
Overall, our RTEA modelling results suggest the potential degree of process improvement through improvement of certain process variables and highlight the future research frontier for lignocellulosic mycoprotein technology, i.e. ionic liquid performance and F. venenatum A3/5 growth on a wider variety of sugar substrates.

Comparative protein sources
In this study, the mycoprotein paste was modelled as the crude protein and compared to other crude protein sourced from livestock. Therefore, lignocellulosic-mycoprotein was compared both economically and environmentally to livestock at the slaughterhouse gate.
From an economic perspective (Figure 7), lignocellulosic-mycoprotein is comparable to Lamb and Beef on a mass basis, however on a protein basis (Table S.7) the cost of lignocellulosic-mycoprotein is approximately twice that of beef, and almost 5 times more expensive than chicken. At supermarket prices, the predicted consumer price of the final mycoprotein product ($21.80/kg / $173.02/kg-protein) follows a similar trend compared to the cost of livestock protein (Table S.8). On the other hand, when we compare this price to an average price across a range of current Quorn products, the cost of lignocellulosic-mycoprotein is predicted to be around 1.6 times the current average Quorn supermarket selling price.   Figure 8A, from a sustainability perspective, lignocellulosic-mycoprotein delivers significant advantages over beef in all impact categories except water consumption and delivers similar environmental footprint to chicken on global warming and terrestrial acidification. However, in comparison to animal-sourced protein and plant-based protein (Tofu), lignocellulosic-mycoprotein shows significant environmental benefits in arable land-use. Thus, lignocellulosic mycoprotein offers a promising protein solution which is largely independent of arable land in comparison to landdemanding protein sourced from animal and arable crops (e.g. meat and tofu).
The performance-limiting steps over the lignocellulosic-mycoprotein life cycle vary between impact category. Straw production dominates the impacts on water consumption (75.7%), due to rice irrigation. A less water-demanding feedstock would offer a mitigation solution to reduce water consumption.
For global warming, total GHGs of 516 kg CO 2 eq/tonne protein are attributed to process emissions, after accounting for biogenic emissions (Supplementary S.8), that is to say, 516 kg CO 2 eq/tonne of is emitted that does not originate from biogenic sources, representing only 2.4% of overall contribution to global warming. The main contributor is the emissions due to external electricity use and production representing 72.9% of emissions. Therefore, a potential decarbonisation solution would be to integrate renewable energy resources in lignocellulosic mycoprotein system to target over a third of energy-related emissions.
For terrestrial acidification the largest contributions to overall acidification are electricity and process emissions which contribute 58.4% and 26.3% respectively. In the case of marine eutrophication, 67.58% is due to straw production.
In terms of arable land, over 95% is associated with rice straw production which is due to allocating some of the rice growth burden to the rice straw itself (1.56% of total impact of rice production). However, as the straw is considered a waste product of rice growth, it is worth considering the change in impact if no burden is allocated to the straw. Therefore, sensitivity analyses were preformed to understand the implications of the allocation approach on LCA results. With zero burden of rice production allocated to rice straw, a significant reduction of the impact of lignocellulosic mycoprotein were observed on land-use (209 ha/tonne-protein), significantly less than the 4390 ha/tonne-protein when the straw is still considered to have an impact associated with its production. This analysis highlights the difficulties in assigning an exact arable land consumption due to allocation of burden to byproducts. Furthermore, the allocation sensitivity analysis also revealed a significant reduction in water consumption (542 m 3 water/tonne-protein, 76% reduction) (Figure S.5) due to water use in the rice production stage being mitigated in a zero burden scenario. This result highlights the advantage of utilising lignocellulosic feedstocks compared to starch-derived glucose, as much of the impact from crop growth is mitigated.
Finally, to understand the sensitivity of the LCA results due to process parameters, a sensitivity analysis was conducted for a range of process parameters ( Figure 8B). The results revealed that across all impact categories, the maximum variation is +23/-16% of the base case scenario. Improvements in saccharification yield, and xylose utilisation would provide the greatest benefits to the sustainability of the lignocellulosic mycoprotein process due primarily to large reductions in feedstock required to produce the same output of protein. Although the electricity grid mix was not considered in the sensitivity analysis, based on the assumed electricity production mix (Natural gas -20.47%, Coal -41.55%, Oil -3.95%, Nuclear -11.27%, Hydroelectric -18.46%, Wind -3.22%, Biomass -0.42%, Biogas -0.31%, Peat -0.03% (Table S.9)), this work can be used as a baseline to understand how the sustainability of the process may change depending on the location of its implementation or changes in the grid mix due to future shifts in electricity generation sources. With the continuing progress towards an increasing share of renewably-sourced electricity, a significant improvement in the sustainability could be attained. On the other hand, if this technology is to be implemented in parts of the world still heavily reliant on fossil fuel resources, the sustainability of this process could be reduced. Therefore, future work should address the uncertainty regarding locationbased parameters, such as the electricity grid mix, and their effect on the sustainability of the technology.
Lignocellulosic mycoprotein performance could be further optimised by utilising alternative lignocellulosic resources instead of rice straw ( Figure 8C). Benefiting from higher cellulose composition, three other feedstocks (switchgrass, miscanthus and corn stover) delivered an improved MSP between 83-91% that of rice-straw derived mycoprotein (Table  S. 38). However, significant reductions in water consumption would be realised for these alternative feedstocks due to their less water demanding growth. Alternative lignocellulosic resources (miscanthus and switchgrass) offer significant reduced arable land use compared with lignocellulosic agricultural residues. Finally, improvements in GHGs can be achieved with alternative feedstocks. Overall, non-food crops (miscanthus and switchgrass) represent environmentally competitive lignocellulosic resources for future exploratory research on lignocellulosic mycoprotein.

Conclusion
This study presents a new lignocellulosic mycoprotein solution, which has been demonstrated as a potential technically viable and environmentally sustainable meat substitute. A considerable amount of cellulose resources are available globally, offering potential underutilised resources for lignocellulosic mycoprotein production. Assuming the estimated annual global availability of rice straw of 520 million tonnes(49), 8 million tonnes of rice strawderived mycoprotein could be produced, which would represent 4% of the current predicted global protein demand (1).
Our experimental results suggested the importance of ionic liquid pretrement in lignocellulosic mycoprotein production. In contrast with choice of cellulase enzyme, the ionic liquid produced greater impacts on the glucose release. Non food-grade IL [TEA][HSO 4 ] outperformed its food-grade counterpart ([Ch][HSO 4 ]), leading to significantly higher (over 50%) glucose release. The food-safety of [TEA][HSO 4 ] is unknown but potentially can be certified as foodgrade ionic liquid.
Modelling results demonstrate that the untextured lignocellulosic mycoprotein paste could be produced at a competitive price of $5.04/kg ($40.02/kg-protein). Although more expensive than meat-protein, lignocellulosic-mycoprotein offers significant potential for improving the sustainability of protein for human consumption. Lignocellulosic mycoprotein offers decarbonisation potential in comparison with protein derived from beef. However, the highlight of this technology is the minimum arable land usage required due to the main feedstock being agricultural residue.
This work can be seen as an initial study highlighting from both an economical and sustainability perspective the potential for this technology. Further work in this area is needed before full commercialisation could be envisaged. Primarily, further process optimisation should be undertaken from both an experimental and computational perspective. For example, increasing pretreatment yields through improving IL performance or selection of the ionic liquid/biomass combination, optimised separation design for cost-effective recovery of the ionic liquid by exploring alternative separation technologies and specifications. Ionic liquid has been researched as a promising pretreatment technology due to the customisable nature of the IL (selecting the cation-anion pair) and process safety offered by their low volatility. Notably, the food-grade ionic liquid (Choline Hydrogen sulphate) investigated in our research, can also be used to enhance F. venenatum fermentation; thus there exists a scope for future indepth research to optimise the separation through synergistic process integration. Furthermore, instead of ILs, alternative food-safe pretreatment technologies offer another research frontier for future investigation.
Overall, our research suggests that lignocellulosic mycoprotein sourced from food-safe lignocellulose represents a transformative solution to animal and plant-sourced proteins which are carbon-intensive and natural resource-demanding. Both animal sourced and plant sourced proteins not only cause significant environmental concerns on climate change and arable land use, but also are constrained by long-production cycles; thus, their supply chains are vulnerable to global arable land scarcity and extreme events e.g. public health emergency. In contrast, lignocellulosic mycoprotein not only enables decarbonised protein supply, which is largely independent of arable land, but also offers future protein security due to its manufacturing in controlled environments and very short production cycles.