Deep dive: Fermentation upstream bioprocess design

Types of upstream fermentation technologies

Submerged and solid-state fermentation

Many different bioreactors can be classified in several ways, such as by gas transfer, agitation type, water content, feedstock, and temperature.  In general, bioreactors can be classified as either submerged or SSF systems. In submerged fermentation, the substrate to be fermented is always in a liquid phase along with the nutrients required for microorganism growth, and gas exchange occurs via mixing. By contrast, in SSF, the microorganism grows on a predominantly solid substrate with gas delivery through aeration and a reduced liquid phase (Sun, Shahrajabian, and Lin 2022). There can be technology overlaps between the two fermentation systems, though, such as with liquid-air interface fermentation. See Table 1 for key features and differences in submerged and SSF systems.

Submerged fermentation, the most common industrial fermentation process, enables a tightly controlled environment for optimal production with diverse microorganisms and feedstocks (Table 1). (Paulová, Patáková, and Brányik 2013). This includes bacteria, microalgae, yeast, and other fungi across many metabolism types. 

Beer production is an example of a traditional submerged fermentation process in which fermented sugars are converted to alcohol under anaerobic and often unmixed conditions. The majority of precision fermentation products, like enzymes, are produced through submerged fermentation because of the tight, reproducible control needed over the system. However, the energy costs, water usage, and feedstock costs can be much higher in submerged than in SSF processes.

Submerged	Solid-state
Water is main component of culture	Reduced or no free-flowing water
2-phase system: liquid – gas Liquid is the primary phase for temperature control and gas exchange. Gas is delivered through the liquid phase for mixing and gas transfer.	3-phase system: solid – gas – liquid  Gas is the primary phase for temperature control and gas exchange.  Gas is delivered across solid and/or liquid interfaces for heat and gas transfer.
Homogenous system; good mixing	Heterogeneous system; poor mixing
Dissolved nutrients in water; lower average concentration of nutrients	Complex insoluble substrate; higher average concentration of nutrients in feedstock
High feedstock material cost (usually); Able to utilize gas feedstocks	Lower feedstock material cost; Able to utilize different agricultural sidestreams
Applicable to a wide range of microorganisms	Limited range of microorganisms used
Microorganisms grown in solution; Absorb nutrients from the liquid medium; Little to no nutrient gradient when mixed.	Microorganisms adsorb to or penetrate substrate surface (typically);  Absorb nutrients from the solid substrate; Nutrient gradient exists
High energy usage	Lower energy usage
Good process control and monitoring  (e.g., temperature, pH, DO, nutrients)	Reduced and poor process control and monitoring
Better heat and mass transfer	Limited heat, nutrient, and gas transfer
Final harvest is a liquid with lower product concentrations	Final harvest is a wet substrate with high product concentrations
Robust DSP options	Difficult to separate substrate from the final product

The other type of bioreactor fermentation used in the food industry is solid-state fermentation (SSF), also called solid substrate fermentation. SSF is a fermentation process where the microorganisms are cultivated on moist, solid, non-soluble organic material with little to no free water. The solid substrate, such as grains, seeds, legumes, husks, peels, or other organic materials, acts as a support and nutrient source for microorganism growth (Yafetto 2022). SSF originated in traditional food fermentation to produce foods like bread, tempeh, miso, and koji. For example, tempeh results from steamed soybean fermentation by the fungus Rhizopus, while koji is the product of cooked rice fermentation by the fungus Aspergillus. 

The industry has seen an increased interest in SSF for producing many bioproducts, including enzymes, organic acids, mushrooms, single-cell proteins, secondary metabolites, and more. The microorganisms used are typically fungi, which are suited for SSF because their hyphae can grow into the solid substrate, penetrating the inter-particle space and digesting cellulosic material (Yafetto 2022). SSF can be a valuable food processing approach for bioconverting organic agro-industrial wastes like husks, straw, and peels, which cannot be as readily implemented in submerged fermentation, into bioproducts. Fungi are predominantly used in SSF to produce enzymes, citric acid, fermented animal feed, single-cell protein, and mycoprotein. Only some yeast and bacteria are used in solid-state fermentation to produce organic acids like acetic acid (Sun, Shahrajabian, and Lin 2022). In the alternative protein space, SSF holds great promise for creating whole-cut meat alternatives and improving plant-based feedstocks’ organoleptic and nutritional properties. There is untapped potential for SSF use in precision fermentation to reduce commercial costs. 

The primary advantages of SSF are the simpler design, lower capital cost, reduced contamination risk, lower water usage, lower energy usage, and the usage of underutilized feedstocks. However, SSF is limited by substrate heterogeneity, slower organism growth, feedstock sterilization, and limited gas and heat transfer rate (Table 1). In addition, imprecise process control over parameters like temperature, pH, dissolved oxygen, and nutrient availability is a key limitation of SSF (Paulová, Patáková, and Brányik 2013). SSF is also partially limited by the types of microorganisms that thrive in low-moisture environments (Sun, Shahrajabian, and Lin 2022). 

It is difficult to cover the diverse bioreactors used for microbial fermentation comprehensively. Below are the typical bioreactor types used in the alternative protein sector, though, there is always room for new bioreactor designs or the application of other bioreactors to the alt protein sector.

Solid-state fermentation bioreactors

The many advantages of solid-state fermentation (SSF) suggest it has abundant economic potential. However, its industrial applications are not as diverse as for submerged fermentation, and a key limitation is a lack of simple and practical automated fermenters for SSF processes. Since the predominant phases in SSF are gas and solid, the system is not as homogeneous for monitoring and control, with the potential for poor mixing and heat removal (Table 1) (Chen 2013). Successful SSF application relies heavily on bioreactor design and precise operation due to the heterogeneous mixture and poor heat and gas mass transfer. 

Traditionally, SSF bioreactors can be classified into static and dynamic systems based on their mixing. These can also be classified as stationary, intermittent, or continuous stirred processes with ventilation as either natural convection or forced air (Chen 2013). Both agitation and ventilation have a strong impact on temperature control and mass transfer. Gas ventilation is a primary mechanism to drive temperature and humidity or moisture control in SSF bioreactor systems, though some SSF reactors use jacketed cooling systems for heat removal (Chen 2013). 
The four predominant types of SSF bioreactors (Figure 3) are (1) reactors without mixing and forced aeration (tray reactors), (2) reactors without mixing but with forced aeration (packed-bed bioreactors), (3) reactors with mixing but no forced aeration (stirred drum bioreactors), and (4) reactors with mixing and forced aeration (stirred-aerated bioreactor) (Ge, Vasco-Correa, and Li 2017).

Productivity & optimization

Microorganism growth and productivity in a bioprocess

The upstream fermentation goal is to grow the microorganisms to a suitable and productive biomass density with sufficient cells or cell-derived ingredients for harvest. For biomass products, more cells equal more product. For precision fermentation, more cells equal higher biological production of the target molecule, such as a protein, lipid, or other biomolecule. Understanding the growth kinetics of the microbe is crucial for large-scale bioprocess so that the inoculum, feedstocks, and conditions can be optimized for productivity. 

When grown in a closed system with limited nutrients, the growth of any microorganism can be broken into four phases: (1) inoculation or lag phase, (2) exponential growth log phase, (3) stationary phase, and (4) decline or death phase (Figure 6). At the beginning of cultivation, viable cells from the seed train are inoculated into the bioreactor filled with a sterilized medium. The lag phase occurs after inoculation with slow growth as cells adapt to a new environment. Once adapted, cells grow proportionally to cell mass exponentially until nutrients or other limiting factors slow the growth during the deceleration phase. Growth can be slowed due to many factors, such as nutrient or gas transfer limitation, toxic compound accumulation, or product inhibition. Once slowed, cells enter a stationary phase where cell division and cell death reach equilibrium. Finally, cells reach the decline phase, where the rate of cell death is greater than cell generation (Gutiérrez-Correa and Villena 2010). In a typical process, the late log phase and early stationary phase are the most productive due to the high number of healthy cells.

For biomass or precision fermentation, commercial productivity metrics can be broken into biomass/product titer, rate, and yield (TRY).

Increasing the titer, productivity, and yield in a process significantly improves process OPEX and CAPEX. Thus, these values are the primary benchmarks for upstream efficiency in submerged and SSF bioprocesses. These are routinely monitored in submerged fermentation, however, it can be more difficult to track in SSF due to substrate and product heterogeneity.  

Titer, productivity, and yield can be increased through technological advancements. These include improved expression systems, strain selection, strain development, medium optimization, feedstock delivery strategies, improved bioprocessing through automation or continuous cultivation, improved process analytical technology via improved sensors or machine learning, and through model-driven process optimization.

Cultivation operation modes

Improving bioprocesses through different cultivation modes is a hallmark of upstream bioprocess development. To achieve higher fermentation performance for either biomass or product titer, cultivations can be run in several operation modes, such as batch, fed-batch, continuous, and semi-continuous (Figure 7). Each operational mode has its advantages and disadvantages.

A typical fermentation run across all industries is a batch process. In a batch run, a single fixed volume of medium is provided for the microbial culture for each harvest (Figure 7a). As the microorganism grows and uses the nutrients, biomass and by-products are generated, and growth slows as it enters stationary growth as previously described (Paulová, Patáková, and Brányik 2013). Many industrial processes are run in batch mode due to the relative simplicity of the process. However, a batch can be limited by substrate availability and/or oxygen mass transfer during the exponential growth phase, which limits maximum productivity and yield. 

A fed-batch process is similar to a batch. However, once the microorganism consumes one or more substrates in the medium, the additional substrate is aseptically added to the bioreactor while the product accumulates in the bioreactor (Figure 7b). This extends the active growth phase and productivity rate, which can be further tailored by adjusting the feed rate over time (Bolmanis et al. 2023). Feeding systems can be implemented for both the carbon substrate and/or nitrogen source used. 

The advantages of fed-batch include a prolonged productivity period, the ability to achieve higher maximal cell density, the capacity to increase productivity through controlled nutrient additions, prolonged productive run time, and avoiding substrate inhibition at higher concentrations (Paulová, Patáková, and Brányik 2013). Another advantage of fed-batch operation is the ability to change substrates during a production run, which can be used to modulate metabolism and toggle gene expression for target pathways and products. This is often implemented in production using the methylotrophic yeast Komagaetella phaffii (previously Pichia pastoris) by initially growing on glycerol for high biomass production and then switching to methanol at high cell densities to turn on recombinant protein production via the AOX promoter (Nieto-Taype et al. 2020). 

At the end of the batch growth phase, a prolonged steady-state phase equilibrium can be obtained through a continuous culture process. A continuous process includes the addition of fresh medium continuously to the bioreactor while a proportionate volume of culture broth containing cells, product, and spent medium is removed from the system. The harvest and feed rates are typically set to keep the cells proliferating at a fixed constant rate (Figure 7c) (Nieto-Taype et al. 2020). These can be operated as either (a) chemostat systems where a defined media addition and removal rate is used to control cell growth, or as (b) turbidostat systems where media addition and removal is controlled on a feedback loop based on the measured cell density (Paulová, Patáková, and Brányik 2013).

The benefit of a continuous mode is the possibility of long-term product synthesis, the ability to achieve stable product quality through continuous harvest, and a reduction in unprofitable downtime between production runs, all of which improve operating efficiency costs. However, the continuous process is operationally complex with constant DSP needs, which can increase contamination risk and create genetic drift over the course of long-term operation (Paulová, Patáková, and Brányik 2013; Li et al. 2014).

The final operation mode that is often considered is a semi-continuous or repeated fed-batch process where harvesting all but a small portion of the completed batch culture leaves the remaining cells for use as inoculum in the next filling cycle (Figure 7d). It is a simple concept but potentially difficult in practice. This mode allows for batched, higher-volume harvested material that can be processed in a single DSP run. This mode also increases operational ‘up-time’ by reducing seed train cycle numbers and CIP cycles between batches. However, a semi-continuous operation can also increase the operational costs and risk of contamination and genetic instability as the number of cycles progresses (Li et al. 2014). 

Operation in fed-batch and (semi)continuous cycles mostly applies to submerged fermentation systems, which allow for the straightforward addition and removal of liquid media. Most SSF bioreactors are limited to batch mode operation due to the nature of the substrate and end product. For example, Nature’s Fynd produces its Fusarium mycelial mats through a batch tray reactor process. While most SSF is a batch process, some SSF bioreactors have been designed that allow for (semi)continuous operation. In fact, Nature’s Fynd has developed a now patented continuous or semi-continuous surface fermentation bioreactor system to produce their fungal protein, Fy. The process uses a roller to harvest the fungal biomats when mature, which provides increased automation and throughput over a more simplified tray-based  air-liquid surface fermentation process. Other researchers have demonstrated semi-continuous SSF of stale bread by edible fungi (Neurospora crassa), which used a plug-flow bioreactor (PFBR) to feed the substrate continually and then return a portion of the product back to the bioreactor as inoculum into the system (R. Wang et al. 2021).

In addition to fed-batch and (semi)continuous fermentation for improved productivity and operation costs, processes can include other strategies like media recycling and gas recycling to improve operating cost efficiency and sustainability. As the name suggests, gas recycling is the reuse of exhaust gasses back into the system as gas inputs. This can improve operating costs and efficiencies in gas fermentation systems where the gas feedstock needs to be continuously supplied (Richter, Martin, and Angenent 2013). A gas recycling system has increased equipment needs and operational complexity but can greatly improve gas use efficiency and sustainability. 

Media recycling can similarly be implemented to improve feedstock use efficiency and is often applied to (semi)continuous systems where biomass is the final product. After biomass dewatering (removal of water), a portion of the ‘spent’ broth can be returned to the fermentation process since key substrates may still be present at high concentrations. Media recycling also helps to improve the water-use efficiency of a fermentation process. Overall, though, this approach comes with higher operational complexity, increased risk for contamination, inhibitor accumulation, and the potential for reduced productivity depending on the recycle ratio and nutrient replenishment rate.

Process monitoring and control

Bioprocesses must be monitored and controlled over time for maximum productivity. These processes can be monitored on-line or in-line directly connected to the bioreactor, or at-line and off-line outside the bioreactor (Figure 8). Process analytical technology (PAT) allows for bioreactor process control in real time by monitoring the physical, chemical, and biological parameters. This can improve microbial or cell growth, product formation, and yield, and it enables batch-to-batch reproducibility. Overall, PAT allows for critical process parameters (CPP) and key performance indicators (KPI) to be continually monitored and controlled during biomanufacturing (Gerzon, Sheng, and Kirkitadze 2022).

Many types of PAT sensors are used to monitor bioprocesses. These can be grouped into three categories by variable type: (1) physical, such as temperature, vessel pressure, agitation rate, and medium viscosity, (2) chemical, such as pH, nutrient concentration, and gas concentration like dissolved oxygen (DO), and (3) biological, such as biomass concentration and titer (Figure 8) (Biechele et al. 2015). In commercial systems, the most commonly used online or inline sensors are temperature, pressure, pH, and dissolved oxygen probes. Continuous substrate monitoring for nutrients like glucose has been advanced through on-line electrochemical and biosensor monitoring (Pontius et al. 2020). Novel optical sensors, such as UV/Vis, RAMAN, and fluorescence spectroscopy probes, have been major developments in bioprocess monitoring and continue improving the monitoring accuracy of many parameters, such as cell density (Biechele et al. 2015).

At-line or off-line monitoring builds data knowledge on a process or measures a production run for KPIs by pulling samples and measuring a parameter in the lab. In-line or on-line process monitoring provides continuous, real-time feedback that then allows for subsequent process and KPI control. However, not all target variables can be directly quantified with in-line/on-line sensors during a bioprocess. Software sensors, also called soft sensors, are predictive software models built from bioprocess data sets and hard sensor data that can indirectly quantify challenging target variables (Rösner et al. 2022). 

Soft sensors combine previous process input and output data with a model that uses the real-time PAT input data to predict a target output (Brunner et al. 2021). The underlying models can be built from process data using mechanistic models, which have a detailed understanding of physical process mechanics (white box model) or empirical models using simple input and output observation data. Empirical models are data-driven models that can be built out using an artificial neural network (ANN or black-box model). Using ANNs can be especially helpful in bioprocesses where the underlying biological mechanisms are highly complex and difficult to mechanistically model. ANN soft sensors estimate and predict the targeted variable based on the input and output data used for training the model (Rösner et al. 2022). These models require large process datasets to train, need to be retrained at different scales, and have range limitations based on training data (Bolmanis et al. 2023).

Soft sensor models can be especially useful in predicting biochemical parameters for which suitable on-line and in-line sensors do not exist. For example, online intracellular lipid sensors are not yet available, and monitoring microbial cells’ lipid content for fat production requires laborious off-line sample measurement. Wang et al. 2023 used an ANN model to predict lipid content based on proxy parameters like glucose, dissolved oxygen, and base addition. The model was trained on prior process input data (time, DO, glucose, and NaOH) and output data (glucose, offline cell dry weight, and lipid content), which resulted in a soft sensor that could accurately predict lipid content from proxy real-time data (Wang et al. 2023). These models can be adapted to various hard sensor data. In practice, spectroscopic or fluorescence sensors offer a range of data that can be modeled to different biological variables for soft sensor training (Rösner et al. 2022). 

The combination of sensors, recorder, set-point, controller, and control element create a control loop, which can be open-loop or closed-loop. Open control loops involve measuring output, such as temperature, with no automatic or feedback control. Open control loops might involve simple human intervention to control the process, such as through manual temperature adjustment or manual agitation changes, which is not an ideal commercial strategy. Open-loop controls can also be used for feeding systems when a precalculated substrate feed rate is provided for a controller to deliver the nutrient to the bioreactor based on that schedule. Feed rates can be set linearly or determined empirically through many cultivations (Bolmanis et al. 2023). 

A closed or feedback control loop operates through real-time monitoring and automatic control of a parameter based on the set-point. All closed-loop control systems respond to nonlinear system dynamics with variable control actions. Closed-loop control methods include proportional integral derivative (PID) control and model-predictive control (Bolmanis et al. 2023). A PID controller is the most commonly used closed-loop controller, such as for dissolved oxygen, linked agitation control, or acid-base feeding for pH control. A PID continuously calculates error signal values as the difference between the set-point and the measured process variable. Model-predictive controls (MPC) are very promising methods for fed-batch feeding system controls. MPC integrates a mechanistic or empirical model (such as ANN models) to predict process control feed rates or other output parameters based on measured input reference values (Helleckes et al. 2022).

Model-based systems are gaining popularity for high-throughput screening and bioprocesses, such as PAT soft sensors, digital twins (modeled digital bioreactor), machine learning models, and model-predictive control. All these areas are driven by adoption and advancements in machine learning, which allows for analyzing large, complex datasets that can be difficult to model mechanistically. These models and process control techniques aim to increase and predict production titers, rates, and yields (TRY) (Helleckes et al. 2022). While process control has advanced, it’s important to screen and assess bioprocess at various scales and expected conditions, especially for model training at a commercial scale.

Scaled-up and scaled-down bioprocess testing

Bioprocesses need to be incrementally scaled from lab (0.25-10 L) to pilot (25-1000 L) to demonstration (1-25 m³) to commercial scales (25-1000 m³) to assess and optimize commercial conditions as well as assess the economic viability of the process. Despite progress in modeling and software approaches, there is no substitute for an integrated pilot process. Many processes are fit for purpose, which requires testing at scale to understand yields and areas for improvement. 

After the lab scale, bioprocesses are first assessed in a pilot system with scale-matched DSP to translate the lab-scale process into a scaled-down commercial manufacturing process. After the pilot, the bioprocess is further scaled to a demonstration scale to assess the techno-economics of the process at even larger volumes, which helps to de-risk the large capital investment required for full commercial scale. Companies can either build pilot and demonstration facilities or utilize contract manufacturers (or toll manufacturing) for demonstration. See our manufacturing capacity landscape for more on scaling up a commercial process.

It’s important to consider the commercial-scale production conditions when optimizing the bioprocess. This can be accomplished via scaled-down testing throughout the scale-up process. Fermentation bioprocesses have complex system operations with many parameters impacting performance and likely changing during volumetric scaling. Scaled-down testing assesses a biological process at lab and pilot scale under commercially relevant conditions, feedstocks, and unit processes (Tajsoleiman et al. 2019).

The objective of scaled-down testing is to maintain physical and physiological similarity between large and small scales. These conditions can be mimicked/modeled, to an extent, at lab scale to assess productivity. Examples of scale-dependent parameters include differences in feedstock quality from lab to commercial, vessel hydrostatic pressure differences, changes to maximum gas-liquid mass transfer, and mixing differences at scale (Crater and Lievense 2018). See Table 2 for a list of common scale-dependent parameters and their implications. It’s important to understand and expect productivity changes at the commercial scale due to scale-dependent factors. Moreover, assessing commercially relevant conditions in scale-down testing and techno-economic models based on productivity results can reduce scaling risk. 

Additionally, mimicking fermentation conditions in strain screening protocols will ensure that selected strains ultimately perform as expected in commercial conditions (i.e. “you get what you screen for”). Strains can be initially screened in high throughput methods (i.e. plates, flasks) in relevant conditions and then scaled to small bioreactors for further scaled-down productivity assessment.

Parameter	Deviation	Impact
Feedstock quality	Grade (industrial vs reagent), purity, concentration, variability	Impurities and unfermentable components can impact titer, rate and yield
Feedstock and medium sterilization	Batch vs continuous sterilization (lab vs commercial)	Sterilization mode may impact the stability or degradation of medium components
Mixing	Increased mixing time or decreased homogeneity	Lower productivity due to gradients that may form for parameters like temperature, pH, DO, and substrate
Gas-liquid volumetric mass transfer	Changes in the gas distribution or mixing due to equipment and cost limitations	Slower or unequal gas distribution leads to lower productivity
Liquid hydrostatic pressure	Large vessel increases pressure gradient	Increased gas concentrations impact productivity and cell stress
Shear stress	Increased shear in some conditions to maintain mixing	High shear stress affects cell morphology, productivity, or DSP yield

Commercial case studies

This deep dive discusses upstream bioprocesses from reactor type, growth, cultivation modes, process control, and more. But what does an upstream bioprocess look like commercially for those already producing food products? Many companies are employing numerous strains, medium formulations, bioreactor types, and cultivation strategies. Below are a few commercial examples that implement the above upstream process concepts.

The “simple case” of baker’s yeast

Let’s start with simple baker’s yeast (Saccharomyces cerevisiae). This yeast strain is valued for its dough-leavening ability and has been used extensively in beer and bread production. Commercial fermentation of baker’s yeast to produce a compressed, dried, or instant biomass was established to meet the growing demand of the baking industry. The fermentation typically occurs in a few 100 m³ large-scale aerobic bubble column fermenters with molasses as the substrate (Paulová, Patáková, and Brányik 2013; Noorman and Heijnen 2017). 

The process should be simple to produce yeast biomass. However, due to the ‘Crabtree effect’, which forms alcohol under aerobic conditions in the presence of excess substrate (i.e. glucose), the process must use a fed-batch strategy to control glucose inflow to balance biomass generation and prevent alcohol formation (Paulová, Patáková, and Brányik 2013). Maintaining the dissolved oxygen and precisely feeding the molasses substrate in a fed-batch mode requires tight process control and feeding regiment. This can be controlled through empirical feed-forward control strategies or by implementing process sensors that are tied to the substrate and dissolved oxygen concentrations.

The fed-batch baker’s yeast process is limited by mixing and sufficient oxygen transfer at large scales when trying to maximize biomass production versus alcoholic fermentation (H. J. Noorman et al. 2018). Biomass yields can easily be reduced by 5-10% upon commercial scale-up, which is a significant hit in productivity for a process with tight margins. This is why even simple baker’s yeast has been tested under scaled-down conditions for many decades to mimic commercial conditions, such as mixing, glucose gradients, and oxygen concentration, and to assess optimal operating conditions like fed-batch feeding regimes.

Filamentous fungi in airlift bioreactors

Filamentous fungi are well known for producing antibiotics, organic acids, enzymes, and recombinant proteins. A well-known food application of filamentous fungi is the protein-rich mycoprotein produced by Marlow Foods (Quorn), which has a fibrous texture that mimics meat. Today, Quorn’s mycoprotein is sold as burgers, sausages, and other alternative meat products, and is produced through their submerged fermentation process.

In the Quorn process, the filamentous fungi Fusarium venenatum is cultivated in a 155 m³ external-loop ALR using a continuous flow culture process designed by Marlow Foods (Figure 9). The external-loop ALR design has several advantages for the Quorn process. At large scale the ALR, versus a stirred tank, allows for more efficient OPEX costs and is gentle on the filamentous fungi hyphae, which help create the meat-like fibrous texture. If grown in an STR, the rotating blades and high shear would damage the hyphae. Additionally, the external-loop design provides improved liquid flow and mixing of fresh substrate (glucose and ammonium), which enters at the bottom of the reactor and circulates with the cell suspension through the upward rise of the gas. Finally, the tall liquid height causes immense hydrostatic pressure at the bottom of the bioreactor, allowing oxygen to dissolve into the medium quickly. The low pressure at the top of the reactor riser allows carbon dioxide waste to degas and exhaust before the culture returns through the downcomer (Moore, Robson, and Trinci 2019).

Quorn’s continuous culture process allows for higher productivity versus a batch culture, and is one of the largest continuous systems in the world (Moore, Robson, and Trinci 2019). The fermenter is inoculated with low-density Fusarium venenatum culture, which rapidly grows and enters the continuous cultivation phase after only four days of initial growth. The process runs continuously for up to 1000 hr (6 weeks), after which the process is terminated and the system cleaned and reset for another run. While the continuous culture could keep growing, the Quorn process is terminated due to the accumulation of highly branched mutant strains of the F. venenaturm. While these mutants are nutritionally identical, their morphology impacts the desired final texture of the mycoprotein meat alternative (Moore, Robson, and Trinci 2019). 

During the continuous culture, carbon as glucose is delivered through fresh medium input, and nitrogen as ammonia is supplied through the gas inlet with the compressed air input. Delivery of ammonia (a weak base) is regulated through a pH feedback control loop to maintain a culture set-point pH. Glucose is added during continuous cultivation by maintaining the culture in a chemostat system with a biomass concentration of 10 – 15 g L^-1. In the chemostat process, biomass is removed and fresh medium with glucose is added at a constant rate of 0.17-0.2 hr^-1, which equates to a 30,000 L hr^-1 or 375 kg-biomass hr^-1 output rate. The harvested biomass is then sent to a DSP that includes a heat treatment for RNA reduction followed by dewatering (Moore, Robson, and Trinci 2019). 

The Quorn process is a technologically mature bioprocess and the result of over thirty years of R&D and commercial production. ENOUGH (formerly 3F Bio) produces a mycoprotein product (ABUNDA) from a similar process and the same strain of fungi, Fusarium venenatum. Their process uses a modified submerged fermentation process using hydrolyzed starch glucose feedstock and incorporating medium recycling after each continuous harvest. Recycling the water and leftover glucose in the medium further improves the economics and sustainability of the process. Additionally, their co-location with Cargill’s starch facility in the Netherlands improves glucose feedstock availability and reduces transportation costs.

Gas fermentation: single-cell protein production

A promising fermentation technology is gas fermentation, which uses gasified feedstocks for microbial bioprocesses. This includes C1 carbon gasses like methane (CH₄), carbon monoxide (CO), and carbon dioxide (CO₂). There is a growing and evolving gas fermentation sector for biomass and biochemical production due to the sustainable process potential by using gas feedstocks versus organic sugar-based fermentation, which can have a higher environmental impact over gas feedstocks and may pull from food crops. One area of commercialization is the production of microbial biomass, also called single-cell protein (SCP), through gas fermentation bioprocesses.

SCP production has strong potential as both an end-product food and feed source. Gas fermentation companies like Solar Foods and Air Protein are using hydrogenotrophs (hydrogen oxidizing bacteria) to produce meat substitutes for human consumption (Marcellin et al. 2022). By contrast, companies like Unibio and Calysta are commercializing SCP production for feed production using the methylotrophic bacteria Methylococcus capsulatus to convert methane to protein. These companies are also looking to expand into the human and pet food sectors using this technology. While methane-based processes currently use natural gas for feed production, the process could also be run off methane sourced from waste anaerobic digesters (Jain et al. 2023).

The process deployed by Unibio and Calysta implements the U-loop style bioreactor, continuous fermentation, and medium recycling operation mode (Figure 10) (Olsen et al. 2020). These processes are crucial for high bioprocess output and sustainability. The U-loop bioreactor with counter-current gas delivery and highly pressurized headspace helps to increase gas mass transfer due to the low solubility of the methane or other C1 gasses by driving the gas bubbles down into the loop (Banks et al. 2022; Marcellin et al. 2022). In the U-loop SCP process, methane, oxygen, ammonia, and minerals are added to the bioreactor in a chemostat mode to maintain biomass at a steady state (~20 g L^-1 ) while it’s continuously harvested (Olsen et al. 2020). The harvested biomass is dewatered by a centrifuge, concentrated by ultrafiltration, then homogenized and dried to create the feed protein. The water separated during the DSP is also recycled back into the fermenter to reduce water and nutrient loss during the process. Theoretically, the gas exhaust waste can be recycled back into the reactor or used as an energy source for either steam boilers or spray dryers, since some of the methane likely is unused in the loop (Banks et al. 2022; Marcellin et al. 2022).

The mighty yet simple tray reactor: Solid-substrate or solid-state?

The most commonly used solid-state reactor system for edible fungi is the tray reactor described above. For traditional koji production, this typically takes place in a climate-controlled room where steamed rice inoculated with Aspergillus orzyae (koji spores) is incubated in a chamber and then transferred to trays with ventilated systems. These trays then require intermittent manual mixing since the koji will heat up in the middle of the bed due to microbial metabolism. Modern-day koji production reduces manual labor by incubating in static flatbed reactors with agitators or by using larger SSF commercial systems with advanced air handling, process control, mechanical mixers, and sanitation systems like rotary-drum reactors or stirred bed reactors. 

Several mycelial fungi alternative protein companies, such as MyForest Foods and Nature’s Fynd, have adapted the traditional tray systems for mycelial biomass production. As mentioned, SSF suffers from a number of disadvantages, such as the produced biomass being tightly entwined with the solid substrate, making separation difficult if feedstock material is undesirable in the final product. To address this, Nature’s Fynd doesn’t use a solid substrate in their tray reactor system. Their process uses a surface liquid fermentation method that they call liquid-air interface fermentation (LAIF) adapted for mycelial mat production in solid-state tray reactor systems (Kozubal, Macur, and Avniel 2022). 

In Nature’s Fynd’s patented process, a liquid growth medium is inoculated with filamentous fungi, such as Fusarium novum yellowstonensis, Fusarium venenatum, or Rhizopus oligosporus, and then incubated in a controlled chamber until a thick fungal biomat forms (Kozubal, Macur, and Avniel 2022). This LAIF process has inherent advantages over traditional SSF in harvesting, quality control, and contamination control. The liquid medium can undergo a typical batch sterilization process before inoculation, and the medium is acidic, which reduces opportunistic microorganism contamination during biomat growth. The trays are incubated on a tray rack in a chamber until the biomats reach a density of 50 – 180 g-biomass L^-1 media and are 1-33 mm thick. Nature’s Fynd has found that incubation at a liquid air-interface method produces fungal biomats that have more dense, tensile, and long-aligned filaments than via SSF (Kozubal, Macur, and Avniel 2022).

MyForest Foods (formerly AtLast and a spin-off of Ecovative Designs) is another company using a tray system to produce a mycelial bacon alternative called ‘MyBacon’. Their process requires a substrate, a fungus (Pleurotus ostreatus, the species that produces oyster mushrooms), and a growth environment optimized to produce mycelium with a meat-like texture. They improved upon the tray system by creating a patent-pending incubation environment that maintains moisture, oxygen, and carbon dioxide concentration to grow a mycelium mat on a sterilized substrate that can be solid and/or liquid.

In the growth chamber, the moisture content is controlled by a humidity sensor and misting system delivered through the top of the chamber and recycled back from the bottom (Winiski et al. 2022). The process analytical controls also maintain oxygen and carbon dioxide concentrations in the chamber. Maintaining oxygen levels improves aerobic metabolism, while higher than ambient carbon dioxide levels (2-7%) modulate aerial growth on the tray top and prevent fruiting body (mushroom) formation. Fungal hyphae often grow toward areas of high carbon dioxide, typically an indicator of decomposing organic matter in nature, while fruiting bodies form in areas of low, ambient carbon dioxide. Finally, airflow across the top of the trays promotes heat exchange as well as lateral rather than downward growth of the fungal hyphae (Winiski et al. 2022). This example highlights how even the “simple” tray bioreactor can be modified to control the desired growth of a mycoprotein product.

Precision fermentation: Protein production in Komagaetella phaffii 

Impossible Foods was the first to introduce a plant-based burger with an ingredient derived from precision fermentation. Leghemoglobin H (LegH) is a soy protein, expressed in roots, that binds to heme and when added to food creates an organoleptic experience that mimics the myoglobin-heme complex in animal muscle protein. To produce LegH, Impossible Foods engineered the budding yeast Komagaetella phaffii (formerly called Pichia pastoris) to produce LegH and the concomitant upregulation of heme biosynthesis. K. phaffii is a workhorse biotechnology microbe also used by TurtleTree, The Every Company, ReMilk, Vivici, and Oobli to produce food proteins (Eastham and Leman 2024). Originally isolated for its ability to metabolize waste methanol from the petrochemical industries, it was soon demonstrated to produce impressive quantities of recombinant protein, such as human insulin for diabetes (Cregg et al. 2000; Cereghino and Cregg 2000; Polez et al. 2016). Indeed, K. phaffii is an important chassis organism for protein expression across biotechnology industries.

While many aspects of these companies’ bioprocess are confidential, the general framework for USP of recombinant protein production in K. phaffii  is very well-characterized, as are the general growth requirements of the yeast. Temperatures from 28-30ºC are preferred by most K. phaffii strains, along with pH 5 to 6, and the presence of molecular oxygen diffused in the liquid substrate. Most importantly, a high-cell density fermentation (HCDF) is a key strategy for growth and recombinant protein production in production using K. phaffii, which is accomplished through a fed-batch submerged fermentation process. The production batch can be divided into two phases: (1) glycerol batch phase and/or fed-batch phase, and (2) methanol induction fed-batch phase for protein production.

In this strategy, K. phaffii cells are grown to high density using a feedstock, such as glycerol, that is fed-batch dosed until the cells reach the desired peak production density. Scale-up production begins with stored strains held in frozen glycerol stocks at -80ºC. Next, successive seed trains from frozen stock-inoculated shaker flasks to thousands of liters has been described, with initial reactor volumes and inoculums determined during process scale-up (Liu et al. 2016). The goal is to achieve HCDF in the production bioreactor. Typically, in the bioreactor, an initial bolus of glycerol-containing medium is provided, and at a predetermined time and cell concentration, a feeding rate of glycerol is established. Once the desired cell concentration is achieved, the proliferation stage is completed.

Once HCDF is achieved, the high number of cell factories can then be triggered, known as induction, to produce the engineered protein of interest, which in the Impossible Foods’ case is LegH (Trinh, Ngo, and Truong 2023). In K. phaffii, the cellular induction is triggered by switching the carbon feedstock from glycerol to methanol, which induces the pAOX1 promoter that has been tied to the expression of the recombinant protein. For some strains this shift occurs gradually, but depending on the genetic modifications, for others it can be accelerated by adding methanol directly. For example, in Mut- strains, the methanol does not provide a carbon source for cell growth and proliferation ends; however, the presence of methanol activates the Mxr1 transcription factor that triggers promoters, pAOX1 and pAOX2. Genes driven by this promoter are activated efficiently and the proteins are produced, such as in the heme biosynthesis.

All phases of growth must have optimized feeding controls based on biomass production so that typically there is no residual glycerol remaining during the induction phase. The methanol fed-batch induction and production must be optimized to maintain high protein productivity.  Integrated sensor systems can help improve phase monitoring and soft sensors have been developed for K. phaffii to optimize production (Brunner et al. 2020). Moreover, the metabolism of glycerol (and methanol) occurs only in the presence of oxygen, so the bioreactor oxygenation/aeration and the mixing systems must be optimized to provide and distribute oxygen to all of the cells in the system while preventing overly aggressive mixing that may lead to cell shear stress and lysis.

In Impossible Food’s process, the products are accumulated intracellularly and are thus protected from the high-oxygen environment of the extracellular medium and any proteases produced by the yeast. However, extracellular secretion can be desirable in many bioprocesses to streamline downstream processing. Once the production of LegH has been determined to be sufficient, such as through off-line laboratory testing or on-line production sensors (Shao et al. 2022), the USP can be ended and the cells containing LegH can be removed from USP for downstream processing of the LegH-heme complex.