In a 10 L bench fermenter in Singapore, a yeast strain secreting an egg-white protein holds dissolved oxygen above its set point throughout, and the cooling jacket barely switches on. The company now plans a 40 m³ production vessel. The biology will not change at scale, but the physics will. Oxygen must cross from bubbles into broth as fast as a dense culture consumes it, heat must leave through far less wall area per litre, and the feed must mix before cells meet a pocket of sugar. Precision fermentation scale-up is the engineering of those three limits.
In short
- Precision fermentation uses engineered microorganisms to make specific food ingredients. Its scale-up is mostly limited by oxygen transfer, heat removal and mixing, not by the strain.
- Oxygen transfer rate is OTR = kLa (C* − C). At steady state it must equal the culture’s oxygen uptake rate (OUR) while dissolved oxygen stays above a safe set point.
- No single criterion can be held constant: constant power per volume roughly preserves kLa but lengthens mixing time and raises impeller tip speed; constant tip speed sharply reduces kLa.
- Aerobic cultures release about 460 kJ of heat per mole of oxygen consumed, and wall area per volume falls as vessels grow, so cooling often caps productivity before oxygen does.
- The feed rate is the main lever, because it sets growth rate, oxygen demand and heat load together.
What is precision fermentation scale-up?
Precision fermentation scale-up is the transfer of a process that makes a specific food ingredient from laboratory or pilot vessels to production vessels, often tens of cubic metres or more, while keeping yield, quality and safety. Precision fermentation means using engineered microbial hosts to produce a defined ingredient, such as an enzyme or a milk or egg protein, rather than the microbial biomass itself.
Common hosts include the yeasts Komagataella phaffii (formerly Pichia pastoris) and Saccharomyces cerevisiae, filamentous fungi such as Trichoderma reesei and Aspergillus species, and Bacillus species. The approach is not new: fermentation-produced chymosin has been used in cheesemaking since around 1990.
Most of these processes are aerobic fed-batch fermentations. Fed-batch means the substrate, usually a sugar or glycerol, is fed during the run at a controlled rate instead of being added at the start. The feed holds the specific growth rate, μ (biomass increase per unit of biomass, in h⁻¹), at a chosen value. That avoids overflow metabolism, such as ethanol formation by S. cerevisiae when sugar uptake exceeds respiratory capacity, and keeps oxygen demand within what the vessel can supply.
How do you calculate oxygen transfer rate and kLa?
Oxygen transfer rate is calculated as OTR = kLa (C* − C), where kLa is the volumetric mass transfer coefficient, C* the dissolved oxygen concentration in equilibrium with the gas, and C the actual dissolved oxygen concentration.
OTR = kLa × (C* − C)
kLa is the product of kL, the liquid-film mass transfer coefficient, and a, the gas-liquid interfacial area per unit volume; because the two are hard to separate, kLa is measured as one number, in h⁻¹ or s⁻¹. Oxygen is poorly soluble: water at 30 °C in equilibrium with air at atmospheric pressure holds only about 7.5 mg/L, roughly 0.23 mmol/L, and broth slightly less. A dense culture uses that inventory within seconds, so transfer must be continuous.
The culture’s demand is the oxygen uptake rate, OUR = q_O₂ × X, where q_O₂ is the oxygen uptake per gram of cells per hour and X the biomass concentration. At steady state OTR = OUR. Dissolved oxygen must stay above the organism’s critical concentration, below which respiration slows, so plants control at a set point safely above it, for example 20 % of air saturation. That leaves a small driving force (C* − C). The designer has three levers:
- Raise kLa with more agitation power, more gas flow and better gas dispersion.
- Raise C* with higher headspace pressure or oxygen-enriched air; C* is roughly proportional to the oxygen partial pressure in the gas.
- Reduce OUR by slowing the feed, and with it the growth rate.
For water-like liquids in stirred tanks, the van ‘t Riet correlation is widely cited: kLa = 0.026 × (P/V)^0.4 × u_s^0.5, with P/V the gassed power per unit volume in W/m³, u_s the superficial gas velocity (gas flow divided by vessel cross-section) in m/s, and kLa in s⁻¹. It was derived mainly in small vessels with water-like liquids and is accurate only to within tens of percent, so measure kLa in the real vessel and broth, ideally by gas balancing, which calculates OUR from inlet and outlet oxygen.
How much oxygen and cooling does a production vessel need?
Work backwards from the culture’s demand and compare it with what the vessel can transfer and cool, using illustrative assumptions rather than data for a particular strain.
Worked example
A 40 m³ (working volume) stirred fermenter must hold a yeast culture at X = 50 g/L dry cell weight and μ = 0.04 h⁻¹. Assume a biomass yield on oxygen of 1.0 g cells per g O₂, C* = 0.22 mmol/L in broth at 30 °C and 1 atm air, and a dissolved oxygen set point of 20 % of air saturation. The vessel offers up to 2 kW/m³ gassed power and u_s = 0.02 m/s.
Demand: OUR = μX / Y = 0.04 × 50 / 1.0 = 2.0 g O₂/L·h = 2.0 / 32 × 1000 = 62.5 mmol O₂/L·h. This ignores maintenance respiration, which adds to the real figure.
Required kLa at 1 atm: C = 0.2 × 0.22 = 0.044 mmol/L, so the driving force is 0.176 mmol/L and kLa = 62.5 / 0.176 ≈ 355 h⁻¹ (0.099 s⁻¹).
Available kLa: 0.026 × 2000^0.4 × 0.02^0.5 = 0.026 × 20.9 × 0.141 ≈ 0.077 s⁻¹, about 277 h⁻¹. Maximum OTR ≈ 277 × 0.176 ≈ 49 mmol/L·h, short of the 62.5 needed. Raising headspace pressure to about 1 bar gauge (2 bar absolute) roughly doubles C* to 0.44 mmol/L. The set point stays at 0.044 mmol/L, so the driving force becomes 0.396 mmol/L and the required kLa falls to about 158 h⁻¹, within capacity.
Heat: 62.5 mmol O₂/L·h equals 62.5 mol/m³·h. At about 460 kJ per mole of O₂, that is 62.5 × 460 / 3600 ≈ 8.0 kW/m³, or about 320 kW for 40 m³. Agitation adds up to 2 kW/m³ × 40 m³ = 80 kW, which also becomes heat, for about 400 kW in total. If the cooling system can remove 250 kW, heat, not oxygen, is the binding constraint.
Halving μ to 0.02 h⁻¹ halves metabolic heat to about 160 kW, about 240 kW in total, just inside the limit. The price is a longer production phase, and maintenance respiration narrows the margin, so test whether specific productivity holds at the lower growth rate.
Which scale-up criterion should you keep constant?
No criterion keeps everything the same, so hold the one that protects the tightest constraint, usually oxygen transfer, and then check what happens to everything else. For aerobic fermentations, constant power per unit volume (P/V) or constant kLa is the usual starting point.
| Criterion held constant | What happens at larger scale (geometric similarity, turbulent flow) |
|---|---|
| Power per unit volume (P/V) | kLa roughly maintained; impeller speed falls; tip speed and maximum shear rise; mixing time lengthens |
| kLa or OTR | Usually needs more gas flow, pressure or oxygen enrichment as well as power |
| Impeller tip speed | Similar maximum shear, but P/V and kLa fall sharply |
| Mixing time | Needs impractically large increases in power |
| Gas flow per liquid volume (vvm) | Superficial gas velocity rises with liquid height, risking impeller flooding and foaming |
Worked example
Scale a geometrically similar vessel from 20 L to 20 m³, a 1000-fold volume increase and a linear scale factor of 1000^(1/3) = 10. In turbulent flow, power P ∝ N³D⁵ (N impeller speed, D impeller diameter), so P/V ∝ N³D². Holding P/V constant means N falls to 10^(−2/3) = 0.215 of its pilot value. Tip speed (π × N × D, the speed of the blade tips) rises 0.215 × 10 = 2.15 times. Mixing time, the time for an addition to become uniform, scales as 1/N and rises 4.6-fold: a 3 s pilot mixing time becomes about 14 s. Holding tip speed constant instead would cut P/V to one tenth of the pilot value.
Longer mixing times create gradients. Cells circulate between a sugar-rich zone near the feed point and sugar-starved regions, and between oxygen-rich and oxygen-poor zones; yeasts meeting a local sugar excess can briefly switch to overflow metabolism. Hydrostatic pressure adds about 1 bar for every 10 m of liquid depth, which raises dissolved carbon dioxide at the bottom of tall vessels.
Why does heat removal often limit the process first?
Heat removal often binds first because metabolic heat rises with oxygen uptake, at roughly 460 kJ per mole of O₂, while jacket area per cubic metre falls in inverse proportion to the linear scale factor.
Q = U × A × ΔT
Here Q is the heat removed, U the overall heat transfer coefficient, A the heat transfer area and ΔT the mean temperature difference between broth and coolant. With broth at 30 °C, cooling water in a warm climate may be only a few degrees colder, so ΔT is small. Large vessels therefore add internal coils or external loop heat exchangers. Every watt of agitation also becomes heat, so adding power to raise kLa raises the cooling load.
How do you reduce the risk of a failed first plant batch?
Test the strain under plant-like conditions before the plant trial, measure the real vessel, and design the feed backwards from the plant’s oxygen and cooling limits.
- Scale-down models: laboratory set-ups, such as two connected vessels or an oscillating feed, that expose cells to the substrate and oxygen fluctuations expected at the plant’s mixing time.
- Vessel characterisation: measure kLa and mixing time in the production vessel with water, then confirm with broth.
- Gas analysis: track OUR, carbon dioxide evolution and the respiratory quotient (RQ, moles of CO₂ produced per mole of O₂ consumed). A rising RQ in a yeast fermentation can signal overflow metabolism.
- Foam, sterility and downstream: check antifoam effects on kLa, keep the extra valves and seals of a large vessel sterile, and check how a longer run affects recovery and product quality.
A milk protein made by precision fermentation should be assumed to carry the same allergenic potential as the cow’s milk protein it copies, and the Codex Guideline for the Conduct of Food Safety Assessment of Foods Produced Using Recombinant-DNA Microorganisms (CXG 46-2003) sets out principles for assessing the production strain and the product. National approval routes differ, so check your national legislation. The Advanced Food Technology course works through a full scale-up review of this kind, and equipment qualification and technology transfer are covered in Food Technology for Industry Professionals.
Frequently asked questions
What is a good kLa for a fermenter?
The required kLa is the oxygen uptake rate divided by the driving force, kLa = OUR/(C* − C). A dense aerobic yeast culture can need several hundred per hour at atmospheric pressure; a slow-growing or dilute culture needs far less. Calculate your own requirement, then compare it with kLa measured in the actual vessel and broth.
Why does dissolved oxygen fall when a fermenter is scaled up?
Usually because the large vessel transfers less oxygen per litre, or because averages hide gradients. Holding tip speed constant cuts power per volume and kLa. Even at constant power per volume, longer mixing times leave zones far from the impeller poorly oxygenated, and a probe measures only one point.
Does increasing pressure in a fermenter improve oxygen transfer?
Yes. Oxygen solubility is roughly proportional to the oxygen partial pressure in the gas, so raising headspace pressure from atmospheric to about 1 bar gauge roughly doubles C*, and more than doubles the driving force if the set point concentration stays the same. Limits are the vessel’s pressure rating and rising dissolved carbon dioxide, which can inhibit some organisms.
Why are most precision fermentation processes run as fed-batch?
Fed-batch lets the operator set the growth rate through the feed. That prevents overflow metabolism, such as ethanol formation by yeasts when sugar is in excess, and keeps oxygen demand and heat release within the vessel’s capacity. Continuous culture runs for long periods, raising the risk of contamination and strain instability, so it is used mainly for biomass products such as mycoprotein.
Can a faster-growing strain increase plant output?
Not necessarily. If the plant is limited by oxygen transfer or cooling, the feed must still hold oxygen demand and heat within capacity, so a strain able to grow faster gains nothing. Output improves only if the strain makes more product per gram of cells or per mole of oxygen.
Next step. The Advanced Food Technology course covers Monod kinetics, oxygen transfer and kLa correlations, scale-up criteria and downstream processing of secreted proteins, including a scale-up review from 50 L to 40 m³ for a secreted dairy protein. It finishes with a proctored final assessment and an ASC certificate, and you can see all eleven food science and technology courses.
Sources. Pauline M. Doran, Bioprocess Engineering Principles, 2nd edn (Academic Press, 2013). K. van ‘t Riet, Review of measuring methods and results in nonviscous gas-liquid mass transfer in stirred vessels, Industrial and Engineering Chemistry Process Design and Development (1979). Codex Alimentarius Commission, Guideline for the Conduct of Food Safety Assessment of Foods Produced Using Recombinant-DNA Microorganisms (CXG 46-2003), available from the Codex Alimentarius website. FAO, Thinking about the future of food safety: a foresight report (2022).
This article is general guidance and is not a substitute for the applicable standard, legislation or the advice of a qualified bioprocess engineer.