ScaleSC-01
Bigger Is a Different Process
Scale-up is not arithmetic. Every physical parameter that kept your small run alive behaves differently at volume, and the cell does not care that your spreadsheet said it would be fine.

The Illusion of Proportionality
There is a seductive logic to the idea that a bioreactor is just a bucket. Double the bucket, double the product. The chemistry is the same, the cells are the same, the medium is the same — surely the maths is straightforward. This is the illusion that scale-up dismantles, expensively and reliably, at every step of a programme's life.
The problem is that the physical world does not scale the way a recipe does. A recipe says: ten times the ingredients, ten times the food. But a bioreactor is not a recipe — it is a system of competing physical phenomena, each governed by its own relationship with geometry. Volume scales as the cube of a linear dimension. Surface area scales as the square. The ratio between them — surface-to-volume — collapses as the vessel grows. That single fact cascades through heat transfer, oxygen delivery, pH homogeneity and shear stress in ways that are not intuitive and are not trivially corrected.
At laboratory scale, say a 2-litre spinner flask, none of this matters much. The vessel is small enough that diffusion alone can carry oxygen to most of the liquid, temperature equilibrates with the environment in minutes, and a stir bar moves the contents uniformly. Move to a 20-litre bioreactor and you are still close enough to benchtop behaviour that the culture usually survives the transition without drama, provided you have thought carefully about the impeller. Move to 2,000 litres and you are in a categorically different engineering problem.

What the Physics Actually Does
Oxygen is the first thing that breaks. Cells consuming glucose also consume oxygen, and they do it at a rate that does not change with tank size — each cell has the same metabolic demand at 2,000 litres as it did at 20. What changes is the difficulty of getting oxygen into the liquid. The rate at which a gas transfers from a bubble into solution is described by a term called the volumetric oxygen transfer coefficient, kLa. It depends on bubble size, agitation intensity, the gas-liquid interfacial area, and the liquid's physical properties. In a small vessel, you can achieve a high kLa relatively easily. In a large vessel, achieving the same kLa requires either sparging more aggressively, agitating harder, or both — and both have consequences.
Agitation is the second thing that breaks. Impellers in large vessels create zones. The region near the impeller sees intense fluid motion; the region far from it, near the liquid surface or the vessel wall, sees relative stagnation. This non-uniformity — the engineer's word is inhomogeneity — means that cells in different parts of the tank experience different concentrations of oxygen, CO₂, pH, glucose and lactate at any given moment. In a 20-litre vessel, the mixing time (the time for a pulse of tracer to fully homogenise) might be measured in seconds. In a 20,000-litre vessel it can be measured in minutes. A mammalian cell whose environment oscillates between oxygen-rich and oxygen-depleted zones on a timescale of minutes is a stressed cell, and stressed cells do not perform consistently.
Heat is the third. Metabolism generates heat, and at laboratory scale the vessel surface dissipates it passively. As tanks grow, the surface-to-volume ratio falls, and the heat load per unit volume — generated by millions of metabolising cells and by the mechanical energy put in by the impeller — cannot be shed through the walls fast enough. Large bioreactors require active cooling via jacket systems or internal coils. Those coils take up volume, alter flow patterns, and introduce their own engineering constraints.
Lifted out · 01
The physics that break at scale
- Oxygen transferkLa falls as vessel volume grows; matching it at large scale requires harder sparging or agitation, both with side effects
- Mixing homogeneitymixing time rises from seconds to minutes; cells cycle through zones of different dissolved gas and nutrient concentration
- Heat dissipationmetabolic and mechanical heat cannot shed through vessel walls; large tanks need active cooling coils that alter flow patterns
- CO₂ accumulationhydrostatic pressure at tank depth raises dissolved CO₂, suppressing pH and potentially altering glycosylation
CO₂ removal is the fourth, and arguably the least intuitive. Dissolved CO₂ is a metabolic waste product. In a small vessel it strips out easily into the headspace. In a large vessel, particularly one that is tall relative to its diameter, CO₂ can accumulate at depth because the hydrostatic pressure at the bottom of the tank raises the partial pressure of dissolved gases, reducing the driving force for stripping. Elevated dissolved CO₂ depresses culture pH, alters cellular metabolism, and can affect glycosylation — the pattern of sugars the cell adds to the protein, which is part of the product's biological identity. Controlling CO₂ at scale requires dedicated sparging strategies and sometimes physical vessel geometry choices that the small-scale developer never had to think about.
What Has to Be Re-established
None of this means scale-up is impossible. It means that each transition to a larger scale is a genuine process-development exercise, not a dial-turn. The parameters that were set at small scale — agitation rate, sparge rate, gas mixture, pH setpoint and control band, temperature — do not transfer numerically. They transfer conceptually, and must then be re-optimised around the engineering realities of the new vessel.
The engineering approach begins with dimensionless numbers: ratios that capture physical relationships in a form that is, in principle, scale-independent. The Reynolds number describes the ratio of inertial to viscous forces and tells you something about the flow regime. The power-per-unit-volume delivered by an impeller is a common basis for attempting to match mixing conditions across scales. The kLa target travels with the process. None of these equivalences is perfect — matching one dimensionless group often forces another out of range — but they provide a rational starting point for what would otherwise be purely empirical searching.
Beyond the fluid mechanics, the vessel itself changes character. Large stainless-steel bioreactors have dead legs in their piping, complex spray-ball cleaning systems, longer steam-in-place cycles, and cleaning validation requirements that scale non-linearly with surface area and geometry. A single-use bioreactor in a smaller format avoids some of this — the bag is replaced rather than cleaned — but single-use has its own physical constraints: bag integrity under agitation, leachable compounds from the plastic into the culture medium, and the fact that the largest commercially available single-use formats top out well below the volumes that a high-demand commercial process may eventually need.
There is a seductive logic to the idea that a bioreactor is just a bucket.
The cell itself is not a passive bystander in this transition. Mammalian cells — Chinese hamster ovary cells are the dominant production host — are large, fragile, and acutely sensitive to their microenvironment. The hydrodynamic stresses generated near a large impeller or at a sparger where bubbles burst can damage cell membranes. This phenomenon is called bubble-induced cell damage, and it is one reason that large-scale processes typically use pluronic F-68, a polymer that reduces surface tension and cushions cells against bubble rupture. The concentration of that additive, and its interaction with the rest of the medium, may itself need adjustment at scale.
Then there is the question of inoculum. A 2,000-litre run does not start at 2,000 litres. It starts at a few millilitres of thawed cell bank vial, which is expanded through a series of progressively larger vessels — shaker flasks, seed bioreactors, intermediate seed bioreactors — before inoculating the production vessel. Each of those steps is a mini scale-up in its own right, with its own oxygen transfer and mixing characteristics. The inoculum train, as this sequence is called, must be designed so that cells arrive at the production vessel in the right physiological state: the right density, the right viability, the right metabolic profile. An inoculum that is too old, too dense, or has already begun producing lactate aggressively will not perform in the production vessel as the process development data predicted.
The Regulatory Dimension
Scale-up has a regulatory shadow that is easy to underestimate. A drug's manufacturing process, once filed with a regulatory agency, is part of the product's approval. Changing scale is a manufacturing change, and manufacturing changes must be evaluated for their potential to alter the product itself — not just the yield, but the molecule. A larger vessel that runs hotter, strips CO₂ differently, or exposes cells to different shear patterns may subtly alter the glycosylation profile, the charge variant distribution, or the proportion of aggregated molecules. These are not hypothetical concerns. They are the reason that comparability studies — systematic analytical comparisons of material made before and after a change — are required every time the process moves to a new scale or a new site.
The practical implication is that scale-up cannot be deferred to the end of development and then executed in a rush before commercialisation. It must begin early enough that the process at commercial scale generates data that feeds the clinical and regulatory files. Material made at clinical scale and material made at commercial scale must be shown to be comparable. If they are not, the programme has a problem that no amount of manufacturing optimisation will quickly resolve.
This is what makes scale-up genuinely hard and genuinely expensive. It is not one experiment. It is a series of engineering campaigns, each requiring dedicated vessel time that is in perpetual competition with other programmes at the same facility, each generating data that must be integrated into a regulatory narrative, and each carrying the latent risk that the product at the new scale is not quite the same product as the one that showed efficacy in the clinic. The number on the side of the vessel is the least interesting thing about it.

Lifted out · 02
Scale-up vocabulary
- kLavolumetric oxygen transfer coefficient; the key number for matching oxygen delivery across vessel sizes
- Reynolds numberdimensionless ratio of inertial to viscous forces; one basis for comparing flow regimes at different scales
- Power-per-unit-volumeimpeller energy input per litre; common engineering basis for attempting to match mixing intensity across scales
- Inoculum trainthe sequence of progressively larger seed vessels used to expand cells from a frozen vial to full production volume
- Comparabilitythe regulatory and analytical demonstration that product made after a process change (including scale change) is equivalent to the prior material
- Pluronic F-68polymer added to culture medium to protect cells from bubble-induced damage; concentration may need re-optimisation at scale

Also in Scale