When lab-scale biology meets manufacturing reality: where most scale-up goes wrong
Reading time: 7 minutes
A T-75 flask is a piece of polystyrene with a flat growing surface of seventy-five square centimetres. It holds roughly fifteen millilitres of medium, sits still in an incubator, and requires a human to open, feed, and observe it every day or two. Most of what we know about how stem cells behave comes from cells grown under such static conditions or something close to them.
A stirred-tank bioreactor is a pressurised vessel, typically stainless steel or single-use plastic, containing litres to hundreds of litres of medium agitated by a mechanical impeller. Oxygen is delivered through sparging or surface aeration. Temperature, pH, and dissolved gas are monitored by probes and controlled by automated feedback loops. The cells inside experience a physical environment that bears little resemblance to the flat, still conditions of the flask.
The gap between these two systems is where most scale-up efforts lose cell quality, and where many ancillary TechBio companies discover that their products do not perform as expected. This article is part of the Pillar 2 series on why ancillary technologies for stem cell science fail. It follows directly from the companion article on reproducibility, which examines the variability problem at the level of individual batches and operators.
The physics changes
Scaling from millilitres to litres introduces physical forces that simply do not exist in static culture. The most consequential is hydrodynamic shear stress, the force exerted on cells by moving fluid. In a stirred bioreactor, the impeller creates turbulence to keep cells in suspension and distribute nutrients evenly. That turbulence also applies mechanical force to the cells, and stem cells are notably sensitive to it [1].
The challenge is quantitative, not just qualitative. Research using computational fluid dynamics (CFD) modelling has shown that in conventional horizontal-blade bioreactors, shear stress is highest near the impeller tips and drops off steeply with distance. This creates a non-uniform environment where some cells experience damaging levels of mechanical force while others experience almost none [2]. Pluripotent stem cells grown in this environment without any substrate to adhere to or in form aggregates of widely varying sizes, because the physical conditions acting on them differ depending on where they happen to be in the vessel at any given moment.
Aggregate size matters. Larger aggregates develop nutrient-deprived cores where oxygen and glucose cannot penetrate adequately. Cells at the centre of an oversized aggregate receive different signals from those at the periphery, leading to heterogeneous differentiation and, in extreme cases, necrotic regions. Smaller aggregates may experience more shear and lose pluripotency markers faster. The distribution of aggregate sizes, not just their average, determines whether a bioreactor run produces a usable population or a mixed collection of cells in various states [2,3].
Oxygen and nutrient gradients
In a flask, the medium layer is shallow enough that dissolved oxygen reaches the cell surface by diffusion alone. In a bioreactor, oxygen must be actively delivered, typically by bubbling gas through the liquid or across its surface. The efficiency of this transfer, measured by a parameter called kLa, changes with reactor geometry, agitation rate, and liquid volume [1].
As reactor volumes increase, maintaining adequate oxygen supply becomes more difficult. Faster agitation improves gas transfer but increases shear. Sparging air or oxygen directly into the medium is effective for oxygen delivery but introduces bubbles, and bubbles present their own problems. Cells can become trapped on bubble surfaces and removed from the bulk medium. When bubbles burst at the liquid surface, they generate localised forces that can damage surrounding cells. In conventional biopharmaceutical production, where the cells are simply hosts for protein expression and are discarded afterwards, this is a tolerable loss. In cell therapy manufacturing, where the cells are the product, it is not [2].
Anti-foaming agents can mitigate bubble-related damage, but these are hydrophobic chemicals that can incorporate into cell membranes. Their effects on stem cell behaviour and patient safety are not fully characterised, and their use in GMP manufacturing of therapeutic cells is therefore strongly undesirable [2].
Nutrient gradients present a related problem. In a well-mixed small vessel, every cell experiences roughly the same concentration of glucose, amino acids, and growth factors. In a larger vessel, mixing may be incomplete, creating zones of nutrient depletion and waste accumulation. Cells in different regions of the reactor effectively grow in different media, and they respond accordingly.
What changes at scale
Focusing here on human pluripotent stem cells, several specific failure modes emerge during scale-up that are not visible at laboratory scale.
Loss of pluripotency. Mechanical and biochemical stresses can cause stem cells to begin differentiating spontaneously. This is invisible to visual inspection in the early stages but can be detected by changes in gene expression or surface marker profiles. A manufacturing batch that begins as a uniform pluripotent population can arrive at harvest with a substantial fraction of partially differentiated cells that do not meet release specifications [3].
Genomic instability. Extended culture under stress, including the mechanical and metabolic stresses inherent in bioreactor environments, increases the risk of chromosomal abnormalities. Karyotypic changes that are rare at early passage become more frequent with continued expansion, and some of these changes may confer a growth advantage, meaning they accumulate preferentially. Monitoring for genomic stability across serial passages in bioreactor conditions is essential but adds cost and complexity to the manufacturing process [4]. The Pillar 1 article on characterisation and quality control and the Pillar 2 article on characterisation gaps examine what current assays can and cannot detect.
Process sensitivity to input variables. Laboratory-scale optimisation typically tests one variable at a time: this medium versus that medium, this seeding density versus that density. At manufacturing scale, variables interact. The choice of medium affects cell metabolism, which affects oxygen demand, which affects the agitation rate needed to maintain adequate gas transfer, which affects shear stress, which affects aggregate size, which in turn affects how efficiently cells consume the medium. This interdependence means that an optimisation performed at small scale may not predict behaviour at larger volumes [4,5].
How ancillary TechBio products are affected
For the company building tools that interface with cell manufacturing, the implications are direct.
If your product is a culture substrate, coating, or scaffold, it must perform under flow conditions, not just static attachment. Adhesion that holds under gravity in a flask may fail under the shear forces present in a stirred vessel.
If your product is an analytical instrument or monitoring system, it needs to measure parameters in a closed, pressurised environment, through sampling ports or inline sensors, rather than requiring flasks to be removed from the incubator and placed on a bench.
If your product is a medium formulation or supplement, its performance must be validated across the range of reactor geometries and agitation regimes that potential customers use. A medium that supports healthy growth in a rocking-motion bioreactor may perform differently in a stirred tank, because the hydrodynamic environment, gas transfer characteristics, and mixing profiles are fundamentally different.
If your product is a software platform or data system, it needs to ingest data from bioreactor sensors, process analytical technology, and downstream quality control in formats that vary between equipment manufacturers. The absence of data standards in this space is a known problem and an unsolved one. This is explored further in the Pillar 2 article on toolchain fragmentation.
What can be done differently
Companies that navigate scale-up successfully tend to share several practices.
Early engagement with manufacturing conditions. Testing your product in bioreactor environments during development, not after commercial launch, eliminates the worst surprises. Bioreactor access is increasingly available through contract development and manufacturing organisations (CDMOs) and academic bioprocessing centres. The cost of a few bioreactor runs during product development is substantially less than the cost of a failed partnership evaluation.
Computational modelling before physical testing. CFD simulations can predict the hydrodynamic environment in a bioreactor at a given scale and agitation rate, and these predictions have been validated experimentally [2]. Using modelling to understand where shear stress will be highest, where nutrient gradients will form, and how aggregate size will respond to operating parameters is a cost-effective step before committing cells and consumables.
Process characterisation across scales. The critical message from recent bioprocess research is that scale-up is not a single event but a characterisation exercise. Parameters that produce optimal results at one volume need to be re-evaluated at the next [4]. For the ancillary technology company, this means your product validation data should include performance at multiple volumes, or at minimum, a clear statement of the scale range within which your data are valid.
Design for the manufacturing environment. GMP manufacturing facilities operate under constraints that research laboratories do not: closed systems, limited manual intervention, material traceability, environmental monitoring. A product designed for the bench needs to be compatible with these constraints, or its market is limited to research applications, which is a viable business but a different one from the one most TechBio companies opt to build towards. The Pillar 1 article on GMP and Quality by Design covers what these constraints require in practical terms. For background on the biology of stem cell culture methods that underpin scale-up, see the Pillar 1 article on key methods.
References
[1] Borys BS, Roberts EL, Le A, Bhatt S, et al. Overcoming bioprocess bottlenecks in the large-scale expansion of high-quality hiPSC aggregates in vertical-wheel stirred suspension bioreactors. Stem Cell Res Ther. 2021;12(1):18. DOI: 10.1186/s13287-020-02109-4
[2] Lee B, Jung S, Hashimura Y, Lee M, Borys BS, Dang T, Kallos MS, Rodrigues CAV, Silva TP, Cabral JMS. Cell Culture Process Scale-Up Challenges for Commercial-Scale Manufacturing of Allogeneic Pluripotent Stem Cell Products. Bioengineering (Basel). 2022 Feb 25;9(3):92. doi: 10.3390/bioengineering9030092.
[3] Yehya, H., Raudins, S., Padmanabhan, R. et al. Addressing bioreactor hiPSC aggregate stability, maintenance and scaleup challenges using a design of experiment approach. Stem Cell Res Ther 15, 191 (2024). https://doi.org/10.1186/s13287-024-03802-4
[4] Borys BS, Dang T, Worden H, et al. Robust bioprocess design and evaluation of commercial media for the serial expansion of human induced pluripotent stem cell aggregate cultures in vertical-wheel bioreactors. Stem Cell Res Ther. 2024;15(1):232. DOI: 10.1186/s13287-024-03819-9
[5] Vunjak-Novakovic G, Scadden DT. Biomimetic platforms for human stem cell research. Cell Stem Cell. 2011;8(3):252-261. DOI: 10.1016/j.stem.2011.02.014
About StemCells.Help
StemCells.Help is an advisory consultancy that aids innovation and real-world impact of life science applications built on developmental and stem cell biology. Founded by Dr Paul De Sousa, it draws on over four decades of experience spanning early embryo development, animal cloning, pluripotent stem cell manufacturing, and technology commercialisation. If you build tools for these domains or work in an emerging application where the biology is the enabling technology, StemCells.Help can provide experienced scientific counsel to ground your decisions. To discuss your needs, talk to Paul.
ORCID: 0000-0003-0745-2504
Web: stemcells.help
