Model for GS-CHO cell culture in perfusion bioreactor with spent media recirculation
23 January 2018
Clare Horgan with her parents, John and Louise Horgan, at the UCD EGA Awards Ceremony 2017
Monoclonal antibodies (mAbs) are complex recombinant proteins (Figure 1) which are industrially produced by large-scale culture of mammalian cells. mAbs are commonly prescribed to treat many life-threatening illnesses, such as multiple types of cancer, auto-immune disorders (e.g. rheumatoid arthritis, psoriasis and Crohn’s disease) and, potentially, neurological disorders, including Alzheimer’s.
Production of mAbs is a complex process, resulting in high capital and manufacturing costs and, consequently, a high end-user price per gram. With multi-billion-dollar sales each year, biopharma companies are continually searching for more efficient means of production.
Large scale fed-batch cell culture is, by far, the most commonly used technology for upstream processing, due to its simplicity and relatively low cost. Fed-batch culture involves adding nutrients to the bioreactor without removing any of the contents until the run is finished.
As such, fed-batch processing has several limitations, including accumulation of by-products, higher process variability in terms of product quality and yield, as well as limited overall product yield, which can only be increased by scaling-up, volumetrically.
Continuous culture, with emphasis on perfusion operation, provides an attractive alternative to batch processing, by offering higher volumetric productivity and high flexibility in terms of target product yield. Although perfusion has existed for decades, it has returned to the limelight due to recent advances in continuous downstream purification processes.
Perfusion operation consists of continuously feeding and withdrawing liquid from the bioreactor. The inlet stream contains fresh culture media, whereas the outlet contains mAb product and spent media including undesirable by-products. Cells are either retained inside the bioreactor by means of a filter unit or flow out of the bioreactor, recovered using external filtration and returned to the bioreactor.
As a continuous process, perfusion operation offers high flexibility in terms of product yield, lower bioreactor volumes (which curbs capital expenditure and space requirements) and higher volumetric productivity. Despite these advantages, perfusion bioprocesses suffer from extremely high media consumption which results in high operating costs, which may discourage companies from using this operation mode (Klutz et al., 2016).
Concept underlying the project
The concept underlying this project involves recycling spent media back into the bioreactor using an ideal separation unit which allows for complete removal of harmful by-products from nutrients in the bioreactor outlet stream. By reducing the amount of fresh media fed into the bioreactor, it may be possible to alleviate the high operating costs associated with perfusion culture.
Three unit operations (UOs) were modelled in this study: (i) the bioreactor, (ii) the filtration unit and (iii) the ideal separation unit. All UOs were modelled, computationally, on the dynamic process modelling platform gPROMS ModelBuilder. The system was defined as shown in Figure 2. The feed medium is considered to contain glucose and glutamine (a nitrogen source), which are essential components of cell culture media for cellular growth.
The bioreactor contents are to be continuously withdrawn from the bioreactor and fed into the filtration unit, where the cells are retained and the liquid permeates. This liquid contains spent media and mAb product. The cells are then returned to the bioreactor, and the permeate is sent to the separation unit. The unused glucose and glutamine are returned to the bioreactor, while by-products – namely ammonia and lactate, which inhibit cell growth – are removed; for the purpose of the model, ideal (100%) separation efficiency is assumed.
The product, which is also separated from the by-product and recycling streams, is transferred to downstream purification UOs. The filtration device was modelled as a tangential-flow hollow fibre unit, which is commonly used in large-scale perfusion processes.
The hypothetical separator was modelled in terms of inputs and outputs, with the internal workings of the unit specified, assuming a ‘black box’ approach. Full recovery was assumed for the modelling of this unit. This allowed, in theory, for all unused nutrients to be recycled, with no loss of glucose or glutamine and no transfer of ammonia and lactate to the bioreactor via the recycling stream.
A recycle ratio, representing the fraction of spent media recycled back into the bioreactor was defined, with a value of between 0 and 1.
Reference cell culture performance and base case
The target capacity for the proposed process was tens of kilograms of mAbs produced per perfusion run. For reference, a study by Klutz et al. (2016) compared the performance of a fed-batch cell culture process yielding 10kg of mAb per batch and 20 batches per year for a total of 200kg of mAb per annum, with a ‘conventional’ perfusion process yielding 40kg of mAb per perfusion run and 5 runs per year, to also yield a total of 200kg of mAb.
Klutz et al. (2016) included detailed cost analyses and concluded that the perfusion process led to a much higher cost of goods (CoG) than the fed-batch process, due to tincreased cell culture media costs.
For the base case (BC), a standard continuous process was modelled with no recirculation of spent media, i.e. the recycle ratio (RR) was set to 0. This confirmed that the model was working correctly, accurately modelling cellular growth behaviour.
For this model, the process yielded 45.95 kg of mAb per batch, showing it had the capacity to produce comparable amounts to those reported in literature. Using a culture medium cost structure based on the concentration of glucose and glutamine, the CoG for this base case was calculated as €98.66/gmAb.
This value was higher than expected, compared with €63/gmAb for perfusion processing, as reported by Klutz et al. (2016). The lower CoG reported by Klutz et al. (2016) can be attributed to the fact that these authors considered both a flat-rate media cost, irrespective of substrate concentrations and constant cell density (i.e. no accumulation) throughout the perfusion run.
Introduction of recycle
Utilising the multiple vector parameterisation algorithm available in gPROMS ModelBuilder, three dynamic optimisations (Figures 3 & 4) were conducted, to minimise CoG, while simultaneously satisfying the three minimum production constraints set by the model: 20 kg (Opt1), 40 kg (Opt2) and 60 kg (Opt3). Constraints on the lower limit of mAb yield were defined for each optimisation, as minimising CoG alone resulted in very low product yields.
Overall, the CoG was reduced by a factor of 2 when compared to the base case standard continuous process, thus demonstrating the potential economic benefits of recycling spent media. Relative to standard perfusion, Opt 1 showed (Figure 4) the largest (50%) reduction in CoG, identifying Opt1 as the most cost-effective of the conditions investigated.
Comparing Opt2 with standard continuous, with both producing 40-45kgmAb (Figure 3), there was a 47% reduction in CoG. This model demonstrated that a yield of 60kgmAb could be achieved with a 44% reduction in CoG as a result of spent media recirculation.
Comparing these results to those reported by Klutz et al. (2016) (Figures 5 & 6) reveals perfusion with recirculated spent medium as a competitive alternative to both standard continuous and fed-batch processing, based on an increase in mAb yield and a reduction in CoG. Klutz et al. (2016) reported €59/gmAb and €63/gmAb for fed-batch and continuous processing, respectively.
A reduction of 18% in CoG was observed when comparing the resuts of Klutz et al. (2016) (continuous) with Opt2 (continuous with recycle), with both processes producing 40kgmAb (Figure 5). The most cost-effective method, Opt1 (recycle ratio = 0.5, 20 kg mAb produced) showed CoG reductions of 17% and 22%, relative to to fed-batch (Klutz) and continuous (Klutz), respectively (Figure 6), thus further supporting the economic advantages of the proposed spent medium recycling concept.
The model developed in this study demonstrated the achievability of target yields of tens of kgmAb per batch, reaching up to 60kgmAb per batch. The model proved that recirculation of spent media significantly reduced the CoG, relative to standard continuous processing. The CoG for standard continuous with no recirculation was calculated to be €98/gmAb, compared to recirculation case costs of €49/gmAb-€56/gmAb.
With reference to results reported in the literature (Klutz et al., 2016), similar mAb yields were achievable with this model, at CoG competitive with both continuous and fed-batch operation. The results are encouraging and suggest an approach which, with further development work, could reduce the upstream costs of mAb production, making perfusion a competitive alternative to fed-batch. Overall, the model proved to be both technically and economical feasible with promising potential for future work.
Future work will focus on the development of technology to perform the duties of the separator modelled in this study. Currently no such technology is available. If this hypothetical separator could be designed to achieve full nutrient recovery and by-product removal, large savings in media cost and overall cost of production could be achieved.
Further work would involve increasing the recycle ratio in an attempt to recirculate as much spent culture media as possible, to reduce fresh media inlet requirements. However, reducing the inlet flowrate will, in turn, reduce system productivity, since mAb production depends on the permeate flowrate.
Thus, an optimum must be identified, with the inlet flowrate decreased to reduce cost, without affecting productivity. With these improvements, the cost reductions could be even more significant.
Clare Horgan is a 2017 BE graduate in chemical and bioprocess engineering from University College Dublin. This paper describes the modelling of a process to produce commercially valuable recombinant proteins, for use as biopharmaceuticals, from mammalian cell culture. It evaluates the feasibility of recirculation of spent media as a potential route to reducing high operating costs.
On the basis of this work, Clare achieved the highest mark in the Chemical & Bioprocess Engineering Final Year Research Project 2017 and was awarded the Carthy Graduate Research Project Award 2017, kindly endowed by Mark Carthy, a UCD chemical engineering graduate from the class of 1982 and currently managing partner at Orion Healthcare Equity Partners.
- del Val IJ. (2013) ‘Assessment of the interactions between bioprocess conditions and protein glycosylation in antibody-producing mammalian cell cultures’ (doctoral dissertation). Imperial College London, UK
- Klutz S, Holtmann L, Lobedann M, Schembecker G. ‘Cost evaluation of antibody production processes in different operation modes.’ Chemical Engineering Science 141:63-74, 2016
Research project supervisor: Assistant Professor Ioscani Jimenez Del Val, UCD School of Chemical & Bioprocess Engineeringhttp://www.engineersjournal.ie/2018/01/23/mathematical-model-gs-cho-cell-culture-perfusion-bioreactor-spent-media-recirculation/http://www.engineersjournal.ie/wp-content/uploads/2018/01/clare-horgan.pnghttp://www.engineersjournal.ie/wp-content/uploads/2018/01/clare-horgan-300x300.pngChembiopharma,biotechnology