Raman Spectroscopy with Ambr® 15 Mini Bioreactors and MVDA Offers Faster, More Cost-Effective Process Control and Development
What’s the secret to developing innovative processes for monoclonal antibody (mAb) production that can be scaled up efficiently? You would be right if you said predictive process control at the bioreactor level using multivariate data analysis (MVDA) combined with in situ Raman spectroscopy data.
A good example of this type of application comes from the Biopharm Process Research Group at GlaxoSmithKline (GSK) in the UK who are using Ambr® 15 Mini Bioreactors with integrated BioPAT® Spectro and SIMCA® MVDA software to create models based on RAMAN spectroscopy data.
This article is posted on our Science Snippets Blog.
The Biopharm Process Research Group is part of a leading global healthcare company with an annual turnover of £34 billion. The group is focused on medicines discovery using a state-of-the-art monoclonal antibody (mAb) discovery platforms and technologies such as automation. They wanted to find innovative methods to improve the process development of their therapeutic mAb proteins by putting the large amounts of data generated in early development for commercial-scale monitoring and control to good use.
They are using Raman spectroscopy in mini bioreactors as an online method of monitoring key analytes and product concentration. With the Ambr® system it is possible to automatically measure Raman spectra and also do reference analysis. The references values and spectral data are afterward combined in the MVDA SIMCA® software to create a predictive model. This predictive model was later applied to monitor and control a key analyte at the pilot-scale to determine if this could overall improve their process control and development. The results from this study are available in an article published by the Biopharm Process Research Group (1).
Using PAT for Bioreactor Monitoring and Control
Process Analytical Technology (PAT) tools such as Raman spectroscopy in combination with multivariate data analysis allow to non-invasively measure and predict in real-time various analytes such as glucose, lactate and product titer.
Raman spectroscopy combined with multivariate data analysis is a valuable PAT tool. This enables process monitoring at a very high frequency compared to offline measurements which are typically carried out once a day. When a process can be monitored and controlled so closely it will increase its safety, efficacy and robustness, and to come a step closer to Quality by Design, QbD.
Using Raman Spectroscopy to Control Critical Process Parameters
Raman spectroscopy is based on inelastic scattering of photons from a monochromatic light source, usually from a laser source. When the photons interact with the molecules in a sample, the photons are scattered at a frequency higher or lower than the incident frequency. This Raman shift depends on the chemical bonds within the molecules and sample. The result is a Raman spectrum which contains a “fingerprint” of the measured sample. As an indirect method Raman spectra have been correlated to respective references values. This is accomplished by using MVDA software, which searches for correlations between changes in the references values and the Raman spectra.
Creating Predictive Models
Raman-based process control relies on developing predictive models that correlate a spectral signal with an analyte or other measurements. To build such a model the Biopharm Process Research Group at GSK measured multiple analytes, based on a Design of Experiments (DOE) approach to cover as many expected process variations as possible. Previously the group built models by using benchtop bioreactors. The general workflow includes mostly manual sampling in multiple bioreactors to include enough process variation to build a robust model. At the end reference analysis values and spectral data for each bioreactor must be consolidated, either by a software or manually. This entire process is very time and labor-intensive.
Using data from commercial manufacturing and production to generate a robust calibration model is not feasible. The main goal in production is to generate as little variation as possible.
The GSK group wanted to implement an integrated online Raman spectroscopy solution that could be used for miniature and commercial-scale bioreactors alike. Generating Raman spectra in mini bioreactors in a process development area allows using a larger design space. The automated Ambr® systems with multiple bioreactors, integrated reference analysis and BioPAT Spectro® Raman integration is therefore faster, as well as more cost-effective than running bench-top bioreactors.
Technology Solutions Used
The GSK group used a number of Sartorius products to develop their optimized bioreactor process.
The products are as followed:
Ambr® 15 cell culture system (48 vessels/batches per system) with integrated BioPAT® Spectro
SIMCA® Multivariate Data Analysis (MVDA) software
Biostat® STR bioreactor and single-use Flexsafe® STR bag (50L) with integrated BioPAT® Spectro
Data Correlation and Model Building using Ambr® 15 Mini Bioreactors
In Ambr® 15 system with integrated BioPAT® Spectro and a Raman spectrometer (Tornado Spectral System) was used for on-line data collection. The Ambr® software automatically merges the data of the Raman spectrometer and reference analysis. Hence offline reference data were collected simultaneously. The Raman spectra and reference data were analyzed with SIMCA® MVDA software to generate Orthogonal Partial Least Square Regression (OPLS) models.
The model-building included:
- Standard fed-batch process
- Four CHO-K1A derived cell lines producing high and low titres of IgG1 or IgG2 molecules
- Seeding density 1 - 1.4 x106 cells/mL
- DOE approach to include variation during the cultivations such as samples spiked with known concentrations of glucose, lactate, glutamine, glutamate, and protein
- Three 48 Ambr®15 mini bioreactor runs to acquire Raman spectra and reference data (generated by a Cedex system [Roche]) for spiked and non-spiked analytes and protein titer
Results
OPLS models showed:
- Lactate, glucose and protein titer concentration from Raman spectroscopy and reference data had a high correlation.
- Glutamine and glutamate concentration from Raman spectroscopy and reference data demonstrated a good correlation
Building and Testing a Predictive Glucose Model
To build a predictive Raman model for glucose, a Biostat® STR bioreactor and single-use Flexsafe® STR bag (50L) with integrated BioPAT® Spectro and a Raman spectrometer were used to collect online data. At the same time off-line reference data for glucose concentration was acquired. The previously created OPLS models from the Ambr® 15 runs were used to predict glucose from the Raman spectra acquired in the 50L Biostat STR®.
The model-building included:
- Standard fed-batch process
- Four CHO-K1A derived cell lines producing high and low titres of IgG1 or IgG2 molecules
- Final viable cell concentration up to 31 x 106/mL and 98% viability
Results
OPLS models showed:
- High correlation of glucose concentration using Raman spectroscopy from three Ambr® 15 mini bioreactor runs and the first 50L single-use Biostat® STR bioreactor
- Predicted glucose concentration had a good correlation with measured glucose concentration in a second 50L single-use Biostat® STR bioreactor
Conclusion
The results from these studies are promising because they show that using data from small-scale Ambr® 15 mini bioreactors and BioPAT® Spectro, the Biopharm Process Research Group at GSK can produce accurate models for a range of analytes. Additionally, using Raman spectral data from Ambr® 15 mini bioreactors and just one 50L single-use Biostat® STR bioreactor, a predictive model can be used to accurately measure glucose concentration at a 50L scale. Therefore, implementing Ambr® 15 mini bioreactor and BioPAT® Spectro technology could enable automated spectral acquisition across scales and facilitate model transfer with the potential to deliver faster, more cost-efficiently process control of GSK’s therapeutic mAb portfolio from development through to commercialization.
Reference
1. Rowland-Jones RC, Graf A, Woodhams A, et al. Spectroscopy integration to miniature bioreactors and large scale production bioreactors–Increasing current capabilities and model transfer. Biotechnol Progress.2021;37:e3074. https://doi.org/10.1002/btpr.3074