This study analyzed the performance of the HyperFlux™ PRO Plus Raman Spectroscopy System, from Tornado Spectral Systems, for the quantitative analysis of biochemical constituents in a simplified chemical orientated pseudo growth medium in mammalian cell cultures. An experiment was created to evaluate how effective Raman Spectroscopy is indirectly measuring individual components within a complex mixture, at concentrations at or below the limit of quantification of traditional Raman spectrometers. Different samples with varying amounts of glucose, lactate, glutamine, glutamate, ammonium, arginine, histidine, leucine, and phenylalanine were prepared so that the value of covariance between the components was almost zero. It also only took one day to complete the spectral collection and model development. The sample spectra were collected in the morning using fast acquisition times, and the afternoon was used to develop promising calibrations using basic pre-treatments such as derivatives and normalization.
The literature extensively describes the real-time analysis of biochemical biomarkers in bioreactors using Raman Spectroscopy. [1,2,3,4,5,6] Possessing the ability to effectively monitor the usage of starting materials, the formation of by-products, as well as estimating the rates of cell growth, apoptosis (viable cell count and total cell count) and protein expression in real-time would provide excellent opportunities to increase control in the process, and thereby improve product output and uniformity.
To date, Raman spectroscopy has not been widely used within industrial bioreactors because of the complexity of the sample being analyzed. There are many signals that can interfere with the analysis, which arise from a cocktail of metabolites that make the calibration model development challenging.7 A second factor occurs within in the bioprocesses where there are high levels of collinearity between variables to be predicted, making it hard to sometimes understand whether the compounds of interest are being measured directly or through an inferential means, which in turn creates issues towards scalability. Finally, the sensitivity of many Raman instruments can sometimes be an issue as the sample concentrations tend to be heavily diluted and Raman spectrometers are often operated at the lower end of their capabilities. The literature quotes a measurement time of 10-13 minutes to obtain a signal. [1,5]
In this analysis, a design space was created so that a multivariate calibration set could be produced for a chemically defined growth medium with no correlations between the variables. The samples were then synthesized and analyzed using the HyperFlux™ PRO Plus 785 nm Raman Spectroscopy System.
Ensuring that the experimental design has orthogonality between the variables, [8,9] a design space was mapped out using nine variables and five levels. A sample covariance plot between two variables is shown in Figure 1. Similar graphs can also be plotted for all variable combinations.
Figure 1. An Example Covariance plot for Lactate vs. Glucose
The variable components within the samples were glucose, lactate, glutamine, glutamate, ammonium, arginine, histidine, leucine, and phenylalanine. The invariant components in the samples were buffer and serum solutions and these were presented at high concentrations with the pH being maintained at 7.0. Individual stock solutions were prepared for the nine different components by dissolving a specified amount of chemical in a buffer solution at pH 7.0. Two sample sets were created by mixing the required volume of stock solutions together. 10% of serum was added to the samples in the second set. The final concentration of the samples ranged from 0.9 to 4.5 g/L (900-4,500 ppm) for glucose, from 0.2 to 1.8 g/L (200-1,800 ppm) for lactate and from 0.1 to 1 g/L (100-1,000 ppm) for glutamine, glutamate, ammonium, arginine, histidine, leucine, and phenylalanine.
Triplicate spectra were collected using the HyperFlux™ PRO Plus Raman Spectroscopy System in conjunction with the acquisition parameters below:
- Laser Intensity (at source): 495 mW
- Exposure Time: 3 seconds
- Accumulations: 15
- Total acquisition time: 45 seconds
Raw spectra between 1800 and 800 cm-1 are showcased in Figure 2 and it shows that the signal to noise ratio is high despite the short acquisition time of 45 seconds.
Figure 2. Raw spectra collected using Tornado’s HyperFlux™ PRO Plus Raman Spectroscopy System
Calibration and Development
Individual PLS calibration models were developed for each of the nine components using Bruker OPUS Software.
Results and Discussion
The calibration performance is summarised in Table 1, and the test set data are plotted in Figure 3.
Table 1. R2 and RMSEP of the calibration models
||RMSEP [g/L (ppm)]
Figure 3. Predicted vs. True diagram of the test set validation models for Glucose (A), Lactate (B), Glutamine (C), Glutamate (D), Histidine (E), Leucine (F), Arginine (G), Phenylalanine (H) and Ammonium (I).
Phenylalanine and ammonium salts were found to be the best performing calibrations, while the calibration for arginine had the largest error. It is possible to further improve the precision of these measurements by increasing the number of measurements taken. For this analysis, the measurement time was 45 seconds. For use in a bioreactor, a sampling frequency of 15 minutes or under would be required, so there is a wide scope to increase the acquisition time. Critically, for dynamic processes, an acquisition time of 3 seconds should be possible.
Raman Spectroscopy possesses the ability to be used as a multicomponent sensor to monitor substrates and products within metabolism bioreactors. In this study, we used an experimental design that ensured orthogonality between the different variables. We have demonstrated that with the HyperFlux™ PRO Plus Raman Spectroscopy System, it is possible to directly and accurately analyze many critical constituents down to at least 0.1 g/L (100 ppm) concentration in complex mixtures.
- B. N. Berry, T. M. Dobrowsky, R. C. Timson, R. Kshirsagar, T. Ryll and K. Wiltberger, “Quick generation of Raman spectroscopy based in-process glucose control to influence biopharmaceutical protein product quality during mammalian cell culture,” Biotechnology Progress, vol. 32, no. 1, pp. 224-234, 2016.
- T. E. Matthews, B. N. Berry, J. Smelko, J. Moretto, B. Moore and K. Wiltberger, “Closed loop control of lactate concentration in mammalian cell culture by raman spectroscopy leads to improved cell density, viability, and biopharmaceutical protein production,” Biotechnology and Bioengineering, vol. 113, no. 11, pp. 2416-2424, 2016.
- N. . R. Abu-Absi, B. . M. Kenty, M. Ehly Cuellar, M. C. Borys, S. Sakhamuri, D. J. Strachan, M. C. Hausladen and Z. J. Li, “Real Time Monitoring of Multiple Parameters in Mammalian Cell Culture Bioreactors Using an In- Line Raman Spectroscopy Probe,” Biotechnology and Bioengineering, vol. 108, pp. 1215-1221, 2011.
- J. Moretto, J. P. Smelko, M. Cuellar, B. Berry, A. Doane, T. Ryll and K. Wiltberger, “Process Raman Spectroscopy for In-Line CHO Cell Culture Monitoring,” American Pharmaceutical Review, vol. 14, pp. 18-25, 2011.
- H. Mehdizadeh, D. Lauri, K. M. Karry, M. Moshgbar, R. Procopio-Melino and D. Drapeau, “Generic Raman-Based Calibration Models Enabling Real-Time Monitoring of Cell Culture Bioreactors,” Biotechnology Progress, vol. 31, no. 4, pp. 1004-1013, 2015.
- J. Whelan, S. Craven and B. Glennon, “In Situ Raman Spectroscopy for Simultaneous Monitoring of Multiple Process Parameters in Mammalian Cell Culture Bioreactors,” Biotechnology Progress, vol. 28, pp. 1355- 1362, 2012.
- H. J. Butler, L. Ashton, B. Bird, G. Cinque, K. Curtis, J. Dorney, K. Esmonde-White, N. J. Fullwood, B. Gardner, P. L. Martin-Hirsch, M. J. Walsh, M. R. McAinsh, N. Stone and M. L. Francis, “Using Raman spectroscopy to characterize biological materials,” Nature Protocols, pp. 664-687, 2016.
- J. Aybar Muñoz and R. G. Brereton, “Partial factorial designs for multivariate calibration: extension to seven levels and comparison of strategy,” Chemometrics and Intelligent Laboratory Systems, vol. 43, p. 89–105, 1998.
- R. G. Brereton, “Multilevel Multifactor Designs for Multivariate Calibration,” The Analyst, vol. 122, p. 1521–1529, 1997.
This information has been sourced, reviewed and adapted from materials provided by Tornado Spectral Systems.
For more information on this source, please visit Tornado Spectral Systems.