- Email: [email protected]
In situ quantification of microcarrier animal cell cultures using near-infrared spectroscopy
Process Biochemistry 45 (2010) 1832–1836
Contents lists available at ScienceDirect
Process Biochemistry journal homepage: www.elsevier.com/locate/procbio
Short communication
In situ quantification of microcarrier animal cell cultures using near-infrared spectroscopy夽 Emma Petiot a, Patrick Bernard-Moulin b, Thierry Magadoux c, Cécile Gény c, Hervé Pinton c, Annie Marc a,∗ a
Laboratoire des Sciences du Génie Chimique, UPR CNRS 6811, Nancy-Université, 2 avenue de la Forêt de Haye, F-54505 Vandoeuvre-lès-Nancy Cedex, France ThermoFisher Scientific, 91963 Courtaboeuf Cedex, France c Sanofi pasteur, 1541 avenue Marcel Mérieux, F-69280 Marcy L’Etoile, France b
a r t i c l e
i n f o
Keywords: Adherent animal cell process Vero cell Near-infrared spectroscopy Microcarriers In situ monitoring
a b s t r a c t In-line monitoring tools are still required to understand and control animal cell processes, particularly in the case of vaccine production. Here, in situ near-infrared spectroscopy (NIRS) quantification of components in culture media was performed using microcarrier-based cultivations of adherent Vero cells. Because microcarriers were found to interfere with NIRS spectra acquisition, a suitable and innovative in situ calibration was developed for bioreactor cultures. A reliable and accurate NIRS technique for the quantification of glucose and lactate was established, with a calibration standard error of 0.30 and 0.21 g l−1 , respectively. The robustness of this method was evaluated by performing NIRS calibration with operating conditions similar to those of industrial processes, including parameters such as microcarrier concentrations, cell seeding states and changes in analyte concentration due to feed and harvest strategies. Based on this calibration procedure, the predicted analyte concentrations in unknown samples was measured by NIRS analyses with an accuracy of 0.36 g l−1 for glucose and 0.29 g l−1 for lactate. © 2010 Elsevier Ltd. All rights reserved.
1. Introduction A large number of human and veterinary viral vaccines are currently produced by large-scale cultures of adherent Vero cells cultivated on microcarriers [1–4]. Despite this important use, a lack in the understanding of Vero cell metabolism in serum-free media remains a hindrance for either the development of new industrial processes, or the optimization of already established ones [5]. Moreover, regulatory agencies have recently encouraged the development of the Process Analytical Technology (PAT) approach for biological production using animal cells, mainly to improve the in situ process monitoring through timely measurements [6]. To date, only a few process-related variables, including pH, temperature, pO2 , biovolume and cell density, can be monitored using in situ probes that can be sterilized [7,8]. Therefore, to access to the
DOI of original article:10.1016/j.procbio.2010.05.005. 夽 An error resulted in this article appearing in a previous volume of this journal. The article is reprinted here for the reader’s convenience and for the continuity of this special issue. The details of the original publication are as follows [Petiot et al., 2010. In situ quantification of microcarrier animal cell cultures using near-infrared spectroscopy. Process Biochemistry 45 (2010) 1427–1431. doi:10.1016/j.procbio.2010.05.005]. ∗ Corresponding author. Tel.: +33 383 595 785; fax: +33 383 595 804. E-mail address: [email protected] (A. Marc). 1359-5113/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.procbio.2010.08.010
real-time evolution of cell metabolism and physiology, the development of new PAT methods for on-line measurement of other key parameters, including nutrients, products or metabolic byproducts, is needed and could support large-scale production by cell culture. NIR spectroscopy (NIRS) is a promising alternative in situ PAT tool. NIRS is an absorption spectroscopy technique, providing a spectrum representative of the “signature” of all components present in the analyzed solution. It possesses numerous advantages when compared to classical analytical methods: rapid simultaneous detection of a large number of molecules, no requirement for chemical reagents or disposable materials, and the availability of a probe that can be sterilized. To date, in the field of biological processing NIRS has been mainly used for bacterial [9–12], microalgae [13] and fungal [14] cultures. In the case of animal cell culture processes, NIRS is currently not routinely used, with only few reports describing NIRS analysis of soluble components in cell culture supernatants [15–18]. All these studies were performed with CHO and insect cells cultivated in suspension mode, and similar work has not been reported for adherent cells. Only two publications report results using NIRS-implemented in situ in bioreactors [16,19], whereas the remaining reports were performed off-line using quartz cuvettes and were mainly focused on the calibration procedure [15,18,20–22]. The quality and robustness of calibration are highly dependent on the standard matrix composition and the evolution of the cell culture supernatant composition during the
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
culture process. Therefore, the off-line calibration and the generation of a standard formulation are often a challenge. In the present work, we evaluated NIRS as an in situ PAT tool to monitor glucose and lactate concentrations during bioreactor culture processes of Vero cells cultivated on microcarriers. A suitable and innovative in situ NIRS calibration procedure dedicated to adherent cell culture processes was developed. This method addresses matrix variations introduced by routine procedures used for adherent cell culture processes, such as different feeding strategies, cell trypsinization as well as microcarrier and analyte concentration variations. Therefore, this method is useful for the monitoring of industrial Vero cell culture processes.
1833
data set was composed of all calibration samples, while for the cross-validation step all samples were used as standards, except for two samples randomly chosen by the data treatment software. The second procedure was repeated to determine the standard error average of all the cross-validation correlations. The quality of the calibration models was then evaluated by several parameters: the correlation coefficients (r2 ) between concentrations obtained from NIRS analyses and from enzymatic analyses, calculated for both calibration and cross-validation data processing, and their respective standard errors (calibration (SEC) and cross-validation (SECV)). The glucose and lactate concentrations were also directly evaluated in samples not previously integrated in the calibration procedure. The prediction quality of NIRS analyses was then quantified by the standard error of prediction (SEP).
2. Materials and methods
3. Results and discussion
2.1. Vero cell cultures The Vero cell line used in this study was provided by Sanofi Pasteur (Marcy L’Etoile, France) and was adapted in a homemade, serum-free culture medium (the composition was described in a previous work) [5]. Cultures were performed using two concentrations of Cytodex 1 microcarriers (GE Healthcare Bioscience-AB, Sweden) in a 2 l bioreactor (Pierre Guérin, France) kept at pH 7.2, 37 ◦ C, with an agitation rate of 90 rpm. Oxygen concentration was set at 25% of air saturation. Seeding cells for bioreactor cultures (2.75 × 105 cell ml−1 ) were previously expanded either from a static culture in 1720 cm2 Hyperflasks® (Corning incorporated, USA) seeded with 3.15 × 105 cell ml−1 , from 500 ml spinner flasks, or from a bioreactor culture. Cultures in Hyperflasks® or spinner flasks were performed in an incubator controlled at 37 ◦ C and 5% CO2 . 2.2. Reference analyses For the quantification of viable cells attached to microcarriers, samples were washed twice with phosphate buffer saline (PBS) after microcarrier settling and then treated with Crystal Violet (Sigma, France) for at least 1 h at 37 ◦ C. The quantification of released nuclei was performed on a Fuchs-Rosenthal haemocytometer (Preciss, France) with 15% accuracy. Glucose and lactate concentrations in filtered cell culture supernatants were quantified by using the Vitalab selectra E enzymatic analyzer (Vital Scientific, The Netherlands). Analyses were carried out in triplicate with a standard error of about 0.07 g l−1 . 2.3. Near-infrared spectroscopy 2.3.1. NIR spectra acquisition Spectra acquisition was performed with a Fourier transform near-infrared (FTNIR) analyzer (Antaris II spectrometer; Thermofisher Scientific, France) for both off-line and in situ analyses. Spectra were collected in triplicate with an 8 cm−1 optical resolution, a spectral point spacing of 3.856 cm−1 , and with each spectra representing 128 co-added scans. A single air spectrum acquired prior to cell culture seeding was used as a reference spectrum. Spectra were obtained in a spectral range from 4000 to 9000 cm−1 . Off-line NIRS analyses were measured in quartz cuvettes with a 2-mm optical path. In situ measurements in bioreactor cultures were performed using a single optic fiber transflectance probe with a variable optical path (Series 625 transflectance probe from Precisions Sensing Devices, USA). The maximal path length of the probe was 20 mm, and NIRS radiations reflected by a mirror covered twice the fixed optical path. The numerical aperture was 0.22 and the collection efficiency was about 30%. The probe path length was fixed at 1 mm, leading to an effective optical path of 2 mm. 2.3.2. Development of NIRS calibration models Spectra processing was performed using the TQ Analyst software (Thermofisher Scientific, France). This software allows the user to perform both principal component analysis (PCA) and partial least square (PLS) regression, to optimize the wavelength range and to calculate the root mean square (RMS) noise values [23]. The spectra RMS noise was calculated using the following equation in a spectral region:
RMSnoise =
2
¯ − (yi − y)
2
¯ (xi − x¯ ) (yi − y)
2
2 /
(xi − x¯ )
2
n−2
¯ average of y yi : absorbance intensity at data point i; xi : value of the data point i; y: data values within the spectral range; x¯ : average of x data values within the spectral range; n: number of data points within the spectral range. Spectra were processed as absorbance spectra and no mathematical treatment was needed, except for the baseline correction. NIRS calibration models were established for glucose and lactate by using a spectral range and a number of partial least square (PLS) factors specific and optimal for each analyte. Two data processing steps were performed to establish the calibration models. For the first step the
3.1. Influence of microcarriers on NIRS analyses of Vero cell culture supernatants Previously the calibration procedure of NIRS has been mainly performed off-line using standards from various chemical formulations, including samples with non-correlated concentrations of the quantified molecules [18–21]. In some cases, these off-line calibrations were then used for in situ monitoring of animal cell culture suspensions [19]. However, the analyzed matrix is composed of both culture medium and beads when cultivating adherent cells growing on microcarriers. The presence of microcarriers in the NIRS probe sensing space could then modify the real optical path length and interfere with the spectra quality and intensity. To evaluate this effect, microcarriers were added to culture supernatants before off-line NIR spectra acquisition in quartz cuvettes. Comparison of the spectra recorded with or without the addition of the microcarriers showed a variation in absorbance intensity, even though they had similar profiles (Fig. 1: spectra A and B). By comparing the path length in filtrated sample (PN ), it could be hypothesized that the optical path length in the quartz cuvette was increased due to the presence of microcarriers (PM ). This hypothesis could explain the higher spectra intensity observed in sample containing beads. For an accurate in situ quantification of analyte concentrations, it is recommended to calibrate using samples that best represent the matrix of interest, i.e., the supernatant during cell culture. During off-line analyses in quartz cuvettes, the rapid settling of microcarriers resulted in an increased bead concentration in the NIRS sensing space. This bead concentration was significantly higher than the bead concentration in agitated cultures. Indeed, agitation condition is an important element contributing to variations in matrix composition, especially in the case of a heterogeneous system such as microcarrier cultures. To evaluate the effect of sample agitation, three spectra acquired in presence of microcarriers were compared (Fig. 1). The spectra acquired, either off-line in static cuvettes and in a stirred 10 ml sample tube, or in situ in a stirred bioreactor culture, showed different intensities. First, these results demonstrated that off-line calibration in static conditions did not allow for a correct calibration procedure using samples containing beads which have to be maintained in a homogeneous suspension. Secondly, the use of different agitation systems provoked various trajectories and bead movement speeds within the optical field. This variation led to an observed gap in intensity between the two spectra acquired with the fiber optic probe in stirred conditions (spectra C and D). Thus, an off-line calibration procedure, either in static or stirred conditions, is not ideal for cultures performed with cells attached to microcarriers. Therefore, it was necessary to develop an in situ calibration procedure which can be performed directly in the bioreactor vessel, to ensure that calibration reflects the medium-bead matrix during cultivation.
1834
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
Fig. 1. Off-line spectra (acquired in quartz cuvette) of cell culture supernatants containing (A) or not containing (B) microcarriers. Off-line spectrum (acquired in a sample tube) of a supernatant containing microcarriers (C). In situ spectrum (acquired in bioreactor) of a supernatant from cells cultured on microcarriers (D). Schematic representations of path length with (PM ) or without (PN ) microcarriers.
3.2. In situ NIRS calibration for Vero cell bioreactor cultures 3.2.1. Kinetics of cultures performed under various operating conditions To develop in situ calibration models that can be used for largescale culture processes, some specific aspects had to be addressed. Important changes in the matrix composition result from classical large-scale culture operations. In the first place, industrial processes often use various medium feeding strategies, such as fed-batch mode or punctual medium replacement, to increase the final cell density and product level. This could dramatically change the matrix composition within few minutes. Similarly, in adherent cell culture processes, cell propagation protocols required to detach cells, contribute to the variation of the physiological state of cells seeded in the reactor, and consequently of cell nutrient consumption and molecule release in the culture supernatant. Additionally, process scale-up and cell amplification generally require an increase of bead concentration to allow for higher cell densities. All these events highly affect the medium-bead matrix composition. To deal with these sources of variability and ensure the quality and robustness of the NIRS in situ monitoring, calibrations were
performed for glucose and lactate using various cultivation conditions. These included various microcarrier concentrations, cell propagation protocols and medium feed-harvest. A first cultivation (batch 1) was inoculated with cells trypsinized from a static Hyperflask® culture. After 3 days of culture, a medium feed-harvest operation was performed by replacing 80% of the spent medium by fresh medium and the culture was prolonged for 2 additional days. Batch 2 was then started at higher microcarrier concentration (1.7 times) with cells trypsinized at the end of the first culture and continued for 4 days without medium feed-harvest. Cell culture supernatants were sampled all over the process while medium NIR spectra were acquired in situ at the same time. Cell growth, glucose consumption and lactate production kinetics for both cultures are shown in Fig. 2. During batch 1, the cell concentration increased until 4.5 × 105 cell ml−1 after 3 days of culture, while glucose was consumed and lactate accumulated. After medium replacement, the cell growth continued, reaching a maximum cell density of 9.6 × 105 cell ml−1 . Growth during batch 2 was similar at the beginning of the culture, but the maximum cell density stabilized at around 6.2 × 105 cell ml−1 . During this stationary phase, glucose consumption was prolonged, leading to non-correlated glucose and lactate concentrations during the last phase of this culture.
Fig. 2. Kinetics of cell growth ( ), glucose consumption () and lactate production (䊉) over two successive batch cultures of Vero cells in a bioreactor. Batch 1: cells from static culture. Batch 2: cells trypsinized from batch 1 microcarriers. Glucose and lactate concentrations were obtained by enzymatic assays.
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
1835
between glucose and lactate concentrations within the calibration models.
Fig. 3. Glucose and lactate concentrations measured by enzymatic methods in bioreactor cultures; batch 1 (), batch 2 before (䊉) and after ( ) medium exchange, batch 3 (). Linearly correlated concentrations are circled in white, whereas uncorrelated data are circled in grey.
3.2.2. Vero cell culture with non-correlated concentrations of glucose and lactate In mammalian cell culture, glucose consumption is generally coupled to lactate production, because an important part of glucose is converted into this metabolite through glycolysis. A high degree of correlation between lactate and glucose concentrations in the cell culture supernatants can occur, as shown for the concentrations of lactate and glucose assayed enzymatically from the two previous batch cultures (Fig. 3). Consequently, as NIRS calibrations are based on the detection of evolution in the spectra, the TQ analysis with PCA/PLS processing is not able to independently quantify glucose and lactate concentrations. To address this problem, several authors reported NIRS calibration with mixtures containing non-correlated concentrations of the studied analytes, but all of these studies are based on off-line methods and are consequently not relevant for the present work. It was therefore investigated whether this problem could be overcome by artificially creating non-correlated analyte concentrations inside the bioreactor. The goal was to train the NIRS calibration model to accurately quantify non-correlated glucose and lactate concentrations during the process. Consequently, a third batch culture was performed with punctual and separate additions of known amount of glucose and lactate after the medium replacement (Fig. 3). The white zone clearly indicates a strong linear correlation between the two solutes, which was also observed in routine cultures. In contrast, the grey zones pointed out non-correlated data. These were representing either the end of the second batch or punctual analyte additions in the third batch. The r-square values before and after addition of these new standards to the data set were 0.94 and 0.65, respectively. The incorporation of in situ NIRS spectra acquired for these non-correlated points thus reduced the degree of colinearity
3.2.3. Calibration and prediction of analyte concentrations by in situ NIRS monitoring NIRS analyses were performed at different time points (n = 85) during the three batch cultures. From these data, randomly chosen analyses with various medium compositions were processed (n = 73) to establish the calibration model for glucose and lactate concentrations. As shown in Table 1, the calibration resulted in high correlation coefficients of glucose and lactate calibration (r2 > 0.90). Standard errors of calibration (SEC) were in the same range as reported by other studies for suspension cell cultures using in situ probes or on-line measurements [16,19,24]. Similarly, the standard errors obtained in cross-validation models (SECV) were slightly increased for the two analytes [19,24]. The calculated RMS noise values for different spectral ranges were 0.9 × 10−3 (4300–4700 cm−1 ), 0.4 × 10−3 (5400–6500 cm−1 ) and 1.1 × 10−3 (7500–8000 cm−1 ). A maximum increase of 0.3 × 10−3 of RMS noise values could be observed when a higher concentration of microcarriers was used. This could have been caused by a reduction of the radiant power at the detector. Indeed, scattering lowers the signal-to-noise ratio of the measurement, thereby increasing the RMS noise values. Our results demonstrated the feasibility of in situ NIRS calibration for glucose and lactate determination, resulting in satisfying correlation coefficients, standard errors and numbers of PLS factors, independent of the bioreactor operating conditions. Moreover, calibration quality can be further improved by additional NIRS monitoring resulting from future cultures. Our novel calibration procedure, although using a quite complex calibration scheme, was very efficient because only 73 samples provided a reliable calibration model. Based on these results, NIRS can be considered a useful tool for in situ monitoring of Vero cell cultures. The last step to validate the NIRS method for adherent cell cultures was to evaluate its prediction accuracy for glucose and lactate quantification during the whole cultivation. Thus, 12 randomized samples that were not introduced in the calibration procedure were analyzed. Glucose and lactate concentrations in these samples were quantified directly from the NIR spectra using the calibration models previously established. Prediction accuracy of the estimated values was then evaluated on the basis of the standard error of prediction (SEP) for each analyte and by comparison with the reference data. Despite the introduced variations in culture conditions, such as medium feed-harvest, microcarrier concentration or punctual addition of glucose and lactate for concentration de-correlation, the obtained SEP values were low, with values of 0.36 and 0.29 g l−1 for glucose and lactate, respectively, and in the same range as the SEC values calculated during the calibration procedure. Moreover, to date, none of the studies could demonstrate in situ data acquisition for concentrations below 0.3 g l−1 (glucose) and 1 g l−1 (lactate) [16,19,25]. Our work demonstrates that NIRS presents potentialities to monitor low glucose concentration often found in fed-batch cultures.
Table 1 Calibration and prediction results of in situ glucose and lactate NIRS quantification from bioreactor cultures with Vero cells on microcarriers (73 standards for calibration models and 12 standards for prediction). Spectral range (cm−1 )
PLS factors
Calibration
Cross-validation
Prediction
r2
SEC (g l−1 )
r2
SECV (g l−1 )
SEP (g l−1 )
Glucose
4308–4784 5418–6513
6
0.95
0.30
0.86
0.36
0.36
Lactate
4346–4755 5349–6452 7504–8775
8
0.94
0.21
0.88
0.27
0.29
1836
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
4. Conclusion This paper reports, for the first time, a new PAT technique using a NIRS fiber optic probe for in situ monitoring of the concentration of key analytes, such as glucose and lactate, during bioreactor processes performed with animal cells attached to microcarriers. This tool was able to predict glucose and lactate concentrations in situ with good accuracy within the concentration ranges routinely observed in Vero cell culture supernatants. The method was also reliable using various operating conditions and applicable for scale-up. These results were particularly interesting because, they are useful not only to monitor conventional Vero cell culture processes, but this in situ NIRS tool is also useful in controlling unusual concentrations and for the development of risk-mitigation plans for the specific case of adherent cell cultures. Further applications could also be envisioned, such as real-time monitoring of the cell viability by quantification of the lactate dehydrogenase enzyme released during cell death or of the production itself. As a consequence, in situ near-infrared spectroscopy is a very promising tool for current and further applications of the PAT strategy for adherent cell culture processes. Acknowledgements The authors are grateful to Dr Emmanuel Guedon (LSGC-CNRS, Nancy) and to Sven Ansorge (Ecole Polytechnique - CNRC, Montreal) for the help they provided in manuscript writing. References [1] Barrett PN, Mundt W, Kistner O, Howard MK. Vero cell platform in vaccine production: Moving towards cell culture-based viral vaccines. Expert Rev Vaccines 2009;8:607–18. [2] Liu C-C, Lian W-C, Butler M, Wu S-C. High immunogenic enterovirus 71 strain and its production using serum-free microcarrier Vero cell culture. Vaccine 2007;25:19–24. [3] Rourou S, van der Ark A, van der Velden T, Kallel H. A microcarrier cell culture process for propagating rabies virus in Vero cells grown in a stirred bioreactor under fully animal component free conditions. Vaccine 2007;25:3879–89. [4] Toriniwa H, Komiya T. Long-term stability of Vero cell-derived inactivated Japanese encephalitis vaccine prepared using serum-free medium. Vaccine 2008;26:3680–9. [5] Petiot E, Fournier F, Gény C, Pinton H, Marc A. Rapid screening of serum-free media for the growth of adherent Vero cells by using a small-scale and noninvasive tool. Appl Biochem Biotechnol 2010;160:1600–15. [6] Mandenius C-F, Graumann K, Schultz TW, Premstaller A, Olsson I-M, Petiot E, et al. Quality by design (QbD) for biotechnology-related pharmaceuticals. Biotechnol J 2009;4:600–9.
[7] Ansorge S, Esteban G, Schmid G. On-line monitoring of infected Sf-9 insect cell cultures by scanning permittivity measurements and comparison with off-line biovolume measurements. Cytotechnology 2007;55:115–24. [8] Rudolph G, Lindner P, Gierse A, Bluma A, Martinez G, Hitzmann B, et al. On-line monitoring of microcarrier based fibroblast cultivations with in situ microscopy. Biotechnol Bioeng 2008;99:136–45. [9] Arnold SA, Gaensakoo R, Harvey LM, McNeil B. Use of at-line and in-situ nearinfrared spectroscopy to monitor biomass in an industrial fed-batch Escherichia coli process. Biotechnol Bioeng 2002;80:405–13. [10] Scarff M, Arnold SA, Harvey LM, McNeil B. Near infrared spectroscopy for bioprocess monitoring and control: current status and future trends. Crit Rev Biotechnol 2006;26:17–39. [11] Zhang L, Small GW, Arnold MA. Multivariate calibration standardization across instruments for the determination of glucose by Fourier transform nearinfrared spectrometry. Anal Chem 2003;75:5905–15. [12] Tosi S, Rossi M, Tamburini E, Vaccari G, Amaretti A, Matteuzzi D. Assessment of in-line near-infrared spectroscopy for continuous monitoring of fermentation processes. Biotechnol Prog 2003;19:1816–21. [13] Sandnes JM, Ringstad T, Wenner D, Heyerdahl PH, Kallqvist T, Gislerod HR. Real-time monitoring and automatic density control of large-scale microalgal cultures using near infrared (NIR) optical density sensors. J Biotechnol 2006;122:209–15. [14] Vaidyanathan S, Harvey LM, McNeil B. Deconvolution of near-infrared spectral information for monitoring mycelial biomass and other key analytes in a submerged fungal bioprocess. Anal Chim Acta 2001;428:41–59. [15] Riley MR, Rhiel M, Zhou X, Arnold MA, Murhammer DW. Simultaneous measurement of glucose and glutamine in insect cell culture media by near infrared spectroscopy. Biotechnol Bioeng 1997;55:11–5. [16] Arnold SA, Crowley J, Woods N, Harvey LM, McNeil B. In-situ near infrared spectroscopy to monitor key analytes in mammalian cell cultivation. Biotechnol Bioeng 2003;84:13–9. [17] Harthun S, Matischak K, Friedl P. Determination of recombinant protein in animal cell culture supernatant by near-infrared spectroscopy. Anal Biochem 1997;251:73–8. [18] Riley MR, Crider HM, Nite ME, Garcia RA, Woo J, Wegge RM. Simultaneous measurement of 19 components in serum-containing animal cell culture media by Fourier transform near-infrared spectroscopy. Biotechnol Prog 2001;17:376–8. [19] Roychoudhury P, O’Kennedy R, McNeil B, Harvey LM. Multiplexing fibre optic near infrared (NIR) spectroscopy as an emerging technology to monitor industrial bioprocesses. Anal Chim Acta 2007;590:110–7. [20] Riley MR, Arnold MA, Murhammer DW, Walls EL, DelaCruz N. Adaptive calibration scheme for quantification of nutrients and byproducts in insect cell bioreactors by near-infrared spectroscopy. Biotechnol Prog 1998;14:527–33. [21] Riley MR, Okeson CD, Frazier BL. Rapid calibration of near-infrared spectroscopic measurements of mammalian cell cultivations. Biotechnol Prog 1999;15:1133–41. [22] Rhiel M, Cohen MB, Murhammer DW, Arnold MA. Nondestructive near-infrared spectroscopic measurement of multiple analytes in undiluted samples of serum-based cell culture media. Biotechnol Bioeng 2002;77:73–82. [23] Teixeira AP, Oliveira R, Alves PM, Carrondo MJT. Advances in on-line monitoring and control of mammalian cell cultures: supporting the PAT initiative. Biotechnol Adv 2009;27:726–32. [24] Card C, Hunsaker B, Smith T, Hirsch J. Rapid, simultaneous monitoring of multiple components in mammalian cell culture with near-infrared spectroscopy. Thermo Fisher Scientific; 2007 [Application Note: 51449]. [25] Kennedy M, Krouse D. Strategies for improving fermentation medium performance: a review. J Ind Microbiol Biotechnol 1999;23:456–75.
Contents lists available at ScienceDirect
Process Biochemistry journal homepage: www.elsevier.com/locate/procbio
Short communication
In situ quantification of microcarrier animal cell cultures using near-infrared spectroscopy夽 Emma Petiot a, Patrick Bernard-Moulin b, Thierry Magadoux c, Cécile Gény c, Hervé Pinton c, Annie Marc a,∗ a
Laboratoire des Sciences du Génie Chimique, UPR CNRS 6811, Nancy-Université, 2 avenue de la Forêt de Haye, F-54505 Vandoeuvre-lès-Nancy Cedex, France ThermoFisher Scientific, 91963 Courtaboeuf Cedex, France c Sanofi pasteur, 1541 avenue Marcel Mérieux, F-69280 Marcy L’Etoile, France b
a r t i c l e
i n f o
Keywords: Adherent animal cell process Vero cell Near-infrared spectroscopy Microcarriers In situ monitoring
a b s t r a c t In-line monitoring tools are still required to understand and control animal cell processes, particularly in the case of vaccine production. Here, in situ near-infrared spectroscopy (NIRS) quantification of components in culture media was performed using microcarrier-based cultivations of adherent Vero cells. Because microcarriers were found to interfere with NIRS spectra acquisition, a suitable and innovative in situ calibration was developed for bioreactor cultures. A reliable and accurate NIRS technique for the quantification of glucose and lactate was established, with a calibration standard error of 0.30 and 0.21 g l−1 , respectively. The robustness of this method was evaluated by performing NIRS calibration with operating conditions similar to those of industrial processes, including parameters such as microcarrier concentrations, cell seeding states and changes in analyte concentration due to feed and harvest strategies. Based on this calibration procedure, the predicted analyte concentrations in unknown samples was measured by NIRS analyses with an accuracy of 0.36 g l−1 for glucose and 0.29 g l−1 for lactate. © 2010 Elsevier Ltd. All rights reserved.
1. Introduction A large number of human and veterinary viral vaccines are currently produced by large-scale cultures of adherent Vero cells cultivated on microcarriers [1–4]. Despite this important use, a lack in the understanding of Vero cell metabolism in serum-free media remains a hindrance for either the development of new industrial processes, or the optimization of already established ones [5]. Moreover, regulatory agencies have recently encouraged the development of the Process Analytical Technology (PAT) approach for biological production using animal cells, mainly to improve the in situ process monitoring through timely measurements [6]. To date, only a few process-related variables, including pH, temperature, pO2 , biovolume and cell density, can be monitored using in situ probes that can be sterilized [7,8]. Therefore, to access to the
DOI of original article:10.1016/j.procbio.2010.05.005. 夽 An error resulted in this article appearing in a previous volume of this journal. The article is reprinted here for the reader’s convenience and for the continuity of this special issue. The details of the original publication are as follows [Petiot et al., 2010. In situ quantification of microcarrier animal cell cultures using near-infrared spectroscopy. Process Biochemistry 45 (2010) 1427–1431. doi:10.1016/j.procbio.2010.05.005]. ∗ Corresponding author. Tel.: +33 383 595 785; fax: +33 383 595 804. E-mail address: [email protected] (A. Marc). 1359-5113/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.procbio.2010.08.010
real-time evolution of cell metabolism and physiology, the development of new PAT methods for on-line measurement of other key parameters, including nutrients, products or metabolic byproducts, is needed and could support large-scale production by cell culture. NIR spectroscopy (NIRS) is a promising alternative in situ PAT tool. NIRS is an absorption spectroscopy technique, providing a spectrum representative of the “signature” of all components present in the analyzed solution. It possesses numerous advantages when compared to classical analytical methods: rapid simultaneous detection of a large number of molecules, no requirement for chemical reagents or disposable materials, and the availability of a probe that can be sterilized. To date, in the field of biological processing NIRS has been mainly used for bacterial [9–12], microalgae [13] and fungal [14] cultures. In the case of animal cell culture processes, NIRS is currently not routinely used, with only few reports describing NIRS analysis of soluble components in cell culture supernatants [15–18]. All these studies were performed with CHO and insect cells cultivated in suspension mode, and similar work has not been reported for adherent cells. Only two publications report results using NIRS-implemented in situ in bioreactors [16,19], whereas the remaining reports were performed off-line using quartz cuvettes and were mainly focused on the calibration procedure [15,18,20–22]. The quality and robustness of calibration are highly dependent on the standard matrix composition and the evolution of the cell culture supernatant composition during the
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
culture process. Therefore, the off-line calibration and the generation of a standard formulation are often a challenge. In the present work, we evaluated NIRS as an in situ PAT tool to monitor glucose and lactate concentrations during bioreactor culture processes of Vero cells cultivated on microcarriers. A suitable and innovative in situ NIRS calibration procedure dedicated to adherent cell culture processes was developed. This method addresses matrix variations introduced by routine procedures used for adherent cell culture processes, such as different feeding strategies, cell trypsinization as well as microcarrier and analyte concentration variations. Therefore, this method is useful for the monitoring of industrial Vero cell culture processes.
1833
data set was composed of all calibration samples, while for the cross-validation step all samples were used as standards, except for two samples randomly chosen by the data treatment software. The second procedure was repeated to determine the standard error average of all the cross-validation correlations. The quality of the calibration models was then evaluated by several parameters: the correlation coefficients (r2 ) between concentrations obtained from NIRS analyses and from enzymatic analyses, calculated for both calibration and cross-validation data processing, and their respective standard errors (calibration (SEC) and cross-validation (SECV)). The glucose and lactate concentrations were also directly evaluated in samples not previously integrated in the calibration procedure. The prediction quality of NIRS analyses was then quantified by the standard error of prediction (SEP).
2. Materials and methods
3. Results and discussion
2.1. Vero cell cultures The Vero cell line used in this study was provided by Sanofi Pasteur (Marcy L’Etoile, France) and was adapted in a homemade, serum-free culture medium (the composition was described in a previous work) [5]. Cultures were performed using two concentrations of Cytodex 1 microcarriers (GE Healthcare Bioscience-AB, Sweden) in a 2 l bioreactor (Pierre Guérin, France) kept at pH 7.2, 37 ◦ C, with an agitation rate of 90 rpm. Oxygen concentration was set at 25% of air saturation. Seeding cells for bioreactor cultures (2.75 × 105 cell ml−1 ) were previously expanded either from a static culture in 1720 cm2 Hyperflasks® (Corning incorporated, USA) seeded with 3.15 × 105 cell ml−1 , from 500 ml spinner flasks, or from a bioreactor culture. Cultures in Hyperflasks® or spinner flasks were performed in an incubator controlled at 37 ◦ C and 5% CO2 . 2.2. Reference analyses For the quantification of viable cells attached to microcarriers, samples were washed twice with phosphate buffer saline (PBS) after microcarrier settling and then treated with Crystal Violet (Sigma, France) for at least 1 h at 37 ◦ C. The quantification of released nuclei was performed on a Fuchs-Rosenthal haemocytometer (Preciss, France) with 15% accuracy. Glucose and lactate concentrations in filtered cell culture supernatants were quantified by using the Vitalab selectra E enzymatic analyzer (Vital Scientific, The Netherlands). Analyses were carried out in triplicate with a standard error of about 0.07 g l−1 . 2.3. Near-infrared spectroscopy 2.3.1. NIR spectra acquisition Spectra acquisition was performed with a Fourier transform near-infrared (FTNIR) analyzer (Antaris II spectrometer; Thermofisher Scientific, France) for both off-line and in situ analyses. Spectra were collected in triplicate with an 8 cm−1 optical resolution, a spectral point spacing of 3.856 cm−1 , and with each spectra representing 128 co-added scans. A single air spectrum acquired prior to cell culture seeding was used as a reference spectrum. Spectra were obtained in a spectral range from 4000 to 9000 cm−1 . Off-line NIRS analyses were measured in quartz cuvettes with a 2-mm optical path. In situ measurements in bioreactor cultures were performed using a single optic fiber transflectance probe with a variable optical path (Series 625 transflectance probe from Precisions Sensing Devices, USA). The maximal path length of the probe was 20 mm, and NIRS radiations reflected by a mirror covered twice the fixed optical path. The numerical aperture was 0.22 and the collection efficiency was about 30%. The probe path length was fixed at 1 mm, leading to an effective optical path of 2 mm. 2.3.2. Development of NIRS calibration models Spectra processing was performed using the TQ Analyst software (Thermofisher Scientific, France). This software allows the user to perform both principal component analysis (PCA) and partial least square (PLS) regression, to optimize the wavelength range and to calculate the root mean square (RMS) noise values [23]. The spectra RMS noise was calculated using the following equation in a spectral region:
RMSnoise =
2
¯ − (yi − y)
2
¯ (xi − x¯ ) (yi − y)
2
2 /
(xi − x¯ )
2
n−2
¯ average of y yi : absorbance intensity at data point i; xi : value of the data point i; y: data values within the spectral range; x¯ : average of x data values within the spectral range; n: number of data points within the spectral range. Spectra were processed as absorbance spectra and no mathematical treatment was needed, except for the baseline correction. NIRS calibration models were established for glucose and lactate by using a spectral range and a number of partial least square (PLS) factors specific and optimal for each analyte. Two data processing steps were performed to establish the calibration models. For the first step the
3.1. Influence of microcarriers on NIRS analyses of Vero cell culture supernatants Previously the calibration procedure of NIRS has been mainly performed off-line using standards from various chemical formulations, including samples with non-correlated concentrations of the quantified molecules [18–21]. In some cases, these off-line calibrations were then used for in situ monitoring of animal cell culture suspensions [19]. However, the analyzed matrix is composed of both culture medium and beads when cultivating adherent cells growing on microcarriers. The presence of microcarriers in the NIRS probe sensing space could then modify the real optical path length and interfere with the spectra quality and intensity. To evaluate this effect, microcarriers were added to culture supernatants before off-line NIR spectra acquisition in quartz cuvettes. Comparison of the spectra recorded with or without the addition of the microcarriers showed a variation in absorbance intensity, even though they had similar profiles (Fig. 1: spectra A and B). By comparing the path length in filtrated sample (PN ), it could be hypothesized that the optical path length in the quartz cuvette was increased due to the presence of microcarriers (PM ). This hypothesis could explain the higher spectra intensity observed in sample containing beads. For an accurate in situ quantification of analyte concentrations, it is recommended to calibrate using samples that best represent the matrix of interest, i.e., the supernatant during cell culture. During off-line analyses in quartz cuvettes, the rapid settling of microcarriers resulted in an increased bead concentration in the NIRS sensing space. This bead concentration was significantly higher than the bead concentration in agitated cultures. Indeed, agitation condition is an important element contributing to variations in matrix composition, especially in the case of a heterogeneous system such as microcarrier cultures. To evaluate the effect of sample agitation, three spectra acquired in presence of microcarriers were compared (Fig. 1). The spectra acquired, either off-line in static cuvettes and in a stirred 10 ml sample tube, or in situ in a stirred bioreactor culture, showed different intensities. First, these results demonstrated that off-line calibration in static conditions did not allow for a correct calibration procedure using samples containing beads which have to be maintained in a homogeneous suspension. Secondly, the use of different agitation systems provoked various trajectories and bead movement speeds within the optical field. This variation led to an observed gap in intensity between the two spectra acquired with the fiber optic probe in stirred conditions (spectra C and D). Thus, an off-line calibration procedure, either in static or stirred conditions, is not ideal for cultures performed with cells attached to microcarriers. Therefore, it was necessary to develop an in situ calibration procedure which can be performed directly in the bioreactor vessel, to ensure that calibration reflects the medium-bead matrix during cultivation.
1834
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
Fig. 1. Off-line spectra (acquired in quartz cuvette) of cell culture supernatants containing (A) or not containing (B) microcarriers. Off-line spectrum (acquired in a sample tube) of a supernatant containing microcarriers (C). In situ spectrum (acquired in bioreactor) of a supernatant from cells cultured on microcarriers (D). Schematic representations of path length with (PM ) or without (PN ) microcarriers.
3.2. In situ NIRS calibration for Vero cell bioreactor cultures 3.2.1. Kinetics of cultures performed under various operating conditions To develop in situ calibration models that can be used for largescale culture processes, some specific aspects had to be addressed. Important changes in the matrix composition result from classical large-scale culture operations. In the first place, industrial processes often use various medium feeding strategies, such as fed-batch mode or punctual medium replacement, to increase the final cell density and product level. This could dramatically change the matrix composition within few minutes. Similarly, in adherent cell culture processes, cell propagation protocols required to detach cells, contribute to the variation of the physiological state of cells seeded in the reactor, and consequently of cell nutrient consumption and molecule release in the culture supernatant. Additionally, process scale-up and cell amplification generally require an increase of bead concentration to allow for higher cell densities. All these events highly affect the medium-bead matrix composition. To deal with these sources of variability and ensure the quality and robustness of the NIRS in situ monitoring, calibrations were
performed for glucose and lactate using various cultivation conditions. These included various microcarrier concentrations, cell propagation protocols and medium feed-harvest. A first cultivation (batch 1) was inoculated with cells trypsinized from a static Hyperflask® culture. After 3 days of culture, a medium feed-harvest operation was performed by replacing 80% of the spent medium by fresh medium and the culture was prolonged for 2 additional days. Batch 2 was then started at higher microcarrier concentration (1.7 times) with cells trypsinized at the end of the first culture and continued for 4 days without medium feed-harvest. Cell culture supernatants were sampled all over the process while medium NIR spectra were acquired in situ at the same time. Cell growth, glucose consumption and lactate production kinetics for both cultures are shown in Fig. 2. During batch 1, the cell concentration increased until 4.5 × 105 cell ml−1 after 3 days of culture, while glucose was consumed and lactate accumulated. After medium replacement, the cell growth continued, reaching a maximum cell density of 9.6 × 105 cell ml−1 . Growth during batch 2 was similar at the beginning of the culture, but the maximum cell density stabilized at around 6.2 × 105 cell ml−1 . During this stationary phase, glucose consumption was prolonged, leading to non-correlated glucose and lactate concentrations during the last phase of this culture.
Fig. 2. Kinetics of cell growth ( ), glucose consumption () and lactate production (䊉) over two successive batch cultures of Vero cells in a bioreactor. Batch 1: cells from static culture. Batch 2: cells trypsinized from batch 1 microcarriers. Glucose and lactate concentrations were obtained by enzymatic assays.
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
1835
between glucose and lactate concentrations within the calibration models.
Fig. 3. Glucose and lactate concentrations measured by enzymatic methods in bioreactor cultures; batch 1 (), batch 2 before (䊉) and after ( ) medium exchange, batch 3 (). Linearly correlated concentrations are circled in white, whereas uncorrelated data are circled in grey.
3.2.2. Vero cell culture with non-correlated concentrations of glucose and lactate In mammalian cell culture, glucose consumption is generally coupled to lactate production, because an important part of glucose is converted into this metabolite through glycolysis. A high degree of correlation between lactate and glucose concentrations in the cell culture supernatants can occur, as shown for the concentrations of lactate and glucose assayed enzymatically from the two previous batch cultures (Fig. 3). Consequently, as NIRS calibrations are based on the detection of evolution in the spectra, the TQ analysis with PCA/PLS processing is not able to independently quantify glucose and lactate concentrations. To address this problem, several authors reported NIRS calibration with mixtures containing non-correlated concentrations of the studied analytes, but all of these studies are based on off-line methods and are consequently not relevant for the present work. It was therefore investigated whether this problem could be overcome by artificially creating non-correlated analyte concentrations inside the bioreactor. The goal was to train the NIRS calibration model to accurately quantify non-correlated glucose and lactate concentrations during the process. Consequently, a third batch culture was performed with punctual and separate additions of known amount of glucose and lactate after the medium replacement (Fig. 3). The white zone clearly indicates a strong linear correlation between the two solutes, which was also observed in routine cultures. In contrast, the grey zones pointed out non-correlated data. These were representing either the end of the second batch or punctual analyte additions in the third batch. The r-square values before and after addition of these new standards to the data set were 0.94 and 0.65, respectively. The incorporation of in situ NIRS spectra acquired for these non-correlated points thus reduced the degree of colinearity
3.2.3. Calibration and prediction of analyte concentrations by in situ NIRS monitoring NIRS analyses were performed at different time points (n = 85) during the three batch cultures. From these data, randomly chosen analyses with various medium compositions were processed (n = 73) to establish the calibration model for glucose and lactate concentrations. As shown in Table 1, the calibration resulted in high correlation coefficients of glucose and lactate calibration (r2 > 0.90). Standard errors of calibration (SEC) were in the same range as reported by other studies for suspension cell cultures using in situ probes or on-line measurements [16,19,24]. Similarly, the standard errors obtained in cross-validation models (SECV) were slightly increased for the two analytes [19,24]. The calculated RMS noise values for different spectral ranges were 0.9 × 10−3 (4300–4700 cm−1 ), 0.4 × 10−3 (5400–6500 cm−1 ) and 1.1 × 10−3 (7500–8000 cm−1 ). A maximum increase of 0.3 × 10−3 of RMS noise values could be observed when a higher concentration of microcarriers was used. This could have been caused by a reduction of the radiant power at the detector. Indeed, scattering lowers the signal-to-noise ratio of the measurement, thereby increasing the RMS noise values. Our results demonstrated the feasibility of in situ NIRS calibration for glucose and lactate determination, resulting in satisfying correlation coefficients, standard errors and numbers of PLS factors, independent of the bioreactor operating conditions. Moreover, calibration quality can be further improved by additional NIRS monitoring resulting from future cultures. Our novel calibration procedure, although using a quite complex calibration scheme, was very efficient because only 73 samples provided a reliable calibration model. Based on these results, NIRS can be considered a useful tool for in situ monitoring of Vero cell cultures. The last step to validate the NIRS method for adherent cell cultures was to evaluate its prediction accuracy for glucose and lactate quantification during the whole cultivation. Thus, 12 randomized samples that were not introduced in the calibration procedure were analyzed. Glucose and lactate concentrations in these samples were quantified directly from the NIR spectra using the calibration models previously established. Prediction accuracy of the estimated values was then evaluated on the basis of the standard error of prediction (SEP) for each analyte and by comparison with the reference data. Despite the introduced variations in culture conditions, such as medium feed-harvest, microcarrier concentration or punctual addition of glucose and lactate for concentration de-correlation, the obtained SEP values were low, with values of 0.36 and 0.29 g l−1 for glucose and lactate, respectively, and in the same range as the SEC values calculated during the calibration procedure. Moreover, to date, none of the studies could demonstrate in situ data acquisition for concentrations below 0.3 g l−1 (glucose) and 1 g l−1 (lactate) [16,19,25]. Our work demonstrates that NIRS presents potentialities to monitor low glucose concentration often found in fed-batch cultures.
Table 1 Calibration and prediction results of in situ glucose and lactate NIRS quantification from bioreactor cultures with Vero cells on microcarriers (73 standards for calibration models and 12 standards for prediction). Spectral range (cm−1 )
PLS factors
Calibration
Cross-validation
Prediction
r2
SEC (g l−1 )
r2
SECV (g l−1 )
SEP (g l−1 )
Glucose
4308–4784 5418–6513
6
0.95
0.30
0.86
0.36
0.36
Lactate
4346–4755 5349–6452 7504–8775
8
0.94
0.21
0.88
0.27
0.29
1836
E. Petiot et al. / Process Biochemistry 45 (2010) 1832–1836
4. Conclusion This paper reports, for the first time, a new PAT technique using a NIRS fiber optic probe for in situ monitoring of the concentration of key analytes, such as glucose and lactate, during bioreactor processes performed with animal cells attached to microcarriers. This tool was able to predict glucose and lactate concentrations in situ with good accuracy within the concentration ranges routinely observed in Vero cell culture supernatants. The method was also reliable using various operating conditions and applicable for scale-up. These results were particularly interesting because, they are useful not only to monitor conventional Vero cell culture processes, but this in situ NIRS tool is also useful in controlling unusual concentrations and for the development of risk-mitigation plans for the specific case of adherent cell cultures. Further applications could also be envisioned, such as real-time monitoring of the cell viability by quantification of the lactate dehydrogenase enzyme released during cell death or of the production itself. As a consequence, in situ near-infrared spectroscopy is a very promising tool for current and further applications of the PAT strategy for adherent cell culture processes. Acknowledgements The authors are grateful to Dr Emmanuel Guedon (LSGC-CNRS, Nancy) and to Sven Ansorge (Ecole Polytechnique - CNRC, Montreal) for the help they provided in manuscript writing. References [1] Barrett PN, Mundt W, Kistner O, Howard MK. Vero cell platform in vaccine production: Moving towards cell culture-based viral vaccines. Expert Rev Vaccines 2009;8:607–18. [2] Liu C-C, Lian W-C, Butler M, Wu S-C. High immunogenic enterovirus 71 strain and its production using serum-free microcarrier Vero cell culture. Vaccine 2007;25:19–24. [3] Rourou S, van der Ark A, van der Velden T, Kallel H. A microcarrier cell culture process for propagating rabies virus in Vero cells grown in a stirred bioreactor under fully animal component free conditions. Vaccine 2007;25:3879–89. [4] Toriniwa H, Komiya T. Long-term stability of Vero cell-derived inactivated Japanese encephalitis vaccine prepared using serum-free medium. Vaccine 2008;26:3680–9. [5] Petiot E, Fournier F, Gény C, Pinton H, Marc A. Rapid screening of serum-free media for the growth of adherent Vero cells by using a small-scale and noninvasive tool. Appl Biochem Biotechnol 2010;160:1600–15. [6] Mandenius C-F, Graumann K, Schultz TW, Premstaller A, Olsson I-M, Petiot E, et al. Quality by design (QbD) for biotechnology-related pharmaceuticals. Biotechnol J 2009;4:600–9.
[7] Ansorge S, Esteban G, Schmid G. On-line monitoring of infected Sf-9 insect cell cultures by scanning permittivity measurements and comparison with off-line biovolume measurements. Cytotechnology 2007;55:115–24. [8] Rudolph G, Lindner P, Gierse A, Bluma A, Martinez G, Hitzmann B, et al. On-line monitoring of microcarrier based fibroblast cultivations with in situ microscopy. Biotechnol Bioeng 2008;99:136–45. [9] Arnold SA, Gaensakoo R, Harvey LM, McNeil B. Use of at-line and in-situ nearinfrared spectroscopy to monitor biomass in an industrial fed-batch Escherichia coli process. Biotechnol Bioeng 2002;80:405–13. [10] Scarff M, Arnold SA, Harvey LM, McNeil B. Near infrared spectroscopy for bioprocess monitoring and control: current status and future trends. Crit Rev Biotechnol 2006;26:17–39. [11] Zhang L, Small GW, Arnold MA. Multivariate calibration standardization across instruments for the determination of glucose by Fourier transform nearinfrared spectrometry. Anal Chem 2003;75:5905–15. [12] Tosi S, Rossi M, Tamburini E, Vaccari G, Amaretti A, Matteuzzi D. Assessment of in-line near-infrared spectroscopy for continuous monitoring of fermentation processes. Biotechnol Prog 2003;19:1816–21. [13] Sandnes JM, Ringstad T, Wenner D, Heyerdahl PH, Kallqvist T, Gislerod HR. Real-time monitoring and automatic density control of large-scale microalgal cultures using near infrared (NIR) optical density sensors. J Biotechnol 2006;122:209–15. [14] Vaidyanathan S, Harvey LM, McNeil B. Deconvolution of near-infrared spectral information for monitoring mycelial biomass and other key analytes in a submerged fungal bioprocess. Anal Chim Acta 2001;428:41–59. [15] Riley MR, Rhiel M, Zhou X, Arnold MA, Murhammer DW. Simultaneous measurement of glucose and glutamine in insect cell culture media by near infrared spectroscopy. Biotechnol Bioeng 1997;55:11–5. [16] Arnold SA, Crowley J, Woods N, Harvey LM, McNeil B. In-situ near infrared spectroscopy to monitor key analytes in mammalian cell cultivation. Biotechnol Bioeng 2003;84:13–9. [17] Harthun S, Matischak K, Friedl P. Determination of recombinant protein in animal cell culture supernatant by near-infrared spectroscopy. Anal Biochem 1997;251:73–8. [18] Riley MR, Crider HM, Nite ME, Garcia RA, Woo J, Wegge RM. Simultaneous measurement of 19 components in serum-containing animal cell culture media by Fourier transform near-infrared spectroscopy. Biotechnol Prog 2001;17:376–8. [19] Roychoudhury P, O’Kennedy R, McNeil B, Harvey LM. Multiplexing fibre optic near infrared (NIR) spectroscopy as an emerging technology to monitor industrial bioprocesses. Anal Chim Acta 2007;590:110–7. [20] Riley MR, Arnold MA, Murhammer DW, Walls EL, DelaCruz N. Adaptive calibration scheme for quantification of nutrients and byproducts in insect cell bioreactors by near-infrared spectroscopy. Biotechnol Prog 1998;14:527–33. [21] Riley MR, Okeson CD, Frazier BL. Rapid calibration of near-infrared spectroscopic measurements of mammalian cell cultivations. Biotechnol Prog 1999;15:1133–41. [22] Rhiel M, Cohen MB, Murhammer DW, Arnold MA. Nondestructive near-infrared spectroscopic measurement of multiple analytes in undiluted samples of serum-based cell culture media. Biotechnol Bioeng 2002;77:73–82. [23] Teixeira AP, Oliveira R, Alves PM, Carrondo MJT. Advances in on-line monitoring and control of mammalian cell cultures: supporting the PAT initiative. Biotechnol Adv 2009;27:726–32. [24] Card C, Hunsaker B, Smith T, Hirsch J. Rapid, simultaneous monitoring of multiple components in mammalian cell culture with near-infrared spectroscopy. Thermo Fisher Scientific; 2007 [Application Note: 51449]. [25] Kennedy M, Krouse D. Strategies for improving fermentation medium performance: a review. J Ind Microbiol Biotechnol 1999;23:456–75.