White light spectroscopy for T-cell culture growth monitoring: towards a real-time and sampling free device for ATMPs production

Authors: Naïs Vaccari, Bruno Wacogne, Afi Claudia Koubevi, Marine Belinger-Podevin, Alain Rouleau, Annie Frelet Barrand
14 October 2022https://hal.science/hal-03814565/ 

Abstract

Nowadays, Advanced Therapy Medicinal Products (ATMPs) are new generation of pharmaceuticals which offer novel and revolutionary answer to be given to patients at the end of their therapeutic options. However, because of their prohibitive cost, they can only be proposed to a reduced number of patients. One possibility to reduce ATMP’ price is to automate their production as much as possible. Among the numerous possibilities, a real-time and closed loop monitoring of the T-cell growth during the expansion phase would be appreciated. In this short paper, we propose a simple white light spectroscopy method which can be easily implemented in such an autonomous device and able to monitor T-cell number evolution with a higher accuracy as compared to current cell counting methods.

Introduction

Advanced Therapy Medicinal Products (ATMPs) recently emerged as new treatments for patients facing therapeutic dead ends. Genetic modification or tissue engineering is used to give living cells new physiological, biological characteristics or reconstruction properties.

However, complex technologies of cell sorting, amplification, genetic transduction, amplification-division and activation are required to produce these new drugs. The whole process takes place in a controlled environment and numerous quality controls are performed throughout the production which can extend over one week. Consequently, the price of these promising therapeutic solutions restricts the possibilities to democratize their use for the greatest number of people. Devices developed during the recent years are not optimal due to the lack of online tracking technologies. Only few parameters such as temperature, pH or dissolved O2 are monitored using sterile probes placed inside the bioreactor. The PAT project (Process Analytical Technology) was born from this observation by the FDA in 2004. This project encourages research and development of new analysis technologies allowing a real time monitoring of all production stages of biopharmaceutical drugs.

Concerning ATMPs, the whole production process is quite complex and the above-mentioned quality controls are frequently performed, especially during the expansion phase [2,3]. Multiplying these controls, and therefore samplings, also increases the risk of introducing new contaminations. All this emphasizes the need to develop monitoring solutions which can easily be transposed in a close-loop system in order to provide a real-time cell growth monitoring and quality control.

Historically in cell culture, cell concentration estimation relies on directly measuring cell number, by counting them one by one, because the cell size makes it possible to directly observe them with common microscopes. On the contrary, absorption-based methods like turbidimetry or Beer-Lambert law derived techniques (optical density measurements for example) are usually preferred when considering smaller biological entities such as bacteria. Malassez cells are probably the most well-known technique used despite the fact that the visual and manual counting can be difficult and poorly reproducible due to the very small cell volumes sampled and therefore not representative of the culture flask. Alternative and commercial automated methods are available to facilitate cell counting.

Automatic cell counters have been developed and are commercially available. The LUNATM system (LOGOS BIOSYSTEMS) used in [4-6] requires 10 μL of cell solution. It is based on conventional imaging and image processing is used to count the imaged cells. Other systems developed by IPRASENSE are based on lensless imaging [7] in which diffraction figures of cells are recorded and analyzed to assess cell concentration. NORMA is a cell counter using 10μL whereas CYTONOTE is preferred to be used with adherent cells on larger volumes. They can reach higher accuracy thanks to the large investigating area (29.4 mm2) that lensless imaging technique allows.

In addition, INCUCYTE. (SARTORIUS), used with both adherent and non-adherent cells [8,9], is an in-situ microscopy system based on holographic imaging designed to be used in an incubator (this is also the case for the HoloMonitor. system from PHI [10]). Here, cell counting can directly be performed within different flasks including 96 wells plates for high throughputs and/or multiple simultaneous experiments. Despite their easy use, these commercial systems seem difficult to integrate in a close-loop and real-time environment.

Other methods can also be used for both cells and subcellular entities qualification. Some of them are based on the capture of the biological entity at the surface of a biosensor in a ligand-analyte reaction. This is the case with the ELISA technique [11,12], SPR methods [13,14] and also Quartz Crystal Microbalances [15,16]. However, for all these methods, the need for a biological interface makes their transposition to a real-time measurement system difficult. Indeed, sensors’ surfaces must be regularly regenerated which prohibits their use in such a configuration. There exist other methods which can potentially be used without the need for a bio-chemical interface. Impedance spectroscopy (or dielectric spectroscopy) has been widely used during cell culture processes, particularly in the monitoring of mammalian cells [17]. This technique makes it possible to know the cell concentration thanks to their polarization following the application of an alternating electric field. It presents several significant advantages such as an in situ analysis of the cell culture and rapid measurements. However, this method suffers from some drawbacks such as the need for calibration and a decrease of the accuracy during the stationary phase of growth

[18]. Raman spectroscopy could be more suitable and has already been demonstrated for agronomy or biology purposes [19], in particular during quality controls carried out on cell culture [20] and for pathogen detection [21]. However, Raman characteristics of cultured cells may not be required for cell concentration monitoring purpose. Flow cytometry can also be envisaged for cell counting [22,23] and cell activation detection [24]. Depending on the optical detection scheme employed together with the flow cytometer, counting cells and assessing some of their biological properties for quality control could be performed simultaneously. Adaptation in a close-loop system is straightforward if a derivation is added to the bioreactor as proposed in [2,3].

Indeed, cell counting methods described above all imply considering cells one by one in order to assess culture concentrations and most require sampling of small volumes poorly representative of what really occurs in the bioreactor. But, as long as only non-adherent cells are concerned, methods based on optical absorption measurement can potentially be performed on large sample volume [2,25,26]. The only condition is that the measurement does not require sampling which is possible using either derivation or sterilized optical probes. Using such methods means considering the global “light-culture” interaction rather than estimating concentration counting individual particles. In [2,3], we proposed proofs of concept using white light spectroscopy to measure B-cell concentration and possibly detect contaminations in real-time (the latter was not yet experimentally demonstrated).

Detecting contamination is part of quality controls regularly performed during ATMP production. Another aspect of quality control concerns cell viability which can be optically estimated using statistical methods applied to absorbance spectra measurements in the UV-VIS range [27].

In this paper, we focus on cell growth monitoring and we present a simple white light spectroscopy system to monitor the evolution of T-cell concentrations over 1 week. 8 independent experiments of daily concentration measurements were performed using either the LUNATM automatic cell counter or white spectroscopy. On the last day, dilution ranges were prepared and used for determining an optical based absorption model. Accuracy and reproducibility of the spectroscopy optical method was then compared to that obtained using the automatic cell counter. This paper is organized as follows. The next section of the paper presents the material and methods used in our experiments from the biology and optics point of view. Experimental results will then be presented together with the description of the simple optical absorption model we used to estimate cell concentrations. After a section devoted to discussing our results, a conclusion and future work will be proposed.

Materials and methods

CEM preparation: CEM cells (ATCC. CRL-2265TM) are T lymphoblasts supplied by the French Blood Agency (EFS Etablissement Fran.ais du Sang). They were grown in RPMI-1640 medium (P04-16515, PAN-Biotech.) supplemented with 25 mM HEPES (P05-01500, PAN Biotech.), 10% heat inactivated FBS (10270 -106, FischerScientific.) and 1% penicillin (10 kU.mL-1)/streptomycin (10 mg.mL-1)(FG101-01, TransGen Biotech.). The cells were maintained at 37ÅãC in a humidified atmosphere containing 5% CO2.

Monitoring experiment protocol: Eight cell culture monitoring experiments were conducted. They last for a week according to the protocol presented in Figure 1. At the end of the 5 first experiments, dilution ranges were prepared for establishing the optical spectroscopic absorption model. CEM cells were seeded on Day 1 and then resuspended in a T75 flask at a concentration of 5Å~105 cells.mL-1 in 15 mL of supplemented RPMI (RSM). 3 mL of cell suspension were taken from the flask and transferred to sterile plastic cuvettes (C0793-100EA, Merck.) for spectral measurements; 3 cuvettes were prepared in order to have triplicates. Cell counting was carried out in the LUNA-II Automated Cell Counter (Logos Biosystems.) with trypan blue (V/V) (15250061, Fischer Scientific.) with 10 μL of cell suspension contained in the flask (3 independent measurements) as well as cell suspensions contained in each of the 3 cuvettes in order to measure the average concentration and determine a standard deviation. Spectral measurements were subsequently carried out on the 3 cuvettes. On Day 1, spectral measurements were performed on solutions directly collected from the flask. On days 2 to 5, sampled solution was diluted 3 times so that the absorption of the cuvette remains in the measurable spectroscopy range (between 20 and 80% absorption). After measurements, the volumes of cell suspensions contained in the cuvettes were returned to the original flask and then kept in the incubator. 24 hours later, 1 mL of cell suspension present in the flask was introduced into a cuvette (3 times in order to obtain triplicates). 2 mL of RSM medium was added to each cuvette in order to lower cell concentration making the absorption spectra measurable.

Cell counting as well as spectral measurements were carried out following the same protocol as the day before. This operation was repeated every 24 hours until Day 5.

Concentration ranges for optical absorption modeling: Different concentrations were prepared by diluting cuvettes in RPMI medium in order to obtain the following concentrations: 106, 9Å~105, 8Å~105, 7Å~105, 6Å~105, 5Å~105, 4Å~105, 3Å~105, 2Å~105 and 106 cells.mL-1. The goal was to obtain data for estimating the mathematical absorption properties of the cells at various concentrations. The cell suspension from the “monitoring cells” was recovered on Day 5 at the end of the experiments and used for this purpose. 3 cell counts and one spectral measurement were performed with each cuvette for mathematical modeling purposes.

Spectroscopic absorption measurements: Spectral absorption measurements of CEM suspensions were performed using the experimental set-up shown in figure 2. The spectroscopy measuring system consists of a light source (AvaLight-DH-S-BAL, Avantes.), connected by optical fibers (Thorlabs M25L01) to a cuvette holder (Avantes CUV-UV/VIS). The white light source was switched on about 30 min before measurements to allow temperature and spectral characteristics stabilization. After propagation through the cuvette, the light was transmitted to the spectrophotometer (Ocean Optics USB 4000 UV-VIS-ES) for spectra acquisition. Before each measurement, a reference spectrum was acquired using a cuvette containing RPMI medium. Spectra were recorded in transmission, in the wavelength range 177 nm and 892 nm with a step of 0.22 nm using the OceanView software.

Spectral data processing: The spectral data were recorded into a text file then transposed to Excel. The data obtained in transmission were converted into absorption percentage and all calculation were performed using MatlabTM R2020b software. Only wavelengths between 330 nm and 860 nm were considered in order to eliminate measurements with a high background noise. Artefacts due to energetic emission peaks of the deuterium lamp were numerically removed.

Regularly, absorption spectra of neutral densities (THORLABS NE05B and NE10B) were recorded and compared to the supplier’s data to ensure correct absorption spectra measurements. Spectra maxima were determined after spectra were slightly smoothed using cubic spline MATLABTM functions. Other spectroscopy signal treatments were performed on raw spectra

Experimental results

This section illustrates the experimental results obtained during the 8 experiments. Concentration ranges were prepared during the 5 first weeks; monitoring experiments were conducted during 8 weeks. For simplicity purpose, only results concerning week 15 are reported here because they are representative of what the results obtained every week. Data obtained during all weeks are available on demand. Cell culture monitoring using the automatic cell counter: As previously mentioned, T-cells were grown during a week and daily concentration measurements were performed. Concentrations were first measured using 10 μL sampled directly from the culture flask. Sampling and measurements were repeated 3 times. Second, concentrations were measured using 10 μL sampled from 3 spectroscopy cuvettes (Figure 1 and Figure 2). The results obtained during week 15 are shown in figure 3a where black circles and blue crosses correspond to data related to the flask and the cuvettes respectively. It can clearly be observed that data exhibit dispersion increasing with time, i.e. with the number of cells. Moreover, data dispersion is larger when sampling from the cuvette than from the flask. This is clearly visible on figure 3b where standard deviations (SDs) are reported as Box and Whiskers plots. To obtain the SD relative to the flask, the SD on day 1 was measured and expressed in terms of percentage, therefore generating 5 SD values for the flask at the end of the week. They were used to draw the “flask” Box and Whiskers plot. The same calculation was performed for the cuvette. On each box, the central red mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to the most extreme data points. The mean SD and dispersion are smaller for the flask than for the cuvette (about 8% against 20%; figure 3b). In order to compare data obtained with the automatic cell counter to those obtained using white light spectroscopy, it is necessary to establish the relation between cell concentration and spectroscopic properties of the cell suspensions. Concentration range and spectroscopic behavior: The goal was to measure the absorption spectra of cell solutions at different concentrations and to express the cell concentration as a function of the spectral behavior. In what follows, we refer to it as “optical concentration model (OCM)”. Overall 5 concentration ranges were used to establish the OCM. An example of such absorption spectra is shown on figure 4a. Data indicated in the legend correspond to mean concentrations measured with the cell counter. Black stars correspond to the maxima positions and values of each individual spectrum. Values of the maxima and their positions were computed from slightly smoothed spectra although raw spectra are drawn on figure 4a. It should be noted that the positions of the maxima evolve with the maxima values. This was observed in all experiment as shown on figure 4b. Here, data are fitted with a linear regression with R2=0.61. This means that this evolution is not necessarily linear. Black stars reported on figure 4a highlight spectra maxima values used to establish the OCM regardless of their position. This slightly differs from calculating the optical density at a fixed wavelength and correlating it to the concentration (from Beer-Lambert law) or to correlate the concentration to the solution transmission (turbidity). The OCM is established as follows. The Beer-Lambert law relates the optical density to the concentration of a species in solution by means of a simple linear relationship:

Conclusion

In this paper, an easy-to-implement white light spectroscopy method to monitor cell cultures was presented. Examples concerning T-cell culture concentration measurements showed accuracy and reproducibility higher than using automatic cell counters, mainly because measurements are performed on large volumes. For this, a white light spectroscopic model of the culture absorption has been developed. It relies on measuring cell optical absorption spectra maxima. Cell culture monitoring performed during 8 experiments highlighted an exponential increase of cell number and dividing times correlated to those already published and possibly further reduced by agitating cell culture. This can be automatically done if a derivation of the bioreactor is considered. Because white light spectroscopy is performed in the visible range, the use of compact and low-cost components can be used to propose a highly integrated method to monitor cell cultures.

This spectroscopic monitoring method can be easily transposed to a real-time measurement and closed loop system which avoid regularly sampling of bioreactor content to measure cell concentration and to check the existence of possible contaminations. Current developments deal with the fabrication of a demonstrator and the use of information from the white light absorption spectra in order to establish a real time quality control. Indeed, monitoring and quality controls could be performed simultaneously since both rely on the only acquisition of the absorption spectra. This would allow a real time detection of ATMP production dysfunction and could be used to stop an useless production early, hence saving a considerable amount of money. Having such a system would be of great interest in terms of research, industrial manufacturing and more importantly in terms of benefit to patients.