Capillary and microchip electrophoresis method development for amino acid monitoring during biopharmaceutical cultivation

Authors : Saara Mikkonen, Leila Josefsson, Meeri E.-L. Mäkinen, Veronique Chotteau, Åsa Emmer
23 March 2022


The increased use of biopharmaceuticals calls for improved means of bioprocess monitoring. In this work, capillary electrophoresis (CE) and microchip electrophoresis (MCE) methods were developed and applied for the analysis of amino acids (AAs) in cell culture supernatant. In samples from different days of a Chinese hamster ovary cell cultivation process, all 19 proteinogenic AAs containing primary amine groups could be detected using CE, and 17 out of 19 AAs using MCE. The relative concentration changes in different samples agreed well with those measured by high-performance liquid chromatography (HPLC). Compared to the more commonly employed HPLC analysis, the CE and MCE methods resulted in faster analysis, while significantly lowering both the sample and reagent consumption, and the cost per analysis.


The biopharmaceuticals market is extremely fast-growing, and monoclonal antibody drugs produced by Chinese hamster ovary (CHO) cells are increasingly used to treat diseases such as cancers, autoimmune, neurodegenerative, and cardiovascular diseases. There is a need for improved means of bioprocess monitoring in terms of speed, throughput, and amount of gained information. High throughput real-time quantification of important compounds or quality attributes can reduce the process development in terms of time and workload, while assuring that a high quality of the product of interest (POI) is maintained. Likewise, manufacturing can highly benefit from this at-line detection as driven by process analytic technology (PAT) initiatives. The culture medium is a complex mixture of nutrients and additives to create optimal conditions for the cells, and components produced or released by the cells, for example, metabolites, POI, host cell proteins, and nucleic acid. Monitoring the concentrations of free amino acids (AAs) is important to obtain information about the culture’s state, and prevent depletion or unfavorable excess. It is often performed using chromatography-based methods (e.g., high-performance liquid chromatography [HPLC]) combined with optical or mass spectrometry (MS) detection. As an alternative to chromatography, AAs can be analyzed using electrophoretic methods, currently less utilized than liquid chromatography for historical reasons.

Electrophoresis exploits differences in charge and size to separate analytes in solution under the influence of an electric field. Electrophoretic separations in microfluidic scale are often performed in narrow capillary tubes, that is, capillary electrophoresis (CE), or in microfluidic devices using microchip electrophoresis (MCE). Since CE and MCE can achieve fast, high-resolution separations consuming only nanoliters of sample, they are promising tools for bioprocess monitoring. Furthermore, the MCE format allows easy multiplication of these detection methods at low cost and small footprint. It is well amenable to automation, which can greatly support high-throughput screening. Its simplicity is also attractive for validation in GMP environment. Alhusban et al. combined a microfluidic H-filter with sequential injection CE and capacitively coupled contactless conductivity detection to monitor the concentrations of glucose, glutamine (Gln), leucine (Leu)/isoleucine (Ile), and lactate during a 100-h period. Gilliland and Ramsey developed an MCE high pressure MS platform for cell growth monitoring, and applied it to quantify four selected AAs in lysogeny broth during cultivation of bacterial cells. After filtration and dilution as sample preparation, the AAs were separated in less than 3 min, and their concentrations were monitored every 30 min for 24 h. While MS is a powerful analytical technique, it generally requires large and expensive instrumentation that can be difficult to integrate in micrototal analysis systems. In comparison, optical detection is typically performed on-capillary/chip. Makeeva et al. analyzed 18 AAs and lactic acid during Lactobacillus helveticus cultivation using CE-UV. By exploiting complexation with Cu2+, direct UV-detection of native AAs was achieved. More commonly, direct optical detection of AAs is realized by chemical derivatization with a UV absorbing or fluorescent molecule. Since derivatization increases the similarity of the AAs, the separation mode micellar electrokinetic chromatography (MEKC), providing an additional separation mechanism through the partitioning into micelles, is particularly powerful for their analysis. MEKC with laser induced fluorescence (LIF) detection was employed to analyze eight AA metabolites for Corynebacterium glutamicum after microwave-assisted fluorescein isothiocyanate derivatization. Celá et al. developed MEKC methods employing CE with fluorescence detection to analyze AAs in culture media for human embryos. In samples of spent medium, 21 naphthalene-2,3-dicarboxyaldehyde (NDA) derivatized AAs were separated in a 45 cm effective length capillary using CE with light-emitting diode-induced fluorescence detection. Compared to CE, MCE can provide faster analysis and enable integration of sample pretreatment procedures in a micrototal analysis system format. For example, 10 AAs from murine and human islets of Langerhans were on-chip derivatized and analyzed using MCE combined with fluorescence detection. In another contribution, five AAs in soil samples relevant to planetary exploration were on-line derivatized, separated, and detected by LIF using a portable and fully automated MCE analyzer.

This work is part of the EU project iConsensus – integrated control and sensing platform for biopharmaceutical cultivation process high-throughput development and production – which has the overall aim of developing an at-line integrated analytical system (ALIAS) for bioprocess monitoring, including MCE methods for the analysis of AAs, monosaccharides, and vitamins ( Herein, we present CE and MCE methods for the analysis AAs, and their successful application on samples of cell culture media. We show that in comparison to traditional chromatographic methods, AA analysis by CE and MCE can increase the throughput while significantly lowering the cost and consumption of chemicals per analysis.


2.1 Cell line and media

Samples of FMX-8 culture medium (kit for 10L, Cell Culture Technologies) were reconstituted according to manufacturer instructions. AA free cell culture medium (M_AA_free) was prepared from individual components (see Table S1 in the Supporting Information).

The monoclonal antibody (Rituximab) producing CHO TurboCell cell line (kindly provided by Rentschler Biopharma, Laupheim, Germany) was used in the experiments. Chemically defined, serum-, AA-, and glucose-free basic medium mixture (IC-0), kindly provided by FUJIFILM Irvine Scientific (Santa Ana, CA, USA), was supplemented with proprietary mixtures of AAs to obtain IC-1 and IC-2. The final media used in cell cultivations (IC-1 and IC-2) had therefore different concentrations of AAs. During their expansion, the cells were kept at 37°C and 5% CO2 in Minitron Incubators (Infors HT, Bottmingen-Basel, Switzerland). The cells were expanded in IC-1 medium and passaged every 3–4 days to maintain exponential growth phase.

2.2 Cultivation

A 4-L stirred tank bioreactor (Belach Bioteknik, Stockholm, Sweden) with 3.4 L final working volume was used to perform a fed-batch cultivation. The bioreactor was inoculated with centrifuged pooled cell broth from shake flask expansion (0.5 × 106 viable cells per mL, MVC mL−1) in chemically defined proprietary medium (IC-1). The temperature was kept at 37.0°C ±0.2. Air, CO2, and O2 were mixed online to maintain dissolved oxygen at 50% oxygen saturation. pH was controlled at setpoint 7.0 by automatic addition of bicarbonate solution (0.5 m Na2CO3, Fisher Scientific, Loughborough, Leicestershire, UK) or CO2 gassing. Data acquisition and process control were performed by Biophantom software (Belach Bioteknik) using Wonderware (Aveva, Cambridge, UK). Chemically defined proprietary feed medium (IC-2) was daily added starting from day 1, in bolus as 9% of final working volume until day 6, after which as 1.5% per day until day 8. Glucose and Gln were added daily based on calculated consumption to maintain the target levels of 6 and 0.5 mm, respectively. The stirring speed was kept at 100 rpm until day 7 when it was increased to 150 rpm. Samples were harvested once per day and the cell count, viability and cell size were analyzed with Norma XS (Iprasense, Clapiers, France).

2.3 Electrophoresis

2.3.1 Amino acid derivatization

Alanine (Ala), arginine (Arg), asparagine (Asn), aspartic acid (Asp), glutamic acid (Glu), glycine (Gly), histidine (His), isoleucine (Ile), leucine (Leu), lysine (Lys), methionine (Met), phenylalanine (Phe), serine (Ser), threonine (Thr), tryptophan (Trp), tyrosine (Tyr), valine (Val) from Sigma–Aldrich (St Louis, MO, USA) and Cysteine (Cys) and Gln from Merck (Darmstadt, Germany) were used as AA standards. The AAs were derivatized with NDA (Molecular Probes, Leiden, The Netherlands), yielding fluorescent N-substituted 1-cyanobenz[f]isoindoles. The derivatization was performed in 5 mm sodium tetraborate (Merck) at room temperature in a total volume of 100 μL, including 20 μL sample (AA standards, FMX-8 medium or spent medium). 0.1 mm α-aminobutyric acid (AABA, Sigma–Aldrich) was included as internal standard (IS). NDA was dissolved in acetonitrile (Sigma–Aldrich) and added as last component to the reaction vial. For the samples analyzed using CE, NDA, and sodium cyanide (NaCN, Merck) concentrations of 2 mm were used, and for the samples analyzed using MCE both concentrations were increased to 5 mm. The derivatization reaction was complete within 10 min.

2.3.2 Capillary electrophoresis

An Agilent 7100 CE system (Agilent Technologies, Waldbronn, Germany) with a built-in UV detector and equipped with a 24.5/33 cm effective/total length capillary (Polymicro Technologies, Phoenix, AZ, USA, 375/50 μm outer diameter/inner diameter) was used. The capillary was rinsed with 1 m NaOH (Merck), 0.1 m NaOH, water, and background electrolyte (BGE) for 20 min each before its first use. Each new analysis day, the capillary was rinsed with 0.1 m NaOH and water for 10 min each, and BGE for 15 min. In between runs, 2 min flushing with 0.1 m NaOH followed by BGE for 3 min was employed. The used BGEs were prepared from sodium tetraborate and sodium dodecyl sulfate (SDS, Sigma–Aldrich), with or without 1-butanol (Merck) as additive. CE runs were performed at 25°C and 12.5 kV (electric field 379 V cm−1, current ≈50 μA), and samples were injected hydrodynamically (50 mbar, 3 s). For CE runs with the sample injected from the back end of the capillary, the effective length was reduced to 8.5 cm. Sample injection was performed by immersing the outlet end of the capillary into the sample vial and applying −50 mbar for 3 s, and the separation voltage was set to −12.5 kV. UV detection was performed at 254 nm. Data treatment was performed using OpenLAB CDS ChemStation (Agilent).

2.3.3 Microchip electrophoresis

The MCE system (Figure S1, Supporting Information) included commercial glass chips (T8050, Micronit, Enschede, The Netherlands) with a separation length of 7.9 cm, and a channel depth of 20 μm and width of 50 μm (total separation channel volume was 0.07 μL), a microchip holder, a four-channel output high-voltage sequencer (eDAQ, Denistone East, Australia), and a fluorescence detection module (RD5900, Micronit), which utilized a laser diode at excitation wavelength 458 nm with an emission filter at wavelength 500 ± 20 nm. Electropherograms were recorded using HP Chemstation Rev A.06.03[509] (Hewlett-Packard, Palo Alto, CA, USA) utilizing a manual trigger.

Before each use, the microchips were manually flushed with 1 m NaOH, 0.1 m NaOH, and twice with water using approximately 3.5 μL of solution each time. The separation channel was filled with BGE before the microchip was mounted in the microchip holder. The position of the microchip was visualized using a USB camera (DCC1545M, Thorlabs, Newton, NJ, USA), and corrected using a manual multi-axis stage (Thorlabs). Detection occurred 2.5 mm from the outlet. Fifty microliters of BGE was added to all wells and the microchip was electrokinetically flushed (Table S2, Supporting Information). Thereafter, BGE solutions were exchanged to fresh ones, and 50 μL of sample was added to the sample well by aspiring the solution three times using a micropipette (Eppendorf, Hamburg, Germany) before injection. After each analysis, the electrokinetic flushing step was repeated, and after each set of analysis, the microchip was manually flushed as described above.

2.4 HPLC

An external standard of AAs was prepared from a commercial stock standard solution of 17 AAs (Waters, Milford, MA, USA), which was mixed (1:1) with 2.5 mm AABA stock standard. A solution with 0.83 mm Asn and 0.83 mm Trp was prepared and mixed with the commercial standard and AABA (2:3). Three replicates were prepared from each harvest cell culture sample collected on cultivation days 1, 4, and 8. Prior to derivatization, the samples were treated with trichloroacetic acid (Merck) to precipitate protein. AAs were derivatized with AccQ-Fluor Reagent kit (Waters). Derivatized AAs were separated with reversed-phase HPLC. Sample injection and liquid handling was performed with Waters Alliance 2695 system. The separation was performed using a Waters AccQ-Tag Column (60 Å, 4 μm, 3.9 mm × 150 mm), and the detection with a 2996 PDA detector at 248 nm (all from Waters). Three eluents were used during the HPLC run. 13.61 g L−1 sodium acetate trihydrate (Merck Millipore, Burlington, MA, USA), 0.57 g L−1 triethylamine (Sigma–Aldrich), and 0.10 g L−1 sodium azide (Sigma–Aldrich) were dissolved into water. The pH was adjusted with phosphoric acid (Merck Millipore) to 5.50 (eluent I) and 6.80 (eluent II). Acetonitrile (99.8%≥, Fisher Scientific) was used as eluent III.


3.1 Capillary electrophoresis

NDA can be used to derivatize primary amines, that is, all standard AAs except proline (Pro), in the presence of the cyanide anion. NDA itself is not fluorescent but absorbs UV. Thus, while optimizing the separation, derivatizations were performed with a molar excess of AAs in relation to NDA to avoid interfering peaks of excess NDA and its degradation products. Based on the literature and initial experiments, an MEKC system comprising borate buffer and SDS was selected to separate NDA derivatized AAs. The shortest possible capillary length for the CE instrument was selected, resulting in an effective separation length of 24.5 cm. In a BGE of 25 mm sodium tetraborate (pH 9.3) and 50 mm SDS, herein referred to as method A, 13 AAs out of 19 (excluding the IS) could be separated, as shown in Figure 1A. In this system, comigration of the late migrating, less polar, AAs – Leu, Phe, and Trp, and Arg, Cys, and Lys, respectively, was, however, observed. To improve the resolution between these AAs, solvent modified MEKC was employed. In Figure 1B, electropherograms of the AA separation in a BGE of 96% (25 mm sodium tetraborate, 50 mm SDS) and 4% 1-butanol, herein referred to as method B, are shown. The six AAs that were not separated using method A were resolved using method B.

Capillary and microchip electrophoresis method development for amino acid monitoring during biopharmaceutical cultivation
CE-UV of NDA derivatized AAs at 0.1 mm in a 24.5 cm capillary. (A) Method A, BGE: 25 mm sodium tetraborate, 50 mm SDS, top: AA standards, bottom: FMX-8 cell culture medium. (B) Method B, BGE: 96% (25 mm sodium tetraborate, 50 mm SDS), 4% 1-butanol, top: AA standards, bottom: FMX-8 cell culture medium. 1: Ser, 2: Asn, 3: Thr, 4: Gln, 5: His, 6: Gly, 7: Glu, 8: Ala, 9: Asp, 10: Tyr, 11: AABA (IS), 12: Val, 13: Met, 14: Ile, 15: Leu, 16: Phe, 17: Trp, 18: Arg, 19: Lys, 20: Cys. AA, amino acid; AABA, α-aminobutyric acid; BGE, background electrolyte; CE-UV, capillary electrophoresis-ultraviolet; NDA, naphthalene-2,3-dicarboxyaldehyde; SDS, sodium dodecyl sulfate

During method development, the cell culture medium FMX-8 was used as model sample. FMX-8 has a known composition, and contains AAs in a concentration range of 0.09–0.9 mm, which was diluted five-fold during the derivatization reaction, to 18–180 μm. Combining the results from methods A and B, all standard AAs except Cys were detected in FMX-8 (bottom electropherograms of Figure 1 A and B). In the alkaline sample and BGE conditions, Cys is present in its oxidized dimer form cystine, which contains two primary amine groups. Thus, the derivatization of Cys likely results in the formation of multiple products, complicating the detection at low concentrations. In addition to AAs, the culture medium includes inorganic salts, vitamins, fatty acids, glucose, and an antifoaming agent. To assure the detected peaks corresponded to AAs, a culture medium of the same composition but without AAs was also prepared and analyzed, and no overlaps with the AA peaks were detected (data not shown). However, a presumed degradation product of NDA was found to have a similar migration time as Ala. The potential interference caused by this will be avoided in the ALIAS by using fluorescence detection instead of UV detection, since NDA does not fluoresce.

The average RSD of migration times was calculated, based on a series of three replicates repeated for two sets of BGE performed at three different days using a sample of AA standards. The obtained RSD values were: intraday 1 – 0.49%, intraday 2 – 0.31%, and intraday 3 – 0.68% (n = 6 each day) for method A, and intraday 1 – 2.0%, intraday 2 – 0.51%, and intraday 3 – 1.2% for method B. RSD values when all 3 days were included (interday) were 8.5% (range 6.9%–10%) for method A, and 15.1% (range 14.7%–15.9%) for method B (n = 18). Accordingly, effective mobilities were 1.49% (range 0.34%–2.5%) for method A, and 2.03% (range 1.3%–2.55%) for method B. RSD values of migration times, effective mobilities, and peak areas for the individual AAs are included in Table S3 in the Supporting Information. The limit-of-detection (LOD) values were in the range of 5–10 μm using method A, and 5–20 μm for method B, which is at or below the lower range of interest for quantitative detection.

3.2 Microchip electrophoresis

As a step toward transferring the CE methods to the microchip, injection of the sample from the capillary outlet, resulting in an effective separation length of 8.5 cm, well-comparable to the 7.9 cm in the microchip, was performed. The resulting electropherograms for AA standards and cell culture medium, respectively, are shown in Figure S2 in the Supporting Information. Combining the results of the two methods allowed complete or partial separation of 17 out of 19 AAs when analyzing AA standards (resolution between Asn and Thr was lost due to the short separation length), and 16 of 19 AAs for the FMX-8 culture medium samples (Cys was not detected).

Transferring the CE methods to the microchip revealed similar separation of the AAs. Figure 2 shows the electropherograms for the standard AA solution and the FMX-8 cell culture medium using both separation methods A and B. As for the short end-CE results (Figure S2, Supporting Information), using method A, Asn and Thr comigrated. Also, at higher AA concentrations comigration of Ala and Asp was observed, and Glu was found to migrate before Gly (compared to after Gly in CE). The latter could possibly be explained by the different surface properties of the fused silica in CE and the borosilicate glass in MCE. A difference in migration order was also observed for method B, where Asp comigrated with Gly in MCE (compared to after Ala in CE). In method B, detection of Lys was complicated by peak dispersion, and low sensitivity, probably due to the formation of both mono- and di-derivatized compounds in the derivatization reaction. The di-derivatized Lys has a low quantum efficiency and was not detected. Thus, the MCE analyses were terminated after the detection of Arg.

Capillary and microchip electrophoresis method development for amino acid monitoring during biopharmaceutical cultivation
MCE-fluorescence (flu) of NDA derivatized AAs in a 7.9 cm chip. (A) Method A (BGE: 25 mm sodium tetraborate, 50 mm SDS), sample: AA standards at 20 μM. (B) Method B (BGE: 96% [25 mm sodium tetraborate, 50 mm SDS], 4% 1-butanol), sample: AA standards at 20 μm. (C) Method A, sample: FMX-8 cell culture medium. (D) Method B, sample: FMX-8 cell culture medium. 1: Ser, 2: Asn, 3: Thr, 4: Gln, 5: His, 6: Gly, 7: Glu, 8: Ala, 9: Asp, 10: Tyr, 11: AABA (IS), 12: Val, 13: Met, 14: Ile, 15: Leu, 16: Phe, 17: Trp, 18: Arg, 19: Lys, 20: Cys. AA, amino acid; AABA, α-aminobutyric acid; BGE, background electrolyte; IS, internal standard; MCE, microchip electrophoresis; NDA, naphthalene-2,3-dicarboxyaldehyde; SDS, sodium dodecyl sulfate

Some of the AAs that were comigrating in method A could, however, be resolved using method B and vice versa. Combining the results from methods A and B, 18 AAs (i.e., all except Lys) in the standard solution and 17 AAs in FMX-8 could be analyzed using MCE (as for the CE runs, Cys was not detected). To ensure full derivatization, the derivatization of the samples analyzed with MCE was performed with a higher amount of NDA and NaCN than for the samples analyzed by CE, since no interference of the NDA degradation product was visible with fluorescence detection. As shown in Figure S3 in the Supporting Information, analyzing the M_AA_free using both methods did not result in any interfering peaks. The LOD for the AAs using the fluorescence module was 0.2–0.5 μm, which is well below the lower range of interest for quantitative detection of AAs in cell culture medium.

The variability in migration times and peak areas on-chip is often larger than in CE, but the variability can be reduced by using preconditioning of the microchips. Here, a 10 min electrokinetic flushing of the separation channel with BGE was utilized as preconditioning. The repeatability of the migration times for three consecutive runs for the individual AAs in a standard solution and for the FMX-8 solution are included in Table S4 in the Supporting Information. The average RSD of migration times for standards and FMX-8, respectively, were 2.3% and 0.4% for method A, and 1.5% and 1.1% for method B, which is below the requested precision from iConsensus. Also, the implementation of the MCE in the ALIAS system will eliminate all operator caused errors, thus further improve the repeatability of the methods.

3.3 Cell culture supernatant samples

To demonstrate applicability to real samples containing by-products of cell metabolism, in addition to the medium components, including the AAs, cell culture supernatant from a fed batch process in a stirred tank bioreactor was analyzed with methods A and B in the CE capillary and on the microchip. Samples were chosen from different time points along the 8-day cultivation, namely days 1, 4, and 8. The cell density and the cell viability of the process are shown in Figure 3.

Capillary and microchip electrophoresis method development for amino acid monitoring during biopharmaceutical cultivation
Viable cell density (solid line, left y-axis) and cell viability (dashed line, right y-axis) during the fed batch process in a stirred tank bioreactor

CE results for the cell culture supernatant samples from day 4 are shown in Figure 4A and B, for methods A and B, respectively. The electropherograms show few interfering peaks, despite the complex sample composition and the use of UV instead of fluorescence detection. For method A, a non-AA peak next to His, and in method B, one or several peaks between Arg and Lys, were seen. The electropherograms obtained using MCE (Figure 4C and D) showed similar separation patterns as the CE electropherograms. For MCE, the peak areas for Asn, Thr, and Asp were calculated by combining the results for the two methods. An HPLC chromatogram for the same sample is shown in Figure S4 in the Supporting Information. With the employed HPLC method, 20 AAs could be analyzed within 55 min. In some chromatograms, Ala and Arg were, however, comigrating and an overlap between the His peak and NH3 was also observed.

Capillary and microchip electrophoresis method development for amino acid monitoring during biopharmaceutical cultivation
Electropherograms of the cell culture supernatant sample from day 4. (A) CE, BGE of method A, (B) CE, BGE of method B, (C) MCE, BGE of method A, (D) MCE, BGE of method B. 1: Ser, 2: Asn, 3: Thr, 4: Gln, 5: His, 6: Gly, 7: Glu, 8: Ala, 9: Asp, 10: Tyr, 11: AABA (IS), 12: Val, 13: Met, 14: Ile, 15: Leu, 16: Phe, 17: Trp, 18: Arg, 19: Lys, 20: Cys. BGE, background electrolyte; CE, capillary electrophoresis; MCE, microchip electrophoresis

The AA concentration changes, measured as the peak area ratios between the individual AAs and the IS, during the cultivation observed using CE and MCE were compared to those determined with HPLC. Figure 5 depicts the peak area ratio changes for HPLC (n = 3 for days 1 and 8, n = 2 for day 4), CE (n = 3 for each day), and MCE (n = 3 for each day). For the chip analyses, the IS was only baseline resolved in method B, and the IS peak areas from method B were thus used as external standard for the AAs analyzed using method A.

Capillary and microchip electrophoresis method development for amino acid monitoring during biopharmaceutical cultivation
Concentration changes of AAs, expressed as peak area ratios, in cell culture supernatant sampled, 1 (red), 4 (gray), and 8 days (yellow) into the cultivation process. For CE and MCE, amethod A, bmethod B, and calculated AA. AA, amino acid; CE, capillary electrophoresis; MCE, microchip electrophoresis

The changes in peak area ratio for days 1, 4, and 8 for each AA determined by the different methods were compared. For the MCE data Lys was not included. Comparing the three techniques, the resulting average Pearson correlation coefficients, that is, R values, were 0.87 (CE vs. HPLC), 0.88 (CE vs. MCE), and 0.73 (HPLC vs. MCE). These datasets, however, contained some clear outliers tested with the Z-score; Val and Cys showed Z-scores higher than ±3, which for Cys could be related to the formation of multiple products in the derivatization reaction and the poor detectability using CE/MCE.[22] Excluding the two AAs from the datasets resulted in R-values of 0.96 (CE vs. HPLC), 0.93 (CE vs. MCE), and 0.90 (HPLC vs. MCE). The correlation coefficients for each AA are shown in Table S5 in the Supporting Information.


Whereas HPLC is an established technique for PAT, CE has been less used. During the CE method development, several matters were taken into consideration to create a robust method. The BGE compositions were kept as simple as possible, and only chemicals that were compatible with the polymer materials of the microchip holder were used. Care was also taken to keep the generated current at ≈50 μA, which is within the linear range of Ohm’s law for the methods (data not shown), to limit the formation of irreproducible temperature gradients in the separation channel. Performing a major part of the method development using CE and not MCE can save development time and effort, and we have shown that results obtained from a commercial CE setup can be readily transferred to the employed glass microchips.

Separation of all NDA derivatized proteinogenic AAs using MEKC often requires rather long separation lengths and has, to the best of our knowledge, not been performed on-chip. Developing two methods, one for the more polar and one for the less polar AAs, and combining the results, however, allowed detecting most AAs of interest, even on-chip. Despite the use of two methods, the throughput of samples could be increased compared to the current HPLC based procedure. The sample preparation protocol before the CE/MCE analyses is fast and simple, requiring only filtration or centrifugation to collect the supernatant, and an ≈10 min one-step derivatization at room temperature. Including sample preparation, the total analysis time per sample is ≈50 min (≈35 min separation time) for CE and ≈35 min (≈15 min separation time) for MCE. In comparison, employing the HPLC based approach results in a total analysis time of roughly 115 min for a single sample (≈55 min separation time). The sample preparation time per sample can be shortened by preparing several samples simultaneously. However, for at-line monitoring, the intended approach would be to analyze one sample at a time.

Another major advantage of switching from HPLC to a CE/MCE based setup is the reduction of costs. An HPLC column that can be used ≈300 times costs ≈900 USD (i.e., ≈3 USD per analysis), while a CE capillary that can be used for roughly the same number of runs only costs ≈5 USD, generating an ≈150 times lower cost per analysis. An MCE glass chip could be used for ≈50 runs, resulting in a cost of 1–3.5 USD per analysis, largely depending on the chip type and the quantity of chips purchased. Since CE and MCE do not require a continuous flow of a mobile phase for the separation, the required volume of chemicals and the generated waste is minimized. For CE, a few mL of BGE is sufficient for at least 10 consecutive runs, and for MCE this volume can be further reduced by approximately a factor of 10. In comparison, HPLC consumes ≈1 mL min−1 of mobile phase. Also, the small sample volumes employed in CE and MCE (<10 μL is sufficient for several sample injections) further enables application of these techniques for bioprocess monitoring when microbioreactors are used.

Comparing the results of AA analysis of cell culture medium using the CE, MCE, and HPLC methods, the correlation coefficients, excluding outliers, were all above 0.9. Since the correlation factors were calculated based on analyses performed using real bioprocess samples (and not standards), these can be considered as high values. The fact that the samples analyzed by HPLC were stored at −80°C, whereas the samples analyzed by CE and MCE were partly stored at −20°C, could also affect the correlation between CE/MCE and HPLC, since AAs can degrade at −20°C. Inspecting the results in Figure 5, the concentrations of most AAs increased from day 1 through 8. Generally, the AA feed is adjusted to keep the AA concentrations at stable levels. However, considering that the bioprocess from which these samples were taken was still in development, the increasing concentrations were not unexpected. The AAs in the samples analyzed by HPLC were quantified, based on a previously established calibration protocol, and the concentrations were in the range of 60 μm–11 mm. According to the results presented in Figure 5 and Table S5, this range is feasible also for CE and MCE analyses. Future work includes the implementation of the MCE in the ALIAS system, where an automated liquid handling system will be coupled to the microchip. The final systems performance needs to be evaluated with a higher number of bioprocess samples. Compared to CE, MCE brings increased possibilities for not only at-line analysis and automation but also the possibility of parallelization of several analytical targets. Implementation of the MCE setup in the ALIAS system would result in a powerful tool for at-line bioprocess monitoring.


The Innovative Medicines Initiative 2 Joint Undertaking (grant agreement No 777397) is acknowledged for funding. This Joint Undertaking receives the support from the European Union’s Horizon 2020 research and innovation programme and EFPIA partners Bayer, Byondis, GSK, Pfizer, Rentschler Biopharma, Sanofi, and UCB. Elwin Vrouwe, Micronit BV, Enschede, the Netherlands, is acknowledged for assembling the fluorescence detection module, and Atefe Shokri, Department of Industrial Biotechnology, KTH Royal Institute of Technology, Stockholm, Sweden, for the preparation of AA free medium. The authors thank FUJIFILM Irvine Scientific, Santa Ana, CA, USA for providing the basic medium mixture IC-0. S.M.: Conceptualization, data curation, formal analysis, investigation, methodology, project administration, validation, visualization, writing – original draft, writing – review, and editing. L.J.: Conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing – original draft, writing – review, and editing. V.C.: Funding acquisition, resources, supervision, writing – review, and editing. Å.E.: Conceptualization, funding acquisition, resources, supervision, writing – original draft, writing – review, and editing. M.M: formal analysis, investigation, writing – review, and editing.