Localized Sampling Enables Monitoring of Cell State via Inline Electrospray Ionization Mass Spectrometry

Abstract

Nascent advanced therapies, including regenerative medicine and cell and gene therapies, rely on the production of cells in bioreactors that are highly heterogeneous in both space and time. Unfortunately, advanced therapies have failed to reach a wide patient population due to unreliable manufacturing processes that result in batch variability and cost prohibitive production. This can be attributed largely to a void in existing process analytical technologies (PATs) capable of characterizing the secreted critical quality attribute (CQA) biomolecules that correlate with the final product quality. The Dynamic Sampling Platform (DSP) is a PAT for cell bioreactor monitoring that can be coupled to a suite of sensor techniques to provide real-time feedback on spatial and temporal CQA content in situ. In this study, DSP is coupled with electrospray ionization mass spectrometry (ESI-MS) and direct-from-culture sampling to obtain measures of CQA content in bulk media and the cell microenvironment throughout the entire cell culture process (~3 weeks). Post hoc analysis of this real-time data reveals that sampling from the microenvironment enables cell state monitoring (e.g., confluence, differentiation). These results demonstrate that an effective PAT should incorporate both spatial and temporal resolution to serve as an effective input for feedback control in biomanufacturing.

1. Introduction

The treatment of life-threatening ailments is being transformed by advanced therapies including gene therapies for inherited diseases, therapeutic cell treatments (e.g., immunotherapies) for cancers and autoimmune disorders, and tissue engineered medical products to restore, maintain, and replace damaged organs.^1–4 The outlook for these therapeutics is promising, with over 900 new investigational drug applications for cell and gene therapy products reported by the FDA as of January 2020. However, even idealized allogenic models for CAR-T production estimate approximately 25% of the cost to manufacture is associated with quality control.⁵ A shortage in suitable real-time quality control methods capable of monitoring cell bioreactors for feedback control results in large batch-to-batch variability and ad hoc approaches to cell culturing that make scalable and high yield manufacturing difficult and costly.⁶ Since advanced therapy workflows depend on the growth of cells in bioreactors, the process analytical technologies (PAT) for real-time monitoring of cell secreted biomarkers are essential for cost effective biomanufacturing of high quality therapeutic products.

As cell cultures mature, the biomolecules they secrete as signaling and paracrine factors serve as the critical quality attributes (CQAs) for cell biochemical state and final therapeutic potency.^7–9 Faley et al. demonstrated the detection of these secreted biomarkers is challenging because the secretions are rapidly diluted in the relatively large amounts of media used in most culture processes.¹⁰ As shown in figure 1, the concentrations of CQAs vary significantly with both space and time in a cell culture/bioreactor. The amount of secreted molecules is highest close to the cell surface with the concentration rapidly decreasing further away into the bulk media. As an illustrative example of secretion from an isolated cell, the concentration of secreted molecules quickly reaches the steady-state with a profile featuring a 1/R (where R is a characteristic radius of the perturbation, proportional to the cell radius) decay away from the cell surface. Even if the size of the perturbation is greater than that of a single cell (e.g., corresponds to a multi-cell aggregate), the domain within which the secreted molecules are enriched is very small, highlighting the critical need for local probing. The actual length scale for this analysis may vary depending on the particular culture conditions, as would the boundary condition at the cell surface (e.g. secretion vs uptake), but these scaling arguments indicate that (1) significant secretome variation is expected in the cell culture and (2) local sampling gives access to a highly enriched state of the secreted biomolecules prior to their dilution in the bulk. A complementary aspect of the locality of the perturbation is the importance of fast sampling to capture the transient evolution of the secretome, as the time scale at which a new steady-state is established is fast and necessitates a probe capable of rapid sampling. This time scale is proportional to the square of the perturbation domain and is inversely proportional to the species diffusion coefficient. Considering a secretome evolution caused by heterogeneity at the length scale of 35 × 10⁻⁶ m (cell diameter for the MC3T3 cells used in this study) and using a characteristic diffusion coefficient of ~10⁻¹⁰ m²/s for a 10 kDa protein in water, the perturbation is dissipated on the time scale of ~ 10 seconds.^{11, 12} In other words, the microenvironment represents an instantaneous CQA composition while the bulk provides only a temporal average.

Figure 1:

Motivation for localized sampling and the analytical framework for the Dynamic Sampling Platform (DSP). While cell cultures mature, secreted biomolecules (linked to cell status and growth trajectory) are in highest abundance near the cells’ surface. Traditional analysis techniques analyze bulk media, which include temporally averaged biomolecular content and higher relative concentrations of molecules contained in stock media (e.g., fetal bovine serum (FBS), salts) that can mask the signature of quality indicating analytes. Rapid, small volume, localized sampling is an advantageous approach because it enables spatial and temporal probing of the cell culture, detecting the molecules when and where they are secreted. After sampling, DSP rapidly treats the sample for inline CQA characterization. As a platform technology, DSP is easily integrated with a broad spectrum of analytical tools, but for this study DSP was coupled with inline electrospray ionization mass spectrometry (ESI-MS).

Most available real-time PATs, such as those that measure temperature, pH, dissolved oxygen, and glucose are based on analytical outputs that lack the specificity and sensitivity required to discover and detect biochemically complex, low concentration CQAs. These PATs indicate general culture viability but are not useful for predicting the final products’ quality. This has motivated considerable efforts to develop other approaches to non-invasive, real-time monitoring including volatile species mass spectrometry, Raman spectroscopy, and infrared or near-infrared spectroscopy.^13–15 Yet, these approaches are still lacking in their utility for cell bioreactor monitoring, in part due to their poor specificity, limited range of detectable molecules, and low sensitivity for dynamic secretome characterization. A clear and continuing need exists for real-time quality control measurement techniques that are highly sensitive to secreted biomarkers with complex biochemical signatures. An ideal PAT should also be label free or untargeted to enable broad biomolecular detection such that unanticipated or unidentified biomarkers may still be characterized for better process understanding. Electrospray ionization mass spectrometry (ESI-MS) is a particularly promising PAT candidate due to its broad molecular weight coverage, sensitivity, and ability to preserve structure/folding and non-covalent interactions of biomolecular complexes through “soft-ionization”.^16–18 Recently, we demonstrated continuous ESI-MS sensing with a ~1 minute response time for detecting the biomolecules serving as proxies to target CQA species.¹⁹

The Dynamic Sampling Platform (DSP, Fig. 1) is a multi-functional analytical platform for cell bioreactor characterization that can be integrated into therapeutic cell manufacturing quality control approaches. The DSP samples very small volumes (~1 μL) of liquid from the reactor microenvironment and then processes the sample for real-time analytics. As a platform technology, DSP can integrate with the optimal analytical tool chosen for a specific application: in this work DSP is coupled with ESI-MS for its capability as a discovery tool.

Using DSP coupled to ESI-MS, we show that CQA heterogeneities exist even within 2D cell cultures. These heterogeneities are important because they may impact aspects of cell metabolism, final yield, and product quality. As DSP direct-from-culture sampling does not affect culture sterility or cell growth trajectory during the 3 week culture process, DSP is found to be a suitable candidate for continuous, real-time characterization of CQA content in bioreactors. Collectively, these results constitute a vital set of capabilities and biochemical data that i) demonstrate that the local CQA content is critical to detecting cells in their various developmental states (e.g., proliferative, confluent, differentiated) and ii) establish DSP as a viable analytical platform for in situ monitoring of bioreactor state and cell development.

2. Materials and methods

2.1. Dynamic Sampling Platform Design & Sampling Methodology

The Dynamic Sampling Platform (DSP) for ESI-MS incorporates aspects of the previously reported Dynamic Mass Spectrometry Probe (DMSP) to create a system for direct-from-culture monitoring. Full details regarding DSP microfabrication, operation, sample treatment efficiency, and comparison to other technologies for real-time ESI-MS have been described elsewhere.^19–21 As shown in Figure 2, the DSP includes 1) a spatially resolved sampling interface for sterile, direct-from-culture media uptake, 2) an optimized “cross flow” sample treatment mass exchanger for inline sample preparation (see supplementary information for details) and 3) an inline ESI emitter for online MS analysis. The DSP is a microfabricated mass exchanger with an integrated 360 μm OD, 50 μm ID PEEK (Upchurch Scientific) inlet for sample intake (probing tip) and a fused silica nanoESI emitter with a 360 μm OD and a 75 μm ID tapered to a 30 μm emitter (New Objective) outlet for inline electrospray ionization of the sample analytes for MS analysis. The device is enclosed in a sealed fluidic package (Fig. 2) which allows for easy integration with MS. The package is designed for the introduction of high flow rate (50 mL/hr) conditioning flow to the mass exchanger section of the DSP, which is an important aspect enabling rapid molecular separation.

Figure 2:

Dynamic Sampling Platform (DSP) ESI-MS setup for direct 2D cell culture analysis. The DSP is contained in a fluidic package which allows for continuous sample treatment in a tangential flow mass exchanger orientation (right inset image) and is positioned with the ESI emitter directly in front of the MS inlet. The DSP is connected to the sampling interface, which enables direct-from-culture sampling. Localized sampling is carried out by positioning the intake capillary directly above the bottom of the cell culture, within ~50 μm of the cells, while bulk sampling is carried out by submerging the capillary just below the media surface (bottom inset image).

Sample is extracted from the cell culture and injected into the microfabricated mass exchanger sample channel at a flow rate of 30 μL/hr. The sample channel is fluidically coupled to the high flow rate conditioning channel through a thin (~5 μm) nanoporous alumina membrane. The pore size (50 nm) impedes the transport of larger biomolecules of interest while allowing smaller molecules to diffuse rapidly into the conditioner channel. Simultaneously, chemicals which have been shown to enhance ESI-MS performance (acetic acid or AA, and 3-nitrobenzyl alcohol or m-NBA) are injected into the sample channel to improve ionization and aid in MS detection of biomolecules.^{19, 20}

The total volume of the mass exchanger is ~20 nL, the ESI emitter ~200 nL, and the PEEK tubing ~250 nL. Given its microfluidic design, DSP is remarkably fast: at a sample flow rate of 30 μL/hr the sample transmission time is ~1 minute. Direct-from-culture sampling was achieved through a sterilized (via autoclaving) ~10 cm length of 360 μm OD, 50 μm ID PEEK tubing (IDEX) submerged directly into the cell culture. A 1 μL sample was drawn into the sampling port at 50 μL/hr and then infused through the DSP system at 30 μL/hr. Two sampling methodologies were employed with identical results: the sampling port was either attached to a switching valve to allow for switching between sample uptake and infusion (as shown in Figure 2) or attached to the DSP system via quick disconnect fitting for infusion. In both cases an identical volume was sampled and then infused through the DSP system. The integrated capillary probe enabled precise control of the inlet positioning ~50 μm to the cell surface (monitored optically via Dino-lite AM7515MT4A microscope). This small volume extraction approach ensures that only the local microenvironment is sampled and allows for more frequent sampling without affecting culture viability by removing excessive amounts of media.

2.2. Cell Culture Method

Murine preosteoblastic cells MC3T3-E1 (ATCC CRL-2593) were obtained and expanded according to established protocols.^{22, 23} Cells were seeded on T-150 flasks (Corning) at 5000 cells/cm² and expanded in a growth media consisting of MEM ⍺ with nucleosides (Gibco) with 10% fetal bovine serum (Atlanta Biologicals, Lot #E15052) and 1% penicillin/streptomycin (Corning). Media was replenished with clean growth media every 2–3 days until cells reached 80% confluence. Once confluent, cells were washed with phosphate-buffered saline (PBS, Gibco) and detached from flasks with 0.25% trypsin-EDTA (Gibco). Dissociated cells were counted and replated onto 6-well plates at 5×10⁵ per well (53,000 cells/cm²). After overnight adhesion, MC3T3s were separated into two groups, one group subjected to osteogenic differentiation (i.e., differentiated group) using the In Vitro Osteogenesis Assay Kit (ECM810, EMD Millipore) and the other cultured in growth media (i.e., undifferentiated group). For cells undergoing osteogenic differentiation, media was aspirated and replaced with 2.5 mL growth media supplemented with 0.2 mM ascorbic acid 2-phosphate (EMD Millipore), and 10 mM glycerol 2-phosphate (EMD Millipore) while non-differentiated cells were aspirated and given 2.5 mL growth media with no additives. Every 3 days, media was replaced after DSP sampling. Immediately after samples of conditioned media from both differentiated and non-differentiated MC3T3s were collected, the media was replaced. After 6 days in culture, the differentiation media was also supplemented with 50 nM melatonin (EMD Millipore). Both cell groups were cultured for 18 days total, being sampled 6 times total. After the final DSP experiment (day 18) cells were washed with PBS and fixed with 4% paraformaldehyde (Sigma Aldrich) in PBS at room temperature for 15 minutes. Fixative was removed and the cells were carefully rinsed three times with distilled water. Next, differentiated and undifferentiated MC3T3s were incubated in 2 mL alizarin red staining solution (EMD Millipore) for 20 minutes. Staining solution was then removed and cells were washed 4 times with deionized water. Osteogenic differentiation was confirmed in the differentiated group by red staining of calcium deposition, while no staining was observed in the non-differentiated group, as expected (Fig. S7).

2.3. DSP ESI-MS Application to 2D Cell Culture

The MC3T3 cells were cultured in a lab separate from the lab where the mass spectrometer used for online analysis was located, requiring a transport and sterile sampling method that preserved the cell growth processes throughout the entire culturing period. When ready for sampling, the cells were removed from the incubator and placed into a sterile fume hood. The 6 well plate cover was removed and replaced by autoclaved aluminum foil, which was taped down in a manner that allowed for visual inspection through the side of the culture plate. During sampling, a digital microscope was positioned orthogonally to the cell culture plate, allowing for visual confirmation of sampling mode (i.e., local or bulk) as shown in the Figure 2 inset image. After the cells were ready for transport, they were placed in a Styrofoam cooler and brought to the mass spectrometer lab, where they were immediately placed on a hotplate to maintain the media at ~37 °C throughout the entire sampling process.

Direct from culture DSP-ESI-MS analysis was carried out every 3 days immediately before media changes. The MS (Bruker™ MicrOTOF) was tuned and calibrated using Bruker™ tune mix at the start of each experiment to facilitate accurate mass identification. To remove all air bubbles, which interfere with continuous ESI, the entire fluidic system was primed with sterilized DI water. A 22 gauge sterilized hypodermic needle was used to puncture the foil and the sampling inlet was inserted through the hole for sampling. Directly before each sampling event, 2 μL of the sterilized water was pumped out of the sampling inlet at 100 μL/hr to purge the sample inlet line, reducing sample carry over effects. Samples of 1 μL volume were drawn into the sampling interface at 50 μL/hr and subsequently infused through the DSP for direct ESI-MS analysis at 30 μL/hr. Conditioning flow (1% acetic acid, 1% m-NBA) was run continuously at 50 mL/hr throughout the entire experiment. The conditioning flow channel also served as the electrical connection for a picoammeter, which was connected to a stainless steel wire submerged in the conditioning liquid reservoir. The electrospray current during all experiments was maintained at 10–20 nA so that the ESI characteristics remained consistent between experiments. Between each sample, DI water was infused through the DSP and into the MS via ESI to remove residual signal before starting another sampling event. 6 samples were captured from each cell group (i.e., differentiated/undifferentiated) at each time point, 3 local and 3 bulk (Fig. 2, inset), with each sample taken at a spatially disparate location to probe heterogeneities throughout the volume.

2.4. Mass Spectra Data Processing for Principal Component Analysis (PCA)

Bruker MicrOTOF output files were parsed and converted to “mzML” file type²⁴ using the open source data conversion software, ProteoWizard.²⁵ Principal component analysis (PCA) was completed using two open source software utilities. The first was ms-alone, a python-based utility for preprocessing and peak extraction used on the raw mzML files prior to importing the data to the second utility, multiMS-toolbox, an R based software for PCA.²⁶ Preprocessing was accomplished using ms-alone on the raw data including baseline subtraction and peak smoothing. Within the ms-alone setup, the signal to noise ratio threshold of data was set to 0.5, and a Savitzky-Golay smoothing method was used to reduce noise.²⁷ Intensity based (i.e., no data normalization) PCA was run on the entire spectrum (m/z 0–2500) and then winnowed down according to raw loadings plots. The range of m/z values with the highest contribution to variance were chosen as the reduced data window size. Without winnowing, the large amount of low intensity noise at the low (m/z 0–500) and high (m/z 1500–2500) ends of the spectra masked the contribution of higher signal to noise ratio features in the middle ranges (Fig. S3). The mass ranges which were removed during winnowing are unlikely to have featured critical biomarkers because lower m/z ranges are not probed with DSP due to the removal of small molecules during sample treatment, while the high m/z range would contain contributions from high molecular weight media additives that are not part of cell secretome.^{19, 20}

3. Results

A total of 72 spectra were obtained, which are included in the supplementary information. Because of the unsupervised nature of the technique, principal component analysis (PCA) was selected for data analysis that focuses on discovery aspects of unanticipated trends. PCA only reveals group structure when within-group variation is less than between-group variation.²⁸ On the other hand, supervised techniques, such as partial least squares discriminate analysis (PLS-DA), are useful for many omics studies, including metabolic fingerprinting experiments, but there is concern that this type of analysis is prone to overfitting and providing false discoveries.²⁹ Such supervised methods for data analysis could be employed in future studies once the discovery-focused PCA has revealed sufficient group structure to help unambiguously identify which specific features in the raw data contribute most to between-group variation. For this study, PCA was employed to reduce the chance of false identification of between-group variation. PCA identified subtle yet significant differences between the spectra and specific m/z values that contributed most to variance between data.²⁷ These selected m/z values were then later used for comparison with high-performance liquid chromatography (HPLC) data to explore whether the spectral features identified via PCA corresponded to CQAs tentatively identified via HPLC.

Due to the small number of sample replicates at each time point, data from multiple time points was grouped to increase the robustness of PCA. The cells in the differentiated group are expected to have begun differentiation by time point 4 and completed by time point 5, while the cells in the undifferentiated group remained undifferentiated throughout the experiment. Time points 1 and 2 were grouped to represent cells in their initial state while time points 5 and 6 were grouped to represent cells in their final state. Subsequent staining (Fig. S7) of both the undifferentiated and differentiated cell lines confirmed that the cells had either remained in the undifferentiated state or had completed differentiation after time point 6, as expected.

The first analysis compared cells in their final states from the undifferentiated and differentiated cultures, i.e., comparing time points 5 and 6 for both cultures. The resulting PCA cluster plots based on bulk samples, taken far from the cells, are shown in Figure 3A. The groupings in these bulk sampling plots are not as well segregated as the groupings for the localized samplings shown in Figure 3B. This suggests that localized sampling is important to being able to detect differences in cell differentiation state. Although both cell groups were given different media throughout the culture process, these differences are not revealed by the bulk samples which suggests that the differences observed in the local samples are due to secreted biomarkers captured near the cells, and not due differences in the cell culture itself.

Figure 3:

Principal component analysis (PCA) cluster plots for time points 5 and 6 of the undifferentiated cell group vs time points 5 and 6 of the differentiated cell group. A) Bulk sampling reveals minimal clustering B) Localized sampling reveals clusters for the two groups.

In order to observe if the same cell culture exhibited differences with time, and to remove the contribution of cell culture conditions to the variance in data, the differentiated cell culture was analyzed alone. For this PCA approach, time points 1 and 2 were grouped and compared to time points 5 and 6. These groupings compare the cells from the same culture in early time points, when they are still undifferentiated cells, to the cells in a fully differentiated state. Since all of the data was taken from the same cell culture in the same continuously performed cell growth and development experiment, this approach also removes the possibility that different culture conditions (e.g., seeding density, media type, etc.) contributed to the differences observed. Figure 4 shows the resulting PCA plots for these groupings. Once again, the spectra do not exhibit significant clustering for bulk sampling (Fig. 4A) but show a stronger clustering for the localized sampling (Fig. 4B). These PCA results (Fig. 3B and 4B) indicate that localized sampling provides an enhanced capability to detect differences between the cells due to secreted biomarkers, which are in highest concentration near the cell membrane.

Figure 4:

Principal component analysis (PCA) cluster plots for the differentiated cell line at time points 1 and 2 versus time points 5 and 6 (same analysis done for undifferentiated group in supplementary information). A) Bulk sampling does not exhibit data clustering, and B) With localized sampling, clusters are observed. Along PC1 with localized sampling, subgroup separation is observed between the time points 1 and 2.

4.1. Discussion

The ability to determine, in real-time, cells’ health, therapeutic potential, or developmental state is paramount to developing effective feedback control for advanced therapy production. From a manufacturing standpoint, detecting cells in their proliferative, confluent, or differentiated states is critical to understanding the processes governing their production.^{30, 31} As a first step towards real-time monitoring of primary cell cultures, DSP was applied to MC3T3 cells for rapid ESI-MS, followed by PCA of the collected spectra to determine if features in the raw spectra corresponded to differences in the secretome, and therefore the cell state. This study was used to assess whether the known spatial heterogeneities in biomolecular content affect a PAT output. In this sense, the negative results from bulk samples (Fig. 3A and 4A) are just as important as the positive results from localized sampling (Fig. 3B and 4B). Depending on where a sample is taken in a bioreactor, the resulting PAT output is altered. An effective PAT should then capture these heterogeneities in the culture environment.

PCA of input spectra from bulk samples, in Figures 3A and 4A, shows no clustering, which indicates that the input spectra are not statistically distinguishable despite the fact that cells had undergone proliferation to confluence and then differentiated between time points 1 and 6: the cells’ secretome was expected to have changed. The data in Figs. 3A and 4A indicate that ESI-MS is unable to distinguish between cell types using sampling from the bulk away from the cells. Only DSP analysis of locally garnered samples produced clustering in ESI-MS output. This suggests that PAT output depends on spatial context, and that these spatial variations are not negligible and should be accounted for when designing sensor systems.

Unlike bulk analysis, PCA using spectra from localized sampling as an input revealed clusters which are associated with cells in their undifferentiated and differentiated states in figures 3B and 4B. An important feature within figures 3B and 4B is the great deal of variation amongst the same cell group along the horizontal axes (principal components 1 and 2). One explanation for this variation is that every sample taken from the cell culture was taken from a spatially disparate location. The variation observed along the horizontal axes in these cluster plots may reveal subtle changes in the secretory profile that are revealed when sampling at different locations in the well. Of course, there are a number of other factors that could contribute to this variation within the same group, including the fact that the cells in the undifferentiated and differentiated cultures experienced different culture conditions, and that the cells were expanding during these experiments.

In the PCA scatter plots in figure 3B and 4B, separation between cells in their undifferentiated and differentiated states is along the third, orthogonal, principal component 3 (vertical axis). The separation along a single axis suggests that even with variation within subgroups, the dimensionality can be reduced to a single component that captures a significant amount of variation in the input spectra which can be used to predict cells in their undifferentiated or differentiated state. The additives used to induce differentiation in the MC3T3 cells are expected to have been removed due to their small size (MW <300 Da). Therefore, it is unlikely that the presence of the additives contributed to the separation observed between the different cultures shown in figure 3B.^{19, 20} In order to remove the potential variation due to different culture conditions, the differentiated cell group was analyzed in an early time point vs a late time point to again compare cells in their undifferentiated and differentiated states. This comparison mitigates the effect that different media compositions could have on the analysis. In fact, each sample was taken from the same well, so this experimental design removes any changes in cell culture conditions that could contribute to variation in the data. The resulting PCA cluster plot with input spectra from early and late time points from the differentiated cell group (Fig. 4B) separates well along the vertical axis (principal component 3), corroborating the assumption that the clustering observed in figure 3B was most likely due to cell differentiation and not culture conditions. Furthermore, this analysis emphasizes a central finding of this work which is that cell bioreactors are highly heterogeneous in nature, and PAT measurements are dependent on where a sample is taken.

Along principal component 1 in figure 4B the undifferentiated cell group (red dots) separates into two subgroups, corresponding to time points 1 and 2, which are demarcated with ovals highlighting different samples. The cells were seeded such that they were proliferative early in the culture process and later reached confluence. Since the subgroup separation in figure 4B correlated with proliferation, it was expected that similar variation could be observed in the undifferentiated culture. Therefore, PCA was applied to time points 1 and 2 versus time points 5 and 6 for the undifferentiated cell group (Fig. S3). In this case, bulk sampling once again resulted in no separation, and for local sampling the separation between the two groups is along principal component 1. This suggests that the variation along principal component 1 may be due to a change in the cell secretome during expansion, which has been shown to correlate with preosteoblasts in a state of proliferation (early) or confluence (late).³² In other words, while principal component 3 correlates well with cell differentiation, principal component 1 correlates with cell proliferation. Future studies will be designed to create spatial maps/scans of the cell culture which will help to investigate cell-to-cell heterogeneity throughout the culture as well as suspected sources of heterogeneities, such as edge effects in 6-well plates.

In addition to the experiments to probe spatial effects in more detail, it would be of fundamental interest to identify target biomolecules which are known to correlate with cells in specific states. One drawback of the experimental design used here is the MS used for real-time analysis was not capable of MS/MS (tandem mass spectrometry) for potential chemical IDs. Coupling DSP with an MS system capable of data-dependent-analysis (DDA) for feature identification will significantly increase the potential for CQA identification. However, the PCA output did enable some exploratory work towards chemical identification. Offline HPLC-MS was performed on aliquots of conditioned media which were gathered during media change and frozen at −40 °C until the end of the study. HPLC was carried out on media from both cell types at time points 1, 2, 5, and 6 to identify candidate differentiation biomarkers and to quantify differences in biomolecules between timepoints. The candidate biomolecules identified with HPLC were then manually compared to m/z values with the highest contribution to variance in the PCA data that are hypothesized to correlate with cell state (i.e., differentiated vs undifferentiated). Since the mass spectrometer used for DSP analysis was not equipped with tandem mass spectrometry, fragmentation patterns could not be matched to the HPLC-MS data. Therefore, the m/z values from the PCA loading data were matched to potential chemicals from HPLC-MS based on accurate mass alone, resulting in tentative IDs only. Multiple candidate molecules were identified using this approach, which suggests that DSP may not only generate cell culture “fingerprint” spectra, but also help to identify which detected biomolecules correlate with cell state (supplementary information, table S1). For more complete identification of biomolecules, DSP can be used with inline MSⁿ for fragmentation data to generate candidate IDs before HPLC is carried out. Further studies will elucidate how PCA loading data correlates with quantitative HPLC data for independent identification of CQA biomarkers using DSP based analysis.

4.1. Conclusions

Current advanced therapy manufacturing faces severe limitations in process control because existing PATs provide analytical outputs based on bulk measurements that do not represent the true state of the product. The Dynamic Sampling Platform (DSP) ESI-MS is a sensitive analytical platform for probing heterogeneities in a bioreactor. DSP incorporates highly localized sampling direct from culture, inline sample treatment, and real-time analysis that can be adapted for a range of cell cultures (e.g., 3D bioreactors) and sensing techniques such as ESI-MS (and others, e.g. Raman, NMR). For this study, DSP was coupled to a spatially resolved sampling inlet for sampling directly from a 2D cell culture bioreactor. The minimal dead volume DSP system enabled inline preparation of a sample for direct ESI-MS analysis via a significantly improved and modified DMSP (Dynamic Mass Spectrometry Probe).¹⁹ DSP with real-time, untargeted, MS sensing capability aims to capture and detect low concentration biomolecules secreted by cells. These biomarkers in the microenvironment correlate to cell states relevant for commercial cell production (e.g. differentiation, proliferation, confluence). Capturing and characterizing these CQAs in real-time is critical to enabling feedback control necessary to scaleup and scale out advanced therapeutics.

Application of the DSP enabled in situ ESI-MS analytics to preosteoblast MC3T3 cells revealed that localized sampling was requisite for detection of differences between the cells in an undifferentiated vs a differentiated state. The mass spectrometer in this study was a time-of-flight (TOF) style, which provides fast analysis and accurate mass of biomolecules but cannot fragment these molecules for database matching. Exploratory work on matching the PCA based m/z values to tentative chemical IDs from offline HPLC-MSⁿ of the same samples resulted in multiple matches, which suggests that DSP may have utility as a discovery tool. Future work aims to incorporate DSP into workflows with tandem mass spectrometry capabilities, allowing for real-time identification of detected biomolecules with both accurate mass and fragmentation patterns, facilitating the complete identification of CQAs in real-time. Importantly, in its current design, DSP can be used in conjunction with traditional approaches such as HPLC-MS to bring an important new dimension of dynamic monitoring to clinically relevant workflows. Ultimately, DSP should allow for better understanding of how secreted biomolecules correlate with cell state and growth trajectory (e.g., MSC differentiation or T-Cell CD4 vs CD8 sub-type concentration) and, when developed further, enable the real-time monitoring of CQAs for improved manufacturing within all cell-based workflows.^33–35

Abbreviations:

DSP

dynamic sampling platform

CQA

critical quality attribute

PAT

process analytical technology

ESI-MS

electrospray ionization mass spectrometry

MC3T3

mus musculus calvaria derived osteoblast

MSC

mesenchymal stromal cell

CAR-T

chimeric antigen receptor t-cell

OD

outer diameter

ID

inner diameter

PCA

principal component analysis

PLS-DA

partial least squares discriminate analysis

HPLC

high performance liquid chromatography

AA

acetic acid

3 m-NBA

nitrobenzyl alcohol