Download PDF

Oksana Sirenko, Angelina Chopoff, Krishna Macha,
Marco Lindner, Felix Spira | Molecular Devices

Summary

While cell-based models and assays are basic tools for drug development and biological research, tissue culture remains the most routine, hands-on, and time-consuming part of biological studies. Because of protocol complexity, sensitivity to errors, and substantial training requirements, these processes are difficult to automate and to ensure reproducibility and quality.

To alleviate the limitations that come with labor-intensive cell culture protocols, we developed the CellXpress.ai™ Automated Cell Culture System. The CellXpress.ai system contains four essential components for automated cell culture: liquid handler, automated incubator, and automated imager, plus integrated AI-powered software to automate complex protocols, scheduling, and imagebased analysis. Critically important, the software provides an efficient decision-making tool that allows researchers to fully automate the entire cell culture process.

Here, we share results from a fully automated standard cell culture process. We developed protocols that provide automated maintenance and expansion of HCT116 cell lines. The maintenance and expansion of cell cultures were controlled by image-based automated decision making. Cells were cultured in 6-well plate format with automated media exchange every 48 hours. Images of cells were taken every 8 or 12 hours in transmitted light and the areas covered by cells were determined by image analysis.

Based on the selected threshold for cell-covered area, notifications were sent to the user via e-mail stating that selected plates or wells had reached the threshold of cell density. Cell passaging was then triggered by the user via system software. In the other scenario, upon reaching the trigger measurement, the passaging was triggered automatically without user interaction, while notification was sent to the user.

Introduction

Cell culture is a labor-intensive task that require constant attention to details and pose substantial demands on scientists’ time during all hours of the week—including weekends. Automation can not only alleviate these time-sensitive and highly routine tasks from the scientist, but can also reduce human error and inconsistencies associated with manual processes. While automated liquid handling has previously been developed for complex processes with liquid handler devices, automating the entire process is far more challenging. Decision making based on cell density or growth rates is notoriously hard to automate and to ensure the process quality. Monitoring cell culture development using imaging and image-based decision making presents a unique opportunity to fully automate the entire cell culture process—not just the liquid handling steps. Automated decision making enables not only media exchange and passaging, but also enables cell culture control in the absence of scientists in the lab.

Here, we describe the methods to enable consistent and reliable cell culture passaging and maintenance with minimal user interaction.

Materials

Materials used include HCT116 cells, media with McCoy, 10% FBS, 1% Pen Strep, Corning 500 mL McCoy’s 5A (product number 10-050-CV), ThermoFisher PenicillinStreptomycin (catalog number 15070-063), FBS (VWR 97068-085), Corning 6-well clear plate (product number 3516), 1000 µL tips (CO-RE Hamilton, sterile and nonfiltered), Corning 0.25% Trypsin (product number 25-053-Cl), and the CellXpress.ai system instrument and software.

All the required materials were preloaded into the liquid handler at the time of experiment.

Methods

We used CellXpress.ai Automated Cell Culture System (Figure 1) to automate the maintenance and passaging of HCT116 cell culturing as well as all processes involved in experimental setup, cell analysis measurements, and automated decision making.

In the CellXpress.ai system software, we created a new protocol using the “Feeding with Passaging” phase for the fully or partially automated experiments. This protocol included a feeding step every 48 hours using the McCoy media. When the feeding step activated, 1400 µL of liquid was aspirated from the experiment’s active plate and replaced with 1400 µL of McCoy. Imaging occured every 8 hours using a 4X imaging in transmitted light protocol. Typically, we used 2x2 sites in the center of each well of the 6-well plate, however, the number of sites per well can be increased for more accurate cell density assessment. The protocol “4x TL HCT116 confluence” from IN Carta® Image Analysis Software was used for image analysis. This Image analysis protocol determines the area covered by cells using thresholding of the transmitted light images. The total cell area (cell area covered by cells) is plotted over time and graphs were created showing cell density and proliferation over time.

See the protocol screenshot in Figure 2. Importantly, the imaging analysis rules were designed to inform the user when to trigger cell passaging or, alternatively, to automatically passage cells without user participation.

Figure 1. CellXpress.ai Automated Cell Culture System.

Figure 2. A screenshot of the Feeding with Passaging protocol phase used for automated cell culture.

Results

To develop protocols for the fully automated culture of cells, we used a “Feeding with Passaging” phase and set a protocol that conducted media exchanges every 48h, took images of plates in transmitted light every 8 hours, and performed cell passaging when triggered automatically or manually (see a screenshot above). In the passaging step we set the cell splitting ratio as 1 to 10.

In the first experiment, we triggered the passaging event manually by setting the software to “Inform User” when the plates are ready to be passaged. When users received this message, they triggered passaging by activating the “Passage” button. Based on image analysis, the software annotated the wells that surpassed the arbitrary selected threshold of 3E7 square microns for the total area sum. The selected threshold approximately corresponded to cell density at which the scientist would typically passage cells manually (Figure 3). Such annotation was happening after the analysis of images taken at the last time point was complete. Typically, analysis and decision making will take approximately 10 min. At this point we set the software to send an e-mail notification to the user about wells reaching that cell density threshold.

Once we received notification that a few wells reached the cell area /confluence threshold, we made the decision to activate the Passage button, which triggers passaging of the plate. “Passage” button is present in the review window of the ongoing experiment. Triggering an active Passage step is possible remotely if all the needed consumables are present on the deck. Consumables include empty 6-well destination plates, tips, McKoy media, and Trypsin. After pressing the Passage button, a dialog box lists consumables that are needed vs. those currently present on the deck. If everything is present, selecting “continue” will start the passaging. If something is missing (e.g. not enough tips, media, or plates) that item will appear in red in the dialog and the passaging step will not proceed until the instrument deck is refilled with consumables.

Prior to the experiment, we placed enough consumables and media in the deck. We started the experiment after placing confluent 6-well plate with HCT116 cells into the incubator. Then the experiment was created in the software, the protocol was chosen, and the source plate was selected from the list of 6-well plates available in the incubator (Figure 4). When we “approved” the protocol, the experiment started by imaging the source plate and image analysis. Images were viewed in the Custom view and the plot showed the Total Area Sum over time. Since cells were already confluent in the source plate, notification was sent right after image analysis and the results appeared in the experimental log. At that point we activated the “Passage“ button and proceed with “Continue” since all consumables were present on the deck.

Figure 3. HCT116 cells with low or high confluency (cell area sum 1.6E7, 3.1 E7, and 3.6E7 of um2, respectively); and the analysis mask in IN Carta covering cell area

The passaging step started by taking the destination plate from the appropriate place on the deck and filling it with media. Then, the source plate was taken out of the incubator and the liquid handler began the process of passaging. During passaging, cells were washed with trypsin then incubated with 600 µL of Trypsin on a warm position in the liquid handler for 5 minutes. 1400 µL of media was then added to this source plate followed by mixing several times to dislodge the cells. Meanwhile, the instrument prepared a destination plate with 1400 µL of media in its target wells. 200 µL of the cell suspension was then added to the destination plate from the source plate, thus resulting in a 1:10 dilution. In this partially automated experiment, one well of the source plate would go into three new target wells so the selected protocol would expand the culture from 1 to 3 with each passaging step. To limit the cell expansion, we manually unselected wells which we did not want to pass (wells can be unselected in the Incubator Plate menu tab) (Figure 5).

Figure 4. Creating a cell passaging rule based on image analysis in the protocol.

Figure 5. Schematic diagram of culture expansion and cell maintenance procedures.

Cell culture was carried out for seven rounds of passaging. The growth curves presented on Figure 6 show the plot of Total Cell Area for each well or a plate average for different imaging time points. Some missing dots/timepoints were due to pausing the instrument for training or demos. When the process was un-paused, the protocol resumed as expected.

In the second experiment, we conducted a fully automated passaging trigger. This experiment used a very similar protocol and also had a 1:10 dilution when passaging and expanded one well into three destination wells. However, in this protocol the action triggered by the decision-making rule was to “Proceed with passaging”. The trigger based on cell area remained the same, however passaging would automatically start when the user-defined percent of wells reached the designated threshold—without any interaction by the user required. The only remote user interaction was related to de-selection of wells that were not desired to passage, to limit the cell expansion.

This experiment was performed for seven passages. When cell areas were reaching the threshold, passaging was triggered automatically and notifications were sent by e-mail. Figure 7 shows a plot for cell growth and expansion averaged for the entire plate; the bar graphs shows averages and error bars for Std Dev between the wells.

The third experiment was entirely automated. We used the same general protocol as the previous experiments. This time, the only alteration was changing the number of target wells per source well to one. This eliminated the need to deselect the wells that were not intended to be passaged. One-to-one passaging maintains the same number of wells/plates. Figure 8 shows the cells growth curves for seven rounds of passaging.

Figure 6. Growth patterns of HCT116 cells of each well on a Corning 6 well plate during a semi-automated experiment on the CellXpress.ai system. Graph A shows plots per well and per plate respectively. Graph B shows averages and Std Dev between wells during the passaging process.

Figure 7. Growth patterns of HCT116 cells on a remotely automated experiment on the CellXpress.ai system. Total Area Sum in square microns vs. Time where error bars are the standard deviation of the total area sum µm2 of each well of the remotely automated experiment. Note: missing time points in graphs were due to pausing when the instrument was used for customer demos.

Figure 8. Growth patterns of HCT116 cells in a fully automated “cell maintenance” experiment.

Conclusion

We demonstrated fully automated culture of HCT16 cells using CellXpress.ai system. The system is very efficient at reliably maintaining and/or expanding cultured cells. A similar protocol was tested using CHO cells and can be adopted for any cell line.

Download PDF