Characterization of linear speed, shear, mixing, and gas transfer in the iCELLis™ 50 bioreactor

The linear speed (also commonly referred to as linear velocity or volumetric flow rate) inside the iCELLis™ 50 bioreactor vessel is the velocity of liquid media flowing through the fixed bed. Linear speed is an important parameter because it influences mass transfer, cell attachment, and shear properties inside the fixed bed.

In this study, we investigated the relationship between linear speed, agitation speed, working volume, and falling film height for all six vessel sizes in the iCELLis™ 50 bioreactor. Before characterizing the linear speed, we determined the relationship between working volume, agitation, and falling film height.

To characterize this relationship, we tested all iCELLis™ 50 fixed bed sizes with a 2 cm, 6 cm, and 10 cm falling film height, at agitation set points between 200 and 450 rpm with 50 rpm increments. For each experiment, we installed a vessel on the control system and filled it with water. We initiated agitation at the desired set point, and added water until the desired falling film height was reached. We used the falling film height markers etched onto the outside of the fixed bed for visual confirmation of the correct falling film height. Then, we used the load cell measurements to determine the volume at which this falling film height was achieved. These tests were conducted in either triplicate or duplicate.

Fig 2. Correlations describing the relationship between vessel volume, agitation speed, and falling film (FF) height for all six vessel sizes. Markers indicate individual data points of vessel volume at different agitation speeds, with a linear regression of best fit.

From the results of these tests, we can draw two important conclusions (Fig 2). First, the relationship between falling film and working volume is not substantially impacted by the agitation speed. This is because the difference in working volume required to maintain the same falling film height between 200 and 450 rpm is less than 100 mL in almost all cases. Second, the volume of media displaced by the carriers is small, as there is only approximately a 200 mL difference between the smallest (6 m²) and largest (50 m²) vessel sizes.

From this data we defined the minimum and maximum working volumes of the vessel—the minimum working volume was defined as the volume at 10 cm falling film, and the maximum working volume was defined at 0 cm falling film. These limits are listed in Table 1 for each vessel size.

Table 1. Minimum and maximum working volumes of each iCELLis™ 50 vessel

Once we defined the relationship between working volume and falling height, we proceeded to characterize the linear speed. To test this, we evaluated each vessel size (6, 10, 13, 20, 33 m², and 50 m²) at the agitation set points of 200, 250, 300, 350, 400, and 450 rpm at both the 6 and 10 cm falling film heights. Shorter falling film heights (i.e., 0 cm falling film) could not be tested due to physical limitations in the test setup. We performed each test condition with three technical replicates, and repeated this process on a second set of vessels to account for vessel-to-vessel packing variability. In total, this testing strategy resulted in six data points for each combination of vessel size, falling film height, and agitation speed. Reverse osmosis (RO) water was used as the fluid cycled through the fixed bed for testing.

For linear speed measurement, we removed the top lid of the vessel to gain access to the fixed bed. To maintain a constant volume during the testing, we positioned a flow catcher on the outer edge of the fixed bed of the vessel to collect any liquid overflowing from the fixed bed. Overflown liquid was then pumped out of the flow catcher and through a flow meter, and was immediately pumped back into the bioreactor vessel.

To determine the linear speed through the fixed bed, we used the pump to match the flow rate of the water entering the flow catcher from the fixed bed such that the liquid level in the flow catcher remained constant. We recorded this steady state flow rate, as measured by the inline flow meter, as a volumetric flow rate (L/min) and calculated the linear speed.

This process was repeated at each agitation speed tested, from the lowest to the highest agitation speed. Since it was necessary to remove the vessel top lid during testing, the dip tubes and the integrated sensors attached to the top lid were not present during testing. We assume that this difference would have a negligible effect on flow and therefore pose minimal impact to linear speed measurements. Before measurement, the fixed bed was circulated with water for at least 2 h to hydrate the carriers.

The linear speed was calculated using the following formula:

Table 2. Definitions of terms used for linear speed calculation

Figures 2 and 3 illustrate the relationship between linear speed and agitation speed for each vessel size, at both 6 and 10 cm falling film heights, respectively.

Fig 3. Linear speed characterization as a function of the agitation speed for each vessel size at 6 cm falling film height. Error bars represent the standard deviation calculated from six data points.

Fig 4. Linear speed characterization as a function of the agitation speed for each vessel size at 10 cm falling film height. Error bars represent the standard deviation calculated from six data points.

From these data, we observed that linear speed depends on multiple variables (Fig 3 and 4). These variables include:

• Agitation speed (rpm)—at higher agitation speeds, there is more force driving media flow through the fixed bed, resulting in a higher linear speed.
• Fixed bed size—macrocarriers act as resistance to media flow, so larger fixed beds with more densely packed carriers have lower linear speeds compared to smaller fixed beds at the same agitation speed. Figures 3 and 4 show that vessels with more carriers (higher carrier surface area) correspond to higher resistance to flow.
• Volume of media inside the bioreactor—linear speeds are higher at the same bed size and agitation speed at a 6 cm falling film (Fig 3) than at a 10 cm falling film (Fig 4). Media volume in the reactor exerts a hydrostatic head pressure on the impeller, aiding the media flow. Vessels filled to higher volumes have higher linear speeds at the same agitation speed. Note that working volume also directly defines the height of the falling film.

To determine the agitation rate and working volume required for a specific linear speed, we used the calculator of the mPath™ software in the iCELLis™ 50 bioreactor system (Fig 5). After inputting the fixed bed size, desired linear speed, and falling film height, the calculator determined the required agitation rate and the target working volume.

Fig 5. Screenshot of the agitation faceplate, including the agitation speed calculator, in the iCELLis™ 50 mPath™ Link software. 1. The user inputs the bioreactor size (6 to 50 m²), the desired linear speed (0 to 3 cm/s), and the desired falling film height (0 to 10 cm). 2. The calculator then calculates and displays the new set points for agitation speed (calculated rpm) and the calculated volume (L).

Mixing time is important in bioreactors because it provides an indication of homogeneity, which is crucial for consistent and efficient cell growth and product yield. Proper mixing prevents uneven distribution of nutrients, oxygen, and other compounds, while also controlling temperature and shear forces. The faster the mixing time, the more homogeneous are the bioreactor contents.

For these tests, we evaluated the mixing time by measuring changes in pH as a result of base addition. However, in a live cell culture, vessel homogeneity depends on the mixing of additional substances, such as cells, media feeds, and transfection complexes.

We performed the mixing experiments in two phases. For the first phase, we installed a 33 m² iCELLis™ 50 vessel on a standard iCELLis™ 50 control system. For these tests, we used the standard, single-use Hamilton pH sensor to measure the pH value at the standard pH sensor location, at the top of the fixed bed. We filled the vessel with media simulant (6.4 g/L sodium chloride, 2.0 g/L sodium bicarbonate, 1.0 g/L Pluronic F68) and controlled the temperature at 37°C using the iCELLis™ 50 heating mat and standard control settings. We adjusted the vessel volume to the desired falling film height, and selected the agitation set point to create the desired linear speed. Three falling film heights were characterized: 2, 6, and 10 cm. At each falling film height, we measured the mixing time at linear speeds of 0.5, 1.0, 1.5, and 2.0 cm/s. Each condition was repeated three times.

For every experiment, we adjusted the pH of the media to a stable pH value between 6.95 and 7.05 with HCl. Next, we manually added to the vessel a single bolus of 5 M NaOH solution, at 0.2% of the volume in the vessel, through the needleless swab-able valve on the sampling line. The mixing time for each experiment was defined as the time required for the measured pH value to reach 95% of the change towards the final stable pH value from the moment of base addition. We calculated the mixing time separately for each sensor and used the average of both as the experiment result. Trending of the pH measurements in the mPath™ Link software was used to perform these calculations. Additionally, we subtracted the probe response time (21.7 s) from each test result to determine the true mixing time. Finally, we removed excess media from the vessel to achieve the desired falling film for the next experiment.

Fig 6. Average mixing times for 2, 6, and 10 cm falling film heights on the 33 m² fixed bed at different linear speeds. Error bars correspond to the standard deviation.

The results show that the mixing times decreased as the linear speed increased from 0.5 to 2 cm/s (Fig 6). The rate of change decreased as the linear speed increased at each falling film height, and started to level off between 1.5 and 2.0 cm/s.

The mixing times at each linear speed were very similar at each falling film height and became more similar as the linear speed increased. This result suggests that the falling film height had a greater impact on mixing time at lower linear speeds, where the taller the falling film (and less total volume in the reactor), the shorter the mixing time, particularly at 0.5 and 1.0 cm/s. This difference is not observed at the higher linear speeds of 1.5 and 2.0 cm/s, suggesting linear speed becomes the more dominant variable in determining mixing time at higher linear speeds.

Then, we performed a second series of tests to assess the impact of the pH sensor and base addition location on the mixing time. This was done to mimic the mixing speed of other elements added to the bioreactor like inoculum, transfection, or glucose feed. These tests were all conducted at the 10 cm falling film height and at identical linear speed, but by varying the location of base addition and pH measurement (Fig 7):

• Test a—we added the base through the sample line to the top of the fixed bed, as described previously.
• Test b—we added the base to the media in the outer chamber of the vessel at the bottom of the falling film. This is close to where the base, and other addition lines are located during a standard cell culture.
• Tests c and d—we varied the location of the pH sensor at the top of the fixed bed to be either directly next to the point of base addition, or on the opposite side of the vessel.

Fig 7. The pH sensor and base addition configurations. The location of the pH sensor is indicated by the blue probe symbol, while the base addition location is indicated by the green teardrop symbol. These icons are superimposed on a schematic of the iCELLis™ 50 vessel. This schematic includes a representation of the pH probe as it appears in the standard vessel configuration. The four configurations (from left to right) were: (A) standard location of both pH sensor and base addition, (B) standard pH sensor location, new base addition location, (C) new pH sensor location, standard base addition location, and (D) new pH sensor and base addition location.

Figure 8 displays the average mixing time after varying the location of the pH sensor and base addition. We observed most variation in mixing time between the different configurations at the lower linear speeds, while the variation decreased as the linear speed increased.

Fig 8. Average mixing times for four different experimental setups (as described in Fig 7), at 10 cm falling film height. Error bars correspond to the standard deviation calculated from three replicates.

The data displayed in Figure 6 and 8 indicate that mixing times were very similar despite different vessel configurations. Most importantly, this data suggests that all additions—including base, cells, media feeds, or transfection complexes—can be dispersed quickly and evenly, creating a homogenous and stable environment throughout the reactor within a biologically reasonable amount of time.

Determining the k_La value (volumetric mass transfer coefficient) is important for bioreactors because it quantitates the efficiency of oxygen transfer from the gas phase to the liquid phase. Since oxygen has low solubility in culture media, the rate at which the media can be oxygenated is determined by the gas-liquid interfacial area. This plays a significant role in the maximum cell density achievable.

Therefore, we performed k_La experiments at 0, 2, 6, and 10 cm falling film heights and at agitation speeds of 200, 250, 300, and 350 rpm. Because we performed this experiment before the linear speed characterization was completed, the corresponding linear speeds were retrospectively calculated after the linear speed characterization was finalized. We tested each experimental condition in triplicate.

For all k_La experiments, we installed a 33 m² iCELLis™ 50 vessel on a standard iCELLis™ 50 control system. To calibrate the standard, single-use dissolved oxygen (DO) sensors in the vessel we defined 100% as the DO at full air saturation. Then, we used these sensors to measure the DO values at the two standard locations in the vessel, both positioned at the top of the fixed bed. We filled the vessel with media simulant (6.4 g/L sodium chloride, 2.0 g/L sodium bicarbonate, 1.0 g/L Pluronic F68) and controlled the temperature at 37°C using the iCELLis™ 50 heating mat and standard control settings. Agitation and gas flow rates were controlled via a manually specified set point in the control software. We used the live trending feature in the control software to trend the DO process value and to export the time-stamped DO measurements for analysis.

The experiments were carried out using the gassing out method for determination of k_La. Before each experiment, we stripped DO from the agitating media simulant by flowing nitrogen (N₂) gas into the headspace of the vessel until the DO levels were between 0% and 10%. The N₂ flow was then stopped, and air flow was turned on at a flow rate of 300 mL/min until the DO levels reached 85% to 100%. We repeated this process for each test condition and each technical replicate.

The DO process value was recorded, starting at the point at which the air flow was initiated. Once the DO reached 20%, we measured the time taken for the DO to increase to 80%. Therefore, all k_La values reported were k_La_20-80, defined as the oxygen transfer coefficient measured from 20% to 80% DO.

k_La was calculated using the following equation:

Table 3. Definition of terms for the k_La equation

Fig 9. Calculated k_La values for a 33 m² iCELLis™ 50 vessel as a function of linear speed and falling film height. Error bars correspond to the standard deviation of three replicates.

The results show that the average k_La values ranged from 0.5 to 3.0 h^-1, depending on the conditions tested (Fig 9). In the iCELLis™ 50 vessel, gas exchange between the culture medium and the headspace occurs predominantly through the falling film. These results indicate that the falling film height is one of the important factors impacting k_La, as the k_La values at 0 cm falling film were lower than the k_La values obtained at other falling film heights.

A second factor impacting k_La is the linear speed, as it determines the turnover rate through the falling film and therefore the mixing of dissolved gases. The results indicate that k_La increased moderately as the linear speed increased, as expected (Fig 9).

CO₂ stripping

The CO₂ evacuation capabilities of the system are important for maintaining optimal culture conditions and process consistency. CO₂ accumulation can be toxic to the cells, and CO₂ removal is important to control media pH without adding substances such as alkali solutions. Therefore, characterizing CO₂ stripping capabilities ensures that CO₂ removal is predictable and reproducible across batches and scales.

Unlike oxygen, CO₂ has a high solubility in culture medium. Therefore, the rate of CO₂ removal is not as strongly dependent on the gas-liquid surface area, but rather the volume of inert gas in the headspace, which is a constant for all iCELLis™ bioreactors. Consequently, CO₂ removal is less dependent on factors like the falling film height and not dependent on vessel surface area. As a result, we evaluated fewer combinations of these parameters in this testing.

To conduct this testing, we removed the pH probe from its respective port on a 33 m² vessel and replaced it with a Hamilton CO₂ sensor of the same length and diameter. This sensor was connected to the Hamilton ArcAir software for CO₂ data logging, providing CO₂ levels in parts per million (ppm). We added media simulant (6.4 g/L sodium chloride, 2.0 g/L sodium bicarbonate, 1.0 g/L Pluronic F68) to the vessel at a working volume sufficient to achieve a 6 cm falling film. Air flow was set to 300 mL/min and agitation turned on. This condition was defined as the baseline 0% dissolved CO₂ (DCO₂) level, by convention. Next, we turned off the air flow, and we set the CO₂ gas flow to 150 mL/min until the DO reading dropped to 0% and CO₂ measurement plateaued. At this point, by convention, DCO₂ was defined as 100%. We turned off the CO₂ gas flow, and set the air flow to 300 mL/min until DCO₂ dropped and stabilized at 0%. The CO₂ stripping rate was calculated between the time required for the DCO₂ to drop from 80% to 20%. We repeated this process for agitation speeds of 200, 250, 300, and 350 rpm, and tested each condition in triplicate. The same test at 224 and 271 rpm was also conducted in three replicates at a 10 cm falling film height.

Fig 10. CO₂ stripping results at 6 cm falling film height and 10 cm falling film height on the 33 m² fixed bed. Error bars show the standard deviation.

These results indicate that the CO₂ stripping rate is sufficient to meet the needs of a high-density cell culture (Fig 10). It has been shown that a stripping rate of 6 pmol/d is required per cell (1). For a culture of approximately 10 million cells/mL (or 200 000 cells/m²), this corresponds to a stripping rate of 2 mmol/L/h. The iCELLis™ 50 bioreactor exceeds that requirement.

Further, it appears that the CO₂ stripping rate has a minor dependency on the falling film height, and no dependency on the linear speed. This observation is consistent with the mechanics of CO₂ stripping observed in stirred-tank bioreactors.

Maintaining precise control over temperature, pH, and DO is critically important in bioreactors because these parameters directly affect the health, growth, metabolism, and productivity of the cells, as well as the quality and yield of the final product. Furthermore, tightly controlling these parameters at optimized values promotes consistency from batch to batch and across scales.

Process control is often achieved through the use of a closed control loop. A control loop consists of several components:

1. Sensor/measurement: Instrument that measures the current state of the process variable (e.g., a temperature probe).
2. Controller: The "brain" of the loop. It compares the measured value (the process variable) to the desired value (the set point). Using proportional, integral, and derivative control (PID), it calculates an appropriate correction signal to minimize the difference.
3. Actuator: Instrument that directly influences the process based on the output of the controller (e.g., a heater mat).

To implement this control strategy successfully, each control loop requires a series of PID values and actuator limits to maintain tight control. These values are dependent on the process, however, here we characterized a default set of values with a high probability of working reasonably well for a common process in the iCELLis™ 50 bioreactor. These values are provided in the following sections.

Please be aware that these were considered optimal at the time of publishing but may have changed since. Contact a Cytiva representative for any updates.

Temperature control

In the iCELLis™ 50 bioreactor, good temperature control consists of the following behaviors:

• Temperature is maintained at the set point ± 0.1°C while media is recirculating to and from the vessel
• Temperature does not overshoot the set point by more than 0.5°C during an extended heat-up phase
• If the temperature set point is changed, the new set point can be achieved rapidly.

All these behaviors were achieved using the set of temperature control loop parameters provided in Table 4 and the media preheater control loop parameters provided in Table 5.

Table 4. Optimized parameters of the temperature control loop

Table 5. Optimized parameters of the media preheater control loop

Using the parameters in Table 4 and Table 5, we performed a series of simple verification experiments. First, to show temperature control stability at a set point during recirculation, we filled an iCELLis™ 50 vessel with 8 L of water and heated the vessel to 37°C. This represented the worst-case cell culture volume. Then, we initiated recirculation from a tote containing room temperature water at 40 mL/min. We initiated the weight control on bioreactor feed out with a set point of 8 kg and the temperature control with a set point of 37°C. The temperature was monitored for 60 min while agitating. During this time, after an initial overshoot to 37.2°C, the measured temperature process value did not deviate by more than 0.1°C from the 37°C set point (Fig 11).

Fig 11. Temperature trend line resulting from an experiment designed to illustrate that the temperature can be controlled at its set point during media recirculation, with less than a ± 0.1°C temperature oscillation.

This performance was confirmed in a subsequent cell culture batch. For more details on this batch, including the specific methodology, please consult our scalability application note (2). For this batch, we monitored the temperature over the course of the first 4 d of the batch (Fig 12). Initially, the media temperature dropped by 0.7°C after the recirculation loop was initiated and the control loop responded to the addition of colder media. After about 3 h, the media temperature stabilized, and the measured temperature process value did not deviate by more than 0.1°C from the 37°C set point.

Fig 12. Temperature trend line from a cell culture batch illustrating that the temperature can be controlled at its set point during media recirculation, with less than a ± 0.1°C temperature oscillation.

Next, to demonstrate well-controlled heating, we filled an iCELLis™ 50 vessel with 8.5 L of chilled water. Once the temperature sensors were submerged during filling, we initiated the agitation and temperature control at a set point of 37°C. It required about 2 h and 25 min for the vessel contents to heat from 13°C to the 37°C set point (Fig 13). The temperature reached a maximum value of 37.3°C, which was within the target of less than 0.5°C overshoot.

Fig 13. Temperature trend line resulting from an experiment designed to illustrate that vessel contents can be heated but do not overshoot 37°C by more than 0.5 °C.

This performance was again confirmed in a cell culture batch. For more details on this batch, including the specific methodology, please consult this application note. In this batch, we monitored the temperature process value approximately 2 d before cell inoculation (Fig 14), after the vessel was initially filled with media. It took about 3 h for the media to warm from room temperature to the set point of 37°C. During this time, the maximum temperature recorded was 37.1°C. This is 0.1°C from the 37°C set point, which was within the target of less than 0.5°C overshoot.

Fig 14. Temperature trend line from a cell culture batch illustrates that the temperature can be controlled during a prolonged heating period, with less than a 0.5°C overshoot.

Finally, to show the temperature control response to step changes, we made 1°C changes to the temperature set point and measured the response time and accuracy of the temperature control loop. When the temperature set point increased by 1°C, it took about 25 min for the heater mat to achieve the new set point (Fig 15). When the set point was decreased by 1°C, it took about 35 min for the new set point to be achieved by passive cooling.

During the cooling, we often observed a setpoint undershoot, sometimes by as much as 0.6°C. This was likely due to the fact that the optimized PID settings were prioritized on heating instead of on cooling. Therefore, for processes that implement a temperature shift to lower temperature set points you can consider further PID tuning if this undershoot is undesirable.

Fig 15. Temperature trend line resulting from an experiment designed to illustrate that the temperature can quickly be adjusted to a new set point.

pH control

In the iCELLis™ 50 bioreactor, good pH control consists of the following behaviors:

• pH is maintained at its setpoint with oscillations ≤ 0.05.
• If the pH point is changed, the new set point can be achieved rapidly.

All these behaviors were achieved using the set of pH control loop parameters provided in Table 6.

Table 6. Optimized parameters of the pH control loop

Similarly to the temperature testing, we performed a series of simple verification experiments using the parameters in Table 6. First, to show pH control stability during an external downward impact on pH, we filled an iCELLis™ 50 vessel with 9 L of media simulant. We started agitation to create a 1 cm/s linear speed and set the temperature control to 37°C. The base manifold was prepared with an alkali solution of 0.1 M sodium bicarbonate and connected to the vessel via the base pump, for automated base addition. Once the system stabilized, two-sided pH control was turned on with a setpoint of 7.0 and a deadband of 0.15. We activated manually the CO₂ gas flow at 150 mL/min to simulate the generation of CO₂ by a growing cell culture.

Fig 16. Trend line of upward pH control by automated base addition to illustrate that the pH can be controlled at its set point with less than a ± 0.05 units of oscillation. For this experiment, the pH set point was 7.0 with a 0.15 deadband. CO₂ was set to a constant flow of 150 mL/min. The base pump was under the pH control loop and the system automatically added 0.1 M NaHCO₃ to remain above the lower deadband limit of 6.85.

Initially, the pH dropped from 7.15 to 6.85, at which point the pH process value reached the lower threshold of the dead band (Fig 16). At this point, the base pump was activated by the controller, maintaining the pH at 6.85 with oscillations of less than 0.01 pH units. This achieved the requirement of oscillations smaller than ± 0.05 pH units.

Next, to show pH control stability during an external upward impact on pH, we repeated this test under similar conditions. For this test, we turned on the base pump manually at a flow rate of 1 mL/min, and the CO₂ mass flow controller was placed under pH control to maintain pH within the deadband. The system was initially equilibrated with active pH control at a pH 8 set point and a temperature set point of 37°C.

Then, while the base pump was running, we changed the pH set point to 7.0. As a response, the pH control loop was activated and the CO₂ flow was started, which brought the pH down to 7.15, the upper threshold of the deadband (Fig 17). At this point, the CO₂ flow was activated periodically by the controller to maintain the pH below the upper deadband limit of 7.15 with oscillations of less than 0.01 pH units. The total response time to achieve the new set point was 20 min, which is suitable for most applications.

Fig 17. Trend line of downward pH control by automated CO₂ addition to illustrate that the pH can be controlled at its set point with less than a ± 0.05 units of oscillation. For this experiment, the pH set point was 7.0 with a 0.15 deadband. The base pump was set to manual mode at 1 mL/min with CO₂ under the pH control loop.

Dissolved oxygen (DO) control

In the iCELLis™ 50 bioreactor, good DO control consists of the following behaviors:

• DO is maintained at its set point with oscillations ≤ 5%.
• If the DO point is changed, the new set point can be achieved rapidly.

All these behaviors were achieved using the set of pH control loop parameters provided in Table 7.

Table 7. Optimized parameters of the DO control loop

Using the parameters in Table 7, we performed a series of simple verification experiments. First, to show DO control stability during an external downward impact on DO, we filled an iCELLis™ 50 vessel with 9 L of media simulant. We started agitation to create a 1 cm/s linear speed and set the temperature control to 37°C. Once the system stabilized, one-sided DO control was turned on with a set point of 40% (no deadband). We initiated the nitrogen gas flow manually at a flow rate of 150 mL/min to simulate the effects of oxygen depletion from a growing cell culture. At this point, the oxygen gas flow was activated by the controller, maintaining the DO at 40% with oscillations less than 5%, achieving the requirement of oscillations smaller than ± 5% (Fig 18).

Fig 18. Trend line of upward DO control by automated oxygen gas addition to illustrate that the DO can be controlled at its set point with less than 5% oscillation. For this experiment, the DO set point was 40% (no deadband). Nitrogen gas was set to a constant flow of 150 mL/min. Oxygen gas was under the DO control loop.

To show two-sided DO control stability, we filled an iCELLis™ 50 vessel to a 6 cm falling film with media simulant and agitated. Once the system stabilized, oxygen and nitrogen mass flow controllers were placed under DO control and DO control was turned on with a set point of 50% (no deadband). The control loop was able to maintain the DO process value at 50% ± 4%, meeting the requirement of ± 5% (Fig 19).

Next, to show the DO control response to step changes, we adjusted the DO set point to 40%. It took about 15 min for the nitrogen gas flow to achieve this new set point. Once the new set point was achieved, oscillations around the set point were only ± 2%. Finally, the DO set point was adjusted back to 50%. It took 8 min for oxygen flow to achieve this new set point, and once the new set point was achieved, oscillations around the setpoint were ± 4%. We repeated these set point changes a second and third time, yielding similar results.

Fig 19. DO trend line resulting from an experiment designed to illustrate the DO control in split mode with both positive (oxygen) and negative (nitrogen) control with less than a ± 5% oscillation.

This performance was again confirmed in a cell culture batch. For more details on this batch, including the specific methodology, please consult [insert reference of scalability application note]. In this batch we monitored the DO process value and oxygen gas flow rate over the duration of the culture (Fig 20). We operated the process with a DO set point of 50%, no deadband, and single-sided (O₂ only) control.

Initially, the DO was measured above the set point, as the nitrogen gas was intentionally not used for control. During the first 2.5 d, cell growth gradually depleted oxygen in the media until the DO process value dropped to the 50% set point. At this point, the controller initiated the oxygen gas flow, maintaining the DO at the 50% set point with oscillations of less than 1% for the remainder of the batch. The DO control loop was tuned to avoid rapid changes in oxygen flow rate and therefore maintain a stable DO. Also important to note is that the oxygen demand did not exceed the capabilities of the system. This demonstrates that the iCELLis™ 50 bioreactor k_La performance and maximum oxygen flow rate of 700 mL/min allowed the system to provide a stable environment for high-density cell cultures. On day five, the vessel was drained for transfection and media exchange, which accounts for the “break” in the DO measurement.

Fig 20. DO trend line from a cell culture batch illustrates that DO can be controlled to a set point of 50% with less than 1% oscillations.

Volume and weight control

In the iCELLis™ 50 bioreactor, good weight control of the media volume inside the vessel consists of the following behaviors:

• Bioreactor volume is maintained at its set point with less than ± 0.1 L oscillations.
• If the bioreactor weight set point is changed, the new set point can be achieved rapidly.

All these behaviors were achieved using the set of optimized weight control loop parameters provided in Table 8. The weight control loop parameters were optimized for one-sided weight control, meaning that one pump (usually Feed In) is automatically adjusted by the control loop, while the other pump (usually feed out) is under manual control.

Table 8. Optimized parameters of the weight control loop

Similar to the temperature, pH, and DO control experiments, we performed an experiment to verify stable volume control. We filled an iCELLis™ 50 vessel with 8.5 L of media simulant. This filling step was conducted manually, not using the bioreactor weight control loop, as the loop was tuned for steady-state control during recirculation and not for filling and draining steps. When the control loop was used for filling, an overshoot occurred (data not shown).

Next, we turned on the agitation to create a 1 cm/s linear speed and the temperature control was set to 37°C. Once the system stabilized, we initiated recirculation by manually turning on the feed out pump, and one-sided weight control was initiated with a set point of 8.5 kg. At this point, the feed in pump was activated by the controller, maintaining the vessel volume at 8.5 L, with oscillations of less than 0.01 L, achieving the requirement of oscillations less than ± 0.1 L (Fig 21).

Fig 21. Bioreactor weight trend line (a surrogate indication of vessel volume) resulting from an experiment designed to illustrate stable single-sided volume control with oscillations less than ± 0.1 L.

This performance was again confirmed in a cell culture batch. In this batch, we monitored the vessel weight (vessel volume) over the duration of the culture (Fig 22). The vessel volume set point was varied, and single-sided (feed in pump only) control was used to maintain the set point.

Initially, the vessel weight was measured above the set point, as the recirculation loop was off during cell inoculation. However, once recirculation and weight control were initiated shortly after inoculation, the control loop established stable weight control within a few minutes. During the process, three weight set point changes were initiated, and in each case, the controller established stable control within a few minutes, with less than 0.01 L oscillations. On day five, the vessel was drained for transfection and media exchange, which accounts for the “break” in the measurements.

Fig 22. Bioreactor weight (volume) trend line from a cell culture batch illustrates that the vessel volume can respond rapidly to a new set point, resulting in stable control with less than 0.01 L oscillations.

1. Goudar CT, Piret JM, Konstantinov KB. Estimating cell specific oxygen uptake and carbon dioxide production rates for mammalian cells in perfusion culture. Biotechnol Prog. 2011;27(5):1347-1357. doi:10.1002/btpr.646
2. Application note: Scale-up of AAV production in iCELLis™ fixed-bed bioreactors