Scale-Up and Comparison Studies Evaluating Disposable Bioreactors and Probes

Abstract

Acceleron Pharma has based its manufacturing platform on disposable technology for cell culture, harvest, and purification. Cell culture was originally performed in GE WAVE bioreactors; however, as the demand for material increased, the need for a more robust, more tightly controlled system became essential. Acceleron chose to maintain disposability and purchased the HyClone stirred-tank single-use bioreactor (SUB) for process development with the goal of implementing the SUB into a new manufacturing facility based on disposable technology. Four 2-L Applikon and two 5-L Sartorius Stedim glass vessel stirred-tank reactors were used for small scale development, process optimization, and comparison between the traditional bioreactor and the disposable technology. Temperature, pH, and dissolved oxygen (DO) were controlled, and nutrients, metabolites, and gases were monitored off-line using a Nova BioProfile Flex. The 50-L and 250-L SUBs were used for scale-up and pilot runs with the optimized process (a 14-day fed-batch culture). Data on cell density, fractional viability, and protein concentration were collected and compared across the scales and types of reactors. The oxygen mass transfer coefficient (k_La) was also compared from vessel to vessel to aid in scaling up from the glass vessels to the SUBs. More recently, a 1,000-L SUB, integrated with a Finesse controller, was used for engineering runs. In addition to disposable vessel technology, studies were performed that compared fiber optic DO probes and sensors (which have the option of being disposable), to autoclavable polarographic probes. The preliminary studies indicate that the fiber optic technologies are interchangeable with the conventional polarographic probes. Based on this work, disposable technology was determined to be a desirable option for biopharmaceutical manufacturing at Acceleron Pharma.

Acceleron Pharma is a privately held company based in Cambridge, MA, developing novel bio-therapeutics focused on the growth and differentiation factor (GDF) family of proteins. To minimize capital and validation costs, Acceleron implemented disposable equipment—mainly disposable bioreactors and harvest and purification equipment—into the development and manufacturing platform. Originally, cell culture was performed in WAVE bioreactors (GE Healthcare). However, because of limitations on control and scalability, a more robust system became necessary. The stirred-tank single-use bioreactor (SUB) by HyClone (Thermo Fisher) provided a stirred-tank design with the ability to control, monitor, and log data throughout cell culture when integrated with a control system. Using glass vessel reactors as a comparison, the SUBs were evaluated in scale-up studies to determine the applicability of the disposable design.

Acceleron Pharma

In addition, to simplify the manufacturing process and create a completely disposable environment for cell culture, disposable dissolved oxygen (DO) probes and the fiber optic technology associated with them were evaluated. The fiber optic probes were compared directly with the traditional polarographic DO probes.

Materials and Methods

Cell Culture and Bioreactor Operation

The cells used for protein production were Chinese hamster ovary (CHO) cells, and all data presented are from one protein, product 1. All cell culture was performed in custom media and feed developed by Irvine Scientific. Atypical seed train was used for manufacturing runs. The original seed train consisted of WAVE reactors up to the 100-L and 500-L working volume scale for manufacturing. The current seed train in the new facility uses WAVE bioreactors for scale-up and SUBs for production vessels at the 50-L, 250-L, and 1,000-L scale. Development runs were performed in four 2-L Applikon glass vessel reactors with ADI 1030 controllers and two 5-L Sartorius Stedim bioreactors with A+ and B+ controllers. Scaled-up runs were performed in 50-, 250-, and 1,000-L SUBs. The 50-L SUB was integrated with an Applikon I controller, and the 250- and 1,000-L SUBs were integrated with a Finesse TruLogic RDPD controller. The parameters that were logged and optimized included feeding strategy (i.e., schedule, volumes), pH, temperature, DO, and agitation set points, and flow rates. Partial pressure of carbon dioxide (pCO₂) levels also were monitored and controlled. Viable cell density and viability were measured using a Cedex automatic cell counter (Innovatis) or a BioProfile Flex (Nova Biomedical). Protein production was determined using a Protein A assay on an Agilent 1100 series high performance liquid chromatography (HPLC) instrument. Some nutrients, metabolites, and electrolytes (L-glutamine, glucose, glutamate, lactate, ammonium, and sodium, calcium, and potassium ions) were monitored offline using the BioProfile Flex.

Multiple scale-up parameters were analyzed to transfer the optimized process from the glass vessels to the SUBs. Initial values for gas flow rates were calculated using vessel volumes per minute. For agitation speed, tip speed, power per volume, and torque, calculations were performed and evaluated. Power per volume was chosen to scale up the agitation speed. After the flow rates and agitation speed had been approximated, oxygen mass transfer coefficient (k_La) studies were performed in the 50-LSUB to determine the efficiency of the calculated values compared with small-scale studies.

Fiber Optic Autoclavable and Disposable Probes

Different configurations of fiber optic technology were evaluated. Autoclavable oxygen probes for in-line measurement from PreSens and Finesse (TruDO Optical) were studied in the 5-L and 2-L glass vessels, respectively. Measurements were taken every minute and compared directly to the Mettler-Toledo (InPro6800) polarographic probes integrated with the controllers. Preliminary studies also were performed in shaker flasks using the PreSens non-invasive oxygen sensors based on the same fiber optic technology. The values from these disposable probes were compared to offline measurements of the BioProfile Flex. Lastly, the disposable option available from Finesse (TruFluor DO) was evaluated in the 50-L SUB and compared directly with autoclavable polarographic DO probes. The TruFluor DO is available in a sleeve that is manufactured with HyClone's bioprocess container (bag) for the SUBs. The sleeve contains the disposable sensor, and inserted into the sleeve is a non-invasive optical reader connected to a transmitter.

Results

Process Optimization

The pH set point was optimized for the cell culture process. Set points in 0.2 increments were tested and the one that resulted in the highest peak viable cell density and final protein concentration was chosen. For product 1, the natural tendency of the cell line is an increase in pH toward the end of the culture. To maintain a lower set point, more CO₂ must be added to the culture, which increases the pCO₂ in the culture. High pCO₂ levels have been shown to negatively affect the quality of product 1. To decrease the amount of pCO₂ in the culture, experiments were run allowing the pH to drift after day nine. The pH drift was performed by removing the base control and increasing the set point, to ensure that the pH did not drift too high. Viable cell density, viability, and protein concentration were monitored throughout the culture and compared to runs in which the pH was maintained at the set point. Figure 1 shows the normalized pCO₂ values over time in culture for 2- and 5-L reactors with and without a pH drift. Initially, the pH of the media is higher and must be brought down to the set point, accounting for higher pCO₂ values in the vessels in the first few days. Around day nine, as the pCO₂ increases again as a result of the natural tendency of the cells, the cultures with the drift show a decrease in pCO₂. Cultures maintained at the set point show increased pCO₂ over the same time. Table 1 shows average normalized peak viable cell density values, final viabilities, and normalized final protein concentrations for both conditions. Those values do not vary significantly, which indicates that letting the pH drift does not have a negative effect on the cell density or protein concentration, although it does decrease the pCO₂ in the culture.

Table 1. Data for a pH drift or no pH drift, and for a temperature shift mid-culture or no temperature shift, averaged over six runs of 2-L and 5-L glass vessel reactors for each condition.

Another parameter that was optimized was temperature. Changing the temperature set point to a lower value mid-culture has been shown to sustain higher cell viabilities.¹ Higher cell viabilities make harvest easier because there is less cell debris. Experiments with a temperature shift mid-culture were performed and compared to runs with the temperature maintained at the set point. Figure 2 shows fractional cell viabilities over time in culture for runs with and without a temperature shift. The viabilities trend similarly through the exponential growth phase at the start of the culture. As the cell density achieves its maximum value, however, the viability begins to drop for the experiments with temperature maintained at the set point. For those that received a temperature shift, the viability is sustained at higher values through day 14, most likely because of slower cell growth. Table 1 displays average values for normalized peak viable cell density, final cell viability, and normalized final protein concentration for both sets of runs. The normalized peak viable cell density and normalized final protein concentration values did not vary significantly. However, the final viabilities were significantly different (p <0.01), showing that a temperature shift mid-culture is preferred.

Figure 1. Normalized partial pressure of carbon dioxide (pCO2) versus time in culture for 2-L and 5-L reactors with and without a pH drift after day nine

In addition to process optimization, scale-up studies were performed to transfer from the 2- and 5-L scale up to the 1,000-L scale. The oxygen mass transfer coefficient (k_La) is an important value in scale-up studies and can vary depending on sparger type (i.e., pinhole or micro). By adjusting agitation speeds and air flow rates, k_La can be compared across scales. Studies were performed at the 2-L, 5-L, and 50-L scales. Optimal flow rates and agitation speeds were determined by achieving similar k_La values across scales.

Figure 2. Fractional viability versus time in culture for 2-L and 5-L reactors with and without a temperature shift mid-culture

Taking into account all parameters optimized in the 2- and 5-L vessels, the final process to transfer and scale-up to the SUBs was a 14 day fed-batch culture with a day 2 feed of 12% of the working volume, a temperature shift mid-culture, and a pH drift after day 9 when necessary. The SUBs have a dual sparger option (open pipe or frit). The open pipe sparger provides air to the culture while stripping excess CO₂. If excess CO₂ is stripped, the pH drift may not be necessary because the pCO₂ levels will be decreased.

Process Transfer

The final process was transferred to the 50-L SUB integrated with an Applikon I controller. Initially, the culture was run pre-optimization without the temperature shift, pH drift, or the final feed schedule. The optimized process was run and resulted in an increase in both the normalized protein concentration (Figure 3) and the fractional viability over the pre-optimization run (Figure 4).

Figure 3. Normalized protein concentration versus time in culture for 50-L stirred-tank single-use bioractor runs pre- and post-optimization

The 250-L SUB was integrated with the Finesse TruLogic RDPD controller. An engineering run was performed to test the SUB and the controller. The engineering run was performed without the temperature shift or final feed schedule. The fractional viability for this run is atypical, and the rapid drop seen in Figure 4 is likely a result of the fact that this was the pilot run at Acceleron. The optimized process was then run with the fractional viability and normalized protein concentration curves seen in Figures 3 and 4. The optimized process maintained higher viabilities and increased the normalized protein concentration. The pH drift was not necessary for the optimized run because the dual sparger option was used. This run was performed in the new manufacturing facility.

Figure 4. Fractional viability versus time in culture for the 50-L and 250-L stirred tank single-use bioractor runs pre- and post-optimization

Finally, the process was transferred to the 1,000-L SUB. The optimized process was compared across the three scales. Figures 5 and 6 show the normalized viable cell density, the cell viability, and the normalized protein concentration. The normalized viable cell density and viability were similar across scales. However, the 50-LSUB shows slower growth, which has been observed consistently over multiple 50-L runs. The cause of this difference is still under investigation, but it has been hypothesized that it is related to the seed train. The 50-L bioreactor is seeded directly from a WAVE bag, whereas the 250- and 1,000-L bioreactors are seeded from other SUBs. For seeding the 50-L SUB, the inoculum cell culture has to remain longer in a less controlled vessel (WAVE) to reach the densities required for seeding compared to the inoculum culture required to seed the 250- and 1,000-L SUBs (seeded from 50- and 250-L SUBs, respectively).

Figure 5. Normalized viable cell concentration (VCD, closed symbols) and fractional viability (FV, open symbols) versus time in culture for the 50-L, 250-L, and 1,000-L stirred-tank single-used bioreactors (SUBs) using the optimized process

On the other hand, the normalized viable cell density for the 1,000-L bioreactor appears higher than the data for the 50- and 250-L bioreactors. The cell counts for the 1,000-L reactor were performed using the Cedex, whereas cells in the 50- and 250-L SUBs were counted using the BioProfile Flex. The Cedex and the Flex consistently drift apart, and the Cedex tends to overestimate viable cell density toward the end of the culture. The normalized protein concentration is consistent across all scales (Figure 6).

Figure 6. Normalized protein concentration versus time in culture for 50-L, 250-L, and 1,000-L stirred-tank single-use bioreactors (SUBs) using the optimized process

Comparing Traditional and Fiber Optic Probes

The PreSens in-line fiber optic DO probe was compared directly with the polarographic DO probe integrated with the 5-L controller. When the data are overlaid, the trends for the two technologies are identical. The data from one run are shown in Figure 7A. This comparability has proven to be reproducible over many runs.

Figure 7. Comparison of disposable versus traditional dissolved oxygen (DO) probes. A). PreSens in-line fiber optic dissolved oxygen probe compared to the traditional polarographic probe. B). Finesse TruDO Optical DO probe compared to the traditional polarographic probe. C). Finesse TruFluor DO disposable DO probe compared to the traditional polarographic probe

Finesse's TruDO Optical uses technology similar to that of the PreSens probe. However, the trends between the traditional probe and the fiber optic probe do not overlay exactly. As shown in Figure 7B, data from the TruDO Optical show a 5–8% upward shift in the pO₂ level compared to the trend captured by the polarographic sensor. Based on discussions with Finesse, this shift is caused by the temperature at which the two different probes were calibrated. The TruDO Optical comes with a factory calibration at room temperature, whereas the polarographic probe is calibrated in-house at 37 °C. In addition, the TruDO Optical data appear to be more noisy. This may come from the number of readings (reading are taken every minute) compared with the number of readings taken from the traditional probe (every three minutes). Additional studies must be run to explain or correct for these differences.

Finally, the TruFluor DO was evaluated in the 50-L SUB. Figure 7C shows the pO₂ levels over time in culture for both the disposable TruFluor DO and the traditional probe. The trends are identical between the two probes. Further studies are being performed to confirm direct comparability. These preliminary data show that disposable probe options available are interchangeable with the traditional technology.

Discussion

An optimized and transferable process was developed in the 2- and 5-L glass vessels by monitoring, controlling, and adjusting process parameters such as pH, temperature, and partial pressure of carbon dioxide (pCO₂). The final process, a 14 day fed-batch culture with a day 2 feed of 12% of the working volume, a temperature shift mid-culture, and a pH drift after day 9 when necessary, was transferred to 50-, 250- and 1,000-L stirred-tank single-use bioreactors (SUBs). The performance of the optimized process was comparable across all scales, determining that the HyClone SUB was a suitable disposable bioreactor for manufacturing at Acceleron.

In addition to implementing disposable bioreactors, fiber optic dissolved oxygen probes and sensors were evaluated. Autoclavable probes from both PreSens and Finesse were compared to the traditional polarographic dissolved oxygen probes, and the disposable sensor from Finesse was also evaluated. Although additional studies need to be performed, initial results showed that the fiber optic technology used in the disposable sensors was comparable to the technology used in the traditional sensors and was feasible for implementation in Acceleron's biopharmaceutical manufacturing process.

Alissa Fernald is a cell culture engineer, Anna Pisania is a senior cell culture engineer, Esam Abdelgadir is a manufacturing associate 2, Jesse Milling is a manufacturing associate 2, Tod Marvell is a cell culture manager, manufacturing, and Bob Steininger is the senior vice president of manufacturing, all at Acceleron Pharma, Cambridge, MA, 617.649.9326,afernald@acceleronpharma.com

REFERENCES

1. Trummer E, Fauland K, Seidinger S, Schriebl K, Lattenmayer C, Kunert R, et al. Process parameter shifting: Part I. effect of DOT, pH, and temperature on the performance of Epo-Fc expressing CHO cells cultivated in controlled batch bioreactors. Biotechnol Bioeng. 2006 Aug 20;94(6):1033–44.