IVD Technology Magazine
IVDT Article Index

Increasing antibody production with hollow-fiber bioreactors

Yuan Shi, Julie Ploof, and Alberto Correia

Using a conditioned growth medium can give hollow-fiber bioreactors a head start toward higher yields at significantly reduced costs.

The use of cell culture hollow-fiber bioreactors is well established for the production of monoclonal antibodies, recombinant protein products, viruses and viral antigens, and viable cell mass.^1—6 In such systems, the bulk of the culture medium is separated from the cell mass by means of hollow-fiber walls, allowing production of high-density cultures of viable cells (> 10⁸ cells/ml) in the extracapillary space (ECS).⁷ This results in highly concentrated antibody and protein harvests.⁸ Antibody harvests commonly reach 1 to 3 mg/ml in the harvested supernatant, but yields in excess of 17 mg/ml have been reported.^9—11 Yields such as these, together with their simple and relatively inexpensive operation, make hollow-fiber bioreactors an attractive means of producing antibodies from mammalian cells.

Electron micrographs of hollow fibers and cells in an inoculated bioreactor, showing growth of monoclonal antibodies in the extracapillary space. PHOTO COURTESY UNISYN

Antibody production using research-scale hollow-fiber bioreactors is generally not as efficient as using mice; quantities of antibody from a few to 100 milligrams are more often produced in shorter periods of time by mice.^12,13 However, it is possible that hollow-fiber bioreactors can be made to support more productive cell growth and antibody production by optimizing the formulations of culture media, the configuration of the hollow-fiber bioreactor, and operating protocols.^3,8,14

With regard to the development of culture media, it is well known that the growth of mammalian cells is principally regulated not by nutrient levels but by highly specialized small molecular proteins known as growth factors, which interact with cell-surface receptors with great specificity and high affinity.^15,16 Investigators have offered the notion that cells may themselves synthesize and secrete growth factors that diffuse back onto receptors on the surfaces of the cell-producing population, serving to stimulate proliferation.¹⁷ A number of such specific autocrine growth factors have been isolated and identified.^{18, 19} The use of conditioned medium (CM) containing such factors has been investigated as a method for accelerating the growth of murine and human hybridomas.²⁰ CM contains many metabolites, and their secreted autocrine factors can stimulate cell proliferation.

Unlike homogeneous suspension cultures, in which the medium surrounding cells contains an even concentration of nutrients and growth factors, hollow-fiber membranes provide a heterogeneous environment by separating the cells from the circulating medium into compartments of extra- and intracapillary spaces (ECS and ICS). Cells are grown in the ECS and are nourished by nutrients circulating in the ICS medium that readily diffuse across the hollow-fiber membrane. Metabolic waste products similarly diffuse away in the opposite direction and are diluted into the medium circulating in the ICS. However, other small molecular proteins can also pass through the hollow fibers into the much larger volume of circulating medium—including growth factors secreted by the cells or provided in serum supplemented in the ECS.¹⁴ This process can lead to a shortage of growth factors in the microenvironment surrounding the cells, with the amount of such proteins so diluted that they no longer contribute to regulating cell growth. To obtain optimal cell growth in membrane-based hollow-fiber bioreactors, it is therefore critical to maintain adequate amounts of growth factors, including autocrine growth factors, within the ECS.

To determine the effect of CM on the growth of cell cultures produced in hollow-fiber bioreactors, the authors conducted a series of experiments over the course of 30 days. The objective of these experiments was to investigate the effect of using medium conditioned by resident hybridoma cells on cell growth and antibody production in static culture and in hollow-fiber bioreactors.

This article describes the results of those studies on the use of CM in both the ICS and ECS, and its regulation of cell growth and antibody production. The regulatory action of CM was quantitatively examined for its ability to support cell growth and the antibody production of three hybridomas. The results demonstrate large differences in the productivity of cells with and without the use of CM. To demonstrate the dialyzing effect of small molecular proteins on cell growth and antibody production, this article also compares results from hollow-fiber bioreactors with 10 to 70 kDa molecular weight cut-offs (MWCOs).

Materials and Methods,

The materials and methods used in this set of experiments enabled the authors to investigate production of several types of antibodies, to distinguish static culture results from those of hollow-fiber bioreactor products, and to evaluate the effects of using membranes with differing dialysis rates.

Cell Lines. Three hybridomas were used in this study: 3F6 (ATCC HB 8512), 3C11 (ATCC HB 8511), and HFN 7.1 (ATCC CRL 1605). Hybridomas 3F6 and 3C11 secrete mouse IgM and IgG^₁ antibodies, respectively. Both antibodies react directly against bacterial cell wall peptidoglycans. Hybridoma HFN 7.1 secretes a mouse IgG^₁ antibody directed against human fibronectin.

Table I. Features of hollow-fiber bioreactors.

All cell lines were maintained in Dulbecco's Modified Eagle Medium (DMEM; Life Technologies, Gaithersburg, MD) containing 4.5 g/L glucose, 4 mM glutamine, and 10% volume-to-volume (V/V) supplemented fetal bovine serum (FBS; Hyclone, Logan, UT) as a complete medium.

To provide enough cells in log-phase growth for both the static culture and hollow-fiber bioreactor experiments, cells were expanded by standard static culture techniques in 175 cm² tissue culture flasks (Becton Dickinson; Lincoln Park, NJ). Cells were counted on a hemocytometer using trypan blue dye exclusion. All IgG_¹ and IgM antibody products of interest were quantified by radial immunodiffusion assay (RID; Binding Site, San Diego).

Static Culture Assays. Static suspension cultures were performed in standard 25 cm² tissue culture flasks using complete medium. For each cell line, 30 ml of early log-phase cells at a concentration of 1 * 10⁶ cells/ml with 90% viability was centrifuged at 400* G for 5 minutes in a tabletop centrifuge.

Following centrifugation, the clarified supernatant was transferred and filtered through a 0.2-µm filter to be used as CM for the experiments, and the cell pellet was resuspended in 30 ml fresh complete medium. The preparation of CM was quantitatively controlled by ensuring that 65% of total glucose and 50% of glutamine remained unconsumed, and that it contained low levels of waste products (e.g., < 5 mM lactate, and < 2 mM ammonia). The experimental media were prepared by mixing complete medium with CM at various percentages of 0%, 12.5%, 25.0%, and 50.0% (V/V). To every experimental flask was added 1 ml of resuspended cells, then 19 ml of each experimental medium. The final cell concentration was 0.5 * 10⁵ viable cells/ml in a total of 20 ml.

All flasks were incubated at 37°C in a CO^₂ incubator with 5% to 9% CO^₂ to maintain a pH between 7.0 and 7.4. During the course of the 5-day cell culture, one milliliter of sample was taken each day to measure the concentration and viability of the cells as well as their metabolic activity, including the rates of glucose use, lactate production, and antibody level. To minimize statistical deviation, each experimental condition was run in triplicate, and an average cell concentration was used to estimate growth rate.

Hollow-Fiber Cell Culture System. Evaluations of antibody production in hollow-fiber bioreactors with 0.14 m² fiber surface area were performed with and without CM. Thirty-day cell culture experiments were conducted for each hybridoma by injecting 2 * 10⁸ total cells (> 90% viability) into flushed hollow-fiber bioreactors contained within presterilized, disposable Micro Mouse systems (Unisyn, Hopkinton, MA). Table I lists most of the parameters that describe the configuration of the bioreactors. The bioreactor system shown schematically in Figure 1 was placed in CO_² incubators in order to maintain a constant temperature and pH. Oxygenation occurred while the ICS medium stream passed through thin-walled silicone tubing within the Micro Mouse system.

Figure 1. Schematic of a single hollow fiber in a bioreactor with flow path that includes an oxygenator, a medium reservoir, and a peristaltic pump for circulating ICS medium through the hollow fiber's lumen.

For the control run, media used in the ICS and ECS consisted of basal DMEM supplemented with 2.5% and 10% FBS, respectively. Spent ICS medium was gradually replaced by basal DMEM to maintain nutrients above certain levels, such as keeping glucose concentration greater than 2 mg/ml. The initial ICS medium used in the study consisted of 75% basal DMEM and 25% 0.2-µm—filtered CM. Spent medium was replaced by basal DMEM as described above. The ECS medium was 100% CM throughout the 30-day culture period. ICS medium was circulated through the inside of the fibers and back to the ICS medium reservoir at a flow rate of 100 ml/min. Beginning on the fourth day of the experiment, 10 ml of medium containing antibody secreted by cells was harvested from the ECS every other day. The harvest was accomplished using two 10-ml syringes. One empty syringe was used to collect the harvested medium while a second syringe containing 10 ml of fresh ECS medium was used to displace the harvested medium. To monitor glucose and lactate levels, samples from the ICS medium were taken daily using a YSI 2300 Stat Plus analyzer (YSI Inc.; Yellow Springs, Ohio). A blood gas analyzer (Ciba Corning Diagnostics Corp.; Medfield, MA) was employed to measure dissolved oxygen and carbon dioxide levels as well as pH.

The Dialysis Effect of Serum Proteins in Hollow Fibers. To determine the dialysis effect of proteins traveling from the ECS to the ICS, three Micro Mouse systems were assembled with 0.14-m² bioreactors having MWCO membranes of 10, 30, and 70 kDa, respectively. All experiments were performed without inoculating cells.

PBS, a chosen ICS solution, was circulated throughout the system at a rate of 100 ml/min. PBS supplemented with 20% FBS (V/V) was added to the ECS of the bioreactor. At time intervals ranging from 0 to 60 minutes, ICS samples were taken and assayed for leaked serum proteins using a characteristic absorbance of 280 nm, at which the spectrophotometer was initially blanked by PBS. A BCA protein assay reagent was purchased from Pierce Chemical (Rockford, IL) and employed to determined the total amount of protein both in the serum solution and in the samples.

Results

Comparison of Cell Growth Rates in Static Cultures. To determine whether CM can play a significant role in increasing hybridoma cell growth, we used CM at concentrations of 12.5%, 25.0%, and 50.0% (V/V), mixed into normal complete medium for each of the three hybridoma cell lines. The experimental control was run using complete medium only. Cell-growth kinetics were investigated for the 3F6, 3C11, and HFN 7.1 cell lines using a specific growth rate defined as

µ = 1 dX > 1X

Xdt Xt

where µ is the specific growth rate (1/day); X is the viable cell concentration (cells/ml); X is the difference in viable cell concentrations in the period of one day (cells/ml); X is the average viable cell concentration within one day (cells/ml); and t is equal to a period of one day (see Figures 2—4).

The 3F6 cells took a full day to adapt to the control medium that contained no CM, initially causing viable cell populations to decline (see Figure 2). In all three concentrations of the media containing CM, however, these cells immediately showed positive growth rates. This shortened lag phase after inoculation of the bioreactor is a measure of the efficiency of using CM. These experimental results suggest that using CM can compensate for the shortage of small molecular proteins required for cell growth and antibody production in the ECS of a hollow-fiber bioreactor system.

Figure 2. Comparison of the specific growth rates (SGR) of 3F6 viable cells cultured in T-flasks in triplicate with the use of conditioned medium (CM) at concentrations ranging from 0 to 50% (V/V).

By day 4, cell growth was in a stationary phase, and measurements taken on that day indicated declining growth rates. Similar results were found for the static cultures of the 3C11 and HFN 7.1 cell lines. The sole measurement taken during the first day (delayed by prolonged set-up for taking samples) indicated that the cell lines being grown in the control medium initially suffered a decline (see Figures 3 and 4). Investigators believe that the improved survival rate of such low-density cultures nourished with CM is the result of a "feeder layer" of cells that releases small molecular metabolites and macromolecules (growth factors and hormones), making CM beneficial to cell growth.^21—23

Figure 3. Comparison of the SGR of 3C11 viable cells cultured in T-flasks in triplicate using CM at concentrations ranging from 0 to 50% (V/V).

Another observation was that the growth rate of the 3F6 and 3C11 cell cultures using 50% CM (V/V) appeared to be at the same level as those using a lower percentage of CM. Since the 50% CM medium contained a lower proportion of nutrients and a higher proportion of waste metabolites than the other media, this result suggests that balancing growth factors in the medium may be more important than balancing nutrients—at least for cell populations growing at a slower rate. Since the HFN 7.1 cell lines had a higher growth rate, supplying nutrients appeared to be the priority after the first day of cell growth.

Figure 4. Comparison of the SGR of HFN 7.1 viable cells cultured in T-flasks in triplicate using CM at concentrations ranging from 0 to 50% (V/V).

Monitoring the Leakage of Serum Proteins. The molecular weights of serum proteins used in cell culture—such as albumin, transferrin, and the globulins—range from 60 kDa to more than 100 kDa. By contrast, bovine growth factors and hormones that have been characterized have molecular weights ranging from 5 kDa to 30 kDa. But autocrine growth factors also have molecular weights similar to those of bovine growth factors, making it very difficult to distinguish the proteins, growth factors, and hormones used in experimental media. For the purposes of these experiments, a serum protein was defined as any substance with a light absorbance at 280 nm.

The dynamics of leaking serum proteins detected in ICS samples were investigated for hollow-fiber bioreactors with three MWCOs of 10, 30, and 70 kDa (see Figure 5). The percentage of serum proteins transferred in the course of 60 minutes was 26.9% through the 10-kDa MWCO fibers; 30.7% through the 30-kDa MWCO fibers; and 38.2% through the 70-kDa MWCO fibers.

Figure 5. Comparison of the leakage of serum proteins from hollow-fiber bioreactors with molecular weight cut offs of 10, 30, and 70 kDa. The flow rate in ICS circulation was set at 100 ml/min.

The volume of media available to the bioreactor system has a direct effect on the transfer of serum proteins between the ECS and ICS. The 0.14-m² hollow-fiber bioreactors used in these experiments have a 10-ml ECS capacity and a capacity of 2 L in the ICS medium reservoir. After 60 minutes had elapsed in the experiment, the BCA assay of a 70-kDa MWCO bioreactor indicated an initial protein concentration of 7.4 mg/ml in PBS, with 20% supplemented FBS, and only 4.2 mg/ml of protein in the ECS. In short, leaked serum proteins were rapidly diluted into the large volume of ICS medium, and their concentrations at equilibrium were related to the volumetric ratio of ICS medium to ECS medium. Normally, the balance for small-molecular-weight proteins will settle at a very dilute concentration.

ICS circulation rate is another factor that affects the leakage of serum. An increase in flow rate can change the pressure distribution along the length of the bioreactor's hollow fibers and significantly increase diffusion and the Starling effect.²⁴ Both movements accelerate the rate of protein dialysis, increasing the amount of protein passing through the membrane from the ECS to the ICS of the hollow-fiber bioreactors. Since all of these experiments used a fixed flow rate of 100 ml/min., the effect of flow rate on leakage of growth factors remains to be examined.

Cell Culture in Bioreactors with Conditioned Medium. Results of a 30-day study of antibody production from three hybridomas in 0.14-m² hollow-fiber bioreactors are shown in Figures 6—8. In the ECS, the control run used complete medium without CM. To maintain a concentration of serum proteins in the ICS, the control run used basal DMEM supplemented with 2.5% FBS; fresh basal medium was used to gradually replace the spent ICS medium.

Figure 6. Profiles of cumulative monoclonal IgM antibody production in a control run and a run using CM in Micro Mouse BR110 from 3F6 hybridoma cultured for 30 days.

To evaluate whether CM was beneficial in antibody production, the initial ICS medium was prepared by mixing 250 ml of filtered CM with 750 ml basal medium. Since the CM contained 10% FBS when used in static culture, the ICS medium apparently retained 2.5% FBS by volume although the serum material had been used.

The 30-day cumulative total of antibody production for the 3F6 cell line is shown in Figure 6. The total amount of IgM antibody produced by the control bioreactor without CM was 14.5 mg, while the bioreactor using CM produced 22.9 mg—an increase of 57.9%. For the control run the average IgM concentration per harvest was 0.12 mg/ml, while the run using CM produced an average harvest of 0.19 mg/ml.

Results of the 30-day experiment were even more impressive for the 3C11 cell line. In this case, the control run produced 107 mg of IgG, while the experimental bioreactor using CM produced a total of 214 mg (see Figure 7).

Figure 7. Cumulative monoclonal antibody production by 3C11 hybridoma in Micro Mouse BR110 hollow-fiber bioreactors with and without the use of CM during a 30- day culture time.

The HFN 7.1 cell line was the highest antibody producer in the study. The control run accumulated 390 mg of IgG^₁ antibody over the 30-day period, while the experimental bioreactor using CM produced 525 mg (Figure 8). Antibody concentrations per harvest increased from an average of 3.2 mg/ml in the control run to 4.4 mg/ml in the experimental unit.

When using hollow-fiber bioreactors it is not possible to measure metabolic activity directly by counting cells, but the glucose uptake rate (GUR) can be used as an indirect gauge of the culture processes. A variety of influences can affect the GUR, including mechanical factors such as the configuration of the bioreactor and the flow and perfusion rates of the medium, as well as biological factors such as the cell line, medium formulation, and oxygenation rate. Unless data indicate that too-rapid cell growth has resulted in oxygen depletion, a high GUR generally indicates rapid growth and metabolically active cells.

Throughout the experiments presented here, the GUR was used as an indicator of cell growth and cellular metabolism (see Figure 9). Measurements of the dissolved oxygen level did not indicate oxygen depletion, and the ratio of lactate production to glucose uptake was kept between 0.70 and 0.85; thus any changes in the glucose level would be the result of cell consumption and would reflect the metabolic activity of the cell culture. GURs were also compared to evaluate the activity of cells in the bioreactor culture with and without CM. The peak GUR of 68 mg/hour was detected in a bioreactor using CM-modified medium, representing an increase of more than 100% over the highest rate detected in the control runs.

Costs. A benefit of using CM is the potential for reducing costs by using smaller amounts of medium and serum. This is possible, in part, because CM contains not only the autocrine growth factors secreted by its cells, but also specific serum proteins.

In fact, the cost and productivity of the culture medium used in a hollow-fiber bioreactor has a direct effect on the cost-efficiency of the system for antibody production. Table II indicates estimated costs for such a system, based on the costs of culture medium and disposable hollow-fiber bioreactors (labor and facility costs are not included). It is estimated that the cost of making monoclonal antibodies could be reduced from $33.72 to $20.04/mg for the 3F6 cell line; from $4.64 to $2.18/mg for the 3C11 cell line; and from $1.28 to $0.90/mg for the HFN 7.1 cell line. The savings related directly to the secretion capacity of the antibodies is somewhere in the range from 30.8% to 52.2%, depending on the cell line in question.

Table II. Comparative costs of producing monoclonal antibodies using normal culture medium and conditioned medium (CM) protocols. The cost of the medium was estimated at $4.00/L for DMEM, $0.30/ml for FBS, and $5.00/filter used for CM preparation.

Conclusion

In this study, operating conditions were standardized and no attempt was made to optimize production parameters for any of the three cell lines under investigation. Nevertheless, it is recognized that antibody yields could be increased by optimizing any number of operating variables, including the flow rate of the ICS medium, the frequency and volume of the ICS medium exchange, and the frequency and volume of harvest.

Increasing the flow rate of the ICS medium can raise the pressure within the system, thus enhancing nutrient exchange. However, doing so can also strengthen harmful Starling effects. In addition, because of their small molecular sizes, many useful autocrine factors could be irreversibly dialyzed away from the ECS to the ICS medium. Experiments performed in this study used an initial ICS medium containing 25% CM by volume. This proportion minimized diffusion of small molecular proteins through dialysis, and enabled the system to retain a higher concentration of such proteins at equilibrium. In the authors' experience, the initial volume of ICS medium should be one liter or less to avoid excessive dilution of growth factors. We recommend gradually exchanging the ICS medium as a way to maintain the level of growth factors necessary for cell growth. A gradual change of 50% by volume, for instance, could be acceptable until the bioreactor is confluent with cells.

Harvests were taken using 100% CM to replace the entire ECS medium. Normally, a 0.14-m² hollow-fiber bioreactor is configured with an ECS capacity of 10 ml. Removing all spent medium in the ECS enables operators to harvest the greatest amount of antibody, and to clean up such growth inhibitors as proteases and TGF-ß_¹ that are produced by some hybridoma cell lines.²⁵ Depending on the antibody secretion rate of the cell line, a harvest frequency of once a day or every other day is recommended.

Figure 8. Profiles of cumulative monoclonal antibody production from HFN 7.1 hybridoma, a higher MAb producer, in a control run and a run using CM in Micro Mouse BR110, cultured for 30 days.

Finally, operators should note that the quality of CM preparation is the key to launching a successful hollow-fiber cell culture. But in hollow-fiber bioreactors, the use of CM may not be required once full-scale antibody production is under way.

Figure 9. Time course of glucose uptake rate (GUR) in Micro Mouse BR110 for the 3C11 hybridoma with CM as 25% of initial ICS medium and 100% of the ECS medium, versus the control run with no CM.

References

1. TL Evans and RA Miller, "Large Scale Production of Murine Monoclonal Antibodies Using Hollow Fiber Bioreactors," BioTechniques 6 (1988): 762—767.

2. AH Hiefetz et al., "Monoclonal Antibody Production in Hollow Fiber Bioreactors Using Serum-Free Medium," BioTechniques 7 (1989): 192—199.

3. D Lowrey, S Murphy, and RA Goffe, "The Effect of Intracapillary Media Feed Protocols on Hollow Fiber Cell Culture," Biotechnology Letters 15 (1993): 1025—1030.

4. EP McKillip et al., "Growth of Transformed C-127 Cell in Bioreactor for Large Scale T-PA Production," Bio/Technology 9 (1991): 805—810.

5. J Rosenberg et al., "Use of Hollow Fiber Technology for Large Scale Production of Viruses and Viral Antigens," American Biotechnology Laboratory 8 (1990): 34—39.

6. FJ Volenec et al., "Automated Large Scale Culturing for LAK Cell Generation," (paper presented at the Conference on Immunobiology, International Symposium on Immunotherapy and Immunology, Nice, France, April 1987).

7. J Hopkinson, "Hollow Fiber Cell-Culture Systems for Economical Cell Product Manufacturing," Bio/Technology 3 (1985): 225—230.

8. LR Jackson et al., "Evaluation of Hollow Fiber Bioreactors as an Alternative to Murine Ascites Production for Small-Scale Monoclonal Antibody Production," Journal of Immunological Methods 189 (1996): 217—231.

9. R Kurkela, E Fraune, and P Vihko, "Pilot Scale Production of Murine Monoclonal Antibodies in Agitated, Ceramic Matrix, or Hollow Fiber Cell-Culture Systems," BioTechniques 15 (1993): 674—683.

10. GL Altshuler et al., "Continuous Hybridoma Growth and Monoclonal Antibody Production in Hollow-Fiber Reactors-Separators," Biotechnology and Bioengineering 28 (1986): 646—658.

11. JM Piret and CL Cooney, "Mammalian Cell and Protein Distributions in Ultrafiltration Hollow Fiber Bioreactors," Biotechnology and Bioengineering 36 (1990): 902—910.

12. RB Brodeur, PS Tsang, and Y Larose, "Parameters Affecting Ascites Tumour Formation in Mice and Monoclonal Antibody Production," Journal of Immunological Methods 71 (1984): 265—272.

13. RB Brodeur and PS Tsang, "High-Yield Monoclonal Antibody Production in Ascites," Journal of Immunological Methods 86 (1986): 239—241.

14. RJ Czirbik et al., "Factors Affecting Antibody Production Efficiency in Hollow Fiber Bioreactors," IVD Technology 2(4) (1996): 56—61.

15. AL Boynton and HL Leffert, Control of Animal Cell Proliferation, vol. 1 (New York City: Academic Press, 1985).

16. G Guroff, Oncogenes, Genes, and Growth Factors, (New York City: Wiley, 1987).

17. D Lauffenburger and C Cozens, "Regulation of Mammalian Cell Growth by Autocrine Growth Factors: Analysis of Consequences for Inoculum Cell Density Effects," Biotechnology and Bioengineering 33 (1989): 1365—1378.

18. A Burgess and N Nicola, Growth Factors and Stem Cells, (London: Academic Press, 1983): 355.

19. WH Burgess and T Maciag, "The Heparin Binding Fibroblast Growth Factor Family of Proteins," Annual Review of Biochemistry 58 (1989): 575—606.

20. GH Bock et al, "Characterization of A New IL-6-Dependent Human B-Lymphoma Cell Line in Long Term Culture," Cytokine 5, no. 5 (1993): 480—490.

21. TT Puck and PI Marcus, "A Rapid Method for Viable Cell Titration and Clone Production with HeLa Cells in Tissue Culture: The Use of X-Irradiated Cells to Supply Conditioning Factors," Proceedings of the National Academy of Science USA 41 (1955): 432—437.

22. M Stampfer, RG Halcones, and AJ Hackett, "Growth of Normal Human Mammmary Cells in Culture," In Vitro 16 (1980): 415—425.

23. K Takahashi and TS Okada, "Analysis of the Effect of 'Conditioned Medium' Upon the Cell Culture at Low Density," Development of Growth Differentation 12 (1970): 65—77.

24. EH Starling, "On the Absorption of Fluids from the Connective Tissue Spaces," Journal of Physiology 19 (1996): 312—326.

25. WR Kidwell, "Filtering Out Inhibition," Bio/Technology 7 (1989): 462—463.

Yuan Shi is manager for cell culture R&D and contract services, Julie Ploof was a cell culture R&D technician, and Alberto Correia is vice president for operations at Unisyn. (Hopkinton, MA).