The Potential Application of Hollow Fiber Bioreactors to Large-Scale Production

May 2, 2011
William G. Whitford|John J.S. Cadwell

BioPharm International

Volume 2011 Supplement, Issue 4

A hollow fiber matrix allows for efficient harvest of secreted proteins.


Hollow fiber bioreactors have been in use for more than 30 years for the production of secreted proteins and antibodies from mammalian cells. Advances in fiber construction and methods for oxygenation have the potential to bring the advantages of hollow fiber cell culture to biomanufacturing.

In order to more closely approximate in vivo cell growth conditions, Richard Knazek developed the hollow fiber bioreactor (HFBR) in 1972 (1). The HFBR system is a high-density, continuous-perfusion culture system. It is a well-established system for the production of secreted proteins from hybridoma, Chinese hamster ovary cells (CHO), human embryonic kidney (HEK) 293 and other mammalian and insect cells. The primary advantage is that it is possible to culture large numbers of cells at high density, resulting in product accumulating to very high concentrations. Up until now, the application of HFBR technology has been limited to laboratory scale applications. HFBRs provide some features that are advantageous in biopharmaceutical manufacturing, such as continuous, modular, flexible manufacturing formats employing compact and disposable systems. There is also growing interest in perfusion-based bioproduction protocols (2). FiberCell Systems has developed a large-scale HFBR system that brings the benefits of hollow fiber cell culture to the bio-production arena (see Figure 1).

Figure 1: Cells growing at high density in Fibercell Systems hollow-fiber bioreactor (HFBR).

HFBRs consist of thousands of semi-permeable hollow fibers in a parallel array within a tubular housing or cartridge, fitted with inlet and outlet ports. These fiber bundles are potted at each end so that any liquid entering the ends of the cartridge will necessarily flow through the interior of the fibers. Cells are generally seeded within the cartridge, but outside of the hollow fibers in what is referred to as the extra capillary space (ECS) (see Figure 2). Culture media is pumped inside the hollow fibers, allowing dissolved gas, nutrients, and waste products to diffuse across the fiber walls. Once having passed through the cartridge, the culture medium is oxygenated and returned to the cartridge.

Figure 2: Schematic of the growing conditions within a HFBR.

Unlike current popular batch-style manufacturing formats, including such disposable systems as rocked bags and bags supported by stainless steel containers, HFBRs do not subject cells to significant variations in primary and secondary metabolites (such as glucose and lactate), oxygen, or product levels. A HFBR significantly cuts the operating footprint and reduces the product harvest volume by factor of 10 to 100 times. Furthermore, it avoids such scale-determined discrepancies as impeller and sparge-induced hydrodynamic forces, and reactor column height-induced differential gas solubility. The high cell density and low-sheer environment permits the use of simplified serum-free, protein-free media that can be less expensive and that does not require such components as synthetic membrane protective surfactants or antifoams. Attached cultures are supported, therefore no lengthy adaptation of cells to suspension growth is required. The 10–100 times more concentrated secreted product, the absence of high molecular weight medium components, and the greatly reduced levels of host-cell proteins and DNA facilitate downstream processing, a major cost in bioproduction.


Bringing the advantages of HFBRs to large-scale bio-manufacturing has long been a goal for those familiar with hollow fiber systems. The operational scale for HFBR systems has been limited by several factors:

  • Solubility of gas in cell culture medium decreases as the temperature is increased to 37° C.

  • It is difficult to design a pump that will provide 6–10 L/min flow rate without denaturing biological components or generating significant backpressure.

  • The gross filtration rate of cellulosic hollow fibers (i.e., the rate of nutrient and waste product exchange) have typically been too low to effectively support the culture of adherent cell lines such as CHO and HEK293.

  • Fiber geometry is such that the fibers were not uniformly distributed within the cartridge housing, resulting in areas where a large inter-fiber distance led to necrotic areas within the cartridge.

FiberCell Systems has developed a unique approach to address these issues. Instead of using pumps and systems to perform gas exchange with the cell culture medium, a gravity-based system for medium flow that also directly oxygenates the bioreactor cartridge from the inside out has been developed (see Figure 3). A new generation of high flux fibers is employed, and they are also cast with "waves" in them to evenly distribute the fiber bundle within the hollow fiber cartridge. With a cartridge volume of 1 L, this fully disposable system has the potential to replace up to 10,000 L stirred tank reactors for mammalian protein production over a 100–day period of culture. This large-scale system offers all of the advantages outlined above, along with others that are specific to bio-production in a disposable system.

Figure 3: Fibercell Systems prototype large-scale HFBR.


There are four fundamental characteristics that differentiate hollow fiber cell culture from classic suspension culture:

  • Cells are bound to a porous matrix much as they are in vivo.

  • Cells are cultured in a gently perfused, non-shear environment.

  • The molecular weight cut off of the support matrix can be controlled.

  • The system possesses an extremely high surface area to volume ratio (150 cm2 or more per mL).

Cells are bound to a porous support much as they are in vivo.

There is no requirement to split cells. Cells in this perfusion system maintain viability and production-relevant metabolism in a post-confluent manner for extended periods of time, months or longer. Passage number is basically irrelevant and a certain percentage of the cells will become quiescent, that is, they will produce protein but do not divide. For example, one hybridoma was reported to maintain efficacious productivity for over one year of culture while a glioma cell line was cultured for nearly two years of continuous HFBR culture.

Cells are cultured in a gently perfused, non-shear environment.

Unlike suspension culture, the cells are constrained within a non-shear compartment with gas, nutrition and waste transport moving across the fibers. Lack of shear simplifies the medium composition, and perfusion permits a more homeostatic, in vivo-like environment for the cells. This can result in improved protein translational fidelity. These growth conditions and lack of shear also result in significantly reduced apoptosis (3). The majority of cells that become necrotic will not release host cell proteins or DNA into the culture medium, resulting in a product that is cleaner and easier to purify from the bulk harvest

The molecular weight cut off of the fiber can be controlled.

Desired products can be retained to 10–100 times higher concentrations. Toxic or unstable products can be selectively removed, and the effects of cytokines controlled. An example of this is hybridoma culture, in which the inhibitory cytokine transforming growth factor beta can be selectively removed from the culture, while the secreted antibody is retained.

The system possesses an extremely high surface area to volume ratio.

HFBRs have an extremely high surface area to volume ratio (in the range of 100–200 cm2 /mL of volume). Coupled with the high gross filtration rate of FiberCell Systems' polysulfone fibers, the exchange of nutrients and waste products is rapid. Cell densities of 1–2 × 108 or more are achieved, close to in vivo tissue-like densities. A module with a volume of 1 L would support as many cells as a 100 L stirred tank, as well as providing continuous production. High cell densities support higher volumetric productivities, and facilitate adaptation to lower serum concentrations or serum free culture. In fact, a simplified, protein-free, chemically defined, animal component-free serum replacement, CDM–HD has been developed by FiberCell Systems and is optimized for use in hollow fiber systems. The use of a protein-free medium such as CDM–HD provides cleaner harvests of product, easier regulatory compliance and simplified downstream purification.

The more in vivo-like cell culture conditions can also result in improved protein folding and more uniform glycosylation patterns over time. Since it is a continuous perfusion system, the amount of protein produced is determined as much by the length of time the culture is maintained as by such parameters as the clone's specific productivity or the size of the cartridge.

The combination of an unlimited nutrient supply and the ability to de-bulk the culture through the cartridge ports allows the system to be maintained at relative equilibrium for several months or longer. This continuous production over long periods of time, rather than the severely batch-style results from other systems, provides several benefits, including: consistency in culture condition, dramatically increased production per unit footprint and culture volume, continuous or daily product harvest allowing timely and convenient stabilizing treatment or storage conditions, and products that might be toxic or inhibitory to cells can be selectively removed from the culture.


The system included the following components:

  • A disposable, preassembled, and sterilized package including the cartridge, two media bags with fittings, and two disposable sensor "windows" with reactive pads bound inside.

  • Gravity pump with motorized cartridge angle controller.

  • Environmental enclosure.

  • Control unit with monitoring devices.

The diameter of the fibers has been increased to reduce the resistance to flow while still presenting a large surface area to volume ratio. This allows the use of a gravity pumping mechanism that can generate a flow of 5–10 L of media per minute through the interior of the hollow fiber bundle. The gravity pump will then simply reverse direction. The bags contain both medium and gas, and when the liquid is finished flowing through the cartridge, the gas is drawn through the interior of the fibers, directly oxygenating the cartridge from within. The disposable bags (Thermo Scientific HyClone BPCs employing CX5-14, a pentalaminate animal component-free film) are gas-impermeable, and the gas composition inside the bags is tightly controlled for optimum cell culture performance. The environmental enclosure is a simple, non-pressurized, temperature controlled incubator, but can be modified to include further environmental control such as HEPA filters to create a Class 100 clean room environment within the system for cGMP operation. Sensors for pH, O2, and potentially real-time glucose monitor the culture environment. The absence of cells in the medium bags to be monitored supports more robust and sensitive measurements. These sensors will support the control of medium and gas in the media bags. More advanced monitoring is possible using any one of a number of new approaches, including capacitance based monitoring probes. This exciting new technology employs passive electrical (dielectric) radio frequency-based sensors to measure the overall capacitance, and therefore viability, of the cell mass within the cartridge. The low volume in both the cartridge and the medium present in the system at any one time makes both monitoring and process control more precise and responsive, even though the system may consume 50–100 L/day. Stacking like-sized cartridges with their media bags on top of each other can provide additional scale.

Figure 4: Hollow fiber cartridge on Fibercell Systems prototype large-scale HFBR.

The FiberCell Systems HFBR allows the production of about 20–200 mg of antibody per week using the 20 mL cartridge (FiberCell Systems catalog # C5011). The product is harvested in a concentrated form 500–5000 µg/mL in a volume of 20–40 mL and production of the antibody can be accomplished in a standard CO2 incubator.

As an example, two monoclonal antibodies produced in DMEM + CDM-HD are shown in Figure 5. Length of culture was three weeks, total antibody produced was in excess of 140 mg each, and average concentration was 3.0 mg/mL. The dialyzed supernatants had their protein concentration quantified and purity check on a 12% sodium dodecyl sulfate polyacrylamide gel. A total of 12 L of medium per antibody was consumed.

Figure 5: Raw supernatant harvest was briefly dialyzed to reduce sample tonicity. No other purification was performed. The heavily loaded samples demonstrate low levels of contaminating proteins. (Data courtesy of Dr. Erin Bromage, U. Mass., Amherst)

Extrapolating this data to a 1 L sized hollow fiber bioreactor cartridge would result in total antibody production of approximately 3 g/day in a volume of 1 L, consuming 50 L of DMEM/CDM-HD per day.

473 mg of purified recombinant protein recovered from an rCHO cell line was harvested from the FiberCell Systems 20kd MWCO cartridge (FiberCell Systems catalog # C2018). Medium was Dulbecco's Modified Eagles Medium (DMEM) with 2% fetal bovine serum, each harvest was 70 mL in volume, and total harvest volume was 4.8 L, for an average protein concentration of approximately 100 µg/mL/d. The protein was a very complex, hexamerized immunoglobulin G (IgG) consisting of 6 IgG1 subunits held together with three IgA tails. The variable fragment (Fv) region was modified to contain the CD4 receptor (2). The cartridge consumed an average of 2 L of medium per day over a 60–day period of production.

An interesting observation was the comparison of protein produced using T–flasks versus the hollow fiber cartridge (see Figure 6). When produced in flasks, approximately 40 % of the protein was secreted as an unfolded monomeric subunit. Placing the same cells into the HFBR cartridge resulted in nearly 95 % of the protein being produced as a properly folded hexamer (3). Better cell culture conditions resulted in better protein expression fidelity.

Figure 6: Gel filtration chromatography of a hexeramized recombinant immunoglobulin (IgG). When cultured in flasks (top trace), approximately 40% of the protein is expressed as an improperly folded monomeric subunit (B) rather than the hexamer (A). These cells when transferred to a hollow fiber bioreactor exhibit 95% complete folding ( C ) as a result of the improved cell-culture conditions.

Expression levels were significantly below those found in commercially optimized CHO cell lines, although the results are still impressive. A 1 L sized HFBR cartridge would have a total harvest volume of 100 L, total protein recovered would be 10 g and total medium consumed would be 2,400 L.

246 mg of purified recombinant IgG1 from a CHO (DG44) cell line was harvested from the FiberCell Systems 20kd MWCO cartridge (FiberCell Systems catalog #C2011). Medium was a serum-free, protein-free formulation similar to CDM-HD. Each harvest was 20 mL in volume; total harvest volume was 320 mL for an average concentration of over 800 µg/day/mL. The cartridge consumed 1 L of medium every three days and the culture was maintained for a total of 35 days. For technical reasons, the cell viability in this run dropped rather low, however, the harvested protein produce was remarkably uncontaminated, as demonstrated by the gel of the unpurified harvest in Figure 7. In a 1 L sized HFBR, the total protein recovered would be 12.5 g in a volume of 16 L with 500 L of medium consumed.

Figure 7: Harvest from bioreactor rCHO cells. Raw supernatant was briefly dialyzed to reduce sample tonicity, no further purification was performed.

To extrapolate these results to a commercially prepared recombinant CHO cell line with expression in the range of 1 g/L in a standard stirred reactor, the 1 L sized HFBR cartridge should be capable of producing 10 g or more of protein per day, while consuming 100 L of medium.


This system offers the following advantages:

  • Completely disposable system components

  • Product can be accumulated at extremely high concentrations.

  • The system supports the production of enzymes, monoclonal antibodies, therapeutic proteins and a variety of vaccine components

  • Cell-culture conditions are optimized for protein production over long periods of time

  • Simplified, low cost, serum-free medium can be used, reducing expense, enhancing regulatory compliance, and simplifying downstream processing

  • Cells do not require adaptation to suspension culture, saving time

  • Further productivity scale-up can be accomplished by longer culture times, not new equipment

  • Culture monitoring sensors are located in the medium reservoir bags, not in the area containing cells, making process monitoring more robust

  • Product quality such as protein fidelity may be improved based upon optimized cell culture conditions

  • Apoptosis is reduced, further removing cellular proteins and DNA from the harvest supernatant, simplifying downstream processing

  • Intensive cleaning involved in traditional stir-tank bioreactors is not an issue

  • The small volume of medium present at any time renders control of parameters rapid and efficient.

The large-scale HFBR has the potential to bring all of these advantages to the manufacture of proteins from mammalian cells. If the oxygen transfer rates are high enough, this system could also be used for some fermentation processes as well. A disposable, compact replacement for large stirred tanks that produces protein at high concentrations in a simplified, chemically-defined medium represents a possible paradigm shift in the production of biopharmaceuticals.


The authors gratefully acknowledge the assistance of LaDonna Connors and Gwendolyn Gainer in the preparation of this article.

WILLIAM G. WHITFORD is senior manager, bioprocessing market, Thermo Fisher Scientific, Logan UT, and JOHN J.S. CADWELL* is president and CEO of FiberCell Systems, Frederick, MD,


1. R.A. Knazek, et. al., Science178 (56), 65–67 (1972).

2. W.G. Whitford and J.J.S. Cadwell, Bioprocess Int.7(9), 54–63 (2009).

3. P-C. Liao et. al., J. Proteome Res.8, 5465–5474 (2009).

4. A.S. Fauci et. al., JBC277(13), 11456–11464 (2002).