Biotechnology Advances 32 (2014) 514–525

Contents lists available at ScienceDirect

Biotechnology Advances

j ourna l homepage: www.e lsev ie r .com/ locate /b iotechadv

Research review paper

Multi-stage continuous high cell density culture systems: A review

Ho Nam Chang a,⁎, Kwonsu Jung a, Jin-dal-rae Choi a,1, Joon Chul Lee b, Hee-Chul Woo c

a Department of Chemical and Biomolecular Engineering, KAIST, 291 Daehak-ro, Daejeon 305-701, Republic of Koreab KITECH, 143 Hanggaul-ro, Ansan-si, Gyeonggi-do 426-910, Republic of Koreac Department of Chemical Engineering, Pukyong National University, 365 Sinseon-ro, Busan 608-739, Republic of Korea

⁎ Corresponding author. Tel.: +82 42 350 3912; fax: +E-mail address: hnchang@kaist.edu (H.N. Chang).

1 Present address: GS Caltex Corporation, 359, Expo-ro

0734-9750/$ – see front matter © 2014 Published by Elsehttp://dx.doi.org/10.1016/j.biotechadv.2014.01.004

a b s t r a c t

a r t i c l e i n f o

Article history:Received 3 July 2013Received in revised form 31 December 2013Accepted 15 January 2014Available online 21 January 2014

Keywords:Multi-stage continuous bioreactorsHigh cell density cultureProduct titer improvementProductivity improvement

Amulti-stage continuous high cell density culture (MSC-HCDC) systemmakes it possible to achieve high produc-tivity together with high product titer of many bioproducts. For long-term continuous operation of MSC-HCDCsystems, the cell retention time and hydraulic retention time must be decoupled and strains (bacteria, yeast,plant, and animal cells) must be stable. MSC-HCDC systems are suitable for low-value high-volume extracellularproducts such as fuel ethanol, lactic acid or volatile fatty acids, and high-value products such as monoclonal an-tibodies as well as intracellular products such as polyhydroxybutyric acid (PHB), microbial lipids or a number oftherapeutics. Better understanding of the fermentation kinetics of a specific product and reliable high-densityculture methods for the product-generating microorganisms will facilitate timely industrialization of MSC-HCDC systems for products that are currently obtained in fed-batch bioreactors.

© 2014 Published by Elsevier Inc.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5151.1. Product yield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5151.2. Product titer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5151.3. Bioreactor productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515

2. High cell density culture systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5152.1. Immobilized cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515

2.1.1. Cell entrapment in polymer matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5162.1.2. Cell retention by membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5162.1.3. Self-immobilized cells by flocculation or aggregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516

2.2. Suspended cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5172.2.1. Membrane bioreactors in wastewater treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5172.2.2. Monoclonal antibody production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5172.2.3. Ectoine production using Halobacteria elongata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5172.2.4. Cell recycling with upflow packed-bed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517

3. Single stage continuous HCDC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5173.1. History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5173.2. Cell growth and product formation kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517

3.2.1. HCDC with continuous cell mass production (CMP) reactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5183.2.2. Membrane fouling and long-term operation of HCDC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5183.2.3. High purity oxygen-based cell culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518

4. Multistage continuous high cell density culture (MSC-HCDC) systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5194.1. History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5194.2. Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5194.3. High productivity and high product titer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520

82 42 350 3910.

, Daejeon 305-380, Republic of Korea.

vier Inc.

515H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

5. Applications of MSC-HCDC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5205.1. Volatile fatty acids production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5205.2. Lactic acid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5215.3. Ethanol production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5215.4. Intracellular products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5225.5. Simulation of MSC-HCDC bioreactors for high titer and productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522

6. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524

1. Introduction

The term bioreactor refers to a device, or system that containssubstrates and enzymes or cells as biocatalysts and provides an environ-ment in which the biocatalysts can perform their functions.

The characteristics of enzyme biocatalysts resemble more or lessthose of chemical catalysts in that their activities degrade with time,whereas cells are self-multiplying living systems. Both types ofbiocatalysts have undergone successful developments in producingvarious products. However, this review will not include discussion onenzymes but instead focus on microbial cells including bacteria, yeastsand fungi, and on plant and animal cells, grown in a reactor or a systemof reactors where bioreactions occur efficiently (Asenjo and Marchuk,1995; Brauer, 1985; Cooney et al., 1985; Nielsen and Villadsen, 1994;Shuler and Kargi, 2002).

Bioreactions can be conducted in bioreactor systems with many di-verse characteristics. Once a product is selected, we must consider var-ious aspects of production using microbial cells, plant or animal cells,including the sterilization processes, bioreactor operationmodes, prod-uct location (intracellular or extracellular) and separation methods ormay even compare bioreactor production with competing manufactur-ing methods by biological or chemical means or a combination of both.

Themanufacturing cost of a bioproduct consists of (1) rawmaterials,(2) utilities (e.g., steam and electricity), (3) labor, and (4) depreciationof the capital investment per unit quantity of a product. In addition,~15% profit may be added. Depreciation is evaluated by dividing thetotal capital cost by the sum of the total manufactured quantities overthedepreciation-years. Theparameters thatmost affect themanufactur-ing costs are product yield and titer. In the case of high-value low-quantity products, the purification costs can exceed the culture costs.

1.1. Product yield

The cost of a bioproduct from feedstock consumption dependsmainly on the cost of the raw materials, the conversion rate and theproduct yield:

$=kg−product ¼ $=kg‐raw materialConv: %ð Þ � Yield kg‐product=kg‐raw materialð Þ : ð1Þ

The contribution of feedstock consumption to thefinalmanufacturingcost declines when developing processes for less costly raw materials,higher conversion efficiency andproduct yield. For example, low-cost lig-nocelluloses have an advantage over sugarcane or grains (e.g., corn orwheat) for lowering fuel ethanol production cost. However, lignocellu-loses have a lower substrate conversion and product yield than grains.Substituting grains with low-cost raw materials is not a simple solution.

1.2. Product titer

Weneed a high-titer product (kg/m3 or g/L) becausewemust removewater or separate a product from broth, which has a water content ofnearly 90%. Distillation or extraction is frequently used to concentratewater-soluble products from brothwith low product titers. For example,we could obtain pure acetic acid from 1 m3 of fermentation broth

containing 3.5% (w/w) acetic acid by distilling or removing ~965 kg ofwater. In practice, it is very difficult to obtain pure acetic acid by distilla-tion because of the water-acetic acid azeotrope. Having a higher producttiter greatly reduces the cost of water removal because a 1% (w/w) solu-tion of the product requires the removal of 990 L of water, while a 10%(w/w) solution needs the removal of 900 L of water. Currently, fermen-tation products are enriched by distillation or extraction, requiring ex-pensive heat energy or solvents. However, if we could remove waterfrom the fermentation broth using a non-phase-changing membranetechnology, such as reverse osmosis or forward osmosis, the enrichmentcostswould be greatly reduced (McCutcheon et al., 2005;Mulder, 1996).

1.3. Bioreactor productivity

Productivity measures bioreactor efficiency in terms of kg-product/(m3 bioreactor volume per unit time), which depends on biocatalystssuch as bacterial cells, yeast, fungi, plant or animal cells and the modeof the bioreactor operation being batch, fed-batch, continuous or highcell density cultures. The volumetric productivity of a bioreactor (Qp)can be expressed as the product of the specific productivity of cells(qp/x) and the cell mass per unit volume, X, within the bioreactor(Cooney, 1983). However, various factors such as supplementation ofcarbon source and other nutritional components, C/N ratio, dissolvedoxygen, and formation of products and byproducts that are inhibitoryto cells also affect Qp through their impact on cell physiology and me-tabolism, which consequently affect either qp/x or X or both.

Qp can be increased by using a high performance strain (qp/x) and alarger cell mass (X). In addition, sufficient oxygen supply is very impor-tant in aerobic high cell density culture tomeet its requirement that sig-nificantly affects cellular physiology and metabolism. In the batch andfed-batch culture systems such as ethanol, lactic acid, or penicillin, thesubstrate for cell mass formation is usually minimized through optimi-zation. A smaller cell mass reduces substrate consumption for biomassaccumulation to improve product yield, although the less cell biomasslowers the productivity. Multistage continuous high cell density culture(MSC-HCDC) systems consist of n-serially connected continuous stirredtank reactors with either hollow fiber cell recycling or cell immobiliza-tion for high cell density culture.

The objective of this review is to introduce a MSC-HCDC bioreactorsystem that increases the productivities bymaintaining high product ti-ters for batch and fed-batch fermentations (Chang, 2011; Chang et al.,2011a). In short, we seek amethod that can replace the current conven-tional fed-batch method with high productivity (Fig. 1). Bioprocessingfor fuels and chemicals from biomass can be quite different fromsugar-based fermentation products in terms of bioreactor productivityand titer (Lee et al., 2012).

2. High cell density culture systems

2.1. Immobilized cells

HCDC refers to approximately 10 times the normal cell density of asimple batch culture-. If 5–10 g/L of Escherichia coli cells is considereda normal cell density, 50–100 g/L would be called high cell density

Product

Pro

duct

Tite

r

Residence Time

Fed-batch: one reactorwith multiple feedings

MSC-HCDC: multiple reactorswith individual feedings

Substrate

Substrate

Substrate

SubstrateFed-batch

(MSC-HCDC)0 20 (2) 40 (4) 60 (6) 80 (8) 100 (10)

Fig. 1. Curve showing the time course of product titer in a MSC-HCDC bioreactor. The cyl-inders below the curve represent a tanks-in-series system, a typical configuration of thefed-batch system. The residence time of a MSC-HCDC (θ msc − hcdc) bioreactor can beas short as the 1/10th that of a batch or fed-batch bioreactors. Multiple feedings are nec-essary only when substrate inhibition kinetics are dominant (Chang et al., 2011a).

516 H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

(Lee, 1996). There are two ways of obtaining high cell density cultures:immobilizing cells and decoupling the solid retention time (SRT) ofsuspended cells from the hydraulic retention time (HRT) of broth in acontinuous culture.

(1) In simple continuous culture, SRT = HRT, and there is no celldensity enrichment.

(2) In cell recycling culture, SRT N HRT, and cell enrichment occurs.(3) With immobilized cells, SRT = ∞, ≫HRT, and no cell washout.

The maximal cell densities per unit volume (cm3 or L) can be esti-mated as follows:

• Suspended cells: a 1 L volume is fully filled with wet cells (solidcontent = 20%), Xmax = 1000 g wet cells/L × 0.2 dry cell weight(dcw)/wet cell weight = 200 g/L If the cells have less than 80%water content, Xmax may exceed 200 g/L.

• Animal cells: 106–107 cells/mL: Maximal cell density = 1.92 × 108

cells (of 10 μm diameter), 2.39 × 107 cells (of 20 μm diameter) or7.08 × 106 cells (of 30 μm diameter) Cells with shapes other thanspheres can have a larger volume per unit surface area and thereforeprovide less cells/mL.

• Plant cells: A 1 kg ofwet seaweed yields roughly 0.1 kg of dry seaweedalthough this yield may change with seaweeds and drying methods.Some plant cells in a confined space can have a cell density of 325 g-dcw/L (Kim et al., 1989). A cell density (ρdcw) of algae (Chlorellavulgaris) is 30.0 g-dcw/L (Javamandian and Palsson, 1991). 1 mL ofwet cell volume consists of 10% dry cell weight and 90% water. Thus30.0 g-dcw/L occupies approximately 30% of the bioreactor volume.

2.1.1. Cell entrapment in polymer matrixWilliams and Munnecke (1981) immobilized Saccharomyces

cerevisiae cells on calcium alginate beads for continuous ethanol pro-duction to investigate various physiological conditions including tem-perature, pH, ethanol concentration, cell density and alginate beaduploading. The data were compared to the optimal parameters for eth-anol fermentation by non-immobilized cells. Purwadi and Taherzadeh(2008) performed ethanol fermentation by S. cerevisiae cells immo-bilized on alginate beads. The cell density near the surface was 129 g/Land the average cell density was 38 g/L. Considering the void volumeof the packed bed as approximately 0.5, the effective average cell densi-ty per liter of the bioreactorwould be 19 g/L, one-half of the average celldensity. Furthermore, the occupancy of the packed beads reduced theresidence time in the reactor to the half that of empty bioreactors. Cer-tainly immobilized cells have an advantage in preventing washout, butit is not considered as a favorable method for long-term operation of

MSC-HCDC systems because of immobilization costs on supports, re-duced residence time, mass transfer resistances, lack of oxygen and nu-trient supply to immobilized cells, and the stability of immobilized cells.

2.1.2. Cell retention by membranes

2.1.2.1. Hollow fiber membrane bioreactor. A hollow fiber membrane bio-reactor was used first in the 1970s as an enzyme immobilization devicein which the enzyme reaction was conducted with the substrate solu-tion in the lumen and the enzyme outside the shell (Rony, 1974;Waterland et al., 1974; Waterland et al., 1975). A small molecule sub-strate would pass through the membrane while the large molecularweight enzyme was impermeable.

Inloes et al. (1983a, 1983b) immobilized Saccharomyces and E. colicells on the shell side of hollow fibers to produce ethanol and enzymes.Unfortunately, the hollow fiber bioreactor did not operate well becauseof the overgrowth of the immobilized cells. Running the bioreactor witha nitrogen-deficientmediumwas not successful because of the nutrientimbalance (Inloes et al., 1985).

2.1.2.2. Dual hollow fiber bioreactors. Common hollow fiber systems arenot suitable for growing aerobic microbial cells because of insufficientoxygen supply. In order to make aerobic cells grow, dual hollow fiberbioreactors (DHFBRs) are used: a silicone tubing system for oxygen sup-ply (gaseous substrates) and polypropylene tubing for supplying liquidnutrients (Robertson and Kim, 1985).

An extremely high cell density of 500–600 g/L is obtained, which ismuch higher than the liquid phase anaerobic (200–300 g/L) or faculta-tive (200 g/L) cultures, and aerobic culture (less than 100 g/L). But thestrength of the microbial culture (fungal) is high enough to distort thetubes, which makes continuous operation almost impossible. Industrialapplications may not be possible unless porous ceramic or stainlessstructures are used. If this becomes successful, perhaps a cell free fer-mentation product may be obtained. This technology may have someapplications in biomedical tissue cultures (Chung and Chang, 1988;Chung et al., 1987; Kim et al., 1989).

2.1.2.3. Depth filter perfusion systems. The nutrients are delivered directlyto cells that are partially surrounded by depth filter perfusion systems(DFPSs) where the membrane pore size is 2–3 times larger than the di-ameter of the cells. A DFPS has high density cells suspended in themacro-pores of the depth filters. The cell density could be as high as 1− 3 × 108 cells/cm3 of the DFPS volume (Lee et al., 2008a; Oh et al.,1994). Scale-up is easy by installing several units of commercially avail-able DFPS units as compared to hollow fiber or dual hollow fiberimmobilized cell systems. DFPS maintains a better operational stabilitythan the other hollow fiber and DHFBR systems (Lee et al., 2008a).The advantage of the immobilized cells is washout-free while insolublesubstrates cannot be utilized and oxygen delivery is still poor in com-parison with the cells in suspension (Chang and Moo-Young, 1988).

2.1.3. Self-immobilized cells by flocculation or aggregationAnother way to separating the solid retention time from the hydrau-

lic retention time is to use cells self-immobilized due to flocculation oraggregation. If this modewere generalized or applied tomanymicrobialcell systems, it would be one of themost desirablemethods for immobi-lization (Zhao and Bai, 2009). This technology has been successfully ap-plied to aworking volume of 1000m3 in commercial ethanol plantswithan annual production capacity of 200,000 tons (Bai et al., 2011). This isthe largest commercial bioreactor system using immobilized cells,which overcomes all the disadvantages of immobilized cells and mem-brane cell recycle systems but still maintains ease of cell immobilization.The limitation canbewhether self-immobilization byflocculation can beapplicable to all other microbial fermentation systems in general.

517H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

2.2. Suspended cells

2.2.1. Membrane bioreactors in wastewater treatmentThe current commercial use of high cell density cultures of

suspended cells is inmixed culturemembrane bioreactors for wastewa-ter treatment (Thobanoglous et al., 2002). The mixed liquor suspendedsolids (MLSS), a mixture of cells and suspend solids, ranges from10,000 mg/L to 20,000 mg/L, whereas conventional activated sludgehas aMLSS of 5,000mg/L. The hollow fibermodules are placed inwaste-water treatment reactors as large as 3000m3. Its BOD removal efficiencyis 2–10 times that of a conventional activated sludge process and doesnot require a settling tank. The effluent water is of very high quality(Wikipedia, 2013a). ThefirstMBRprocess usingflatmembranemoduleswas introduced externally to an activated sludge reactor in the late1960s by Dorr-Liver, Inc. when ultrafiltration (UF) and microfiltration(MF) membranes became commercially available. In 1989, Yamamotoand coworkers proposed placing membrane modules in an internal/submerged mode inside the activated sludge tank. This internal modeoperation uses only 0.3 kWh/m3 of membrane filtrate while the formerexternalmode uses 10 kWh/m3. It is nowwidely used formunicipal andindustrial wastewater treatment of 48,000 m3/day (80,000 people) ormore (Wikipedia, 2013b).

It is interesting that large-scale hollowfibermodules have been usedsuccessfully in commercial wastewater treatment plants by employingsimple and chemical backwashing (Thobanoglous et al., 2002).

2.2.2. Monoclonal antibody productionChu and Robinson (2001) reviewed the industrial choices for protein

production in scaled up cell cultures including standard fed-batchstirred tank reactors and hollow fibers for various low-volume and spe-cialty applications.

Refine Technology developed the ATF™ System, a cell retentiondevice that is known for being able to routinely generate extremelyhigh cell concentrations (Bonham-Carter, 2011). Fed-batch reactorsyielded a monoclonal antibody titer and productivity of 2 g/L and0.15 g/L/day, respectively; concentrated fed-batch reactors yielded17 g/L and 0.94 g/L/day antibody titer and productivity, while concen-trated perfusion reactors yielded 0.8 g/L and 1.6 g/L/day antibody titerand productivity The highest titer of mAb is obtained in concentratedfed (17 g/L)-batch followed by fed-batch(2 g/L) and concentratedperfusion (0.8 g/L) while the highest productivity 1.6 g/L/h of mAb isobtained in concentrated perfusion followed by 0.94 g/L/day (concen-trated fed-batch) and 0.15 g/L/h (fed-batch). A high titer and mediumproductivity may be obtained in MSC-HCDC systems.

2.2.3. Ectoine production using Halobacteria elongataLentzen and Schwarz (2006) introduced the production of

extremolyte ectoine using H. elongata, which is common in environ-ments where large amounts of salt, moisture, and organic materialsare available. The use of this bacterium eliminates the chance of con-tamination during culturing because other common bacteria cannotlive in high salt conditions, and the downstream processing of intracel-lular products from this bacterium is simple because the cell walls arepartially or completely disrupted when distilled water is added to thebacterial paste, creating a very low osmotic condition. This process is agood method for obtaining intracellular HCDC products. The cells aregrown in a membrane bioreactor to a high cell density and lysed by os-motic pressure change by suspending them in distilled water (Bitop,2012). Further product processing is accomplished with electrodialysis,chromatography, drying and crystallization.

2.2.4. Cell recycling with upflow packed-bedSettling or packed-bedfiltration ismost frequently used inwastewa-

ter treatment. Chang et al. (2008a) used thismethod in high cell densitycultures of yeast cells for ethanol production and for separating thesludge particles in food-waste treatment (Chang et al., 2010a). Although

the retention efficiency is less than 100% (80–95%), this method retainsmost of the cells by settling and packed-bed filtration. Hollow fiber fil-tration of high cell density cultures becomes difficult and of limiteduse as the cell densities increase and the medium becomes more vis-cous. Using up-flow packed-beds for pre-filtration reduces the mem-brane loading by decreasing the cell mass.

The membrane system has 100% of cell separation efficiency, but itslong-term operation is rather difficult while centrifugation and up-flowpacked bed has less than 90% efficiency (60–80%), but its long-term operation is better than the membrane system. A better under-standing of the relationship between the cell density and efficiency ofseparation will make it possible to carry out long-term high cell densityfermentation.

3. Single stage continuous HCDC

3.1. History

HCDC of microbial cells is based on separating the SRT of microbialcells from the HRT of the culture system. The degree of this separationcan vary significantly. If we use membranes with pore sizes smallerthanmicrobial cells, we can achieve a complete separation that may re-sult in a maximal cell density of nearly 200 g/L (Lee and Chang, 1987).Othermeans of separation such as the gravity settling are used inwaste-water treatment systems but cell density is low. The cell density inactivated sludge bioreactors is only 1,500 mg/L, but it can be increasedto 6,000 mg/L by using pure oxygen instead of air (Rittmann andMcCarty, 2001). Applying the membrane bioreactor (MBR) process toactivated sludge bioreactors (Wikipedia, 2013a) can increase the celldensity to 10,000–25,000mg/L.With the introduction of UFmembranesand hollow fibers, membrane devices were used in enzyme bioreactorsin the early 1970s and in microbial bioreactors in the late 1970s andearly 1980s, but none of them remain in use at this time. In the 1980s,membrane cell recycling systems were also introduced and the kineticsof high cell density culture was first formulated (Lee and Chang, 1987).In the 1990s, high cell density culture for PHB (polyhydroxybutyric acid)production using pure oxygenwas carried out, yielding 160 g/L in a fed-batch culture (Kim et al., 1994). In the late 1990s and early 2000s, therewere a number of studies of two-stage high cell density culture systemsfor ethanol, acetic acid and lactic acid production. High density cell cul-tures using membrane-based cell recycling systems were reviewed(Chang and Furusaki, 1991; Chang et al., 1994).

3.2. Cell growth and product formation kinetics

A schematic diagram of a cell recycling system and its operation isshown in Fig. 2 (Lee and Chang, 1987).

The rate equations for cell mass (X), substrates (S), and metabolicproduct (P) are

dX=dt ¼ μ–BDð ÞX ð2Þ

dP=dt ¼ νX–DP ð3Þ

dS=dt ¼ D So−Sð Þ–μX=Yx=s ¼ D So−Sð Þ−nX=Yp=s: ð4Þ

In Eq. (5) regarding substrate consumption, the substrates are usedfor cell growth, product formation and maintenance. That is,

D So−Sð Þ ¼ ΔSgrowth μX=Yx=s−g

� �þ ΔSmaintenance X=Ys=x−m

� �

þ ΔSproduct nX=Yp=s−p

� �: ð5Þ

Because we do not have a good understanding of how much of thesubstrates is exactly consumed for each part of cell growth, product for-mation and maintenance, especially for maintenance purposes, Eq. (6)

Fig. 2. Schematic diagram of a membrane cell recycle fermenter (a); B, bleed ratio: S,glucose concentration; F, feed flow rate; X; cell concentration; P; product concentration;V; fermenter volume. Fermentation kinetics in a membrane cell recycle reactor (Lee andChang, 1987).

Table 1Kinetics of batch, simple CSTR and continuous HCDC bioreactors (Chang et al., 2011a).

Batch CSTR C-HCDC Remark

dX/dt μX (μ − D)X (μ − BD)XdS/dt −μX Yx/s D(S0 − S)−μX Yx/s D(S0 − S)−μX Yx/sdPex/dt vX −DPex + vX −DPex+ vX Extracellular PdPin/dt vX −DPin + vX −BDPin + vX Intracellular PdPip/dt Pip = P0 D(I0 − I) = 0 −BDI + DI0 =

I0 − BI DPip increasesin C − HCDC

At t = 0, X = X0, S = S0, Pex= 0, Pin = 0, and I = I0.X cell density, S substrate concentration, Pex extracellular product concentration, Pinintracellular product concentration, I inert particle concentration, μ cell growth rate, vproduct formation rate, B bleed rate, 0 ≤ B ≤ 1, D dilution rate.

518 H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

cannot be used for process control and optimization. Instead, Eq. (7) canbe developed based on the total substrate consumed.

D So−Sð Þ ¼ ΔSgrowth μX=Yx=s−t

� �¼ ΔSmaintenance X=Ym=s−t

� �

¼ ΔSproduct nX=Yp=s−t

� �: ð6Þ

At steady-state:

μ ¼ BD ð7Þ

n ¼ DP=X ð8Þ

Yx=s ¼ μX=D So−Sð Þ ¼ BX= So−Sð Þ ð9Þ

Yp=s ¼ vX=D So−Sð Þ ¼ P= So−Sð Þ: ð10Þ

To avoid cell washout, D should be kept lower than μ/B

D ≤ μ=B ð11Þ

μX ¼ BDX ¼ Yx=sD So−Sð Þ: ð12Þ

Eqs. (2) and (7) do not tell us how the high cell densities can be ob-tained. Eq. (12) predicts that in a cell recycling bioreactor produced bio-mass (μX) is balanced by that removed (BDX). Naturally cell productionshould be supported by substrate consumption D(So − S).

If all substrates are consumed, the maximal biomass density can bepredicted.

Xmax ¼ So Yx=s

h i=B or DSo Yx=s

h i=μ ð13Þ

μXmax ¼ BDð Þ SoYx=s

� �ð14Þ

The key relationship for maintaining a desired cell density isexpressed by Eqs. (2) and (4). Eq (2) shows that whatever number ofcells is produced in the bioreactor (μX) should be removed by BDX(Lee and Chang, 1987). In other words, μ can be manipulated by chang-ing D, B or both. Eq. (2) does not reveal how much biomass can be, butEq. (4) does, in which μX can be balanced by BDX, and μX productionshould be supported by substrate consumed Yx/sD(So − S). Highervalues of μX (in the exponential growth phase and at high cell density)require higher values of Yx/sD(So − S), whereas low values of μX(in the stationary phase or at low cell density) require low values ofYx/sD(So − S) so that the rest of supplied substrates can be used forproduct formation. If we maintain cells in a stationary state, a very

small amount of cells is removed from the reactor, which means thatB should be very small. In this case, a large amount of broth should befiltered by membranes. Otherwise, HCDC becomes very difficult.

Table 1 summarizes kinetics for batch, simple CSTR, and continuousHCDC systems in terms of cell growth, substrate consumption, productformation (extracellular and intracellular) and inert particles.

3.2.1. HCDC with continuous cell mass production (CMP) reactorA cellmass production reactor system is designed to prepare cells for

the next bioreactors withoutmassive cell recycling. Fig. 3 shows a sche-matic diagram of a cell mass production (CMP) reactor consisting of anin-house cell production reactor operating at exponential growth phasefor maximal cell mass production followed by a cell enrichment systemthrough packed-bed and centrifuge or a hollow fiber device, respective-ly. For extracellular products such as lactic acid and ethanol, the externalcell supply (concentrated) comes from the last reactor, where the cellsare enriched with the aforementioned methods if necessary, andreturned to the CMP reactor system. The percentage of the cells pro-duced in the CMP reactor and the cells recycled from the last reactor de-pends on the viability of the cells being returned from the last reactor.

For intracellular products such as polyhydroxybutyric acid (PHB)and microbial lipids for biodiesel, all of the cells are discharged in thelast reactor without cell recycling. One certainty is that a CMP operatedwith a low cell density culture in the exponential growth phase in-creases the amount intracellular products obtained in the fermentationreactor by several fold.

3.2.2. Membrane fouling and long-term operation of HCDCWepresently usemembranes for high cell density cultures ofmicro-

bial cells. Again, the problems we have to solve are how to avoid mem-brane fouling by proteins, microbial cells, and microbial extracellularpolysaccharides. Currently, the high cell density cultures of animalcells used for commercial monoclonal antibody production incorporatemembrane cleaning by pressure pulsing (Bonham-Carter, 2011;Furusaki et al., 1977). Backwashing by ultrafiltration swing (Kim andChang, 1983a, 1983b) and packed-bed filtration with settling appearsto reduce membrane fouling more effectively than before, allowingthe long-term operation of high cell density membrane bioreactors(Fig. 4).

3.2.3. High purity oxygen-based cell cultureHigh density anaerobic cell culture can be obtained using cell

recyclingwith UF/MFmembranes, but aerobic cultures require high pu-rity oxygen instead of air. By using high purity oxygen instead of air,nearly 4 times of product concentration is obtained for intracellularand extracellular products such as PHB, human growth hormone andhuman serum albumin (Kim et al., 1994; Shang et al., 2009; Younet al., 2010). Because the residence time of high purity oxygen in the fer-menter is short, its consumption is low. If we regenerate this off-gas, theeconomics of the process can improve (Chang et al., 2010b).

Fig. 3. Cell mass production reactor system. This reactor system supplies cells to the product formation bioreactors. The cell mass production reactor (CMP) may receive cells from any ofthe subsequent bioreactors or enough cells may be generated in the CMP itself. The presence of a CMP allows second and subsequent bioreactors to avoid cell washouts because of thecontinual supply of cells. CEU-1: cell mass enriching unit-1, removes water from cell broth by upflow packed-bed, centrifuge or hollow fibers; CEU-2: separation of cells from the broth.

519H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

4. Multistage continuous high cell density culture(MSC-HCDC) systems

4.1. History

To study MSC-HCDC systems, Nishiwaki and Dunn (1997) per-formed simulations of amultistage fermenterwith cell recycling for eth-anol production. Chang et al. (2003) obtained a US patent on two-stageand amultistage HCDCmembrane cell recycle for lactic acid production.In the late 2000s, an experimental two-stage HCDC for animal cellculturewas carried out successfully and a patent for theMSC-HCDC sys-tem was filed and registered in Korea in 2007 (Chang et al., 2008b).Disposer-ground food waste from a collective residence facility suchas an apartment complex was successfully treated in a three-stage con-tinuous stirred tank bioreactor with a packed-bed cell retention devicefor almost one year without extra maintenance (Chang et al., 2010a).Chang et al. (2011a) published a theory of the MSC-HCDC systems pro-posing that fed-batch bioreactors can be replaced with this new systemformuch higher productivity. The application of theMSC-HCDCmethod

Fig. 4. Proposed high cell density culture (HCDC) bioreactor system. Gravity settling and packfibers. Backwashing by pressure and/or ultrafiltration, and sometimes chemical treatment of tseveral years in wastewater treatments. ① Continuous stirred tank reactor (CSTR),② packed-⑥ backwashing based on pressure or ultrafiltration swing (Furusaki et al., 1977; Kim and Chan

for the production of monoclonal antibodies and microbial lipids wasproposed (Chang et al., 2011a; Chang et al., 2011b). Cell retentionwith an upflow packed-bed was proposed (Chang et al., 2008a) andfouling-free animal cell HCDC has been reported (Bonham-Carter,2011). Cell retention with an upflow packed-bed is not perfect, but80–90% separation is possible with a low cost, which can improve pro-ductivity. For aerobic operation, the higher oxygen uptake rate (OUR)resulting from the higher cell mass should be compensated by a higheroxygen transfer rate (OTR).

4.2. Kinetics

Fig. 5 shows a schematic diagram of MSC-HCDC bioreactors withparallel and series connections. The general equations for multi-stagehigh cell density cultures are shown as follows in Fig. 6.

In addition, the system can be simplified to one HCDC reactorfollowed by series of several HCDC reactors. Only the first reactorneeds HCDC operation and the other reactors receive cells from the pre-vious reactor. A washout phenomenon occurs only in the first reactor. If

Back-washing

tank

ed-bed filtration reduce the cell density of the packed-bed effluent and fed to the hollowhe fouled membrane surface, allow long-term operation of membrane bioreactors, up tobed filtration, ③ solids settling zone,④ hollow fiber filtration,⑤ backwashing tank, andg, 1983a, 1983b).

(1)

(2)

Fig. 5. Series-connection showing single and multiple feedings of qij. qij: feed flow rates(i) to j-th reactor, there can bemany feed flows to j-th reactor. Single andmultiple extrac-tions from these reactors are also possible (Chang et al., 2011a).

Fig. 6.General equations for anMSC-HCDC system. AnMSC-HCDC system can receive nu-trients from side feed streams and remove products through an external cell filtration fer-menter; qj is the feed stream and q0 is the main feed flow at the first stage. If the celldensity in a certain stage is lower than Xmax, a value selected by the operator, cells fromthe preceding stage are added. If not, the same amount of cells is discharged to the nextstage (Chang et al., 2011a).

520 H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

the reactors are connected in series, we can use the formula given inFig. 6.

4.3. High productivity and high product titer

The highest productivity will be achieved in the first bioreactor withthe lowest product titer. As the reactor number increases, theproductiv-ity will decrease and the product titer will increase. Inmost biochemicalreactions, cell growth and product formation will be inhibited by theirown products. Thus, as the reactor number increases, this product or in-hibitor concentration will also increase, until the product formation orcell growth ceases. However, monoclonal antibody production is notinhibited by the product protein but by ammonia or lactate. Thus, ifwe can control the formation of these byproducts, we can increase theproduct titer to a value higher than that of fed-batch bioreactors. Thenumber of reactors needed to achieve a given product titer dependson howmuch the product titer in the first reactor differs from the targetproduct titer. There is no appropriate kinetics with which MSC-HCDCsystems can be simulated for products of interest. If their production ki-netics is known in a fed-batch reactor, a prediction inMSC-HCDC is pos-sible, but the kinetics for a high cell density culture may be quitedifferent from those systems with low product titer and low cell mass.

As a rule of thumb, a MSC-HCDC system with two or three bioreac-tors can be a substitute for a fed-batch bioreactor to obtain extracellularand intracellular products of higher product titer withmuch higher pro-ductivity (e.g., 10 times).

5. Applications of MSC-HCDC

5.1. Volatile fatty acids production

Volatile fatty acids (VFAs), such as acetic, propionic, butyric and lac-tic acids, can be produced anaerobically from a variety of biomasses in-cluding organicwaste biomasses, sludge, animalmanure, andmicrobial,plant, and animal biomass sources (Chang et al., 2010c). These acids areconverted to methane, H2 and CO2 biogas by naturally occurring mixedcultures. The methane forming step can be blocked to allow the accu-mulation of VFAs, as shown below.

These processes are conducted naturally in aseptic conditions bymixed cultures. The strategy is to maximize the yields of VFAs at highproduct titer with high productivity by blocking biogas formation asmuch aspossible. The VFAs or their ketone forms can be used to producenormal alcohols by the following reactions:

VFAs (C2\C4) + 2H2 ➔ VFA-primary alcohols (C2\C4)

VFA-salts➔ heat-decomposition ➔ VFA ketone (C3\C7) + H2➔

VFA-secondary alcohols (C3\C7).

Fig. 7 shows a production of VFAs in 4-stage HCDC bioreactors usingfood waste (Chang et al., 2010a; Kim, 2010), which showed an incre-mental increase in product titers with the number of reactors. Table 2shows that the results in MSC-HCDC system were much better thanthose of our previous studies by other investigators (Chang et al.,2010a; Lim et al., 2008a). If the above VFAs are fed to lipid-forming mi-crobial cells, the cells will accumulate biolipids as intracellular products(Fei et al., 2011). The earlier application of VFAs was as electron donorfor denitrification (Lim et al., 2008b). The utilization of VFAs fromwaste biomass in fuels and chemicals will establish a new type of bio-mass conversion platform (Chang et al., 2010c; Wilke et al., 2006).

Biomass Biomass Biomass Biomass

CompressorRoom

Gas bag

VFA Broth

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0

5

10

15

20

25

30

35

40

0 5 10 15 20 25

Yie

ld (

g/g

bio

mas

s)

VF

As

con

cen

trat

ion

(g

/L)

Time (day)

1stage2stage3stage4stageYield (g/g biomass)

a

b

Fig. 7.Volatile fatty acid production in anMSC-HCDC system. (a) Each reactor volume is 1 L and the total volume is 4 L. The reactor is operated at a dilution rate of 0.2/day (feeding rate=0.8 L/day). The pH is controlled at 7–8. The substrate: food waste is 60% carbohydrate and was collected from the KAIST student restaurant every day. The reactions were mixed with gasrecycled from the reactors once a day. (b) The effects of increasing product concentration in each chamber are clearly shown. Themaximal VFA titer was 34.6 g/L and its composition was5:1:5 (acetic, ropionic, and butyric acids, respectively). The productivity= 6.92 g/(L/day). The yield= 0.49 g-VFA/g-feed (dry foodwaste) when the reactor was operatedwith a dilutionrate of 0.2/day.

521H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

5.2. Lactic acid

Fig. 8 (a) shows a simulation of lactic acid production in an 8-stageC-HCDC system. The general trend of product titer P and productivityPD depends on the number of stages, with P rapidly increasing and PD

Table 2Comparison of VFA production in MSC-HCDCs and other systems (Chang et al., 2010a,2010b, 2010c).

Agbogso(2005)

Thanakoseset al. (2003)

Moody(2006)

Lim et al.(2008a,2008b)

Kim(2010)

Raw material Rice straw/CM CS/PM OP/CM FW FWPretreatment Lime Lime Lime Lime Grinding GrindingNum. stages 4 4 4 4 1 4pH (avg) 5.8 6.4 6.0–6.5 5.8 5.5 7.0Temp (°C) 40 40 40 40 35 40PD (g/L/day) 1.69 0.172 0.67 2.65 3.1 6.95TVFA (g/L) 32.4 42.3 16.0 34.5 25 34.6Yield1) 0.286 0.61 0.55 0.118 0.37 0.49AA (wt.%) 44.0 73.6 39.8 39.6 49.2 44.2PA (wt.%) 5.46 4.12 20.4 8.9 23.5 7.2BA (wt.%) 29.9 10.3 16.6 26.8 20.7 46.9Other acids 20.64 11.98 23.2 24.8 6.5 1.7

slowly decreasing as the stage number increased. This is expected be-cause high product inhibition occurs at the latter stages. Fig. 8 (b) com-pares the productivity and titer of various productionmethods for lacticacid fermentation. The productivity in the batch and the fed-batch bio-reactors was similar, while the product titer was quite different. Thehighest productivity of 318.0 g/L/h was obtained at P = 18.7 g/L andD= 17.0/h. A lower D or a longer residence time in a SSC-HCDC systemincreased P from 18.7 to 155.1 g/L, but a SSC-HCDC system alone doesnot attain the P value of the fed-batch system because a single CSTR(continuous stirred tank reactor) cannot follow a multi-CSTR operationthat resembles a PFR (plug-flow reactor) type operation. The 8-stageC-HCDC yielded a PD of 10.8 g/L/h with a P of 212.9 g/L. A more im-proved optimal operation scheme may be developed as further investi-gation onMSC-HCDC is carried out. The SSC-HCDC had the same DwithDt as theMSC-HCDC, showing that SSC-HCDC had a lower P and a lowerPD than those of MSC-HCDC (Chang et al., 2011a; Kwon et al., 2001).

5.3. Ethanol production

Ethanol is a typical high-volume low-value biofuel product.The amount produced in 2010 was 20 billion gallons, worth US$480 billion. There has always been a need for high productivity contin-uous process, if possible, with high product titer. Table 3 shows a few

0

50

100

150

200

250

0 1 2 3 4 5 6 7 8

Co

nce

ntr

atio

n, P

rod

uct

ivit

y

Number of stage

S_in (g/L)

S_out (g/L)

LA (g/L)

P (g/L/hr)

125

210

18.7

79.3

115.1

212.9

2.5 2.47

318

31.7

7.8 10.6

0

50

100

150

200

250

300

350

Batch Fed-Batch SSC-HCDC1 SSC-HCDC2 SSC-HCDC3 MSC-HCSC

LA (g/L)

PD (g/L/h)

a

b

Fig. 8. Simulation of lactic acid production in anMSC-HCDC system. (a): General trend of the concentration and productivity of the fermentation product, lactic acid, in anMSC-HCDC sys-tem (Chang et al., 2011a). (b): Comparison of the values for other production systems: batch, fed-batch, SSC-HCDC1, SSC-HCDC2, and MSC-HCDC.

522 H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

experimental examples of high cell density cultures for ethanol produc-tion using glucose as the substrate. Two experiments (Chaabane et al.,2006; Lee and Chang, 1987) used hollow fiber cell recycling systems,while the last one (Chang et al., 2011a) used packed-bed filtration anddid not obtain an appropriately large cell mass because the filtrationmethod was not suitable for controlling the bleed ratio B.

5.4. Intracellular products

Microbial polyester PHB and polyhydroxyalkanoates (PHAs) areproduced as intracellular products by fed-batch fermentations (Kimet al., 1994; Vaz Rossell et al., 2006). Table 4 showsmicrobial productionof PHB and biolipids in HCDC systems,mainly in fed-batch bioreactors. Ifwe exchange the fed-batch system for aMSC-HCDC system, the bioreac-tor productivity would be more than 5 g/L/h.

Table 3Multistage continuous HCDC for ethanol production (Chang et al., 2011a).

Stages Cell/subs/dilution Conv/titer/PD (PDx)

Lee and Chang (1987) 1 100 − 150/14%/1.3 100/65/85 (0.85)1 100/195/0.36 95/90/32.4 (0.32)

Chaabane et al. (2006) 2 59− 159/ND/0.57 100/71/41 (0..41)Chang et al. (2011a) 2 10/200/0.138 100/95/13.2 (1.32)

Furthermore, if we use a very low-cost substrate such as volatilefatty acids, the cost of lipid production may be under $1.00/kg-lipid(Chang et al., 2013). In both cases, the product yield per substrateweight (PHB or lipid-g/g-glucose consumed for product formation)was 0.3 (Kim et al., 1994). Because some substrate in this productionsystem was consumed in forming the cell mass, the overall yield maybe as low as 0.180 g-lipid/g-glucose where “glucose” here refers tototal glucose consumption for growth and product. The VFA platformhas an advantage over the glucose platform in terms of product forma-tion in terms of cost (Chang et al., 2011b). Empty cell mass after extrac-tion of PHB or lipids can be converted to VFAs that can be used formaking more PHBs or lipids.

5.5. Simulation of MSC-HCDC bioreactors for high titer and productivity

Chang and his associates conducted a number of animal cell cultureexperiments using depth filter perfusion systems (DFPSs) (Kim, 2006;Kim et al., 2006; Lee et al., 2005, 2008a, 2008b). The surface-to-volume ratio of DFPS is 450–600 cm2/cm3 and the cell density is esti-mated to be 107 cells/mL. Table 5 compares the productivity of theDFPS with those of batch, continuous suspension, and dual hollowfiber bioreactors, as well as that of calcium-alginate capsules.

Two different strains of monoclonal antibody-producing recombi-nant were used: CHO 26*-320 cells characterized by a low product

Table 4Intracellular production of PHB and biolipids in an HCDC system (Chang et al., 2011c).

Product Strain Substrate Xmax/PHB (g/L) % saturation Productivity (g/L/h)

Kim et al. (1994) (fed-batch, 2 L) PHB Ralstonia eutropha Glucose 164/121 73.7 2.42Ryu et al. (1997) (fed-batch, 60 L) PHB Ralstonia eutropha Glucose 281/232 82.5 3.14Li et al. (2007) (fed-batch,15 L) Biolipid Rhodosporidium toruloides Glucose 106.5/71.8 67.5% 0.54Fei et al. (2011) (flask culture) Biolipid Cryptococcus albidus VFA 1.2/0.334 27.8 3.4 × 10−4

Chang et al. (2011b)1 Biolipid C. albidus VFA 150/112.5 75 N5

Table 5Comparison of reactor performances in culturing the hybridoma cell-line Alps 25-3 (Ohet al., 1994).

Cell density(106 cells/mL)

Productivity(mg/L/day)d

Long-termculture

Batch suspension 1.5–2.5* 10 NoContinuous suspension 1.0–2.0* 20–30 YesDual hollow fiberbioreactor (DHFBR)

8.54* 205 Yes187a

Ca-alginate capsule 15* 650 No150b

Depth filter perfusionsystem (DFPS) (20 μm)

15* 744 Yes30c

DFPSe 10* 600 81 days

Cell densities:ain extracapillaryspace; bin capsules; cin depth filter matrix;d(mg/day per 1 L of the total working volume); eLee et al., 2008a.⁎Based on the total working volume.

523H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

titer and non-sticky nature of their DUKX origin (Lee et al., 2005,2008b); and 13*-1.00 cells by a high product titer and the sticky natureof their DG44 origin (Lee et al., 2008a). The former strain was used intwo-stageHCDC (DFPS) experiments, attaining an antibody titer similarto that of the batch system. However, the productivity of the two-stagesystemwas approximately 50 times that of the batch system. No mem-brane clogging was observed with the 26*-320 cells, whereas the 13*-1.00 cells clogged the DFPS system. The clogs were dissolved with tryp-sin,whichmade it possible to prolong the production in the single-stageDPFS system for 81 days.

Monoclonal antibody production in an 8-stage MSC-HCDC bioreactorwas simulated based on the data of an average titer of 100 mg/L in asingle-stage DFPS with the CS*13-1.00 strain cells. The product titer in afed-batch bioreactor in a separate experiment was approximately250mg/L. The productivity of the 8 stageMSC-HCDC systemwould be ap-proximately 10 times that of the fed-batch product titer of 250mg/L (Leeet al., 2008a, 2008b; Chang et al., 2011a). The number of cells in the DFPSwas nearly the total number of cells in the DFPS reactor systemmeaningthat few cells were present in the broth of the reactor (Lee et al., 2008b).

Table 6Simulation of mAb production in MSC-HCDC systems (Chang et al., 2011b).

Concentration change Mabs conc. (mg/L) Productivity (g/L/day)

Fed-batch-1 296.2 0.32⋮ ⋮ ⋮10q and1χm 2154.5 2.3210q and2χm 2492.8 2.6910q and3χm 2678.3 2.8910q and4χm 2790.6 3.0110q and5χm 2865.2 3.09⋮ ⋮ ⋮10q and10χm 3033.1 3.27

q: Specific, average productivity of cells; χm: maximum cell density.10q and 10χm = 10 fold Mab conc and 10 fold cell density of Fed-Batch 2.

a Productivity ratio of MSC-HCDC system to Fed-batch 2.

Kim et al. (2006) conducted a fed-batch experiment using theCS*13-1.00 strain and obtained approximately 296 mg/L of antibody,and later, using a different medium, the product titer reached406 mg/L. Table 6 shows the simulation results of two different MSC-HCDC systems based on different batch experiments. The first systemhas a final fed-batch antibody titer of 296.2 mg/L and the other has atiter of 406.4 mg/L. We assumed that an industrial version of this strainhas a 10-fold greater activity than the laboratory version (fed-batch)and increased the cell density from 1-fold (107/mL) to 10-fold(108/mL). Interestingly, these two MSC-HCDC systems yielded morethan 3 g/L of antibody, with antibody titer and productivity of 3.27 g/Land 4.28 g/L/day, respectively. The current highest known antibody ti-ters and productivity are 4.8 g/L and 0.4 g/L/day (Wang, 2011). Thissimulation results show that the MSC-HCDC bioreactor could yieldsuch a high product titer and an approximately 10-fold greater produc-tivity (Chang et al., 2011b).

6. Concluding remarks

MSC-HCDC systems consist of several (approximately 2–6) highcell density bioreactors that can outperform fed-batch productionsystems in terms of productivity and product titer. Realizing MSC-HCDC systems for industrial production depends on our under-standing on the principles of their kinetics and our efforts to solveproblems such as the fouling of cell retention devices andmaintain-ing the long-term stability of microbial strains. The potential appli-cations of MSC-HCDC bioreactors are in the area of producing fuelsand chemicals from biomasses and the mass-producing therapeuticmonoclonal antibodies.

Acknowledgments

This work was financially supported by the Ministry of Oceans andFisheries (contract no. 20131039449). They would like to extend theirsincere appreciation to Dr. Fengbu Bai, for his kind support, advice,and guidance throughout the editing process.

Concentration change Mabs conc. (mg/L) Productivity (g/L/day)

Fed-batch-2 406.4 0.44⋮ ⋮ ⋮10q and 1χm 3181.3 3.43 (7.8 folda)10q and 2χm 3629.5 3.9110q and 3χm 3778.6 4.0710q and 4χm 3850.7 4.1510q and 5χm 3892.8 4.20⋮ ⋮ ⋮10q and 10χm 3973.6 4.28 (9.7 folda)

524 H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

References

Agbogso K. Anaerobic fermentation of rice straw and chicken manure to carboxylic acids.Texas, USA: Texas A&M University; 2005 [Ph.D. Thesis].

Asenjo JA, Marchuk JC, editors. Bioreactor system design. New York: Marcel Dekker; 1995.Bai FW, Zhao XQ, Xu JF. Immobilization technology: cells. Comprehensive biotechnology,

vol. 2; 2011477–89.Bitop AG. http://www.bitop.de/cms/website.php?id=/en/index.htm; 2012.Bonham-Carter J. Factory of the future — analysis of upstream intensification using a cost

of goods model. BIT-ICA (International Conference on Antibody 2011, 0323-25).Beijing, PR China: CNCC; 2011.

Brauer H. Fundamentals of biotechnical engineering. In: Rehm HJ, Reed G, editors. Bio-technology, vol. 2. Weinheim: VCH; 1985.

Chaabane FB, Aldiguier AS, Alfenore S, Cameleyer X, Blanc P, Bildeaux C, Guillouet SF,Roux G, Molina-Jouve C. Very high ethanol productivity in an innovative continuoustwo-stage bioreactor with cell recycle. Bioproc Biosyst Eng 2006;29:49–57.

Chang HN. Multistage continuous high cell density culture. In: Moo-Young M, Butler M,Webb C, Moreira A, Grodzinski B, Cui ZF, Agathos S, editors. Comprehensive biotech-nology. 2nd ed. Amsterdam: Elsevier; 2011.

Chang HN, Furusaki S. Membrane bioreactors: presents and prospects. Adv Biochem EngBiotechnol 1991;44:27–64.

Chang HN, Moo-Young M. Estimation of oxygen penetration depth in immobilized cells.Appl Microbiol Biotechnol 1988;29:107–12.

Chang HN, Yoo I-K, Kim BS. High density cell culture membrane-based cell recycle.Biotechnol Adv 1994;12:467–87.

Chang HN, Chang HN, Kwon SK, Lee, WG, Lee PC, Yoo IK, Lim SJ. Method for manufactur-ing organic acid by high efficiency continuous fermentation, US Patent 6,596,521 B1(July 22, 2003).

Chang HN, Kim BJ, Kang JW, Jeong CM, Kim NJ, Park JK. High cell density ethanol fermen-tation in an upflow packed-bed cell recycle bioreactor. Biotechnol Bioproc Eng2008a;13:123–35.

Chang HN, Kim BJ, Jeong CM, Kang JW, Park JK. Multi-Stage CSTR Bioreactor SystemEquipped with Cell Recycle Unit (Korea patent 10–0834110, 2008b.05.26). PCT WO2008/091113 A1 (US Patent application 12/523.279).

Chang HN, Jang ST, Jung HM, Kang JW, Jeong CM. Method of treating foodwaste at collec-tive residences. Korea Patent 10-0946368(2010a).

Chang HN, Kim MI, Fei Q, Choi JDR, Shang LA, Kim NJ, Kim JA, Park HG. Economicevaluation of off-gas recycle pressure swing adsorption (PSA) in industrialscale poly(3-hydroxybutyrate) fermentation. Biotechnol Bioproc Eng 2010b;15:905–10.

Chang HN, Kim NJ, Jeong CM, Kang JW. Biomass-derived volatile fatty acid platform forfuels and chemicals. Biotechnol Bioproc Eng 2010c;15:1–10.

Chang HN, Kim NJ, Kang JW, Jeong CM, Choi JDR, Fei Q, Kim BJ, Kwon SH, Lee SY, Kim JB.Multi-stage high cell continuous for high productivity and titer. Bioproc Biosyst Eng2011a;34:419–31.

Chang HN, Fei Q, Choi JDR, Jung KS. Economic evaluation of heterotropic microbial lipid(C. albidus) production using low-cost volatile fatty acids in MSC-HCDC bioreactorsystem. BIT’s first annual congress of bioenergy. Dalian, PR China:World Expo Center;2011b. p. 0425–9.

Chang HN, Jung KS, Choi JDR. MSC-HCDC vs fed-batch culture in monoclonal antibodyproduction. BIT-ICA (International Conference on Antibody). Beijing, PR China:CNCC; 2011c 0323–5.

Chang HN, Fei Q, Park GW, Choi JDR, Fu RZ, Sinskey AJ, Kim YC. Economic assessment ofmicrobial lipids for biodiesel production, WCBE-2013 0426 0428. PR China: Nanjing;2013.

Chu L, Robinson DK. Industrial choices for protein production by large-scale cell culture.Curr Opin Biotechnol 2001;12:180–7.

Chung BH, Chang HN. Aerobic fungal cell immobilization in a dual hollow-fiber bioreactor:continuous production of a citric acid. Biotechnol Bioeng 1988;32:205–12.

Chung BH, Chang HN, Kim IH. Rifamycin B production by Nocardia mediterraneiimmobilized in a dual hollow fibre bioreactor. EnzymMicrob Technol 1987;9:345–9.

Cooney CL. Bioreactors, design and operation. Science 1983;219:728–33.Cooney CL, Humphrey AE, editors. The Principles of biotechnology: engineering consider-

ations. In: Moo-Young M, editor-in-chief. Comprehensive biotechnology. Oxford:Pergamon Press; 1985.

Fei Q, Chang HN, Shang LA, Choi JDR, Kim NJ, Kang JW. The effect of volatile fatty acids as asole carbon source on lipid accumulation by Cryptococcus albidus for biodiesel pro-duction. Bioresour Technol 2011;102:2695–701.

Furusaki S, Kojima T,Miyauchi T. Reaction by the enzyme entrapped by the UFmembraneby UF membrane with acceleration of mass transfer by pressure swing. J Chem EngJpn 1977;10:233.

Inloes DS, Taylor DP, Cohen SN, Michaels AS, Robertson CR. Ethanol production bySaccharomyces cerevisiae immobilized in hollow-fiber membrane bioreactors. ApplEnviron Microbiol 1983a;46:264–78.

Inloes DS, Taylor DP, Cohen SN, Michaels AS, Robertson CR. Hollow fiber membrane bio-reactor using immobilized E. coli for protein synthesis. Biotechnol Bioeng 1983b;25:2653–81.

Inloes DS, Michaelis AS, Robertson CR, Matin A. Ethanol production by nitrogen-deficientyeast cells immobilized in a hollow-fiber membrane bioreactor. Appl MicrobiolBiotechnol 1985;23:85–91.

Javamandian M, Palsson BO. High-density photoautotrophic algal cultures: design,construction, and operation of a novel photobioreactor system. Biotechnol Bioeng1991;38:1182–9.

Kim DY. Process development for production of recombinant antibody by rCHO cells.Korea: KAIST; 2006 [PhD Thesis].

Kim NJ. Comparative research of sugar and volatile fatty acid platform for biofuels andbiochemicals production. Korea: KAIST; 2010 [PhD thesis].

Kim IH, Chang HN. Variable volume enzyme reactor with ultrafiltration swing: a theoret-ical study on CSTR case. AIChE J 1983a;29:645–51.

Kim IH, Chang HN. Variable volume hollow-fiber enzyme reactor with pulsatile flow.AIChE J 1983b;29:910–4.

Kim DJ, Chang HN, Liu JR. Plant cell immobilization in a dual hollow fiber bioreactors.Biotechnol Tech 1989;3:139–44.

Kim BS, Lee SC, Lee SY, Chang HN, Chang YK, Woo SI. Production ofpoly(3-hydroxybutyric acid) by fed-batch culture of Alcaligenes eutrophus with glu-cose concentration control. Biotechnol Bioeng 1994;43:892–8.

Kim DY, Lee JC, Chang HN, Oh DJ. Development of serum free media for a recombinantCHO cell line producing recombinant antibody. Enzyme Microb Technol 2006;39:426–33.

Kwon SH, Yoo IG, LeeWG, Chang YK, Chang HN. High-rate continuous production of lacticacid by Lactobacillus rhamnosus in two stage membrane cell recycle bioreactor.Biotechnol Bioeng 2001;73:25–34.

Lee SY. High cell density culture of Echerichia coli. Trends Biotechnol 1996;14:98–105.Lee CW, Chang HN. Kinetics of ethanol fermentations in membrane cell recycle fermen-

ters. Biotechnol Bioeng 1987;29:1105–12.Lee JC, Chang HN, Oh DJ. Recombinant antibody production by perfusion cultures of rCHO

cells in a depth filter perfusion system. Biotechnol Prog 2005;21:134–9.Lee JC, Kim DY, Oh DJ, Chang HN. Long-term operation of depth filter perfusion

systems (DFPS) for monoclonal antibody production using recombinant CHO cells:effect of temperature, pH, and dissolved oxygen. Biotechnol Bioproc Eng 2008a;13:401–9.

Lee JC, Kim DY, Oh DJ, Chang HN. Two-stage depth filter perfusion culture for recombi-nant antibody production by recombinant Chinese hamster ovary cell. BiotechnolBioproc Eng 2008b;13:560–5.

Lee SU, Jung KS, Park GW, Seo C, Hong YK, Hong WH, Chang HNKorean J Chem Eng2012;29:831–50.

Lentzen G, Schwarz T. Extremolyte natural compounds from extremophile for versatileapplications. Appl Microbiol Biotechnol 2006;72:623–34.

Li YH, Zhao ZB, Bai FW. High-density cultivation of oleaginous yeast Rhodosporidiumtoruloides Y4 in fed-batch culture. Enzyme Microb Technol 2007;41:312–7.

Lim SJ, Kim BS, Jeong CM, Choi JDR, Ahn YH, Chang HN. Anaerobic organic acid productionof food waste in once-a-day feeding and drawing-off bioreactor. Bioresour Technol2008a;99:7866–74.

Lim S-J, Kim E-Y, Ahn Y-H, Chang HN. Biological nutrient removal with volatile fattyacids from food wastes in sequencing batch reactor. Korean J Chem Eng 2008b;25:129–33.

McCutcheon JR, McGuinnis RL, Elimelech M. A novel ammonia-carbon dioxide forward(direct) osmosis desalination process. Desalination 2005;174:1–11.

Moody AG. Pilot-scale fermentation of office paper and chicken manure to carboxylicacids. Texas, USA: M.S. Texas A&M University; 2006.

Mulder J. Basic principles of membrane technology. 2nd ed. Amsterdam: Kluwer AcademicPublishers; 1996.

Nielsen J, Villadsen J. Bioreaction engineering principles. New York: Plenum Press;1994.

Nishiwaki A, Dunn IJ. Performance of a multistage fermentor with cell recycle for contin-uous ethanol production. Chem Eng Commun 1997;162:179–98.

Oh DJ, Choi SK, Chang HN. High density continuous cultures of hybridoma cells in a depthfilter perfusion system. Biotechnol Bioeng 1994;44:895–901.

Purwadi R, TaherzadehMJ. The performance of serial bioreactors in rapid continuous pro-duction of ethanol from dilute-acid hydrolyzates using immobilized cells. BioresourTechnol 2008;99:2226–33.

Rittmann BE, McCarty PL. Environmental biotechnology. McGraw-Hill; 2001330.Robertson CR, Kim IH. Dual aerobic hollow-fiber bioreactor for cultivation of Streptomyces

aureofaciens. Biotechnol Bioeng 1985;27:1012–20.Rony PR. Enzyme reactors. J Amer Chem Soc 1974;94:23.Ryu HW, Hahn SK, Chang HN, Chang HN. Production of poly(3-hydroxybutyrate) by high

cell density fed-batch culture of Alcaligenes eutrophus with phosphate limitation.Biotechnol Bioeng 1997;55:28–32.

Shang LA, Tian PY, Kim NJ, Chang HN, Hahm S. Effect of oxygen supply modes on the pro-duction of human growth hormone in different scale bioreactors. Chem Eng Technol2009;32:600–5.

Shuler ML, Kargi F. Bioprocess engineering: basic concepts. 2nd ed. Upper Saddle River:Prentice Hall; 2002.

Thanakoses P, Black AS, Holtzapple MT. Fermentation of corn stover to carboxylic acids.Biotechnol Bioeng 2003;83:191–200.

Thobanoglous G, Burton FL, Stensel HD, Metcalf, Eddy. Wastewater engineering: treat-ment and reuse. 4th ed. New York: McGraw-Hill; 2002 1104–37.

Vaz Rossell CE, Mantelatto PE, Agnelli JAM, Nascimento J. Sugar-based biorefinery —

technology for integrated production of poly(3-hydroxybutyrate), sugar, and ethanol.In: Kamn B, Gruber PR, Kam M, editors. Biorefineries — industrial processes andproduct. Wiley-VCH; 2006. p. 209–24.

Wang DIC. Current challenges in mAb productions ACB-2011 0512-0515. PR China:Shanghai; 2011.

Waterland LR, Michaels AS, Robertson CR. A theoretical model for enzymatic catalysisusing asymmetric hollow fiber membranes. AIChE J 1974;20:50.

Waterland LR, Michaels AS, Robertson CR. Enzymatic catalysis using asymmetric hollowfiber membranes. Chem Eng Commun 1975;2:37–47.

Wikipedia. Membrane bioreactor; 2013, August 22a [Wikipedia.com. Retrieved August31, 2013 from http://en.wikipedia.org].

Wikipedia. Activated sludge; 2013, August 20b [Wikipedia.com. Retrieved August 31,2013 from http://en.wikipedia.org].

525H.N. Chang et al. / Biotechnology Advances 32 (2014) 514–525

Wilke T, Prusse U, Vorlop K-D. Biocatalytic and catalytic routes for the production ofbulk and fine chemicals from renewable resources. In: Kamm B, Gruber PR, KammM, editors. Biorefineries — industrial processes and products. Wiley-VCH; 2006.p. 389–93.

Williams D, Munnecke DM. The production of ethanol by immobilized yeast cells.Biotechnol Bioeng 1981;23:1813–925.

Youn JK, Shang LA, Kim MI, Jeong CM, Chang HN, Hahm MS, Rhee SK, Kang HA.Enhanced production of human serum albumin by fed-batch culture ofHansenula polymorpha with high purity oxygen. J Microb Biotechnol 2010;20:1534–8.

Zhao XQ, Bai FW. Yeast flocculation: new story in fuel ethanol production. Biotechnol Adv2009;27:849–56.