JOURNAL OFBIOSCIENCE AND BIOENFINMRING

Vol.95, No. 1, 13-20.2003

Continuous Plant Cell Perfusion Culture: Bioreactor Characterization and Secreted Enzyme Production
WE1 WEN SU* AND RENEE ARIAS
Department of Molecular Biosciences & Bioengineering, Honolulu, HI 96822, USA University of Hawaii.

Received 16 January 2002iAccepted20 August 2002 Culture perfusion is widely practiced in mammalian cell processes to enhance secreted antibody production. Here, we report the development of an efficient continuous perfusion process for the cultivation of plant cell suspensions. The key to this process is a perfusion bioreactor that incorporates an annular settling zone into a stirred-tank bioreactor to achieve continuous cell/medium separation via gravitational sedimentation. From washout experiments, we found that under typical operating conditions (e.g., 200 rpm and 0.3 wm) the liquid phase in the entire perfusion bioreactor was homogeneous despite the presence of the cylindrical baftle. Using secreted acid phosphatase (APase) produced in Anchusa officinulis cell culture as a model we have studied the perfusion cultures under complete or partial cell retention. The perfusion culture was operated under phosphate limitation to stimulate APase production. Successful operation of the perfusion process over four weeks has been achieved in this work. When A. officinalis cells were grown in the perfusion reactor and perfused at up to 0.4 wd with complete cell retention, a cell dry weight exceeding 20 gll could be achieved while secreted APase productivity leveled off at approximately 300 units/l/d. The culture became extremely dense with the maximum packed cell volume (PCV) surpassing 70%. In comparison, the maximum cell dry weight and overall secreted APase productivity in a typical batch culture were lo-12 g/l and 100-150 units/Z/d, respectively. Operation of the perfusion culture under extremely high PCV for a prolonged period, however, led to declined oxygen uptake and reduced viability. Subsequently, cell removal via a bleed stream at up to 0.11 wd was tested and shown to stabilize the culture at a PCV below 60%. With culture bleeding, both specific oxygen uptake rate and viability were shown to increase. This also led to a higher cell dry weight exceeding 25 g/l, and further improvement of secreted APase productivity that reached a plateau fluctuating around 490 units/l/d. [Key words: perfusion bioreactor, plant cells, protein secretion, high cell density] Higher plants have been shown in recent years to be suitable hosts for large-scale recombinant protein production (1). In fact, crop-based foreign protein production, or molecular farming (2), has become a growing enterprise with demonstrated successes in producing a variety of products including antibodies, therapeutic proteins and industrial enzymes. For producing specialty protein products with high values, plant cell cultures have been proposed as an appealing alternative to transgenic plants growing in open fields (24). The main advantages offered by plant cell cultures include faster growth rates than their whole-plant counterparts, their ability to grow using simple and inexpensive media in a well-controlled environment such as a bioreactor, and their capacity in complex posttranslational processing (2-4). Furthermore, GM0 issues associated with field-grown transgenic crops are avoided with the plant-cell-culture production systems. In addition to these more genera1 points, de Wilde et al. (5) pointed out a less obvious distinction between bioreactor-cultured plant cells and whole plants. * Corresponding author. e-mail: wsu@hawaii.edu phone: +I-808-956-3531 fax: +l-808-956-3542
I3

These researchers suggested that transgene silencing, which is triggered in a limited number of cells, might not spread throughout the bioreactor (5). In whole plants, the signal for systemic acquired silencing is believed to be transmitted through the plasmodesmata and the vascular tissue, whereas cultured plant cells are not interconnected in that way (5). de Wilde et al. (5) further reasoned that cultured plant cells are commonly grown in a hemizygous state, which could be advantageous because in some cases transgene silencing is triggered only when the transgenes are present in a homozygous state. In our study of green fluorescent protein (GFP) expression (6), as well as in a study of P-glucuronidase (GUS) expression (7), both using tobacco suspension cultures, the transgene silencing phenomenon was not observed. To improve downstream recovery efficiency, it is desirable to allow secretion of the target protein into the surrounding medium. Cultures operated at a high cell density using a perfusion system can potentially increase the productivity of secreted products (8, 9). In a perfusion system, cell-free spent medium containing secreted products is constantly removed and harvested, while the culture is simulta-

14

SU AND ARIAS

J. BIOSCI. BIOENG.,

neously replenished with fresh medium. There have been only a handful of published studies on perfusion plant cell cultures (10-13). Findings from these studies, along with other aspects of perfusion plant cell cultures have been reviewed by Su (8, 9). Previously, we reported a perfusion bioreactor design in which cell retention was accomplished by incorporating an internal cell-settling zone into the downcomer region of an external-loop air-lift bioreactor (14). Using this reactor, we were able to achieve a two-fold increase in cell concentration, over batch cultures, with continuous perfusion ( 14). In an attempt to further enhance the perfusion bioreactor performance, the same concept for cell retention was applied to a stirred-tank bioreactor in this study. Stirred-tank type reactors are generally superior to air-lift reactors in their mixing and mass transfer capacity, especially under high cell densities. For the perfusion bioreactor described in this work, an annular settling zone was incorporated into a stirred-tank reactor by inserting a cylindrical baffle. Here, we report on the liquid mixing characteristics of this stirredtank perfusion bioreactor, and show that such a reactor could be used for the continuous perfusion cultivation of plant cell suspensions. Our ultimate goal is to use the system for the efficient production of secreted recombinant proteins. In this pilot study, the production of a non-recombinant protein, the secreted acid phosphatase (APase) from A. officinalis cell suspension culture, was chosen as a model to investigate the bioreactor performance and the perfusion process. APase was also selected because of the potential biotechnological applications of its promoter for inducible expression of recombinant proteins. APase is the most abundant enzyme found in the extracellular compartment of A. officinalis cultures (Liang, Ph.D. thesis, Univ. of Hawaii, Honolulu, 1998). The promoter and putative signal sequence of a member of the Anchusa APase multigene family have been cloned in our laboratory (Liang, Ph.D. thesis, Univ. of Hawaii, Honolulu, 1998). An APase promoter system that is inducible by phosphate starvation has also been cloned from Arabidopsis by Hat-an et al. (15). By understanding the kinetics of Anchusa APase synthesis and secretion in a perfusion process, this study also provides information useful for applying the APase promoter in conjunction with high-density perfusion cultures to improve the production of high-value recombinant protein products by transgenie plant cells.

natant was filtered through a 0.2~pm membrane filter and then assayed using a protein assay reagent (Bio-Rad, Hercules, CA, USA) with bovine serum albumin as the standard (19). For intracellular phosphate measurement, plant cells were ruptured by sonification, followed by extraction of free ortho-phosphate into deionized water. For extracellular phosphate measurement, the culture supernatant was collected by centrifugation. The molybdate method (20) was used to determine phosphate concentrations. Oxygen uptake rate (yo) was measured on-line using the dynamic method (21). Acid phosphatase activity was measured by a colorimetric method usingp-nitrophenyl phosphate as the substrate (22). Cell viability was assessed by detecting the cell respiratory effrciency using 2,3,5triphenyltetrazolium chloride (TTC) according to a modified protocol developed by Su and Arias (unpublished). By counting the formazan-producing cells, rather than measuring the total formazan produced, the modified TTC method circumvents the major drawbacks of the traditional TTC reduction assay (23), namely the heterogeneity of formazan production in a cell population and the inconsistency in formazan extraction efficiency. Dissolved oxygen (D.O.) was measured using a polarographic oxygen electrode (Ingold, Lemexa, KS, USA). Northern analysis was performed using total RNA isolated from cultured A. oJkinaZis cells. Here, RNA extraction and gel blotting procedures similar to those described elsewhere (6) were used, except that a 1.2-kb Anchusa APase cDNA clone (AP32; Liang, Ph.D. thesis, Univ. of Hawaii, Honolulu, 1998) was used to synthesize the 32P-1abeled DNA probes. Sugars (sucrose, glucose, and fructose) were measured by HPLC using a Supelcosil LC-NH2 column (i.d., 4.6 mm; length, 25 cm; cat no. 58338; Supelco, Bellefonte, PA, USA). Separation was achieved using an aqueous solvent with 80% acetonitrile at a flow rate of 1 ml/min. The perfusion stirred-tank reactor Perfusion bioreactor (PSTR) system is presented schematically with the major dimensions in Fig. 1A (note the reactor has a dished bottom). This glass bioreactor uses a New Brunswick Scientific BioFlo III head-plate

(4

Cell Bleed

MATERIALS AND METHODS

The routine culture of A. ofJicinalis cell susCell culture pensions has been described elsewhere (14). Briefly, stock cultures of A. of$cinalis were maintained as cell suspensions in a liquid Gamborg B5 medium supplemented with 1 mg/l2,4-dichlorophenoxyacetic acid (2,4-D), 0.1 mg/l kinetin, and 30 g/Z sucrose (16). The suspension was maintained on a gyratory shaker at 120 rpm and 25C and subcultured every 10 d using a 10% inoculum. Cultured A. oJ?kinaZis cells formed tine suspensions with very few large cell aggregates. Assays Analytical procedures for cell dty weight, medium osmolarity, and packed cell volume (PCV) have been described
elsewhere (17, 18). For the measurement of total extracellular protein concentration, the culture sample was centrifuged, the super-

Eb
15 cm t 14 cm 13cm+ +

FIG. 1. (A) Schematic diagram of the PSTR with major dimensions; (B) A. officinalis cells cultured in the PSTR (note the wavy cell sediment in the bottom of the reactor).

VOL. 95,2003 and it has a working volume of 3.3 1 (including the settling zone). A critical component of the reactor is a cylindrical baffle that allows the creation of a stagnant zone within the reactor. This design concept has been used in a perfusion reactor system for culturing mammalian cells (24). Compared with plant cells, mammalian cells are smaller in size and they are not as prone to cell aggregation, and as a result, cell/medium separation with simple sedimentation was effective only under relatively low perfusion rates (24). Gravitational sedimentation nonetheless can be a simple and effective way for achieving perfusion in plant cell culture, as indicated in our previous study (14). The volume of the annular stagnant (settling) zone in the PSTR is 0.6 1. Pressure equilibration between the well-mixed and the settling zones is ensured by connecting the headspaces of the two compartments. Continuous cell-medium separation in the perfusion reactor is achieved by simple gravitational sedimentation. The reactor was jacketed for temperature control. A six-bladed Rushton turbine (on top) and a three-bladed upward pumping marine axial impeller (on the bottom, model E01919-30; Cole Parmer, Vernon Hills, IL, USA) were used for agitation while a sintered glass tube with a mean pore opening of about 140 pm was used as a sparger placed below the lower impeller. The delivery of perfusion medium and culture bleeding were done using ultra-low flow peristaltic pumps (model 77; Harvard Apparatus, Holliston, MA, USA). Cultivation of A. officinalis in the perfusion bioreactor The bioreactor containing B5 medium supplemented with 1 mgil 2,4-D, 0.1 mgil kinetin, and 30 g/l sucrose was inoculated with a 1O-day-old culture of A. officinalis. For the perfusion runs, the culture was allowed to grow in the batch mode until the extracellular APase activity approached 1200-l 500 U/l (which is the maximum APase activity encountered in a typical batch culture). The culture was then perfused using B5 medium (containing 0.5 mM phosphate) supplemented with 3% sucrose and reduced hormone concentrations (0.1 mg/l 2,4-D and 0.01 mgil kinetin). This perfusion medium (except for the reduced phosphate concentration) was used in our previous study to achieve high cell density in A. of$ccinalis perfusion culture carried out in an external-loop air-lift bioreactor (14). The agitation and aeration rates were set at 200 rpm and 0.3 vvm, respectively. To maintain D.O. at 30% air saturation, a PID algorithm was implemented using the LabVIEW programming language (National Instrument, Dallas, TX, USA) to adjust the air/O? ratio in the gas inlet via two mass-flow controllers (Omega Engineering, Stamford, CT, USA) connected to a data acquisition board (National Instrument) supervised by a PC. Foam control was achieved using a silicone-based antifoam solution (Antifoam C emulsion, Sigma Chemical). The culture temperature in the bioreactor was kept at 25C. Characterization of liquid-phase mixing in the perfusion bioreactor Mixing in the perfusion bioreactor was characterized using washout experiments. Phosphate was chosen as the tracer in these experiments primarily because it was the limiting substrate in the perfusion culture medium and it can be measured easily. The reactor was initially charged with 1 mM aqueous phosphate solution. At the onset of the experiment, a continuous water stream was pumped into the reactor with or without the cylindrical baffle at a preset rate using a peristaltic pump (model 77; Harvard Apparatus) while another stream was continuously removed via gravitational overflow through a site port. In another set of experiments, the effluent stream was removed from within the wellmixed zone using a peristaltic pump identical to that used for feeding the input stream into the well-mixed zone. The two pumps were carefully calibrated to ensure delivery of indistinguishable flows under the same setting within the time frame of each experiment. Agitation and aeration rates during these washout experiments were set at 200 rpm and 0.3 vvm, respectively. The temperature in the bioreactor was kept at 25C. The decrease in phosphate

PERFUSION PLANT CELL CULTURE concentration at different locations during the course of the experiments.

15

in the reactor was followed

RESULTS Liquid-phase mixing

AND DISCUSSION in the perfusion bioreactor In

the PSTR, cell/medium separation during culture perfusion is achieved by incorporating a cylindrical baffle into the reactor. By incorporating this baffle an annular stagnant region is created inside the reactor to serve as a settling zone. It is thus necessary to characterize to what extent the baffle affects the overall mixing in the bioreactor. While we have observed from culture experiments that the annular settling zone was essentially free of cells during the perfusion operation (as long as the PCV was kept below 60%), it was unclear how liquid-phase mixing was affected by the presence of the cylindrical baffle. To answer this question, we characterized the residence time distribution (RTD) in the reactor by conducting a series of washout experiments. Medium perfusion was anticipated to cause a higher degree of bulk liquid mixing in the settling zone. As such, two sets of washout experiments were performed - one for simulating the batch operation and the other for the perfusion operation. To simulate reactor washout during the batch phase of the bioreactor operation, the eMuent stream was withdrawn from within the well-mixed zone of the bioreactor, so that the settling zone was not disturbed. Typical washout data from such experiments are presented in Fig. 2. To obtain these washout data, liquid samples were withdrawn from various locations in the reactor as indicated in the figure insert. When collecting these samples, special care was taken to minimize the disturbance to the liquid phase. These data indicate a very uniformed concentration distribution throughout the reactor, and the data match the ideal CSTR model, suggesting some degree of convective mass transfer would have to be considered in the settling zone even during the batch phase of the reactor operation. This finding was not influenced by the washout rates within the range tested (0.02-0.5 h-). In fact, during routine culture experiments, we have repeatedly observed wavy motion at the interface of the liquid medium and culture sediment at the en-

00

0.5

1.0 VT(-)

15

2.0

25

FIG. 2. Tracer (phosphate) washout during simulated batch operation (i.r., effluent stream withdrawn from the well-mixed zone). C, Tracer cont.; C,,, initial tracer cont.; t, time; r, residence time. Experimental data and ideal CSTR model simulation are represented by the symbols and the solid line, respectively.

16

SU AND ARIAS

BIOENG., J. BIOSCI.
1500

trance of the annular settling zone (note the bottom of the reactor as shown in the photo in Fig. 1B). During the perfusion phase of the reactor operation, effluent exited from the settling zone through an overflow tube. It is therefore not surprising that the washout curves are essentially identical in the presence or absence of the cylindrical baffle, and the washout data match the ideal CSTR model, even at a washout rate as low as 0.02 h- (data not shown). Results from these washout experiments suggest that the liquid phase in the 3.3 I perfusion reactor could be considered homogeneous under normal operating conditions despite the presence of the cylindrical baffle. This finding is useful for further analysis of the bioreactor in which dissolved substrate/product concentrations in the liquid phase can now be assumed to be uniform throughout the reactor. Characteristics of acid phosphatase expression In this study, APase was chosen as a model secreted protein primarily for two reasons. First, APase is the major secreted protein in the A. ojkinalis culture, as confirmed by electrophoresis and protein sequencing analysis (Liang, Ph.D. thesis, Univ. of Hawaii, Honolulu, 1998). Furthermore, the APase promoter may be used for effective inducible gene expression. We have cloned the promoter and putative signal sequence of a member of the APase multi-gene family from a cell culture of A. officinalis (Liang, Ph.D. thesis, Univ. of Hawaii, Honolulu, 1998). In addition, Haran et al. (15) recently demonstrated inducible reporter gene expression using an Arabidopsis APase promoter under phosphate starvation. Extracellular acid phosphatases appear to be ubiquitous in roots and plant cell cultures. These extracellular APases may be localized either within the cell wall or secreted to the surrounding environment (25). In cultured tobacco cells, it was found that the synthesis and secretion of an APase were regulated by phosphate (26). In tomato suspension cultures, APase is also secreted and is suggested to function either as a phosphate transport agent or a phosphate-scavenging agent that acts on the phosphorylated compounds in the culture medium (27). When linked to the Arabidopsis APase promoter and signal sequence, a GFP reporter was shown to be secreted by the roots of transformed Arabidopsis plants under phosphate starvation (15). To establish a nutrient (phosphate) regime that supports APase production in the perfusion bioreactor, we first characterized the effect of phosphate on APase expression in shake-flask cultures. We found that the Anchusa APase is encoded by a gene family the members of which share more than 85% homology (Liang, Ph.D. thesis, Univ. of Hawaii, Honolulu, 1998). A 1.2-kb cDNA fragment that covered the 3 part of the full-length gene encoding one of the Anchusa APase isoenzymes (AP32) was used to prepare a 32P-labeled DNA probe. Northern analysis was performed using total RNA isolated from 8-day-old A. of$cinaZis cells grown in medium containing either 0.5, 1 or 4mM phosphate. As shown in the insert of Fig. 3, the hybridization signals obtained with the AP32 probe showed a marked difference in the mRNA levels for the acid phosphatase. The RNA from cells growing in medium with higher levels of phosphate (4 mM) had a very weak hybridization signal whereas the RNA from cells growing in lower levels of phosphate (0.5

m
0

0.5 Ink4 4.0 mM

1000 Q L

$
500 $ % 0 DW APase FIG. 3. Effect of phosphate on cell dry weight, APase transcript level, and intracellular APase activity. Insert: Northern blot of APase transcript from 8-day-old A. officinalis cells cultured in medium containing either 0.5, 1 or 4 mM phosphate.

and 1 mM) had a very strong hybridization signal, indicating that the expression of the acid phosphatase genes in A. of$cinalis cell culture is stimulated by low phosphate concentration. Specific intracellular APase enzyme activity was also inversely correlated with the phosphate concentration as shown in Fig. 3 while the biomass yield was increased by about 50% when the initial phosphate concentration was increased from 0.5 to 4 mM. Cell growth and acid phosphatase production in the perfusion bioreactor One of the main objectives in perfusion cultivation is to achieve high cell density as a means to increase reactor volumetric productivity. The first set of perfusion reactor experiments was therefore conducted with total cell retention to minimize cell loss. We then examined the effect of culture bleeding in another set of experiments. To stimulate APase production, all cultures were operated under phosphate-limited growth. At an initial phosphate concentration of up to 1 mM, the Anchusa culture was shown to be under phosphate limitation based on shake flask experiments (Smith, M. S. thesis, Univ. of Hawaii, Honolulu, 1997). The initial phosphate concentration in the medium for all perfusion culture experiments was thus set at 0.5 mM. Typical results from perfusion experiments in the absence of culture bleeding are presented in Fig. 4. In this and subsequent figures, extracellular and intracellular phosphate concentrations, and extracellular and intracellular APase activities are denoted by P,,, P,,, APase,,, and APase,,, respectively. The goal here was not to achieve steady states and study culture behavior under such conditions. Instead, our aim was to achieve the highest possible cell concentration under phosphate limitation. Accordingly, the perfusion rate was adjusted empirically throughout the culture with the aim of sustaining cell growth while keeping a low intracellular phosphate level to stimulate APase production. In a typical batch culture using a standard B5 medium (containing 1 mM phosphate), the cell dry weight reached about 12 g/l (Smith, M.S. thesis, Univ. of Hawaii, Honolulu, 1997). With culture perfusion, one can achieve and maintain a higher cell density. As shown in Fig. 4, the perfusion culture became very thick (with a PCV exceeding 70%) after about 17 d of cultivation. Once the culture reached such a high PCV, the cells began to enter the annular settling zone,

VOL. 95, 2003

PERFUSION

PLANT CELL CULTURE

17

c30
225 3 20 0.8 F ." g 15 -0.6 G 5 e, -0.4 2 2 Z

s
0.4 ho

&IO
5 0

IO

15

20

25

30

0.0 35 12

80 60 g 40 jg 20
IO
2X perfusion medium

% -0.2 a

15

20

25

30

35

.o.o

0.8

B
F 600 "c ,g 500 .> 2 400 e a 5 300 H 2 200 100 IO z 3.6 v 5 aJ $ B 0.4 g 3 8 'Z 0.2 Lg

10
I I

15

20

25

30

35

s2 1500
~~1000 % 500 0

15

20

r____-_--------------,J
15 20 25 30 : 0.0

Time (d)

FIG. 4. Time course of perfusion culture with complete cell retention. The first and second dashed lines are referring to onset of perfusion and switching to 2x medium, respectively (cf: Fig. 5A and see text for details).

Time (d) FIG. 5. Time course of secreted APase productivity (circles) with corresponding perfusion (solid line) and bleed rates (dashed line): (A) complete cell retention; (B) perfusion with cell bleeding.

which led to a gradual decline in the PCV from days 21 to 25. From day 25, the perfusion rate was halved (from 0.4 to 0.2 vvd) and we switched to a 2 x perfusion medium (Fig. 5A). This was carried out to prevent cell washout while providing a similar level of nutrients to the cells. The upward trend of the PCV resumed after day 25. The cell dry weight followed a similar pattern to the PCV during the batch phase of the culture. However, during the perfusion phase, the change in cell dry weight was subtler and the peak observed for the PCV around days 18-2 1 was not seen with the dry weight. Cell dry weight reached and remained at approximately 18 +2 g/l from days 13 to 2 1 during which the settling zone was practically cell free. A small but steady increase in cell dry weight was observed from day 25 onwards, which is consistent with the increased PCV and oxygen uptake rate during that period. The maximum cell dry weight reached approximately 22 g/l in the perfusion culture without cell bleeding. This is a little lower than that achieved previously in a perfusion air-lift bioreactor in which the cell dry weight reached 27 g/l (14). It should be noted, however, in that previous study the culture was not operated under any nutrient limitation. The phosphate concentration in the medium remained essentially undetectable from day 2 onwards despite medium perfusion being initiated on around day 11. The exceedingly fast phosphate uptake by the cultured Anchusa cells corresponds to a very high maximum specific phosphate uptake rate of approximately 5 1 mg phosphate/g dry weight/d as determined in our previous study (28). The intracellular free phosphate content also stayed at a very low (below 1 mg P/g

DW) and constant level throughout the perfusion phase, indicating a rapid turnover of the absorbed free phosphate into phosphorylated compounds (29). As such, the medium perfusion strategy implemented met the phosphate limitation requirement. By comparing the phosphate profiles with the APase production in Fig. 4, it is apparent that regulation of APase expression by phosphate starvation is governed by intracellular rather than extracellular phosphate levels. The initiation of medium perfusion caused a sharp decrease in the extracellular enzyme activity due to the dilution effect. An increase in extracellular APase activity was nevertheless observed about 2 d afterwards. Although no exact steady state of APase production was observed during the perfusion culture, by taking into account the changes in perfusion rates, the volumetric extracellular APase productivity from day 20 onwards was found to fluctuate around 300 U/lid (f30 U/l/d; Fig. 5A), which is about twice the productivity seen in batch cultures (Smith, MS. thesis, Univ. of Hawaii, Honolulu, 1997). Note that between days 20 and 25, there appeared to be a metabolic shift that led to declined oxygen uptake and intracellular APase activity, despite stepwise increases in perfusion rates from 0.2 to 0.4vvd during this period (Fig. 5A). Judging from the medium osmolarity (data not shown) which dropped to 18 mOSM/l (regular fresh medium has an osmolarity reading of 220 mOSMII) on day 18, the perfusion rate prior to day 19 might have been set too low which led to excessive nutrient starvation. Together with the possible stress caused by the extremely high PCV (the population crowding effect), it could have caused the

18

SU AND ARIAS 35 e30

325 520 'g 15 g10 5 00 5 10 15 20 25 30

J.Brosa. 5
4 3g 2 B Lp 1 35' 12 g 52 3s g

BIOENG.,

reduced metabolic activities seen from days 20 to 25. Also during this period, an increased amount of cells began to enter the settling zone due to the high cell loading (a PCV exceeding 70%). We noted that for the Anchusa culture, the minimum PCV that corresponds to a 100% settled cell volume (the culture biotic volume fraction after the cells are allowed to settle for a sufficient period under gravity) is about 70% (14). Therefore, it is not surprising that as the culture PCV reached 70% in the well-mixed zone the cells started to get pushed into the settling zone. The observation of an extremely high PCV and reduced metabolic activity during the perfusion culture under complete cell retention prompted us to look into ways to alleviate the potential cellular stresses created by the highly crowded culture environment. In microbial chemostats with cell recycling or in perfusion mammalian cell cultures, manipulation of the cell bleed rate has been identified as a viable operating strategy to manipulate specific growth rates and to reduce dead cell accumulation (30). To study how cell removal affects the performance of the plant cell perfusion culture, the perfusion bioreactor was operated with a bleed stream. The results are presented in Figs. 5-7. The cell bleed rate was determined based on a conservative estimation of the cell specific growth rate, and was initially set at 0.04 vvd and later increased to 0.11 vvd (Fig. 5B). The perfusion rate was first set at 0.2 vvd, increased to 0.4 vvd on around day 20 and then adjusted back to 0.2vvd on about day 25 (Fig. 5B). Under these conditions, improved culture performance in terms of a higher cell dry weight, improved culture viability (data not shown), increased extracellular phosphatase production, as well as augmented oxygen uptake and a lower PCV were noted (Fig. 6). Similar to the perfusion culture with complete cell retention, no apparent steady state with respect to extracellular APase productivity was reached; rather the productivity fluctuated between 400 and 600 U/l/d, and averaged around 490 U/l/d from day 20 onwards (Fig. 5B). This productivity is approximately 60% higher than that achieved in perfusion culture with complete cell retention. With cell removal, the PCV could be kept below 60% while the maximum cell dry weight exceeded 25 gll. The increase in oxygen uptake relative to that seen in the case of complete cell retention was also quite marked. A closer examination of the oxygen uptake rate data revealed that for both perfusion cultures the specific oxygen uptake rate (SOUR, oxygen uptake rate divided by cell dry weight) peaked in early to mid exponential growth during the batch phase of the cultures. The SOUR then declined and fluctuated around a steady level (approximately 0.12 and 0.05 g 0,/g DW/d for partial and complete cell retention, respectively) from day 14 onwards during the perfusion phase. Due to the relatively constant SOUR, the course of oxygen uptake closely paralleled that of cell dry weight during the perfusion phase of the culture under either complete or partial cell retention. Oxygen uptake rate is therefore potentially useful as a metabolic indicator for estimating cell growth online in a perfusion culture. Moreover, since the SOUR did not increase after culture perfusion was commenced, it may be possible to further improve the culture performance by initiating the perfusion earlier while the culture is metabolically more active (e.g.,

-60 -60

7 @ -40 > z

&&

-20

10

15

20

25

30

:i5

Time(d)

FIG. 6. Time course of perfusion culture with cell bleeding. The dashed lines represent changes in perfusion rates (cf: Fig. 5B and see text for details).

as indicated by the high ATP production rate and SOUR). We also examined the carbon source utilization during the continuous perfusion cultivation with cell bleeding. During perfusion, sucrose fed into the reactor was rapidly converted to glucose and fructose by hydrolysis, while glucose was preferentially utilized by the Anchusa cells (Fig. 7). By comparing Figs. 5B and 7, it is clear that the perfusion rates had a dominant effect on the sugar profiles. Moreover, medium osmolarity closely mimicked the concentration of the hydrolyzed sugars (Fig. 7). This observation is consistent with our previous study in which an A. officinalis culture was grown in a stirred tank bioreactor operated under discontinuous perfusion via in situ filtration (31). Hence, medium osmolarity might serve as a convenient

100

:o

.o ti
E 6

IO

15

20

25

30

-100

Time(d)

FIG. 7. Time course of sugar concentrations during perfusion culture with cell bleeding (cf Fig. 6).

VOL. 952003

PERFUSIONPLANTCELLCULTURE

19

process parameter for monitoring sugar concentrations in perfusion cultures. Also noted from these data was the presence of a large amount of residual sugars (glucose and fructose) in the medium. Thus decreasing the sugar concentration in the fresh perfusion medium might be considered, although operating at a very low level of residual sugar may have an adverse effect on the culture due to the low medium osmotic pressure. Cell retention and culture mixing in the perfusion The cell retention efficiency of the bioreactor bioreactor was mainly affected by the cell loading, perfusion rate, and the agitation and aeration rates (8). In our previous study (14) where we tested an external-loop air-lift perfusion bioreactor, we observed that it was especially important to alleviate gas bubble entrapment in the settling zone to assure proper reactor performance. Air bubbles not only disturbed cell sedimentation but also carried cells upwards, resulting in excessive biomass accumulation and wall growth above the settling zone. In air-lift reactors, aeration and agitation are coupled. As a result, as mixing becomes more demanding at high cell loading, a higher aeration rate has to be used which results in a higher gas holdup and the bubbles are more prone to escape into the settling zone. This problem is alleviated in a stirred-tank type reactor (such as the PSTR) in which aeration and agitation are decoupled (i.e., they can be controlled independently). Moreover, we incorporated two design considerations to minimize the presence of bubbles in the settling zone of PSTR. First, the cylindrical baffle was extended so that the lower edge of the baffle leveled with the bottom impeller. Second, an upward-pumping axial impeller was used as the bottom stirrer. In this study, it was noted from the washout experiments that while we observed a homogeneous liquid phase in the perfusion bioreactor, mixing of the cell sediment below the cylindrical baffle was clearly very sluggish. After extended operation, the cell sediment apparently became more compact and it could not be readily re-suspended without a substantial increase in the reactor agitation rate that could result in excessive turbulence in the bulk (the well-mixed zone) of the reactor. This design shortcoming could potentially be overcome by installing a separate mixer system for the bottom part of the reactor (e.g., using a magnetically driven stirrer) operated at a lower rpm than the stirrers used to mix the culture in the bulk of the reactor. This study has demonstrated the usefulness of a modified stirred-tank bioreactor for the continuous perfusion cultivation of plant cells. This perfusion bioreactor system shows great potential as an effective alternative to immobilized plant cell bioreactors. The results indicate that it is desirable to operate the perfusion culture with cell bleeding. Further improvement of the culture performance could be achieved by optimizing the perfusion and bleed rates. We suspect that by operating the culture under extremely high packed cell density and prolonged nutrient starvation could lead to excessive physiological stresses that cause culture deterioration. Thus, to further extend and improve the operation of perfusion culture, more study is necessary to better understand these cellular stresses, and ways to alleviate them. Such studies are currently underway in our laboratory.

ACKNOWLEDGMENT The authors are grateful to Hua Liang for conducting the Northem blot analysis, and to Charles Nelson for his excellent technical assistance in constructing the reactor employed in this study. This
work was supported in part by the National Science Foundation (grant no. BES97-12916).

REFERENCES
Giddings, G.: Transgenic plants as protein factories. Curr. Opin. Biotechnol., 12,450454 (2001). Fischer, R. and Emans, N.: Molecular farming of pharmaceutical proteins. Transgenic Res., 9,279-299 (2000). Doran, P. M.: Foreign protein production in plant tissue cultures. Curr. Opin. Biotechnol., 11, 199-204 (2000). James, E. and Lee, J. M.: The production of foreign proteins from genetically modified plant cells. Adv. Biochem. Eng. Biotechnol., 72, 127-156 (2001). de Wilde, C., van Houdt, H., ve Buck, S., Angenon, G., ve Jaeger, G., and Depicker, A.: Plants as bioreactors for protein production: avoiding the problem of transgene silencing. Plant Mol. Biol., 43, 347-359 (2000). Liu, S., Bugos, R., Dharmasiri, N., and Su, W. W.: Green fluorescent protein as a secretory reporter and a tool for process optimization in transgenic plant cell cultures. J. Biotechnol., 87, I-16 (2001). Gao, J., Lee, J. M., and An, G.: The stability of foreign protein production in genetically modified plant cells. Plant Cell Rep., 10, 533-536 (1991). Su, W. W.: Perfusion bioreactors, p. 230-242. In Spier, R. E. (ed.), Encyclopedia of cell technology, vol. 1. Wiley, New York (2000). Su, W. W.: Bioprocessing technology for plant suspension cultures. Appl. Biochem. Biotechnol., 50, 189-230 (1995). Pareilleux, A. and Vinas, R. A.: A study on the alkaloid production by resting cell suspensions of Catharunth~s roseus in a continuous flow reactor. Appl. Microbial. Biotechnol., 19, 316-320(1984). Kim, D. I., 00, G. H., Pedersen, H., and Chin, C. K.: A hybrid bioreactor for high density cultivation of plant cell suspensions. Appl. Microbial. Biotechnol., 34, 726-729 (1991). Shin, M. K., Park K., and Cho G. H.: Ultrasonic cell separator as a cell retaining device for high density cultures of plant cell. Biotechnol. Bioprocess Eng., 4, 264-268 (1999). Su, W. W. and Humphrey, A. E.: Production of plant secondary metabolites from high density perfusion cultures, p. 266-269. In Furusaki, S., Endo, I., and Matsuno, R. (ed.), Biochemical engineering for 2001. Springer-Verlag, Tokyo (1992). Su, W. W., He, B. J., Liang, H., and Sun, S.: A perfusion air-lift bioreactor for high density plant cell cultivation and secreted protein production. J. Biotechnol., 50, 225-233

5.

6.

7.

8.

9. 10.

Il.

12.

13.

14.

(1996).
15. Haran, S., Logendra, S., Seskar, M., Bratanova, M., and Raskin, I.: Characterization of Arubidopsis acid phosphatase promoter and regulation of acid phosphatase expression. Plant Physiol., 124, 615-626 (2000). 16. De-Eknamkul, W. and Ellis, B. E.: Rosmarinic acid production and growth characteristics of Anchusa officinalis cell suspension cultures. Planta Med., 51, 346-350 (1984). 17. Dixon, R.A.: Isolation and maintenance of callus and cell suspension cultures, p. I-20. In Dixon, R. A. (ed.), Plant cell cultures: a practical approach. IRL Press, Oxford (1985). 18. Su, W. W., Lei, F., and Su, L. Y.: Perfusion strategy for rosmarinic acid production by Anchusa officinalis. Biotechnol. Bioeng., 42, 884-890 (1993). 19. Bradford, M. M.: Rapid and sensitive method for the quanti-

20

SU AND ARIAS

J. Broscr. BIOENG.,

20.

21.

22.

23.

24.

25.

tation of microgram quantities of protein utilizing the principle of protein dye binding. Anal. Biochem., 72, 248-254 (1976). Taras, M.: Standard methods for the examination of water and waste water. American Public Health Association, New York (1971). Taguchi, H. and Humphrey, A.: Dynamic measurement of the volumetric oxygen transfer coefftcient in fermentation systems. J. Ferment. Technol., 44, 881-889 (1966). Nahas, E., Terenzi, H. F., and Rossi, A.: Effect of carbon source and pH on the production and secretion of acid phosphatase and alkaline phosphatase in Neurospora crassa. J. Gen. Microbial., 128,2017-2021 (1982). Towill, L. E. and Mazur, P.: Studies on the reduction of 2,3,5-triphenyltetrazolium chloride as a viability assay for plant tissue cultures. Can. J. Bot., 53, 1097-l 102 (1975). Tokashiki, M., Hamamoto, K., Takazawa, Y., and culture of mouse-human hyIchikawa, Y.: High-density bridoma cells using a new perfusion culture vessel. Kagaku Kogaku Ronbunshu, 14,337-341 (1988). Duff, S.M. G., Sarath, G., and Plaxton, W. C.: The role of acid phosphatses in plant phosphorous metabolism. Physiol.

26.

27.

28.

29.

30.

31.

Plant., 90,791-800 (1994). Ueki, K. and Sato, S.: Effect of inorganic phosphate on the extracellular acid phosphatase activity of tobacco cells cultured in vitro. Physiol. Plant., 24, 506-511 (1971). Goldstein, A. H., Baerrlein, D.A., and McDaniel, R. G.: Phosphate starvation inducible metabolism in Lycopersicon esculentum. Plant Physiol., 87, 711-715 (1988). Zhang, J. N. and Su, W. W.: Estimation of intracellular phosphate content in plant cell cultures using extended Kalman filter. J. Biosci. Bioeng., 94, 8-14 (2002). van Gulik, W., ten Hoopen, H., and Heijnen, J.: A structured model describing carbon and phosphate limited growth of Catharanthus roseus plant cell suspensions in batch and chemostat culture. Biotechnol. Bioeng., 41, 771-780 (1993). Hiller, G. W., Clark, D. S., and Blanch, H. W.: Cell retention-chemostat studies of hybridoma cells - analysis of hybridoma growth and metabolism in continuous suspension culture on serum-free medium. Biotechnol. Bioeng., 42, 185195 (1993). Su, W. W., Lei, F., and Kao, N. P.: High density cultivation of Anchusa officinalis in a stirred tank bioreactor with in situ filtration. Appl. Microbial. Biotechnol., 44,293-299 (1995).