Biotechnol. Prog.

2000, 16, 391401

391

Hairy Root Culture in a Liquid-Dispersed Bioreactor: Characterization of Spatial Heterogeneity

Gary R. C. Williams and Pauline M. Doran*
Department of Biotechnology, University of New South Wales, Sydney NSW 2052, Australia

A liquid-dispersed reactor equipped with a vertical mesh cylinder for inoculum support was developed for culture of Atropa belladonna hairy roots. The working volume of the culture vessel was 4.4 L with an aspect ratio of 1.7. Medium was dispersed as a spray onto the top of the root bed, and the roots grew radially outward from the central mesh cylinder to the vessel wall. Significant benefits in terms of liquid drainage and reduced interstitial liquid holdup were obtained using a vertical rather than horizontal support structure for the biomass and by operating the reactor with cocurrent air and liquid flow. With root growth, a pattern of spatial heterogeneity developed in the vessel. Higher local biomass densities, lower volumes of interstitial liquid, lower sugar concentrations, and higher root atropine contents were found in the upper sections of the root bed compared with the lower sections, suggesting a greater level of metabolic activity toward the top of the reactor. Although gas-liquid oxygen transfer to the spray droplets was very rapid, there was evidence of significant oxygen limitations in the reactor. Substantial volumes of non-free-draining interstitial liquid accumulated in the root bed. Roots near the bottom of the vessel trapped up to 3-4 times their own weight in liquid, thus eliminating the advantages of improved contact with the gas phase offered by liquid-dispersed culture systems. Local nutrient and product concentrations in the non-free-draining liquid were significantly different from those in the bulk medium, indicating poor liquid mixing within the root bed. Oxygen enrichment of the gas phase improved neither growth nor atropine production, highlighting the greater importance of liquid-solid compared with gas-liquid oxygen transfer resistance. The absence of mechanical or pneumatic agitation and the tendency of the root bed to accumulate liquid and impede drainage were identified as the major limitations to reactor performance. Improved reactor operating strategies and selection or development of root lines offering minimal resistance to liquid flow and low liquid retention characteristics are possible solutions to these problems.

Introduction
Hairy root cultures are a means of obtaining high yields of phytochemicals and other products from plant cells in vitro. Advantages associated with hairy roots compared with suspension cultures include the stimulatory effect of cell differentiation on secondary metabolite biosynthesis and the improved genetic stability of structurally differentiated tissues. Development of bioreactor technology for hairy roots is aimed at industrial exploitation of these production systems. However, the morphological properties and complex three-dimensional structure of hairy root cultures present new challenges for large-scale processing. The introduction into bioreactors of a solid phase of dense, tangled, but fragile roots restricts the levels of mechanical agitation and aeration that can be applied without damaging the cells (Wilson et al., 1987). Oxygen transfer is a major concern in large-scale root cultures in the absence of vigorous mixing (Kino-oka et al., 1996; Yu et al., 1997; Tescione et al., 1997). A variety of bioreactor configurations has been examined for growth of hairy roots, including stirred tanks (Davioud et al.,
* Fax: +61 (2) 9313-6710. Email: p.doran@unsw.edu.au
10.1021/bp0000306 CCC: $19.00

1989; Hilton and Rhodes, 1990), air-sparged vessels (Kwok and Doran, 1995; Tescione et al., 1997), and rotating drums (Kondo et al., 1989). Because hairy root cultures are likely to become oxygen-limited (Williams and Doran, 1999), liquid-dispersed reactors allowing direct contact between the biomass and gas phase are widely considered to offer significant advantages compared with reactors in which the roots remain submerged in medium. These benefits follow from the expectation that rates of transfer of oxygen and other gaseous components can be greatly increased if the liquid is dispersed as a fine spray or in rivulets trickling through the root bed. Liquid-dispersed reactors applied for hairy root culture have been presented variously as mist or nutrient-mist (DiIorio et al., 1992; Whitney, 1992; Buer et al., 1996; Nuutila et al., 1997; Weathers et al., 1997), droplet (Nuutila et al., 1997; Wilson, 1997), trickle-bed or tricklingfilm (Taya et al., 1989; Flores and Curtis, 1992; Whitney, 1992), and drip-tube (Holmes et al., 1997) reactors. In some of these systems, the fraction of the vessel volume actually used for growth (root volume/volume of growth chamber) is very small (0.012-0.16: DiIorio et al., 1992; Whitney, 1992; Chatterjee et al., 1997), or the volume occupied by the roots has a low (<1) aspect ratio because

2000 American Chemical Society and American Institute of Chemical Engineers Published on Web 05/12/2000

392

Biotechnol. Prog., 2000, Vol. 16, No. 3

the biomass is inoculated onto a single horizontal support screen (Buer et al., 1996; Nuutila et al., 1997). Although such configurations allow good direct penetration of dispersed medium into the root bed, in large-scale operations, the reactor volume must be used more effectively to reduce costs. To address this problem, several dispersedliquid reactors have been constructed with higher rootbed aspect ratios and with internal supports for the roots, such as glass beads, Raschig rings, steel wire scaffolding, polyurethane foam, or horizontal mesh trays, distributed throughout the vessel (Taya et al., 1989; Flores and Curtis, 1992; Whitney, 1992; Holmes et al., 1997; Wilson, 1997). Either liquid droplets or fine spray can be used to disperse the medium in these reactors; liquid coalescence in the upper few centimeters of root bed means that medium contact with most of the biomass is in the form of trickling rivulets. Despite the expected benefits associated with liquiddispersed bioreactors, application of nutrient mist and trickle-bed devices for hairy root culture has met with mixed results. Taya et al. (1989) found that the final biomass of horseradish roots in a trickle-bed bioreactor was 38% lower than in the same vessel operated as an air-sparged reactor. Whitney (1992) reported similar doubling times for growth of Nicotiana tabacum hairy roots in airlift and trickling-film reactors; growth of strawberry hairy root cultures was also similar in mist and airsparged vessels but lower in a droplet reactor (Nuutila et al., 1997). Solavetivone production by Hyoscyamus muticus hairy roots was about 3.5 times higher in a trickle-bed reactor than in submerged culture (Flores and Curtis, 1992). The growth rate of Datura stramonium hairy roots in a droplet-phase bioreactor was lower than in a stirred tank system, although the lag phase was eliminated (Wilson et al., 1990); in other work with this species, growth rates in a drip-tube reactor were very close to those in shake flasks (Holmes et al., 1997). These examples show that, although liquid-dispersed reactors have been applied successfully for hairy root culture, the conceptual advantages of improved gas-phase exposure and gas exchange commonly assumed for such systems have not translated into consistently higher growth rates or culture performance characteristics compared with other reactor configurations. Several reports have highlighted the importance of homogeneous biomass distribution as a critical parameter determining the outcome of hairy root cultures in liquid-dispersed reactors (Ramakrishnan and Curtis, 1994; Wilson, 1997). Reactor performance has also been found to diminish significantly if liquid distribution over and through the root bed is not uniform (DiIorio et al., 1992; McKelvey et al., 1993; Nuutila et al., 1997). The aim of this work was to investigate the performance of a novel liquid-dispersed bioreactor applied for cultivation of Atropa belladonna hairy roots. The reactor design addresses some of the operational difficulties commonly associated with large-scale hairy root culture. A nutrient spray was applied for medium dispersal, while a central vertical mesh tube was employed for support and distribution of the inoculum roots. The experiments provide detailed information about spatial variations in culture properties generated within the reactor as root growth occurred, including biomass density, alkaloid accumulation, liquid holdup, and interstitial liquid composition. These results were used to identify the major problems and limitations associated with operation of liquid-dispersed reactors.

Materials and Methods

Bioreactor Apparatus and Operation. The bioreactor for liquid-dispersed culture of hairy roots consisted of a culture vessel containing the roots connected to a separate 5-L medium reservoir. The culture vessel (Figure 1a) was custom-made from QVF glass pipe with a dished base and had a total volume of 9.1 L. The cylindrical section of the vessel was 15 cm i.d. and 32 cm high, with a custom-made stainless steel headplate. Roots were inoculated into a vertical stainless steel mesh cylinder (1.5 cm diameter, 25 cm long, 1.25 mm aperture), which was supported in the center and about 5 cm above the base of the vessel by a 2-cm diameter plate attached by three arms to an outer ring at the walls of the reactor. The working volume of the culture vessel, defined as the volume between the upper and lower ends of the mesh cylinder, was 4.4 L. Liquid in the medium reservoir was mixed using a magnetic stirrer. Medium was drawn from the reservoir by a valveless, positive displacement pump (QD-2, Fluid Metering Inc., Syosset, NY) and sprayed into the culture vessel through a Unijet spray nozzle (1/4TT, Spraying Systems Co., Wheaton, IL) located in the center of the headplate. The nozzle was fitted with a 100 mesh internal screen strainer and Teejet spray tip (TG0.3, 0.020 in. hole diameter, Spraying Systems) and produced a full-cone circular spray pattern. A constant spray flow rate of 130.5 ( 9.5 mL min-1 (( indicates the maximum variation in flow rate between experiments) was used. Liquid drained from the culture vessel through sintered glass filters that trapped any root debris. Medium was recycled back to the reservoir by a peristaltic pump delivering a flow rate equal to or slightly higher than that of the spray pump. Oxygen diffusion through liquid lines in continuous operation was minimized using marprene tubing. Air humidified by bubbling through sterile water was supplied to the culture and reservoir vessels at flow rates of 1000 ( 50 and 200 ( 25 mL min-1, respectively. Air entered the culture vessel through three ports covered with baffle plates on the underside of the headplate; sintered metal spargers were used in the medium reservoir. Liquid holdup in the root bed was minimized by cocurrent flow of gas and liquid through the culture vessel. Exit gases from the culture vessel and reservoir were passed through condensers to minimize evaporative losses. Polarographic dissolved oxygen probes (Ingold, Urdorf, Switzerland) were used to measure oxygen partial pressures in the reservoir medium and in the medium recycled from the culture vessel. The oxygen probe on the medium recycle line was inserted into a flow cell similar to that described previously (Williams and Doran, 1999). Temperature in the medium reservoir was maintained at 25.0 ( 0.5 C; air entering the culture vessel was humidified and preheated to 25.0 ( 0.8 C by passage through temperature-controlled water. Temperatures in the reservoir vessel and in the inlet medium recycle line to the culture vessel were measured with PT 100 thermistors. The culture vessel was insulated by covering with thick polystyrene foam, which also maintained the roots in darkness. After dehumidification (Komatsu, Hiratsuka, Japan) of the exit gas, an infrared carbon dioxide analyzer (Servomex, Crowborough, U.K.) was used to measure CO2 levels; results for the exit gas streams from the culture and medium reservoir vessels differed by less that 0.02% v/v. Gas pressures in the culture and reservoir vessels were monitored using water-filled glass manometers. The pressure in the

Biotechnol. Prog., 2000, Vol. 16, No. 3

393

Figure 1. (a) Culture vessel containing A. belladonna hairy roots after 26.7 days of growth in the phosphate fed-batch experiment. (b) Vertical mesh cylinder in the center of the culture vessel 4.6 days after inoculation, showing radial growth of the roots. (c) Root section (2.6-cm thick) cut from the biomass harvested after 24.9 days of reactor operation under standard conditions. The roots in the central mesh cylinder were removed separately. The section was located 14.2 cm from the base of the mesh cylinder, and the average local biomass density was 11.9 g L-1 dry weight.

culture vessel was manually regulated to 2.6-4.0 kPa above atmospheric; the relative pressure in the reservoir ranged between 0.2 and 0.6 kPa. Experimental Procedure. Atropa belladonna hairy roots were initiated and maintained as described previously (Williams and Doran, 1999). The Murashige and Skoog (MS) medium (Flow Laboratories, Irvine, Scotland, and ICN, Costa Mesa, CA) used in the bioreactor experiments contained 30 g L-1 sucrose and was adjusted to pH 5.8 before filter-sterilization through a 0.22-m filter. The reactor was inoculated with 13.00 ( 0.01 g fresh weight of roots 2-4 cm in length from 14-day-old shake flask cultures. The roots were transferred into the vertical mesh cylinder in the culture vessel under gravity through wide-bore (1.25 cm i.d.) silicone tubing attached to a steel tube. The inoculum roots filled the mesh cylinder and grew radially outward from the cylinder toward the vessel wall (Figure 1b). The initial medium volume was 4.5 L. Independent culture experiments of durations up to 41.8 days were performed, after which the entire root bed was harvested for analysis. Samples of bulk medium were also taken every 1-3 days from the reservoir vessel. The net specific rate of dissolved oxygen depletion from the liquid phase was determined from the difference in oxygen partial pressures of medium entering and leaving the root bed:

net specific rate of disssolved oxygen depletion ) C*QS(Ci - Co) (mmol min-1 g-1 dry weight) (1) 100MB
where C* is the solubility of oxygen in medium at 25 C and 1 atm air pressure (mmol L-1), QS is the spray flow rate (L min-1), Ci is the dissolved oxygen tension in the spray droplets after dispersion (% air saturation), Co is the dissolved oxygen tension in the liquid leaving the culture vessel (% air saturation), and MB is the biomass dry weight (g). The solubility of oxygen in the medium was confirmed to be within 5% of the solubility in water (Williams and Doran, 1999) and was taken to be 0.272 mmol L-1. The net rate of dissolved oxygen depletion from the liquid is a balance between the rate of oxygen uptake by the biomass and the rate of medium reaeration from contact with the gas phase. Because medium reaeration reduces the net rate of oxygen depletion as defined in the above equation, the combination of terms in eq 1 also provides an estimate of the minimum specific oxygen uptake rate by the roots. Dissolved oxygen depletion rates were determined periodically during the culture experiments. Additional measurements were also carried out at atmospheric pressure to determine the effect of gas and liquid flow rates; dissolved oxygen depletion rates were determined as a function of spray flow rate between 87 and 299 mL min-1 at an air flow rate of 1000 mL min-1 at the end of several experiments, and as a function

394

Biotechnol. Prog., 2000, Vol. 16, No. 3

of air flow rate between 1.0 and 7.2 L min-1 at spray flow rates of 121 and 157 mL min-1 after 38.3 days of culture. Oxygen Enrichment and Phosphate Fed-Batch Experiments. The operating conditions described above using batch culture of hairy roots with oxygenation by air are referred to as standard operating conditions. In addition to these studies, experiments were performed to determine the effects of oxygen enrichment of the inlet air and intermittent feeding of phosphate solution. In the oxygen-enrichment experiment, the culture vessel was supplied with a mixture of air and medical-grade oxygen to provide a gas-phase oxygen level of 158 ( 2.3% that in air, at a total gas flow rate of 1000 mL min-1. The inlet gas oxygen level was monitored using a paramagnetic oxygen analyzer (Servomex), and the roots were cultured for 25.3 days. In the phosphate-enrichment experiment, a concentrated solution of 50 g L-1 KH2PO4 was added intermittently in fed-batch mode to the medium reservoir to maintain the phosphate concentration close to that in MS medium (119 mg L-1 PO43-). The total volume of solution added over the entire experiment was 82 mL, or less than 2% of the initial medium volume. The phosphate fed-batch experiment was carried out for 26.7 days under air. Analytical Procedures. The volume of free liquid in the root bed was determined after gravity draining immediately prior to harvesting and before measuring the wet weight of the roots. The root bed was removed from the cultivation vessel with its shape preserved and cut into 4-12 cross sections (depending on the amount of biomass) of similar thickness (Figure 1c). The height of a section in the root bed was determined as the distance from the support plate at the bottom of the cylinder to the midpoint of the root section. Non-freedraining interstitial liquid was squeezed out of each dense spongelike root section carefully by hand and the volume measured. The weight of the remaining biomass sections was measured as fresh weight. The total nonfree-draining interstitial liquid volume for the entire root bed was determined as the difference between the wet weight and fresh weight of the root bed. Root dry weight for each section was determined by freeze-drying as described previously (Williams and Doran, 1999). Local root densities were based on the cylindrical volume of the reactor occupied by the roots as defined by the outer root extremities (usually the vessel wall). Inorganic phosphate (PO43-) in the spent medium and interstitial liquid was determined using a kit (Sigma, St. Louis, MO), with measurement of UV absorbance at 340 nm and quantification using a KH2PO4 standard solution. Results are reported as the average of duplicate sample preparations. Interference from other ions in the medium was examined: MS medium gave an insignificantly higher (about 1.5%) reading than the standard solution. Sugars were analyzed by HPLC as described previously (Williams and Doran, 1999). Total sugar concentrations are reported as the sum of the constituent sugar concentrations. Biomass yields from sugar were calculated using sucrose equivalents to account for the molecule of water incorporated during sucrose hydrolysis. Atropine concentrations in the biomass were measured by HPLC as described previously (Williams and Doran, 1999). Each cross section of dried roots was mechanically pulverized in a hammer mill for 5-10 min (depending on the quantity of material) at a speed of 6000 rpm to provide samples of root powder for extraction. Samples of free-draining and non-free-draining interstitial liquids were analyzed directly by HPLC after filtering through

Figure 2. Biomass dry weight in the reactor plotted using semilogarithmic coordinates. Each point represents biomass harvested from an independent experiment conducted (b) under standard operating conditions (batch culture with air), (O) with a gas-phase oxygen level of 158 ( 2.3% that in air, or (4) with intermittent phosphate feeding.

0.45-m filters. Total atropine contents in sections of the biomass were calculated from the sum of the alkaloid contents in the roots and non-free-draining interstitial liquid.

Results
Overall and Local Root Growth. Roots inoculated into the vertical mesh cylinder in the culture vessel grew radially through the mesh apertures as shown in Figure 1b. Some roots began touching the vessel wall by day 7, indicating a maximum average linear growth rate of 9.6 mm day-1. The spray droplets coalesced into a liquid film on the upper layers of roots, which increasingly shielded the lower roots from direct spray contact as the biomass increased. As a result, the lower roots became covered in fine aerial root hairs, similar to A. belladonna hairy roots growing above the liquid level in shake flasks (Kanokwaree and Doran, 1997). Spray coalescing on the upper roots trickled down the root bed in small rivulets. Lateral branching on new-growth roots outside the mesh cylinder occurred only after day 10. There was no significant difference in the appearance of the root bed during the oxygen enrichment and phosphate fed-batch experiments compared with operation under standard conditions. Results for overall root growth in the bioreactor are shown in Figure 2. The maximum biomass achieved under standard operating conditions was 49.4 g dry weight after 24.9 days. The root specific growth rate declined progressively from an initial value of 0.37 day-1 (doubling time, 1.9 days). As indicated in Figure 2, final biomass levels in the oxygen enrichment and phosphate fed-batch experiments were not significantly different from those in cultures under standard conditions. Results for local biomass density in the reactor, excluding the core section within the wire mesh cylinder, are shown in Figure 3. Biomass sections with a negative height were formed by roots growing downward from the root bed at the level of the end of the cylinder; low biomass densities at the extreme top and bottom of the root bed reflect the predominance of single root strands in these sections (Figure 1a). As shown in Figure 3a, a gradient of root density developed after 14-16 days, with higher biomass levels near the top of the root bed. This variation in local biomass density became more pronounced with culture time; after 24.9 days, the maximum

Biotechnol. Prog., 2000, Vol. 16, No. 3

395

Figure 3. Local biomass densities in cross sections of the root bed as a function of height above the support plate. (a) Cultures conducted under standard conditions (batch culture with air) after (1) 5.4, (0) 9.8, (b) 13.6, (3) 16.0, (2) 21.0, (9) 24.9, and (O) 41.8 days. (b) Cultures with (O) a gas-phase oxygen level of 158 ( 2.3% that in air after 25.3 days or (b) intermittent phosphate feeding after 26.7 days. The biomass densities do not include the roots in the core section of the culture vessel. Zero height corresponds to the bottom support plate of the root bed.

Figure 4. Typical time-course data for hairy root culture under standard operating conditions (batch culture with air). (a) Concentrations of (9) sucrose and (0) total sugars in the bulk medium and (b) gas-phase carbon dioxide concentration. (b) (9) Phosphate concentration in the bulk medium, (2) dissolved oxygen tension in the recirculation medium, and (O) net specific rate of dissolved oxygen depletion. The data were measured during the 24.9-day experiment.

local biomass density near the top of the bed (22 cm) was 1.9 times that near the bottom (1.4 cm). In addition to a biomass density variation over the height of the root bed, there was also a significant radial gradient, e.g., cross sections of the root bed after 16 days revealed an essentially homogeneous, high-density interior of diameter about 8 cm, with a lower density outer growth to a diameter of 15 cm incorporating a high proportion of void space. The highest local biomass density in all experiments was found in the core, which after 9.8 days was brown in color and morphologically similar to the initial stages of callus formation, then progressively darkened to be black by 24.9 days. The maximum local biomass density in the core measured over the total height of the mesh cylinder was 67.5 g L-1 dry weight after 13.6 days. As shown in Figure 3b, the relationship between height and local biomass density in the oxygen- and phosphatesupplemented cultures after 25.3 and 26.7 days, respectively, was similar to that in Figure 3a; however, especially in the phosphate fed-batch experiment, the maximum values were slightly greater and occurred at lower heights than at a similar time (24.9 days) under standard operating conditions. Overall Time-Course Culture Characteristics. Typical time-course data for bulk medium sugar and phosphate levels, dissolved oxygen tension in the recirculating medium, gas-phase carbon dioxide concentration, and the net specific rate of dissolved oxygen depletion are shown in Figure 4. On the basis of the results from all the experiments, sucrose was completely hydrolyzed to glucose and fructose by day 11; after this time, both monosaccharides were taken up by the roots, with glucose being preferentially consumed. Total sugars were exhausted from the bulk medium after 22-24 days,

coinciding with a sharp decline in the gas-phase carbon dioxide concentration. There was a small distinct peak in the gas-phase carbon dioxide level after 9-10 days; this pattern was reproduced in each of eight independent reactor experiments, including the oxygen- and phosphatesupplemented cultures. Depletion of phosphate from the bulk medium occurred after 8-9 days under standard operating conditions. The dissolved oxygen tension in the recirculating medium decreased from 100% to ca. 65% air saturation when all sugars were depleted. A small peak in dissolved oxygen tension occurred at day 11-12; this transient was evident during each of the experiments conducted under standard operating conditions. The dissolved oxygen tension measured in the reservoir vessel declined throughout the culture, reaching a minimum of 88% air saturation. The net specific rate of dissolved oxygen depletion decreased from a maximum of 1.3 10-3 mmol min-1 g-1 dry weight, except for a distinct transient depression at day 11-12. Time-course data from the oxygen-enrichment experiment followed trends very similar to those shown in Figure 4. Elevating the oxygen partial pressure did not increase the rate of total sugar consumption compared with operation under standard conditions; values for gasphase carbon dioxide concentration were coincident with those shown in Figure 4a for most of the culture period and differed by less than 10% at other times. These results indicate that oxygen enrichment of the gas phase had a negligible effect on rates of metabolic activity in the reactor. In the phosphate fed-batch experiment, uptake of total sugars was slower toward the end of the growth phase compared with standard conditions, and a small sugar concentration (2.0 g L-1) remained in the bulk medium after 26.7 days. Gas-phase carbon dioxide levels decreased gradually after day 19, suggesting that

396

Biotechnol. Prog., 2000, Vol. 16, No. 3

Figure 5. Free-draining and non-free-draining liquid holdup in the root bed. (a) Volume of free-draining liquid measured during experiments conducted (0) under standard operating conditions (batch culture with air), (O) with a gas-phase oxygen level of 158 ( 2.3% that in air, and (4) with intermittent phosphate feeding; volume of non-free-draining interstitial liquid measured at the end of experiments conducted (9) under standard operating conditions, (b) with a gas-phase oxygen level of 158 ( 2.3% that in air, and (2) with intermittent phosphate feeding. (b) Specific volume of non-free-draining interstitial liquid in the root bed measured at the end of experiments conducted (b) under standard operating conditions, (O) with a gas-phase oxygen level of 158 ( 2.3% that in air, and (4) with intermittent phosphate feeding.

Figure 6. Local specific volumes of non-free-draining interstitial liquid in cross sections of the root bed as a function of height above the support plate. (a) Cultures conducted under standard conditions (batch culture with air) after (1) 5.4, (0) 9.8, (b) 13.6, (3) 16.0, (2) 21.0, (9) 24.9, and (O) 41.8 days. (b) Cultures with (O) a gas-phase oxygen level of 158 ( 2.3% that in air after 25.3 days, or (b) intermittent phosphate feeding after 26.7 days. The results do not include the core section of the culture vessel. Zero height corresponds to the bottom support plate of the root bed.

metabolic activity started to shut down about 5 days earlier than in the other experiments. Another difference in the phosphate fed-batch culture was that the results for dissolved oxygen tension did not exhibit the midculture peak shown in Figure 4b. Liquid Holdup and Local Nutrient Concentrations. The liquid volume in the reservoir vessel decreased by ca. 55% during the hairy root growth phase (0-25 days) as a result of liquid uptake by the roots and entrapment within the root bed. The free-draining liquid holdup in the culture vessel increased with time as indicated in Figure 5a; the initial holdup volume of 43 mL in the absence of roots can be attributed to the liquid dispersed in the spray and captured on the vessel surfaces. Figure 5a also shows the non-free-draining interstitial liquid holdup, which could be removed from the root bed only by careful squeezing. Under standard operating conditions, the non-free-draining liquid volume reached a maximum of 1680 mL after 41.8 days, or about 6.3 times the maximum free-draining holdup. Results from the oxygen enrichment experiment were similar to those under standard conditions; however, in the phosphate fed-batch culture, although free-draining volumes were similar, the non-free-draining holdup after 26.7 days was 37% lower than under standard conditions. As indicated in Figure 5b, the specific non-free-draining interstitial holdup, defined as the volume of non-freedraining liquid per mass of fresh roots, reached a maximum of 2.4 mL g-1 fresh weight at 41.8 days. The result in Figure 5b for the oxygen-enriched culture is

similar to those under standard operating conditions; the lower specific holdup at the end of the phosphate fedbatch experiment is consistent with the significantly reduced volume of interstitial liquid in this culture as indicated in Figure 5a. Local specific volumes of non-free-draining liquid are shown in Figure 6a as a function of height in the root bed. The specific holdup increased down the bed, with the differences between the top and bottom becoming more pronounced with culture age, i.e., biomass level. The sharp decline in specific holdup below the support plate at zero height is due to the low density of biomass in that section (Figure 3a) and its inability to entrap significant volumes of liquid. Specific non-free-draining liquid holdup in the core section measured over the total height of the mesh tube did not change significantly during the culture period and was relatively low in all of the experiments (0.24-0.76 mL g-1 fresh weight), probably because of the dense biomass packing in the cylinder. As shown in Figure 6b, the relationship between height and specific liquid holdup at the end of the oxygen- and phosphatesupplemented cultures followed a pattern similar to that under standard operating conditions. Compared with the results at 24.9 days under standard conditions, the maximum specific holdup was greater at the end of the oxygen-enrichment experiment, while the specific holdups at all heights were lower in the phosphate fed-batch experiment. Typical local concentrations of glucose, fructose, and total sugars in the non-free-draining interstitial liquid collected from different heights of the root bed are shown in Figure 7a. These data were measured after 9.8 days of operation under standard conditions; results at other times exhibited similar trends. Sugar concentrations

Biotechnol. Prog., 2000, Vol. 16, No. 3

397

Figure 7. Local sugar concentrations after 9.8 days of hairy root culture under standard operating conditions (batch culture with air). (a) Concentrations of (b) glucose, (O) fructose, and (9) total sugars in the non-free-draining interstitial liquid in cross sections of the root bed as a function of height above the support plate. Sucrose was not present in any of the root sections, and there was no non-free-draining liquid in the top section of the root bed. (b) Concentrations of (b) glucose, (O) fructose, (2) sucrose, and (9) total sugars in the non-freedraining interstitial liquid in the core section, and in the bulk medium in the reservoir vessel. Zero height corresponds to the bottom support plate of the root bed.

Figure 8. Overall atropine content in the biomass. Each point represents biomass harvested from an independent experiment conducted (b) under standard operating conditions (batch culture with air), (O) with a gas-phase oxygen level of 158 ( 2.3% that in air, or (4) with intermittent phosphate feeding.

increased down the bed; levels of glucose and fructose were, respectively, 36% and 16% lower at the top than at the bottom, these relative values being consistent with glucose being preferentially consumed by the roots. As indicated in Figure 7b, sugar concentrations in the interstitial liquid of the core section were similar to the average values over the bed height. However, sucrose and total sugar levels in the bulk medium were substantially higher than in the non-free-draining liquid samples. No sucrose was evident in the non-free-draining interstitial liquid even though 3.3 g L-1 was still present in the bulk medium, while the total sugar concentration in the bulk medium was 11-49% higher than in the interstices of the root bed. The yield of biomass from total sugar consumed from the bulk medium and interstitial liquid reached a maximum value of 0.44 g g-1 dry weight after 13.6 days and then decreased to 0.38 g g-1 at 24.9 days. Final biomass yields from sugar in the oxygen-enrichment and phosphate fed-batch experiments after 25.3 and 26.7 days, respectively, were within 5% of that after 24.9 days under standard operating conditions. Local phosphate concentrations in the non-free-draining interstitial liquid after 26.7 days of phosphate fedbatch culture decreased down the bed from 1350 mg L-1 near the top to a minimum of 710 mg L-1 near the bottom. These values were significantly higher than the concentration of 180 mg L-1 in the bulk medium. A phosphate (PO43-) mass balance was carried out to calculate the difference between the total mass of phosphate provided to the culture in the initial MS medium (530 mg) and added KH2PO4 solution (2865 mg), and the total mass of phosphate recovered from the free-draining and bulk medium (570 mg) and interstitial liquid fractions (710 mg) at the end of the fed-batch experiment. The difference of 2115 mg represents the amount of phosphate taken up by the 50.6 g dry weight of root biomass produced in the culture, giving an average phosphate content in the roots of 41.8 mg g-1. This value is essentially identical to the average phosphate content of 42.1 mg g-1 also calculated by mass balance for roots cultured under standard operating conditions for 9.8 days

when the phosphate concentration in the bulk medium was exhausted (Figure 4b). Overall and Local Atropine Levels. Changes in overall root atropine content during the reactor experiments are shown in Figure 8. Under standard operating conditions, the atropine concentration in the biomass increased from an inoculum value of 2.4 mg g-1 dry weight to a maximum of 3.8 mg g-1 at the end of the growth phase (24.9 days). The atropine level at the end of the oxygen enrichment experiment was similar to that under standard conditions; the final atropine concentration in the phosphate fed-batch experiment was somewhat higher than the maximum obtained under standard conditions. The variation in biomass atropine content over the height of the root bed is shown in Figure 9. An atropine gradient with decreasing levels down the bed was evident toward the end of the cultures under standard operating conditions and at the end of the oxygen- and phosphatesupplemented cultures. At these times, the most significant decreases in atropine content occurred in the upper sections of the root bed; atropine levels near the top of the bed were roughly double those in the middle and bottom sections. After 16 days in the standard cultures and at the end of the oxygen- and phosphate-supplemented cultures, there was also an increase in atropine levels in the biomass developing below the support plate at zero height. Root atropine concentrations were relatively low in the core section of the reactor, declining over the first 9.8 days from 2.4 mg g-1 dry weight in the inoculum to around 1 mg g-1 dry weight, then remaining relatively constant. Atropine was also present in the liquid phase in the hairy root reactor. Most of the liquid atropine was recovered from the non-free-draining interstitial holdup, as concentrations in the bulk medium were negligible. The overall proportion of atropine released into the medium increased with time from 0% at inoculation to 19% by 24.9 days; at 41.8 days, after growth had ceased, the proportion of atropine in the medium was 53%. Results for the distribution of atropine between the roots and culture liquid in the oxygen enrichment and phosphate fed-batch experiments were within 4% of those measured under standard operating conditions. The local distribution of atropine between the roots and non-freedraining interstitial liquid is plotted as a function of bed height in Figure 10. Under most reactor operating

398

Biotechnol. Prog., 2000, Vol. 16, No. 3

Figure 9. Local atropine contents in the biomass as a function of height in the root bed above the support plate. (a) Cultures conducted under standard conditions (batch culture with air) after (1) 5.4, (0) 9.8, (b) 13.6, (3) 16.0, (2) 21.0, (9) 24.9, and (O) 41.8 days. (b) Cultures with (O) a gas-phase oxygen level of 158 ( 2.3% that in air after 25.3 days, or (b) intermittent phosphate feeding after 26.7 days. The results do not include the root biomass in the core section of the culture vessel. Zero height corresponds to the bottom support plate of the root bed.

Figure 10. Distribution of atropine between the biomass and non-free-draining interstitial liquid as a function of height above the support plate. (a) Cultures conducted under standard conditions (batch culture with air) after (1) 5.4, (0) 9.8, (b) 13.6, (3) 16.0, (2) 21.0, (9) 24.9, and (O) 41.8 days. (b) Cultures with (O) a gas-phase oxygen level of 158 ( 2.3% that in air after 25.3 days, or (b) intermittent phosphate feeding after 26.7 days. The results do not include the core section of the culture vessel. Zero height corresponds to the bottom support plate of the root bed.

conditions, including the oxygen- and phosphate-supplemented experiments (Figure 10b), there was a gradual decline in the proportion of atropine found in the biomass down the height of the root bed. As indicated in Figure 10a, the decline was much greater at 41.8 days than at earlier culture times. Effect of Reactor Operating Conditions on the Net Specific Rate of Dissolved Oxygen Depletion. In the absence of roots, deoxygenated liquid medium sprayed into the reactor over a wide range of flow rates reached ca. 95% air saturation during one passage of the culture vessel, demonstrating that gas-liquid oxygen transfer to the fine droplets was very effective. To determine the effect of air and liquid flow rates on overall oxygen transfer, rates of dissolved oxygen depletion were measured in the presence of roots after 38.3 days of growth. The results indicated that dissolved oxygen depletion rates calculated using eq 1 were essentially independent of gas flow rate over the range 1.0 to 7.2 L min-1. However, as shown in Figure 11, there was a direct linear relationship between liquid flow rate and the rate of dissolved oxygen depletion, resulting in a significant increase in oxygen depletion at high spray flow rates. Dissolved oxygen depletion was slowest after 38.3 days, when the culture was in stationary phase.

Figure 11. Net specific rate of dissolved oxygen depletion as a function of liquid spray flow rate in the culture vessel. Measurements were performed at atmospheric pressure after (b) 13.6, (2) 21.0, (9) 24.9, and (O) 38.3 days of hairy root culture at an air flow rate of 1000 mL min-1 in the culture vessel. The net specific rate of dissolved oxygen depletion increased at each biomass level with increasing spray flow rate.

Discussion
Use of a vertical steel mesh cylinder as an inoculation support proved to be an effective method for maintaining a relatively even distribution of roots over the height of the vessel, at least for the first half of the culture period, and for facilitating a unidirectional, radial root growth pattern. This type of support overcomes the need for manual placement of inoculum roots in the reactor (Taya

et al., 1989); unlike systems in which the root inoculum is randomly immobilized by initial operation of the vessel in submerged liquid mode (Ramakrishnan et al., 1994; Wilson, 1997), it also provides a fixed and reproducible inoculum distribution. The absence of other internal structures or packing inside the reactor maximizes the void volume available for occupation by the roots; harvesting and recovery of the biomass is also easier than from reactors containing packing elements such as glass beads to which the roots can adhere (Ramakrishnan et

Biotechnol. Prog., 2000, Vol. 16, No. 3

399

al., 1994; Whitney, 1992). Complete consumption of sugars in the bulk liquid (Figure 4a) represents an improvement over the reactor configurations tested in previous bioreactor studies with A. belladonna hairy roots (Sharp and Doran, 1990; Kwok and Doran, 1995), in which root growth in identical medium ceased in the presence of high residual bulk sugar concentrations. However, an important requirement for success of the liquid-dispersed reactor developed in this work is that the individual roots must have sufficient mechanical strength and rigidity (Carvalho et al., 1997) to support their own radially directed, horizontal growth outside the mesh cylinder (Figure 1b), including carrying the weight of coalescing liquid films. Pronounced bending of the roots under gravity would result in very poor radial biomass distribution in the vessel. Experiments with a preliminary reactor design using a horizontal rather than vertical mesh screen to support the roots showed that horizontal orientation of the screen and its subsequent biomass loading were significant impediments to liquid drainage even in the early stages of root growth. Although horizontal support screens are commonly applied in nutrient-mist bioreactors (DiIorio et al., 1992; Nuutila et al., 1997), installation of the vertical mesh cylinder in this work greatly reduced liquid entrainment in the culture vessel. Cocurrent flow of air and liquid was also found to improve fluid motion within the root bed. Nevertheless, substantial volumes of interstitial liquid accumulated near the bottom of the vessel toward the end of the culture period, as the roots trapped up to 3-4 times their own weight in liquid (Figure 6). This is an undesirable characteristic in a liquid-dispersed reactor (McKelvey et al., 1993), as it reduces contact of the roots with the gas phase. Effective submersion of the biomass in interstitial liquid not only reduces oxygen transfer by creating liquid-solid boundary layers, but also may increase internal tissue resistance to oxygen transfer by flooding the intercellular gas spaces (Ohmura and Howell, 1960; Armstrong and Gaynard, 1976). That oxygen transfer limitations existed in the reactor system is evident from the changes in net specific rate of dissolved oxygen depletion with liquid velocity (Figure 11), while the large drop in dissolved oxygen depletion rate with culture time (Figure 4b) is consistent with oxygen transfer from the medium to the roots becoming increasingly restricted as the culture proceeded. These results suggest that, without the assistance of mechanical or pneumatic agitation in the culture vessel, the spray reactor did not provide sufficient relative velocity between the roots and liquid to overcome liquid-solid oxygen transfer limitations. The large spatial variations in interstitial phosphate and sugar levels observed in this work, and the significant differences between interstitial and bulk medium properties, indicate that liquid mixing within the root bed was poor, especially at high biomass densities. In the absence of roots, gas-liquid oxygen transfer to the spray droplets was found to be virtually instantaneous, irrespective of the rate of spraying within the range tested. Similar results have been reported previously for spray reactors (Wilson et al., 1990) and reflect the very high surface area available for gas-liquid transfer. However, although such findings have prompted considerable interest in liquid-dispersed reactors for hairy root culture, they do not address the problem of liquid-solid oxygen transfer. Accordingly, except in the upper sections of the root bed where the amounts of interstitial liquid were relatively low (Figure 6b), growth was not significantly improved by oxygen enrichment of

the gas phase compared with culture under standard conditions (Figure 3). There was no difference in the total biomass produced (Figure 2), and rates of sugar consumption and carbon dioxide evolution were essentially the same as under air. This result highlights the greater importance of liquid-solid mass transfer processes for oxygen delivery to hairy roots, and confirms previous findings showing no significant improvement in root growth in sparged bioreactors using oxygen-enriched air (Kanokwaree and Doran, 1998). As growth of the roots proceeded, a pattern of spatial heterogeneity developed in the culture vessel. The observed differences in local biomass, nutrient, and product concentrations emphasize the difficulty of characterizing hairy root reactor performance using reactor-average culture properties, which do not accurately reflect the conditions actually developed or experienced by the roots. Compared with the lower sections, the upper regions of the root bed closer to the sources of oxygen and recirculated medium were characterized by higher root densities, lower volumes of non-free-draining interstitial liquid per gram fresh biomass, lower sugar concentrations, higher atropine contents in the roots, and a greater tendency to retain atropine in the biomass. These spatial variations are consistent with roots near the top of the reactor having higher metabolic activity than roots in the lower sections, hence the higher local biomass and atropine levels and lower sugar concentrations near the top. The mechanisms behind development of this particular pattern of spatial heterogeneity were not determined definitively in this work; however, it is probable that the roots were increasingly subject to oxygen limitations down the height of the reactor. Yet, other explanations for the reduced metabolic activity near the bottom of the vessel, e.g. increased pressure due to the weight of the biomass, may also be valid. Liquid entrainment in hairy root cultures has been correlated with the hairiness of the roots (Ramakrishnan and Curtis, 1994), as fewer and/or shorter root hairs can be expected to facilitate liquid drainage through the root bed. Reactors such as that developed in this work may therefore be more suitable for less hairy roots than for roots with prolific hairs, so that careful selection of species could be an important factor affecting the performance of liquid-dispersed systems. The significantly lower non-free-draining liquid volume measured in the phosphate fed-batch culture compared with standard operating conditions (Figure 5a) is consistent with this interpretation of the influence of root hairs on liquid holdup. Studies with intact plants have found that root hair length is inversely related to phosphate concentration in the medium (Bates and Lynch, 1996), and local phosphate concentrations were considerably higher in the fed-batch culture than under standard conditions. Interstitial phosphate levels in the fed-batch culture were 4.0-8.1 times higher than in the bulk medium, depending on position in the root bed. This finding indicates that local concentration effects operated in the interstices of the bed, as water was absorbed for growth of new biomass but phosphate in the medium was either excluded or subsequently excreted. The results of the phosphate mass balance calculations suggest that the high interstitial phosphate concentrations observed in the fed-batch experiment were necessary to maintain physiological conditions inside the cells. The average root PO43- content of ca. 42 mg g-1 dry weight determined from the mass balance is within the range of 38 ( 5.9 mg g-1 reported previously for growing plant cell cultures (Curtis et al., 1991).

400

Biotechnol. Prog., 2000, Vol. 16, No. 3

a nontoxic acoustic window nutrient-mist bioreactor and relevant growth data. In Vitro Cell. Dev. Biol. Plant 1996, 32, 299-304. Carvalho, E. B.; Holihan, S.; Pearsall, B.; Curtis, W. R. Effect of root morphology on reactor design and operation for production of chemicals. In Hairy Roots: Culture and Applications; Doran, P. M., Ed.; Harwood Academic: Amsterdam, 1997; pp 151-167. Chatterjee, C.; Correll, M. J.; Weathers, P. J.; Wyslouzil, B. E.; Walcerz, D. B. Simplified acoustic window mist bioreactor. Biotechnol. Tech. 1997, 11, 155-158. Curtis, W. R.; Hasegawa, P. M.; Emery, A. H. Modeling linear and variable growth in phosphate limited suspension cultures of opium poppy. Biotechnol. Bioeng. 1991, 38, 371-379. Davioud, E.; Kan, C.; Hamon, J.; Tempe, J.; Husson, H.-P. Production of indole alkaloids by in vitro root cultures from Catharanthus trichophyllus. Phytochemistry 1989, 28, 2675-2680. DiIorio, A. A.; Cheetham, R. D.; Weathers, P. J. Growth of transformed roots in a nutrient mist bioreactor: reactor performance and evaluation. Appl. Microbiol. Biotechnol. 1992, 37, 457-462. Flores, H. E.; Curtis, W. R. Approaches to understanding and manipulating the biosynthetic potential of plant roots. Ann. N.Y. Acad. Sci. 1992, 665, 188-209. Hilton, M. G.; Rhodes, M. J. C. Growth and hyoscyamine production of hairy root cultures of Datura stramonium in a modified stirred tank reactor. Appl. Microbiol. Biotechnol. 1990, 33, 132-138. Holmes, P.; Li, S.-L.; Green, K. D.; Ford-Lloyd, B. V.; Thomas, N. H. Drip-tube technology for continuous culture of hairy roots with integrated alkaloid extraction. In Hairy Roots: Culture and Applications; Doran, P. M., Ed.; Harwood Academic: Amsterdam, 1997; pp 201-208. Kanokwaree, K.; Doran, P. M. The extent to which external oxygen transfer limits growth in shake flask culture of hairy roots. Biotechnol. Bioeng. 1997, 55, 520-526. Kanokwaree, K.; Doran, P. M. Application of membrane tubing aeration and perfluorocarbon to improve oxygen delivery to hairy root cultures. Biotechnol. Prog. 1998, 14, 479-486. Kino-oka, M.; Tsutsumi, S.; Tone, S. Oxygen transfer in bioreactor for culture of plant hairy roots. J. Chem. Eng. Japan 1996, 29, 531-534. Knobloch, K.-H.; Beutnagel, G.; Berlin, J. Influence of accumulated phosphate on culture growth and formation of cinnamoyl putrescines in medium-induced cell suspension cultures of Nicotiana tabacum. Planta 1981, 153, 582-585. Kondo, O.; Honda, H.; Taya, M.; Kobayashi, T. Comparison of growth properties of carrot hairy root in various bioreactors. Appl. Microbiol. Biotechnol. 1989, 32, 291-294. Kwok, K. H.; Doran, P. M. Kinetic and stoichiometric analysis of hairy roots in a segmented bubble column reactor. Biotechnol. Prog. 1995, 11, 429-435. McKelvey, S. A.; Gehrig, J. A.; Hollar, K. A.; Curtis, W. R. Growth of plant root cultures in liquid- and gas-dispersed reactor environments. Biotechnol. Prog. 1993, 9, 317-322. Nuutila, A. M.; Lindqvist, A.-S.; Kauppinen, V. Growth of hairy root cultures of strawberry (Fragaria ananassa Duch.) in three different types of bioreactors. Biotechnol. Tech. 1997, 11, 363-366. Ohmura, T.; Howell, R. W. Inhibitory effect of water on oxygen consumption by plant materials. Plant Physiol. 1960, 35, 184-188. Ramakrishnan, D.; Curtis, W. R. Fluid dynamic studies on plant root cultures for application to bioreactor design. In Advances in Plant Biotechnology; Ryu, D. D. Y., Furusaki, S., Eds.; Elsevier: Amsterdam, 1994; pp 281-305. Ramakrishnan, D.; Salim, J.; Curtis, W. R. Inoculation and tissue distribution in pilot-scale plant root culture bioreactors. Biotechnol. Tech. 1994, 8, 639-644. Sharp, J. M.; Doran, P. M. Characteristics of growth and tropane alkaloid synthesis in Atropa belladonna roots transformed by Agrobacterium rhizogenes. J. Biotechnol. 1990, 16, 171-186. Taya, M.; Yoyama, A.; Kondo, O.; Kobayashi, T.; Matsui, C. Growth characteristics of plant hairy roots and their cultures in bioreactors. J. Chem. Eng. Japan 1989, 22, 84-89.

The transient peaks and troughs in the time-course data for gas-phase carbon dioxide concentration, dissolved oxygen tension, and net specific dissolved oxygen depletion rate (Figure 4) were reproducible phenomena during the reactor cultures under standard operating conditions. The reason for the observed transients is unclear, although their timing under standard operating conditions corresponded roughly with depletion of phosphate from the bulk medium (and thus a probable switch to utilization of stored phosphate, Knobloch et al., 1981) and/or complete hydrolysis of sucrose in the bulk medium. However, because there were significant spatial variations in local conditions within the root bed, and because bulk medium concentrations were not the same as the local nutrient levels available to the roots (Figure 7), relationships with bulk medium properties have little meaning when the system is heterogeneous. An alternative possible correlation for the time-course transients is with the onset of root lateral branching at day 10.

Conclusions
This work demonstrates the effectiveness for hairy root culture of a novel liquid-dispersed reactor equipped with a vertical mesh inoculum support. The biomass distribution remained homogeneous down the height of the culture vessel for the first 14-16 days of culture, after which growth in the upper sections close to the sources of oxygen and recirculated medium was greater than at the base. Root growth in the reactor continued until all sugars were depleted, representing an improvement over previous submerged reactor designs applied for culture of A. belladonna hairy roots. Oxygen enrichment of the gas phase had a negligible effect on the cultures, highlighting the greater importance of liquid-solid oxygen transfer resistances in the root bed compared with gas-liquid processes. The tendency of the root bed to accumulate liquid and prevent proper drainage was identified as a major problem with liquid-dispersed reactors for root culture, having the potential to severely reduce rates of liquid-solid oxygen transfer. Strategies aimed at improving liquid drainage are required to further enhance culture performance. Possibilities along these lines for further investigation include reducing the density and length of root hairs either by root line selection or manipulation of culture conditions, better design of the height and diameter of the root bed to avoid generation of excessive pressure drops preventing medium flow in the root bed, and operation with intermittent bursts of air pressure to release entrained liquid.

Acknowledgment
We are grateful to Russell Cail and Malcolm Noble for technical assistance. This work was supported by the Australian Research Council (ARC), the Horticultural Research and Development Corporation, Australia, and the New South Wales Department of Agriculture, Horticultural Stock and Nurseries. P.M.D. acknowledges additional support from an ARC Queen Elizabeth II Research Fellowship.

References and Notes

Armstrong, W.; Gaynard, T. J. The critical oxygen pressures for respiration in intact plants. Physiol. Plant. 1976, 37, 200-206. Bates, T. R.; Lynch, J. P. Stimulation of root hair elongation in Arabidopsis thaliana by low phosphorus availability. Plant Cell Environ. 1996, 19, 529-538. Buer, C. S.; Correll, M. J.; Smith, T. C.; Towler, M. J.; Weathers, P. J.; Nadler, M.; Seaman, J.; Walcerz, D. Development of

Biotechnol. Prog., 2000, Vol. 16, No. 3

Tescione, L. D.; Ramakrishnan, D.; Curtis, W. R. The role of liquid mixing and gas-phase dispersion in a submerged, sparged root reactor. Enzyme Microb. Technol. 1997, 20, 207-213. Weathers, P. J.; Wyslouzil, B. E.; Whipple, M. Laboratory-scale studies of nutrient mist reactors for culturing hairy roots. In Hairy Roots: Culture and Applications; Doran, P. M., Ed.; Harwood Academic: Amsterdam, 1997; pp 191-199. Whitney, P. J. Novel bioreactors for the growth of roots transformed by Agrobacterium rhizogenes. Enzyme Microb. Technol. 1992, 14, 13-17. Williams, G. R. C.; Doran, P. M. Investigation of liquid-solid hydrodynamic boundary layers and oxygen requirements in hairy root cultures. Biotechnol. Bioeng. 1999, 64, 729-740. Wilson, P. D. G. The pilot-scale cultivation of transformed roots. In Hairy Roots: Culture and Applications; Doran, P. M., Ed.; Harwood Academic: Amsterdam, 1997; pp 179-190.

401
Wilson, P. D. G.; Hilton, M. G.; Robins, R. J.; Rhodes, M. J. C. Fermentation studies of transformed root cultures. In Bioreactors and Biotransformations; Moody, G. W., Baker, P. B., Eds.; Elsevier: London, 1987; pp 38-51. Wilson, P. D. G.; Hilton, M. G.; Meehan, P. T. H.; Waspe, C. R.; Rhodes, M. J. C. The cultivation of transformed roots from laboratory to pilot plant. In Progress in Plant Cellular and Molecular Biology; Nijkamp, H. J. J., van der Plas, L. H. W., van Aartrijk, J., Eds.; Kluwer Academic: Dordrecht, 1990; pp 700-705. Yu, S.; Mahagamasekera, M. G. P.; Williams, G. R. C.; Kanokwaree, K.; Doran, P. M. Oxygen effects in hairy root culture. In Hairy Roots: Culture and Applications; Doran, P. M., Ed.; Harwood Academic: Amsterdam, 1997; pp 139-150.

Accepted for publication March 28, 2000. BP0000306