Journal of Biotechnology, 23 (1992) 291-301

291

© 1992 Elsevier Science Publishers B.V. All rights reserved 0168-1656/92/$05.00

BIOTEC 00737

Production of recombinant human interferon-a 1 by Escherichia coli using a computer-controlled cultivation process Xiao-Ming Yang

1, Lun Xu

2 and Lee Eppstein

t

~ R&D Lab, New Brunswick Scientific Co., Inc., Edison, New Jersey, U.S.A. 2 Shanghai Institute of Biological Products, Shanghai, China

(Received 26 August 1991; revision accepted 16 November 1991)

Summary Genetically engineered E. coli K12 BMH-71-18 with plasmid PBV-867 was used for constitutive expression of human interferon-a 1 (IFN) with a defined medium. A manual, time-based, fed-batch cultivation process produced a cell density of 26.3 g 1-1 (OD550 89), an I F N activity of 1.55 x 108 I U 1-t and a specific I F N productivity of 0.65 × 106 I U g-1. An analysis was conducted to characterize the problems involved in the high density microbial processes of recombinant protein production. The strategy suggested by the analysis is to establish a nutrient feeding profile that improves both the plasmid stability and the overall productivity of IFN. The nutrient feeding procedure developed here was based on the growth dynamics and a glucose consumption model. By using this procedure to continuously supply nutrients during cultivations, cell density reached 58 to 80 g 1-t and the specific IFN productivities of these runs were increased over that of the manual process. Nutrient feeding rates were found to affect the specific IFN productivity substantially. The optimized process achieved an I F N activity of 1.26 × 109 I U 1-t, a cell density of 58 g 1-1 and a specific I F N productivity of 2.2 × 107 IU g - 1. More significantly, the overall productivity IU 1- t

Correspondence to: X.-M. Yang, R&D Lab, New Brunswick Scientific Co., Inc., 44 Talmadge Rd.,

Edison, NJ 08818-4005, U.S.A. Nomenclature: ~, specific growth rate of a culture (h- ~); Go, initial glucose consumption before starting

feeding procedure; Gt, total glucose addition at time t; k, time constant in Eqn. (2); t, cultivation time (h); X, total amount of cells (g).

292 h-1 of the optimized, computer-controlled cultivation process was increased 12.9fold over that of the manual cultivation process. Interferon-oq; Recombinant E. coli cultivation; High cell density microbial process; Fed-batch process; Computer control

Introduction

Efficient microbial processes for genetically engineered microorganisms enable mass production of pharmacologically active human gene products which were previously infeasible to produce commercially from natural resources. A key issue in the development of an efficient microbial process for recombinant microorganisms is the optimization of productivity or the ratio of final product concentration to process time (Zabriskie and Arcuri, 1986; Allen and Lull, 1987). For most recombinant Escherichia coli strains which produce intracellular proteins, the final product concentration is equal to the product of final cell density and specific productivity (amount of product produced per cell). Therefore, for intracellular protein production, high productivity requires high cell density, high specific productivity and minimum process time. Initially, microbial process development focused on the optimization of final cell density and biomass productivity using wild-type E. coli and E. coli strains that carried foreign plasmids but did not express recombinant products (Allen and Luli, 1987; Bauer and Shiloach, 1974; Gleiser and Bauer, 1981; Mori et al., 1979). The final cell densities of more than 100 g 1-1 dry cell weight (DCW) were achieved (Mori et al., 1979; Eppstein et al., 1989). However, it was commonly observed that the specific productivity of a recombinant product (amount of product per cell) might decline considerably with the increased cell density (Zabriskie and Arcuri, 1986; Shimizu et al., 1988; Siegel and Ryu, 1985; Mizutari et al., 1986). Thus the overall productivity of a recombinant product may not be increased significantly even though high biomass productivity is achieved. The reduction of the specific productivity may result from one or more of the following causes: (1) plasmid instability including plasmid shedding and expression instability; (2) excessive concentrations of nutrients which inhibit growth, product expression or both; (3) lack of some components required for product synthesis; (4) accumulation of inhibitory metabolic products. For a constitutive expression system, as the same as a single-stage process (Yang, 1991), the attainment of sustained expression stability along with increased cell density is vital to reach high protein productivity. The solution for the first problem relies on recombinant D N A technology and the adjustment of growth conditions that affect plasmid stability. In chemostat cultures, the nature of the growth limiting nutrients had profound effect on the plasmid stability (Godwin and Slater, 1979). Thus, a stabilized nutrient concentration profile during cultivation process is expected to improve plasmid stability. The second and third problems

293 are also solved by providing a continuous nutrient feeding profile during cultivation to maintain the optimum nutrient concentrations for product synthesis. For some recombinant strains whose growth is inhibited by the accumulation of toxic metabolites, a growth rate limiting nutrient feeding schedule was applied to eliminate inhibition (Yang, 1991). This inhibition did not occur with the strain BMH-71-18, and it will not be discussed in detail here. A microbial process in which a stable nutrient concentration profile is maintained during the entire cultivation process is expected to achieve high productivity of recombinant products. In a manual fed-batch cultivation process, the infrequent addition of nutrients leads to sharp changes in the nutrient concentrations of the culture and significant deviation from optimum conditions. But with the aid of a computer, continuous adjustment of the nutrient feeding can be adjusted automatically according to the nutrient consumption. This paper will discuss the development of a supervisory nutrient addition model based on growth dynamics and the results of the manual fed-batch cultivation process, and demonstrate the model's application in the optimization of the productivity of recombinant human interferon-a 1 (IFN-a l) expressed in E. coli.

Materials and Methods

Organism

The strain used for constitutive expression of human interferon-a1 in this study was E. coli K12 BMH-71-18 (A[lac, Pro], F'lacIqZ AM15 Pro+). The plasmid was constructed by p8212, pUR-222, pBV-114 and pKC-30, which has PL promoter, A a and the gene coding for IFN-a~ (Hou et al., 1984). Cultivation media

LB medium supplemented with 500/zg m1-1 ampicillin was used for the seed culture. The defined medium used for the manual fed-batch cultivation contains: 1 g 1-1 KeHPO4; 1.5 g 1-1 KH2PO4; 1.25 g 1-1 (NH4)2SO4;3.4 g 1-1 MgSO4 • 7HeO; 0.07 g 1-1 Vitamin B1 and 3 ml l-1 trace metal, and the feeding solution contains 50% glucose. The results of a manual fed-batch cultivation and the typical elemental composition of E. coli cell (Bauer and Shiloach, 1974) were used to formulate the media in Table 1. This medium composition was designed to support the growth of E. coli up to a cell density of approximately 80 g l- 1 dry cell weight (DCW). It was used in all computer-controlled cultivations. Cuhivation

The strain was stored in LB agar slants with 500 /xg m1-1 ampicillin. It was aseptically transferred to 50 ml of LB ampicillin medium in a 500 ml flask. This

294 TABLE 1 Media composition of computer-controlled cultivation Component

K 2 HPO4 KH 2 PO4 (NH4)2SO 4 MgSO 4 •7H 2° Vitamin B 1 Trace metal (ml) Glucose Antifoam (ml) Volume (I) pH

Concentration (g 1-1) Initial medium

Feeding medium

3.0 5.0 4.0 2.0 0.1 3.5 30.0 0.5

30.0 5.0 2.0 500 -

1.500 7.0

0.5

culture was grown for 11 h at 37°C and 200 rpm in a New Brunswick Scientific (NBS) G-25 incubator shaker. The entire seed culture was then used to inoculate the cultivation medium in a 2.5 I bioreactor. The OD550 values of the seed cultures for the manual cultivation and the computer-controlled cultivation Nos. 1, 2, 3 and 4 were 4.5, 4.0, 2.9, 3.5 and 2.0, respectively. The initial volumes of the medium in the bioreactor were 1.8 1 for the manual process and 1.5 1 for the computer-controlled process. The nutrient feeding procedure started 1 h after inoculation. Other initial parameters included agitation 300 rpm, air flow 2 1 min -~, temperature 37.0°C and DO 50%. During fed-batch process, pH was automatically adjusted to 7.0 by adding concentrated ammonium hydroxide.

Bioreactor and control A BioFlo III 2.5 1 bioreactor (NBS) was used in this study. A general purpose multi-loop controller Model ML-4100 (NBS) was used to regulate the ratio of pure oxygen and air in the sparge air stream. The Advanced Fermentation Software package, AFS (NBS) executed by an IBM P S / 2 personal computer was used to implement all supervisory control tasks and calculations. The bioreactor and control system were previously described in detail (Yang, 1991).

Analytical methods The optical density was measured at 550 nm with a Perkin-Elmer Lambda 4B U V / V I S spectrophotometer. DCW g 1-1 was determined using the method described previously (Eppstein et al., 1989). Glucose concentration of the culture was analyzed enzymatically with an A L P K E M glucose analysis kit. Biological interferon assays were performed on MDBK cells (Familletti et al., 1981) at Dr S. Pestka's lab of Robert-Wood Johnson Medical School, Piscataway, New Jersey.

295

Results and Discussion Manual fed-batch cultivation process The profiles of important parameters during this cultivation are shown in Fig. 1. The agitation speed and air flow rate reached their maximum values at 11.5 h. After this point DO of the culture dropped quickly as shown in Fig. lb. The temperature was gradually reduced from 37.0°C to 25°C to avoid oxygen limitation. Cell growth slowed as the temperature decreased in Fig. la even though there was no limitation of oxygen or glucose as shown in Fig. la. Cell density reached 26.3 g 1-1 DCW (OD550 89). The ratio of DCW to OD550 for this strain is 0.295, which is very close to the ratio of other E. coli strains (Eppstein et al., 1989; Yang, 1991).

I

25

I

I

,

~

,

~

, 100

~

a)

I

~0

0 ~ 0 / 20

•

7

vo

|

./.

v

8o

d o o

~o

15

4o 10

~

__~,.__.~

m

3o

O

~o •

5

~0

~0

•

to

0

~ 0

2

4

6

~

~

Z

L

L

I

I

I

8

10

12

14

16

18

20

22

Time

0

24

(h)

100

40

90

35

80

30 ~

70 25

60

C,

~

50

zo

40

~ ~

30

~. ~

e~

20

5

tO I

0 0

'~

,

I

I

I

I

I

t

I

I

I

4

6

8

10

1~

14

t~

~8

20

Time

(h)

Fig. 1. The manual fed-batch cultivation process.

I

~

0

~4

E-, ,,

296

4 3 ~

2

7 1 0

-2 -3

_41 -5

I

I

I

I

I

I

I

I

I

I

I

2

4

6

8

10

1~

14

16

18

20

22

Time

24

(h)

Fig. 2. Growth dynamicsof the manual cultivationprocess. The growth yield was 0.43 g g - ~. The manual, time-based, glucose addition profile caused sharp changes in glucose concentration as shown in Fig. la. Interferon activities after 12 h of cultivation and at the end of the cultivation were 5.72 × 107 IU 1-~ and 1.55 × 108 IU 1-~ respectively. The specific productivity was 4.3 × 106 IU g-~ and 5.9 × 106 IU g - l , respectively. The overall IFN productivity of this cultivation process was 6.5 × 106 IU I-~ h-1.

Growth dynamics analysis The specific growth rate, /~, is defined as /'~

d(ln X ) dt

(1)

where X is the total amount of cells (g) and t is cultivation time. A plot of In(X) versus t using the data from the manual fed-batch experiment is shown in Fig. 2. The slope of the curve is the specific growth rate ~ according to Eqn. (1). The specific growth rate started to decrease even before temperature shifted at 12 h (Fig. 2). The inability of the culture to maintain a sustained period of exponential growth was assumed to be related to the sharp changes in nutrient concentrations indicated by the glucose concentration shown in Fig. la. These changes were caused by infrequent addition of glucose solution. As discussed previously, a continuous nutrient feeding procedure can be used to maintain a stable specific growth rate in the culture. An exponential glucose addition model (Yang, 1991) based on growth dynamics [Eqn. (1)] and the mass balance of cell synthesis was used for this purpose: G t = G Oexp(kt)

(2)

where G t (g) is the total amount of glucose to be added to the culture, G O is the

297 initial glucose consumption before starting the feeding procedure, k is a time constant whose optimum value is the instantaneous specific growth rate, ~. If k /z, glucose concentration in the culture will increase and excessive concentration might occur. A computer program developed in our previous work (Yang, 1991) allows the glucose addition to follow the profile of Eqn. (2). The feeding solution was composed of glucose, ammonium sulfate, magnesium sulfate and vitamin B 1 as shown in Table 1. However, only glucose concentration was monitored since it was considered to be a representative of all nutrient concentrations. The results in Fig. lb show that DO could not be maintained above 10% even though the agitation speed and air flow rate were adjusted to their maximum values. Reduction of temperature to compensate for oxygen demand dramatically reduced the growth rate as shown in Figs. la and 2, thereby adversely affecting IFN productivity. To overcome oxygen limitation, oxygen-enriched air was used as the sparge gas. An air-oxygen mixing device and control system is described in the Materials and Methods section. A fed-batch cultivation process to realize the above nutrient feeding profile and oxygen supplementation was studied in the following experiments.

Computer-controlled cultivation Using the nutrient feeding strategy, the parameter k in Eqn. (2) is the key variable to be optimized. A series of cultivation runs were conducted to apply the feeding procedure with k of 0.3, 0.4, 0.42 and 0.45. These runs are respectively defined as cultivation No. 1, 2, 3 and 4. The selected k values were based on the specific growth rate between 0.2-0.5 h-1 during the fast growth phase (Fig. 2) in the manual process. The growth curves and glucose concentrations in these cultivations are shown in Fig. 3. The results in Fig. 3a show a constant exponential growth phase during each cultivation run, which is in contrast with the results of the manual process (Fig. 2). And high cell densities of 66, 64, 80 and 58 g 1-1 DCW were achieved in cultivation Nos. 1, 2, 3, and 4 respectively. It could be assumed that during the exponential growth phases of these cultivations, the nutrient conditions for growth were relatively stable and no significant metabolic inhibition occurred. The specific growth rates of cultivation Nos. 1, 2, 3 and 4 are 0.4, 0.46, 0.45 and 0.47 h-1, respectively. These cultivation processes were terminated when their growth rates showed significant decrease. In the first run with a k of 0.3 (