Cytotechnology 6: 55--63, 1991. 9 1991 KluwerAcademic Publishers. Printed in the Netherlands.

On-line monitoring of monoclonal antibody formation in high density perfusion culture using FIA Christel Fenge 1, Elisabeth Fraune 2, Ruth Freitag 3, Thomas Scheper 4 and Karl Schtigerl 5 1,2B~Braun Diessel Biotech GmbH, Schwarzenberger Weg 73, D-3508 Melsungen 1,3-5lnstitut fiir Technische Chemie, Universit~it Hannover, Callinstr.9, D-3000 Hannover, Germany Received 8 August 1990; accepted 25 January 1991

Key words: flow injection analysis, high cell density, hybridoma cells, on-line monitoring of monoclonal antibody formation, on-line glucose and lactate measurements, perfusion cultivation Abstract An automated flow injection system for on-line analysis of proteins in real fermentation fluids was developed by combining the principles of stopped-flow, merging zones flow injection analysis (FIA) with antigen-antibody reactions. IgG in the sample reacted with its corresponding antibody (a-IgG) in the reagent solution. Formation of insoluble immunocomplexes resulted in an increase of the turbidity which was determined photometrically. This system was used to monitor monoclonal antibody production in high cell density perfusion culture of hybridoma cells. Perfusion was performed with a newly developed static filtration unit equipped with hydrophilic microporous tubular membranes. Different sampling devices were tested to obtain a cell-free sample stream for on-line product analysis of high molecular weight (e.g., monoclonal antibodies) and low molecular weight (e.g., glucose, lactate) medium components. In fermentation fluids a good correlation (coefficient: 0.996) between the FIA method and an ELISA test was demonstrated. In a high density perfusion cultivation process mAb formation was succesfully monitored on-line over a period of 400 h using a reliable sampling system. Glucose and lactate were measured over the same period of time using a commercially available automatic analyser based on immobilized enzyme technology. Abbreviations: TIA - Turbidimetric immunoassay; mAb - Monoclonal Antibody

Introduction The industrial scale production of biologically active products has led to the development of new techniques for the on-line estimation of process parameters. Although traditional in-line sensors for monitoring physical and chemical parameters such as temperature, pH and dissolved oxygen are commonly used, many publi-

cations in the last few years have described systems for the on-line analysis of low molecular weight medium compounds ( G a r n e t al., 1989). For on-line analysis of high molecular weight components such as proteins and nucleic acids (Schendel et al., 1989), a highly sophisticated approach is needed. Flow injection analysis (FIA) is an advanced method that can be used for on-line analysis of proteins and was shown to

56 have the advantages of speed, selectivity, and was adaptable to a wide concentration range (Ruzicka et al., 1983). The demand for large quantities of monoclonai antibodies has stimulated the development of perfusion bioreactors. Perfusion systems offer the advantage of operation with high cell densities and thus high volumetric production rates (Takazawa et al., 1988). The development of appropriate on-line techniques to monitor product formation and nutrient supply in continuous cell fermentations offers new and effective optimization strategies to control bioprocesses. Our aim was to establish an on-line IgG determination method and to connect it directly to a real longterm animal cell fermentation. Additionally the levels of some important metabolites should be monitored on-line with already existing automatic analyzers.

~

feed

fl

perfu=~ion module

ho~ut

2

o0ton, f,t

bioreoctor

0oo.b.

,

~

pump

Fig. 1. Experimental setup of perfusion bioreactor with sampling system for on-line IgG determination. TFU: tangential filtration unit, BIOPEM: stirred filtration cell. ment. After an initial batch phase, perfusion was started when cell densities reached 106 cells/ml.

Off-line analysis Materials and methods

High density perfusion cell culture The perfusion experiments were performed in a 1.6 L stirred tank bioreactor with bubble-free aeration via silicon tubing designed for the special requirements of shear-sensitive animal cells (Biostat MC, B. Braun Diessel GmbH). Temperature, pH and dissolved oxygen (DO) were measured via in-line probes. DO was controlled at 20% of air saturation and pH at 7.2. A newly developed static perfusion filter for freely suspended animal cells consisting of hydrophilic microporous tubular membranes was used as an internal filtration unit (Fraune et al., 1988). Medium exchange rates were adjusted by continuous addition of fresh medium and by a harvest pump activated by a level probe. The experimental setup is shown in Fig. 1. The mouse-mouse hybridoma cell line DB9G8 (ATCC HB 124)(Schroer et al, 1983) producing a IgG2a ~:-type monoclonal antibody against human insulin was cultivated in serum-free DME medium. CPSR4 (Controlled Process Serum Replacement 4, Sigma) was used as serum supple-

Samples were taken daily and stored at - 2 0 ~ for further analysis. Cells were counted with a haemocytometer and viability was measured using the Trypan Blue dye exclusion method. Glucose and lactic acid concentrations were determined with an automatic analyzer (YSI 2000, Yellow Springs Instruments). Protein levels were quantified by Coomassie Brilliant Blue staining using Bradford or micro-Bradford assays (Bradford et al., 1976). A sandwich ELISA was applied to determine the IgG concentration. The coating antibody was affinity-purified goat anti-mouse IgG. The same antibody conjugated with alkaline phosphatase (both purchased from Dianova, Hamburg) was used as detection antibody.

On-line analysis Sampling devices. Two different sampling devices were used to obtain a sample stream. The BIOPEM (Fig. 2, B. Braun Diessel GmbH) is a stirred filtration cell especially developed as sampiing device for on-line monitoring (Kroner et al., 1988). The other device is a microporous filtra-

57 ~1~~

retentate

blank 37"C

a jfL fi I trate ~"~

-

-

-

-

photomL~.Jeter

a s

t

i

rrer

membrane fi I trate

b ~~rane inletF~--~.~ ~ ,

/ ,/ / / / / ~ - ~ 2 , 1 ~ * o u t l e t

0

- buffer

j

reactioncoil photometer

(~

pump

0

sampleinjector

carrier

reagent(antibody)

Fig. 2. Schematicdrawings of a) BIOPEMwith magneticstirrer (cell suspensionfrom the bioreactoris recirculatedthrough the stirred filtrationcell), b) tangential filtrationunit. tion probe which was equipped with a section of modified PTFE tubing (pore size 1 l.tm) and inserted into the vessel. In order to guarantee sterility, a disk-shaped tangential filtration unit (diameter 47 mm, equipped with 0.22 ktm PVDF membranes (Durapore type, Millipore)) was used as sterile barrier in combination with one of the other filtration units. Both systems were tested to examine their effectiveness with different membranes. The sampling devices were interfaced to the reactor and the automatic analysers for IgG and glucose/lactate determination as shown in Fig. 1.

FIA System for on-line IgG determination. A stopped-flow, merging zones flow injection analysis system was assembled for on-line monitoring of monoclonal antibody concentrations in real fermentation fluids (Freitag et al, in press). The experimental setup is shown in Fig. 3. The antibody concentration was detected by a turbidimetric immunoassay (ImmunoFIA). A modified Atari computer was used for data processing. The cell free sample stream provided by the sampling system was pumped into the flow injection analyzer. Two coils corresponding to blank and sample were flushed with sample solution. The reagent coil was filled discontinuously with 50 ~tl of reagent solution after removal of the remaining buffer by an air bubble in order to

Fig. 3. Scheme of the FIA system for turbidimetric immunoassay (ImmunoFIA). Sample and reagent were injected into a buffer stream (37~ by injectionvalves and incubatedat 37~ in a 70 cm longteflontube (manifold).Turbiditywas measured photometrically in 10 BL microcuvettes.The operation of the differentvalves was controlledby a combinationof timers. reduce the amount of expensive reagents needed. Reagent and sample were injected into separate reaction buffer streams and after merging incubated for 90 seconds. Therefore the flow was interrupted for the incubation time. During this time, the turbidity of the sample blank was assayed by injection into another buffer stream. The optical density of the sample and blank were measured at 340 nm. The data were transferred to a computer (Atari 1040 ST). A calibration function programmed into the computer was used to convert the OD340 readings into antibody concentrations.

Turbidimetric Immunoassay (TIA). A turbidimetric immunoassay (Freitag, 1989) was used to determine the IgG content of the fermentation samples. The mouse-mouse monoclonal antibody (IgG K-type) that was produced by the cells was incubated with the reagent solution consisting of rabbit anti-mouse antibody against mouse IgG (Sigma M-7023). The reaction buffer was composed of a 0.01 mol/L sodium phosphate buffer, pH 7.2, with 30 g/L polyethylene glycol 6000 and 4.5 g/L sodium chloride. At constant antibody concentration, the amount and size of the immunocomplexes formed, and therefore the turbid-

58

1.2

The analysis principle is based on immobilized enzyme technology. Analysis runs were performed at distinct time intervals.

furbldl~ (OD 340 rim)

Results and discussion

0,8 0,6 0,4 0,2 0

I

0

I

I

I

200 400 600 800 IgG conce~,;r,~;;on [mg/L]

1000

Fig. 4. Calibrationfunctionof the turbidimetricimmunoassay. ity of the sample, depended on the antigen concentration. The optical density was found to increase with increasing antigen concentration (antibody excess) until equivalent concentrations of antigen and antibody were reached. Further increases in the antigen concentration (antigen excess) led to a decrease in turbidity. By selecting an appropriate antiserum dilution, the detectable concentration range of antigen could be extended over several orders of magnitude. In preliminary perfusion experiments, IgG concentrations from 0.2 mg/L up to 1000 mg/L were reached. Therefore, a calibration function over a broad concentration range was needed to avoid complicated dilution steps. Rabbit anti-mouse IgG (Sigma M-7023) was found to cover the necessary concentration range. Figure 4 shows the calibration function used in the perfusion experiments described in this paper. Mouse IgG which was purified from pooled normal serum (Sigma 15381) was used as standard antibody. Glucose and lactate concentration determination. The commercially available automatic analyzer (YSI 2000) was interfaced to the on-line sampiing system for glucose and lactate quantification.

Two similar perfusion experiments with the hybridoma cell line HB 124 (in the further text referred as experiment I and experiment II) were carded out to investigate the potential and problems of on-line determination of IgG in real fermentation fluids. The approach was to develop a flow injection analysis system coupled to a bioreactor. For on-line analysis in animal cell cultivations, a special requirement is long-term stability of sampling devices and assays. To meet this requirement different sampling devices and experimental assay procedures were tested in both experiments.

Cell growth Figure 5 shows the growth kinetics of experiment I which is typical for a hybridoma perfusion culture. After inoculation with 0.14.106 viable cells/mL, the cells grew in the initial batch phase until a cell density of 1.106 viable cells/mL was reached. At this time, perfusion of the suspension culture was started. Exponential growth was observed over a period of two days during the batch mode and could be extended over another two days due to perfusion of the culture with fresh medium. In the exponential growth phase a specific growth rate of lx=l.2 d -1 was determined. Medium exchange rates between D=0.8 d -1 and D 1.1 d -1 led to an extensive stationary growth phase with an apparent growth rate of lxapp---0.07 d -1. In experiment II an apparent growth rate of ktapp=0.18 d -1 was determined in the stationary growth phase. The 2.6-fold increase of growth in experiment II was caused by higher dilution rates (D= 1 to 1.52 d -1) during the perfusion phase. Maximum cell densities of 2.107 cells/mL were reached in both experiments.

59 106celle/mL

dilution

rate

100

[t/d]

5

lacfafe ['g/L]

3

on-line

glucoee [g/L]

5

4 10

3 I I i

--0-

v~b~

--

dillon

2,5

ceil= rate

0,1

2

1

0,01 0

i 100

i 200 time

t 300

i 400

0 5OO

[h]

A

1.5

Fig. 5. Growth of HB 124 cells under perfusion cultivation conditions in DME medium with CPSR4 as serum supplement; CPSR4 concentration was decreased from 5% during initial batch phase to 0.8% at the end of the cultivation. 0,5 [] A

Cell metabolism 0

In order to monitor the nutrient levels in the medium and to obtain information about the metabolic state of the culture, glucose and lactate concentrations were determined on-line. Glucose and lactate. T h e influence of the sampiing procedure on the on-line glucose and lactate measurements was determined by comparing these data to the corresponding off-line values (measured with the automatic analyzer YSI 2000, too) as shown in Fig. 6. The maximum deviation was below 10% and did not depend on the sampling procedure and equipment. Figure 7 illustrates glucose consumption and complementary production of lactate during perfusion cultivation of the HB 124 cells (experiment I). The glucose and lactate concentrations could be maintained at steady-state levels (glucose 2 g/L, lactate 1 g/L) during the stationary phase due to perfusion of the cell suspension. The lower yield coefficient Y(1, g) for the conversion of glucose to lactate during stationary growth reflects a change in the metabolic pathway of energy generation from glucose (Fig. 7). The efficiency of glucose utilization increases, i.e., an increase in glucose oxidation. This phenomenon has also been described by other authors (Reuveny et al., 1986, Glacken et al., 1986).

I

0

I

gluooee laef=to

I

J 0

I

5

1 2 3 4 off-line measurement of oonoenfrafion

Fig. 6. Correlation of on-line and off-line glucose and lactate measurements using the automated analyzer YSI 2000. A correlation coefficient of 0.99 was calculated for glucose determination and a correlation coefficient of 0,93 was calculated for lactate.

8

v0,e)

concentration [g/L]

2 0 ~

glur

-2 -4

2

f

~

~ -6 --8

0

1O0

200

300 time

400

5OO

[h]

Fig. 7. Glucose and lactate concentrations during perfusion cultivation of liB 124 cells. Y(1, g) is the yield coefficient for the conversion of glucose to lactate (Y(I, g) ratio of produced lactate/L/d and consumed glucose/L/d).

On-line

analysis of productformation

Prelminary experiments. In order to compare IgG determination by FIA (on-line) and ELISA (off-

60

250

IgG [mg/L]

140

IgG conc. [mg/L]

120 200 150

100 BO 6O

--e-- FiA

100

~

40

o

'

i

~

D

50 0

_2.2 10

,

,

20 time [hi

30

2O 40

Fig. 8. Correlationbetweenturbidimetricflowinjectionanalysis (FIA) and ELISAfor the determinationof IgG in a simulated fermentation.

line) a cell free culture supematant containing 540 mg/l., IgG (determined by ELISA) harvested after 49 days of perfusion was used to prepare fermentation fluid gradients as a simulation of a real fermentation. The measurement cycle was 2.5 rain. and quantification in triplicate was card e d out every hour. These simulation experiments lasted between 12 and 36 h. As shown in Fig. 8, the correlation between the turbidimetric flow injection analysis and off-line ELISA deterruination of IgG indicated good agreement (correlation coefficient: 0.996). These results demonstrate the ability of this on-line turbidimetric FIA system to accurately determine the IgG concentration in real fermentation fluids. Fermentation experiments. Using the FIA method for on-line IgG determination described above, the kinetics of mAb production during perfusion cultivation of hybridoma cells was monitored. Analysis runs were performed every 8 h in triplicate since the mAb formation kinetics did not affort shorter time intervals. Figure 9 shows the IgG production kinetics measured in two perfusion experiments of HB 124 cells. In experiment II the shown results were measured on-line whereas in experiment I the shown IgG concentrations were obtained using the FIA method off-line with an aliquot of cell free culture supernatant. Maximum IgG concentrations of 110 mg/L were reached at the end of perfusion cultivation in both experiments. In ex-

0

expsr|mentII

zx experimentI 100

200

300

tim~[h]

400

500

Fig. 9. IgG determinationusingthe FIA method.ExperimentI: samples measured off-line; experiment II: samples measured on-line. In experimentI the BIOPEMwas insertedafter 320 h. This causedthe rapid decrease in IgG concentrationdue to the dilution of cell suspension with fresh medium which was necessary to fill the BIOPEM.

perirnent I, the cells required much longer times to produce high levels of IgG, indicating that the lower specific growth rate in that experiment during the perfusion phase caused the decrease in productivity. For experiment II, the following production parameters were calculated using equations derived from the mass balance of the system: the average specific IgG production rate was approximately 6 ~tg/106 viable cells/day, the highest specific production rates were 14 ktg/106 viable cells/d (in the exponential growth phase), the maximum volumetric IgG production rates were greater than 60 mg/L/d and 550 mg IgG were produced in total. As in the preliminary experiments, IgG concentrations in the sample stream were measured by both FIA and ELISA. In contrast to the results from the simulated fermentations, some discrepancies between the two analytical methods were observed. Good agreement was observed at low mAb concentrations, yielding a correlation coefficient of 0.94 for the concentration range 5-15 mg/L. However, for IgG concentrations greater than 15 mg/L (experiment II), the correlation coefficient fell to 0.83. These correlations allow the interconversion of data obtained by the two antibody determination methods. In the simulated fermentations, the changes in all medium components were a function of the same mixing gradient. In real fermentations, these changes can

6]

differ completely from one another, thus affecting the assay accuracy. The fact that a polyclonal standard was used in the FIA and ELISA determinations while a monoclonal antibody was measured should also be taken into account. A better comparison between ELISA and TIA could be achieved using the product mAb in purified form as the standard imrnunoglobulin.

1.2

permeability experiment I

I 0.8 o.$ 0.4 ~M with

0.2

5 k~m~nembrmle

o

z

i

i

I

1 O0

200

300

400

500

time [h]

Sampling procedure Different filter arrangements were tested for their applicability in long-term animal cell fermentations. In perfusion cultivation processes, problems might occur due to clogging of the filter materials by cells and cell debris. In addition, concentration polarization should be considered. This phenomenon might occur due to the formation of a thin protein layer on the surface of the filtration material, resulting in increased retention of the product. Both of these phenomena influence the permeability of the product and, consequently, the applicability of on-line analysis for fermentation monitoring of high molecular weight components.

On-line sampling procedure. During the initial fermentation phase (characterized by cell densities in the range of 105 to 106 cells/mL), the filtration probe was used in conjunction with the tangential filtration unit to take a cell-free sample for on-line analysis. The 0.22 ktm PVDF membrane in the tangential filtration unit guaranteed process sterility during long term operation. This filter combination was chosen to avoid mechanical stress on the cells which might negatively influence cell proliferation in the early fermentation phase. As an alternative sampling system, the BIOPEM was introduced in an external loop. It was equipped with a 0.22 l.tm PVDF membrane (Durapore) for direct connection to the FIA system or with 5 ~tm membranes (modified PTFE) in combination with the tangential filtration unit which acted as a sterile barrier. The different pumps of the sampling system were activated

permeability 1.2@

experlmllnt II

0~ 0,4

0,2

giOP~ ~filtrotion

p r o b e c o m b i n e d with T F U ~

with

, 0,22 ~m membrane

o

J

i

i

50

1oo

150

i

200 time [h]

I

I

250

300

350

Fig. 10. Productpermeabilityof differentfiltrationmembranes of the investigatedsamplingsystems.Permeabilitywas calculated by relating the IgG concentration in the corresponding filtrate to the IgG concentrationin the vessel (TFU: tangential filtration unit). A: permeabilityof 1 p.m membrane (filtration probe, option 1, Fig. 1) or 5 p.mmembrane(BIOPEM,option2, Fig. 1) respectively. B: Permeability of 0.22 p.m membrane (sample stream). automatically via a timer (Fig. 1). The product permeability of filtration membranes of different pore sizes and materials in these filtration modules and filter combinations is shown in Fig. 10. The product retention of the 1 I.tm membrane of the filtration probe was negligible for the first 100 to 150 h, but increases (i.e., decreased permeability after this time). These findings indicate that the filtration process is not only affected by the filtration unit (or combination of units) and the operation parameters but also by the physicochemical nature of the membrane. The larger pore size (5 ~tm) used in the BIOPEM (experiment I) did not perform better than the 0.22 ~tm membranes (experiment I). When membranes were utilized as an additional sterile barrier, no further loss in permeability of product occured. This observation indi-

62 IgG sample stream [mg/L] (IrlA)

140 120

100 80 60

0

E)~D[]

10 20 50 40 IgG vessel ling/L] (ELISA)

50

Fig. H. Comparison of IgG concentrations in the cultivation

vessel (ELISA) and in the sample stream (FIA). The difference of the scales is due to the discrepancies of the two analytical methods which were observed during on-line monitoring of monoclonal antibody formation in real fermentations. cates the important role o f cell debris and large proteins in filter fouling. T h e highest product permeabilities over the entire process were obtained in experiment II. Using the B I O P E M equipped with 0.22 g m P V D F membranes, more than 90% o f the cell culture supematant was recovered in the sample stream for on-line analysis. Due to the optimized sampling procedure and sampling device used in experiment II, the comparison o f the IgG concentration in the culture vessel (by E L I S A ) with the IgG concentration in the cell-free sample stream (by FIA) is in good agreement (correlation coefficient: 0.96, Fig. 11).

Conclusion The experiments described above show that the monoclonal antibody concentration in h y b r i d o m a cell cultivation processes can be monitored online using the flow injection analysis technique.

The turbidimetric immunoassay system developed in our laboratories was successfully applied to high density perfusion cultures during periods as long as 400 h. T h e best results for continuous sampling for determination o f high molecular weight m e d i u m components were obtained using the B I O P E M with 0.22 g m P V D F membranes. The stopped-flow, merging zones FIA technique is a useful tool for real-time monitoring o f macromolecular product formation in bioreactors. This was demonstrated in high cell density hybridoma culture in a newly developed perfusion bioreactor. On-line mAb concentration measurement could also be combined with online monitoring o f nutrient consumption and metabolite formation. The experimental results indicate that the FIA immunoassay system allows on-line monitoring o f product formation in continuous, long term cultivation processes, thus facilitating routine operation and safety supervision. The application o f FIA offers new possibilies for the development o f production strategies.

Acknowledgements This work was partly supported by a grant o f the Bundesminister fOr Forschung und Technologie, grant N o 0318815A, and the D E C H E M A .

References Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein using the principle of protein-dye binding. Anal. Biochem. 72: 248254. Fraune E, Fenge C and Kuhlmann W (1988) High density hybrdoma cell culture in stirred tank pcrfusion bioreactors. Abstract Book, 8th InternationalBioteehnologySymposium, Paris, 115. Freitag R (1989) PhD Thesis. University of Hannover. Freitag R, Scheper T Schiigerl K (in press) Developmentof a turbidimetric immunoassay for on-line monitoring biomolecules of high molecular weight. J. Biotechnol. Gain M, Gisin M and Thommen C (1989) A flow injection analysis system for fermentation monitoring and control. Biotechnol. Bioeng. 34" 423--428.

63 Glacken MW, Heischaker RJ and Sinkskey AJ (1986) Reduction of waste product excretion via nutrient control: possible strategies for maximizing product and cell yields on serum in cultures of mammalian cells. Biotechnol. Bioeng. 28: 13761389 Kroner KH and Papamicheal N (1988) Continuous sampling technique for on-line analysis. Process Biochemistry Probiotech: iii-vi. Reuveny S, Velez D, Miller L and Macmillen JD (1986) Comparison of cell propagation methods for their effect on monoclonal antibody yield in fermentors. J. Immunol. Meth. 86: 61--69. Reitzer I.J, Wice BM and Kenell D (1979) Evidence that glutamine, not sugar, is the major energy source for cultured HeLa cells. J. Biolog. Chem. 254: 2669-2676. Ruzicka J and Hansen EH (1983) Recent developments in flow injection analysis: gradient techniques and hydrodynamic injection. Anal. Chim. Acta 145: 1.

Schendel FJ, Baude EJ and Flickinger MC (1989) Determination of protein expression and plasmid copy number from cloned genes in Escherichia coli by flow injection analysis using an enzyme indicator vector. Biotechnol. Bioeng. 34: 1023-1036. Schroer J, Bender T, Feldmann RJ and Kim KJ (1983) Mapping epitopes on the insulin molecule using monoclonal antibodies. Eur. J. Immunol. 13: 693-700. Takazawa Y, Tokashiki M, Hamamoto K and Murakami H (1988) High cell density perfusion culture of hybridoma cells recycling high molecular weight components. Cytotechnol. I: 171-178.

Address for offprints: Ch. Fenge, c/o Braun Diessel Biotech GmbH, Schwarzenberger Weg 73, D-3508 Melsungen, Germany