Development of a turbidimetric immunoassay for on-line monitoring of proteins in cultivation processes Ruth Freitag, Thomas Scheper and Karl Schiigerl Institut fiir Technische Chemie, Universitdt Hannover, Hannover, Germany
An on-line assay f o r a thermostable pullulanase and antithrombin III (AT III) is described. The assay is based on the formation o f aggregates between the protein to be measured and antibodies raised against this protein. Assay automation was achieved by utilizing the flow injection analysis (FIA) principles. The apparatus, a stopped-flow, merging-zone manifold, is described in detail. Since the reaction used in an FIA system does not have to reach equilibrium, it was possible to reduce the time f o r an assay cycle to 2.5 rain. A method f o r simulating cultivation conditions was developed f o r assay optimization. Using this method, a detection limit o f l mg 1- i together with a standard deviation o f l.5 was found. A sandwich ELISA was used as reference assay in the case o f A T III and an enzymatic activity assay in the case o f pullulanase. Correlation coefficients o f 0.988 (AT III) and 0.976 (pullulanase) were determined. The turbidimetric assay was successfully used f o r pullulanase monitoring during a 240-h cultivation o f Clostridium thermosulfurogenes.
Keywords:On-line immunoassay; turbidimetric assay; bioprocess monitoring;antithrombin III assay; pullulanase assay
Introduction The discriminatory power of antibodies makes an immunoassay the best candidate for monitoring substances of high molecular weight in bioreactor supernatants. The fact that antibodies can be raised against virtually all proteins as well as many other biomolecules contributes to the great attractiveness of this analytical tool for clinical and biotechnological purposes. Immunoassays used in biotechnology often depend on a label (such as an enzyme or a fluorescent dye) linked to one of the reactants for detection. The distribution of the labelled substance between two phases or two types of immunocomplexes is then evaluated. The most commonly used method to assay for highmolecular-weight proteins is ELISA (enzyme-linked immunosorbent assay). ELISAs tend to be time consuming, since many dilution, washing, and incubation steps are included, especially when no routine analysis assays are used. Often samples are drawn throughout a cultivation and analyzed afterwards. Cultivation con-
Address reprintrequeststo Dr. Scheperat the Institutfor Technische Chemie, Universit~tHannover,Callinstrasse 3, D-3000Hannover1, Germany Received 15 February 1991;accepted 20 May 1991 © 1991 Butterworth-Heinemann
trol based on production rates is thus impossible. The low detection range inherent in these assays is of no advantage if substances found in mg 1-1 concentrations or higher are concerned. This will often be the case for biotechnological products. Immunoprecipitation has been used since 1935 for measuring protein concentrations. 1 The method is based on the formation of large aggregates between bivalent antibodies and multivalent antigens. The complexes formed are large enough to cause light scattering of the Mie type.l This method has become the most widely used in clinical analysis when the specific quantification of proteins above the mg 1-1 level is desired. The fact that the method is easily automated has also increased its popularity. An automated immunoprecipitation analyzer of the continuous flow type was developed as early as 1970. 2 Turbidimetric detection, in which the decrease in transmission is recorded, can also be f o u n d ) Most applications utilize a rate measurement (kinetic method), although end point measurements are sometimes t a k e n Y The analytic potential of flow injection analysis (FIA) has been established over the last decade. 6 Although the theoretical treatment of the behavior of heterogeneous samples in an FIA system is difficult,7 turbidimetric detection has been used successfully in such systems. 8 Due to the high reproducibility of sample
Enzyme Microb. Technol., 1991, vol. 13, December
969
Papers dispersion in FIA manifolds, the reaction that is used does not necessarily need to reach equilibrium. This is an important aspect when automating relatively slow reactions such as antibody-antigen reactions. Surprisingly, there are not many applications to be found in the literature. Among the few are a system using naturally occurring binding proteins such as conc a n a v a l i n A 9 and an assay in which IgG concentrations are measured off-line in buffer solutions. 1° To our knowledge, a turbidimetric assay has never been used on-line for process control. This report describes a fully automated on-line assay that uses turbidimetric detection as well as its successful employment for product monitoring during a cultivation. Conditions for assaying two proteins, human antithrombin III (h-AT III) and a thermostable pullulanase, are given.
dium were supplied by Dr. Eberhard, Behring Werke. The medium contained between 2.5% and 4% fetal calf serum (FCS). Purified h-AT III (Kybernin HS I000, Behring Werke Marburg) was used during preliminary experiments. The starch-degrading enzyme pullulanase was produced by thermophilic bacteria (CIostridium thermosulfurogenes EM 1, DSM 3896). I1 Cultivation was carried out at 60°C and the pH was regulated at 6 during continuous culture. A mineral medium described by Antranikian et al. lz was used. Process mixture broths containing approximately 1,000 U 1-1 pullulanase were supplied by A. Spreinat, Georg-August-Universitfit, G6ttingen.
Materials and methods
During the cultivation simulations, a sterile trap (break tube) was used as a sampling device (for both on-line and off-line sampling). During the C. thermosulfurogenes cultivation, a cross-flow module (KF-200-010-SME, Microgon Inc.) was employed for on-line sampling.
Chemicals Chemicals were from Fluka unless indicated otherwise. Deionized distilled water was used to prepare solutions. PEG 6000 and BSA (bovine serum albumin) were from Sigma Chemical Co. A sodium phosphate buffer was used for the on-line assay (AT III assay: Na2HPO4, 1.18 g 1-1; NaH2PO 4, 0.23 g 1-1; NaCI, 4.50 g 1-1; PEG 6000, 40.00 g 1-1. Pullulanase assay: NazHPO4, 5.90 g 1-1; NaHzPO 4, 1.15g 1-1; NaC1, 4.5g i-1; PEG 6000, 40.00 g 1-1). The pH was adjusted to 7.4 with 1 M HCI. The same buffer without PEG 6000 was used for antibody dilution. Buffers were degassed before use.
Biologicals Antibodies against h-AT III were purchased from Behring Werke West Germany [OSAY 08/09; (rb) titer: 1.35 g ! -1 (RID)]. Antibodies were diluted threefold before use. Antibodies (rb) against pullulanase produced by C. thermosulfurogenes were kindly donated by A. Spreinat and G. Antranikian, Institut for Mikrobiologic, Georg-August-Universitat G6ttingen, Germany. The antiserum was diluted 1 : 40 prior to use. A standard solution of 1 g 1-1 h-AT III in buffer was provided by Dr. Eberhard, Behring Werke. Standard dilutions ranging from 1 to 200 mg 1-l AT III were prepared from this using fresh culture medium as dilutant. The same standard was used for the reference assay, for which dilutions covering a range of 10 to 100 mg l- 1 were prepared with pH 7.2 phosphate-buffered saline (Na2HPO 4, 1.18 g l-l; NaHePO4, 0.23 g 1-1; NaC1, 4.50 g 1-~; NAN3, 0.2 g 1-1; Tween 20, 0.5 ml 1-1). Purified pullulanase (50,000 U I-i) was provided by Dr. G. Antranikian, Georg-August-Universit~,t, G6ttingen. Dilutions ranging from 50 to 1,000 U I-i were prepared from this stock using culture medium as diluent.
Source o f proteins Cell-free bioreactor supernatant (containing AT III in concentrations up to 40 mg 1-1) and fresh culture me-
970
TM
Sampling
Reference assay A noncompetitive sandwich ELISA procedure was performed to obtain reference values for AT III. The assay was carried out in a microtiter plate precoated with anti-AT III antibody (Dr. Pelzer, Behring Werke). A horseradish peroxidase-labeled second antibody was employed (Anti-AT III-POD, Behring Werke). O-Phenylenediamine served as substrate for the enzyme. Enzymatic activity was detected using an ELISA reader (Nunc Immunoreader NJ 2000, at 490 rim, detection range: 10- I00 ng 1 ~). The method published by Bergmeyer t3 was used to obtain reference values for the pullulanase assay. Samples were incubated at 65°C with a pullulan-containing reagent solution. The amount of reducing sugars produced during this reaction was measured with dinitrosalicylic acid photometrically. This method utilizes measurements based on enzymatic activity. All off-line samples were analyzed at the end of the cultivation experiments.
Analysis system The FIA manifold is shown in Figure 1. Activation of the different pumps and valves of the system is achieved by a set of interactive time-switch relays. Separate injection ports (Latek rotary valves) are available for sample and reagent (diluted antibody). Teflon tubing (0.5 mm i.d.) was used throughout the system. Peristaltic pumps equipped with color-coded tubing were used to pump buffer (Watson-Marlow pump 502-S/50), sample (Watson-Marlow pump 1010), and reagent (Ismatec pump IPN-8). The loop of each injection port holds 50/xl. Sample was continuously pumped through the injection loop. Detection occurs photometrically (Skalar filter photometer) at 340 nm. This detection wavelength was
Enzyme Microb. Technol., 1991, vol. 13, December
On-line turbidimetric immunoassay: R. Freitag et al. stream and measuring the optical density at 340 nm. The dispersion in the two manifolds was not identical, but the computer compensated for this by multiplying the blank value with an appropriate factor. The corrected blank value was subtracted from the sample turbidity measured after the incubation period. The concentration of the protein to be monitored was then computed.
blank
37~C photometer
\
••••r••
Results and discussion
(37 °C)
(~
~eter
pump
buffer
magnetic air
valve
antibody
Figure 1 Diagram of the flow injection manifold for turbidimetric detection of immune complexes and sample blank. A merging-zones, stopped-flow system serves for antigen evaluation. The optical density inherent to the sample is determined in a second flow injection analysis manifold of conventional design. Instead of sample, a standard solution may be injected into both systems. Each reagent portion injected is preceded by an air bubble to prevent dilution by residual buffer
chosen because the intensity of light scatter increases with decreasing incident light wavelengths. At the same time, it was necessary to avoid the absorption wavelength of proteins (280 nm). The photometer was equipped with a 10-mm, flow-through microcuvette (volume 35 /~1). A similar FIA manifold was set up for the detection of the optical density intrinsic to the sample. In this case, five miniature magnetic valves were connected to form the injection port. A modified ATARI computer was used to collect and process the data. The program was activated after the absorbance reading of the photometer had passed a threshold. An A/D converter transferred the data to the computer. Typically, 1,000 readings were taken during each peak at 100-ms intervals. The program automatically compensated the signal if the baseline value found after the peak differed from that preceding it. A calibration curve is implemented into the program. Both the signal peak height and the area can be used for correlating turbidity and protein concentration. After loading, sample and reagent were injected into separate buffer streams. Merging of antibody and sample was accomplished using a T-connection piece, which allows the reactant plugs to meet head on. This ensured adequate mixing while keeping the overall dispersion at a minimum. After the reactants were combined, their transport towards the detector was stopped for 30-120 s by redirecting the buffer streams (incubation). During this time, the sample blank was established by injecting a sample segment into a third buffer
The bond between antigen and antibody depends on the synergistic interactions among a number of noncovalent types of binding. Buffer composition therefore strongly influences the kinetics of the reaction and the size of the immunocomplexes formed. It was established by optimization experiments that an increase in the ion content of the buffer lowers the speed of the antibody-antigen reaction, although there is no significant change in the final optical density reading. Sodium phosphate buffer used for both antibody and antigen dilution gives the fastest reaction. An increase in PEG concentration increases both reaction speed and end point value. The strongest influence was generally observed at small antibody excess. While the effect of cation species (sodium vs. potassium) was no longer observed for media-diluted samples, the influence of ion and PEG concentration was found for samples diluted with buffer and with medium alike (although the effect was modified in the latter). When a medium containing more than 4% FCS was used for sample dilution, unspecific protein precipitation was observed for buffers containing 60 g 1 i PEG. The accelerating influence of both decreased ion concentration and increased PEG concentration is well documented in the literature. 3'~4 We found a 0.01 M sodium phosphate buffer containing 4.5 g 1- J NaC1 and 40 g 1-1 PEG best for AT III.
Flow system parameters for the A T III assay During the incubation period, when most of the immune complexes form, the reaction mixture does not move within the flow system. Static experiments were therefore appropriate for kinetic measurements. Since heterogeneous samples show complex dispersion and mixing patterns during transport, further development of the turbidimetric immunoassay was carried out under flow conditions. After an optimum antibody dilution of 1 : 3 had been established, dose-response curves were recorded for AT III concentrations between 0 and 50 mg 1- ~(Figure 2). Varying concentrations of FCS and/or pH indicators such as phenyl red, often used in hybridoma cell cultivations in concentrations between 15 and 75 mg 1-1, had small effects on the signals and could be excluded by measuring the sample blank value. Small amounts of AT III (10-400/~1 -~) gave readings below the blank value. The optimal concentration range is above 1 mg 1- ~. Since dispersion increases with length of manifold
Enzyme Microb. Technol., 1991, vol. 13, December
971
Papers
average dispersion was 3.7 for this assay. The optimal assay conditions are listed for the flow system in Table 1. The costs for each assay were below the comparable ELISA costs for AT III monitoring in these concentration ranges.
ca o~
t o ,t~o
Simulation of cultivation conditions for the A T III assay
O
,o,
15 o o
o o
1i0
oO
Zt0
3L0 AT III
4i0
SO
[rag.l-l]
Figure 2 D o s e - r e s p o n s e curves for AT Ill in 0.01 M sodium phosphate buffer, containing 4.5 g 1-1 NaCI, as a sample dilutant. The buffer is used in both cases for antiserum dilution. The values were measured after 60 s incubation. Mean values of three different assays are given (standard deviation: 2.13%)• The lowest detectable antigen concentration in an undiluted sample is 1 mg 1-1 (see also Reference assay)
o
~
t~
Qo
F.
¥ ,o
oJ
7
14
Time [ h ]
~I
28
3 5°
Figure 3
Simulation of cultivation of AT Ill-producing cells. AT Ill is determined hourly by the turbidimetric assay (TIA) and reference ELiSA. Values given for the on-line assay are the mean value of three measurements, w h i l e the ELISA data are averages of t w o assay measurements
tubing and causes the detector signal to decrease, the stopped-flow principle was used to increase residence time with minimal dispersion. During time-response studies in the flow system, a flow pause of 80 s proved to be optimal. Under these circumstances, AT III concentrations above 1 mg l f gave a distinct reading from noise. For longer incubation times, increasing levels of protein precipitation within the tubing were observed. This resulted in reduced reproducibility and clogging became a problem. Repeated measurements of a sample containing 50 mg l-z AT lII gave a standard deviation of 2.13%, while repeated injection of a dye solution showed that the
972
Development and optimization of any assay are simplified if performed during a simulation instead of a real bioprocess. With a simulated process, product concentrations can be changed from zero to the final value within hours instead of weeks. However, a simulation cannot completely replace experience gained during an actual cultivation for a number of reasons (e.g. the effect of sampling devices on the whole analysis procedure is not tested; the influence of compounds produced and consumed during a cultivation on the accuracy of the analysis is hard to determine). The following system was used to simulate the changes in product concentration and sample matrix that would be expected during a bioprocess. A homemade, sterilizable gradient mixer was connected via a sampling device to the FIA system. Linear gradients of cell-free bioprocess supernatants replacing fresh culture medium were mixed over a period of 12 to 16 h. In some cases, several gradients were run in succession over a period of 36 to 48 h. On-line protein determinations were performed in triplicate every hour. During the incubation period before the second reading, offline samples were withdrawn and stored at 4°C for protein evaluation by a reference assay. The sample matrix changes significantly during a cultivation. In the case of AT III production, an increase in optical density of 30% was observed for media containing 4% FCS. Since it had been established by the experiments described above that a difference in protein content resulted mainly in a shift of the calibration curve, it was necessary to monitor the optical density intrinsic to the sample. Optimization was undertaken by repeatedly using the FIA system for on-line monitoring with several gradients in succession. The data obtained are depicted in Figure 3. AT III concentrations in corresponding
Table 1 Parameters for on-line AT Ill m e a s u r e m e n t during simulation of cultivation using AT Ill-producing cells Flow rates Sample: 0.6 ml rain -1 Buffer: I ml rain -~ Reagent: 0.1 ml min I Assay cycle (a) Reagent p u m p on for 48 s (b) Sample injection after 76 s (c) Reagent p u m p off and incubation for 78 s (d) Blank injection after 104 s Time for one assay cycle: 154 s
Enzyme Microb. Technol., 1991, vol. 13, December
On-line turbid/metric immunoassay:
off-line samples that were determined with the ELISA are included. While the correlation at lower AT III concentrations was excellent, higher deviations were observed at higher concentrations because of dilution problems with the ELISA. A correlation coefficient of 0.988 was found. Recalibration of the system with the 50 mg 1- ~standard after the simulation gave a deviation of 2%. During the experiment, a slight baseline drift was observed. Data collecting and processing by the computer were not affected, since the computer program automatically took the absorbance value prior to the peak as the new baseline value.
The pullulanase to be detected was produced by thermophilic bacteria. Several bioprocess parameters, such as temperature (60°C) and pH (4-6), differed from the normal range of immunoassay conditions, where a temperature of 37°C or less and a pH range of 7.0-7.6 is usually found. ~ The effect of temperature and pH on the precipitation reaction was investigated in a series of kinetic measurements. For 10 pullulanase concentrations between 0 and 4,000 U l- ~, the influence of temperatures between 4°C and 70°C and pH values between 4.0 and 9.0 was explored with an antibody dilution of I : 40; the indicated pullulanase concentrations cover the entire antibody excess as well as the equivalence range. Temperature influence was investigated using buffers of pH 7.4, while pH influence was investigated at 37°C. Best results were obtained at a pH of at least 7, which has to be maintained in the reaction mixture. To ensure this, the phosphate concentration of the buffer was raised to 0.05 M. The temperature dependence of the reaction was not as marked. No influence could be observed below 40°C. Above 40°C, the data obtained were highly irregular and not reproducible. The cultiva-
0
o 0
I 240
I
480
I
7ZO
I
960
1200
P u l l u l a n a s e [ U . I "1 ]
Figure 4 Calibration curve for the turbidimetric pullulanase assay. The curve was obtained from 16 standard dilutions ranging from 0 to 1,000 U 1-1
o
o
E ltO
¢t~
¢-
0
"5
"-1
n 13.
I'N c)
•
"r"--
I
0
50
1
O
I
1
t00
150
200
250
Fermentation time [h] Figure 6 Mean cultivation
pullulanase
concentration
measured
on-line
(TIA) and off-line
every 2 h during the first 20 h (batch culture). Lower activity data were obtained during 60-150 h, because the off-line samples were analyzed with a long time delay (longer incubation times) during this period. After a batch growth a continuous culture was started at 20 h. The dilution rate was increased from 0.075 to 0.1 h -] at 132 h. At 196 h the feed was changed from a starch-containing medium to a glucose-containing medium, thus decreasing the pullulanase concentration. When comparing on-line and off-line data for pullulanase production (Figure 6), one must consider that the protein acts as an antigen in one assay whereas its enzymatic activity is measured in the other. These two functions do not necessarily correspond identically in all concentration ranges. However, since the average correlation factor was 0.975 during simulation and 0.974 during the cultivation, this difference can be assumed to be small. The detection limit of 1 mg 1- t (corresponding to 100 U 1-1) was not a problem, since this concentration was surpassed after only 10 h. It was possible to estimate the product concentration range to be expected because the production rates of the bacteria were well known beforehand. The second disadvantage of precip974
(reference
assay) during
a C. t h e r m o s u l f u r o g e n e s
itation immunoassays, the fact that two antigen concentrations might cause identical turbidities, could thus be eliminated by selecting an appropriate antibody concentration.
Conclusions Automatic on-line monitoring of proteins in the complex medium of cultivation processes is possible with the turbidimetric immunoassay system described in this paper. It can be considered an interesting alternative to the more labor-intensive manual monitoring of proteins. An automated system for the continuous monitoring of viruses in infection bioreactors is under development. During batch culture, the manual method has the advantages of requiring less sample volume and possessing a lower detection limit, since even low enzymatic activities can be detected by extending reaction time. On the other hand, automatic control of production by changes in media composition or exchange rates is not possible with the time-consuming off-line analysis. The on-line immunoassay should therefore be most useful during continuous culture, especially for production monitoring during stationary phases.
Enzyme Microb. Technoi., 1991, vol. 13, December
On-line turbidimetric immunoassay: R. Freitag et al.
Acknowledgements
2
We wish to thank Dr. U. Eberhard, Behring Werke Marburg, for donating fresh culture media as well as bioreactor supernatants containing AT III. Dr. G. Antranikian, Institut ft~r Mikrobiologie, Georg-August Universit~t G6ttingen supplied puUulanase in various forms. A. Spreinat, Institut for Mikrobiologie, GeorgAugust Universit~it G0ttingen not only donated antipullulanase antibodies but also made part of this research possible by carrying out the C. t h e r m o s u l f u r o g e n e s cultivation. This work was supported by a grant from the Deutsche Gesellschaft for Chemisches Apparatewesen ( D E C H E M A ) to R. Freitag and the B M F T within the " S c h w e r p u n k t : Grundlagen der BioprozeBtechnik."
3 4 5 6 7 8 9 10 11 12 13
References 1
Heidelberger, N. and Kendall, F. E. J. Exp. Med. 1935, 51, 563-591
14
Larson, C., Orenstein, P. and Ritchie, R. F. Advances in Automated Analysis. Technicon International Congress, 1970, 1-9 Mtiller-Matthesius,R. and Opper, C. J. Clin. Chem. Clin. Biochem, 1980, 18, 501-510 Ebeling,H. J, J. Clin. Chem. Clin. Biochem. 1978, 16, 191-195 Wojciechowicz,D., De Falco, M. and Muller-Eberhard, U. J. J. lmmunol. Meth. 1988, 106, 57-61 Ruzicka,J. and Hansen, E. H. Anal. Chim. Acta 1980, 114, 19-44 Janata, J. and Harrow, J. J. Anal. Chim. Acta 1986, 180, 14-16 Krug,F. J., Bergamin, F. H., Zagatto, E. A. G. and J6rgensen, S. S. Analyst 1977, 102, 503-508 Worsfold,P. J. and Hughes, A. Analyst 1984, 109, 339-341 Worsfold,P. J., Hughes, A, and Mowthorpe, D. J. Analyst 1985, 110, 1303-1305 Freitag, R., Scheper, T., Spreinat, A. and Antranikian, G. Appl. Microbiol. Biotechnol. 1991, 35, 471-476 Antranikian, G., Zablowski, P. and Gottschalk, G. Appl. Microbiol. Biotechnol. 1987, 27, 75-83 Bergmeyer,H. U. Methoden der enzymatischen Analyse 3rd ed. Verlag Chemie, Weinheim, 1974 Buffone,G. J., Savory, J. and Hermans, F. Clin. Chem. 1975, 21/12, 1735-1746
Enzyme Microb. Technol., 1991, vol. 13, December
975
An on-line assay f o r a thermostable pullulanase and antithrombin III (AT III) is described. The assay is based on the formation o f aggregates between the protein to be measured and antibodies raised against this protein. Assay automation was achieved by utilizing the flow injection analysis (FIA) principles. The apparatus, a stopped-flow, merging-zone manifold, is described in detail. Since the reaction used in an FIA system does not have to reach equilibrium, it was possible to reduce the time f o r an assay cycle to 2.5 rain. A method f o r simulating cultivation conditions was developed f o r assay optimization. Using this method, a detection limit o f l mg 1- i together with a standard deviation o f l.5 was found. A sandwich ELISA was used as reference assay in the case o f A T III and an enzymatic activity assay in the case o f pullulanase. Correlation coefficients o f 0.988 (AT III) and 0.976 (pullulanase) were determined. The turbidimetric assay was successfully used f o r pullulanase monitoring during a 240-h cultivation o f Clostridium thermosulfurogenes.
Keywords:On-line immunoassay; turbidimetric assay; bioprocess monitoring;antithrombin III assay; pullulanase assay
Introduction The discriminatory power of antibodies makes an immunoassay the best candidate for monitoring substances of high molecular weight in bioreactor supernatants. The fact that antibodies can be raised against virtually all proteins as well as many other biomolecules contributes to the great attractiveness of this analytical tool for clinical and biotechnological purposes. Immunoassays used in biotechnology often depend on a label (such as an enzyme or a fluorescent dye) linked to one of the reactants for detection. The distribution of the labelled substance between two phases or two types of immunocomplexes is then evaluated. The most commonly used method to assay for highmolecular-weight proteins is ELISA (enzyme-linked immunosorbent assay). ELISAs tend to be time consuming, since many dilution, washing, and incubation steps are included, especially when no routine analysis assays are used. Often samples are drawn throughout a cultivation and analyzed afterwards. Cultivation con-
Address reprintrequeststo Dr. Scheperat the Institutfor Technische Chemie, Universit~tHannover,Callinstrasse 3, D-3000Hannover1, Germany Received 15 February 1991;accepted 20 May 1991 © 1991 Butterworth-Heinemann
trol based on production rates is thus impossible. The low detection range inherent in these assays is of no advantage if substances found in mg 1-1 concentrations or higher are concerned. This will often be the case for biotechnological products. Immunoprecipitation has been used since 1935 for measuring protein concentrations. 1 The method is based on the formation of large aggregates between bivalent antibodies and multivalent antigens. The complexes formed are large enough to cause light scattering of the Mie type.l This method has become the most widely used in clinical analysis when the specific quantification of proteins above the mg 1-1 level is desired. The fact that the method is easily automated has also increased its popularity. An automated immunoprecipitation analyzer of the continuous flow type was developed as early as 1970. 2 Turbidimetric detection, in which the decrease in transmission is recorded, can also be f o u n d ) Most applications utilize a rate measurement (kinetic method), although end point measurements are sometimes t a k e n Y The analytic potential of flow injection analysis (FIA) has been established over the last decade. 6 Although the theoretical treatment of the behavior of heterogeneous samples in an FIA system is difficult,7 turbidimetric detection has been used successfully in such systems. 8 Due to the high reproducibility of sample
Enzyme Microb. Technol., 1991, vol. 13, December
969
Papers dispersion in FIA manifolds, the reaction that is used does not necessarily need to reach equilibrium. This is an important aspect when automating relatively slow reactions such as antibody-antigen reactions. Surprisingly, there are not many applications to be found in the literature. Among the few are a system using naturally occurring binding proteins such as conc a n a v a l i n A 9 and an assay in which IgG concentrations are measured off-line in buffer solutions. 1° To our knowledge, a turbidimetric assay has never been used on-line for process control. This report describes a fully automated on-line assay that uses turbidimetric detection as well as its successful employment for product monitoring during a cultivation. Conditions for assaying two proteins, human antithrombin III (h-AT III) and a thermostable pullulanase, are given.
dium were supplied by Dr. Eberhard, Behring Werke. The medium contained between 2.5% and 4% fetal calf serum (FCS). Purified h-AT III (Kybernin HS I000, Behring Werke Marburg) was used during preliminary experiments. The starch-degrading enzyme pullulanase was produced by thermophilic bacteria (CIostridium thermosulfurogenes EM 1, DSM 3896). I1 Cultivation was carried out at 60°C and the pH was regulated at 6 during continuous culture. A mineral medium described by Antranikian et al. lz was used. Process mixture broths containing approximately 1,000 U 1-1 pullulanase were supplied by A. Spreinat, Georg-August-Universitfit, G6ttingen.
Materials and methods
During the cultivation simulations, a sterile trap (break tube) was used as a sampling device (for both on-line and off-line sampling). During the C. thermosulfurogenes cultivation, a cross-flow module (KF-200-010-SME, Microgon Inc.) was employed for on-line sampling.
Chemicals Chemicals were from Fluka unless indicated otherwise. Deionized distilled water was used to prepare solutions. PEG 6000 and BSA (bovine serum albumin) were from Sigma Chemical Co. A sodium phosphate buffer was used for the on-line assay (AT III assay: Na2HPO4, 1.18 g 1-1; NaH2PO 4, 0.23 g 1-1; NaCI, 4.50 g 1-1; PEG 6000, 40.00 g 1-1. Pullulanase assay: NazHPO4, 5.90 g 1-1; NaHzPO 4, 1.15g 1-1; NaC1, 4.5g i-1; PEG 6000, 40.00 g 1-1). The pH was adjusted to 7.4 with 1 M HCI. The same buffer without PEG 6000 was used for antibody dilution. Buffers were degassed before use.
Biologicals Antibodies against h-AT III were purchased from Behring Werke West Germany [OSAY 08/09; (rb) titer: 1.35 g ! -1 (RID)]. Antibodies were diluted threefold before use. Antibodies (rb) against pullulanase produced by C. thermosulfurogenes were kindly donated by A. Spreinat and G. Antranikian, Institut for Mikrobiologic, Georg-August-Universitat G6ttingen, Germany. The antiserum was diluted 1 : 40 prior to use. A standard solution of 1 g 1-1 h-AT III in buffer was provided by Dr. Eberhard, Behring Werke. Standard dilutions ranging from 1 to 200 mg 1-l AT III were prepared from this using fresh culture medium as dilutant. The same standard was used for the reference assay, for which dilutions covering a range of 10 to 100 mg l- 1 were prepared with pH 7.2 phosphate-buffered saline (Na2HPO 4, 1.18 g l-l; NaHePO4, 0.23 g 1-1; NaC1, 4.50 g 1-~; NAN3, 0.2 g 1-1; Tween 20, 0.5 ml 1-1). Purified pullulanase (50,000 U I-i) was provided by Dr. G. Antranikian, Georg-August-Universit~,t, G6ttingen. Dilutions ranging from 50 to 1,000 U I-i were prepared from this stock using culture medium as diluent.
Source o f proteins Cell-free bioreactor supernatant (containing AT III in concentrations up to 40 mg 1-1) and fresh culture me-
970
TM
Sampling
Reference assay A noncompetitive sandwich ELISA procedure was performed to obtain reference values for AT III. The assay was carried out in a microtiter plate precoated with anti-AT III antibody (Dr. Pelzer, Behring Werke). A horseradish peroxidase-labeled second antibody was employed (Anti-AT III-POD, Behring Werke). O-Phenylenediamine served as substrate for the enzyme. Enzymatic activity was detected using an ELISA reader (Nunc Immunoreader NJ 2000, at 490 rim, detection range: 10- I00 ng 1 ~). The method published by Bergmeyer t3 was used to obtain reference values for the pullulanase assay. Samples were incubated at 65°C with a pullulan-containing reagent solution. The amount of reducing sugars produced during this reaction was measured with dinitrosalicylic acid photometrically. This method utilizes measurements based on enzymatic activity. All off-line samples were analyzed at the end of the cultivation experiments.
Analysis system The FIA manifold is shown in Figure 1. Activation of the different pumps and valves of the system is achieved by a set of interactive time-switch relays. Separate injection ports (Latek rotary valves) are available for sample and reagent (diluted antibody). Teflon tubing (0.5 mm i.d.) was used throughout the system. Peristaltic pumps equipped with color-coded tubing were used to pump buffer (Watson-Marlow pump 502-S/50), sample (Watson-Marlow pump 1010), and reagent (Ismatec pump IPN-8). The loop of each injection port holds 50/xl. Sample was continuously pumped through the injection loop. Detection occurs photometrically (Skalar filter photometer) at 340 nm. This detection wavelength was
Enzyme Microb. Technol., 1991, vol. 13, December
On-line turbidimetric immunoassay: R. Freitag et al. stream and measuring the optical density at 340 nm. The dispersion in the two manifolds was not identical, but the computer compensated for this by multiplying the blank value with an appropriate factor. The corrected blank value was subtracted from the sample turbidity measured after the incubation period. The concentration of the protein to be monitored was then computed.
blank
37~C photometer
\
••••r••
Results and discussion
(37 °C)
(~
~eter
pump
buffer
magnetic air
valve
antibody
Figure 1 Diagram of the flow injection manifold for turbidimetric detection of immune complexes and sample blank. A merging-zones, stopped-flow system serves for antigen evaluation. The optical density inherent to the sample is determined in a second flow injection analysis manifold of conventional design. Instead of sample, a standard solution may be injected into both systems. Each reagent portion injected is preceded by an air bubble to prevent dilution by residual buffer
chosen because the intensity of light scatter increases with decreasing incident light wavelengths. At the same time, it was necessary to avoid the absorption wavelength of proteins (280 nm). The photometer was equipped with a 10-mm, flow-through microcuvette (volume 35 /~1). A similar FIA manifold was set up for the detection of the optical density intrinsic to the sample. In this case, five miniature magnetic valves were connected to form the injection port. A modified ATARI computer was used to collect and process the data. The program was activated after the absorbance reading of the photometer had passed a threshold. An A/D converter transferred the data to the computer. Typically, 1,000 readings were taken during each peak at 100-ms intervals. The program automatically compensated the signal if the baseline value found after the peak differed from that preceding it. A calibration curve is implemented into the program. Both the signal peak height and the area can be used for correlating turbidity and protein concentration. After loading, sample and reagent were injected into separate buffer streams. Merging of antibody and sample was accomplished using a T-connection piece, which allows the reactant plugs to meet head on. This ensured adequate mixing while keeping the overall dispersion at a minimum. After the reactants were combined, their transport towards the detector was stopped for 30-120 s by redirecting the buffer streams (incubation). During this time, the sample blank was established by injecting a sample segment into a third buffer
The bond between antigen and antibody depends on the synergistic interactions among a number of noncovalent types of binding. Buffer composition therefore strongly influences the kinetics of the reaction and the size of the immunocomplexes formed. It was established by optimization experiments that an increase in the ion content of the buffer lowers the speed of the antibody-antigen reaction, although there is no significant change in the final optical density reading. Sodium phosphate buffer used for both antibody and antigen dilution gives the fastest reaction. An increase in PEG concentration increases both reaction speed and end point value. The strongest influence was generally observed at small antibody excess. While the effect of cation species (sodium vs. potassium) was no longer observed for media-diluted samples, the influence of ion and PEG concentration was found for samples diluted with buffer and with medium alike (although the effect was modified in the latter). When a medium containing more than 4% FCS was used for sample dilution, unspecific protein precipitation was observed for buffers containing 60 g 1 i PEG. The accelerating influence of both decreased ion concentration and increased PEG concentration is well documented in the literature. 3'~4 We found a 0.01 M sodium phosphate buffer containing 4.5 g 1- J NaC1 and 40 g 1-1 PEG best for AT III.
Flow system parameters for the A T III assay During the incubation period, when most of the immune complexes form, the reaction mixture does not move within the flow system. Static experiments were therefore appropriate for kinetic measurements. Since heterogeneous samples show complex dispersion and mixing patterns during transport, further development of the turbidimetric immunoassay was carried out under flow conditions. After an optimum antibody dilution of 1 : 3 had been established, dose-response curves were recorded for AT III concentrations between 0 and 50 mg 1- ~(Figure 2). Varying concentrations of FCS and/or pH indicators such as phenyl red, often used in hybridoma cell cultivations in concentrations between 15 and 75 mg 1-1, had small effects on the signals and could be excluded by measuring the sample blank value. Small amounts of AT III (10-400/~1 -~) gave readings below the blank value. The optimal concentration range is above 1 mg 1- ~. Since dispersion increases with length of manifold
Enzyme Microb. Technol., 1991, vol. 13, December
971
Papers
average dispersion was 3.7 for this assay. The optimal assay conditions are listed for the flow system in Table 1. The costs for each assay were below the comparable ELISA costs for AT III monitoring in these concentration ranges.
ca o~
t o ,t~o
Simulation of cultivation conditions for the A T III assay
O
,o,
15 o o
o o
1i0
oO
Zt0
3L0 AT III
4i0
SO
[rag.l-l]
Figure 2 D o s e - r e s p o n s e curves for AT Ill in 0.01 M sodium phosphate buffer, containing 4.5 g 1-1 NaCI, as a sample dilutant. The buffer is used in both cases for antiserum dilution. The values were measured after 60 s incubation. Mean values of three different assays are given (standard deviation: 2.13%)• The lowest detectable antigen concentration in an undiluted sample is 1 mg 1-1 (see also Reference assay)
o
~
t~
Qo
F.
¥ ,o
oJ
7
14
Time [ h ]
~I
28
3 5°
Figure 3
Simulation of cultivation of AT Ill-producing cells. AT Ill is determined hourly by the turbidimetric assay (TIA) and reference ELiSA. Values given for the on-line assay are the mean value of three measurements, w h i l e the ELISA data are averages of t w o assay measurements
tubing and causes the detector signal to decrease, the stopped-flow principle was used to increase residence time with minimal dispersion. During time-response studies in the flow system, a flow pause of 80 s proved to be optimal. Under these circumstances, AT III concentrations above 1 mg l f gave a distinct reading from noise. For longer incubation times, increasing levels of protein precipitation within the tubing were observed. This resulted in reduced reproducibility and clogging became a problem. Repeated measurements of a sample containing 50 mg l-z AT lII gave a standard deviation of 2.13%, while repeated injection of a dye solution showed that the
972
Development and optimization of any assay are simplified if performed during a simulation instead of a real bioprocess. With a simulated process, product concentrations can be changed from zero to the final value within hours instead of weeks. However, a simulation cannot completely replace experience gained during an actual cultivation for a number of reasons (e.g. the effect of sampling devices on the whole analysis procedure is not tested; the influence of compounds produced and consumed during a cultivation on the accuracy of the analysis is hard to determine). The following system was used to simulate the changes in product concentration and sample matrix that would be expected during a bioprocess. A homemade, sterilizable gradient mixer was connected via a sampling device to the FIA system. Linear gradients of cell-free bioprocess supernatants replacing fresh culture medium were mixed over a period of 12 to 16 h. In some cases, several gradients were run in succession over a period of 36 to 48 h. On-line protein determinations were performed in triplicate every hour. During the incubation period before the second reading, offline samples were withdrawn and stored at 4°C for protein evaluation by a reference assay. The sample matrix changes significantly during a cultivation. In the case of AT III production, an increase in optical density of 30% was observed for media containing 4% FCS. Since it had been established by the experiments described above that a difference in protein content resulted mainly in a shift of the calibration curve, it was necessary to monitor the optical density intrinsic to the sample. Optimization was undertaken by repeatedly using the FIA system for on-line monitoring with several gradients in succession. The data obtained are depicted in Figure 3. AT III concentrations in corresponding
Table 1 Parameters for on-line AT Ill m e a s u r e m e n t during simulation of cultivation using AT Ill-producing cells Flow rates Sample: 0.6 ml rain -1 Buffer: I ml rain -~ Reagent: 0.1 ml min I Assay cycle (a) Reagent p u m p on for 48 s (b) Sample injection after 76 s (c) Reagent p u m p off and incubation for 78 s (d) Blank injection after 104 s Time for one assay cycle: 154 s
Enzyme Microb. Technol., 1991, vol. 13, December
On-line turbid/metric immunoassay:
off-line samples that were determined with the ELISA are included. While the correlation at lower AT III concentrations was excellent, higher deviations were observed at higher concentrations because of dilution problems with the ELISA. A correlation coefficient of 0.988 was found. Recalibration of the system with the 50 mg 1- ~standard after the simulation gave a deviation of 2%. During the experiment, a slight baseline drift was observed. Data collecting and processing by the computer were not affected, since the computer program automatically took the absorbance value prior to the peak as the new baseline value.
The pullulanase to be detected was produced by thermophilic bacteria. Several bioprocess parameters, such as temperature (60°C) and pH (4-6), differed from the normal range of immunoassay conditions, where a temperature of 37°C or less and a pH range of 7.0-7.6 is usually found. ~ The effect of temperature and pH on the precipitation reaction was investigated in a series of kinetic measurements. For 10 pullulanase concentrations between 0 and 4,000 U l- ~, the influence of temperatures between 4°C and 70°C and pH values between 4.0 and 9.0 was explored with an antibody dilution of I : 40; the indicated pullulanase concentrations cover the entire antibody excess as well as the equivalence range. Temperature influence was investigated using buffers of pH 7.4, while pH influence was investigated at 37°C. Best results were obtained at a pH of at least 7, which has to be maintained in the reaction mixture. To ensure this, the phosphate concentration of the buffer was raised to 0.05 M. The temperature dependence of the reaction was not as marked. No influence could be observed below 40°C. Above 40°C, the data obtained were highly irregular and not reproducible. The cultiva-
0
o 0
I 240
I
480
I
7ZO
I
960
1200
P u l l u l a n a s e [ U . I "1 ]
Figure 4 Calibration curve for the turbidimetric pullulanase assay. The curve was obtained from 16 standard dilutions ranging from 0 to 1,000 U 1-1
o
o
E ltO
¢t~
¢-
0
"5
"-1
n 13.
I'N c)
•
"r"--
I
0
50
1
O
I
1
t00
150
200
250
Fermentation time [h] Figure 6 Mean cultivation
pullulanase
concentration
measured
on-line
(TIA) and off-line
every 2 h during the first 20 h (batch culture). Lower activity data were obtained during 60-150 h, because the off-line samples were analyzed with a long time delay (longer incubation times) during this period. After a batch growth a continuous culture was started at 20 h. The dilution rate was increased from 0.075 to 0.1 h -] at 132 h. At 196 h the feed was changed from a starch-containing medium to a glucose-containing medium, thus decreasing the pullulanase concentration. When comparing on-line and off-line data for pullulanase production (Figure 6), one must consider that the protein acts as an antigen in one assay whereas its enzymatic activity is measured in the other. These two functions do not necessarily correspond identically in all concentration ranges. However, since the average correlation factor was 0.975 during simulation and 0.974 during the cultivation, this difference can be assumed to be small. The detection limit of 1 mg 1- t (corresponding to 100 U 1-1) was not a problem, since this concentration was surpassed after only 10 h. It was possible to estimate the product concentration range to be expected because the production rates of the bacteria were well known beforehand. The second disadvantage of precip974
(reference
assay) during
a C. t h e r m o s u l f u r o g e n e s
itation immunoassays, the fact that two antigen concentrations might cause identical turbidities, could thus be eliminated by selecting an appropriate antibody concentration.
Conclusions Automatic on-line monitoring of proteins in the complex medium of cultivation processes is possible with the turbidimetric immunoassay system described in this paper. It can be considered an interesting alternative to the more labor-intensive manual monitoring of proteins. An automated system for the continuous monitoring of viruses in infection bioreactors is under development. During batch culture, the manual method has the advantages of requiring less sample volume and possessing a lower detection limit, since even low enzymatic activities can be detected by extending reaction time. On the other hand, automatic control of production by changes in media composition or exchange rates is not possible with the time-consuming off-line analysis. The on-line immunoassay should therefore be most useful during continuous culture, especially for production monitoring during stationary phases.
Enzyme Microb. Technoi., 1991, vol. 13, December
On-line turbidimetric immunoassay: R. Freitag et al.
Acknowledgements
2
We wish to thank Dr. U. Eberhard, Behring Werke Marburg, for donating fresh culture media as well as bioreactor supernatants containing AT III. Dr. G. Antranikian, Institut ft~r Mikrobiologie, Georg-August Universit~t G6ttingen supplied puUulanase in various forms. A. Spreinat, Institut for Mikrobiologie, GeorgAugust Universit~it G0ttingen not only donated antipullulanase antibodies but also made part of this research possible by carrying out the C. t h e r m o s u l f u r o g e n e s cultivation. This work was supported by a grant from the Deutsche Gesellschaft for Chemisches Apparatewesen ( D E C H E M A ) to R. Freitag and the B M F T within the " S c h w e r p u n k t : Grundlagen der BioprozeBtechnik."
3 4 5 6 7 8 9 10 11 12 13
References 1
Heidelberger, N. and Kendall, F. E. J. Exp. Med. 1935, 51, 563-591
14
Larson, C., Orenstein, P. and Ritchie, R. F. Advances in Automated Analysis. Technicon International Congress, 1970, 1-9 Mtiller-Matthesius,R. and Opper, C. J. Clin. Chem. Clin. Biochem, 1980, 18, 501-510 Ebeling,H. J, J. Clin. Chem. Clin. Biochem. 1978, 16, 191-195 Wojciechowicz,D., De Falco, M. and Muller-Eberhard, U. J. J. lmmunol. Meth. 1988, 106, 57-61 Ruzicka,J. and Hansen, E. H. Anal. Chim. Acta 1980, 114, 19-44 Janata, J. and Harrow, J. J. Anal. Chim. Acta 1986, 180, 14-16 Krug,F. J., Bergamin, F. H., Zagatto, E. A. G. and J6rgensen, S. S. Analyst 1977, 102, 503-508 Worsfold,P. J. and Hughes, A. Analyst 1984, 109, 339-341 Worsfold,P. J., Hughes, A, and Mowthorpe, D. J. Analyst 1985, 110, 1303-1305 Freitag, R., Scheper, T., Spreinat, A. and Antranikian, G. Appl. Microbiol. Biotechnol. 1991, 35, 471-476 Antranikian, G., Zablowski, P. and Gottschalk, G. Appl. Microbiol. Biotechnol. 1987, 27, 75-83 Bergmeyer,H. U. Methoden der enzymatischen Analyse 3rd ed. Verlag Chemie, Weinheim, 1974 Buffone,G. J., Savory, J. and Hermans, F. Clin. Chem. 1975, 21/12, 1735-1746
Enzyme Microb. Technol., 1991, vol. 13, December
975
