542
Fermentation Cootrol with Biosensors in Flow-injection Systems: Problems and Progress
U. PRINZING, I. OGBOMO, C. LEHN and H.-L. SCHMIDT
Lehrstuhl fir Allgemeine Chemie und Biochemie, Technbche Universitcir Uiinchen, D-8050 Freising- Weihenstephan (F.R.G.)
Abstract
Efforts to develop automatic and sterilizable
samplers for the liquid phase are briefly presented.
Special problems associated with the use of
biosensors in on-line fermentation control are
demonstrated and possibilities to overcome them
are discussed. The assay system presented for the
determination of volatile substances includes,
prior to enzymatic analysis, a continuous purifica-
tion by pervaporation. Examples are given for the
monitoring of ethanol in a yeast extractlpeptone
solution and diacetyl in buffer solutions.
Introduction
Flow injection analysis (FIA) based on immo-
bilized dehydrogenases is a very versatile method
for process control in biotechnology. Many of the
highly specific oxidoreductases are commercially
available, and the determination of the common
product of their substrate conversion, NADH, is
easily possible, either electrochemically or by
fluorimetry and luminometry. Predictable correla-
tions exist between the concentration of the sub-
strate to be determined and the NADH signal,
thus providing a basis for an automated calibra-
tion and response control [ 11.
With immobilized enzymes in FIA systems, a
continuous substrate and/or product monitoring
in fermentation broths is only possible when
macromolecular impurities and other inhibiting
substances are eliminated. All pretreatments in use
are based on membrane separation processes,
which imply two problems: ‘concentration polar-
ization’ and ‘fouling of the membrane’. The
purpose of ultrafiltration is to retain the macro-
molecules of the solution. If the convective flow of
these macromolecules is faster than their diffusion
back to the bulk solution, an accumulation near
the membrane surface occurs. Fouling of the
membrane due to its clogging by macromolecular
substances may be influenced by membrane and/
or solute properties and processing parameters.
Concentration polarization and fouling of the
Sensorsand Actuators, Bl (1990) 542-545
membrane lead to a drastic flux decline as well as
to a time dependence of flux and a different
rejection behaviour by filtration processes.
Additional constraints of sampling devices
based on membrane separation processes are
the long response time, the presence of dead
volumes, large sampling volumes and the problem
of maintaining aseptic conditions in the culture
medium. These problems caused the develop-
ment of sophisticated filter systems in the fer-
mentor and by-pass loops. To avoid fouling of
the membranes, among many other techniques,
there are also efforts to change the properties
of the membrane surface by sulfonation of
polystyrene or polysulfone, modification by en-
zymes, carbon coating and surfactant pretreat-
ment [2-61.
Depending on the fermentation at hand, prob-
lem-orientated pretreatments such as dialysis and
electrodialysis units [ 71 or pervaporation modules
should be integrated to guarantee long stability of
the biological component of the sensor as well as
optimal reaction conditions for the immobilized
enzymes. In the following contribution, examples
will be given for the determination of diacetyl and
ethanol in flow systems containing a pervapora-
tion module.
Experimental
Enzymes and Chemicals
Alcohol dehydrogenase (EC 1.1.1.1, from
yeast) was from Boehringer, Mannheim/F.R.G.
Diacetyl reductase (EC 1.1.1.5, from Lactobacillus
kejir) was a gift of Professor M.-R. Kula, Institute
of Enzyme Technology, University of Diisseldorf.
The dialysis membrane ( 12 pm cellulose mem-
brane) was a gift of Gambro Dialysatoren GmbH,
Hechingen/F.R.G. The pervaporation membrane
was a PTFE membrane filter (Sartorius, Gottin-
gen/F.R.G.; pore size 0.45 pm) with an effective
membrane area of 0.17 cm2. The support for en-
zyme immobilization was controlled porous glass
(CPG, 120-200 mesh, pore size 14OOA; Fluka,
Neu-Ulm/F.R.G.). All other chemicals (analytical
grade) originated from E. Merck, Darmstadt/
F.R.G. or Serva, Heidelberg/F.R.G.

Enzyme Immobilization
Diacetyl reductase was covalently bound to
porous glass using a coupling method described
by Weetall [8]. The porous beads were alkylated
with aminopropyltriethoxysilane and the product
was activated with glutardialdehyde. The acti-
vated carrier ( 150 mg) was incubated at 4 “C with
1 ml diacetyl reductase solution (40 U) in Tris/
HCl buffer (pH 7) containing 100 mM diacetyl.
Alcohol dehydrogenase ( 1200 U) was incubated
with 150 mg of the same activated carrier in 1 ml Fig. 1. Schematic of the pervaporation module.
0.1 M phosphate buffer, pH 7. The enzyme reac-
tors used in the flow systems were small glass
tubes (i.d. 4 mm, length 20 mm) filled with these netted with fouling of the membrane and also
immobilized oxidoreductases. protects the biological component of the sensor.
Figure 2 shows the flow diagram of the experi-
Instruments mental set-up used for pervaporation experiments.
The equipment for pervaporation experiments The gas-liquid separation system can be de-
(Fig. 2) comprised a peristaltic pump (Ismatec scribed as follows. The volatile component present
mv-ge, Zurich/Switzerland), a thermostated water- in the donor liquid stream diffuses through a
bath (Thermomix 1420, Braun Melsungen/ PTFE membrane with an effective membrane
F.R.G.), a homemade pervaporation module, a area of 0.17 cm2 into an acceptor buffer stream.
fluorimeter (Hitachi FlOOO, E. Merck, Darm- The latter transports the sample to the detection
stadt/F.R.G.) with a 60 ~1 flow-through cell and a system.
flow-injection system (FIAstar 5020 annlyser,
Tecator, HiiganHs/Sweden). FIA for the Detection of Ethanol and Diacetyl
in Complex Media
Ethanol and diacetyl are volatile compounds
Results and Discussion occurring during the production of beer. Ethanol
is the main component aimed at, diacetyl is an
Pervaporation Module unwanted byproduct, namely the most common
A very good method to withdraw volatile com- off-flavour in beer. In order to attain an ethanol
pounds continuously from complex media, like level in the final beer of about 4.5 g/l, the fermen-
fermentation broths, is prevaporation. It is a pro- tation is slowed down at an ethanol concentration
cess in which volatile substances in a heated donor of 3.5 g/l by cooling the fermentation tub from 10
stream evaporate through a porous hydrophobic
membrane. The vapour condenses on the surface
of a cool acceptor stream on the other side of the
membrane. Surface tension forces withhold the
liquids from the pores and prevent direct contact
between them. The temperature difference, caus-
ing a corresponding vapour pressure difference
across the membrane, provides the driving force
for the membrane process. Evaporation will occur
at the solution surface if the vapour pressure is
greater than the vapour pressure at the condensate
surface.
Volatile components of the fermentor broth,
like ethanol or diacetyl, are able to diffuse
through such a hydrophobic membrane driven by
a temperature difference. The advantage of our
homemade pervaporation module (Fig. 1) is given Fig. 2. Flow system for the determination of diacetyl and
ethanol. PI = peristaltic pump (Ismatec mv-ge); P2 =
by an additional air gap between the process fluid peristaltic pump and Vi = injection valve (included in FIAstar
and the membrane, thus avoiding any contact 5020); W = thermostated waterbath; D = detector (Hitachi
between both. This diminishes the problems con- FlOOO);M = pervaporation module.
544
to 4 “C. The diacetyl concentration, on the other
hand, can rise up to 0.4 mg/l during the first stage
of the fermentation but has to be below 0.1 mg/l
after the second period (storage). Therefore an
on-line control of the ethanol as well as of the
diacetyl concentration would be very promising.
Diacetyl reductase catalyzes the reaction
H,C-CO-CO-CH, + NADH/H+
Z$ H,C-CHOH-CO-CH, + NAD+
In the presence of diacetyl, NADH/H+ is con-
sumed, thus the decrease of the NADH/H+ peak
is a measure of the diacetyl concentration. The
gas-liquid separation system for the determina-
tion of diacetyl can be described as follows.
Acetate buffer (0.1 M, pH 5.4) was pumped as
an acceptor stream and as a donor stream to
establish a baseline. Then pulses of 50 ~1 of a
130 pmol NADH/H+ solution were injected into
the acceptor stream to produce a constant
NADH/H + peak measured by fluorimetry
(E,,, = 465 nm; E, = 340 nm). After reaching con-
stant NADH/H+ peaks, a diacetyl standard as a
donor stream was pumped through the lower
chamber of the pervaporation module. In relation
to the diacetyl concentration, a proportional part
pervaporates into the acceptor stream. The de-
crease of NADH/H+ consumed in the enzymatic
reaction was measured.
The calibration graph (Fig. 3) was linear in the
range from 0.7 to 0.8 mM diacetyl and was paral-
lel to that of diacetyl without the pervaporation
module. Since the diacetyl level during the pro-
duction of beer lies between 0.001 and 0.006 mM,
higher enzyme activities and more efficient immo-
bilization techniques have to be tested to meet the
concentration range of diacetyl in beer. Investiga-
tions are in progress.
[mvl
1000,
I ration module for 54 h.
100
/ q
a-
,
0.01 1
Dlace~~l
imM1
Fig. 3. Calibration curves for diacetyl with ( q ) and without
(*) pervaporation module by means of the assay system illus-
trated in Fig. 2.
I
100
0.01 1
mM &not
Fig. 4. Calibration curves for ethanol with (*) and without
( + ) pervaporation module determined with the assay system
illustrated in Fig. 2.
For ethanol measurements, phosphate buffer
(0.1 M, pH 9) was pumped as an acceptor stream
and a solution of yeast extract (3.5 g/l), peptone
( 1.Og/l) and ethanol (0.05 M) as a donor stream.
The temperature of the waterbath was 70 “C.
After 15 min, 50 ,~l portions of 10mM NAD+
were injected into the acceptor stream. In contrast
to the diacetyl determination where the con-
sumption of NADH/H+ was a measure of the
diacetyl concentration, here the NADH/H+
formed after the enzyme catalyzed reaction was
measured. Standard calibration graphs with and
without pervaporation module show a parallel
course (Fig. 4). This proves that in relation to a
given ethanol concentration, a proportional part
evaporates through the membrane into the accep-
tor stream.
With this arrangement it was possible to mea-
sure the ethanol concentration in a complex
medium for days without fouling of the mem-
brane or loss of enzymatic activity. Figure 5
shows the values after injection of NAD+ every
30 min. The donor stream was pumped continu-
ously through the lower chamber of the pervapo-
I I
0
0 lo au
ti: [hl
40 a0 a0
Fig. 5. Measurement of ethanol in complex medium (yeast
extract 5 g/peptone 3 g/50 ml ethanol in 3 1 distilled water).

Conclusions
The present work emphasizes again the impor-
tance of the integration of sample pretreatments
and separation procedures into the flow systems. It
also demonstrates the capability and versatility of
a new system, pervaporation, for special problems.
The advantage of our homemade pervaporation
module is given by an air gap between the process
fluid and the membrane, thus avoiding any contact
between the analyte and the semipermeable mem-
brane, hence excluding fouling. At present, univer-
sal automatic and sterilizable samplers for the liquid
phase of fermentors are not commercially available.
This is due to problems associated with clogging of
the membranes by microbial cells and other solids.
Further work is in hand to overcome these
problems and to develop flow systems for multiple
substrate/product monitoring [ 7J In addition, mi-
croprocessor techniques will be applied to allow the
simultaneous monitoring of NADH/H+ consump-
tion and NADH/H+ formation with the same flow
injection system and thus giving the possibility of
determining diacetyl and ethanol as well as other
dehydrogenase substrates with one system.
Acknowledgements
The authors would like to thank Mr U. Prinzing
Sr. for construction of the pervaporation module.
545
The authors acknowledge financial support from
the Bundesministerium fiir Forschung und Tech-
nologie (BMFT).
References
1 H. Huck, A. Schelter-Graf and H.-L. Schmidt, Measure-
ment and calculation of the calibration graphs for flow
injection analysis using enzyme reactors with immobilized
dehydrogenases and an amperometric NADH-detector,
Bioelectrochem. Bioenerg., 13 (1984) 199-209.
2 D. A. Noordegraaf, C. A. Smolders, R. De Boer and D. J.
Romijn, Preparation and properties of a composite charged
membrane, Desalination, 41 (1982) 249-261.
3 A. G. Fane, C. J. D. Fell and K.-J. Kim, The effect of
surfactant pretreatment on the ultrafiltration of proteins,
Desalination, 53 (1985) 37-56.
4 H. Bauser, H. Chmiel and N. Stroh and E. Walitza, Inter-
facial effects with microfiltration membranes, J. Membr.
Sci., 11 (1982) 321-332.
5 H. Bauser and H. Chmiel, Improvement of the biocompati-
bility of polymers through surface modification, Pofym. Sci.
Technol., 23 (1983) 297-309.
6 H. Reihanian, C. R. Robertson and A. S. Michaels, Mecha-
nisms of polarization and fouling of ultrafiltration
membranes by proteins, J. Membr. Sci., 16 (1983) 237-
258.
7 R. Kittsteiner-Eberle, I. Ogbomo and H.-L. Schmidt,
Biosensing devices for the semi-automated control of dehy-
drogenase substrates in fermentations, Biosensors, 4 (1989)
75-85.
8 H. H. Weetall, Covalent coupling methods for inorganic
support materials, in K. Mosbach-(ed.), Method in &zy-
mology, Vol. 44, Academic Press, New York, 1976, pp.
134-148.