Beruflich Dokumente
Kultur Dokumente
4, 1983 113
The major impact of instability is on 4 Shepard, H. M., Yelverton, E. and and W. Goebel, eds), pp. 19-29,
the yield and efficiency of the fermenta- Goeddel, D. V. (1982) DNA 1, Springer Verlag
tion process. However for pharma- 125-131 14 Jones, I. W., Primrose, S. B. and
ceuticals which have to be manufactured 5 Gelfand, D. H., Shepard, M., Ehrlich, S. D. (1982)Mol. Gen. Genet
under GMP (Good Manufacturing O'Farrell, P. N. and Polisky, B. (1978) 188, 486-489
Practice) conditions the drug regu- Proc. Natl Acad. Sci. 75, 5869-5873 15 Rood, J. I., Sneddon, M. K. and
6 Sninsky, J. J., Uhlin, B. E., Gustafsson, Morrison, J. F. (1980)J. Bacterio1144,
latory authorities consider instability to P. and Cohen, S. N, (1981) Gene 16, 552-559
be undesirable. Where the recombinant 275-286 16 Tomizawa, J. and Itoh, T. (1981)Mol.
protein is for industrial use, e.g. 7 Uhlin, B. E. and Nordstrom, K. (1977) Gen. Genet. 178, 525-533
enzymes for inclusion in detergents Plasmid 1, 1-7 17 Cesarini, G., Muesing, M. A. and
etc., errors in the amino acid sequence 8 Gentz, R., Langer, A., Chang, A. C. Y., Polisky, B. (1982)Proc. NatlAcad. Sci.
introduced by mistranslation are of Cohen, S. N. and Bujard, H. (1981) 79, 6313-6317
little importance unless they signifi- Proc. Natl Acad. Sci. 78, 4936-4940 18 Tomizawa, J. and Itoh, T. (1982) Cell
cantly decrease the specific activity. 9 Swamy, K. I-I. S. and Goldberg, A. L. 31, 575-583
Again, however, they are of great sig- (1982)J. Bacterio1149, 1027-1033 19 Stueber, D. and Bujard, H. (1982)
nificance for drugs intended for human 10 Simon, L. D., Randolph, B., Irwin, N. EMBOffournal 1, 1399-1404
and Binkowski, G. (1983) Proc. Natl 20 Edelmann, P. and Gallant, J. (1977)
use.
Acad. Sci. 80, 2059-2062 Cell 10, 131-137
11 Jones, I. M., Primrose, S. B., Robinson, 21 Grosjean, H. and Fiers, W. (1982) Gene
References A. and Ellwood, D. C. (1980)214ol. Gen. 18, 199-209
1 Hawley, D. K. and McClure, W. R. Genet. 180, 579-584 22 Chavancy, G., Daillie, J. and Garel,
(1983) Nucl. Acid. Res. 11, 2237-2255 12 Kreft, J. and Hughes, C. (1982) in Gene J. P. (1971)Biochimie 53, 1187-1197
2 de Boer, H. A., Comstock, L. J. and Cloning in Organisms other than E. coli 23 Nath, K. and Koch, A. L. (1971)
Vasser, M. (1983)Proc. NatlAcad. Sci. (P. H. Hofschneider and W. Goebel, ft. Biol. Chem. 246, 6956-6967
80, 21-25 eds), pp. 1-17, Springer Verlag 24 Itakura, K., Hirose, T., Crea,
3 Nugent, M. E., Primrose, S. B. and 13 Ehrlich, S. D., Niaudet, B. and Michel, R., Riggs, A. D., Heyneker, H. L.,
Tacon, W. C. (1983) Develop. Ind. B. (1982) in Gene Cloning in Organisms Bolivar, F. and Boyer, H. W. (1977)
MicrobioL 24 (in press) other than E. coil (P. H. Hofschneider Science 198, 1056-1063
be limited by inadequate mass transfer 10 -2 mol m -~ s -1. For small vessels, up OUR (mot m-3s-I)
even at intermediate growth rates and to about 1 m 3, the vessel wall provides Ex (rno[ m-3)
cell densities. Cell volume will not enough surface area for cooling. For
generally be limiting. The cell volume, larger vessels cooling coils or outside 100 _
derived from Fig. 1 for realistic OUR heat exchangers are needed. There is
values, does not exceed a value of 50% normally no engineering difficulty in
by volume, even at /~ - 0 where it providing the required surface of 1 m 2 10-1 _ 5000
reaches its maximum. It must be m -3 and thus considerations of heat
remembered that, in the case of fungal transfer will not restrict the design of
OUR= 2.8 x 10-2
growth, the viscosity is strongly related fermenters. i0 -2_
to cell density. This can lead to high Another restricting parameter in
viscosities2 and related problems in large scale fermenters might be mixing.
mixing and mass transfer, even for cell If the substrate is injected at only one 10-3_ so
densities corresponding to the region position into a continuous culture or
below OUR -- 2.8 x 10 -2 given in fed batch, then a volume element has to
Fig. 1. traverse the whole circulation loop I0 -~_
The rate of heat production H P R (W before it returns to the injection point.
m -3) is straightforwardly coupled to This circulation time can easily be 0 iC)-s 5x~O-S i~-~
OUR (Ref. 1): 10-100 s in large fermenters 5. The l.t (s -I)
concentration of substrate in the
H P R -- 4.7 x 105OUR (4) Fig. 1. Example of OUR values calculated
volume element will decrease during
The heat transfer rate H T R (W m -3) this circulation. A characteristic time by means of Eqn 3.
from the broth to the wall or cooling for substrate depletion, t~a, can be For a substrate consumption rate equal
coil can be calculated from: derived as follows. Assume a volume to that of the original one at t -- 0 with
HTR = a x 6T x A (5) element leaves the site where the Cs --- C,, the time needed to deplete the
substrate is added at t = 0. After a time substrate can be derived from Eqns 6
where: t, the decrease, ACa, of the substrate and 7, with AC, -- Cs introduced in Eqn
= heat transfer coefficient ( W m -2 concentration Cs, is given by 6:
K-') rat = ACs V (6) Y~C,+K~
A T = temperature difference (K) tcs - (8)
V = reactor volume (m 3) /amax Cx
A = cooling surface area (m2 m -3) When t < < tea then ACa < < Ca. At very
When maintenance is neglected, the
From Refs 3 and 4 the order of application of the Monod equation in low growth rates whereby Cs<< Ks, a
magnitude of a can be estimated as Eqn 2 leads to: minimum value of tcs is achieved, given
103Wm-2K-L Assuming AT = 10 K, by:
ra=l ~ CxV=~ Ca CxV(7 )
the surface area required for adequate ta _ YaxKa
cooling can be calculated to be of the Y,~ Ka+Ca Y,a /~m~x (9)
order of 1 m 2 m -3 for OUR = 2.8 x Ks = Monod constant (mol m -3) (, << ~ )
In many cases the time derived from
Symbols and their units Eqn 9 may be of the order of 1-100 s.
Symbol Definition Units Symbol Definition Units This means that, for large scale
A surface area m 2 m -3 OTR oxygen transfer rate mol m -3 fermenters with a single substrate
CoL,CoG concentration of oxygen s-~
P stirrer power
injection point, the organisms will
in the liquid (CoL) or gas
(CoG) phase mol m -3 consumption W experience severe variations in the con-
C~)L equilibrium COL value at ro oxygen consumption centration ofsubstrate in their environ-
the gas-liquid interface mol m 3 rate mol s- i ment. There is not much published
COL S concentration of oxygen rs CH20 consumption rate mol s information about the influence of fluc-
in the liquid in the stirrer rx biomass production rate mot s-j
region mol m 3 Sh Sherwood number no units
tuations in substrate concentration on
C biomass concentration mol m -3 tcO critical time for oxygen
the growth and metabolism of
/~L diffusion coefficient in the consumption s organisms. However, some articles6'7
liquid phase m 2 s -I tcs critical time for substrate show that there can be an effect.
db bubble diameter m consumption s Fluctuations in concentration,
g gravitational acceleration m s-2 vb bubble rise velocity m s-I
HPR heat production ratio W m -3
however, are inherently linked to large
vL liquid velocity m s-I
HTR heat transfer rate W m-3 Ysx yield of biomass on
scale operations and cannot be
kL liquid side mass transfer substrate mol eliminated completely by better
coefficient ms i mol -~ mixing. A simple method of decreasing
m partition coefficient of o~ heat transfer coefficient W m -2 the variations is to use multiple
oxygen between gas and K-I
liquid no units
injection points and well-defined
AT temperature difference K
ms maintenance coefficient mot gas hold up no units
circulation loops7.
mol -I I~ growth rate s- I Another restriction caused by poor
s-I blL liquid viscosity Pa s mixing is local oxygen depletion. This
OUR oxygen uptake rate mol m -3 60 density difference can happen when the transfer of oxygen
s-I between liquid and gas kg m -3
from the gaseous into the liquid phase
Trends in Biotechnology, Vol. 1, No. 4, 1983 115
effective mass transfer requires a con- bacteria) is beyond the scope of this
~< siderable power consumption and mass article*. In conclusion it can be said that
transfer determines part of the variable the kL and A values of the gas-liquid
manufacturing cost. In the next sec- interface in a fermentation broth play
o tions the relevant mass transfer data the dominant role in oxygen transfer.
g~ COL will be briefly reviewed and the costs The kL value is determined mainly by
worked out. The transfer of oxygen the character of the liquid interface
from air bubbles to the liquid phase will around the bubble. It can be 'rigid' or
be used as a model system. 'moving'. With a rigid interface there
exists a stagnant layer of liquid around
G a s - l i q u i d m a s s transfer the bubble. Oxygen is transferred by
The formula for the transfer of molecular diffusion along a concentra-
oxygen from the gaseous to the liquid tion gradient that can be assumed to be
A
phase is (see Fig. 2): linear, as shown schematically in Fig. 2.
organism
bubble
diffusion OTR = kLA (C~L - COL ) (11) The thickness of the diffusion layer is of
the order of 10 -~ m. Outside this layer
layer where: there exists well-mixed bulk liquid.
Fig. 2. Schematic representation of oxygen
COL, CoG = concentration of oxygen in The kL value will always be larger than
mass transfer from a bubble to an organism.
the liquid (L) or gas (G) phase (tool m -3) that for a bubble in completely
takes place in only one section of the C~)L ---- CoG/m = equilibrium C O L value stationary surroundings, given by the
vessel (for instance, the stirrer region) at the gas-liquid interface (molm -3) Sherwood number, Sh:
while long circulation loops transport m = partition coefficient of oxygen
dissolved oxygen to other sections of between gas and liquid (no units) Sh - kLdb _ 2 (12)
the fermenter. A critical oxygen deple- kL = liquid side mass transfer coeffi- DL
tion time too can be derived as: cient (ms -1) where
A = area of the gas-liquid interface (m2 db = bubble diameter (m)
tcO- COLS (10) m -3)
OUR D L = diffusion coefficient in the liquid
This formula assumes that the limita- phase (m2 s -1)
where: tion to mass transfer is located in the
COLS = oxygen concentration in the liquid layer around the gas bubble, kLA Experimentally determined kL values
stirrer region (mol m -3) is a reciprocal of the resistance to mass for bubbles with a rigid film and a well-
t r a n s f e r ; C~) L - C O L is the driving force
mixed bulk liquid are (Ref. 8):
Assuming CoLs to be the value that is
for mass transfer. The resistance in the Sh - kLdb _ 2 + 0 31(db~33APg~'3 (13)
in equilibrium with air at atmospheric
gas phase can be neglected due to the
conditions, then &o is of the order of
large value o f m ( "-~35 for oxygen). The
10 s. For large fermenters circulation where:
resistance in the liquid layer around
times can be of the order of 10-100 s, so
the organism can be neglected since the A# = density difference between liquid
local zones with oxygen depletion may
organism is relatively small and thus and gas (kg m -3)
be present. This phenomenon can be
only an extremely small driving force is g = gravitational acceleration (m s -2)
reduced by improving the mixing
necessary to transfer O2 from the bulk ~.1L = liquid viscosity (Pa s)
characteristics, increasing the static
liquid into the organism. Mass transfer
pressure in the stirrer region, using
to closely packed organisms (e.g. t Seep. 120forreviewon mouldsgrownin pellet
oxygen-enriched air and using multiple
moulds in pellet form or immobilized form.
impeller systems.
The order of magnitude calculations
presented so far show that heat transfer '0
and cell volume should not be regarded
as limiting factors in fermenters. The
distribution of substrate and oxygen E
may be problematic, but some relative- J
ly simple measures can solve these 2':/+-
problems. It should be noted that all
these calculations have been based on ~.:.:.:.:.:.:.: ............. eq. 1/-,
an OTR of 2.8 x 10 -2 mol m -3 s -1, ..............
..............
assuming that to be a rather high value. _ .,......
..........
..., ... i::i::iiiii::iii::iilregion wifh mectsu red
Fig. 1 shows, however, that at high
growth rates and moderate cell density eq. 13 values
values much higher values of OTR are k~eq.12
possible. If indeed the OTR is deter-
mined by the capacity of the fermenter, , ~ , 1~)-2
then this is the limiting parameter and
hence it largely determines the invest-
d b m)
ment. Later it will alsobe shown that Fig. 3. k L values dependent of bubble diameter.
116 Trends in Biotechnology, VoL 1, No. 4, 1983
A bubble with a moving interface rises until they become large enough to be
through the liquid without any liquid 'w 10 o- dispersed by turbulent and shear
being entrained in the film layer. Mass IE forces. An equilibrium bubble size is
transfer is then described by penetra- established, resulting from continuous
tion theory9. This describes non- coalescence and dispersion processes.
stationary mass transfer into an infinite This equilibrium size is about 6 mm
medium which initially has an oxygen for tap water. This means that within
concentration equal to the bulk concen- 0.1-1 m above sparger the bubble size
tration. The liquid is exposed to the gas is independent of the bubble size near
bubble for a very short time, equal to sparger or stirrer.
that of the displacement of the gas In a non-coalescing liquid, the
10-z , . , .
bubble over its diameter. It can theoret- bubbles remain at the size initially pro-
10-4 10-3 10-2
icaUy be derived that for such a system: duced at the sparger or stirrer and thus
db(m) it is possible for much smaller bubbles
kL = /4_ JgL%~5 (14) Fig. 4. Bubble rise velocity. to exist in the bulk liquid than in a
shown in Fig. 4. It will be clear from the coalescing liquid (Table 1). A typical
where: kL, A, and vb formulas that the bubble non-coalescing liquid is a strong salt
% = bubble velocity diameter plays a dominant role in solution; a typical coalescing liquid is
gas-liquid mass transfer. This is shown distilled water. The addition of salt to a
The relationships 12, 13 and 14 can also in Fig. 5 for an idealized system. vigorously stirred vessel with distilled
be verified by the vast amount of Unfortunately it is just this parameter water decreases the bubble size
measured data available from the for which quantitative information is drastically"; in this case the very small
literature4; these are presented in Fig. 3. lacking. bubbles produced by the stirrer are
Although the scatter is rather large, it The bubble size is determined by two maintained in the bulk liquid.
can be seen that the diameter has a separate processes: formation at the
major influence on the ,/eL value. Very sparger or stirrer and coalescence/ Oxygen transfer in large scale
small bubbles do have a rigid interface. dispersion phenomena in the bulk fermenters
Eqn 13 describes the data; Eqn 12 liquid. Air entering the reactor is Commercial fermentation liquids
predicts much lower values in dispersed by the stirrer or the sparger. have very complex compositions.
accordance with the difference between The resulting bubble size can range Besides a third phase, the ceils, there is
absolutely static bubbles in a stagnant from very small (0.1-1 mm) up to large considerable amounts of medium, sub-
liquid and the rigid diffusion layer of (> 6 mm)4,u, depending on process strates, products and cell debris. These
limited thickness on which Eqn 12 is conditions. The collision frequency of all influence coalescence characteristics
based. Very large bubbles have a fully the bubbles will be high in the bulk of to some - mainly unknown - extent,
moving interface and Eqn 14 predicts the reactor, even at low gas flow rates making it impossible to predict the
the values rather well. For the region and hold up values. The result of the bubble size in the fermenter. This
between small and large bubbles the collision is determined by the character uncertainty is increased by the presence
data points are rather scattered. In this of the liquid. Two types of liquid can be of chemicals often added to control the
region the bubble surface can be either distinguished, coalescing and non- foaminess of fermentation liquids. Even
rigid or mobile, depending on the coalescing. In a coalescing liquid each traces of these chemicals can affect
presence of surface active materials in collision leads to the formation of a coalescence and kLA considerably~z.
the liquid. In general for large bubbles: single bubble. In this way bubbles grow Although coalescence behaviour and,
kL ~ 4 X 10-4 ms -1 and for small
bubbles k L --, 10-4 ms 1. A
with a product separation that is rather scales, rather accurate data can be 6 Leegwater, M. P. M., Neijessel, O. M.
insensitive to broth density. This case is obtained by combining general and Tempest, D. W. ( 1981) SecondEur.
comparable to example B in Fig. 7. formulae for model liquids with a Congr. of Biotech. 5-10 April 1981,
Examples might be single cell protein limited number of measurements on Eastbourne England. Abstracts of
(SCP) and waste water treatment. For the specific fermentation broth. Large Comm. p. 22
7 Schiigerl, K. (1983) Chem. Ing. Teehn.
fermentations requiring expensive scale mass transfer may even be pre-
55, 123
fermenters with a recovery that is very dicted quite accurately in this way; 8 Calderbank, P. H. and Moo-Young, M.
sensitive to product concentration, a however, a number of complications B. (1961) Chem. Eng. Sci. 16, 39
highly loaded stirred reactor is to be due to Co~, CoL and kLA differences in 9 Higbie, R. (1935) Trans. A.LCh.E. 31,
preferred. Even then P/Vvalues > > 10 the vessel can be expected. These can 365
kW/m 3 will rarely occur in practice. be partly overcome by modelling gas 10 Wallis, G. B. (1969) One-dimensional
The very high fermenter costs could be and liquid mixing. two-phase flow. McGraw Hill Book
caused by the necessity of expensive Oxygen transfer determines a signi- Company
construction materials, of (volume ficant part of manufacturing costs 11 Lee, J. C. and Meyrick, D. L. (1970)
dependent) safety measures or of high and, to a certain extent, even the type Trans. Inst. Chem. Engrs. 48, T37
pressure. of fermenter that will be chosen. In 12 Adler, I., Buchholz, H., Voigt, J.,
Wittier, R. and Schfigerl, K. (1980)
many cases, a non-stirred low-loaded
Eur. J. Appl. Microbiol. Biotechnol. 9,
fermenter is to be preferred. 249
Conclusion 13 Van 't Riet, K. (1979) Ind. Eng. Chem.
In large commercial fermenters a References Proc. Des. Dev. 18, 357-364
number of restricting parameters are 1 Heynen, J. J. and Roels, J. A. (1981) 14 Oosterhuis, N. M. G., Groesbeek, N.
manifest. Calculation of critical times Biotechnol. and Bioeng. 23, 739 M., Olivier, A. P. C. and Kossen, N.
shows that substrate addition and 2 Roels, J. A., Van den Berg, J. and W. F. (1983 ) Biotech. Letters 5, 141-146
oxygen mass transfer are usually the Voncken, R. M. (1974)Biotechnol. and 15 Deckwer, W. D., Hallensleben, J. and
main restrictive factors. To prevent Bioeng. 16, 181 Popovic, M. (1980) Can. ,7. Chem. Eng.
variations in concentrations substrates 3 Henzler, H. J. (1982) Chem. Ing. Techn. 58, 190
must be injected at several positions. 54, 461 16 Merchuk, J. C. and Stein, Y. (1981)
4 Heynen, J. J. and Van 't Riet, K. (1982) Biotechnol. Bioeng. 23, 1309
When predicting OTR for fermenta- Proc. BHRA Fourth Eur. Conf. on 17 Finn, R. K. (1967)Biochem. and Biol.
tion liquids it can be argued from Mixing April 27-29, 1982, Noord- Eng. Sci. (N. Blakebrough, ed.)
fundamental reasoning that simple, wijkerhout, The Netherlands p. 195 Academic Press p. 69
accurate correlations do not exist for Paper F1. 18 Quicker, G., Schumpe, A., KSnig, B.
gas-liquid mass transfer in fermenta- 5 Einsele, A. (1978) Pract. Biochem. July, and Deckwer, W. D. (1981)Biotechnol.
tion broths. For small and intermediate 13 Bioeng. 23, 635