vi: éfl"\\\\ s

     
     

‘ “\fl’” ,
...1. :u.‘

 

W:
25¢ per day per item

RETURNIM LIBRARY MATERIALS:

Place in book return to remove
charge from circulation records

 

 

DIALYSIS CONTINUOUS PROCESSES FOR
MICROBIAL FERMENTATIONS:
MATHEMATICAL MODELS, COMPUTER

SIMULATIONS, AND EXPERIMENTAL TESTS

BY

Robert William Stieber

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Microbiology and Public Health

1979

ABSTRACT

DIALYSIS CONTINUOUS PROCESSES FOR MICROBIAL FERMENTATIONS:
MATHEMATICAL MODELS, COMPUTER SIMULATIONS, AND EXPERIMENTAL TESTS

BY

Robert William Stieber

Dialysis continuous culture processes were mathematically
modeled, computer simulated and experimentally tested, and the
literature was reviewed, to assess the potential for application
of membrane technology to industrial fermentations. The ammonium-
lactate fermentation of whey was used as a model system for the
experiments because of its commercial potential. The simulated
and experimental results both showed ways that dialysis continuous
processes can maximize substrate utilization and produce high
cell populations and cell-free metabolites at high rates. The
dialysis processes were a considerable improvement over conventional
continuous processes.

In one system for dialysis continuous fermentation, whey of
high concentration was fed into a fermentor while the fermentor
contents were dialyzed through a membrane against water to relieve
the inhibitory effect of metabolites. The fermentation was
modeled and a solution for the steady state was used to simulate
the fermentation on a digital computer. The results showed the
effects of various parameters on the process and predicted that
it could be operated efficiently. To test the predictions, dried
sweet-cheese whey was rehydrated to contain 230 mg of lactose per

ml, charged into a S-liter fermentor, adjusted in pH (5.3) and

Robert William Stieber

temperature (44°C), and inoculated with Lactobacillus bulgaricus.

 

The fermentor and dialysate circuits were connected, and a series
of steady—state conditions was managed without sterilization or
asepsis for 94 days. As time progressed, the fermentation remained
homofermentative and increased in conversion efficiency, although
membrane fouling necessitated dialyzer cleaning about every 4
weeks. With a retention time of 19 h, 97% of the substrate was
converted into products and cells. Relative to nondialysis
continuous or batch processes for the fermentation, the dialysis
continuous process enabled the use of more concentrated substrate,
was more efficient in the rate of substrate conversion, and
additionally produced a second effluent of less concentrated but
purer ammonium lactate.

The model was improved by incorporating separate terms for
substrate limitation and product inhibition into the equation
describing the rate of cell growth. The improved model was
validated using the results from previous experiments with whole
whey and also in subsequent experiments with deproteinized whey.
The latter showed that the cell accumulation, cell productivity,
and molar growth yield were greater than those obtained from a
conventional continuous process for the fermentation and that the
dialysis process could produce a continuous effluent containing
50 mg of cell-free ammonium lactate per ml. The whole whey
proved to be a better substrate than the deproteinized whey.

In a second system for dialysis continuous fermentation with
a prefermentor, simulations predicted that the addition would not

bring improvement. In a third system with 100% recycling (feedback)

Robert William Stieber

of cells, simulations predicted that the retention time would be
reduced 9-fold and the conversion of substrate (deproteinized
whey) to product would be increased by 25%, accomplished almost
entirely by maintenance metabolism.

In a fourth system, the use of dialysis for substrate feed
as well as product removal also was modeled and simulated. The
simulations predicted that the feeding of substrate into the
fermentor circuit via the dialysate circuit has potential for
production and recovery of both cell mass and cell-free metabolite.
The experimental tests indicated that the fourth system would be
a useful method for immobilizing microbial cells to produce a
cell-free metabolite at a rate greater than the maximum specific
growth rate of the microbe, and additionally would serve to

filter-sterilize the substrate.

To my Mother and Dad
and Bill, Tom, John, Pat,

Mark, Dan, Tim, and Kimberly

ACKNOWLEDGMENTS

I wish to express my sincere gratitude to Dr. Philipp Gerhardt
for his encouragement, insistence on excellence, and the opportunity.
I would also like to thank my guidance committee members,

Dr. C. A. Reddy and Dr. Ralph N. Costilow, for their help; and
especially Dr. George A. Coulman for his expertise and guidance with
the mathematical modeling.

I also wish to express my appreciation to Marilyn Lautzenheiser
for her excellent secretarial services.

I also acknowledge the financial support from grant ENG 76-17260
of the National Science Foundation, the Department of Microbiology
and Public Health of Michigan State University, and the National

Institutes of Health training grant GM 01911-10.

iii

TABLE OF CONTENTS

LIST OF TABLES O O O O O O O 0 O O C O O O O 0 O O O O O O O 0

GENERAL INTRODUCTION. . . . . . . . . . . . . . . . . . . . .

GENEML LITEMTURE REVIEW 0 O O O O O O O O O C O O O O O O O

Dialysis culture of microbial and mammalian cells:
applications, 1969-79. . . . . . . . . . . . . . . . . .

MATERIALS AND METHODS . . . . . . . . . . . . . . . . . . . .

RESULTS AND DISCUSSIONS . . . . . . . . . . . . . . . . . . .

I.

II.

III.

VI.

VII.

GENERAL

SUMMARY

Dialysis continuous process for ammonium-lactate
fermentation of whey: mathematical model and
computer simulation. . . . . . . . . . . . . . . . .

Dialysis continuous process for ammonium-lactate
fermentation of whey: experimental tests. . . . . .

Dialysis continuous process for ammonium-lactate
fermentation: improved mathematical model and
use of deproteinized whey. . . . . . . . . . . . . .

Production of lactobacillus cells by dialysis
continuous fermentation of deproteinized whey. . . .

Continuous process for ammonium-lactate fermentation
Of Whey O O O O O O O O O O O O O O O O O O O O O O O

Dialysis continuous process for ammonium-lactate
fermentation: simulated prefermentor and
cell-recycling systems . . . . . . . . . . . . . . .

Dialysis continuous process for ammonium—lactate
fermentation: simulated and experimental
dialysate-feed systems . . . . . . . . . . . . . . .

DISCUSSION 0 O O O O O O O O O O O O O O O O O O O O O

- ABSTMCT O O O O O I I O O O O O O O O O O O O O O 0

LITERATURE CITED. . . . . . . . . . . . . . . . . . . . . . .

iv

Page

.123

.150

.174
.203
.207

.210

TABLE

LIST OF TABLES

Page

Genera of microorganisms studied with dialysis
culture in Vitro. O O O O O O I O O O O O C O O O O O 12

Genera of microorganisms studied with dialysis
culture in vivo and ex vivo . . . . . . . . . . . . . l3

Mammalian cells and tissues studied with
dialysis culture in vitro . . . . . . . . . . . . . . 14

Mammalian cells and tissues studied with
dialysis culture in vivo and ex vivo. . . . . . . . . 15

Diffusible and nondiffusible products studied
with dialysis culture in vitro. . . . . . . . . . . . 16

Index of materials and methods. . . . . . . . . . . . 63

GENERAL INTRODUCTION

0f the various separation processes that involve membranes
(dialysis, microfiltration, ultrafiltration, and reverse osmosis),
dialysis is most easily applied to the production of microbial
cells and their metabolites. Microbial cultivation can be enhanced
by introducing a membrane dialyzer between a culture chamber and
a dialysate reservoir with the contents of each continuously
circulated on opposite sides of the membrane, enabling diffusional
exchange of nutrients and products while confining the cells.
The culture and dialysate circuits of such a system of dialysis
culture can be operated batchwise, continuously or in a combination
of these modes: and dialysis culture can be applied in a number
of ways to the production of cells or metabolites from a variety
of substrates. A schematic diagram of the generalized design of

a dialysis culture system is as follows:

 

 

 

 

 

 

 

 

 

 

 

 

L v ) L Y J
CULTURE, DIALYSATE
CIRCUIT CIRCUIT

In this study, the potential for application of dialysis
culture to continuous fermentation was investigated by using the
ammonium-lactate fermentation of whole or deproteinized whey as a
model process. The process involves the fermentative conversion
of the whey lactose to lactic acid and its neutralization by
ammonia. The fermentation can be managed without sterilization
or asepsis as a result of the restrictive conditions of low pH,
high concentration of undissociated acid, high temperature and
anaerobiosis. The biochemistry of the conversion is straightforward
and well described in kinetic terms. Thus, the fermentation is
representative of other homofermentations, can be easily managed
in the laboratory, and is amenable to mathematical analysis and
computer modeling.

The substrate whey is a residue of cheese manufacture and
has become a pollution problem as well as an economic and nutrient
loss, so that useful disposal of whey is important. Whole whey
can be processed by ultrafiltration to obtain a concentrate of
milk protein of high quality and considerable market potential,
but the remaining residue of deproteinized whey retains all of
the lactose and a substantial disposal problem still remains.

The rationale for an ammoniated organic acid fermentation is
applicable also to other carbohydrate-rich agricultural residues.
Ammonium salts of organic acids can be used directly as a nitrogenous
feed supplement for ruminant animals, or can be converted into

lactic acid for use as an industrial chemical and ammonium.sulfate
for use as a fertilizer. Lactobacillus cells can provide high-

quality single-cell protein for animal or human consumption.

Consequently, the fermentation is useful for its practical potential
as well as for fundamental study.

Dialysis can be used to improve a continuous fermentation by
means of two interrelated yet different system designs. In one,
the dialysate circuit is employed primarily to remove inhibitory
metabolites from the culture, resulting in increased extent and
rate of substrate conversion and permitting the use of a relatively
high concentration of substrate. As applied to the ammonium-
lactate fermentation, the substrate is fed into the fermentor
circuit, the contents of which are dialyzed through the dialyzer
membrane against water in the dialysate circuit. Thereby, the
small molecular product (lactic acid) is removed from the immediate
environment of the cells, thus relieving the inhibition by the
product that normally regulates its production. As more product
is withdrawn by dialysis, more substrate is consumed and more
product is made: i.e., the fermentation becomes more efficient.

The dialysis separation of product additionally represents a
purification and recovery step.

In a second system design, the fermentor contents are dialyzed
against substrate fed into the dialysate circuit and is employed
primarily to obtain greatly more concentrated cell populations in
the fermentor circuit. Such a design also can be used to increase
the throughput of substrate for conversion purposes. The second
design can be applied to fermentations that are not inhibited by
metabolite products as well as to those that are. In this design,
since the substrate enters the fermentor filtered through a

membrane, prior sterilization of the substrate is unnecessary.

Similarly, since metabolite products leave the fermentor filtered
through a membrane, the product from the dialysate circuit is
free of cells. This design can be operated for the continuous
production of a high concentration of cells and of a cell-free
metabolite, both at high rates.

The broad goal of this study was to assess the feasibility
of applying membrane technology to industrial fermentation processes.
Among the several membrane techniques, dialysis was selected for
study not only because of its easy application but also because
of the existence of a theoretical basis for its application to
bacterial culture (Schultz and Gerhardt, 1969). Continuous
rather than batch operation was chosen because a continuous
process generally is more productive than a batch process.
Practically, continuous operation appeared desirable not only for
best use of membrane technology but also because whey residue is
generated and the fermentation products could be used continuously.

The specific objectives of the study were to (1) develop
mathematical models of dialysis continuous processes for the
ammonium-lactate fermentation, in order to define the processes
and to predict by computer simulation the outcome of various
modes of operation, (2) experimentally test the best predicted
modes of operation by use of conventional laboratory equipment
and, (3) project the simulated and experimental results in order
to evaluate the potential for use of membranes in fermentations
industrially.

The investigative procedure was to first develop a mathematical

model of a particular dialysis continuous process design specifically

for the ammonium-lactate fermentation of whey. The fermentation
was modeled as a set of equations representing material balances
and rate relationships in the fermentor and dialysate circuits.
The model was generalized for application to other fermentations
by combining the equations and using dimensionless parameters.
The generalized model was then solved for the steady state and
used to simulate the ammonium-lactate fermentation on a digital
computer. The results predicted the effects of changes in
various material, operational and kinetic parameters. Laboratory
experiments then were conducted at critical points to test the
theoretical predictions. The simulated and experimental results
were correlated to demonstrate the validity of the model, establish
values for the bacterial metabolic constants, and evaluate the
process.

This thesis was prepared as a series of articles that have
been or are in process to be published separately in scientific
journals. The general literature review is intended to be part
of a review examining dialysis culture techniques used in studies
during the ten years since the 1969 review of the subject (Schultz
and Gerhardt, 1969). The materials-methods, results and discussions
are presented as a series of articles on specific aspects of the
study, each of which is organized in the usual journal format.

For convenience of the thesis reader, the materials and methods

in all of the articles are indexed in a composite table. In the
thesis section on results and discussions, the articles are
presented in entireties. Articles I and II describe the development

of a mathematical model, simulated predictions, and experimental

tests of a dialysis continuous process for the ammonium-lactate
fermentation of whole whey using the first system design. The

model was improved and tested and was extended in applicability

to the use of deproteinized whey (Article III) and the production

of lactobacillus cells (Article IV). A nondialysis continuous
process for the fermentation was modeled and experimentally

tested to further validate the model and to obtain cell mass data
for the dialysis studies (Article V). Dialysis continuous processes
with a prefermentor or a cell-recycle system were investigated
theoretically (Article VI). Dialysis substrate-feed and product-
removal systems were modeled and also experimentally tested

(Article VII). In the concluding thesis section, a general
discussion is provided about the long-range potential for application

of membrane technology to industrial fermentations.

GENERAL LITERATURE REVIEW

Dialysis Culture of Microbial and Mammalian
Applications, 1969-79

INTRODUCTION 0 O I O O O O O O O O I O O O O O C

MEDICAL APPLICATIONS O O O O I O O O O O O O O O

2.1 In vitro . . . . . . . . . . . . . . . . .

2.1.1 Microbial cells . . . . . . . . . .
2.1.2 Mammalian cells . . . . . . . . . .

2 O 2 In Vivo O O O O C O O O O O O O O O O O O 0

2.2.1 Microbial cells . . . . . . . . . .
2.2.2 Mammalian cells . . . . . . . . . .

2.2.2.1 Antibody-producing cells .

Cells:

2.2.2.2 Blood and bone marrow cells. . .

2 2 2 2 l Rat cells . . .
2 2 2 2 2 Mouse cells . .
2.2.2.2.3 Goat cells. . .
2 2 2 2 4 Rabbit cells. .
2 2 2 2 5 Human cells . .

2.2.2.3 Tumor cells. . . . . . . .
2.2.2.4 Other cell types . . . . .

ECOLOGICAL APPLICATIONS . . . . . . . . . . . .

3.1 Interbiosis. . . . . . . . . . . . . . . .
3 O 2 In Situ O O O O O O O O O O O O O O O O O O
3 O 3 In Vitro O O O O O O O O O O O O O O O O O

FERMENTATIVE APPLICATIONS . . . . . . . . . . .

4.1 Production of cells. . . . . . . . . . . .
4.2 Production of nondiffusible compounds. . .
4.3 Production of diffusible compounds . . . .
4.4 Dialysis sensors . . . . . . . . . . . . .

DISCUSS ION O O O O O O O O O O O O O 0 O O O O O

5.1 In vivo. . . . . . . . . . . . . . . . . .
2 In vitro . . . . . . . . . . . . . . . . .
3 Electrodialysis. . . . . . . . . . . . . .
.4 Microfiltration. . . . . . . . . . . . . .
5 Dialysis vs. Microfiltration . . . . . . .

Page

17
17

17
21

25

25
31

32
33

33
34
36
36
37

37
39

42

42
43
47

48

49
52
53
56

57

57
58
59
60
61

1 . INTRODUCTION

Dialysis is a process for separation of solute molecules by
means of their unequal diffusion through a semipermeable membrane
because of a concentration gradient. The process is applied to
the growth and maintenance of living cells by a technique called
dialysis or diffusion culture.

The principle of dialysis culture is fundamental to the
growth of microbial and mammalian cells in nature. Unlike the
usual batch or closed situation in the laboratory, cells free-
living in an ecological niche or parasitic on or part of a host
usually exist in a continuous or open situation in which the flux
of nutrients or metabolites is regulated by a membrane. Metabolically
active transport supplements passive diffusion in the membrane
regulation of living systems, but only the passive component
presently can be simulated in the laboratory cultivation of
cells.

There are other passive processes that involve membranes,
e.g., ultrafiltration, microfiltration, and reverse osmosis.
However, dialysis is the most easily applied to the production of
cells and their metabolites. To employ the technique of dialysis
culture, a membrane is introduced between a culture chamber and a
dialysate reservoir. For dialysis culture to be effective, the
volume of the dialysate must be large compared to that of the
culture, or else the dialysate must be replenishable. Further,
the permeability and area of the membrane must be sufficient to
permit useful diffusion, both in rate and quantity. Such a

system of dialysis culture can be operated in vitro or in vivo

and batchwise, continuously, or in a combination of these modes.

The system can be designed for nutrients to diffuse from the reservoir
to the culture and for metabolic products to diffuse from the

culture to the reservoir with the result that their respective
inhibitory effects are lessened.

Specific advantages and reasons for using dialysis for cell
culture include: the prolongation of cell growth, which allows the
attainment of high densities of viable cells; the extension of the
stationary phase of the growth cycle, thus permitting increased
production of metabolites associated with this phase; the relief of
product inhibition by removal or dilution of diffusible growth
products, thus causing greater production; the production of metabolites
free of cells and conversely the production of culture liquors free of
macromolecules from the medium: the means to study and recover a cell
population placed in an in situ or in vivo environment such as a
river or animal system: and the capability for study of interactions
between separated populations of cells.

Dialysis greatly can enhance the study of medically important
microbial and mammalian cells and their metabolites, and the
technique can establish the effects of certain drugs and other
therapeutic agents on the cells. Dialysis has been useful
ecologically to examine interactions between different microorganisms,
to study an organism in its natural environment and, to characterize
an organism and its metabolites in the laboratory. Fermentative
uses of dialysis principally have included the production and

recovery of cells and of both diffusible and nondiffusible compounds.

10

The history of dialysis culture can be traced to 1896 when
Metchnikoff et al. (1896) inserted a collodion sac containing
cholera bacteria into the peritoneum of a guinea pig to learn if the
bacteria produced a diffusible toxin. This history, the generalized
design, a mathematical theory, and the practical applications of
dialysis culture were brought together into a comprehensive review and
preview article by Schultz and Gerhardt in 1969 (Schultz and
Gerhardt, 1969). In a thesis afterwards, Quarles (1973) updated
their review of applications. Quesnel (1969) published a review on
methods for microculture of mammalian cells, including an extensive
discussion on perfusion (dialysis) chambers used for this purpose.
Recently in a review on the immobilization of microorganisms,
Abbott (1978) included a discussion of the fermentative uses of
dialysis culture which have appeared in the literature since the review
of Schultz and Gerhardt (1969). The present article reviews a few
papers on dialysis culture techniques published before 1969 which had
not been cited by Schultz and Gerhardt (1969) but mostly those during
the last ten years.

Dialysis culture has numerous applications. References to
more than 50 genera of microorganisms, including viruses, studied
with dialysis culture in vitro are contained in Table 1. Many
genera also have been studied with dialysis in vivo (Table 2).
Mammalian cells, more than 50 types in all, have been examined
with dialysis in vitro (Table 3) or in vivo within a variety of
hosts, including man (Table 4). Besides the cells themselves,
dialysis has been used to study both diffusible and nondiffusible

metabolites (Table 5). During the last ten years, more than 200

11

papers have been published reporting studies that have used the
dialysis technique for medical, ecological and fermentative

purposes.

 

 

TABLE 1. Genera of microorganisms studied with
dialysis culture in vitro
Microorganism Selected reference
Algae:

11a

   
  
    

green alga)

trum
Prototheca
zameeuua
Skaletonema
Thalassionema
Thalassiosira (Cyclotella)
Bacteria:
Aerobacter
Aeromonas
Arthrobacter
erase:
Brevibacterium
Clostridium
Escherichia
Bxdrggenomonas
Klebsiella
Lactobacillus
Methanomonas
gycabacterium
lasma
:isseria
Ltrobacter
Ltrosomonas
aracolobactrum

F

I. In I?

 

'0

 

 

    

8
Candida

Kluyggrggyces
licrggglyspgra
gycotorula
Saccharomyces
Schizochytrium
Thannidiun
Thraustochytr idima
Protozoa:
Entadinu-
Eudiplodiniua
Leah-Ian 1a
Ochramonas
gzypanosona
Viruses:
lchovirus
Casachievirus
Poliovirus

 

Sakshaug, 1977

Powers et a1., 1976
Jensen et a1., 1977
Prakash, 1973

Jensen et a1., 1972
Payton and Brenton, 1970
Jensen et a1., 1972
Jensen and Rystad, 1973
Pan and Umbreit, 1972b
Jensen et a1., 1972
Payton and Brenton, 1970
Dar, 1975

Jensen and Rystad, 1973
Jensen et a1., 1972
Jensen and Bystad, 1973

Hendricks and Morrison, 1967
McPeters et a1., 1974
Duxbury, 1977

Clive and Landman, 1970
Basket and Lulves, 1974
Iillardsen et a1., 1977
Landvall and Holme, 1977a
Pen and Umbreit, 1972b
Anderson and Shanmugan, 1977
Priedman and Gaden, 1970
laguib, 1975

Rightsel et a1., 1978
Hassover and Hayflick. 1974
Russinsky, 1975

Pan and umbreit, 1972b

Pen and Umbreit, 1972a
Hendricks and Morrison, 1967
Hendricks and Morrison, 1967
Abbott and Gerhardt, 1970a
Bednarski and Reporter, 1978
Ficsor and Huthiana, 1971
easel et a1., 1971

Pagarty and Griffin, 1973
Hendricks and Morrison, 1967
Damon and Hague, 1970
Collins and Tillion, 1977
xaainek, 1975a

Borichevski, 1967
lhattacharyya, 1973

Pan and Unbreit, 1972b
Lane, 1977

Edwards, 1972

Aida and Yamaguchi, 1969
Pan and Umbreit, 1972b
Payton and Stanton, 1970
Payton and Branton, 1970
Payton and Brenton, 1970

lake-urn and Kurihara, 1978
lake-urn and Kurihara, 1978
Crook et a1., 1969

Payton and Brenton, 1970
Dusanic, 1969

Hetcalf and Stiles, 1967
O'Brien and Newman, 1977
O'Irien and Newman, 1977

IEAEHHB 2.

13

Genera of microorganisms studied with

dialysis culture in vivo and ex vivo

 

 

 

 

Microorganism Host Selected reference
Bacteria:
Bacillus goat Gerhardt et a1. (1977)
Bacteroides goat Gerhardt et a1. (1977)
Clostridium guinea pig Suzuki et a1. (1971)
Escherichia goat Gerhardt et a1. (1977)
Haemophilus goat Gerhardt et a1. (1977)
Listeria mouse Osebold and DiCapula (1968)
goat Mohan et a1. (1977)
gycobacterium mouse Rightsel and Wigul (1971)
Neisseria rabbit Arko (1972)
Pseudomonas goat Gerhardt et a1. (1977)
Serratia goat Gerhardt et a1. (1977)
Staphylococcus goat Gerhardt et a1. (1977)
Streptococcus goat Gerhardt et a1. (1977)
Tre nema guinea pig Rathlev (1973)
goat Gerhardt et a1. (1977)
Protozoa:
Plasmodium chicken, turkey, mouse, Huff et a1. (1960)
duck, chick embryo
3312323253 mouse Petithory and Rousset (1969)
Yeast:
Candida goat Gerhardt et a1. (1977)

 

14

TABLE 3. Mammalian cells and tissues studied
with dialysis culture in vitro

 

 

Cell type Source Selected reference
bone marrow human Golde and Cline, 1973
fibroblasts mouse Horng, 1971
gingiva human Rose and Cattoni, 1974
lung mouse Rose and Yajima, 1977
lymphocytes rat Trowell, 1963
ovary rat Rose et a1., 1970
pancreas rat Rose et a1., 1970
salivary gland rat Rose et a1., 1970
spleen mouse Marbrook, 1967
thoracic duct lymphocytes mouse Kearney and Reade, 1974
thyroid rat Rose et a1., 1970

 

TABLE 4 .

15

Mammalian cells and tissues studied with

dialysis culture in vivo and ex vivo

 

 

Cell type Source Host Selected reference
Aortic endothelium rabbit rabbit Kitsukawa, 1969
Baby kidney hamster goat Gerhardt et a1., 1977
3°“. marrow 9°Ct goat Laissue et a1., 1974
Bone marrow human mouse Boyum et a1., 1972a
Bone marrow mouse mouse Marmot et a1., 1975
Bone marrow rabbit rabbit Millemze et a1., 1978
Bone marrow rat mouse Petersen et a1., 1974
Bone marrow rat rat Rosin et a1., 1963
Breast human rat, hamster Eugenjiva, 1970
Buffy-coat rat rat Rasmussen and Hjortdal, 1969
Distal radius mouse mouse Vilpo et a1., 1979
Epidermis mouse mouse Laerum and Boyum, 1970
Erythrocytes goat goat Gerhardt et a1., 1977
Granulocytes mouse mouse Benestad, 1970
BeLa mouse Meck et a1., 1976
Hodgkin human mouse Boecker et a1., 1975
Leukemia human rat Laerum et a1., 1973
Leukemia mouse chick Tucker and Owen, 1969
chorioallantois
Leukemia rat rat Laerum et a1., 1973
Leucocytes goat goat Gerhardt et a1., 1977
Leucocytes human goat Gerhardt et a1., 1977
Liver human guinea pig Germain et a1., 1966
Liver newt newt Grillo and Spink, 1968
Lung human rat, hamster Eugenjiva, 1970
Lymph node human mouse Gengozian, 1964
Lymphocytes human human Green, 1966
Macrophage mouse ‘mouse Tyler et a1., 1972
Macrophage
polykaryon mouse mouse Mariano and Specter, 1974
Mamary rat rat Laerum et a1., 1973
Melanoma hamster hamster, mouse Schieferstein and
Laerum, 1974
Mononuclear rat rat, mouse Benestad et a1., 1971
Myeloma mouse goat Gerhardt et a1., 1977
Myleoid elements mouse rat, mouse Niskanen et a1., 1974
Ovary human rat, hamster Bygenjiva, 1970
Peripheral
erythrocytes human mouse Gengozian, 1964
Peritoneal exudate guinea pig Guthrie and Munez, 1970
Pluripotent stem mouse mouse Miskanen et a1., 1974
Spleen guinea pig mouse Spertzel and Pollard, 1970
Spleen human mouse Gengozian, 1964
Spleen mouse mouse Borella, 1972
Spleen mouse goat Gerhardt et a1., 1977
Spleen newt newt Grillo and Spink, 1968
Spleen rabbit mouse Urso and Makinodan, 1963
Stomach human rat, hamster ngenjiva, 1970
Thymus mouse mouse Schneiberg et a1., 1968

 

16

 

 

 

TABLE 5. Diffusible and nondiffusible products studied
with dialysis culture in vitro
Product Organism Selected reference
Diffusible
cycloheximide Streptococcus griseus xominek, 1975a
Lactic acid Lactobacillus (L1) delbrueckii Friedman and Gaden, 1970

Salicylic acid

Urocanic acid

 

Pseudomonas fluorescens

 

Pseudomonas fluorescens

 

Abbott and Gerhardt, 1970

Kan and Shuler, 1978

 

Nondiffusible
Antigens
11 untyped antigens

“farmer's lung
disease” antigen

Enzymes
Amylase
Lysozyme and others
Proteinase
Proteinase
Proteinase
Proteinase
Proteinase

Protein
Protein A

Taxin
Botulinum toxin
Enterotoxin
Enotosin

Slime

Leishmania mexicana

 

Micropolyspora faeni

 

Clostridium acetobutylicum
Staphlococcus aureus
Bacillus polymyxa,

Sarcina strain (coccus P)
Serratia marcescens

sgggptococcus faecalis

 

Streptococcus thermophilus

Staphlococcus aureus A676

Clostridium botulinum
Escherichia coli
Vibrio cholerae

Pseudomonas aeruginosa

Crook et a1., 1969

Edwards, 1972

Pagarty and Griffin, 1973
Bamen and Hague, 1970
Pogarty and Griffin, 1973
Sarner et a1., 1971
Pogarty and Griffin, 1973
Miller, 1969

Pogarty and Griffin, 1973

Landwall, 1978

Ciccarella et a1., 1977
Landwall and Mollby, 1978
Bhattacharyya, 1973

Goto et a1., 1971

 

l7

2. MEDICAL APPLICATIONS
2.1 In vitro

2.1.1 Microbial cells

 

The dialysis or diffusion culture technique has been used in
many studies of a medical nature, to cultivate microorganisms and
produce their metabolites. The technique was useful for several
purposes which include the separation of cells from the macromolecules
of a complex medium, the production and recovery of compounds such as
antigens and exotoxins, and the extension of the growth and viability
of cells for prolonged study.

Rightsel and coworkers used dialysis culture as a method for an
in vitro evaluation of antileprosy drugs without the influence of a
multiplicity of host factors (Rightsel et a1., 1978). They grew
Mycobacterium lepraemurium within macrophage cultures inside diffusion
chambers that were maintained on monolayer cultures of macrophages.
The method was used to evaluate the effect of three sulfone derivatives
and rifampin on the growth of the bacterium. The results showed the
effects of these drugs on the organism and demonstrated that the
method may be useful for evaluating the chemotherapeutic effects of
drugs or their analogs when outside the host. In earlier studies, the
bacterium was inoculated into diffusion chambers on a complex medium
containing bovine serum and mouse brain extracts (Dhople and Hanks,
1972a: Dhople and Hanks, 1972b; Dhople and Hanks, 1973); during
17 days of incubation with periodic renewal of the medium, both the
cell count and the energetics of the bacterium declined.

A double dialysis method was used to produce antigenic material

from the mold that causes "farmer's lung disease" (Edwards, 1972).

18

Nutrient medium within dialysis tubing was equilibrated with an

aqueous solution of sodium chloride, and mycelia of Micropolyspora

 

fagni_were inoculated into the resulting dialysate. After growth,
the dialysate (containing organisms, dialyzable material from the
original nutrient broth, and products of microbial metabolism) was
placed in other dialysis tubing and a second dialysis was conducted
to separate the dialyzable molecules from the mycelium and the
macromolecular products. The resultant organisms and macromolecules
constitutes the antigens for further studies.

Neisseria gonorrhoeae of colony type T produced good cell yields

 

2
and maintained above 92% T colony type morphology for 12 h in a

2
modified Marbrook chamber (Ruezinsky, 1975). The chamber had two
parts, one for the medium and the other for the culture, which were
separated by a semipermeable membrane. After incubation, the viable
cell yields were not as high as those obtained from a biphasic agar-
1iquid culture system (Gerhardt and Hedén, 1960; Hunter and McVeigh,
1970; LaScolea et a1., 1975), which also is based on diffusion

principles.

L-phase variants of Streptococcus faecium were inoculated on

 

membrane filters placed on solid agar and the organisms were found to
pass through the filters, which had 0.22 um pores (Wyrick and Gooder,
1971). The reversion of protoplasts also was studied, with the
finding that the protoplasts could not pass through the filters and
form colonies under it. In some cases, however, covering the filters
with solid L-phase medium before inoculation of the protoplasts gave
rise to colonies, indicating that the three-dimensional effect of the

agar was important. In similar studies, the reversion of protoplasts

19

of Bacillus subtilis was enhanced when cell walls of the organism were

 

layered on membrane filters (Clive and Landman, 1970); the reversion
by the walls was nonspecific since the effect was also obtained with
wall preparations from other species.

A microperfusion chamber was designed by Poyton and Branton (1970)
for many purposes including the dialysis culture of tissues and
microorganisms. Duxbury (1977) developed a microperfusion chamber for
studying the growth of bacterial cells with phase contrast microscopy.
He demonstrated use of the chamber with observations on the effect

of penicillin on Arthrobacter globiformis. The Powell microperfusion

 

chamber was modified for the observation and manipulation of growing
bacteria (Issac et a1., 1975).

Vibrio cholerae was cultured in dialysis culture flasks to
produce exotoxin (Bhattacharyya, 1973). Though diffusion of nutrients
from the medium reservoir was rate-limiting, the method resulted in
increased toxin yield.

A microbial population and a portion of substrate contained in
dialysis tubing was introduced into food (Willardsen et a1., 1977).
The method eliminated the problem of retrieving a representative
portion of the population after incubation and enabled containment
of the microbe in a sample without isolation from the environment.
Results using membrane-contained strains of Clostridium perfringes
introduced into beef indicated that the method has potential for use
in studies on various food systems.

Trypanosoma lewisi was cultured in dialysis tubing containing
a medium composed of a "gel phase" of protein, agar, and defibrinated

rabbit blood (Dusanic, 1969). Agar was necessary for good growth

20

and presumably acted as a sink to absorb inhibitory products from the
medium. In addition, the system provided a good way to study amino
acid utilization.

For recovery purposes, a cellophane film was placed on an agar

plate and a broth culture of Pseudomonas aeruginosa was spread on

 

the film (Goto et a1., 1971). The film enabled the easy removal of
the bacteria and extracellular slime after growth for study of the
invasive role of the slime and its immune reactions.

Lysed blood in dialysis tubing was used for growing Leishmania

 

mexicana to produce antigen (Crook et a1., 1969). Dialysis culture was
considered the method of choice for eliminating contamination of the
antigen by medium components.

In search for a virulence factor, the technique of growing
bacteria on dialysis membranes spread over agar medium provided a
good yield of essentially all the extracellular products of
Staphlococcus aureus (Bamen and Hague, 1970). Although not all of
the products were present in a concentrated form, the significant
finding was that the entire complement of products could be detected.

Massover and Hayflick (1974) used a batch dialysis culture of
T—strain mycoplasmas to observe growth and urea metabolism. Dialysis
relieved the growth inhibition of the organisms caused by elevated
pH and increased ammonium ion concentration. The authors also used
a dialysis continuous process in which the reservoir medium was
continuously fed and replaced; the culture reservoir of 5 ml volume
was operated batchwise. The process maintained a stable and viable
culture of the organism for 10 days. Further extension of this work

should result in better characterization of the nutritional

21

requirements of T-strain mycoplasmas, which may be agents of human
disease.

Dialysis culture was used to produce toxin from Clostridium

 

botulinum type G (Ciccarella et a1., 1977). The method yielded
30,000 mouse 50% mean lethal doses of the toxin per ml, which was
750 times greater than that produced in regular broth cultures.

A diffusion chamber was used to determine the mutagenicity of test
compounds or their metabolites localized in the tissues and organs of
animals (Ficsor and Muthiana, 1971). The test organism was a

histidine-requiring strain of Salmonella typhimurium. The organism

 

was enclosed in a chamber which was exposed to blood, liver, and
testes homogenates from mice previously injected with the mutagen,
streptozotocin. After exposure, the organism was plated in an
appropriate medium.to determine the frequency of histidine mutations.
Frequencies increased substantially as compared to controls. The
method was simple and should be suitable for testing other chemicals
suspected of being mutagenic.

2.1.2 Mammalian cells

 

The dialysis culture technique has been used to cultivate a
number of mammalian cell and tissue types in vitro for reasons
similar to those for microbial cells. Ingenious variations of the
technique have been particularly important for study of cell and
tissue differentiation.

Rose and colleagues, over a period of 25 years, have developed
and refined equipment for the in vitro dialysis growth of various
types of tissue cultures (Rose, 1954; Rose et a1., 1958; Rose, 1967;

Rose et a1., 1968; Kumegawa et a1., 1968; Rose et a1., 1969; Rose

22

et a1., 1970; Rose and Cattoni, 1974; Rose and Yajima, 1977; Rose

and Yajima, 1978). Their equipment has evolved from a single multi-
purpose chamber (Rose, 1954) to a lZ—chamber circumfusion system
which allows control of nutrient and gas flow rates, pressure,
agitation and temperature (Rose, 1967) and, finally, to a dual unit
for the rotation of 24 culture chambers which have increased physical
capabilities (Rose et a1., 1970). The systems were designed not only
for growth of tissue cells but also for their differentiation. The
latter system was evaluated by cultivating human gingival explants

in the chambers for 5 to 6 weeks (Rose and Cattoni, 1974). Results
showed that the system provided microenvironments that allowed the
cells to differentiate according to their genetic potential. Fetal
mouse lung and fetal rat ovarian, thyroid, pancreas, and salivary
gland cultures also were maintained in a differentiating state.

The Rose tissue culture chamber (Rose, 1954) was modified by
inserting a membrane to divide the chamber into two compartments
(Batzdorf et a1., 1969). The resulting double chamber could be used
to study the interaction of different cells while the different cells
are physically separated by the membrane.

Marbrook (1967) studied the primary immune response of spleen
cells from unimmunized mice by growing the cells on dialysis membranes
inserted in a reservoir of medium. Three to five days after exposure
to sheep or horse erythrocytes, significant numbers of antibody
producing cells were detected. The dialysis system had the advantage
of allowing the spleen cells to grow undisturbed by agitation or
mixing, so that foci of antibody-producing cells were not disturbed.

The system, because of its simplicity and reproducibility, has

23

been widely used and modified, and is commonly called the
Marbrook chamber.

Using the Marbrook chamber to cultivate suspensions of mouse
spleen cells, a primary immune response was obtained using a polymer

of Salmonella adelaide flagellin as antigen (Diener and Armstrong,

 

1967). The antibody-forming cells increased in number to a peak

at 4 days. Experiments showed there also was an induction of tolerance.
The importance of this model system is that it allowed examination of
the induction of both antibody formation and immunological tolerance in
a defined medium using a pure antigen. A modification of the chamber
was used to determine the capacity of a given spleen-cell dose to
produce a primary antibody response (Halsall and Makinodan, 1974).

The membrane area was manipulated by changing the size of the small
inner chambers.

Mouse thoracic duct lymphocytes responded vigorously to several
mitogens in a dialysis culture system similar to the Marbrook chamber
(Kearney and Reade, 1974). The kinetics of activation caused by
the mitogens suggested that different populations of lymphocytes
were activated by different mitogens. The pokeweed mitogen, considered
to be a B-cell stimulant, caused thoracic duct lymphocytes to
differentiate to plasmacytoid cells.

A modified Marbrook chamber also was used for the purpose of
prolongation of culture conditions, in order to achieve a system
where in vitro cell responses were parallel to in vivo responses
(Maizels and Dresser, 1977). The system provided a means to cultivate
mouse lymphocytes in which resting memory cells could be stimulated

by antigen to differentiate into antibody secreting cells. Cells which

24

secreted IgG required 6 to 10 days to develop their full potential
from being resting memory cells. It was concluded that the chamber has
potential for long term experiments.

A miniaturized diffusion culture system holding 24 culture
chambers was developed by Eipirt et al. (1978). Each chamber
utilized a cell suspension volume of 0.1 ml and a reservoir of 1.0 ml.
The anti-sheep erythrocyte response of spleen cells from mice was
studied to evaluate the system, which was found to yield reliable
and reproducible primary and secondary antibody responses. Also,

a membrane filter of 0.2 pm pore diameter (Nuclepore) was found
preferable to a dialysis membrane and yielded up to a two-fold
higher response.

An artificial capillary device consisting of a bundle of hollow
fiber membranes of two types was used to perfuse both gas and medium
into tissue cultures (Knazek et a1., 1972; Knazek et a1., 1974). The
capillaries also provided a matrix for cell attachment so the
secreted products could be harvested without disturbing the culture.
This technique of combining both nutrient-passing and gas-passing
membranes in a single device allowed cells to attain tissue densities
in vitro similar to those found in vivo.

A dialysis culture system also was used to cultivate normal
and malignant human bone marrow cells in a liquid medium (Golde
and Cline, 1973). For several weeks, normal granulocytes and
macrophages proliferated and differentiated, and the entire
maturational sequence of the granulocyte and macrophage series was
identified. The system provided a convenient means for studying

cell differentiation and function.

25

L—strain mouse-fibroblast cells were cultivated in a coil
dialyzer system which removed inhibitory metabolite products and
thus promoted cell growth for a limited period (Horng, 1971).
Subsequent dialysis with intermittent renewal of the reservoir
medium did not promote growth, probably because of physical
limitations of the system.

The survival of rat lymphocytes was studied using an in vitro
organ-culture system (Trowell, 1959; Trowell, 1963). The cells in
a liquid medium were layered on a sheet of agar, which supplied
nutrients and took up metabolites. The lymphocytes were found to
have a preference for hypotonic medium. Thymus, spleen, and bone
marrow of the rat and tissues from the guinea pig and the rabbit
did not have this preference.

Spleen fragments from mice were explanted in the well of a
Millipore filter assembly which was placed in an organ culture dish
containing tissue culture medium, to study primary immune reactions
(Globerson and Auerback, 1965). The system provided a method which
permited lymphoid maintenance, differentiation and proliferation,
which together may be necessary for the initiation of immune reactions.
Tissue from.mice treated with either phytohemagglutinin or adjuvant
were stimulated to form hemagglutinins and hemolysins. Tissue from
untreated mice was necessary to invoke a response which led to
specific splenomegaly.

2.2 In vivo

2.2.1 Microbial cells

 

Dialysis culture techniques have been used with a variety of

hosts for the cultivation of many microorganisms (Table 2). The

26

technique appears promising especially for the culturing of fastidious,
pathogenic or slow growing parasitic microbes.

The exoerythrocytic stages of malaria were cultured in diffusion
chambers implanted intraperitoneally within chickens, turkeys, ducks,
chick embryos and mice (Huff et a1., 1960). Chickens readily
encapsulated the chambers, which is a major limitation of the technique.

Another finding was that Plasmodium gallinaceum, although about 0.5 to

 

1.5 um in size, could pass through 0.45 um Millipore membranes. It
was concluded that the technique should be useful for the study of
problems in parasitology. In a later study with a parasite, mice

were immunized to Trypanosoma gambiense via growth in cellulose diffu-

 

sion chambers (Petithory and Rousset, 1969).
Relatively fastidious bacteria also have been grown in implanted

chambers. Mycobacterium lepraemurium was cultured in cell-impermeable

 

chambers implanted intraperitoneally in mice, and a relationship was
found between growth of the organism and susceptibility of the host
(Rightsel and Wigul, 1971). Another important finding was that the
organism grew well in the chambers in the absence of other tissue
cells. That is, living cells were not essential and multiplication
occurred in a cell-free environment. However, inclusion of macrophages
in the chamber seemed to yield better results. The organism had a
generation time of 6 to 8 days with macrophages and 11 days in the

cell-free environments. M, lepraemurium also grew in implanted

 

chambers containing cells from a species (human embryonic skin)
other than the natural host (mice). In fact, the human skin cells
enhanced growth. Chambers without skin cells gave greater yields in

mice than in guinea pigs. Other studies confirmed that M, lepraemurium

27

can grow in the absence of host cells within implanted chambers (Dhople
and Hanks, 1972a; Dhople and Hanks, 1972b; Ito and Kishi, 1972; Dhople
and Hanks, 1973).

This model system also has been used to study the growth of

Mycobacterium leprae (Rightsel and Hall, 1976). The organism was

 

inoculated into diffusion chambers containing different mammalian
cell lines. With primary suspensions of neural elements derived
from the cerebral ventricular walls, M, leprae increased 11.3-fold
after 169 days and was infectious when inoculated into the mouse foot
pad.

Two agents of venereal disease, Treponema pallidum and

 

Neisseria gonorrhoeae, also have been grown in implanted dialysis

 

chambers. g} pallidum, inoculated into subcutaneous polyethylene
chambers inserted in guinea pigs, increased in number by about a
factor of 10 by the sixth week after infection (Rathlev, 1973). The
organisms maintained their virulence throughout the period as indicated
by the fact that samples used to infect normal rabbits intratesticularly
produced the usual syphlitic orchitis.

The effects of nitrogen and oxygen on E, pallidum were examined
in chambers implanted subcutaneously in mice (Horvath et a1., 1975).
Compared to a control, nitrogen enhanced both motility and survival
time, whereas the effect of oxygen was deleterious. Similarly,
the effect of dexamethasone and oxsuran on 2, pallidum was
investigated using chambers implanted in rabbits (Tight and Perkins,
1976). For a limited period, the organism multiplied to significant
numbers in rabbits given dexamethasone but decreased in a control

situation. Oxsuran appeared to have little or no effect on the growth.

28

This method of using subcutaneous polyethylene chambers was

developed by Arko in studies on Neisseria gonorrhoeae (Arko, 1972;

 

Arko, 1973; Arko, 1974). It is an interesting variation of the in
vivo dialysis technique. A polyethylene practice golf ball (a hollow
plastic ball with numerous holes) is implanted in the subcutaneous
tissue of rabbits. Encapsulation by connective tissue forms an

artificial chamber and natural membrane. M, gonorrhoeae grew and

 

retained virulence in these chambers, which were usable for several
weeks. The success of this system raised the possibility of studying
in vivo interactions between various factors such as antibody,
leukocytes, complement and bacteria. In addition, a relatively

simple system was provided for culturing virulent M, gonorrhoeae.

 

Disadvantages to the method included the nonspecific immune response
which may be elicited by the artificial chamber exclusive of bacteria,
"walling off" of the chambers making diffusion minimal, and difficult
sampling without sacrificing the host. It should be noted that the
membrane enclosing the organisms was tissue and not an artificial
membrane.

A similar method was employed by Flynn and Waitkins (1973)
to cultivate and study the fate of gonococci in the mouse. Inserting
vinyl cylinders under the skin provided a chamber which allowed
humoral and cellular factors free access to the gonococci. However,
the survival time (1 day) of the gonococci was much shorter than
that observed by Arko (1972). Subsequently Waitkins (1975) showed
that, when gonococci were taken into cells grown in tissue culture
and then put in chambers implanted in mice, the survival time of

the intracellular bacteria was extended. Using another subcutaneous

29

chamber technique, Scales and Kraus (1974) showed that the immunity

developed by guinea pigs infected with M, gonorrhoeae can be

 

transferred to normal susceptible guinea pigs by serum or certain
serum fractions.

The chorioallantoic membrane (a natural diffusion membrane)
of chick embryos was infected with gonococci to investigate
differences in virulence of the four gonococcal colony types
(Buchanan and Gotschlich, 1973). Type 1 and 2 organisms produced
infection significantly more often than types 3 and 4, establishing
in the animal model the same correlation between colony types and
infectivity observed in humans. As with Arko's method, a disadvantage
of this system was that sampling was difficult without sacrificing the
host.

Placement of gonococci and meningocci in Millipore diffusion
chambers implanted subcutaneously in mice provided a growth system
which allowed diffusion of humoral factors but not of cells
similarly to the method of Arko (Hafiz and McEntegart, 1977). M,

gonorrhoeae grew for only a limited period. The meningcocci survived

 

for 49 days.

Less fastidious organisms frequently have been grown in dialysis
chambers. Implanted chambers containing bursa from chicks resulted
in enhanced antibody response in surgically bursectomized chickens
immunized with human erythrocytes (Dent et a1., 1968). Bacterial
cultures were used to confirm that chambers containing the bursa were
contaminated and thus the enhanced response may have been a function

of bacterial products released from contaminated chambers.

30

Osebold and colleagues conducted a series of experiments on

cellular immunity to Listeria monocytogenes in diffusion chambers

 

implanted in the peritoneal cavities of mice (Osebold and DiCapua,
1968; Osebold et al., 1970a; Osebold et al., 1970b). This model
system was useful for studying resistance mechanisms to infections

by facultative intracellular parasites. They reported that live
bacteria must contact macrophages for development of good cellular
immunity. Humoral substances produced in response to diffusible
antigens were not able to inactivate the organism. In several
instances, sham chambers without bacteria increased the host's
resistance to later challenge with Listeria, thus indicating that
nonspecific resistance was associated with the presence of the
foreign body. This was an important finding, for it showed that such
resistance studies must be carefully interpreted to separate the role
of the chamber from that of the enclosed organism. They also found
that tissue reactions progressed around the chambers to the point
that a chronic abscess was formed, and that pleomorphic mutants of
Listeria appeared during prolonged cultivation in implanted chambers.
They further studied the relationship of antimicrobial cellular
immunity to delayed hypersensitivity (Osebold et a1., 1974).

By use of spores and cells of Clostridium botulinum contained in

 

implanted chambers within guinea pigs, it was shown that body fluids
do not attack spores but that bacteriolytic enzymes in the fluids
attack vegetative cells which release toxin (Suzuki et a1., 1971).
Polymorphonuclear leukocyte engulfment was necessary for germination

and release of the spore-bound toxin.

31

A health hazard is associated with high levels of gram negative
bacteria contaminating the fluids associated with hemodialysis
systems (Favero et a1., 1975). The bacteria in the dialysate of the
artificial kidney machines can cause pyrogenic reactions and sepsis.

Gerhardt and coworkers (Quarles, 1973; Quarles et a1, 1974;
Gerhardt et a1., 1977; Mohan et a1., 1977; Belding et a1., 1976)
purposely managed an ex vivo hemodialysis culture to investigate its
use as a technique for growing microorganisms entirely on nutrients
from the blood stream of a living animal, and yet separate from the
macromolecular and cellular defense mechanisms of the blood. One
system they used consisted of a goat, an artificial-kidney hemo-
dialyzer, and a modular fermentor (Quarles et a1., 1974). A second
system consisted of a goat and a small culture chamber (3.3 ml volume),
designed to allow use of various membranes and to remain attached to
the neck of the goat for weeks (Gerhardt et a1., 1977). In both
systems, the blood stream was shunted surgically via prosthetic
tubing from a carotid artery, through the hollow—fiber membranes in
the hemodialyzer or to the reservoir side of the membrane in the
small chamber, and back to a jugular vein. Experiments with 16
pathogenic organisms and 5 types of mammalian cells indicated that
most aerobes grew well; obligate anaerobes did not grow. They also
studied the synergism of penicillin and gentamicin against Listeria
monocytogenes (Mohan et a1., 1977).

2.2.2 Mammalian cells

Antibody-producing cells, blood and bone marrow cells, tumor

cells and various other cells from mammals have been cultivated

in vivo within diffusion chambers. Mainly, this technique is used

32

because such cells grow poorly or not at all in vitro, and the
technique can establish the existence of humoral influences on
various cellular processes. Other uses of the technique include:
the quantitation of in vivo cell growth, the determination of the
effects of chemotherapeutic drugs and their host-mediated metabolites
on the implanted cells, the study of the host response to diffusible
products from the cells within the chambers and, the study of cell
and tissue differentiation during extended cultivation periods.
2.2.2.1 Antibody-producing cells

Four human hematopoietic tissues (lymph node, spleen, bone
marrow and peripheral white blood cells) were cultivated in diffusion
chambers implanted in mice to determine their capacity to form

antibody against a test antigen, Salmonella typhosa (Gengozian, 1964).

 

Lymph node and spleen cells were induced for active antibody synthesis
whereas bone marrow and white blood cells were negative.

The effects of Rauscher Leukemia Virus (RLV) on antibody
production and induction of cellular changes was studied with a
system in which mouse spleen cells were cultivated in diffusion
chambers implanted in isogenic hosts (Borella, 1969; Borella, 1971;
Borella, 1972). It was an ideal system for study of RLV induced
immunosuppression at the cellular level. Spleen extracts from infected
mice inhibited antibody formation in the spleen cells cultured in the
chambers. Antigen stimulation of the infected cell culture was
found to alter the cellular pathway induced by the virus.

Makinodan and coworkers have published a series of articles
on antibody-forming cells (Urso and Makinodan, 1963; Capalbo and

Makinodan, 1964; Makinodan et a1., 1965; Nettesheim et a1., 1966;

33

Makinodan et a1., 1967; Vann, 1969; Vann and Makinodan, 1969; Sado,
1969; Sado et a1., 1970; Groves et a1., 1970). An implanted diffusion
chamber was used in the studies because it represented a closed
system, resulting in no problem of cells entering or leaving the
compartment being assayed. In most of their studies the immune
response of mouse spleen cells was examined using sheep erythrocytes
as antigens. Their later results showed that the growth of the
antibody-forming cells occurs non-randomly in a "staircase" manner,
and that transformation of precursor cells into antibody-forming
cells occurs shortly after mitosis.
2.2.2.2 Blood and bone marrow cells

Experimental results of using diffusion chambers in vivo for
culturing blood and bone marrow cells were reviewed in detail by
Stohlman, Quesenberry and Tyler (1973). The cells from rat, mouse,
goat, rabbit and human origin have been cultured in a variety of
hosts (Table 4).
2.2.2.2.1 Rat cells

Rat marrow cells have been cultivated within diffusion chambers
inserted in the abdomen of the rat (Rosin et a1., 1963) and in
the peritoneal cavity of the mouse and rat (Petersen et a1., 1974).
The former study found that bone formation occurred only in the
presence of osteogenic cells. The latter demonstrated that
granulopoiesis was enchanced in mouse hosts and remained at a steady-
state equilibrium in rat hosts, indicating the importance of host
factors on growth in diffusion chambers. Mononuclear cells isolated
from rat blood were cultured in chambers to study the possible

conversion of lymphocytes to macrophages (Benestad et a1., 1971).

34

Microscopic examination of homologous blood and buffy-coat cells,
cultured for 3 week periods in diffusion chambers in the peritoneum
of rats, showed that neither fibroblasts nor connective tissue fibers
developed inside the chambers if contamination with extraneous
connective tissue was prevented (Rasmussen and Hjortdal, 1969).
Consequently, the development of fibroblasts in blood and buffy-coat
culture is due to contamination with connective tissue cells during
sampling of the blood.
2.2.2.2.2 Mouse cells

Granulocytopoiesis also was examined by culturing mouse bone
marrow cells in diffusion chambers for extended periods of time
(Marmor et a1., 1975). Cells proliferated logarithmically for 7 days
and then the population stabilized for as long as 14 days. This latter
period probably occurred because diffusion became limiting, i.e., the
ratio of cell number to membrane area increased to a critical point.
The state of the cells during this period resembled granulocyte growth
in normal mouse bone marrow. Earlier it was shown that the marrow
cells can retain their biological activity within chambers inserted
intraperitoneally for periods up to 30 days (Berman and Kaplan,
1959; Berman and Kaplan, 1960). Also, in vivo there is stimulation
of granulocytopoiesis by one or more diffusible factors (Rothstein
et a1., 1971). Similarly, chamber marrow incubated for six days in
mice and stimulated with a specific antigen had increased eosino-
philopoietic activity when compared to an unstimulated control
(McGarry and Miller, 1974). This was interpreted as evidence for
the humoral stimulation of eosinophil granulocytopoieses. A study

of mouse bone marrow colonies in agar diffusion chambers indicated

35

cyclophosphamide causes the elaboration of a stimulating factor acting
on colony precursor cells (Gordan and Blackett, 1975). Diffusion
chambers also were used to examine the osteogenic potential of
irradiated marrow (Kuralesova, 1968). After 10 days of implantation,
osteogenesis began to be restored and by day 15 it was close to
normal. Diffusion chambers with Millipore or Nuclepore filters
were compared for use by culturing murine marrow (Carsten et a1.,
1975). The proliferation and differentiation of the marrow were
comparable in the two types of chambers but cell recovery was
greater from that employing the Nuclepore filters.

Murine marrow cultured in diffusion chambers implanted in mice
at concentrations of 1.1 x 106 and 1.0 x 105 cells resulted in
decreased cell growth at the higher concentration as compared to that
of the lower (Quesenberry et a1., 1974). It was suggested that the
decreased growth could be a reflection of a type of feedback inhibi-
tion or the release of various nucleases and proteases as cells were
destroyed. The high concentration also may have altered the
environment by limiting diffusion of nutrients, thus causing the
decrease. It would be interesting to determine whether this relatively
poor growth would occur if a greater membrane area was employed for the
same cell number and culture volume.

Diffusion chambers have been and should continue to be very
useful in the study of the kinetics and regulation of myelopoiesis
for the purpose of better understanding disorders of granulocytopoiesis
such as leukemia (Stohlman et a1., 1973). Studies appear to show that
the growth of mouse marrow consisting of myleoid elements, macrophages,

and pluripotent stem cells are influenced by humoral factors and also

36

the cell concentration in the chambers (Tyler et a1., 1972; Niskanen
et a1., 1974). However, stem cell growth has not been ruled out as
responsible for the increased myelopoiesis in the chambers (Tyler

et a1., 1976).

Haematopoietic cells inoculated and incubated in chambers
showed that mononuclear leucocytes give rise to granulocytes and
macrophages which grow and differentiate; however, lymphocytes from
the inoculum were gradually lost (Benestad, 1970; Breivik et a1.,
1971; Breivik, 1971; Benestad and Reikvam, 1975). Granulopoiesis
was stimulated more in the chambers than in normal steady-state
conditions (in situ) and its rate could be manipulated (Benestad,
1972; Breivik and Benestad, 1972).
2.2.2.2.3 Goat cells

Autologous bone marrow cells cultured in diffusion chambers
implanted in goats provided evidence for diffusion of hematopoietic
stimulators from the host into the chamber (Laissue et a1., 1974;
Laissue et a1., 1975). Irradiation appeared to intensify the
stimulation. Goat erythrocytes and leukocytes maintained their
numbers in an ex vivo hemodialysis culture system (Gerhardt et a1.,
1977).
2.2.2.2.4 Rabbit cells

Diffusion chambers containing bone marrow cells were inserted
into the peritoneal cavity of rabbits to evaluate factors influencing
granulopoiesis (Willemze et a1., 1978). The rabbit enabled use of
autologous bone marrow and up to 30 chambers in the cavity.
Experimental results showed that mature granulocytes inhibited

both myeloid and erythroid cell production.

37

2.2.2.2.5 Human cells

Boyum and coworkers have conducted several studies on the growth
and differentiation of bone marrow in implanted chambers (Boyum and
Borgstrom, 1970; Boyum et a1., 1972a; Boyum et al., 1972b; Boyum et
a1., 1976). In one study, human bone marrow cells were implanted in
the abdominal cavity of mice and as with mouse bone marrow cells
(Boyum et a1., 1972b), both granulocytic cells and macrophages grew
in the chambers (Boyum et al., 1972a). In a more recent study,

Boyum et al. (1976) found granulopoiesis was inhibited when mature
granulocytes from human blood or syngeneic mouse peritoneal fluid
were added to mouse bone marrow cells. Instead, there was a stimula-
tion of macrophage formation. The inhibition appeared to be tissue
specific and caused by a diffusible factor.

Other studies indicated that blood cells proliferated and were
capable of differentiating into granulocytic, erythrocytic, mega-
karyocytic and macrophagic lines in diffusion chambers, establishing
that the blood of normal man contains progenitor cells (Boecker
et a1., 1971; Barr et a1., 1975; Chikkappa et a1., 1978).

Human marrow cells cultivated in diffusion chambers, implanted
intraperitoneally in normal and irradiated mice, demonstrated that
the proliferation and maturation of granulopoietic cells was greatest
in heavily irradiated hosts (Squires, 1975). A linear relationship
existed between the number of cells inoculated and the number of
cells harvested after an 8 or 10 day period of incubation.
2.2.2.3 Tumor cells

With diffusion chambers implanted in rats, Laerum et a1. (1973)

cultivated four types of malignant cells: mammary cells of the rat,

38

melanoma cells of the golden hamster (ascitic form), leukemic cells
of the rat, and leukemic cells from two humans. All the cell types
proliferated in the chambers for periods of 8 to 13 days. Greater
proliferation of the hamster cells occurred when the cells were
cultured in hamster and mouse hosts, apparently indicating that the
cells grow better in iso- and heterologous host animals (Schieferstein
and Laerum, 1974). Four human tumors (carcinomas of the stomach,
lung and ovary, and adenocarcinomas of the breast) were found to
grow in diffusion chambers implanted in rats and hamsters (Evgenjiva,
1970). The tumors maintained their histological specificity in the
chambers. This technique may have potential for evaluating the
effects of chemotherapeutic drugs.

Host irradiation did not effect the growth of HeLa cells cultured
in diffusion chambers (Meck et a1., 1976). This observation was in
contrast to the stimulated growth of hemic cells observed in chambers
hosted by irradiated mice or goats (Boyum et al., 1972b; Laissue
et a1., 1974). Murine ascites tumor and leukemia cells proliferated
in chambers implanted in chick chorioallantois (Tucker and Owen,
1969). An advantage of this system was the inability of the
embryo to produce an immune reaction against transplanted tumor cells.
Transplantation immunity reactions also were investigated with a
dialysis technique, and the results showed that a diffusible toxic
antibody was involved in tumor transplantation rejection (Ambrose,
1969). This factor appeared to be an IgG immunoglobulin and present
only in hamsters immunized specifically to SV40 tumor specific

transplants. Adenovirus 31 tumor cells in diffusion chambers were not

39

inhibited in hamsters immunized against SV40 tumor specific
transplantation antigen.

The diffusion chamber technique has been important for the
study of human leukemic cells. The growth of peripheral blood cells
from human leukemic and control patients was characterized by such a
technique (Hoelzer et a1., 1974; Hoelzer et a1., 1976). Most types
of leukemic cells increased in number within the chambers, due
mainly to proliferation of blast cells and formation of granulopoietic
cells (Hoelzer et a1., 1977). The growth patterns differed widely
and appeared to depend on the type of leukemia. In a study of the
question of whether the maturation defect in vivo in acute leukemia
is due to environmental or cellular factors, human leukemic cells
were cultured in chambers implanted in the abdominal cavity of mice
(Fauerholdt and Jacobsen, 1975). The defect appeared to be due
to cellular factors. Boecker et a1., (1978) used bone marrow from a
patient with untreated acute promyelocytic leukemia to study leukemic
cell differentiation in diffusion chambers. For this leukemia,
results showed that microenvironmental factors can determine whether
or not gene expression is leukemic. They also cultivated Hodgkins
cells in diffusion chambers implanted in the peritoneum of
irradiated mice and found evidence that the cells probably originate
from B lymphocytes (Boecker et a1., 1975). Leukemic cells from the
rat also have been cultivated within implanted chambers (Vilpo, 1972).
2.2.2.4 Other cell types

Spleen tissues from the rat, guinea pig and mouse, enclosed in
diffusion chambers and implanted in irradiated mice, increased the

survival and hematopoietic recovery rates of the mice (Spertzel and

4O

Pollard, 1970). The results suggested that a humoral factor was
responsible for the therapeutic effect. Earlier in a similar study,
it was found that chambers containing thymus cells implanted in
thymectomized mice restored radiation resistance (Schneiberg et a1.,
1968). Altogether, the diffusion chamber technique has been very
useful for study of thymic humoral influences (Anthens, 1970; Levey
et a1., 1963; Stutman et a1., 1970; Trench et a1., 1966).
Hairless-mouse epidermal cells in diffusion chambers appeared
intact for the first 24 h, but then a significant cell loss occurred
(Laerum and Boyum, 1970). Basal cells either went into differentiation
or proliferation, but the latter process was outweighed by the cell
loss. The differentiating cells remained as such for 7 days. Cells
of rabbit aortic endothelium were cultured in diffusion chambers
implanted in rabbit abdominal cavities, in an attempt to study the
biological nature of the cells (Kitsukawa, 1969). The cells often
showed ring formations of two to three cells, and tended to form
"alveolar-like" arrangements around clusters of erythrocytes. Liver
and spleen homografts in diffusion chambers implanted in the abdominal
cavity of newts grew as indicated by DNA synthesis and cell prolifera-
tion (Grillo and Spink, 1968). This type of work may provide
information about the factors responsible for determining the
pattern of differentiation in the regenerating newt limb. The
histocompatibility interactions between mixed types of mouse spleen
cells was investigated by use of diffusion chambers (Harrison et a1.,
1968). Results showed the interactions to be immunological in nature.
The delayed hypersensitivity to l-fluoro-Z,4-dinitrobenzene (DNFB)

and Mycobacterium.tuberculosis was studied by implanting chambers

41

containing peritoneal exudate cells from guinea pigs sensitized to
these agents into unsensitized animals (Guthrie and Nunez, 1970).

The recipients developed specific skin-test reactions to DNFB but not
to old tuberculin or to a purified protein derivative. Lymphoid
tissue was cultured in implanted chambers to test the chambers made
of 1 X 2 cm envelopes of membrane filters, with 0.1 to 0.3 um pores
(Alekseeva and Yunker, 1969). No thick connective-tissue capsule
formed around the chambers. Human liver cells were cultivated
successfully in diffusion chambers implanted in guinea pigs as an
alternative to organ culture methods (Germain et a1., 1966). By use
of mouse macrophage polykaryons (inflammatory giant cells) cultured in
diffusion chambers, it was demonstrated that the giant cells form at
sites of inflammation by fusion of newly arrived macrophages with
macrophages already in situ (Mariano and Spector, 1974). The anti-
bacterial activity in the fluid obtained from subcutaneously implanted
chambers in rats was similar to that observed in serum after intra-
muscular injection of carbenicillin (Garner et a1., 1973). A
subsequent study showed that therapeutic levels of penicillin G

also were attainable in the chamber fluid after intramuscular
injection (Tight et a1., 1975). The effect of an excess of bone
mass, a hypothetical source of chalone (a diffusible tissue-specific
substance which regulates cellular growth by means of feedback
inhibition), on bone formation also was studied using a diffusion
chamber technique (Videman et a1., 1978). The DNA synthesis of the
osteogenetic cells was inhibited significantly by the excess bone,

suggesting the existence of a chalone. Cells of the mouse distal

42

radius within implanted chambers continued osteogenesis for 22 days
(Vilpo et a1., 1978).

Humans also have been used as hosts for diffusion chambers,
subcutaneously (Brooks et a1., 1960; Green, 1966). Although autologous
and homologous cells grew in such systems, the results were much more
satisfactory with chambers placed intraperitoneally in animals.

An interesting and possibly valuable use of diffusion culture
is as an in vivo test system for mutagenic and carcinogenic assays
(Huang and Furukawa, 1978). The system involves culturing human
cells or Chinese hamster cells in diffusion chambers in mice. After
injecting the host with a test compound, induction of mutations,
sister-chromatid exchanges or chromosome aberrations in the implanted
cells are used as indicators for mutagenicity or carcinogenicity.

The system has particular merit for testing compounds which need to
be metabolically activated.
3. ECOLOGICAL APPLICATIONS

The dialysis technique has been used in ecological studies
primarily for three purposes: to examine the interbiosis between
different cell populations (e.g., mutalism or parasitism); to study a
cell population sequestered in a natural environment, permitting easy
and total cell recovery after incubation; and to characterize the
in vitro growth of an organism and also the metabolites and the
effects of the metabolites.

3.1 Interbiosis studies

A dialysis technique was used for separating free-living

rhizobia from soybean cells to examine the exchange of solutes between

the two cell types (Reporter and Hermina, 1975; Reporter, 1976;

43

Bednarski and Reporter, 1976; Bednarski and Reporter, 1978).
Rhizobia of different species were found to produce chemicals that
induce soybean cells to produce other chemicals which effect the
acetylene reduction activity in test rhizobia.

The growth of autotrophs, phototrophs, and heterotrophs in
mixed cultures was studied with both dialysis and conventional
methods (Pan and Umbreit, 1972b). The stimulatory and inhibitory

effects of the cultures were found to be highly specific. Escherichia

 

coli and Pseudomonas aeruginosa had essentially no effect on

 

Nitrobacter agilis. Streptococcus faecalis caused slight inhibition

 

 

and Hydrogenomonas eutropha caused slight stimulation of Nitrobacter.

 

 

Pseudomonas fluorescens and Saccharomyces cerevisiae enhanced the

 

 

growth of Plectonema boryanum whereas M, coli inhibited the growth.

 

Dialysis culture showed less effect and was considered less
successful for the demonstration of the interactions than the use of
ordinary mixed cultures. But technical limitations such as inadequate
diffusion and aeration may have influenced the results.

For dairy purposes, the use of a double dialysis chamber enabled
the detection of an antagonistic relationship between bacteria

(Collins and Tillion, 1977). Pure cultures of Streptococcus

 

diacetilactis and g, lactis were inoculated into separate compartments

 

(divided by a membrane) of the chamber and incubated. g, diacetilactis

 

produced a diffusible antibiotic, which was indicated by the inhibited
growth of g, lactis as compared to that of a control experiment.
3.2 In situ studies
Hendricks and Morrison (1967) reported the first studies in

which a dialysis culture technique was used in situ, for observing

44

the growth of enteric bacteria in mountain stream water. Dialysis
tubing, containing six test organisms, was suspended in the stream and
samples were taken periodically. The results showed that a clear

cold mountain stream not only can maintain populations but can support
growth of enteric bacteria. The dialysis experiments did not wholly
simulate in situ conditions, however, because natural predators such
as protozoa did not have access to the bacteria.

Several phytoplankton species have been studied with dialysis
culture (Jensen et a1., 1972; Prakash et a1., 1973; Jensen and
Rystad, 1973; Skoglund and Jensen, 1976; Jensen et a1., 1976). In
situ use of the technique allowed the cells to accumulate trace
substances from large volumes of water. Results for a "batch
fermentor-continuous reservoir" dialysis system showed three growth
phases (exponential, linear, and asymptotic) as mathematically
predicted by Schultz and Gerhardt (1969). In laboratory studies, the
nitrogen-limited growth of diatoms (Skoglund and Jensen, 1976) and
the zinc and copper tolerance of three species of marine
phytoplankton (Jensen et a1., 1974; Jensen et a1., 1976) were
examined. Generally, the laboratory dialysis systems provided
conditions which simulated those in situ.

Diffusion chambers specifically were designed and used to
study the survival of coliform bacteria both in situ and in the
laboratory under controlled conditions (McFeters and Stuart, 1972;
Bissonnette et a1., 1975). In experiments with well water, fecal
streptococci and coliform (indicator) bacteria remained viable to
a similar extent (McFeters et a1., 1974). An apparatus consisting

of dialysis tubing suspended from a styrofoam flotation ring was

45

built to study bacterial growth in a farm pond (Basket and Lulves,
1974). The use and sampling ease of the apparatus was demonstrated

by inoculating a Brevibacterium sp. into the tubing and monitoring

 

its growth for 96 h. The permeability of the membranes is the limiting
factor in these ecological dialysis studies. Vargo et a1. (1975),
while using scanning electon microscopy to study the problem, found
that regenerated cellulose membranes provided an ideal substrate for
epiphytic growth in situ, thus influencing solute transport. A
diffusion chamber with an internal battery-powered stirring mechanism
for agitation was developed to study the survival of indicator
bacteria in a marine estuary (Vasconcelos and Swartz, 1976). The
stirring mechanism improved the membrane permeability, the interaction
between bacteria and pollutants and insured a uniform cell suspension
for homogenous sampling. The survival studies showed that

Streptococcus faecalis persisted longer than coliforms. In the

 

experiments, membrane filters were employed (polycarbonate with
0.4 um pores) rather than less porous dialysis membranes.

The adverse effects of chlorinated hydrocarbons on marine
phytoplankton were examined with a dialysis technique in situ
(Powers et a1., 1976; Powers et a1., 1977). The dinoflagellate
Exuviella baltica in the presence of DDE was inhibited in growth
for 2 to 4 days as compared to controls. In general, polychlorinated
biphenyls at relatively low concentrations (10 ul per liter)
adversely affected certain physiological functions (chlorophyll a
levels and carbon fixation per cell) of the algae. A dialysis
technique also was employed to study limiting nutrients and maximum

growth rates for two diatoms, Skeletonema costatum and Asterionella

 

 

46

japonica, in tanks with running sea water (Sakshaug, 1977). The
distribution and survival of Aeromonas spp., which are pathogenic
to many animals, were examined with dialysis in a freshwater lake
that receives heated effluent from a production-nuclear reactor
(Fliermans et a1., 1977; Fliermans and Gorden, 1977). At various

water depths, the survival of M, Mydrophila always was greater when the

 

reactor was in operation.

Survival of enteric viruses in estuarian waters was examined
using cellulose dialysis tubes (Metcalf and Stiles, 1967). The
virus survival was dependent on temperature, pollution levels and
the type of virus. The viruses survived in the water for 56 days in
the winter and 32 days in the summer. The inactivation rates of
polioviruses and coxsachieviruses were determined with dialysis in
situ in the Rio Grande River; the inactivation was exponential,
the rate depending primarily on the water temperature (O'Brien and
Newman, 1977).

A dialysis continuous culture system was developed for main-
taining an in vitro culture of rumen microorganisms (Nakamura and
Kurihara, 1978). For 15 days, the system maintained a concentration of
ciliate protozoa, pH value, and concentrations of ammonia-nitrogen
and volatile fatty acids similar to those found in situ in the rumen.

Interstitial water samplers were constructed using dialysis
membranes to separate water from particulate matter, including
bacteria, in lake waters and sediments (Mayer, 1976; Hesslein, 1976).
With such a sampler, Winfrey and Zeikus (1977) found that sulfate
inhibited methanogenesis in a freshwater lake sediment by altering

normal carbon and electron flow. Anaerobic metabolism of freshly

47

deposited particulate organic matter initially occurred at the
sediment-water interface in a hypereutrophic lake (Molongoski, 1978).
3.3 In vitro studies

Growth of the strictly autotrophic Thiobacillus thiooxidans

 

ceases when the organism metabolizes sulfur and keto acids accumulate
to inhibitory levels. Dialysis greatly increased the growth by
removing the inhibitory acids (Borichewski, 1967). A dialysis
continuous fermentation was used to obtain a substantial increase in
cell density of the methane-oxidizing strain M 102, confirming that
the growth of the organism.was inhibited by its metabolites

(Naguib, 1975).

Dialysis culture techniques also demonstrated that presumably
obligate autotrophic bacteria will grow on glucose rather than
inorganic energy sources if metabolic products are prevented from
accumulating in the culture environment (Pan and Umbreit, 1972a).

Nitrosomonas europaea, Nitrobacter agilis, Thiobacillus

 

denitrificans, 2, neapolitanus, and I, thioparus were grown on

 

 

glucose-salts media in the absence of the specific inorganic
energy sources. Pyruvic acid was the metabolic product found to
inhibit M, agilis, but the toxic product for M, europaea was not

identified. Results with the Thiobacillus species indicated that

 

pyruvate (or related keto acids) may be more inhibitory to the
organism when growing on glucose than when growing on a specific
nutrient. This work was important for it demonstrated a qualitative
as well as a quantitative difference between dialysis and non-
dialysis culture, and indicated that redefinition of the term

"obligate autotroph" may be necessary.

48

Recently, a dialysis system was used to examine the energetics

of nitrogen fixation of (Nif)—derepressed mutants of Klebsiella

 

pneumonia (Anderson and Shanmugam, 1977). The system consisted of a
dialysis bag (containing inoculum and 25 ml of medium) suspended in a
flask containing 250 ml of medium. This system allowed fresh medium
to diffuse to the culture and products (e.g., NH4+) to diffuse out.
Results from monitoring the production of NH + and H showed that

4 2

nitrogenase-cataloyzed H production is a major factor in the economy

2
of nitrogen fixation in vivo.
4. FERMENTATIVE APPLICATIONS

Dialysis has been applied to fermentations in order to enhance
the production and recovery of microbial cells and of their diffusible
and nondiffusible products. Dialysis can improve a fermentation
by two interrelated yet different approaches. In one, a membrane
and a reservoir are employed to remove inhibitory metabolites from
a fermenting culture, resulting in increased rate and extent of
substrate conversion, and permitting use of a relatively high
concentration of substrate. The dialysis separation of small
molecular products additionally represents a recovery step. A
second approach is to dialyze the fermentor contents against
a reservoir of substrate medium primarily to retain cells in the
fermentor (and nondiffusible products), resulting in increased cell
populations for recovery purposes. Such a process applied
continuously also can allow increased throughput of substrate for
conversion purposes. This second approach can be applied to
fermentations not inhibited by metabolite products as well as

to those that are.

49

4.1 Production of cells

Dairy starter cultures of Streptococcus and Lactobacillus sp.

 

 

produce lactic acid which inhibits their growth. Dialysis can be
used to prevent this inhibition by removing lactate from the
bacterial culture (Gerhardt and Gallup, 1963; Friedman and Gaden,
1970). Osborne tested (Osborne, 1977) and patented (Osborne et

a1., 1975) such a fermentation using a completely batch dialysis
culture system. The design was based on that of Gallup and Gerhardt
(1963) and consisted of three components: a fermentor containing
culture, a dialyzer, and a reservoir containing medium. The system
had 5 cm2 of membrane area for every ml of culture, which was

necessary for the lactate-sensitive Streptococcus sp. The fermentor

 

to reservoir volume ratio was 1:15. Operation of the system with

Streptococcus cremoris for 24 h resulted in concentrations between

 

l and 1.5 x 1011 cells per ml. Using another fully autoclavable
system with membrane filters (cellulose acetate with 0.2 um pore
size), a 1:30 fermentor to reservoir ratio and a fermentor circuit
volume of 500 ml, Osborne concluded the system can produce in a
40 week year at least half the annual starter cultures necessary
for a dairy plant utilizing 100,000 gallons of milk per day for
cheesemaking. The resulting starter concentrates had activity
better than conventional starters. The dialysis system also
eliminated the requirement for centrifugation which is costly,
damages the cells, and easily results in contamination.

Another dialysis process for production of concentrates of

Streptococcus lactis was unsuccessful, probably because the system

50

had a small membrane area in relation to the culture volume (less
than 1 cm2 per m1; Bergere and Hermier, 1968).

Friedman and Gaden (1970) also used a batch dialysis system
of three components (fermentor, reservoir and dialyzer) to study

the growth and lactic acid production of Lactobacillus delbrueckii.

 

The results showed the specific acid production rate was independent
of the specific growth rate at low lactate concentrations; this
could not be determined in a conventional fermentation. They also
confirmed the inhibitory effect of lactate on the growth of the
organism. Further, they quantitated the inhibition and incorporated
it into the mathematical model of Leudeking and Piret (1959)
relating the rate of growth to the rate of acid production.
Similarly to the lactic acid bacteria, the growth of a yeast,

Mycotorula japonica, on hexadecane or decane was improved greatly

 

by using dialysis to relieve the inhibitory effects of lauric acid
(Aida and Yamaguchi, 1969).

Single-cell algal protein from sewage was produced by inoculating
Scenedesmus obliquus into dialysis tubing containing tap water
(300 ml) and suspending the tubing in a beaker containing 1.8 liters
of raw sewage (Dor, 1975). Dialysis resulted in a 3-fold increase
in the cell population as compared to a control. Limiting the
growth was a lack of light, membrane permeability, membrane area
and agitation. This fermentation has promise especially
because the reservoir containing sewage can continuously be
replenished.

Lane (1977) investigated dialysis culture as a means to produce

high concentrations of a food yeast, Kluyveromyces fragiles, from

 

51

deproteinized whey, a residue of the dairy industry. He used an
8-1iter fermentor, a fermentor to reservoir ratio of 1:2.5, a
fermentation period of 90 h, and four successive fresh reservoirs of
medium. There resulted a biomass concentration of 90 g per liter and
residual lactose concentration of 5 to 8 g per liter in the
reservoirs. In this fermentation system, dialysis results in a high
cell concentration primarily because, at a given time, the culture
is sequestered to a small part of the medium but, in the long term,
the culture has access to all the medium. This principle differs
from the lactic acid dialysis fermentations where dialysis primarily
serves to enhance cell growth by removing inhibitory products from
the culture.

In the above studies substrate mostly is supplied to the
culture via a membrane from a reservoir containing medium. Cell
growth is limited either by the rate of diffusion of substrate into
the culture, by the accumulation of an inhibitory product into the
culture, or a combination of both. Landwall and Holme (1977a, 1977b)

dialyzed cultures of Escherichia coli B against a medium reservoir

 

to supply substrate and remove toxic products. They ensured that
substrate was not limiting by also feeding substrate directly into
the culture chamber. In this manner, and using a fermentor to
reservoir ratio of 1:11, they obtained biomass concentrations of
140-150 g dry wt. per liter as compared with 30 to 40 g per liter in
a culture without dialysis. Their system utilized dialysis principly
to remove toxic end-products which decrease the molar growth yield

of M, coli on glucose.

52

4.2 Production of nondiffusible compounds
A dialysis method, to obtain concentrates of extracellular

proteinase of Streptococcus faecalis var. liquifaciens consisted

 

 

simply of growing the bacterium within dialysis tubing suspended in
an Erlenmeyer flask containing medium (Millner, 1969). The dialysis
methods for obtaining concentrates of nondiffusible enzymes of course
are like those for obtaining cells. Concentration of two proteolytic
enzymes via dialysis culture also was demonstrated by growing a
Sarcina strain (coccus P) in a dialysis bag suspended in a reservoir
of medium (Bissel et a1., 1971; Sarner et a1., 1971) with a culture
volume to reservoir volume ratio of 1:10. A concentration of

2.4 x 109 cells per ml was obtained with dialysis as compared to

6 x 108 in conventional flasks, and the proteinases and other
nondiffusible compounds were similarly concentrated. A dialysis
apparatus for producing concentrates of extracellular proteases and
amylase consisted of two dialysis bags (Visking tubing), each pro-
tected by a perforated cylindrical glass shield and suspended in a
two-liter glass vessel (Fogarty and Griffin, 1973). The apparatus

was tested by cultivating Bacillus polymyxa, Serratia marcescens,

 

 

Streptococcus thermophilus and Clostridium acetobutylicum within the

 

dialysis bags. The concentration of the enzymes in the bags was
4.8 to 22.7-fold greater than in conventional fermentation flasks.
Considering the total bulk of medium used, however, there was no
increase in the amount of enzyme produced. Mainly, the study
demonstrated that dialysis can serve to benefit enzyme recovery by

restricting the enzymes from the bulk of the medium.

53

The production of heat-labile enterotoxin from M, ggli_by
dialysis culture with a continuous feed of substrate was examined by
Landwall and Mollby (1978). The fermentor to reservoir ratio was
1:10 and the feed (mostly glucose at 600 g per liter) was introduced
into the fermentor at a rate so that the carbon source was not
limiting. The dialysis process produced at least eight times more
enterotoxin than did an ordinary fermentation process. Using
similar dialysis procedures, a lO-fold increase in the concentration
of extracellular protein A (2 g per liter) was achieved from

Staphlococcus aureus A676 as compared to results from a control

 

(Landwall, 1978).
4.3 Production of diffusible compounds

Abbott and Gerhardt (1970a) demonstrated that dialysis culture
can be used to increase the production of a diffusible metabolite.
Pseudomonas fluorescens was used to convert naphthalene to salicylic
acid, which usually accumulates to inhibitory levels. A dialysis
flask system was used with a culture volume of 150 m1 and a dialysate
volume of 1100 ml. The flask was placed on a rotary shaker to
provide aeration and mixing. The fermentation was conducted for 15
days with intermittent (4 times) removal and replenishment of medium
in the dialysate reservoir in order to maintain low levels of
product in the culture chamber. The rate of the fermentation increased
throughout the time, suggesting that the fermentation could be improved
by operating the reservoir continuously. The total amount of salicylic
acid produced, though dilute, was determined to be 20-fold greater as

compared to a control without dialysis. A kinetic study showed that

54

maintenance metabolism accounted for 84 percent of the salicylate
produced (Abbott and Gerhardt, 1970b).
Dialysis also was used in an attempt to alleviate product

control over threonine production by an auxotroph of Escherichia

 

ggli_(Abbott and Gerhardt, 1970c). Inhibition instead of enhancement
was observed. This was explained by depletion of a-s-diaminopimelic
acid as the limiting factor rather than threonine inhibition of its
own synthesis.

Kominek (l975a, l975b, l975c, l975d) developed and patented
a dialysis extraction apparatus for production and recovery of

cycloheximide, an antifungal antibiotic, by Streptomyces griseus.

 

As in the lactate and salicylate fermentations, cycloheximide
inhibits its own synthesis; also, cycloheximide is rapidly degraded
in the absence of glucose. The dialysis system consisted of a S-liter
New Brunswick fermentor (3-1iter volume) with a rubber mesh mat
fastened around the baffles to support dialysis tubing (5500 cm2).
The reservoir included the dialysis tubing and also an extractor,
containing methyl chloride, that was connected by conduits to the
tubing. Operationally, the dialysate from the tubing was bubbled
through the extractor to remove the cycloheximide. The dialysate
then was aerated to remove traces of the solvent, and finally was
returned via conduits to the tubing for collection of more
antibiotic. Glucose continuously was fed into the fermentor to
prevent cycloheximide degradation and so that the carbon source did
not become limiting. Results showed that, by the tenth day of
continuous operation, there was a two-fold increase in titer as

compared to a control. Also, 82 percent of the solids content of

55

the extractor was cycloheximide. Therefore, the dialysis extraction
apparatus increased the titer of the antibiotic, enhanced recovery of
the antibiotic in a relatively pure form, and eliminated the need for
a large reservoir. Kominek concluded that this process should have
application "to any fermentation product that is dialyzable and
possesses a favorable distribution in a suitable water immiscible
solvent." This work was the first to demonstrate the potential of
dialysis for continuous production and recovery of a diffusible
secondary product of microbial metabolism.

Hollow fiber membrane dialyzers (Cordis-Dow artificial kidneys)
were used to immobilize microbial cells for the continuous production
of biochemicals (Kan and Shuler, 1978). In this system, the cells
were immobilized by containing them in the chamber volume (within
the dialyzer jacket) outside the hollow fibers; while solutes freely
diffused across the hollow fiber membranes. The system was tested
by examining the conversion of l-histidine to urocanic acid by a

heat-treated suspension of Pseudomonas fluorescens. All of the

 

continuous production experiments lasted for at least 8 days.
However, the rate of the reactions decreased with time, most likely
as a result of degradation and denaturation of the enzyme. This type
of system may have potential if microbial cells can be maintained in
a living yet nongrowing state. The half-life of a reactor containing
such cells (disregarding membrane fouling) would be indefinite for
the continuous production of a diffusible metabolite.

Promising experiments using a dialysis system to immobilize
living microbial cells for continuous production of ammonium-

1actate, were reported by Stieber and Gerhardt at a recent meeting

56

(Abstract, Amer. Chem. Soc. Meeting, Miami Beach, 1978). In other
work (Stieber, 1976), the Cordis-Dow artificial kidney was found
unsuitable for continuous fermentation use because the design and lack
of agitation in the dialyzer jacket resulted in rapid clogging and
decreased diffusion.

Low pressure ultrafiltration was used to separate liquid from
solids in an activated sludge system (Hardt et a1., 1971). The
membrane flux rate was 4.5 gallons per day per ftz. This fermentation
system consistently removed 97-99 percent of the COD in the influent
waste stream. During operation, biomass was monitored by oxygen
uptake. The results indicated that the biomass decreased from
30 g per liter to 20 g per liter during the first six days, remaining
constant thereafter (48 h). The decrease was attributed to cell lysis.
Ultrafiltration was a very effective way of separating microbes from
effluent, and it enabled effective control of the biomass to substrate
ratio.

4.4 Dialysis sensors

On-line sensors are becoming increasingly important in the opera-
tion of fermentations (Dobry and Jost, 1977). Schultz (Abstract,
2nd Inter. Conf. Computer Appl. Fermentation Technol., Philadelphia,
1978) has used principles of dialysis to develop an affinity sensor
for continuously monitoring glucose levels. The method depends on
the reversible competition of glucose and a fluorescent labeled
polysaccharide (FITC-Dextran) for a specific binding site on a
protein, Concanavalin-A. Operationally, glucose from the
fermentation liquor diffuses via a membrane into a probe and displaces

the labeled polysaccharide from the protein. Both the polysaccharide

57

and the protein are impermeable to the membrane. An optical system
was used to measure the polysaccharide. Stability studies indicated
that the equilibrium reactions involved were stable for up to 100 h.

Cell-free samples were generated from a fermentation broth for
on-line analysis by a dialysis process by Zabriskie and Humphrey
(1978). A baffle in a fermentor was converted into a dialyzer, the
membrane of which separated the fermentation broth from a continuous
stream of fresh water. Water leaving the dialyzer (dialysate) was
channeled to a commercial glucose analyzer. A major limitation of
the process was the relatively slow response time. This method and
that of Schultz (1979) both have application for sensing compounds
other than glucose.

5. DISCUSSION
5.1 In vivo

Since Algire (1954) first used membrane filters as chamber
walls, improvement in the design of in vivo dialysis systems has
progressed minimally. Several problems need solutions and certain
characteristics need study so that experiments are more reproducible
and meaningful. Needed research areas comprise: the non-specific
immune response caused by the implanted chambers; the growth of
macrophages and connective tissue around the chambers, thus limiting
diffusion; the lack of agitation which results in a nonhomogeneous
cell suspension; the difficulty of controlling the concentration of
biologically active compounds within the chambers and; the lack of
knowledge concerning the diffusion capacity and rate for the
chambers. Answers to some of these needs should enhance the

importance of in vivo dialysis culture for study of bacterial and

58

blood diseases, tumors, and immune mechanisms, and for evaluation
of chemotherapeutic strategies for certain of the disorders such as
leukemia.

5.2 In vitro

In vitro dialysis systems for medical and ecological use usually
are of small scale and are designed for a specific purpose so that
the systems used are as many as the purposes. Generally, the membrane
area should be relatively large compared with the culture volume and
the culture should be agitated to prolong membrane utility and to
provide homogeneous conditions. Membrane filters, which separate
only cells from solutes, appear more useful than smaller-pored
dialysis membranes.

To date, the best in vitro dialysis system for fermentative pur-
poses is that built and used by Gerhardt and coworkers (Gerhardt and
Gallup, 1965; Schultz and Gerhardt, 1969; Humphrey, 1970). The
system consists of three separate components, a fermentor, a reservoir,
and a dialyzer, and it can be operated batchwise, continuously or in
a combination of these modes. Since the components are separate,
each may be individually controlled. The dialyzer is of the plate
and frame type and allows variability in membrane type and area,
is sterilizable, and is designed so that there is much turbulence on
both sides of the membrane, preventing membrane fouling for extended
periods (Humphrey, 1970). Though the dialyzer specifically was
designed for microbiological purposes, much improvement seems
attainable. In recently designing a dialyzer for hemodialysis
purposes, Dawids and Boe (1977) tested 27 different injection

molded membrane support patterns. The patterns were judged on

59

several properties, e.g., compliance and rupture level of the membrane,
and 15 patterns were found to be acceptable. Some of the properties
were as much as 200 percent better than those of conventional
patterns. Such considerations were not undertaken before (Schultz
and Gerhardt, 1969).
5.3 Electrodialysis

Refinements of dialysis such as electrodialysis (Young, 1974;
Johnson and Hill, 1976) and ultrafiltration (Michaels, 1968) increase
the flux of certain solutes through a membrane. Electrodialysis is
an electro-chemical process for separating different ionic species
in solution by the use of direct-current electricity and ion-permeable
membranes. The transport of ions across a membrane depends on the
ability of the membrane to conduct electricity. The process separates
various ions by use of different types of membranes (e.g., anion-
selective, cation-selective, or multivalent-selective). Published
studies of fermentation processes employing electrodialysis are few.
Hongo and Iwahara (Chemical Abstracts, Vol. 88, 1978) grew

Brevibacterium fulvum on a glucose medium and the fermentation pro-

 

duct was first dialyzed and the resulting cell suspension was returned
to the fermentor. The dialysate was channeled to an electrodialysis
system to separate glutamic acid from the carbon and nitrogen sources,
which were returned to the dialysate. The yield of glutamic acid

was 51.3 percent with respect to glucose. Recently, electrodialysis

was used for a L-lysine fermentation by Brevibacterium flavum

 

(Iwahara et a1., Abstract, Meeting Am. Soc. Microbiol., Honolulu,
1979). Selective removal of the L-lysine from the culture resulted

in a 50 percent increase in yield as compared to a control.

60

5.4 Microfiltration

Ultrafiltration is a process for separating solutes of different
sizes by means of their unequal passage through a membrane because
of a pressure gradient. The process of using ultrafiltration to
separate only microbial cells from solutes is called microfiltration.
Michaels (1968) was the first to suggest and diagram a microfiltration
system for continuous microbial fermentation. Sortland and Wilke
(1969) and Pirt and Kurowski (1970) devised filtration methods for the
retention and concentration of microbial cells in a continuous culture
of limited duration. Wang et a1. (1970) used an ultrafiltration cell
as a fermentation and filtration device for the retention of

Clostridium histolyticum cells and enzymes, and for removal of small

 

molecular weight products. The process produced 1.55 times the
enzyme units per mg of cell dry weight and 2.5 times the cell
concentrations that were obtained in an ordinary batch fermentor.
Fensom et a1. (1974) used a microfiltration cell to retain non-

growing suspensions of Agrobacterium tumefaciens as part of a

 

continuous-flow apparatus. The process lasted about five days and

was used to study the kinetics of ferricyanide reduction and
3-ketosucrose production. Vieth and Venkatasubramanian (1974)
immobilized microbial cells by use of a hollow-fiber recycle reactor
to eliminate several expensive steps in enzyme purification. Vera

and Wang (Abstract, Meeting Am. Chem. Soc., Chicago, 1977) reported
work on acetic acid production with microfiltration for cell

retention. This work demonstrated that a hollow fiber filter
fermentor with intermittent backflush can operate for extended periods.

A heuristic indication of improved productivity also was presented.

61

5.5 Dialysis vs. microfiltration

The above studies as a whole indicate that microfiltration
systems can improve upon dialysis systems for production of dense
cell populations and recovery of cell-free metabolites. The rates
would be increased, the membrane area per volume needed would be
relatively small, and a large or replenishable reservoir would
not be required with microfiltration. But microfiltration cannot
relieve a culture from the effects of inhibitory products to as
great an extent as dialysis. Thus, for a continuous fermentation
which produces inhibitory and diffusible metabolites, the application
of dialysis enables use of a relatively high concentration of feed
substrate; microfiltration does not. Whereas there is an exchange
of solutes across the membrane in a dialysis process, the flow of
solutes for a microfiltration process is in one direction. Because
of these process characteristics, microfiltration usually must be
managed continuously; dialysis can be managed in a batch or
continuous mode. Accordingly, with respect to the sterilization
requirements of many microbial fermentations, dialysis may have
advantage over microfiltration. As with dialysis, the major
limitation of microfiltration is membrane fouling and consequently
an ultrafilter also needs to be designed specifically for
microfiltration use. Altogether, considerations are many in order
to choose a membrane process and a mode of operation that are best
suited for a microbial system.

The various membrane processes easily lend themselves to
mathematical modeling and computer simulation, greatly aiding

the development of a particular process. The design of a more

62

efficient dialyzer and a microfilter along with the use of such
modeling should further advance the resource of membrane technology
for cell culture and metabolite production.
ACKNOWLEDGMENT
The use of reference material, including discussion,
accumulated in the thesis of John M. Quarles is appreciatively

acknowledged.

63

MATERIALS AND METHODS

The details of materials and methods are presented in Articles I

through VII. A reference index is given below (Table 6):

TABLE 6. Index of materials and methods

 

 

Item Article and Page
MATERIALS
Organism for experimental tests. . . . . . . II - 733
Whole dried sweet-cheese whey. . . . . . . . II - 733
Deproteinized dried cheese whey. . . . . . . III - 490
Dialysis continuous fermentation system. . . II - 733
Modifications of the systems . . . . . . . III - 490
Conventional continuous fermentation system. V - 7
Dialysis immobilized-cell system . . . . . . VII - 10
METHODS
Fermentation operation
Dialysis continuous fermentation system. . II - 733
Conventional continuous fermentation
system . . . . . . . . . . . . . . . . . V - 6
Dialysis immobilized-cell system . . . . . VII - 10
Sampling procedures. . . . . . . . . . . . . II - 734
And. . . . . . . . . . . . . . . . . . . . IV - 8
And. . . . . . . . . . . . . . . . . . . . V - 7
And. . . . . . . . . . . . . . . . . . . . VII - 11
Measurements
Lactic acid. . . . . . . . . . . . . . . . II — 734
And. . . . . . . . . . . . . . . . . . . III - 490
Lactose. . . . . . . . . . . . . . . . . . II - 734
And. . . . . . . . . . . . . . . . . . . III - 490
Specific conductance . . . . . . . . . . . III - 490
Dry weight . . . . . . . . . . . . . . . . IV - 9
And. . . . . . . . . . . . . . . . . . . V - 8
Protein. . . . . . . . . . . . . . . . . . IV - 8
Optical density. . . . . . . . . . . . . . IV - 8

 

64

ARTICLE I

DIALYSIS CONTINUOUS PROCESS FOR

AMMONIUM-LACTATE FERMENTATION OF WHEY:

MATHEMATICAL MODEL AND COMPUTER SIMULATION

BY

G. A. Coulman, R. W. Stieber, and Philipp Gerhardt

Reprinted from Appl. Environ. Microbiol. 343725-732, 1977.

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Dec. 1977, p. 725-732

Copyright © 1977 American Society for Microbiology

Vol. 34, No. 6
Printed in U. S.A.

Dialysis Continuous Process for Ammonium-Lactate

Fermentation of Whey: Mathematical Model and
Computer Simulation'f

G. A. COULMAN,“H' R. W. STIEBER,2 AND PHILIPP GERHARDT’“

Departments of Chemical Engineering‘ and of Microbiology and Public Health,2 Michigan State University,

East Lansing, Michigan 48824
Received for publication 8 April 1977

A mathematical model was developed to describe a dialysis process for the
continuous fermentation of whey lactose to lactic acid, with neutralization to a
constant pH by ammonia. In the process, whey of a relatively high concentration
is fed into the fermentor circuit at a relatively low rate so that the residual
concentration of lactose is low. The fermentor effluent contains ammonium
lactate, bacterial cells, and residual whey solids and could be used as a nitrogen-
enriched feedstuff for ruminant animals. Only water is fed into the dialysate
circuit at a relatively high rate. The dialysate effluent contains purified ammo-
nium lactate and could be converted to lactic acid and ammonium sulfate for
industry. The fermentation was specifically modeled as a set of equations repre-
senting material balances and rate relationships in the two circuits. Dialysis
continuous fermentations, in general, were modeled by combining these equations
and by using dimensionless parameters. The generalized model was then solved
for the steady state and used to simulate the specific fermentation on a digital
computer. The results showed the effects of various material and operational
and kinetic parameters on the process and predicted that it could be operated

efficiently.

 

Whey utilization is a large and worsening
problem to the dairy industry. American
cheesemakers annually discard about 107 metric
tons of cheese whey because of its limited mar-
ket value. Since about half of the solids originally
in milk are left in whey, its discard is an envi-
ronmental burden as well as an economic and
nutrient loss.

A potential solution to this problem lies with
the conversion of whey into feedstuff for rumi-
nant animals, accomplished by the bacterial fer-
mentation of whey lactose into lactic acid and
its neutralization to a constant pH by ammonia.
The primary purpose of the fermentation is to
trap ammonia as ammonium lactate, thereby
greatly increasing the nitrogen content of whey
and decreasing the toxicity of ammonia. After
condensation, the product is stable and can be
used effectively as a ruminant feed supplement.
The background of this development has been
reviewed by Keller and Gerhardt (4) and Reddy
et al. (7).

Practical usefulness is dependent on the abil-
ity to conduct the fermentation as cheaply and

1‘ Journal article 7986 from the Michigan Agricultural Ex-
periment Station.

1") Present address: Department of Chemical Engineering,
Cleveland State University, Cleveland, OH 44115.

725

efficiently as possible. It can be conducted ho-
mofermentatively without sterilization or asep-
sis because of the restrictive fermentation con-
ditions of low pH (5.3 to 5.8), high concentration
of undissociated acid, high temperature (44°C),
and anaerobiosis. The fermentation can be con-
ducted either in batches (7) or continuously (4),
the latter process resulting in increased conver-
sion efficiency with time.

A further increase in efficiency appeared pos-
sible by the application of dialysis (i.e., the sep-
aration of solute molecules by their unequal
diffusion through a semipermeable membrane
because of a concentration gradient) with the
fermentation (l, 2, 8; P. Gerhardt and D. M.
Gallup, US. Patent 3186917, June 1965).
Thereby, the small molecular products are re-
moved from the immediate environment of the
bacterial cell (and ultimately from the intracel-
lular enzyme site), thus relieving the feedback
inhibition by a product that normally regulates
its production. As more product is withdrawn
by dialysis, more substrate is consumed and
more product is made; i.e., the fermentation
becomes more efficient. As the cell population
attains high density, the substrate is increasingly
converted into product by maintenance rather
than by growth metabolism, thus also improving

 

726

fermentation efficiency. The dialysis separation
of much of the small molecular products addi-
tionally represents a purification step towards
alternative uses, such as the conversion of am-
monium lactate into lactic acid for use as an
industrial chemical and into ammonium sulfate
for use as a fertilizer (6). On the other hand,
the application of dialysis increases cost, com-
plicates operation, and eventually reduces effi-
ciency because of membrane fouling. The poten-
tial advantages must compensate the disadvan-
tages if a dialysis process is to be useful.

In our investigations, the goal was to assess
the feasibility of applying membrane technology
to the fermentation. Continuous operation ap-
peared desirable not only for the best use of
membrane technology but also because waste
whey is generated and the fermentation prod-
ucts could be used more or less continuously.
Among the several membrane techniques, di-
alysis appeared desirable for the initial study
because of its simplicity and the existence of a
theoretical basis for its application to bacterial
culture (8).

In the studies reported in this article, our
objective was to develop a mathematical model
of a dialysis continuous process specifically for
the ammonium-lactate fermentation of whey,
but generalized with dimensionless parameters
so that the model could be applied to other
fermentations. The resulting set of equations
was solved for the steady state and used to
simulate the specific fermentation, on a digital
computer. The experimental tests are reported
in a companion article (9).

MATHEMATICAL MODEL
Design of fermentation system. Figure 1

shows a schematic of a completely continuous
dialysis system designed for the ammonium-lac-

3f°»Pf°~ Ff H20» Fa

  
   
 
 
  

  
    
   
 

 

l—_ _—l

I

I |

l |

l |

I.- _VQJ

vapfvxfIFf Sd'Pd'Fd
FERMENTOR DIALYSATE
CIRCUIT CIRCUIT

FIG. 1. Schematic of dialysis continuous fermen-
tation system. Symbols are described in Table I.

COULMAN, STIEBER, AND GERHARIYI‘

APPL. ENVIRON. MICROBIOL.

tate fermentation of whey. The symbols corre-
spond to those used previously (4, 8) and are
listed in Table l. The system differs from the
usual dialysis culture systems (see Fig. 8 in ref-
erence 8) following an earlier idea to dialyze a
fermentation against a stream of water (see Fig.
3 in P. Gerhardt and D. M. Gallup, US. Patent
3186917, June 1965). This provides the greatest
concentration gradient. possible for dialysis and
exploits the greater permeation rate of the prod-
uct. (lactate) rather than that of the substrate
(lactose), thus maximizing efficiency of the fer-
mentation.

The feed into the fermentor circuit contains
the substrate (lactose) in relatively high concen-
tration (8,”) and is maintained at a relatively
low rate of flow (F ,) consistent with a minimal
concentration of unused substrate in the fermen-
tor effluent (S,). The substrate is converted es-
sentially only to a single product (lactic acid)
that remains in the fermentor effluent (P,). A
small concentration of the product may preexist
in the whey feed (Pf’). The liquid volume in
the fermentor (V,) and the cell population (X,)
are maintained at constant levels. The fermentor
contents are thoroughly mixed and are continu-
ously circulated through one side of the dialyzer.
The addition of ammonia solution is not shown
because it does not enter into the modeling.

The usual reservoir vessel for dialysate is elim-
inated. Instead, the dialysate circuit consists
only of the tubing, pump, and dialysate side of
the dialyzer and contains a relatively small vol-
ume (V,,). The feed into the dialysate circuit
consists only of water and is maintained at a
relatively high flow rate (F 0,) consistent with a
very low substrate concentration (80,) and a
practically useful product concentration (Pd) in
the dialysate effluent.

For purposes of mathematical modeling, the
assumptions are made that high mixing and
circulation rates insure homogeneity throughout
the system, liquid turbulence and excess mem-
brane surface insure insignificant fouling of the
dialyzer membranes for a practically useful pe-
riod, bacterial metabolic rates remain constant,
the volume of the dialysate circuit is negligible
relative to that of the fermentor, the rate of
ammonia-solution addition is negligible relative
to that of substrate feed, and pressures are equal-
ized between both circuits.

Material-balance equations. A set of ma-
terial-balance equations was developed analo-
gous to those in previous dialysis-culture theory
(see equations 21, 22, 23, 51, and 52 in reference
8). The equations were formulated in general as
follows: input + production = output + accu-
mulation.

Equations for the substrate, product, and cell

R0810]-

corre-
nd are
m the
in ref-
Llyze a
99 Fig.
Patent
'eatest
is and
pad
Istrate
re fer-

ntains
Incen-
tively
nimal
'men-
ed es-
acid)
9,). A
29th
me in
I (Y)
entor
tinu-
yzer.
Iown
;.

elim-
Isists
le of
vol-
rcuit
at a
th 3
Id 8
,) in

the
and
mat
em-
the
Pe‘
ant.
Ible

of
:iV'e

ral-

na-
110'
012V
109
I as
cu-

~ell

MODEL AND SIMULATION OF DIALYSIS FERMENTATION 727

 

 

 

 

 

 

 

 

 

 

VOL. 34, 1977
TABLE 1. Glossary of mathematical symbols
Symbol Description _ _._. fillnit‘s_ M
A". Area of membrane available for dialysis cm2
F4 Flow rate into and out of dialysate circuit ml/h
F, Flow rate into and out of fermentor circuit ml/h
Kp Product inhibition constant mg/ ml
K. Michaelis-Menten saturation constant mg/ml
Pd Product concentration in dialysate circuit mg/ml
P,° Product concentration in fermentor feed mg/ ml
P, Product concentration in fermentor circuit mg/ ml
Pm, Permeability of membrane to product mg/ch-h
Pm. Permeability of membrane to substrate mg/cmz-h
r, Rate of cell growth mg/ml-h
rp Rate of product formation mg/ml-h
—-r. Rate of substrate utilization mg/ml-h
Sd Substrate concentration in dialysate circuit mg/ ml
Sf Substrate concentration in fermentor feed mg/ml
S, Substrate concentration in fermentor circuit mg/ml
t Time h
Vd Volume of liquid in dialysate circuit ml
V, Volume of liquid in fermentor circuit ml
X, Cell-mass concentration in fermentor circuit mg/ ml
a Substrate/cell ratio mg/ mg
[3 Specific maintenance rate h'1
7 Product/substrate ratio mg/ mg
a", Maximum specific growth rate of cells h‘1
T, Cell retention time in fermentor circuit h
mass in the fermentor circuit are as follows: F,S,° + V,
rs
F1810 + Vfrs = FISI + inAm (1) F! + Pm‘Am F! + PmsAm
d8, = _ P’I'iAl'__ S + S 6
"SI—SI) + “'3? F,+P,...A.. I , "
F,P,° + v,r,, = F,P, + Pm,,A,,, 2 + [___Yr ] (1:31
( ) F, + P,,,,A,,. dt
P V dP,
"PI ‘ d) + , a? FIPR + v, r
p
dX F, + Pm,,A,,, F, + PmpAm
_ f
Vfl‘g _ FIX! + Vf _d_t_— (3) _ PmpAm P P 7
— ———————— d + , I )
F, + P,,,,,A,,,
The symbols in these and subsequent equations V1 dpf
are described in Table l. + [m] at—
For the dialysate circuit, the corresponding ’ "”’ "‘
e uations are: V V dX
q —frg=X,+[—f:|——-l (8)
de F, F, dt
PmaAm (Sf - St!) = Fde + Vd E;— (4)
Maximal retention times can be calculated
91:! from the terms within brackets in the above

PmpAm(PI_Pd)=Fde+Vd (5)

(it

The terms Pm,A,,,(S,- 8,1) and PmpAm(P, — Pd)
are based on Fick’s law and describe the diffu-
sion of substrate and product, respectively,
across the dialyzer membrane.

To establish maximal retention times for the
substrate, product, and cells in the fermentor,
equations 1, 2, and 3 can be modified as follows:

 

equations. If no bacteria existed in the fermen-
tor, the differential equations would be linear
and the actual retention times would equal the
maximal retention times. From these times, the
period necessary to reach a new steady state
after the occurrence of a step change in an input
variable can be estimated. From a change or
start-up, the system will have moved 95% to a
new steady state in three retention time periods,

 

728

and the system will reach steady state in a small
fraction of the expected duration of continuous
operation before membrane fouling becomes ex-
cessive. Consequently, the study was restricted
to steady-state behavior, without integrating the
equations for transient behavior.

Calculations of the terms within brackets in
the above equations also show that the cell
retention time is always longer than that of the
substrate or product. The long cell retention
time and short product retention time enable a
high rate of substrate conversion in dialysis con-
tinuous fermentation.

Rate-relationship equations. The term for
the rate of substrate utilization includes terms
for both cell growth and maintenance (5):

—r,, = or, + BX, (9)

The rate of product formation is proportional
to that of substrate utilization (4):

r,,= —yr, (10)

Although substrate limitation was not consid-
ered significant in a previous modeling of (non-
dialysis) continuous whey fermentation (4), the
higher substrate conversion rates resulting from
dialysis may cause substrate concentrations to
occur at limiting levels. Consequently, the term
for cell growth rate includes terms for substrate
limitation as well as product inhibition (3):

s, 1
= 11
"’ “”(K.+s,)(1+P,/K.lx’ ( ’

Generalized model. The above equations for
rate relationships were combined with those for
material balances and the variables were defined
in dimensionless parameters (Table 2) to obtain
a generalized model for dialysis continuous fer-
mentation.

 

COULMAN, STIEBER, AND GERHARDT

APPL. ENvmoN. Mufnomo1..
The resulting equations for the fermentor cir-

cuit. are as follows:

 

[S
if: [—u + HIS,
S,
‘l" (1h 8) (i «1me) ”2’
+—BTMa]X+ W,~+1]/n
9’5- _ —
(fl —[ (1+IuhP,
,,Sr P _
+ [0(171 + g) (1 + pr/Rp) (13)
+ BT,/a] X', + RIIP’, + F,“]/T,
Q9- 6__§l_ __WLAL
d‘ —i (R: + 3‘) (1 + PM?)
no

— 1]X’,/T,

where T]: Vf/Ff.
For the dialysate circuit, the corresponding
equations are:

4% = [—11 + ans, + 41113,] Fd/vd I151
dpd

Generalized steady-state solution. The
time derivatives of the generalized model were
set at zero to obtain a generalized solution for
the steady state.

TABLE 2. Glossary of dimensionless parameters

 

Description

 

Type Symbol and definition
Material parameters Pd- - I’d/78,"
pf = Pf/YS/O
gd = Sci/Sf”
g/ = Sf/Sf”
= (XXI/S,"
Operational parameters R = Pmp/me
#1 = Fr/Fd
Kinetic parameters R..- = Ks/S,"
K11 = Kp/YSI"
0 = “Ir-Tl

Conversion-efficiency parameter

E=l—S—SM¢

Product factor in dialysate circuit
Product factor in fermentor circuit
Substrate factor in dialysate circuit
Substrate factor in fermentor circuit
Cell factor in fermentor circuit

Ratio of product/substrate membrane
permeabilities

Membrane permeability factor

Flow-rate ratio

Michaelis-Menten saturation factor
Product-inhibition factor

Time factor

Fraction of substrate converted to product

 

 

 

M l(.'R()Hl()l..

mentor cir-

esponding

(15)

V4 (16,)

on. The
)del were
ution {or

'embrane

W

V01. 34, 1977

Equations 15 and 16 were solved for the di-
mensionless substrate and _product factors in the
dialysate circuit (S, and Pd) in terms of S, and
P,, respectively:

§1= 1111.87 (17)
E, = «1,7317, (18)

where 7r,= II/(1+ (MI) and 17p= Bil/(1+ oRH).

Equations 12 and 13 were_cor_nbined t_o yield
an equation in terms of S,, P,, 8.1, and Pd. The
further substitution of equations 17 and 18 re-
sulted in an expression only in terms of S, and
P7. Equation 14 was used to obtain another
expression in terms of S, and P, These two
expressions were combined to obtain one (quad-
ratic) equation for S, and another for P,:

 

S,‘~’ + [K
3111:1de 2):; : Pr" -
1 + 773 ]St (19)
- K[I?,.(1_:‘Jul+1 _+_ P10] = 0
1 + 11',‘
- _ _ #51 ._ _
Pf - K]: [01?” + S—f 1] (20)

Equation 19 has two real and distinct roots, one
positive and one negative. The positive root is
the only physically valid solution.

Equation 14 was used to eliminate the nonlin-
ear term in equation 12, and the resulting expres-
sion was solved for X,as follows:

if = L..Sf + n _(_§,‘; 5’) (21)
l + BT,/a

Equations 17 to 21 comprise a generalized
steady-state solution for substrate, product, and
cells in the fermentor and the dialysate circuits
of the system. This solution was found to be
stable locally.

COMPUTER SIMULATION

The generalized mathematical model was
used to simulate the dialysis continuous process
for the fermentation of whey lactose into am-
monium lactate. The model was programmed
on a digital computer. Because the simulations
were performed before actual experimental fer-
mentations were conducted, approximations of
various conditions were used as shown in Table
3, unless otherwise stated.

The dimensionless parameters that were ma-
nipulated in the simulations are included in Ta-
ble 2. The time factor (0), the flow-rate ratio
(It), and the membrane permeability factor (II)
each can be adjusted experimentally. The time

 

MODEL AND SIMULATION OF DIALYSIS FERMENTATION 729

factor (0) was selected as the primary dimen-
sionless variable because it included the impor-
tant design parameter V,/F, and eliminated the
need for specific values of pm. The permeability
ratio (R) characterizes the molecular sensitivity
of the membrane. The Michaelis-Menten satu—
ration factor (K,) and the product-inhibition
factor (KP) are the only parameters that are
characteristic of a specific system, i.e., the am-
monium-lactate fermentation of whey, but are
generally relative to feed concentration. There-
fore, the model could be applied to other systems
without difficulty.

The simulations were made first with the two
main experimental operating parameters, cell
retention time in the fermentor circuit (T,) and
flow rate into and out of the dialysate circuit
(Fd). To obtain values for T, from the time factor
(0), the maximum specific growth rate (pm) was
set at 0.27 h”1 (see Fig. 5 in reference 4). Figure
2 shows the simulated effects of changes in T,
on the concentration of substrate in the fermen-

TABLE 3. Values used in computer simulations

 

 

__ Symbol" Value
R, 0.2
K. 0.001
P," (mg/ml) 0

R 3.0
Sfo (mg/n11) 230

y 0.9
II 10

 

" Symbols are described in Tables 1 and 2.

 

 

 

 

 

 

 

I A
U .‘I to If) .11

'I‘, (b)

FIG. 2. Simulated effects of changes in cell reten-
tion time (1“,) on substrate concentrations in the fer-
mentor circuit (S,) and the dialysate circuit (8,) at
constant flow rates, when it = 0.25 and II = 0.5 75.

 

 

730 COULMAN, STIEBER, AND GERHARDT

tor circuit (3,) and the dialysate circuit (S,,) at
constant flow rates. Below a critical threshold
value of T,, the values of S, and 8,, remain
constant because of cell washout. Above the
threshold, S, and 8,, decrease as T, increases. In
each circuit, a plateau region is reached at which
only marginal improvements occur with further
increases in T,. The simulations predicted that
satisfactorily low levels of residual substrate
(<5 mg/ml) could be attained by use of a rea-
sonably short retention time in the fermentor
circuit (<20 b). As expected, the general shape
of the curves was the same as that for the
nondialysis continuous process (see Fig. 3 in
reference 4).

Figure 3 shows the simulated effects of
changes in the dialysate flow rate (Fd) on the
concentration of substrate in the fermentor cir-
cuit (S,) and the dialysate circuit (8,) at con-
stant cell retention time. Both of the dependent
variables decrease as F d increases, and satisfac-
torily low levels of residual substrate apparently
could be attained in both circuits with a reason-
ably low flow rate in the dialysate circuit. In
both of the simulations, the concentration of
residual substrate (rather than the concentra-
tion of accumulated product) was used as a
measure of fermentation conversion efficiency
because of greater accuracy and precision in the
experimental analysis for lactose than for lac-

140 r , r

 

120 1'

ICC

 

 

S, or 3d (mg/ml)

 

 

L
‘3 33-3 600

Fd (ml/h)

kL'
(. 1
('5

FIG. 3. Simulated effects of changes in dialysate
flow rate (F4) on substrate concentrations in the
fermentor circuit (3,) and the dialysate circuit (Sd)
at constant cell retention time (T, =- 13 h), when II =-
0.575.

 

 

 

 

Al’l’l.. EM'moN. Micmmum.
ID a 1 I
08» _
e pf ‘
(3.6» _ .
In.13 pd
5
Id13 04~ ,
Kl bd/q:
C2» .
C1,, ‘ ‘ ‘
o 0.5 1.0 1.5 2O

96

FIG. 4. Simulated effects of changes in the flow-
rate rati(_) «Al on the product factor in the fermentor
circuit (P,) and the dialysate circuit (P,,) and on the
dialysate product yield (P,,/(p), when E = 0.995.

tate. Furthermore, substrate use is more appli-
cable in a generalized model because there are
often several products, but there is usually only
one substrate.

Generalized simulations then were made to
illustrate the interrelationships of the dimen-
sionless parameters. Figure 4 shows the simu-
lated effects of changes in the flow-rate ratio
((1)) on _the product factors in the fermentor cir-
cuit (P,) and the dialysate circuit (P,,) and on
the dialysate product yield (P,,/(b), which rep-
resents the fraction of substrate remaining as
p_roduct_ in the dialysate; effluent. As ¢ increases,
P, and P,, increase and P,,/(p decreases. However,
to achieve a preselected level of efficiency (E =
0.995), the time factor (0) must be adjusted for
changes in d) (Fig. 5); i.e., '1‘, must be increased
with increasing changes in (p to maintain a low
level of substrate in both circuits. In the am-
monium-lactate fermentation, it is desirable to
have high values of P,, P,,, and P,,/o. Since 0
varies with d», a compromise probably would be
made considering cost of the facility and relative
marketability of the two effluents.

Figure 5 also shows the effect of changes in
the product-inhibition factor (K ,,) on the dialysis
continuous fermentation. The smaller the value
of If,” the more effective dialysis becomes.

The effect of changes in the membrane perme-
ability factor (II) on the product factor in the
fermentor circuit (P,) and the dialysate circuit
(P,,) at three permeability ratios (R) is shown
in Fig. 6. Each of the curves reaches a plateau
region where a further increase in II does little
to decrease P, or increase P,,. A high II is needed

 

 

(1:)

flow-
,entor
0n the

l.

appu-
Heine
lyomy

made to
dimen-
le simu-
Ite ratio
ntor cir-
) and on
nich rei)’
aining as
Increases.
However,
ancy (E =
ljusted lOT
incre
min a ‘0‘"
[n the am-
lesirabie t0
1/0- Since 0
I\\' would be
and relative

.I changes ‘9
n the diaIySIS
ller the “he
Iecomes-
Ibrane perme'
factor in the
alvsate circuli-
‘. (R) is Show“
I‘heS a 9mg"
in II does 11ttle
3gb ll isneed

l A.

VOL. 34, 1977

7 I r I

 

 

 

 

2 I I I
O 05 IO 1.5 2.0

¢

FIG. 5. Simulated effects of changes in the flow-
rate ratio (a) on the time factor (0) at three values
of the product-inhibition factor (K), when E = 0.995.

 

'0 A ‘1“ 1 m r #r—mr—r

 

 

 

FIG. 6. Simulated effects of changes in the mem-
brane permeability factor (II) on the product factor
in the fermentor circuit (P,) and the dialysate circuit
(P,,) at three permeability ratios (R), when ¢ = 1.0
and E =f 0.995.

only when R is low, and a II value greater than
2.0 probably would prove unnecessary in the
ammonium-lactate fermentation.

The effect of changes in the time factor (0)
on the conversion efficiency parameter (E) at
three membrane permeability factors (II) is
shown in Fig. 7. As 0 increases, E increases until
a plateau region is reached. The relationship
illustrates the tradeoff between the two param-
eters in designing the system.

MODEL AND SIMULATION OF DIALYSIS FERMENTATION

 

 

 

 

731

IO T

0.8»
06 J

E

04
0.2.

0I 2 T '5‘ “—5'i5 cad—‘7

FIG. 7. Simulated effects of changes in the time
factor (0) on the conversion efficiency parameter (E)

at three membrane permeability factors (II), when (1’
-= 1.0.

DISCUSSION

Mathematical modeling and computer simu-
lation provide powerful tools for predicting the
results to be expected in a fermentation process.
Laboratory experiments then need be conducted
only to validate the predictions by using a rela-
tively limited number of changes at preselected
critical points. The experimental results in turn
are used to establish constants and perhaps to
indicate additional terms in the mathematical
equations. By this process of successive theoret-
ical prediction and experimental validation, the
model becomes increasingly accurate and useful
to predict how the fermentation should be con-
ducted for accomplishing objectives.

The foregoing generalized model and specific
simulations clearly predict that a dialysis contin-
uous process for the ammonium-lactate fermen-
tation of whey can be operated efficiently, ena-
bling the rapid and almost complete conversion
of a high concentration of substrate into high
concentrations of product and cells in one ef-
fluent and an adequate concentration of purified
product in another effluent. The stirnulations
also project the effects of changes in various
parameters on the process and the regions for
experimental validation.

Although generalized, the mathematical model
was developed within certain specifications re-
lated to the steady-state conversion of substrate
primarily into a product that exerts feedback
inhibition. Transient as well as steady-state so-
lutions sometimes are obtained from a model
by integration, but this consideration was un-
necessary for the present fermentation.

Cell yield rather than product yield sometimes
is more important than in the present fermen-

 

732 COULMAN, STIEBER, AND GERHARDT

tation. The generalized expression for cells in
the fermentor circuit (equation 21) could be
simulated but would require establishing values
for a and ,6, which depend on the organism used.
In equation 21, the cell population term (X,)
increases as the term BT,/a decreases. Thus, in
situations where substrate utilization is more a
function of growth ((1) than of maintenance (,8),
the cell population would increase with retention
time (T,). Beyond a critical retention time, how-
ever, maintenance supersedes growth and then
the cell population would actually decrease. In
a previous mathematical model of dialysis con-
tinuous culture, X, was predicted to increase
with T, (i.e., with decreasing dilution rate); how-
ever, the theory did not consider values of T, >
10 h (see p. 21—25 in reference 8). Such consid-
erations and simulations of cell yield would be
useful if the conversion of whey were designed
to produce single-cell protein rather than am-
monium lactate as the primary product or if the
process were modeled to produce dairy starter
cultures. The generalized nature of the model
enables such alternative applications.

ACKNOWLEDGMENTS
This work was supported by a grant from the Office of
Research Development of Michigan State University and by
grant ENG 76-17260 from the National Science Foundation.

AWL. ENYlltUN. Mlt'lttflilnl"

LITERATURE CITED

. Friedman, M. R., and E. L. Gaden, Jr. 1970. Growth

and acid production by Lactobacillus- (1..) dclbruechii
in a dialysis culture system. Biotechnol. Bioeng.
12:961—974.

‘. Gerhardt. P.. and D. M. Gallup. I963. Dialysis flask

for concentrated culture of microorganisms. J. Bacte-
riol. 86:919-929.

. Humphrey. A. E. 1972. The kinetics of biosystems: a

review. Adv. Chem. Sen. no. 109. p. 630-650.

. Keller, A. K., and P. Gerhardt. 1975. Continuous lactic

acid fermentation of whey to produce a ruminant feed
supplement high in crude protein. Biotechnol. Bioeng.
17:997—ll'118.

. Marr. A. G., E. H. Nilson, and D. J. Clark. 1963.

Maintenance requirement of E. coli. Ann. N.Y. Acad.
Sci. 102:536—548.

. Poznanski. S., K. Kornacki, Z. Smietana, J. Rymasz-

ewslu', A. Surazynski, and W. Chojnowski. 1973.
Ny metode til fremstilling af maelkesyre ved hjaelp af
ionbytter. Ein neues Verfahren zur Milchsauregewim
nung unter Verwendung von lonenaustauschern. Nord.
Mejeri-Tidssk. 39: 167-170.

. Reddy, C. A., H. E. Henderson, and M. D. Erdman.

1976. Bacterial fermentation of cheese whey for produc-
tion of a ruminant feed supplement rich in crude pro-
tein. Appl. Environ. Microbiol. 32:709—776.

. Schultz, J. S., and P. Gerhardt. 1969. Dialysis culture

of microorganisms: design. theory. and results. Bacte-
riol. Rev. 33:1-47.

. Stieber, R. W., G. A. Coulman, and P. Gerhardt. 1977.

Dialysis continuous process for ammonium-lactate fer-
mentation of whey: experimental tests. Appl. Environ.
Microbiol. 34:733-739.

!‘ ~..J

UC

man.
Kim'-
1 pro-

Jlture.
5acte-
. 1977.
Its fer-

73

ARTICLE II

DIALYSIS CONTINUOUS PROCESS FOR

AMMONIUM-LACTATE FERMENTATION OF WHEY:

EXPERIMENTAL TESTS

BY

R. W. Stieber, G. A. Coulman, and Philipp Gerhardt

Reprinted from Appl. Environ. Microbiol. 335733-739, 1977.

AWLu-zn AND ENVIRONMENTAL Mrcaomouxw, Dec. 1977, p. 733—739

Copyright © 1977 American Society for Microbiology

Vol. 34. NO. 6
Printed in U. SA.

Dialysis Continuous Process for Ammonium-Lactate

Fermentation Of Whey: Experimental TestsT

R. W. STIEBER,l G. A. COULMANHT AND PHILIPP GERHARDT“

Departments of Microbiology and Public Health' and of Chemical Eng-gineering,2 Michigan State University,

East Lansing, Michigan 48824

Received for publication 8 April 1977

Laboratory experiments were conducted to validate theoretical predictions
describing a dialysis continuous process for the fermentation of whey lactose to
ammonium lactate, in which the fermentor contents are poised at a constant pH
by adding ammonia solution and dialyzed through a membrane against. water.
Dried sweet—cheese whey was rehydrated tO contain 230 mg of lactose per ml,
supplemented with 8 mg Of yeast extract per ml, charged into a 5-liter fermentor
without sterilization, adjusted in pH (5.3) and temperature (44°C), and inoculated
with Lactobacillus bulgaricus. The fermentor and dialysate circuits were con-
nected, and steady-state conditions were established. A series of such conditions
was managed nonaseptically for 94 days to study the process and to demonstrate
efficiency and productivity. As time progressed, the fermentation remained hom-
ofermentative and increased in conversion efficiency, although membrane fouling
necessitated dialyzer cleaning about every 4 weeks. With a retention time of 19
h, 97% of the substrate was converted into products. Relative to nondialysis
continuous or batch processes for the fermentation, the dialysis continuous
process enabled the use of more concentrated substrate, was more efficient in
the rate Of substrate conversion, and additionally produced a second effluent of
less concentrated but purer ammonium lactate.

 

In a companion article (3), we have provided
background, rationale, mathematical modeling,
and computer simulation of a dialysis continuous
process for the ammonium-lactate fermentation
Of whey. In this article, we report on experimen-
tal tests of the simulation—predicted modes of
operation, studies of the kinetics and parame-
ters, validation Of the mathematical model, and
demonstration Of the greatly increased efficiency
obtainable by this process as compared with
nondialysis batch and continuous processes for
the fermentation (5, 8).

MATERIALS AND METHODS

Inoculum. Lactobacillus bulgaricus 2217 (Chris
Hanson's Laboratory, Milwaukee, Wis), the inoculum
for the fermentation. was selected (8) on the basis of
its high rate Of acid production in the pH range Of 5.0
to 6.0. The culture was grown at 44°C for 8 to 12 h in
a sterile medium of 10% (wt/vol) skim milk powder
and 90% tap water, stored at 4°C, and transferred
weekly to fresh medium.

Substrate. Dried sweet-cheese whey (Galloway
West Co., Fond du Lac, Wis), used as the fermentation
substrate, was rehydrated to contain 230 mg Of lactose

(Journal article 7985 from the Michigan Agricultural Ex-
periment Station.

fi Present address: Department of Chemical Engineering,
Cleveland State University, Cleveland, OH 44115.

per ml and was supplemented with 8 mg of yeast
extract per ml. The reconstituted whey was made up
in 7.5—1iter batches without sterilization and was held
in a stirred, chilled (6°C) reservoir during use.

Dialysis continuous fermentation system. Fig-
ure 1 shows a diagram of the experimental dialysis
continuous fermentation system. The fermentation
was conducted continuously by metering the whey
from the feed reservoir into the fermentor and remov-
ing overflow into the product reservoir. thus maintain-
ing a constant volume in the fermentor. The fermentor
contents were mixed thoroughly by stirring and baf-
fling. Nitrogen was continuously purged into the fer-
mentor to force overflow into the reservoir. Concen-
trated (58%) NILOH solution was automatically fed
into the fermentor to maintain a constant pH. Con-
currently, the fermentation was dialyzed continuously
by circulating the fermentor contents through the
dialyzer. Solutes diffused through membranes of the
dialyzer to a circulating stream of dialysate, into which
tap water was metered. The dialysate effluent. was
removed into another product reservoir.

The fermentation was conducted in a 5-liter modu-
lar fermentor with automatic temperature control
(Microferm model MF 105, New Brunswick Scientific
Co., New Brunswick, NJ.) and automatic pH control
(model pH-22, New Brunswick Scientific). Contents
of the fermentor circuit (1.4 to 4.0 liters) were main-
tained at a temperature of 44°C :1: 05°C and at pH
5.3 :t 0.05. The plate-and-frame dialyzer (see Fig. 5 in
reference 9) consisted of two stainless-steel plates

733

 

734

bolted to compress a series of silicone-rubber gaskets,
sheet stainless-steel frames, molded silicone-rubber
separators, and regenerated cellulose membranes cut
from Visking tubing (25-cm flat-width, Union Carbide
Corp., New York, N .Y.). Eight membranes, providing
2,300 cm2 of effective area, were used in the dialyzer.
Peristaltic tubing pumps (models T8 and T68, Sig-
mamotor, Inc., Middleport, N.Y.) and polyurethane
tubing were used to meter flow rates of the whey to
the fermentor and of the water to the dialyzer. Maisch
gear pumps (model HQDCC, Tuthill Pump Co., Chi-
cago, Ill.) were used to circulate the fermentor con-
tents and water dialysate through the dialyzer. Flow
rates were determined twice daily for all inputs and
outputs of the system. Operation of the system occa-
sionally was interrupted for 5 to 10 h for cleaning of
the dialyzer or fermentor or for installation of a new
component in the system. After 76 days the operation
was suspended for 6 days and then was resumed for
18 more days. During inoperative periods, the fermen-
tor-circuit contents were stored at 4°C.

Samples of the dialysate and fermentor circuits
were taken at 12-h intervals from the dialysate effluent
and from a glass “'1‘” inserted in the tubing between
the fermentor and dialyzer, respectively. Steady-state

$.31 r""‘°i .
l l ii) i
.l.

termemor »
Of I t fermerflor dialyza
reservor

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

feed, dial sale wotw
7mm! pfdmi fese'VON’
reserve-t

FIG. 1. Diagram of the experimental dialysis con-
tinuous fermentation system.

STIEBER. COULMAN, AND GERHARDT

APPL. ENV’lRON. MICROBIOL.

data were determined from samples taken at five times
the cell retention time or at 48 h after changing a
parameter. Lactose in the samples was determined by
the colorimetric method of Dubois et al. (4) as modi-
fied by Montgomery (6). Lactic acid was determined
by a gas chromatographic procedure (1). The concen-
tration of residual lactose rather than lactic acid was
preferred as a measure of fermentation conversion
efficiency because of better accuracy and precision.

The symbols used in this article have been defined
in the companion paper (3).

RESULTS

Effect of fermentor cell retention time on
conversion efficiency. The results of experi-
mental variation in the cell retention time (T,)
affecting the residual lactose in the fermentor
circuit (8,) and the dialysate circuit ( Sd) showed
that the whey was readily fermented below the
desired level of 5 mg/ml (Fig. 2), as predicted
by the computer simulation (see Fig. 2 in refer-
ence 3). The results also showed the effect of
changes in T, on the process during two different
periods, in each of which a point was reached
where an increase in T, did little to decrease S,
and 8,}.

Culture adaptation. The conversion effi-
ciency increased substantially as time pro-
gressed (Fig. 2), i.e., the culture at days 50 to
75 was able to reduce the residual lactose in the
fermentor to the desired level with a shorter
retention time (T, = 18 h) than did the culture
at days 19 to 24 (T,= 22 h). A more efficient

bacterial population evolves also during pro-

 

OAYS I9“ 24

S, or 5d (mo/ml)
8

$3

 

 

 

T T

DAYS 50 - 75

 

 

 

T, (h)

T, (ll)

FIG. 2. Effects of changes in cell retention time (T,) on residual lactose in the fermentor circuit (8,) and
the dialysate circuit (3,.) during two periods of dialysis continuous fermentation. T, was adjusted by changing
V,. F, and E, were maintained essentially constant (¢ = 0.27 for days 19 to 24 and 0.29 for days 50 to 75).
The objective was to attain <5 mg of residual lactose per ml, as indicated by the dashed abscissa.
Interpretations of the curves were based on computer simulations (see Fig. 2 in reference 3).

 

 

Micki iBioL,

it five times
changing a
emined by
4) as modj-
detemnned
"he concen-
.ic acid was
conversion
)recision.

een defined

[1 time on
of experi-
1 time (Tl,-)
femientor
{,1 showed
below the
predicted
‘2 in refer-
? effect of
0 different
[5 reached
ecrease S,

rsion effi-
time pr0-
lays 50 t0
:ose in the
a shorter
.18 culture
63 efficient
iring pro-

! (SJ 0""
chunflmg
50 to 75)-
absrlb'sa'

 

 

 

Vot... 34. 1977

longed nondialysis continuous fermentation of
whey, as reported by Keller and Gerhardt (5).
In neither study was it determined whether the
change resulted from adaptation, mutation, re-
placement by a contaminant, or symbiosis with
a contaminant. Whatever the change, it im-
proved the fermentation.

This culture change can be shown in kinetic
terms. Assume that the fermentor product con-
centration (P,) is small in relation to the product
inhibition constant KP (i.e., P,-/K,, = 0), and
recall that the maximal specific growth rate (pm)
is the reciprocal of T, at steady-state conditions.
Then the rate-of-growth equation used to de-
scribe bacterial growth in the mathematical
model (equation 11 in reference 3) can be rear-
ranged and a Lineweaver-Burk plot can be con-
structed, with realization that the saturation
constant (K..) so determined reflects both sub-
strate limitation and product inhibition. The
results (Fig. 3) indicated that ,u,,, increased from
0.18 h”1 in the earlier period to 0.27 h"1 in the
later period. The K, (20.8 mg/ml), determined
from Fig. 3, was the same before and after the
adaptation, indicating noncompetitive inhibition
kinetics. Although the maximal concentration
of the product tolerated by the culture did not
change, the culture was able to grow and metab-
olize at a faster rate after adaptation than before.

Effect of dialysate flow rate on conver-
sion efficiency. Figure 4 shows the results of

24 I T T 1 T Y

 

20- DAYS I9-24
Pm = '8

 

 

 

908 -.04 6 0‘4 .68 i2 i6 .20
Sf" (ml/mg)

FIG. 3. Lineweaver-Burh plot of the data in Fig.
2. The maximal specific growth rates of the culture
()1...) were determined from the slopes of the regres-
sions. The Michaelis-Menten saturation constant
was determined from the abscissa intercept common
to both regressions (K, = 20.8 mg/ml).

 

 

EXPERIMENTAL TESTS OF DIALYSIS FERMENTATION 735

:40 ~ ———--
V—_—_fi. —T ‘r
l\
11 ‘
l\
120‘

1
1

 

‘ |
i l
ICC \
\
. . (
E \\
o at. \
E ’1‘ i
\
mu :1 \
L 60)‘ \
‘1 '1 \
a) , \ \
\ \
4o \ \
\Sd \St
= \ s
o, \
(.0 \\ O\O\
\ \
-.‘_ A _____
o 0 603 cm
Fd (ml/h)

FIG. 4. Effects of changes in dialysate flow rate
(F4) on residual lactose in the fermentor circuit (8,)
and the dialysate circuit (8,,) during days 40 to 63,
with cell retention time held constant (T, = 16.4 h).
The objective was to attain <5 mg of residual lactose
per ml, as indicated by the dashed abscissa. Inter-
pretations of the curves were based on computer
simulations (see Fig. 3 in reference 3).

experimental variation in the flow rate of water
through the dialysate circuit (F d) affecting the
residual lactose in the fermentor circuit (8,) and
dialysate circuit (S,,), with the cell retention
time held constant. The results confirmed the
computer-simulated prediction (see Fig. 3 in ref-
erence 3). The desired 5-mg/ ml level of S, was
attained when Fd was greater than 800 ml of
water per h.

Figure 5 shows the concentration of accumu-
lated lactate in the fermentor circuit. (P,) as a
function of the dialysate flow rate (F,,) and of
the flow-rate ratio ¢(F,/F,,). As F,, increased
and 4) decreased (i.e., as the rate of dialysis
increased), P, first. increased to a maximum (F ,,
= 370 ml/h, qb = 0.68) and then decreased be-
cause of dilution by water from the dialysate
circuit. At these levels of F,, and (b, the culture
withstood a much greater concentration of lac—
tate (P, = 116 mg/ ml) than that without. dialysis
(i.e., P,= 82 mg/ml when F,, = 0 ml/h).

Figure 6 shows the residual lactose concentra-
tion (8,) correlated with the accumulated lactate
concentration in the fermentor circuit (P,), with
constant cell retention time. The desired 5-mg/
ml level of S,was attained when P, was less than
60 mg/ml.

Figure 7 shows the effect of the dialysate flow
rate (Fd) on the sum of the lactate and lactose

 

736 S'I‘IEBER, COULMAN, AND GERHARDT

 

120 1 t r

P, (mg/ml)

 

 

 

50o 360 600 900

 

120 r I r

 

..o. l

P, (mg/ml)
A 8
“\7_

 

 

 

80 >- .1
70, -
60F .
50% foe (5.6 0i4 0.2

¢ (Ff / Fd )
FIG. 5. Effects of changes in dialysate flow rate
(F .1) and flow-rate ratio (a) on accumulation of lactate

in the fermentor circuit (P,), with cell retention time
held constant (T, = 16.4 h).

concentrations in the fermentor circuit (P,+ 8,),
the points for which correspond to those in Fig.
6. The sum decreased proportionally as Fd in-
creased. Diffusion of the solutes across the mem-
brane accounted for some of the decrease. How-
ever, net osmosis of water from the dialysate
into the fermentor circuit accounted for most
of the decrease, as indicated by an increase in
flow rate of the fermentor effluent relative to
that of the fermentor feed. As a result of the
increased osmosis with increased dialysate flow
rate, the concentrations of lactose and lactate
both decreased. The former decrease is desirable
and the latter is not. Consequently, a trade off
in the regulation of R, is required in designing
the system.

Dialyzer dependability. The preceding ex-
perimental results were all obtained during op-
eration of the system for 76 days. A week later,
operation was resumed for an additional 18 days.
Although the dialysate compartments of the di-
alyzer stayed clean throughout the total 94 days
of operation, the fermentor compartments be-
came fouled and needed cleaning after about 4

 

APPL. ENVIRON. MIcIioaIOI..

weeks of operation. However, the same mem-
branes were used throughout.

Product quality. Samples from the fermen-
tor circuit were regularly analyzed by gas chro-
matography not only to determine the concen-

 

l40» AY

8, (mg/ml)

 

 

 

P, (mg/ml)

FIG. 6. Effects of changes in lactate accumulation
(P,) on residual lactose in the fermentor circuit (8,),
with cell retention time held constant (’1‘, = 16.4 h).
The objective was to attain <5 mg of residual lactose
per ml, as indicated by the dashed abscissa.

240

 

 

40»- 4

 

 

 

o 360 660 900
FIG. 7. Effects of changes in dialysate flow rate
(F d) on the sum of the lactate and lactose concentra-

tions in the fermentor circuit (P, + 8,), with cell
retention time held constant (T, = I 6.4 h).

l”(r‘m.-.‘~m

ulation
uit (SJ.

16.4

hi.

I lactose

isalt’ flou‘ role
N. mncentra

it
. S.)
4 h).

with C9

VOL. 34, 1977

tration of lactic acid but also to monitor the
possible presence of atypical metabolic products.
The results indicated that the fermentation re-
mained homofermentative, i.e., only negligible
amounts of products other than lactic acid were
found.

Operation at nearly optimal conditions.
A dialysis continuous fermentation next was op-
erated at nearly optimal conditions (T,= 19 h,
4) = 0.4, S, < 5 mg/ml) to demonstrate the
efficiency and productivity of the process. Due
to inadvertent changes in speeds of the metering
pumps controlling flow rates into the fermentor
and dialysate circuits, fluctuations occurred in
the concentration of residual lactose during the
18 days of operation (days 77 to 94). During this
period, a mean of 35% of the fermentor effluent
represented water that osmosed from the dialy-
sate circuit, i.e., a mean 77 ml of water per h
entered the fermentor from the dialysate circuit
(Table 1).

From material-balance data for operating pe-
riods when the parametric conditions were
nearly optimal (Table l), the equivalent sub-
strate concentration in the fermentor-feed
stream (8,") was calculated by using a conser-
vation-of-mass equation. The mean calculated
value (223 mg/ml) agreed well with the mean
experimental value (230 mg/ml) for 8,”. With
allowance for substrate incorporated into bac—
terial cells, the results confirmed that no signif-
icant portion of the substrate was lost to prod-
ucts other than lactic acid.

Validation of the mathematical model.
The experimental results were used to validate
the mathematical model for the fermentation
(3). The values used for the computer simula-
tions are listed in Table 2. The values for qb and
II were calculated directly from the experimen-
tal results. By successive approximations, the
values for um and K, (determined in Fig. 3) and
an exaggerated KP proved best for correlating
the simulations with the experimental data. The

EXPERIMENTAL TESTS OF DIALYSIS FERMENTATION 737

high value for K, eliminates the effect of the
product inhibition term in the mathematical
model. Consequently, the Michaelis-Menten sat-
uration constant (K ,) reflects not only substrate
limitation but also product inhibition, as shown
also by Contois (2). The dilution of the fermen-
tor contents caused by the net osmotic transfer
of water from the dialysate circuit was taken
into account by assuming that the water instead
entered the fermentor with the feed stream.
Accordingly, since the dilution of the fermentor
contents averaged 35%, S,” was taken as 150
instead of 230 mg/ ml.

Figure 8 shows the computer-simulated curves
obtained by using the mathematical model (3)
and the values in Table 2 to describe the corre-
lation between one important operating condi-
tion and the conversion efficiency, e.g., T, versus
S, and S,,. The curves fit closely with the super-
imposed points of experimental results and
thereby demonstrated the validity of the math-
ematical model.

DISCUSSION

The results obtained with the dialysis contin-
uous process at nearly optimal conditions can
be compared with the best results obtained with
the nondialysis continuous and batch processes
for the fermentation (5, 8; Table 3). Relative to
the other two processes and with similar conver-
sion efficiency, the present process enabled the
use of more than three times the concentration
of substrate at a similar retention time and thus
was more efficient in the rate of substrate con-
version; it was similarly efficient in the percent-

TABLE 2. Values used for computer simulation of

 

 

fermentation
K, K, s," ,1.
Days (mg/ml) (mg/ml) (mg/ml) (h"‘) n 4’
19—24 1,000 20.8 150 0.18 0.665 0.27
50-76 1,000 20.8 150 0.27 0.615 0.29

 

TABLE 1. M aterial- balance data for operation of the dialysis continuous fermentation system at nearly
optimal conditions

 

 

 

Sf 8,: P P.) F". . 8”.
Day Tr (h) ‘3 (mg/ml) (mg/ml) (mg/ml) (mg/ml) (ml/h) F1 (ml/h) F" (ml/h) (mg/ml)

79 20.9 0.40 2.3 0.4 65 25 110 189 478 217
81 20.4 0.36 2.0 0.3 60 22 124 193 531 183
85 18.5 0.40 2.0 0.5 79 25 131 214 527 226
86 18.8 0.39 2.3 0.4 72 24 135 210 533 203
89.5 16.9 0.38 3.3 0.4 94 ‘24 150 234 613 243
93.5 18.5 0.33 3.5 0.2 110 23 142 214 640 267
Mean 19.0 0.38 2.6 0.4 80 24 132 209 554 223

 

' F)", Flow rate of feed into the fermentor circuit. F, - F ,° represents the net osmotic transfer of water from

the dialysate to the fermentor circuit.

" Calculated from the formula: S,0 = (F ,S, + F ,P, + R18.) + dedl/Ff" - P,", where P," = 9 mg/ml.

 

 

738

STIEBER, COULMAN, AND GERHARDT

APPL. ENVIRON. MICROBIOL.

 

S, or 84 (mg/ml)

 

 

10 15
1,0.)

     

 

1 v v r

DAYS 50—75

 

 

 

 

15 20

)0
1}(h)
FIG. 8. Computer-simulated effects of cell retention time (T,) on residual lactose in the fermentor circuit
(8,) and the dialysate circuit (Sd) during two periods of dialysis continuous fermentation. The curves were
plotted by use of the mathematical model (3) and the values in Table 2. The points were replotted from Fig.
2 to demonstrate the close fit between the experimental results and the computer simulations. These can be
compared with the preliminary simulations (see Fig. 2 in reference 3).

TABLE 3. Comparison of results obtained at nearly optimal conditions with three processes for the

 

 

fermentation
Lactose in Cell reten- Rate of [ac Conversion Lactose in Lactate in Lactose in Lactate in
Process whey sub- tion time tose conver- of lactose femientor fermentor dialysate dialysate
strate (h) Sion Into prod- product product product product
(mg/ml) (mg/ml-h) ucts (‘5) (mg/ml) (mg/ml) (mg/ml) (mg/ml)
Nondialysis
Batch (refer- 7O 16" 4.3 98 1 .0 75
ence 8)
'I‘wo-stage con- 50 31 1.6 98 1.0 55
tinuous (ref-
erence 5)
Dialysis
Continuous 230 19 11.7 97 2.6 80 0.4 24

 

° Required fermentation time for the batch process.

age conversion of substrate into products and
in the low concentration of residual substrate
and the high concentration of accumulated prod-
uct in the fermentor; and it additionally yielded
a second effluent containing less concentrated
but purer ammonium lactate.

A dialysis culture is able to tolerate more
lactic acid and thus ferment more lactose than
a nondialysis culture (Fig. 5). The metabolic
activity of the culture in the ammonium lactate
fermentation apparently is inhibited mainly by
the concentration of undissociated lactic acid
(see discussion in reference 5), which in turn is
dependent on the pH, temperature, and concen-
tration of salts of the acid. Since the dialysis
membrane is more permeable to the ammonium
than the lactate ions (results not shown), dialysis

may shift the equilibrium from the undissociated
to the dissociated acid by removing the cations
to a greater extent than the anions from the
fermentingculture. This may explain why the
culture tolerated more acid under dialysis than
nondialysis conditions.

The substrate for the dialysis continuous fer-
mentation was greatly concentrated. The cost
of concentrating may be offset by lesser trans-
portation costs, stabilization of the whey against
spoilage, smaller holding capacity needed for
operation of the process, and the eventual need
to concentrate the products for use as feedstuff.

The use of further concentrated substrate fed
into the system might be accomplished by use
of a two-stage fermentation, for the same reasons
that this proved more effective in the nondialysis

 

 

ircuit
: (1‘07?
11 Fig.
can be

'he

___/
mate in
inhsme
pfudUCI
rug 'Illll

 

VOL. 34, 1977

continuous process (5). The present system
would need to be modified only by adding a
continuous prefermentor, without dialysis. The
conversion in this prefermentor would be rela-
tively less affected by the greater concentration
of substrate. However, since contamination is
controlled mainly by a high concentration of
undissociated lactic acid, a lower pH would need
to be maintained in the prefermentor than in
the dialyzed fermentor. Such a two-stage contin-
uous process also might be effective if it were
desirable to accomplish more complete conver-
sion of lactose to lactic acid.

The cell retention time (T,) required to fer-
ment the whey in the dialysis continuous proc-
ess might be reduced by decreasing the value
for the flow-rate ratio ((1)). However, this manip-
ulation would cause the product solids in the
effluents to become more dilute. By increasing
(1) and the permeability factor (II), the dialysate
product concentration could be increased con-
siderably (perhaps doubled) without decreasing
the rate of the fermentation, as predicted by
computer simulation (results not shown). How-
ever, this improvement was not validated exper-
imentally.

The dialysis continuous fermentation yielded
two product streams. The fermentor effluent
containing ammonium lactate, bacterial cells,
and whey proteins may be used as ruminant
feedstuff. The dialysate effluent containing am-
monium lactate and little else may be used as
feedstuff or may be converted into industrial
lactic acid and ammonium sulfate, for which it
is important that the lactate is relatively pure
and highly concentrated. A reasonable goal for
both effluents would be concentrations of at
least 50 mg of lactate per ml, which is similar
to that produced by the nondialysis batch and
continuous processes. A low residual lactose con-
centration (<5 mg/ ml) is also necessary because
lactose may crystallize when storing the
product and form a troublesome sludge. These
product goals appear readily attainable.

A side effect of the dialysis continuous process
was the large osmotic influx of water from the
dialysate into the fermentor, diluting its con-
tents. A more permeable membrane might
lessen the problem by allowing greater diffusion
of solute and thus reducing the differences in
solute concentration on each side of the mem-
brane. Greater membrane area also might help
if turbulence were maintained at the membrane
surfaces. Calculations of the osmotic pressure
due to solutes indicated that the net pressure
resulting from the accumulated lactate is greater
by a factor of at least 50 than that resulting
from the residual lactose. To alleviate the prob-
lem, membranes with greater permeability to

 

EXPERIMENTAL TESTS OF DIALYSIS FERMENTATION 739

lactate might be selected or a greater atmos-
pheric pressure might be maintained on the
fermentor circuit than on the dialysate circuit.

A dialysis membrane is permeable to small
molecules and ions, but not to cells and to mac-
romolecules such as proteins. The difference in
osmotic pressure is much greater in this situa-
tion, especially at low macromolecular concen-
trations, than that predicted by the Van’t Hoff
equation (7). The reason for this “Donnan ef-
fect” is that ions are at a lesser concentration
on the macromolecular side of the membrane,
thus causing an increase in ions of opposite
charge. Removal of proteins from the whey be-
fore fermentation might eliminate most of the
Donnan effect, thus also serving as a way to
alleviate the problem of osmotic influx of water
diluting the fermentor contents. This appears
to be a practical way to improve the process,
because increasingly in industry the valuable
proteins are being recovered from waste whey
by ultrafiltration. The resulting ultrafiltrate re-
mains as a troublesome lactose-rich waste prod-
uct, but it should be ideal as a substrate for
dialysis continuous fermentation into ammo-
nium lactate.

ACKNOWLEDGMENTS

This work was supported by a grant from the Office of
Research Development of Michigan State Unviersity and by
grant ENG 76-17260 from the National Science Foundation.

LITERATURE CITED

1. Carlsson, J. 1973. Simplified gas chromatographic pro-
cedure for identification of bacterial metabolic products.
Appl. Microbiol. 25:287—289.

2. Contois, D. E. 1959. Kinetics of bacterial growth: rela-
tionship between population density and specific
growth rate of continuous cultures. J. Gen. Microbiol.
21:40-50.

3. Coulman. G. A., R. W. Stieber, and P. Gerhardt. 1977.
Dialysis continuous process for ammonium-lactate fer-
mentation of whey: mathematical model and computer
simulation. Appl. Environ. Microbiol. 34:725-732.

4. Dubois. M.. K. A. Gilles, J. K. Hamilton, P. A. Rebers,
and F. Smith. 1956. Colorimetric method for determi-
nation of sugars and related substances. Anal. Chem.
28:350-356.

5. Keller, A. K., and P. Gerhardt. 1975. Continuous lactic
acid fermentation of whey to produce a ruminant feed
supplement high in crude protein. Biotechnol. Bioeng.
17:997-1018.

6. Montgomery. R. 1961. Further studies of the phenol-
sulfuric acid reagent for carbohydrates. Biochim. Bio-
phys. Acta 48:591-593.

7. Moore. W. J. 1962. Physical chemistry, 3rd ed., p.
759-761. Prentice Hall, Englewood Cliffs, NJ.

8. Reddy. C. A.. H. E. Henderson, and M. D. Erdman.
1976. Bacterial fermentation of cheese whey for produc-
tion of a ruminant feed supplement rich in crude pro-
tein. Appl. Environ. Microbiol. 32:769-776.

9. Schultz, J. 3.. and P. Gerhardt. 1969. Dialysis culture
of microorganisms: design, theory, and results. Bacte-
riol. Rev. 33:1-47.

10. Simon, E. W., and H. Beevers. 1952. The effect of pH
on the biological activity of weak acids and bases. New
Phytol. 51:163-189.

 

81

ARTICLE III

DIALYSIS CONTINUOUS PROCESS FOR AMMONIUM-LACTATE

FERMENTATION: IMPROVED MATHEMATICAL MODEL

AND USE OF DEPROTEINIZED WHEY

BY

R. W. Stieber and Philipp Gerhardt

Reprinted from Appl. Environ. Microbiol. 31;487-495, 1979.

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Mar. 1979. p. 487-495

0099-2240/79/03-0487/09S0200/0

Vol. 37, No. 3

Dialysis Continuous Process for Ammonium-Lactate
Fermentation: Improved Mathematical Model and Use of

Deproteinized Whey’r

R. W. STIEBER AND PHILIPP GERHARDT‘

Department of Microbiology and Public Health, Michigan State University, East Lansing, Michigan 48824

Received for publication 5 December 1978

Separate terms for substrate limitation and product inhibition were incorpo-
rated into an equation describing the rate of cell growth for the steady-state
fermentation of lactose to lactic acid with neutralization to a constant pH by
ammonia. The equation was incorporated into a generalized mathematical model
of a dialysis continuous process for the fermentation, developed previously, in
which the substrate is fed into the fermentor and the fermentor contents are
dialyzed through a membrane against water. The improved model was used to
simulate the fermentation on a digital computer, and the results agreed with
previous experimental tests using whole whey as the substrate. Further simula-
tions were then made to guide experimental tests using deproteinized whey as the
substrate. Dried cheese-whey ultrafiltrate was rehydrated with tap water to
contain 242 mg of lactose per ml, supplemented with 8 mg of yeast extract per ml,
charged into a 5-liter fermentor without sterilization, adjusted in pH (5.5) and
temperature (44°C), and inoculated with an adapted culture of Lactobacillus
bulgaricus. The fermentor and dialysate circuits were connected, and a series of
steady-state conditions was managed nonaseptically for 71 days. The fermentation
of deproteinized whey relative to whole whey, with both highly concentrated,

resulted in similar extents of product accumulation but at a lesser rate.

 

Whey utilization continues to be a problem
for the dairy industry. Whey can be processed
by pressure filtration through semipermeable
membranes to obtain protein concentrates (3).
Forty to 70% of the original protein is recovered
(10, 15). The protein has many commercial uses
because of its high nutritional quality (20, 23,
24). However, the capital costs of production are
high and large volumes of lactose-containing
ultrafiltrate are left as residue, which is nearly
as much an environmental burden and an eco-
nomic and nutrient loss as the whole whey.
Thus, feasibility of the process is dependent
upon use of the deproteinized but lactose-rich
ultrafiltrate.

A potential solution to this problem lies with
the conversion of whey ultrafiltrate into feed-
stuff for ruminant animals, accomplished by the
bacterial fermentation of the lactose into lactic
acid and its neutralization to constant pH by
ammonia (7). A background for this develop-
ment exists in studies of the fermentation using
whole whey as the substrate. The fermentation
can be managed as a batch process (18), contin-
uous process (11), or dialysis continuous process

1 Journal article no. 8608 from the Michigan Agricultural
Experiment Station.

(4, 22). The latter process relative to the nondi-
alysis processes enables the use of more concen-
trated substrate, is more efficientin the rate of
substrate conversion, and additionally produces
a dialysate effluent of less concentrated but
purer ammonium lactate.

In the studies reported here, a rate expression
for bacterial growth was developed containing
separate terms for substrate limitation and prod-
uct inhibition. The expression was incorporated
into a mathematical model of the dialysis con-
tinuous process generalized with dimensionless
parameters so that it could be widely applied
(4). The resulting improved set of equations was
employed to simulate the fermentation on a
digital computer, and the results were verified
using previous experimental results with whole
whey as the substrate (22). Further simulations
were then made to guide experimental tests in
which deproteinized whey was used as the sub-
strate. The simulated and experimental results
were used to compare the relative value of the
process applied to the two substrates.

MATHEMATICAL MODEL

Growth rate theory. An expression describing the
rate of cell growth, previously used in a mathematical

487

 

 

 

 

488 STIEBER AND GERHARDT

model of the dialysis continuous ammonium-lactate
fermentation (see equation 11 in reference 4) is as
follows:

(1)

St 1 1
r9 ' "nix. + sfjll + Pf/npjxf

The symbols in this and subsequent equations are
described in Table 1.

In the studies of Stieber et al. (22), an exaggerated
K, value (1,000 mg/ml) was used to correlate the
simulation with the experimental results. The high
value for K, eliminates the effect of the product inhi-

TABLE 1. Glossary of mathematical symbols

 

 

Symbol Description Units

A... Area of membrane avail— cm2
able for dialysis

Fd Flow rate into and out of ml/h
dialysate circuit

F¢° Flow rate into fermentor ml/h
circuit

Ff Flow rate out of fermen- ml/h
tor circuit

K, Product-inhibition con- mg/mg
stant

K. Substrate-limitation con- mg/ml
stant

Pd Product concentration in mg/ml
dialysate circuit

P:° Product concentration in mg/ml
fermentor feed

P,,, Permeability of mem- mg/cmz-h
brane to product

P,,. Permeability of mem- mg/cmz-h
brane to substrate

r, Rate of cell growth mg/ml-h

r, Rate of product formation mg/ml-h

—r. Rate of substrate utiliza— mg/ml-h
tion

8., Substrate concentration mg/ml
in dialysate circuit

SI° Substrate concentration mg/ml
in fermentor feed

Sf Substrate concentration mg/ml
in fermentor circuit

t Time h

Vd Volume of liquid in dialy- ml
sate circuit

V; Volume of liquid in fer- ml
mentor circuit

X: Cell-mass concentration mg/ ml
in fermentor circuit

a Substrate/cell ratio mg/mg

,8 Specific maintenance rate h‘1

7 Product/substrate ratio mg/mg

E Efficiency of lactose con- %
version

)1... Maximum specific growth h"l
rate of cells

T; Cell-retention time in fer- h
mentor circuit

 

APPL. ENVIRON. MICROBIOL.

bition term, and equation 1 is reduced to the following
equation:

3:

r - u [-——-—] x (2)
g In as + Sf f
This equation is based on classical Michaelis-Menten
kinetics and adequately describes the rate of cell
growth for the fermentation. However, the separate
effects of substrate limitation and product inhibition
cannot be determined. Moreover. equation 2 is based
on an enzyme-substrate-product relationship which
implies that there is no product inhibition (21).
These restrictions were considered, and the follow-
ing competitive relationship was proposed:

k k k

1 2 3
"" (3)
x: I sf " xrsr " xrpr " x: + pf
*4

where k1, kg, in. and k. are rate constants. Once the
cells (X!) obtain substrate (Sf), it is metabolized to
product (Pf). Pf cannot be converted to 8;. Further,
since It. 9‘ 0, the cells are inhibited by P: (i.e., they
cease obtaining substrate). In equation 3, assuming
there is an adequate concentration of cells and sub
strate, the rates of substrate utilization and product
formation depend mostly on the concentration of
product in the environment of the cells.

An expression describing the rate of cell growth was
obtained from equation 3 by a steady-state approach
(21):

sf

x
f + Kppfl f

where p... - kzks/(kz + k3). K. - k2lLi/[kilk2 + lb”. and
K, - K2k4/[K1(k2 + kg”. The substrate-limitation con-

(4)

 

...[
g mKs+S

TABLE 2. Glossary of dimensionless parameters
Swnhfland

 

 

Type definition Deacnption
Material Pd - Pd/(ySf) Product factor in dialysate
parameters circuit
Pr- Pg/(ySf) Product factor in fermen-
tor circuit
Pl" - P¢°/(yS¢°)Product factor in fermen-
tor feed
S, - 84/8;o Substrate factor in dialy—
sate circuit
Sq - S./S¢o Substrate factor in fermen-
tor circuit
X; - «flit/Sc" Cell factor in fermentor cir-
cuit
Operational R = P...,/P... Ratio of product/substrate
parameters membrane permeabili-
ties
II = P,,..Aml Membrane permeability
Ff factor
‘1’ - Fd/Fr Flow-rate ratio
Kinetic K. - K./S¢° Substrate-limitation factor
parameters K, a 7K, Product-inhibition factor
0 = mer Time factor

 

ROBIOL.

)UOWIng

(2)

vMenten
of cell
separate
ihibition
is based
p which
I).

9 follow-

(3)

Once the
olized to
Further,
i.e., they
assuming
and sub-
1 product
ration of

owth W85
approach

(4)

ition con-

ameters
/

tion
/
in dialysaw
m fermen-
”1f9nnen-
)r in 01111)"
f in fennen-
mentor CII‘

mum?”
Wmeabm'

’enneflblhty

ation {actor

[ion (“tor

/

 

 

 

VOL. 37, 1979

stant (K.) and the product-inhibition constant (K,)
express the relationships between the actual steady-
state concentrations of the various cell, substrate, and
product states.

Generalized model. The design of the fermenta-
tion system, the assumptions for purposes of modeling.
the material balance equations, and the rate equations
for substrate utilization and product formation were
the same as those used previously (4). The rate equa-
tions and equation 4 were combined with the material
balance equations, and the variables were defined in
dimensionless parameters (Table 2) to obtain a gen-
eralized model for dialysis continuous fermentation.

The resulting equations for the fermentor circuit
are as follows:

43', __ if 81'! _ _ (s)
F. '(1+fl)St- [:-—_-—-_—_-]6—-Xf§llsd§1/Tf
I. 9 8‘ 9 I P 0

pt

5! _ if or, _ _ _o (e)
E;— u -(1 0 I")?! 6 O[:_——_——T_—} + — ‘f 9 ”Pa 9 Pt H,
I. 9 S 0 K P O

2 p f

-—f . [———8f ] 1 it. /'1' (7)
dt _. .. _ _ ' f f
K. 0 St 0 RP?!

where T: - Vr/Fr and is an operational parameter.
For the dialysate circuit, the corresponding equa-
tions are:

as“ _ _
3;.- -(n + “Sci + Itsf rf/vc1 (8)
a?

_ . -(nn 4» med 4» an?! 1?de (9)

This generalized model improves upon that devel-
oped previously (see equations 12 to 16 in reference
4).

Generalized steady-state solution. The equa-
tions of the generalized model were rearranged and
combined and the time derivatives were set at zero to
obtain a generalized solution for the steady state,
comparable to that developed previously (see equa-
tions 17 to 21 in reference 4).

Equations for the substrate, product, and cell mass
in the fermentor circuit are as follows:

DIALYSIS CONTINUOUS FERMENTATION 489

_ E81: +E +EE;
s = P P p (10)

 

 

 

f 1r (9 - 1) - 1r E
p s p
_' "O
_ xsns+e-1+pf(e-1)
Pf = _ (11)
np(6 - 1) - nsxp
_ 1 -§ +n(§ -§)
Xf = f d f (12)

1 + BTf/a.

where w. - HUI/(cs + II) - 1] — l, and 17,, a 1+ RII[1
- fill/(q) + RID].

For the dialysate circuit, the corresponding equa-
tions are:

(I)
ll

nsf/w + n) (13)

Pa

anSf/(s + all) (14)

Equations 10 to 14 thus comprise a generalized
steady«state solution for substrate, product, and cells
in the fermentor and dialysate circuits of the system.

COMPUTER SIMULATIONS

Previous experimental results with whole whey (22)
were used in digital-computer simulations to validate
the improved mathematical model. The values used
are listed in Table 3. The values for ¢ and II were
calculated directly from the experimental results. The
values for am, K., and Kp were determined by succes-
sive curve fitting of the simulated and experimental
results.

A side effect of the dialysis process was a large
osmotic influx of water from the dialysate into the
fermentor. diluting its contents. The dilution was ac-
counted for by assuming that the diluting water en-
tered the fermentor with the feed stream rather than
from the dialysate and by correcting Sr" accordingly.

Figure 1 shows the computer-simulated curves ob-
tained by using the steady-state solution with the
values in Table 3 to describe the correlation between
one important operating parameter and the conver-
sion efficiency, e.g., T; versus S; and Sd. The curves all
fit closely with the superimposed points of the pre-
vious experimental results with whole whey (22).

TABLE 3. Values used for computer simulations of previous fermentations with whole whey (22)

 

 

 

Figure (days) p... (11“) R, R. T; (h) «to II R
1 (19-24) 0.145 0.6 0.0004 3.7 0.66 3.0
1 (50—75) 0.25 0.6 0.0004 3.4 0.62 3.0
2 (40-63) 0.25 0.6 0.0004 16 0.59 3.0

 

 

 

 

 

 

 

-_.~ _. 4.
-‘_«_—_——7.-, _-#7-_

490 STIEBER AND GERHARDT

APPL. ENVIRON. MICROBIOL.

 

|nn
'V I I T I

 

DAYS l9-24

 

20

5:!

 

l0
mm

    

 

 

DAYS 50-75

 

  

 

 

  
     

I
mm

FIG. 1. Computer-simulated effects of cell-retention time (T,) on residual lactose in the fermentor circuit
(Sf) and the dialysate circuit (84) during two periods of dialysis continuous fermentation. The curves were
plotted by use of the steady-state solution with the values in Table 3. The points were replotted from Fig. 2 in
reference 22 to demonstrate the close fit between the experimental results and the computer simulations.

A similar comparison between the simulated and
the experimental results was also made for a second
operating parameter, e.g., F4 versus S, and 8., (Fig. 2).
Although not fitting as closely to the experimental
points, the simulation curves described the trends well.
The discrepancy mostly was caused by decreased
permeability of the membranes.

Altogether, the results demonstrated the validity of
the mathematical model, which then was used to guide
experimental tests for the ammonium-lactate fermen-
tation of deproteinized whey. The improved model
predicted the same effects of changes in the various
parameters on the process and the same regions for
experimental tests as did the previous model (4).

MATERIALS AND METHODS

Inoculum. The inoculum culture was obtained
from the fermentor effluent (day 94) of a previous
dialysis continuous fermentation (22) originally
started with Lactobacillus bulgaricus 2217 (Chris
Hanson’s Laboratory, Milwaukee, Wis).

Substrate. Dried deproteinized cheese whey (pre-
pared by ultrafiltration of whole sweet-cheese wheys
by Stauffer Chemical Co., Rochester, Minn.) was re-
hydrated to contain 242 mg of lactose per ml and was
supplemented with 8 mg Of yeast extract per ml. The
reconstituted whey was made up in 7-liter batches
without sterilization, stored at 4°C, and held in a
stirred, heated (60°C) reservoir to keep the lactose in
solution.

Dialysis continuous fermentation system. The
experimental dialysis fermentation system was con-
ducted continuously at a temperature of 44°C and a
pH of 5.5 with essentially the same equipment as used
previously (22). However, plunger-type reciprocating

 

pumps (type “ P", Bran and Lubbe, Inc., Evanston,
Ill.) were used to meter the whey and water, and the
pH was regulated with a different automatic device
(model pH-40, New Brunswick Scientific Co., New
Brunswick, N. J .). The circulation rates through the
dialyzer, for both the fermentor and dialysate circuits,
were 2 liters per min. Operation of the system was
interrupted once, for 3 weeks, during which the fer-
mentor-circuit contents were stored at 4°C.

Analytical procedures. Samples from the dialy-
sate and fermentor circuits were taken at 12-h inter-
vals from the dialysate effluent and from a glass “T”
inserted in the tubing between the fermentor and
dialyzer. Steady-state data were determined from sam-
ples taken at five times the cell retention time or 48 h
after changing a parameter. Lactose in the samples
was determined by the colorimetric method of Morris
(14). Lactic acid was determined by use of a gas
chromatograph (series 1420, Varian Associates, Palo
Alto, Calif.) with an integrator (model CDS 111, Var-
ian Associates), using a stainless-steel column (6 feet
by Ma inch (on. 1.83 m by 0.32 cm] outside diameter)
packed with 10% SP-1000/1% H3PO. on 100/ 120 Chro-
mosorb WAW (Supelco, Inc., Bellefonte, Pa). Samples
were prepared by the procedure of Holdeman and
Moore (9). Lactic acid also was determined by the
colorimetric method of Pryce (17). Specific conduct-
ance was measured with a conductivity bridge (model
RC-16B2, Beckman Instruments, Inc., Cedar Grove,
N. J .).

RESULTS

Cell-retention time versus conversion ef-
ficiency. The results Of experimental variation
in the cell-retention time (T;) affecting the resid-

“j

 

 

 

 

 

    

VOL. 37, 1979
'50 I I 1
I40 ~ -
IZO - -
DAYS 40-63
100 r -‘
'é
\
o
3 30 I. ..
a?
8
«5' 60 ~ -
4o - —
I
20
O.
O 300 600 900

FIG. 2. Computer-simulated effects of dialysate
flow rate (P,,) on residual lactose in the fermentor
circuit (8,) and the dialysate circuit (8.!) with cell-
retention time held constant (T, = 16 h). The curves
were plotted by use of the steady-state solution and
the values in Table 3. The points were replotted from
Fig. 4 in reference 22 to demonstrate the fit between
the experimental results and the computer simula-
tions.

ual lactose in the fermentor circuit (Sf) and the
dialysate circuit (8..) showed that the lactose
concentrations in both circuits decreased with
increased T: (Fig. 3). All of the curves reached a
point where an increase in Tr did little to de-
crease the lactose concentration.

Dialysis rate versus conversion effi-
ciency. The flow-rate ratio (qb) is an operational
parameter used to manipulate the rate Of di-
alysis. Results of experimental variation in <1)
affecting the residual lactose in the fermentor
circuit (Sr) and the dialysate circuit (8..) are also
shown in Fig. 3. The lactose concentrations were
less with increased dialysis (<1) = 2.5) than with

DIALYSIS CONTIN UOL'S FERMENTATION

491

decreased dialysis (d) = 1.0). The effect was
mostly on Sd.

Cell-retention time versus lactate con-
centration. Figure 4 shows the results of vari-
ation in Tr affecting the lactate concentrations
in the fermentor circuit (Pf) and in the dialysate
circuit (P,,). The lactate concentrations in-
creased tO a maximum with increasing Tr.

Dialysis rate versus lactate concentra-
tion. Figure 4 also shows the results of the flow-
rate ratio affecting Pf and P... ¢ had little effect
on Pf. However, Pd was much greater with de-
creased dialysis (d; = 1.0) than with increased
dialysis (O = 2.5). Values for the dialysate prod-
uct yield (Pdo, which represents the fraction of
substrate leaving as lactate in the dialysate ef-
fluent) were similar at either value Of 4).

Lactose conversion and lactate produc-
tivity. Figure 5 shows that the lactate produc-
tivity decreased with increased Tr, whereas the
percentage Of lactose converted to product in-
creased with increased Tr. High lactate produc-
tivity and high lactose conversion along with
high lactate concentrations are all desirable.

 

'40 I I I I I

I20

 

20

 

 

 

IS 20 25 30
mm

FIG. 3. Computer-simulated effects of cell-reten-
tion time (T,) on residual lactose in the fermentor
circuit (Sf) and the dialysate circuit (S4) at two flow-
rate ratios (o) and during two time periods of dialysis
continuous fermentation. The curves were plotted by
use of the steady-state solution and the values in
Table 5. The points are experimental data and dem-
onstrate the fit between the experimental results and
the computer simulations. The dashed curves and
circle points were obtained at <1) = 1.0 and during
days 26 to 36. The smooth curves and square points
were obtained at 4) = 2.5 and during days 40 to 49.

I
0 5 IO

 

 

 

 

492 STIEBER AND GERHARDT

 

 

 

 

 

l
O 5 IO I5 20 25 30
mm

FIG. 4. Computer-simulated effects of cell-reten-
tion time (T,) on accumulated lactate in the fermentor
circuit (P,) and the dialysate circuit (Pd) at two flow-
rate ratios (¢) and during two time periods of dialysis
continuous fermentation. The curves were plotted by
use of the steady-state solution and the values in
Table 5. The points are experimental data and dem-
onstrate the fit between the experimental results and
the computer simulations. The dashed curves and
circle points were obtained at ¢ =2 1.0 and during
days 26 to 36. The smooth curves and square points
were obtained at ¢> = 2.5 and during days 40 to 49.

 

 

 

 

20 I I I I I '00

I6" "80
E
.L I2~ «so A
E s
o v
5 In
0. 8" «40
L.

4" ‘20

O l l 1 l l

0 5 IO 15 20 25 300
mm

FIG. 5. Experimental effect of changes in cell-re-
tention time (T ,) on lactose conversion (E) and lactate
productivity (r,,) at a flow-rate ratio (<5) of 1.0. Percent
lactose conversion = [I — (ES, + Fde)/(F"} 9)] x
100. Lactate productivity = (P/F, + PdFd — P‘}F‘})/
(TIP/)-

Consequently, trade-offs in the regulation of Tf
are required in designing the system for practical
use.

Product/substrate ratio. Results Of exper-
imental variation in T; and ¢ showed that these
operations] parameters had little effect on the
product/substrate ratio (7, results not shown).
Consequently, for purposes of modeling the am-
monium-lactate fermentation, a fixed value was
assigned to 7 (0.96). The ratio was calculated

APPL. ENVIRON. MICROBIOL.

from rp/r,, where values for r,, were obtained
from Fig. 5 and where r, = (SPFr0 —S;F{ — SdFd)/
(Tth)-

Dialyzer dependability. The preceding ex-
perimental results all were Obtained during un-
interrupted operation of the system for 56 days.
After a hiatus of three weeks, Operation was
resumed for an additional 15 days. The fermen-
tor and dialysate compartments of the dialyzer
stayed clean throughout the total 71 days of
Operation. However, the membrane surfaces in
the fermentor compartments were fouled with a
film of debris after the 56 days, and fresh mem-
branes were used for the final 15 days.

Product quality. Samples from the fermen-
tor circuit were regularly analyzed by gas chro-
matography, not only to determine the concen-
tration Of lactic acid but also to monitor the
possible presence of atypical metabolic products.
The results indicated that the fermentation re-
mained homofermentative i.e., only negligible
amounts Of products other than lactic acid were
found.

Prolonged steady-state Operation. The di-
alysis continuous fermentation was Operated at
set steady-state conditions (T; = 27.2, 4) = 2.5)
during the 15-day period to determine other
characteristics of the process (Table 4). The
mean rate of ammoniumolactate production was
4.7 mg/rnl-h and the lactose conversion was 81%.

By day 58 Of the fermentation, the residual
lactose levels were very low, but then the levels
increased with time. The lactose increase was
possibly caused by a buildup of inhibitory sub-
stance in the fermentor circuit resulting from
membrane fouling, as seen by the greater differ-
ences between Sf and Sd and between Pf and P,,
as time progressed.

During the fermentation, a mean of 36.5% of
the fermentor effluent represented water which
osmosed from the dialysate circuit, i.e., a mean
of 56.4 ml of water per h (Fr -Ff0 in Table 4)
entered the fermentor from the dialysate circuit.

From material-balance data (Table 4), the
equivalent substrate concentration in the fer-
mentor-feed stream (S?) was calculated by use
of a conservation-Of-mass equation. The mean
calculated value (247 mg/ml) agreed well with
the mean analytical value (242 mg/ml) for Sf”.
Thus, the results confirmed that no significant
portion of the substrate was lost to products
other than lactic acid.

Validation Of the mathematical model.
The values shown in Table 5 were used to cor-
relate the experimental and simulated results.
The values for (1) and y were calculated directly
from the experimental data. The values for um,
K3, and Kp were obtained by successive curve
fitting Of simulated with experimental results Of

 

 

amhcmnmr—M

X: {I‘M‘srzo

.. EN'I'lRON. MIL‘ROBIOL.

for r, were obtained
15,’F,-'—S,F1-Sdl‘dl/

y. The preceding ex~
obtained during un-
.= system for 56 days.
eels, operation was
5 days. The fermen-
ents of the dialyzer
9 total 7l days of
nbrane surfaces in
were fouled with a
s. and fresh mem-
l5 days.
from the fermen-
rzed by gas chro-
nine the concen-
to monitor the
abolic products.
annentation re-
only negligible
actic acid were

ation. The di-
IS operated at
27.2. ’ = 2.5)
emiine other
tb/e 4). Th9
tduction W88
ion “21.9819.
he residual
n the levels
crease was
)it0r_\' sub-
’ting from
ter differ-
PI- and Pd

36.5? of
ar which
.9 mean
‘able 4)

circuit.
17. the
re fer-
')_(' use
mean
with
r 51'”-
Rant
ucts

iel.
‘01"
its.
tly
"1!
be
9f

VOL. 37, 197.1

DIALYSIS CONTINUOUS FERMENTATION

493

TABLE 4. M aterial- balance data for operation of the dialysis continuous fermentation system at steady-
state conditions

 

 

Day T. on 4» 5'3” $3” Pm“ ”,3” Fr (ml/h) r. (ml/h) F. (ml/h) S'nffilg/
25.5 2.29 77.0 16.3 22.5 9 4 104.4 164.4 376.8 245.1
57 24.2 2.26 33.0 6.6 55.0 23 6 63.4 172.3 390.0 314.5
25.7 2.36 12.6 2.6 61.5 31 2 35.2 163.2 335.6 293.7
56 26.2 2.69 6.7 0.5 65.0 29.6 34.7 143.3 400.2 265.6
27.4 2.62 7.5 0.4 63.0 23.0 94 2 153.0 400.3 239.3
59 26.4 2.46 10.0 1.1 61.0 29.4 93.2 153.4 339.4 245.1
27.4 2.60 15.0 1.2 67.0 25 2 97.0 153.0 393.4 233.4
60 27.9 2.66 15.3 1.4 65.0 27 0 101.3 150.0 399.0 227.9
27.5 2.62 17.5 2.1 62.5 26 2 96.5 152.4 399.6 233.3
61 27.0 2.53 17.5 2.7 73.0 25.6 101.4 155.4 393.0 243.3
27.0 2.53 21.7 2.3 63.0 22.0 102.9 155.4 393.6 224.3
62 26.9 2.53 25.9 2.6 65.0 22.0 100.3 156.0 394.2 233.0
26.3 2.46 22.2 2.3 66.5 25.0 102.2 159.0 391.3 240.1
63 27.1 2.55 21.2 2.5 67.0 24.6 94.7 154.3 395.4 253.2
27.7 2.61 13.4 2.2 67.5 22.3 99.4 151.2 394.3 224.9
64 27.6 2.61 19.7 2.2 69.0 22.6 97.1 151.3 396.0 235.2
27.6 2.53 21.3 2.1 74.5 24.6 100.0 151.3 391.2 244.4
65 27.2 2.53 22.9 2.1 69.0 26.6 93.0 154.2 369.4 256.5
27.4 2.53 23.6 2.0 30.0 23.6 100.0 153.0 367.6 253.1
66 27.7 2.56 22.9 2.0 74.0 23.4 95.0 151.2 339.4 254.5
27.6 2.59 24.9 1.9 30.0 22.0 97.2 150.6 390.0 254.4
67 23.2 2.61 24.3 1.6 33.5 19.0 96.9 143.3 366.3 244.5
27.9 2.60 24.2 1.6 75.5 21.2 104.0 150.0 390.6 225.1
66 27.3 2.53 23.1 1.5 75.5 20.2 99.5 150.6 333.2 229.6
27.6 2.53 27.7 2.1 69.5 25.4 1 151.3 334.6 251.0
69 27.5 2.54 26.6 2.0 72.5 22.6 99.4 152.4 336.4 243.3
27.2 2.49 27.5 2.2 79.0 21.0 I 154.2 334.6 251.4
70 27.0 2.47 27.5 2.3 77.5 23.6 107.6 155.4 384.6 240.1
27.5 2.53 27.7 1.8 33.5 21.4 93.4 152.4 365.3 259.6
71 27.3 2.52 23.6 2.2 77.5 23.0 99.0 153.6 337.6 259.9
Mean 27.2 2.54 23.2 2.7 69.2 23.7 97.9 154.3 390.9 247.0

 

" Sto 3 (Ftsf + Ftp! + FIX: + Fde + ded)/F{o - Pro, where Pro = 9.1 mg/ml.

TABLE 5. Values used for computer simulations of experimental fermentations with deproteinized whey

 

 

Figure (days) u... (h") K, K. ‘9 II R 7
3 (26—36) 0.35 2.2 0.0004 1.0 0.50 3.0 0.96
3 (40—49) 0.35 2.2 0.0004 2.5 0.40 3.0 0.96
4 (26-36) 0.35 2.2 0.0004 1.0 0.50 3.0 0.96
4 (40—49) 0.35 2.2 0.0004 2.5 0.40 3.0 0.96

 

a nondialysis continuous fermentation (unpub-
lished results). The value for II was obtained
similarly with the experimental results of the
dialysis continuous fermentation. Sr” was taken
as 160.3 mg/ml instead of 237.5 mg/ml (which
was Obtained from material balances of the ex-
perimental data) because a mean 32.5% Of the
fermentor contents represented water which os-
mosed from the dialysate circuit.

Figure 3 shows the computer-simulated curves
obtained by use of the steady-state solution with
the values in Table 5 to describe the relation
between the two principle Operating parameters
(T1 and «11) and the conversion efficiency (8; and
8.4). Figure 4 shows the relation between the
same Operating parameters and the accumula-

tion of product (P1 and P,,). The curves in both
figures fitted well with the superimposed points
Of experimental results and thereby further dem-
onstrated the validity Of the mathematical
model. Moreover, the kinetic constants (11",, K,”
and Kp) used for the simulations in Fig. 3 and 4
were Obtained from the results Of nondialysis
continuous fermentations, adding to the valida-
tion of the model.

Process evaluation by use of the mathe-
matical model. The experimental and simu-
lated results were correlated to evaluate the
fermentation process. The values for u... (0.35
h“), K. (0.0004), and Kr, (2.2) were the same in
both time periods, indicating no change in the
bacterial culture from days 26 to 49. The values

 

494 STIEBER AND GERHARDT

for K. and KP also showed that the fermentation
was not affected by substrate limitation but was
greatly limited by increasing concentrations of
product. The values for 11 (0.5 at days 26 to 36;
0.4 at days 40 to 49) showed that the permeabil-
ity Of the membrane decreased as the fermen-
tation progressed. Calculations of II using a P,,...
Of 0.06 mg/cm2-h showed that II should have
been about 0.9. Thus, by 4 weeks into the fer-
mentation, the permeability factor decreased by
50%. Coulman et al. (4) showed that a II of 2.0
should be used for the process to obtain suitable
relief from product inhibition.

Process monitoring. On-line sensors are be-
coming increasingly irnportant in the Operation
of fermentations (16). In the present process, the
dialysate effluent is a relatively pure solution of
ammonium lactate and should be measurable by
its electrical conductivity. Figure 6 illustrates
the positive correlation between the concentra-
tion of ammonium lactate and the conductance
in the dialysate. Thus, a conductivity bridge
could be readily used for the on-line monitoring
of product concentration. Moreover, if coupled
with the model, this parameter could indicate
other parameters of the fermentation, e.g., cell
concentration in the fermentor and membrane
permeability in the dialyzer.

DISCUSSION

Various models have been used to describe
the growth Of lactic acid bacteria (4—6, 8, 11-13,
19). Although these models correlate with ex-
perimental results, they have only been used for
fermentations of very low substrate and product
concentrations, they contain equations or terms
which have no biological basis, or they lack a

 

SC I I I I

40*-

30’-

Pd (mo/ml)

201-

 

 

1 I 1 1
I5 20 25 30 35 4O

CONDUCTANCE (mmho/cml

 

 

FIG. 6. Regression correlation between concentrac
tion of ammonium-lactate (Pd) and electrical con-
ductance in the dialysate circuit.

APPL. ENVIRON. MICROBIOL.

strong product-inhibition effect. Recently, Abor-
hey and Williamson (1) developed a model
which appeared useful, but the model was com-
plicated for our purposes because it incorporated
the concept Of intracellular substrate concentrav
tion. Since acid production cannot continue once
all of the substrate has been utilized, Rogers et
al. (19) suggested improving Luedeking and Pi-
ret’s (13) rate equation for acid production by
incorporating a substrate-limitation term into
the maintenance term. Their simulation results
agreed with experimental results, but the K.
(0.371 mg/ ml) used for the simulations was high
and probably represents a product-inhibition ef-
fect (2, 11, 22). A better way to improve the
equation would be to incorporate a product-in-
hibition term as well as a substrate-limitation
term.

The model developed in the present study is
simple and correlated well with experimental
tests of dialysis and nondialysis (unpublished)
continuous processes for the ammonium-lactate
fermentation over a wide range of operating
parameters with high substrate and product con-
centrations. Specifically, the model agreed with
experimental results Of substrate utilization (Fig.
1 tO 3), product formation (Fig. 4), and cell-mass
accumulation (results not shown). The model
also contains substrateolimitation and product-
inhibition terms which have a biological basis.
Moreover, the values used for these terms are
realistic. To date, the model has only been cor-
related with steady-state data. The major short-
coming Of the model is that the product-inhibi-
tion effect is not strong enough at lactate con-
centrations greater than 70 mg/ ml. As suggested
above, the incorporation of a product-inhibition
term into the maintenance term may provide a
solution. Another cause Of the problem is that,
since a dialysis culture is able to tolerate greater
lactate concentrations than is a nondialysis cul-
ture (22), there may be an unknown dialyzable
factor which also has an inhibition effect on the
lactic acid fermentation.

The mean product/substrate ratio of the am-
monium-lactate fermentation was determined as
0.96 mg/mg. Thus, 96% of the lactose utilized
was converted to ammonium-lactate and 4% was
incorporated into bacterial cells. On a substrate
efficiency basis, the fermentation thus would be
more useful for production of ammonium-lactate
than for bacterial cells.

The dialysis continuous process has been used
to ferment whole whey (22) and deproteinized
whey. Both fermentations are most efficient
when using a T1 Of about 20 h, i.e., further
increases in T; do little to decrease Sf and 8.,
(compare Fig. 1 and 3). A comparison Of the use
Of these substrates at. similar conditions is shown

 

 

.Iln‘aonIOL.

IIIlI‘. Abor-

a model
‘wascom-
orporated
oncentra-
inue once
Rogers et
I and Pi-
ction by
m into
iresults

the K.
was high
ition ef-
1V8 the
uct-in-
itation

udy is
rental
Shed)
It‘tate
nting
con-
with
'Fig.
nass
)(lel
ICI-
sir).
are
or-

VOL. 37, 1979

DIALYSIS CONTINUOUS FERMENTATION 495

TABLE 6. Comparison of wheys used for production of ammonium-lactate by dialysis continuous

 

 

 

fermentation
Sng/ S(m/ P(m/ St / P
Substrate T h I g t 8 4 m8 4! (mg/
ml) ‘( l 4’ E ‘9") ml) ml) ml) ml)
Whole whey (22) ’ 230 19 2.5 97 2.6 80 0.4 24
Deproteinized whey 242 27 2.5 81 23.2 69 2.7 24

in Table 6. The fermentation of whole whey technol. Bioeng. 19:821-839.

relative to deproteinize' d whey occurred at a 6- Fnedmd “P: “£5.18" PndbE-LL- at“??? 41111920.! 1’0”“ng

greater fermentation rate and resulted in more 9" ac.‘ p. “cm" y ““0 a.“ “8 ( 'l .e "m ‘f

, In a dialySIs culture system. Biotechnol. Bioeng. 12.

complete substrate conversmn. The fermentor 961-974.

contents Of both processes were similarly diluted 7. Gerhardt. P... and C. A. Reddy. 1978. Conversion of

by a net osmosis of water from the dialysate “Ewing'rml Wastes! ‘fnw ”$193“ I???” ,by a";

- - . - - monia organic acr ermen mm a ne revrew an

CII'CUIIE. BOth fementatlons were 811111.1me 3“” preview. Dev. Ind. Microbiol. 19:71—78.

0191“ m. the Percent convefsflon 0f lactose Into 3. Hanson. 'r. P., and G. T. Tsao. 1972. Kinetic studies of

ammomum-lactate only and In the high concen- the lactic acid fermentation in batch and continuous

tration of accumulated ammonium-lactate in the 9 H Crgtures- B£$Chnoh aWES-Clammftiém A be
. o eman, ., an . . . oore. . naero
fementor' The present StUdy also showed that laboratory manual. 3rd ed., p. 113-114. V.P.I. Anaerobe

a high concentration of cell-freeammonium lac- Laboratory, Virginia Polytechnic Institute and State

tate (50 mg/ml) could be maintained in the University. Blacksburg, Va.

- - Membrane processing of cheese whey reaches commer-
The mathematical model showed the perme cm scale. Food Technol. 26:30_35-

ablhty factor '20 be as low 8% 04- Consequeptlyn II. Keller, A. 15.. and P. Gerhardt. 1975. Continuous lactic

the fermentation could be improved conSIder- acid fermentation of whey to produce a ruminant feed

ably (e.g., more efficient conversion Of substrate Sgpplflmemahish in crude Protein Biotechnol- Bioeng-
° ' ' l :99 -101 .

and greater lacuite concentration m the dlaly 12. Lee, 1. IL. A. G. Fredrickson, and H. M. Tsuchiya.

sate effluent) With use Of 8.1110“? permeable 1976. Dynamics Of mixed cultures of Lactobacillus

membrane. Greater converswn Of substrate plantarum and Propionibacterium shermanii. Biotech-

could also be had by increasing the cell-mass "01-3309'18- 13:513-525- . . _

concentration in the dialysis fermentor. This 13' Luedek'qg’ R." and E‘ 91"?“ 1959' A kmem Study Of
. ht b lished b rec clin th cells 0 the lactic-and fermentation. batch process at controlled

mlg e accomp * . y y g e . 1" ' pH. J. Biochem. MicrObIOl. Technol. Eng. 1:393-412.

by employing a nondialySIs prefermentor which 14. Morris, D. L. 1948. Quantitative determination Of carbo-

is Optimized for cell production and the effluent hygrates with Drewood's anthrone reagent Science
' ' ' ' 10 :254-255.

90’." ‘l’hwh.fl9ws mm the willy” ffmentor’ 15. Nielsen, I. 11.. A. G. Bundgaard, O. J. Olsen, and R.

Wthh ls Optll’Illzed I‘Ol‘ conversmn effiCIency and F. Madsen. 1972. Reverse osmosis for milk and whey.

product accumulation. Such fermentation sys— Process Biochem. 7:17-20.

terns can be easily modeled before conducting 16. Perlman. D. (ed.). 1977. Annual reports on fermentation

experimental tests processes, vol. 1, p. 95—114. Academic Press, Inc., New

' York.
17. Pryce, J. D. 1969. A modification of the Barker-Summer-
ACKNOWLEDGMENT son method for the determination Of lactic acid. Analyst
This work was supported by grant ENG 76-17260 from the 94:] 151-1152-

National Science Foundation. 18. Reddy, C. An, H. E. Henderson, and M. D. Erdman.
1976. Bacterial fermentation of cheese whey for produc-
tion Of a ruminant feed supplement rich in crude pro-

LITERATURE CITED tein. Appl. Environ. Microbiol. 32:769-776.
1. Aborhey, S" and D- Williamson. 1977- Modelling 0f 19. Rogers, P. L., L. Bramall, and l. J. McDonald. 1978.
lactic acid production by Streptococcus cremoris HP. Kinetic analysis of batch and continuous culture of
J. Gen. Microbiol. 23:7—21. Streptococcus cremoris HP. Can. J. Microbiol. 24:372-
2. Contois, D. E. 1959. Kinetics of bacterial growth: rela- 380.
tionship between population density and specific 20. Schingoethe, D. J. 1976. Whey utilization in animal
growth rate of continuous cultures. J. Gen. Microbiol. feeding: a summary and evaluation. J. Dairy Sci. 59:
21:40-50. 556-570.
3. Coton. S. G. 1976. Recovery of dairy waste, p. 221-231. 21. Segel, I. H. 1975. Enzyme kinetics, p. 18-37. John Wiley
In G. G. Birch. K. J. Parker, and J. T. Worgan (ed), and Sons, New York.
Food from waste. Applied Science Publishers, Barking, 22. Stieber, R. W., G. A. Coulman, and P. Gerhardt. 1977.
England. Dialysis continuous process for ammonium-lactate fer-
4. Coulman, G. A., R. W. Stieber, and P. Gerhardt. 1977. mentation of whey: experimental tests. Appl. Environ.
Dialysis continuous process for ammonium-lactate fer- Microbiol. 34:733-739.
mentation Of whey: mathematical model and computer 23. Wingerd. W. H. 1970. Lactalbumin as a food ingredient.
simulation. Appl. Environ. Microbiol. 34:725-732. J. Dairy Sci. 54:1234—1236.
5. Driessen, F. M., J. Ubbels, and J. Stadhouders. 1977. 24. Wingerd, W. H., S. Saperstein, and L. Lutwak. 1970.

Continuous manufacture Of yogurt. 1. Optimal condi-
tions and kinetics Of the prefermentation process. Bio-

Whey protein concentrate has excellent nutritional
properties. Food Technol. 24:758-764.

 

91

ARTICLE IV

PRODUCTION OF LACTOBACILLUS CELLS BY DIALYSIS

CONTINUOUS FERMENTATION OF DEPROTEINIZED WHEY

BY

R. W. Stieber and Philipp Gerhardt

Production of Lactobacillus Cells by Dialysis
Continuous Fermentation of Deproteinized Whey

R. W. STIEBER and PHILIPP GERHARDT

Department of Microbiology and Public Health,
Michigan State University,
East Lansing, Michigan 48824

ABSTRACT

A previously developed mathematical model was used to computer-
simulate the continuous conversion of deproteinized whey into bacterial
cells by a process in which the fermentor contents are poised at a
constant pH by the addition of ammonia and are dialyzed through a
membrane against water. The simulations predicted the fermentation
and dialysis parameters for experimental test. Dried deproteinized
cheese whey was rehydrated with tap water to contain 242 mg lactose/ml,
supplemented with 8 mg yeast extract/m1, charged into a 5-liter
fermentor without sterilization, adjusted in pH (5.5) and temperature

(44°C), and inoculated with an adapted culture of Lactobacillus

 

bulgaricus. The fermentor and dialysate circuits were connected and

 

a series of steady-state conditions was managed for 71 days without
sterilization or asepsis. Cell accumulation, cell productivity and
molar growth yield were greater than those obtained from an

ordinary continuous fermentation for the production of Lactobacillus

 

cells.

 

l .
Journal article number 8962 from the Michigan Agricultural

Experiment Station.

INTRODUCTION

The cells of Lactobacillus and other lactic-acid genera apparently

 

are safe and acceptable for human use: these bacteria have been
consumed in large amounts regularly for lifetimes by the various
nationalities of people who use fermented milks as an integral part of

their diet. Furthermore, the cells of Lactobacillus bulgaricus, for

 

example, contain protein of high nutritional quality (Erdman et al.,
1977). These cells have been used in the continuous manufacture of
yogurt milk (Lewis, 1969; Driessen et al., 1977a and 1977b), and also
can be immobilized and used to hydrolyze lactose in foods for intake
by people who cannot digest the sugar (Ohmiya et a1., 1977). However,
the production of E? bulgaricus may be most useful as a model for the
production of cheese starter cultures for the dairy industry.

Such cell-production processes can be managed batchwise (Marshall
and Earle, 1975; Reddy et a1., 1976), continuously (Keller and Gerhardt,
1975; Cox and MacBean, 1977; Stieber and Gerhardt, 1979b), and dialysis-
continuously (Coulman et al., 1977; Stieber et al., 1977; Stieber and
Gerhardt, 1979a). Of these the dialysis continuous process should be
the most useful because it enables the use of concentrated substrate,
is efficient in the rate of substrate conversion, and produces a
useful cell-free permeate (Stieber et al., 1977). Moreover, pre-
liminary theoretical study indicated that the cell mass in a continuous
process is directly proportional to the feed substrate concentration
that can be metabolized; therefore, the cell yield of a dialysis
continuous process should be much greater than the yield obtained
from an ordinary continuous process. Recently, Osborne (1977) used a

dialysis batch process to produce concentrated cheese starter cultures

ll

of Streptococcus, obtaining concentrations of 10 cells per ml. A

 

dialysis batch process also was used for production of yeast of up to
90 g dry weight per liter from deproteinized whey (Lane, 1977).
However, dialysis continuous operation appeared desirable for producing

Lactobacillus cells because waste whey is generated and the fermentation

 

products could be used more or less continuously, and a continuous
process generally is much more productive than a batch process (Pirt,
1975).

Deproteinized whey was used as substrate in the present study.
It contains essentially all of the carbon and low molecular-weight
nitrogen sources that are in milk or whole whey. Albumins and other
milk proteins are removed and concentrated by ultrafiltration of the
whey after making cheese from the milk (Coton, 1976) because the
resulting protein is of high nutritional quality and has commercial
value (Wingerd, 1970; Wingerd et al., 1970; Schingoethe, 1976; De Boer
et al., 1976; Morr, 1976; Delaney, 1977; Mathur and Shahani, 1979).
The residual deproteinized whey, however, represents an environmental
burden and an economic and nutrient loss for which a practicable use
needs be found. We first considered use of this dairy residue as a
substrate for the fermentative production of ammonium-lactate (Stieber
and Gerhardt, 1979a; Stieber and Gerhardt, 1979b), which can be used
as a nitrogen-rich feed supplement for ruminant animals (Gerhardt and
Reddy, 1978; Juengst, 1979).

The present study was undertaken to determine the feasibility of

producing cells of lactic-acid bacteria in general and of l? bulgaricus

 

in particular from deproteinized whey. The usefulness of a dialysis

continuous fermentation was examined and compared with the prior

3

results of an ordinary (nondialysis) continuous fermentation. As in
our previous studies in this general area (Coulman et al., 1977;
Stieber and Gerhardt, 1979a; Stieber and Gerhardt, 1979b) the investi-
gative procedure was to use a mathematical model to simulate, on a
digital computer, the outcome to be expected and the parameters to be
studied. The theoretical predictions then were tested experimentally.
Both the theoretical and experimental results demonstrated the potential
advantages and disadvantages of this unusual type of fermentation
process.

THEORY

System Design. Figure 1 shows a schematic of the dialysis

 

continuous fermentation system design. The symbols correspond to
those used previously and are listed in Table l.

The feed into the fermentor circuit contains the substrate
(lactose) in relatively high concentration (5:) and is maintained at a
relatively low rate of flow (Ff) consistent with a minimal concentra-
tion of unused substrate in the fermentor effluent (Sf). The substrate
is converted essentially only to cells (xf) and product (lactic acid,
Pf) which both remain in the fermentor effluent. A small concentration
of the product preexists in the feed (P2). The liquid volume in the
fermentor (Vf) is maintained at a constant level. The addition of
ammonia solution is not shown because it did not enter into the
theoretical considerations.

The system does not have a reservoir vessel for the dialysate.
Instead, the dialysate circuit consists only of tubing, pump and the
dialysate side of the dialyzer, and contains a relatively small volume
(V6) which is continuously circulated. The feed into the dialysate

4

circuit consists only of water and is maintained at a relatively high
flow rate (Pd), consistent with a very low substrate concentration
(Sd) and a practically useful product concentration (Pd) in the
dialysate effluent.

This system provides the greatest concentration gradient possible
for dialysis and, thus, much of the product is removed from the
immediate environment of the cells, relieving the inhibition of cell
growth by the product. As more product is withdrawn by dialysis, more
substrate is consumed and more cells are formed.

Mathematical Model. A mathematical model previously was developed
for the dialysis continuous fermentation (Stieber and Gerhardt, 1979a).
The generalized solution of the model for steady-state conditions is

as follows:

0 O

K’np + wasf + Pf)
Sf I (1)
6 - 1 w - K r
( ) p Y 9 s

O 0
7K3": 4» (6 - l) hrsf + Pf)

 

 

Pf - - (2)
(9 - 1hrp - ykpw.
s; - st + Insd - sf)
Xf " (3)
a + Be/uIn
8d I fist/w + II) (4)
Pd I “Pf/(O 4' RH) (5)

where "s = H(H/(¢ + H) - l) - l, "p = l + RH(l - RH/(¢ + RH)).

Computer Simulated Predictions. The steady-state solution was

 

used to simulate the effect on cell mass (Xf) of feed substrate concen-

tration (8:) and three operational parameters: cell-retention time

(Tf), flowhrate ratio (¢), and permeability factor (H). Because the

fermentation was simulated before actual experiments were conducted,

approximations of various conditions were used (Table 2). The values

for K K R d 1>°
“my 5’ I Yr I an

p f
Gerhardt (1977). Values for a and B were Obtained from Keller and

were obtained from Stieber, Coulman and

Gerhardt (1975). Values for o and H were chosen arbitrarily. A value

0

of 150 mg/ml for Sf

was used because previous work (Stieber et al.,
1977) indicated that dilution of the fermentor contents occurs in the
dialysis process.

The results in Fig. 2 show the simulated effect of changes in
feed substrate concentration on the cell mass in the fermentor circuit
at constant cell retention time and constant rate of dialysis. The
cell mass is proportional to the substrate concentration that can be
metabolized by the cells. Thus, the dialysis process is most produc-
tive for cell production at high substrate concentrations. For these
results, it was assumed that all of the substrate was metabolized,
i.e., all the lactose in the whey was converted to lactic acid and
cells.

Fig. 2 also shows the effect of changes in the product-inhibition
factor (KP) on the dialysis continuous fermentation. The smaller the
value of HP, the greater the cell mass.

The cell-retention time (Tf = Vf/Ff) is an operational parameter
which is varied to change the average length of time a cell remains in
the fermentor. Fig. 3 shows the simulated effects of changes in T on

f
6

cell mass in the fermentor circuit. At first, with increasing retention
time where growth metabolism (0) predominated, the cell mass increased
to a maximum. But beyond a critical retention time, maintenance
metabolism (8) superseded that of growth and then the cell mass
decreased.

The flow-rate ratio is an operational parameter used to manipulate
the rate of dialysis (¢ = Fd/Ff). Simulated results of variation in
the flowbrate ratio affecting the cell mass also are shown in Fig. 3.
The cell mass increased only a little with a four-fold increase in the
flowbrate ratio (i.e., with a greatly increased rate of dialysis).

Fig. 4 emphasizes that the overall effect of changes in the flow-
rate ratio on the cell mass is small. In fact, the cell mass for a
nondialysis continuous process (¢ = O) is only slightly less than the
cell mass for a continuous process with a high rate of dialysis (¢ = 4),
when both processes are operated at the same feed substrate concentra-
tion. Consequently, the primary advantage of a dialysis continuous
process over an ordinary continuous process for production of cells is
that the former enables the efficient use of concentrated substrate
when an inhibitory and dialyzable product is formed.

The permeability factor is an Operational parameter which is a
measure of the diffusion capacity of the dialysis process

(n=p A/F

ms m_ f). Simulated results of variation in the permeability

factor affecting the cell mass also are shown in Fig. 4. The cell
mass increased a little with decreased permeability factor at low
flowbrate ratios, but almost no change occurred at high flow-rate

ratios.

EXPERIMENTAL MATERIALS AND METHODS
Inoculum. The inoculum culture was obtained from the fermentor
effluent (day 94) of a previous dialysis continuous fermentation

(Stieber et al., 1977) originally started with L, bulgaricus 2217

 

(Chris Hanson's Laboratory, Milwaukee, Wisconsin).

Substrate. Dried deproteinized cheese whey (prepared by ultra-
filtration of whole sweet-cheese wheys by Stauffer Chemical Company,
Rochester, Minnesota) was rehydrated to contain 242 mg of lactose per
ml and was supplemented with 8 mg of yeast extract per ml. The
reconstituted whey was made up in 7.0-liter batches without steriliza-
tion, stored at 4°C, and held in a stirred, heated (60°C) reservoir to
keep the lactose in solution.

Dialysis Continuous Fermentation System. The experimental
dialysis fermentation system was conducted continuously at a temperature
of 44°C and a pH of 5.5 and operated with essentially the same equip-
ment and in the same manner as previously (Stieber et al., 1977), with
only minor changes (Stieber and Gerhardt, 1979a). Operation of the
system was interrupted once for three weeks, during which the
fermentor-circuit contents were stored at 4°C.

Analytical Procedures. Samples of the dialysate and fermentor
circuits were taken at 12 h intervals from the dialysate effluent and
from a glass "T" inserted in the tubing between the fermentor and
dialyzer, respectively. Steady-state data were determined from samples
taken at five times the cell retention time or 48 h after changing a
parameter. Since the whey was deproteinized by ultrafiltration,
optical density measurements (700 nm) could and were made on samples
as an index of cell concentration. Protein was precipitated from

8

samples with hot 5% trichloroacetic acid and measured by the method of
Lowry et al. (1951), with bovine albumen as the standard. Dry weight
measurements were made on samples by centrifuging cells from 5 ml
samples, washing the cells twice with distilled water, and drying the
cells at 105°C for at least 18 h.

EXPERIMENTAL RESULTS

Effect of Cell-Retention Time. The results of experimental

 

variation in the cell-retention time affecting the cell concentration
in the fermentor circuit showed the same form of curves (Fig. 5) as
predicted by the computer simulation (Fig. 3). The curve shapes were
similar for the three indexes of cell concentration (dry weight,
protein, and optical density).

Effect of Flow-Rate Ratio. Fig.5 also shows the results of

 

experimental variation in the flow-rate ratio (¢) affecting the cell
concentration. The results confirmed the simulation (Fig. 3) pre-
dicting that the cell concentration would increase with increasing
flowbrate ratio.

Cell Productivity. The results of experimental variation in the

 

cell-retention time affecting the cell productivity of the process
showed the cell productivity decreased with increased retention time
(Fig. 6). The results also showed that the cell productivity was
greater with increased flow-rate ratio than decreased flow-rate ratio
(¢). However, the degree to which the flow-rate ratio affected the
cell productivity depended on the cell-retention time.

Molar Growth Yield. Fig. 7 shows the results of changes in the

cell-retention time affecting the molar growth yield (Y ), which is

ATP

the number of grams dry weight of cells produced per calculated mole
of ATP generated from catabolism. The molar growth yield decreased
with increased retention time. Thus, as the retention time increased,
the bacteria increasingly used the energy obtained from catabolism for
maintenance and other nongrowth associated purposes.

The results of variation in the flow-rate ratio (¢) affecting the
YATP also are shown in Fig. 7. The molar growth yield was greater

with increased flowbrate ratio than with decreased flow-rate ratio.

But the degree to which the ratio affected the Y

ATP depended on the

cell-retention time.

Prolonged SteadyrState Operation. A dialysis continuous fermenta-

 

tion next was operated at a steady-state (Tf = 27.2, ¢ = 2.5) to
observe the cell mass behavior over an extended period. No cyclic
behavior was observed (Table 3). The cell mass decreased during the
first 5 days of the fermentation and then gradually increased for the
final 10 days. The decrease occurred in the period when the bacterial
population was adjusting to continuous culture at a relatively slow
rate of growth after having reached a high population in a batch
culture, which was necessary to start the fermentation. The increase
may have been due to decreased membrane permeability in this period
(see p. 492 in reference Stieber and Gerhardt, 1979a), which would
confirm the trend predicted by the computer simulation shown in Fig. 4.
Correlation of Simulated and Experimental Results. The simulated
results were correlated with the experimental results to evaluate the
metabolic constants of the model, demonstrate the validity of
the model, and evaluate the nature of bacterial metabolism in the
process.

10

The values used for the computer simulations are listed in Table 5.
The values for ¢ and H were calculated directly from the experimental
results. The values for um, Ks, KP, y, R, and P: were shown to be
valid previously (Stieber and Gerhardt, 1979a). The values for a and
B were obtained by successive curve fitting of the simulated results
with the experimental results. 5: was taken as 160.3 mg/ml instead of
237.5 mg/ml (which was obtained from material balances of the experi-
mental data) because a mean 32.5 percent of the fermentor contents
represented water which osmosed from the dialysate circuit (Stieber
and Gerhardt, 1979a).

In Fig. 8 are shown the computer-simulated curves obtained by
using the mathematical model (Stieber and Gerhardt, 1979a) together
with the values in Table 4 to show the effect of cell—retention time
on accumulation of cell mass. The curves fitted closely with the
superimposed points of experimental results and thereby demonstrated
the validity of the model.

The relative importance of growth and maintenance metabolism in
the fermentation was calculated from knowledge of the values for a
and 8 (Keller and Gerhardt, 1975). The calculations showed that, at
a cell-retention time of 7.6 h and flow-rate ratio of 2.5 where it was
experimentally shown that the cell productivity was greatest (Fig. 6),
77.4 percent of the substrate utilization was due to maintenance
metabolism and 22.6 percent was due to growth metabolism. Further

calculations showed that, as the cell-retention time increased, energy

requirements for maintenance increased and those for growth decreased.

11

DISCUSSION

The experimental results confirmed the computer-simulated pre-
diction that the cell mass accumulation of a dialysis continuous
process is greater than that obtained from an ordinary (nondialysis)
continuous process for the ammoniumrlactate fermentation. The
rationale is that the cell mass accumulation is directly proportional
to the feed substrate concentration which can be metabolized by the
cells. The degree of proportionality depends on the magnitude of the
inhibition effect by products. In the ammonium-lactate fermentation
of whey there is a strong product-inhibition effect. Thus, dialysis
is necessary to relieve the inhibition effect by lactic acid. As more
product is withdrawn by dialysis, the growth rate of cells becomes
higher and more substrate is consumed, which enables the use of
concentrated substrate with the resultant high cell mass concentrations.
In a fermentation with no inhibition effect by product, the present
dialysis continuous process would be unnecessary because an ordinary
continuous process then could accomplish the efficient use of con-
centrated substrate.

Figure 2 shows that, in a fermentation for biomass, it is para-
mount to select an organism with a small product inhibition constant.
At a given rate of dialysis and feed substrate concentration, an
organism with a small product inhibition constant will attain a higher
cell mass concentration than one with a high inhibition constant. The
magnitude of the metabolic constants c and B of an organism also have
a considerable effect on cell mass. The simulated predictions of cell
mass (Figs. 2 to 4) were much greater than the experimental results of
cell mass (Fig. 5) because the values for a and 8 used to make the

12

predictions were relative values and not absolute values (Keller and
Gerhardt, 1975). This comparison demonstrated the considerable effect
of a and B on the cell mass.

The experimental results confirmed the computer-simulated pre-
diction of a maximum.ce11 mass at an intermediate cell-retention time
(Fig. 5). Since the cell productivity was maximum at the shortest
cell-retention time managed, a trade-off exists in the regulation of
the cell-retention time.

The best results of the dialysis continuous process were compared
with those of a nondialysis continuous process for producing cells of

L, bulgaricus (Table 5). The dialysis process produced more than

 

twice the mass of cells per volume, was more than twice as productive
for cells (mg/ml-h), and was more efficient energywise (YATP) than
was the nondialysis process over a wide range of cell-retention times.
Further, the dialysis process used more than twice the concentration
of feed substrate (160.3 mg lactose per ml) than did the nondialysis
process (74.6 mg lactose per ml). Comparisons of cell mass accumula-
tion with concentration of feed substrate agreed with the simulated
predictions (Fig. 2).

In Table 6 results are shown which compare the percentages of
growth and maintenance metabolism of the culture when grown at similar
cell-retention times and different rates of dialysis, including
nondialysis. Interestingly, the percentage of growth metabolism
decreased and the percentage of maintenance metabolism increased
with increased dialysis (increased ¢). Also, the specific
maintenance rate (8) increased, whereas the substrate/cell ratio (a)

of the culture decreased, with increased dialysis.

13

The present dialysis continuous process is useful for production
of a relatively high cell mass accumulation in a fermentation where
metabolic products inhibit microbial growth. The process may be
especially applicable to the production of starter cultures in the
dairy industry. The continuous operation would allow automated
process control, would not require repetitive subcultures and inocula-
tions, and would yield a uniform culture. The dialysis operation
would enable the use of concentrated substrate and result in a great
accumulation of cell mass. Thus the final product would need less
concentration, which subjects the cells to the mechanical stress of
centrifugation. The major disadvantages of using the present dialysis
process for cell production are equipment complexities and cost,
membrane fouling (Stieber and Gerhardt, 1979a), and the concurrent
production of large volumes of dilute permeate. An economical process
probably would entail production of a permeate of value, e.g., in the
present study the ammonium-lactate permeate may have value as a
nitrogenous-feed supplement for ruminants or may be converted to
lactic acid for chemical use and.ammonium sulfate for use as fertilizer.
Yet higher cell accumulations in this system should easily be attained
by use of a more permeable membrane, a greater membrane area, a
greater rate of dialysis (¢), and a higher concentration of feed
substrate.

ACKNOWLEDGEMENT
This work was supported by grant ENG 76-17260 from the National

Science Foundation.

14

10

11

12

13

REFERENCES

Coton, S. G. 1976. Recovery of dairy waste. In_G. G. Birch,
K. J. Parker, and J. T. Worgan (eds.), Food from Waste, Applied
Science Publishers, Barking, England, p. 221.

Coulman, G. A., R. W. Stieber, and P. Gerhardt. 1977. Dialysis
continuous process for ammonium-lactate fermentation of whey:
mathematical model and computer simulation. Appl. Environ.
Microbiol. 34:725.

Cox, G. C., and R. D. MacBean. 1977. Lactic acid production by
Lactobacillus bulgaricus in supplemented whey ultrafiltrate.
Australian J. Dairy Technol. 32:15-22.

 

DeBoer, R., J. N. DeWit, and J. Hiddink. 1977. Processing of
whey by means of membranes and some applications of whey protein
concentrate. J. Soc. Dairy Technol. 30:112-120.

Delaney, R. A. M. 1976. Composition, properties and uses of
whey protein concentrates. J. Soc. Dairy Technol. 29:91-100.

Driessen, F. M., J. Ubbels, and J. Stadhouders. 1977. Continuous
manufacture of yogurt. I. Optimal conditions and kinetics of the
prefermentation process. Biotechnol. Bioeng. 19:821-829.

Driessen, F. M., J. Ubbels, and J. Stadhouders. 1977. Continuous
manufacture of yogurt. II. Procedure and apparatus for continuous
coagulation. Biotechnol. Bioeng. 19:841-851.

Erdman, M. D., W. G. Bergen, and C. A. Reddy. 1977. Amino acid
profiles and presumptive nutritional assessment of single-cell
protein from certain lactobacilli. Appl. Environ. Microbiol.
33:901-905.

Gerhardt, P., and C. A. Reddy. 1978. Conversion of agroindustrial
wastes into ruminant feedstuff by ammoniated organic acid
fermentation: a brief review and preview. Dev. Ind. Microbiol.
19:71-78.

Hargrove, R. E., F. E. McDonough, D. E. LaCroix, and J. A. Alford.
1976. Production and properties of deproteinized whey powders.

Juengst, F. W., Jr. 1979. Use of total whey constituents--
animal feed. J. Dairy Sci. 62:106-111.

Keller, A. K., and P. Gerhardt. 1975. Continuous lactic acid
fermentation of whey to produce a ruminant feed supplement high
in crude protein. Biotechnol. Bioeng. 17:997-1018.

Lane, A. G. 1977. Production of food yeast from whey ultrafiltrate
by dialysis culture. J. Appl. Chem. Biotechnol. 27:165-169.

15

14

15

16

17

18

19

20

21

22

23

24

25

26

27

Lewis, P. M. 1967. A note on the continuous flow culture of
mixed populations of Lactobacilli and Streptococci. J. Appl.
Bact. 30:406-409.

 

 

Lowry, O. H., N. J. Rosebrough, A. L. Farr, and R. J. Randall.
1951. Protein measurement with the folin phenol reagent. J. Biol.
Chem. 193:265-275.

Marshall, K. R., and R. L. Earle. 1975. The effect of agitation
on lactic acid by Lactobacillus bulgaricus in a casein whey medium.
N.Z. J. Dairy Sci. Technol. 10:123-128.

 

Mathur, B. N., and K. M. Shahani. 1979. Use of total whey
constituents for human food. J. Dairy Sci. 62:99-105.

Moor, C. V. 1976. Whey protein concentrates: an update. Food
Technol. 30:18-42.

Ohmiya, K., H. Ohashi, T. Kobayashi, and S. Shimizu. 1977.
Hydrolysis of lactose by immobilized microorganisms. Appl.
Environ. Microbiol. 33:137-146.

Osborne, J. W. 1977. Production of frozen concentrated cheese
starters by diffusion culture. J. Soc. Dairy Technol. 30:40-44.

Pirt, S. J. 1975. Principles of microbe and cell cultivation.
Halsted Press, John Wiley & Sons, Inc., New York, New York, p. 29-
41.

Reddy, C. A., H. E. Henderson, and M. D. Erdman. 1976. Bacterial
fermentation of cheese whey for production of a ruminant feed supple-
ment rich in crude protein. Appl. Environ. Microbiol. 32:769-776.

Schingoethe, D. J. 1976. Whey utilization in animal feeding: a
summary and evaluation. J. Dairy Sci. 59:556-570.

Stieber, R. W., G. A. Coulman, and P. Gerhardt. 1977. Dialysis
continuous process for ammonium-lactate fermentation of whey:
experimental tests. Appl. Environ. Microbiol. 34:733-739.

Stieber, R. W., and P. Gerhardt. 1979a. Dialysis continuous
process for ammonium-lactate fermentation: improved mathematical
model and use of deproteinized whey. Appl. Environ. Microbiol.
37:487-495.

Stieber, R. W., and P. Gerhardt. 1979b. Continuous ammonium-
1actate fermentation of deproteinized whey. J. Dairy Sci. In_

press.

Wingerd, W. H. 1970. Lactalbumin as a food ingredient. J.
Dairy Sci. 54:1234-1236.

16

28 Wingerd, W. H., S. Saperstein, and L. Lutwak. 1970. Whey protein
concentrate has excellent nutritional properties. Food Technol.
24:758-764.

17

TABLE 1. Glossary of mathematical symbols

 

 

Symbol Description Units or Definition
A.In Area of membrane available for dialysis cu2
Fd Flow rate into and out of dialysate circuit ml/h
Ff Flow rate into and out of fermentor circuit ml/h
K? Product inhibition constant mg/mq
Ks Substrate limitation constant mg/ml
Pd Product concentration in dialysate circuit mg/ml
P: Product concentration in fermentor feed mg/ml
Pf Product concentration in fermentor circuit mg/ml
mp Permeability of membrane to product mg/cmz-h
Pms Permeability of membrane to substrate mg/cmz-h
Sd Substrate concentration dialysate circuit mg/ml
S: Substrate concentration in fermentor feed mg/ml
sf Substrate concentration in fermentor circuit mg/ml
Xf Cell-mass concentration in fermentor circuit mg/ml
ATP Molar growth yield mg/mmole
a Substrate/cell ratio mg/mg
8 Specific maintenance rate h-1
Y Product/substrate ratio mg/mg
um Maximum specific growth rate of cells h"1
Tf Cell-retention time in fermentor circuit h
R Ratio of product/substrate membrane permeabilities = Pmp/Pms
9 Time factor - umTf
H Membrane permeability factor = PmSAm/Ff
¢ Flow-rate ratio - Fd/Ff

 

18

 

 

 

om. ~.o ~.m oma H.m o.m nu :. nu mo. 6.0 mm. m
mm. «.6 ~.~ oma H.m o.m o.H u- u: mo. 0.0 mm. m
mm. «.6 ~.~ omH H.m o.m o.H o.H o.oH mo. us mm. a
Ams\mse range Ams\mse Aas\msa Aas\ms. are Aasxmsv Ams\mee Aanr.
m m
> m 5 mm we m e e me x x a: museum
mcofiumucmshmm msozcwucoo mam>HMHo mo meowuowooum ooumdzaamluouzmaoo new con: mosam> .N mqmda

19

TABLE 3. Cell mass produced during prolonged operation of
the dialysis continuous fermentation at steady-state conditions

 

 

Day Cell-retention Flow-rate Cell mass,

time, Tf (h) ratio, ¢ Xf (mg/ml)
57 25.5 2.29 3.04
24.2 2.26 4.44
25.7 2.36 4.06
58 28.2 2.69 3.72
27.4 2.62 3.36
59 26.4 2.46 3.56
27.4 2.60 3.00
60 27.9 2.66 3.74
27.5 2.62 2.78
61 27.0 2.53 2.98
27.0 2.53 2.04
62 26.9 2.53 2.62
26.3 2.46 2.96
63 27.1 2.55 3.06
27.7 2.61 2.66
64 27.6 2.61 2.88
27.6 2.58 2.42
65 27.2 2.53 2.72
27.4 2.53 2.92
66 27.7 2.58 3.28
27.8 2.59 3.28
67 28.2 2.61 3.00
27.9 2.60 3.28
68 27.8 2.58 3.28
27.6 2.53 3.62
69 27.5 2.54 3.46
27.2 2.49 3.54
70 27.0 2.47 3.42
27.5 2.53 3.58
71 27.3 2.52 3.72
mean 27.2 2.54 3.21

 

20

 

om. mm..n c.m m.C©H H.m o.m v. m.m co. N.N mm. mVIov

 

om. om.a o.a m.ooa H.m o.« m. o.H mo. ~.~ mm. omnom
ras\mss inure Amexoss Aaexmsc rHs\ms. Aasxms. Ams\mss reuse
a
> m 5 mm we m = e me x an S8

 

muazmmu Housmaaummxm sue: muasmmu poundsawm wumHmHHoo on com: monam> .o mqmdfi

21

.oovv I ouduouomsou one .m.m I ma

.Ha\ma m I uomuuxm ammo» .Ha\ms o.eh n mucuoea omou sue: Aomnmav uoumnuoo one nanowum mo muazmamua

.m.~ n owumu cumulsloc

 

 

v.” mo. m.a h.o~ o.m ma. m.m m.n~

h.m NH. m.a v.0a o.m AN. 0.? m.ma

h.n we. h.H ~.oa m.v mm. n.m «.ma

~.v ow. m.~ n.m m.v mm. w.m. . v.oa

o.m mm. o.H v.o m.m an. v.m m.»
3-385 «Exes. 2. 05» 2-3me 2595 c: .95...

near >ua>auosvoum Hauo 9:0«93 Sun scaucououuanoo mac» >uw>auosooum Haou unowoz mud cOwuccuouuaaou

 

mmmoomm mHmathDzoz

 

mmuoomm mHmrA<Ho

 

magma Howuouomn saw sceuMucoEHou mzoscwucoo ecu mo muasmou ucmamSchococ can unamaamwo wo canwuomaoo .m mqm<a

22

 

mm.H o.m
ON.“ o.h
oo.H o.HH

.uausaaao.

0.5 m.~
Amausaaacs
«.5. o.“

382 £3.68 93
wonowum oucououou
ea .mam>ndwococ.
v.o o.o

 

3. 3:339: 3. Smeuoneuos

oucecoucwo: occacmucwoa oauwoomm

Ama\oa.a .owumu
Haoo\mumuumnsm

.H-gsa .ouuu

o .o«uau
ounuuaoam

.5 as .33
sawucmuoulaaou

 

«mafia noducouounaaoo noddedm no can mauhaeao uo uouou uncuouwwo no amaaonmuoe ousuaso mo comwuumaoo

.0 mgmdfi

23

5I°vpf°v PI “2°. ‘0

  

F" "1
|
l
I
l

 

u- .341
5: - PI vxf v"! 5a- pa: Fa
¥——w-'——-" T— 4
FERMENTOR DIALYSATE
CIRCUIT CIRCUIT

FIG. 1. Schematic of dialysis continuous fermentation system
(reprinted with permission from reference Coulman et al., 1977).

Symbols are described in Table l.

24

 

 

 

 

60 I I I I
Kp=0

7:: 40- 0.5 -
B
E LG
‘5.
X 20- ‘

O I l I l I

O 50 I00 ISO 200 250

Sfo (mg/ml)

FIG. 2. Simulated effects of changes in the feed substrate
concentration (5;) on the cell mass concentration in the fermentor
circuit (Xf) at three values of the product-inhibition factor

(K ).
P

25

 

.b
O
-I

 

 

X f (mg/ml)
":3

 

 

 

 

 

FIG. 3. Simulated effects of changes in the cell-retention
time (Tf) on the cell mass concentration in the fermentor circuit

(Xf) at two values of the flow-rate ratio (¢).

26

3O

 

3’0 I I I

 

 

 

1'I=I.o

 

 

 

FIG. 4. Simulated effects of changes in the flow-rate ratio
(¢) on the cell mass concentration in the fermentor circuit (Xf)

at two values of the membrane permeability factor (H).

27

Xf (mg/ml)

PROTEIN (mg/ml)

O. 0.

FIG.

optical density (0.D.).
and during days 26 to 36.

and during days 40 to 49.

 

 

 

 

 

 

 

 

 

 

 

 

I I I I I
IAN-I1
/ ’F§ ‘ ‘ ‘5
I— /’.’ a ~I__====.-_
é,
I
II, I L 1 L
If I r I I
_ -l"-I~- _
f ~~..__ —____.__
’ ~———-.-—I
I
All I L 1 1
I :It I I I
‘
’ ‘ ‘1
’ ‘\.~ ‘*~
._ / ~~~--...
0' ‘1>--%
I
ill 1 1 1 1
O 5 IO I5 20 25 30
T, (h)

5. Experimental effects of changes in the cell-retention
time (Tf) and the flow-rate ratio (¢) on the cell concentration in

the fermentor circuit as determined by dry weight (xf), protein, and

the simulated predictions (Fig. 3).

28

The circle points were obtained at o = 1.0
The square points were obtained at o = 2.5

Interpretations of the curves were based on

 

.0
m

 

 

 

 

 

3.5
'E'
‘\. C.
a, 0.4 ..
E
9 a

O I l l l L

0 5 I0 I5 20 25 30

FIG. 6. Experimental effects of changes in the cell-retention
time (Tf) and the flow-rate ratio (¢) on the cell mass productivity
(rg) of the dialysis continuous process. The circle points were
obtained at ¢ = 1.0 and during days 26 to 36. The square points were

obtained at ¢ = 2.5 and during days 40 to 49.

29

 

 

 

 

 

6 I
13
O
E
E 4— —
‘\~
C3)
.5.
a. 2— ..
I'—
)5!
c) L I l l l
O 5 IO I5 20 25 30

FIG. 7. Experimental effects of changes in the cell-retention

time (Tf) and the flow-rate ratio (¢) on the molar growth yield (YATP)’

1.0 and during days 26 to 36.

The circle points were obtained at ¢

2.5 and during days 40 to 49.

The square points were obtained at ¢

30

 

Xf (mg/ml)

 

J)
I

 

 

 

 

O 5 IO IS 20 25 30
mm

FIG. 8. Simulated effects of changes in the cell-retention time
(Tf) and the flow-rate ratio (¢) on the cell mass concentration in the
fermentor circuit (Xf). The curves were plotted by use of the
mathematical model and the values in Table 4. The points were
replotted from Fig. 4 to demonstrate the close fit between the

experimental results and the simulated results.

31

123

ARTICLE V

CONTINUOUS PROCESS FOR AMMONIUM-LACTATE

FERMENTATION OF DEPROTEINIZED WHEY

BY

R. W. Stieber and Philipp Gerhardt

Accepted by the Journal of Dairy Science.

Continuous Process for Ammonium-Lactate Fermentation
of Deproteinized Wheyl

R. W. STIEBER and PHILIPP GERHARDT

Department of Microbiology and Public Health,
Michigan State University,
East Lansing, Michigan 48824

ABSTRACT

A mathematical model was developed to describe a process for the
continuous fermentation of lactose to lactic acid, with neutralization
to a constant pH by ammonia. The equations for the steady state were
used to computer-simulate the effect of cell-retention time (Tf) on
substrate, product and cell-mass concentrations in a single-stage
system. The theoretical predictions then were validated by laboratory
experiments in which reconstituted and supplemented deproteinized whey

was continuously fermented by an adapted culture of Lactobacillus

 

bulgaricus at pH 5.5 and 44 C, without sterilization or asepsis.

 

Lactate concentration was maximum (58.7 mg/ml) at Tf = 27 h, and cell

mass was maximum (1.9 mg/ml) at Tf = 9.3 h. The single-stage con-
tinuous fermentation of deproteinized whey as compared to that of
whole whey, previously described, resulted in similarly high extents
of product accumulation and of substrate utilization at similarly high
conversion rates. The fermentation of deproteinized whey also was

simulated for operation with two stages in series at various retention-

time ratios, the results of which predicted that the combined

 

1Journal article number 8854 from the Michigan Agricultural
Experiment Station.

retention time could be reduced to 17 h with results as good as that
with the single-stage operation.
INTRODUCTION

The whey residues of cheese and casein manufacture continue to be
a disposal problem for the dairy industry. These residues can be
processed by ultrafiltration to obtain concentrates of milk protein
(Coton, 1976) of high quality and considerable market potential
(Wingerd, 1970; Wingerd et al., 1970; Schingoethe, 1976; DeBoer et
al., 1977; Morr, 1976; Delaney, 1976: Mathur and Shahani, 1979). The
remaining residue of deproteinized whey retains all of the lactose,
however, and a substantial disposal problem still remains.

A potential solution to the problem of economically using whole
or deproteinized whey lies with conversion into nitrogen-enriched
feedstuff for ruminant animals, accomplished by the fermentation of
the lactose into lactic acid and its neutralization by ammonia (Gerhardt
and Reddy, 1978; Juengst, 1979). The fermentation can be managed as a
batch process (Marshall and Earle, 1975; Reddy et al., 1976), a con-
tinuous process (Keller and Gerhardt, 1975; Cox and MacBean, 1977), or
a dialysis continuous process (Coulman et a1., 1977; Stieber et al.,
1977; Stieber and Gerhardt, 1979b). The continuous processes can be
managed easily, nonaseptically and indefinitely to produce a uniform
product at a useful concentration.

The present study was undertaken to evaluate the use of deprotein-
ized whey instead of whole whey as substrate for the continuous
fermentative conversion of the lactose to ammonium lactate. Further-
more, cell-mass was studied for use as a product in itself because of
its high nutritional quality (Erdman, et al., 1977). The investigative

2

procedure was first to deve10p a mathematical model and use it to
simulate a single-stage process for the fermentation. Laboratory
experiments then were conducted to test the theoretical predictions.
The simulated and experimental results were correlated to evaluate
the process, establish values for bacterial metabolic constants, and
demonstrate the validity of the model. The validated model finally
was used to predict the even greater usefulness of a two-stage process
for the fermentation.
THEORY

Design of Fermentation System

Figure 1 shows a schematic of a completely continuous fermenta-
tion system designed for the ammonium-lactate fermentation of whey.
The symbols correspond to those used previously (Stieber and Gerhardt,
1979b) and are listed in Table l. The feed into the fermentor, which
contains the substrate (lactose) in relatively high concentration
(5:), is maintained at a relatively low rate of flow (Ff) consistent
with a minimal concentration of unused substrate in the fermentor
effluent (Sf). The substrate is converted essentially only to a
single product (lactic acid) in the fermentor effluent (Pf). A small
concentration of product preexists in the feed (P2). The liquid volume
(Vf) and the cell population (Xf) in the fermentor are maintained at
constant levels. The addition of ammonia solution is not shown because
it does not enter into the theoretical considerations.

For purposes of mathematical modeling, the assumptions were made
that: a high rate of mixing insures homogeneity throughout the
system; the bacterial metabolic values (um, KS, KP, y, a, and B as

identified in Table 1) remain unchanged for a useful period of time;

3

and the rate of ammonia solution addition is negligible relative to
that of substrate feed.
Generalized Mathematical Model

Material-balance equations for the fermentation system were
formulated as follows: input + production = output + accumulation.
Equations for substrate, product and cell mass in the fermentor are

respectively as follows:

0 dsf
1?fo + Vfrs = 5‘fo + V571? (1)
F p° + v r = F p + v'c-13$ (2)
f f f p f f ffdt
o dxf
F x -+ v r . F x + v ——— (3)

f f f g f f ffdt

Rate equations for substrate utilization, product formation and
cell growth were the same as those previously found to be valid

(Stieber and Gerhardt, 1979b) and are respectively as follows:

 

rp 5' "Yrs (5)
I .f

r ' Wu (6)

9 Ks + sf + rppfjxf

The rate equations were combined with the material-balance

equations, and the variables were defined in dimensionless parameters

(Table 2), to obtain a generalized model for the continuous
fermentation. The resulting equations for substrate, product and

cells are respectively as follows:

 

dsf Sf 8T
dt=‘9 _ __]+—f-§-§ +1/T (7)
. Ks + sf + x p f f f

 

 

 

p f a
de Sf 8T
at a 6E +‘ “']+—f§f-Ff+§§/Tf (8)
S Sf+Kpr 0’
dx ‘E
f I -— -
—' 6[_ __ '1X+x°rr 9
a. [K.s.i31 f f’f "
s f p f
4
where Tf = Vf/Ff and is an operational parameter.

Generalized Steady-State Solution

The equations of the generalized model were rearranged and the
time derivatives were set at zero to obtain a generalized solution for
the steady state. The resulting equations for substrate, product and

cell mass are respectively as follows:

—- —-° — — _

sf = (bef + KP + KS)x/(e - x + KPX) (10)
_ -0 _. _

Pf =- [(1 + Pf) (6 - X) - KSX]/(6 - X + KpX) (11)
If =- (l - E} + Ftp/I1 + Be/(aumn (12)

where X = (if - Eta/if = BEE/(Ks + Sf + EDP-f) .
99mputer Simulated Predictions

The steady-state solution was used to simulate the continuous
fermentation of deproteinized whey into ammonium-lactate and cell
mass, by programming on a digital computer. Because the simulations

were performed before actual experimental fermentations were

conducted, approximations of various conditions were used at first.
The resulting simulations for substrate utilization, product formation
and cell-mass accumulation were similar to those described previously
(Keller and Gerhardt, 1975; Coulman et al., 1977; Stieber and Gerhardt,
1979b) and predicted the conditions used for the following experimental
tests.
EXPERIMENTAL MATERIALS AND METHODS

Inoculum

The inoculum, obtained from the fermentor effluent (day 71) of a
previous dialysis continuous fermentation (Stieber and Gerhardt,
1979b), was grown at 44 C for 12 to 24 h in a sterile medium of 10%
(wt/vol) nonfat milk powder and 90% tap water, stored at 4 C, and
transferred biweekly to fresh medium. The original culture,
Lactobacillus bulgaricus 2217 (Chris Hanson's Laboratory, Milwaukee,
WI), had been selected (Reddy et a1., 1976) on the basis of its
high rate of acid production in the pH range of 5.0 to 6.0.
Substrate

Dried deproteinized cheese whey (prepared by ultrafiltration;
Stauffer Chemical Company, Rochester, MN), used as the fermentation
substrate, was rehydrated to contain 62 mg of lactose per ml and was
supplemented with 2 mg of yeast extract per ml. The reconstituted
whey was made up in 7.5-liter batches without sterilization,
stored at 4 C, and held for as long as 10 h in a stirred, heated
(78 C) reservoir during use.
Continuous Fermentation System

The fermentation was conducted continuously be metering the
whey from a feed reservoir into a fermentor and removing overflow

6

into a product reservoir, thus maintaining a constant volume in the
fermentor. The fermentor contents were mixed thoroughly by stirring
and baffling. Nitrogen was continuously purged into the fermentor
(1500 ml per min) to force the overflow into the reservoir.
Concentrated NH4OH solution (58%) was automatically fed into the
fermentor to maintain a constant pH.

The fermentation was conducted in a 5-1iter modular fermentor
with automatic temperature control (Microferm model MF 105, New
Brunswick Scientific Co., New Brunswick, NJ) and automatic pH
control (model pH-40, New Brunswick Scientific). Contents of the
fermentor (2.0 to 3.6 liters) were maintained at a temperature of
44 C i 0.5 C and at pH 5.5 i 0.05. A plunger-type reciprocating
pump (type "P", Bran and Lubbe, Inc., Evanston, IL) was used to
meter the flow of whey into the fermentor. The flow rate of ammonia
solution and the flow rate of the fermentor effluent were determined
at least thrice daily. The system was operated continuously for
17 days.

Analytical Procedures

 

Samples were taken at least every 8 h from the fermentor
effluent. Steady-state data were determined from samples taken at
five times the cell-retention time or 60 h after changing a parameter.
Lactose in the samples was determined by the colorimetric method of
Morris (1948). Lactic acid was determined by use of a gas
chromatograph (series 1420, Varian Associates, Palo Alto, CA)
equipped with an integrator (model CDS lll, Varian Associates),

a thermoconductivity detector, and a stainless steel column

(6 ft x l/8 in 0.D.) packed with 10% SP-1000/1% H PO on 100/120

3 4
7

Chromosorb WAW (Supelco, Inc., Bellefonte, PA). The samples were
prepared by the procedure of Holdeman and Moore (1975).
Dry weight measurements were made by centrifuging cells from
10 m1 samples, washing the cells twice with distilled water, and
drying the cells at 105 C for at least 18 h.
EXPERIMENTAL RESULTS

Conversion Efficiency and Product Accumulation

 

Figure 2 shows the experimental results of variation in the
cell-retention time (Tf) affecting the residual lactose (Sf) and the
accumulation of lactate (Pf) in the fermentor with one-stage
operation. The lactose concentration decreased and the ammonium-
lactate concentration increased with increased retention time.

Cell Concentration

 

Figure 3 shows the results of changes in the retention time
affecting the cell-mass concentration in the fermentor. With increas-
ing retention time where growth metabolism.predominated, the cell con-
centration at first increased rapidly to a maximum. But beyond a
critical retention time where maintenance metabolism superseded
growth metabolism, the cell concentration decreased gradually.

Molar Growth Yield

 

The results of variation in the retention time affecting the
molar growth yield (YATP' the grams dry weight of cells produced per
calculated mole of ATP generated from catabolism) are shown in
Figure 4. The molar growth yield decreased with increased retention
time; i.e., as the retention time increased, the bacteria increasingly
used the energy obtained from catabolism for maintenance and other

purposes not associated with growth.

8

Lactate and Cell Productivities and Lactose Conversion

 

Figure 5 shows that the lactate productivity and the cell
productivity decreased with increased retention time, whereas the
percentage of lactose converted to product increased with increased
retention time. High lactate and cell productivity, high substrate
conversion, and high product concentration are all desirable.
Consequently, trade-offs in the regulation of the retention time
would be required in designing an operational process.

Product Quality

 

Samples were regularly analyzed by gas chromatography not only
to determine the concentration of lactic acid but also to monitor
the possible presence of atypical metabolic products. The results
indicated that the fermentation remained homofermentative, i.e.,
only negligible amounts of products other than lactic acid were
formed.

Correlation of Experimental Results With Simulated Predictions

 

The experimental results were correlated with the simulated
predictions to evaluate the metabolic constants of the mathematical
model, demonstrate the validity of the model, and evaluate the nature
of bacterial metabolism in the process.

The values used for the metabolic constants in the model are
listed in Table 3. The values for um, KS, KP, a and 8 were obtained
by successive fitting of simulated curves with experimental results.

Because the feed reservoir was maintained at 78 C (causing evapora-

o
f

lactose per ml, this value was used instead of 62 mg/ml (which was

tion) and because material balances showed S to be 74.6 mg of

measured before the substrate was heated). The value for Y was
evaluated from the experimental results.

Figure 2 shows the computer-simulated curves obtained by using
the values in Table 3 in the mathematical model to describe the
relationships between the retention time (Tf) and the conversion of
lactose (Sf) and accumulation of lactate (Pf). Figure 3 shows the
relationship between the Tf and the cell mass (Xf). The curves in
both figures fitted with the superimposed points of experimental
results and thereby demonstrated the validity of the model. With
the same values for um, KS, and KP, moreover, simulated predictions
correlated with experimental results also in a dialysis continuous
process (Stieber and Gerhardt, 1979a; Stieber and Gerhardt, 1979b),
additionally validating the model.

The values for Ks and Kp (Table 3) showed that the fermentation
was not affected by substrate limitation but was greatly limited by
increasing concentrations of product. The relative importance of
growth and maintenance metabolism in the fermentation were calculated
from knowledge of the values for a and 8 (Keller and Gerhardt, 1975).
At a retention time of 6.4 h, where it was experimentally shown that
the cell productivity was greatest (Fig. 5), 63.2 percent of the
substrate utilization was calculated to be due to growth metabolism
and 36.8 percent to maintenance metabolism. However, at a retention
time of 26.7 h, where the lactate concentration was highest (Fig. 2),
29.1 percent of the energy of catabolism was calculated to be due to
growth and 70.9 percent to maintenance.

Similar calculations, when correlated with the Y values

ATP
(Figure 4) and extrapolated to 100 percent, resulted in a mean

10

Y??? of 8.0; this value, which describes the maximal cell mass
produced per mole of ATP generated from catabolism, agrees with values
reported by Forrest and Walker (1971). The product/substrate ratio
(y) of the fermentation was .96 mg/mg, i.e., 96 percent of the lactose
utilized was converted to lactate and 4 percent was incorporated

into bacterial cells. On a substrate efficiency basis, the fermenta-
tion thus would be better for production of lactate than of bacterial

cells.

Simulated Two-Stage Operation

 

Keller and Gerhardt (1975) improved the conversion efficiency for
the continuous fermentation of whole whey by the employment of two
equal fermentors in series. In this system the effluent of the first
fermentor, containing residual lactose (Sf), lactate (Pf) and cell
mass (Xf), flows directly into the second fermentor. With the
values for the metabolic constants in Table 3 and the steady-state
solution, the two-stage operation was computer simulated for the
fermentation of deproteinized whey. The effluent concentrations
(Sf, Pf, and Xf) from the first fermentor were used as the feed
concentrations (8:, PE, and X?) into the second fermentor. Also, it
was assumed there was no lag in growth of the cells entering the
second fermentor.

Figure 6A shows the simulated effects of changes in the ratio
of the retention time of the second fermentor (sz) to that of the

first fermentor (Tfl) on the total retention time (T T )

f1+ f2
required to achieve 60 mg of lactate per ml in the effluent of the

11

second fermentor. The shortest time required was 17 h at Tf2/Tf1
1.0 (8.5 h for each stage), which compared to about 26 h for the

single-stage operation (T 0).

fZ/Tfl =

Figure 6B shows the simulated effects of changes in TfZ/Tfl
on the accumulation of cell mass in the effluent from the second
fermentor. The cell mass was greatest in a two-stage operation with
a total retention time of 17 h (8.5 h for each stage) and lowest in
the single-stage operation. Thus, two stages provided the optimal
conditions for increased cell mass, resulting in an increased rate
of converting lactose to lactate.

DISCUSSION

A continuous process for the ammonium-lactate fermentation was
previously studied in our laboratory using unsupplemented whole whey
(Keller and Gerhardt, 1975). A comparison of the prior results
with whole whey and the present results with deproteinized whey, each
operated as a single-stage fermentation and at comparable retention
times, is shown in Table 4. Use of the two substrates resulted in a
similar extent of lactate accumulation and rate of lactose utilization.
The percentage of lactose conversion was greater with whole whey than
with deproteinized whey; however, the latter was more concentrated in
the feed, making the comparison difficult. The concentration of 59 mg
of lactate per ml in the effluent from the fermentor probably was near
the maximum which a continuous culture without dialysis can tolerate,
thus limiting the extent of substrate conversion. Liquid deproteinized
whey, before concentration or drying, contains about 46 mg of lactose
per ml, which should be capable of complete conversion to product by

the continuous process. The rate of lactose utilization with whole

12

whey was similar to that with deproteinized whey, even though the
latter was supplemented with yeast extract. Moreover, the inoculum
for the latter fermentation was obtained from a culture with a 163 day
history of continuous ammonium-lactate fermentation and thus was well
adapted for the fermentation. With these differences taken into
account, whole whey probably is a somewhat better substrate for the
fermentation but either is practicable. With whole or deproteinized
whey, continuous operation of the fermentation is of simple operation,
can be managed for prolonged periods of time (perhaps indefinitely),
does not require sterilization or asepsis, and produces a uniform
product at useful concentration. The ammonium-lactate fermentation is
one of the few processes where continuous operation appears more
practicable than the conventional batch operation.

Mathematical modeling and computer simulation provide powerful
but sometimes unappreciated tools for predicting the results to be
expected in a continuous fermentation process. Laboratory experiments
need be conducted only to validate the predictions by using a rela-
tively limited number of changes at preselected critical points. The
experimental results in turn are used to establish constants and
perhaps to indicate additional terms in the mathematical equations.
The model can then be further used to evaluate the results. By this
process of successive theoretical prediction and experimental vali-
dation, the model becomes increasingly accurate and useful to predict
how the fermentation should be conducted for accomplishing practical
objectives.

The rate equation for substrate utilization in the present mathe-

matical model is the same as in a prior one (Keller and Gerhardt,

13

1975). However, the prior model requires knowledge of the relative
values of a and B (i.e., the ratio of a to 8) to simulate substrate
utilization and product formation, Whereas the present model does not.
The absolute values of these constants are required to simulate cell
mass concentration with either model. In both, the relative values
for the constant were the same (11.0).

The present mathematical model for the fermentation is simple, has
a sound biological basis (Stieber and Gerhardt, 1979b), and has values
for constants which are realistic. The model was particularly useful
for evaluating the metabolic constants of the culture, which were
helpful in simulation studies of a dialysis continuous process (Stieber
and Gerhardt, 1979b) and also in predictions of the usefulness of a two-
stage dialysis or cell-recycle process for the fermentation (Stieber
and Gerhardt, unpublished). The validity of the model was demonstrated
by the correlation of simulated predictions with experimental results in
both the prior studies of continuous fermentation with dialysis (Stieber
and Gerhardt, 1979b) and the present studies without dialysis. Con-
sequently, the simulated predictions of the two-stage process (Fig. 6)
should be realized experimentally. Furthermore, the model should be
useful for on-line control and optimization of such a process for
industrial use.

ACKNOWLEDGMENT
This work was supported by grant ENG 76-17260 from the National

Science Foundation.

14

10

11

12

13

REFERENCES

Coton, S. G. 1976. Recovery of dairy waste. I2_G. G. Birch,
K. J. Parker, and J. T. Worgan (ed.), Food from Waste, Applied
Science Publishers, Barking, England, p. 221.

Coulman, G. A., R. W. Stieber, and P. Gerhardt. 1977. Dialysis
continuous process for ammonium-lactate fermentation of whey:
mathematical model and computer simulation. Appl. Environ.
Microbiol. 34:725.

Cox, G. C., and R. D. MacBean. 1977. Lactic acid production by
Lactobacillus bulgaricus in supplemented whey ultrafiltrate.
Australian J. Dairy Technol. 32:19.

 

DeBoer, R., J. N. DeWit, and J. Hiddink. 1977. Processing of
whey by means of membranes and some applications of whey protein
concentrate. J. Soc. Dairy Technol. 30:112.

Delaney, R. A. M. 1976. Composition, properties and uses of
whey protein concentrates. J. Soc. Dairy Technol. 29:91.

Erdman, M. D., W. G. Bergen, and C. A. Reddy. 1977. Amino acid
profiles and presumptive nutritional assessment of single-cell

protein from certain lactobacilli. Appl. Environ. Microbiol.
33:901.

Forrest, W. W., and D. J. Walker. 1971. The generation and
utilization of energy during growth. Adv. Microbiol. Physiol.
5:213.

Gerhardt, P., and C. A. Reddy. 1978. Conversion of agroindustrial
wastes into ruminant feedstuff by ammoniated organic acid fermen-
tation: a brief review and preview. Dev. Ind. Microbiol. 19:71.

Holdeman, L. V., and W. E. C. Moore. 1975. Anaerobe Laboratory
Manual. VPI Anaerobe Laboratory, Virginia Polytechnic Institute
and State University, Blacksburg, Virginia, p. 113.

Juengst, F. W., Jr. 1979. Use of total whey constituents -
animal feed. J. Dairy Sci. 62:106.

Keller, A. K., and P. Gerhardt. 1975. Continuous lactic acid
fermentation of whey to produce a ruminant feed supplement high
in crude protein. Biotechnol. Bioeng. 17:977.

Marshall, K. R., and R. L. Earle. 1975. The effect of agitation
on the production of lactic acid by Lactobacillus bulgaricus

in a casein whey medium. New Zealand J. Dairy Sci. Technol.
10:123.

 

Mathur, B. N., and K. M. Shahani. 1979. Use of total whey
constituents for human food. J. Dairy Sci. 62:99.

15

14

15

16

17

18

19

20

21

22

Morr, C. V. 1976. Whey protein concentrates: an update. Food
Technol. 30:18.

Morris, D. L. 1948. Quantitative determination of carbohydrates
with Dreywood's anthrone reagent. Science. 107:254.

Reddy, C. A., H. E. Henderson, and M. D. Erdman. 1976. Bacterial
fermentation of cheese whey for production of a ruminant feed
supplement rich in crude protein. Appl. Environ. Microbiol.
32:769.

Schingoethe, D. J. 1976. Whey utilization in animal feeding:
a summary and evaluation. J. Dairy Sci. 59:556.

Stieber, R. W., G. A. Coulman, and P. Gerhardt. 1977. Dialysis
continuous process for ammonium-lactate fermentation of whey:
experimental tests. Appl. Environ. Microbiol. 34:733.

Stieber, R. W., and P. Gerhardt. 1979. Improved models of
dialysis and nondialysis continuous processes for ammonium-
lactate fermentation. Proc. 2nd Internat. Conf. Computer Appl.
Ferm. Technol. Ig_press.

Stieber, R. W., and P. Gerhardt. 1979. Dialysis continuous
process for ammonium-lactate fermentation: improved mathematical
model and use of deproteinized whey. Appl. Environ. Microbiol.
34:487.

Wingerd, W. H. 1970. Lactalbumin as a food ingredient. J. Dairy
Sci. 54:1234.

Wingerd, W. H., S. Saperstein, and L. Lutwak. 1970. Whey

protein concentrate has excellent nutritional properties.
Food Technol. 24:758.

16

{HABIJB 1.

Glossary of Mathematical Symbols

 

 

Symbol Description Units
Ff Flow rate into and out of fermentor ml/h
KP Product inhibition constant mg/mg
Ks Substrate limitation constant mg/ml
P: Product concentration in feed mg/ml
Pf Product concentration in fermentor mg/ml
rg Rate of cell growth mg/ml-h
rp Rate of product formation mg/ml-h
-rs Rate of substrate utilization mg/ml-h
S? Substrate concentration in feed mg/ml
Sf Substrate concentration in fermentor mg/ml
t Time h
vf Volume of liquid in fermentor ml
x: Cell-mass concentration in feed mg/ml
xf Cell-mass concentration in fermentor mg/ml
YATP Molar growth yield mg/mmole
a Substrate/cell ratio mg/mg
B SpeCific maintenance rate h-l
Y Product/substrate ratio mg/mg
E Efficiency of lactose conversion t
um Maximum specific growth rate of cells h.1
Tf Retention time of cells in fermentor h

 

17

TABLE 2. Glossary of Dimensionless Parameters

 

 

Type Symbol and Definition Description
Material parameters P: = pg/(ysg) Product factor in feed
P} = Pf/(YS:) Product factor in fermentor
SE = Sf/S: Substrate factor in fermentor
i: B axg/S: Cell factor in feed

Xi = on/Sf Cell factor in fermentor

Kinetic parameters E; = YKP Product-inhibition factor
E; = Ks/S: Substrate-limitation factor
6 - umTf Time factor

 

18

TABLE 3. Values Used to Correlate Simulated
Predictions with Experimental Results

 

 

Symbola Value
K 2.3 m /m
p 9 9
Ks 0.07 mg/ml
P: 2.8 mg/ml
o
Sf 74.6 mg/ml
X0 0 mg/ml
f
a 11.0 mg/mg
B 1.0 h’1
Y 0.96 mg/mg
u 0.35 h'1
m

 

aSymbols are described in Tables 1 and 2.

19

 

uonuuxo anew»

 

A.nh oa.~ h.mm ~.H~ m.mh n.0m no «S\ma N sue:
83 om.~ 6...... Cam 6.2. «.3 s2: 8.253083
Amped
.uoumzuoo use Madamxv
scauaucoauamman
m.om vn.a mm m.o me o.am usonuaa
«.3 Sun S o; 2. 0.2 >2... Son:

3 a 2-3}... a an}... a .33... “32.. .5

m .c0wuuo>cou u .co«umuaawus m .uoucoauou m .HOucmEuou om .ooou us .oawu
cocoon” no aucowoauwm owouoma mo comm cw cumuooa cw encuomq cw omouomq scaucouou Haoo oucuunnsm

 

scauouodo ocoumIoHoaam new: nao>oq ofiaBIcOwucmuom 039 am when: no
cOaumucoauom ounuomAIauacan< mooscwucou no camwuoaaoo .v mqm<a

20

sgpafi

 

 

Sf,Pf,Xf,Ff

Fig. 1. Schematic of one-stage continuous fermentation system.

Symbols are described in Table l.

21

 

 

 

 

BC) I |
60" "
E e
O
E
" 4o— -
d: I
3 I
If)" ' I
20- ' -
o 1 l
0 IO 20 3O
'1‘f (III

Fig. 2. Computer-simulated effects of cell retention time (Tf)
on residual lactose (Sf) and accumulated lactate (Pf) in the fermentor
with one-stage operation. The curves were plotted by use of the steady-
state solution and the values in Table 3. The points (squares for Sf,
circles for Pf) are experimental data and demonstrate the fit

between the experimental results and the computer simulations.

22

 

 

 

E 2" g -I
\ O

E? ‘““}'---
1 I- ._
X

 

 

 

TfIhI

Fig. 3. Computer-simulated effects of cell-retention time (Tf)
on cell-mass concentration (xf) in the fermentor. The curve was
plotted by use of the steady-state solution and the values in
Table 3. The points are experimental data and demonstrate the fit
between the experimental results and the computer simulation.

23

 

 

 

 

5 I I
4 I- d
a.
'5
)-
2 b -
o J I
0 IO 20 30
T, (h)

Fig. 4. Experimental effect of changes in cell retention time

(Tf) on the molar growth yield (YATP).

24

 

 

 

 

5 l 1 l00
.. E
3
J.
E
:0 3 .. '4 60
g a
U 33
_ kl
O
E
\
o
E
.6
g— Q ~20
O L l O
0 IO 20 30
7, III)
Fig. 5.

Experimental effect of changes in cell retention time
(Tf) on efficiency of lactose conversion E = (1 - S?) x 100 ,
rate of lactate formation rp - (Pf - P:)/Tf , and rate of cell growth

(rg = Xf/Tf) .

25

(ll

 

 

 

 

 

 

X,Imolml)
T
\\\\\\\\
\ \\.\\\V

 

 

 

 

 

 

.\\\\\
I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

30 A
25—‘? "
/ ._
.. g —I '_I ,__1
%I5-/ '
er 4
. / -
“W? Z? 7
°l as 0.4.13 (2 {a

 

TIz’TII

Fig. 6. Computer-simulated two-stage continuous operation for

the fermentation. A - The effects of changes in the ratio of reten-

tion time of the second fermentor to that of the first fermentor

(sz/Tfl) on the total retention time (Tf1 + sz) required to achieve
60 mg of lactate per ml in the effluent of the second fermentor; the

hatched portion of the bars represents Tf1 and the remaining portion
represents sz. B - The effects of changes in sz/Tf1 on the accumu-

lation of cell mass in the effluent of the second fermentor (Xf).

26

150

ARTICLE VI

DIALYSIS CONTINUOUS PROCESS FOR AMMONIUM-LACTATE

FERMENTATION: SIMULATED PREFERMENTOR

AND CELL-RECYCLING SYSTEMS

BY

R. W. Stieber and Philipp Gerhardt

Dialysis Continuous Process for Ammonium-Lactate Fermentation:
Simulated Prefermentor and Cell-Recycling Systems

R. W. STIEBER and PHILIPP GERHARDT

Department of Microbiology and Public Health,
Michigan State University,
East Lansing, Michigan 48824
SUMMARY
A generalized mathematical model, previously developed and
experimentally validated, was modified and used to computer-
simulate the outcome of two systems for increasing the rate and
extent of substrate conversion of a dialysis continuous process
for the ammonium-lactate fermentation of deproteinized whey. The
addition of a nondialysis continuous prefermentor would not
improve the process. However, the addition of cell recycling
(feedback) would greatly improve the process. With 100% recycling
the cell-retention time would be reduced ninefold, and the extent
of converting substrate to product would be increased 25% and
accomplished almost entirely by maintenance metabolism.
INTRODUCTION
A dialysis continuous process can be used effectively for

the fermentive production of ammonium lactate and Lactobacillus

 

cells from whole or deproteinized whey (Stieber et a1., 1977;
Stieber and Gerhardt, 1979a; Stieber and Gerhardt, 1979c). The
products have commercial potential, including use as a nitrogenous
feed supplement for ruminants (Juengst, 1979; Gerhardt and Reddy,
1978; Reddy et al., 1976; Keller and Gerhardt, 1975; Coulman et

a1., 1977). In the process, a high concentration of the whey

(e.g., 240 mg of lactose per m1) is continuously fed into a

fermentor, the contents of which are dialyzed through a membrane
against a continuous stream of water. Relative to nondialysis
continuous or batch processes for the fermentation, the dialysis
continuous process enables the use of more concentrated substrate,

has a higher rate of substrate utilization and additionally produces a
dialysate effluent of cell-free ammonium lactate. With deproteinized
whey, however, the effluent from the fermentor contains a considerable
residuum of lactose which lessens the value of the process.

Increasing the concentration of cell mass in the fermentor may be
a way to increase both the extent and the rate of substrate conversion.
Thus, the objective of this study was to test by computer simulation
the feasibility of improving the dialysis continuous process with
deproteinized whey either by the recycling of cells or by the addition
of a nondialysis prefermentor that is optimized for cell-mass accumu-
lation.

Recycling (feedback) of cells in a continuous fermentation
process is used to retain cells in a fermentor and thereby to increase
the biomass concentration in the fermentor, allowing increased through-
put and decreased retention time (Fencl, 1966; Schroeder, 1971; Pirt,
1975). Importantly for the present study, recycling also can be used
to reduce substrate concentration in the fermentor effluent (Grieves
et a1., 1964). Wall growth, a form of cell retention, was modeled
(Topiwala and Hamer, 1971) and a prediction of increased biomass
productivity was experimentally validated (Wilkinson and Hamer, 1974).
Using filtration to retain cells in a fermentor, Pirt and Kurowski

(1970) obtained a fourfold increase in biomass productivity.

2

Topiwala and Khosrovi (1978) theoretically studied product accumulation
and its effect on process water recycle in a biomass production
process. Weissman and Benemann (1979) demonstrated that recycling can
change the outcome of species competition in a mixed continuous
culture of Spirulina gertleri and a Chlorella sp. Other studies on
recycling with heterogenous populations have been concerned with an
activated-sludge process (Ramanathan and Gaudy, 1969; Ramanathan and
Gaudy, 1971; Gaudy and Srinivasaraghaven, 1971).

The use of a nondialysis continuous prefermentor from which the
effluent goes into a dialysis fermentor is analogous to the "multi-
stream" two-stage process modeled by Pirt (1975). He also reviewed
the results obtained with other such processes and concluded that they
extend the applicability of continuous culture. In particular, such
processes may have use for producing secondary metabolites and for
obtaining steady-state conditions that sometimes are unobtainable with
single-stage systems.

In the present study, a generalized mathematical model of the
dialysis continuous process for the fermentation, previously developed
and experimentally validated (Stieber and Gerhardt, 1979a; Stieber and
Gerhardt, 1979c), was modified to incorporate a second feed stream
for cells, substrate and product into the fermentor. The modified
model was used to computer-predict the outcome of adding cell-recycling
or a prefermentor to the dialysis continuous process with

deproteinized whey as substrate.

MATHEMATICAL MODEL

Design of Fermentation System

 

Figure 1 shows a schematic of a dialysis continuous process for
the ammonium-lactate fermentation of whey with two feed flows into the
fermentor circuit. The symbols correspond to those used previously
(Stieber and Gerhardt, 1979a) and are listed in Table 1. A usual
feed flow (Ff) contains concentrated substrate (S?), a small
concentration of product (P2) and no cells. A second cell-feed flow
(Fcf) contains cells (ch) obtained either from recycling, (i.e., the
return from the fermentor-circuit effluent) or from the effluent of
a prefermentor (i.e., a single-stage nondialysis continuous fermentor
with feed whey of normal concentration), and may or may not contain
residual lactose (Scf) and accumulated ammonium-lactate (Pcf).
Further rationale for the design of the fermentation system and the
assumptions for purposes of mathematical modeling were detailed by
Coulman et a1. (1977), but an additional assumption for the present
process is that there is no lag-growth phase for the cells entering

the fermentor circuit.

Generalized Model

 

From Figure l, a set of material-balance equations was developed
for substrate, product and cell mass in the fermentor circuit and for
substrate and product in the dialysate circuit (not shown). Rate-
relationship equations for substrate utilization, product formation
and cell growth were the same as those found to be valid previously
(Stieber and Gerhardt, 1979a; Stieber and Gerhardt, 1979b; Stieber

and Gerhardt, 1979c). The rate equations were combined with the

material-balance equations and the variables were defined in
dimensionless parameters (Table II) to obtain a generalized model for
the fermentation.

The resulting equations for the fermentor circuit are as follows:

 

 

 

.3, _ [ z: 1 _
--- -(l+II)s-6[_ __ __ +8'r/ax +11? +A§+l-A/T (1)
dt f K +5 +KP) f f :3 cf f
I f pf
:3- -(1+an)3+e '5" I+srm§+an§ +AF+1 A’° 2
a: r E +§ +-—-j : '2 a of ‘ " ”f”: H
K!’
I f pf
a? E
dtf- e[_ _f __} -1+Ax3?£/'rf (3)
x.I-sf+-fi;2
where Tf = Vf/Ff and is an operational parameter.
For the dialysate circuit, the corresponding equations are:
Ina ' _ _
ag- - -(¢ + msd + nsf Ff/Vd (4)
_ r
-£- -( +Im" ’
dt 0 )Pd 4' NIP: Pf/Vd (5)

 

Generalized Steady:§tate Solution
The equations of the generalized model were rearranged and
combined, and the time derivatives were set at zero, to obtain a

generalized solution for the steady state.

Generalized equations for substrate, product and cells in the

fermentor circuit are as follows:

 

_ - 05, - rpm [1 + pf + Mpcf+ see“ pf - 1)] - (x3 - xsaxunpnn 1 an)

5: .. _ (6)
(1+H-I‘HHKp-KPAXI-(O-l+Ax)(wafl-l-Rfl)
F . [AIl+P£- scf- cf) -l-P£](6 - l+Ax) + (x8 -KSAX)(1+ n -u8n)

 

(7)
f ._ ..
(9 1 AX)(1IPRH 1 RA ) (KP KPAX) (1 + II 1‘11)

_ 1-‘s‘ +nI's' -§)+A(" -1)
x- f a f 5t: m

f 1 + art/a - A:

where NS = H/(¢ + H) and up = RH/(¢ + RH).
For the dialysate circuit, the corresponding generalized equations

are the same as those used previously (Stieber and Gerhardt, 1979a):

5d . n's'f/M + n) (9)

Fa - unit/I; + am (10)

Equations 6 to 10 comprise a generalized steady-state solution
for substrate, product and cells in the fermentor and dialysate
circuits of the system.

COMPUTER SIMULATIONS
Dialysis Continuous Process With Recycling of Cells

The generalized steady-state solution was used to simulate the
dialysis continuous process with recycling of cells for the fermen-
tation of deproteinized whey lactose to ammonium lactate and cells.
The feed-cell ratio (x) and the feed-flow ratio (A) are dimensionless
operating parameters not contained in the previous dialysis model

(Stieber and Gerhardt, 1979a) and are defined in Table II with the

6

other operating parameters. The use of X enables the simulation of
substrate utilization and product formation without knowledge of cell
concentration. However, the model cannot predict cell-concentration
data unless values are given for the substrate/cell ratio (a) and the
specific maintenance rate (8). Operationally, A should be kept small
because the smaller that it is, the less that the feed-substrate
concentration (5:) is diluted. The product of A and x represents the
fraction of cells from the fermentor-circuit effluent which is recycled.
Thus, A can easily be maintained small if the cells from the fermentor-
circuit effluent are concentrated.

The simulations were made by arbitrarily setting the feed-flow
ratio (A) at .2 and the substrate (Scf) and product (Pcf) concentrations
in the cell-feed f1ow (Fcf) at 0, and by varying the feed-cell ratio
(x) and the time factor (6). All other parameters and conditions
(Table III) were the same as those used before (see Table 5 in
reference Stieber and Gerhardt, 1979a; and Table 4 in reference
Stieber and Gerhardt, 1979c), which closely simulated the experimental
results.

Figure 2 shows the simulated effects of changes in the time
factor (6) on the substrate factor in the fermentor circuit (SE) at
various fractions of cell-recycle (AX). S. decreased with increased

f

8. More importantly, S. can be reduced completely and at a much

f
smaller 6 with total recycling of cells (AX = 1.0) than with none
(AX = 0). The latter prediction (at AX = 0) corresponded with
experimental results of 75% conversion of lactose to products at a

retention time of 27.5 h (Stieber and Gerhardt, 1979a). When given

dimensions, the simulated outcome with total recycling of cells

predicted practically complete conversion of lactose (an increase of
25%) at a retention time of only 2.9 h. Furthermore, calculations of
growth-associated metabolism (up) and nongrowth-associated
(maintenance) metabolism (8) showed that, with total recycling of
cells, the conversion would be accomplished solely by the latter.

The effect of changes in the time factor (6) on the cell factor
in the fermentor circuit (X?) at various fractions of cell-recycling
(AX) is shown in Figure 3. The highest value of Xk occurred with
total recycling of cells at 6 = 1.0 and corresponded to 10 mg dry
weight of cells per ml for the ammonium-lactate fermentation.

Dialysis Continuous Process With Prefermentor

A two-stage dialysis continuous process with deproteinized
whey was simulated by use of the experimental results from a nondialysis
continuous process at parameters where the cell mass concentration
in Figures 1 and 2 of reference Stieber

was maximum (8 P and XC

cf' of f

and Gerhardt, 1979b). The first (nondialysis) stage is optimized for
cell-mass accumulation and the effluent from it flows into the second
(dialysis) stage, which is also fed with concentrated substrate. The
process is simplified in that there is no concentration or separation
step between the prefermentor and the dialysis fermentor. With this
system, the only way to manipulate the amount of cells entering the
dialysis fermentor (ch) is to change the feed-flow ratio (A).

Unlike the simulation of cell recycling, a specific value for the
feed-cell ratio (X) was required initially to simulate the effects of
changes in the feed-flow ratio (A) and the time factor (6) on the
dialysis continuous process with a prefermentor. Whereas X does not

vary with 9 for the simulation of cell recycling, X does vary with 9

8

for simulating the dialysis process with a prefermentor because ch
is constant and X; varies with 6 (Fig. 3). Thus, to begin the
prefermentor simulation, the results from cell-recycling simulation
were used to choose the proper value for X (with ch = 1.85 mg/ml) at
each value of A.

It was thought that the feed of cells (ch) into the dialysis
fermentor would increase the cell mass in the fermentor circuit (Xf)
and favorably affect the rate of lactose utilization (mg/ml-h) and the
extent of lactose conversion. Previous work (Stieber and Gerhardt,
1979a) had established the parameters that would accomplish adequate
product concentrations. Consequently, the present simulations
compared the dialysis process without cell input to the dialysis
process with increasing input of cells from a prefermentor with respect
to the rate of substrate utilization (-rs), the extent of lactose
conversion (E), and the cell-mass accumulation (Xf). The simulations
were made by varying the time factor (8) over a range of feed-flow
ratios (A), with initial feed—cell ratios (X) determined as described
above. All other parameters and conditions (Table III) were the same
as those used previously (Stieber and Gerhardt, 1979a; Stieber and
Gerhardt, 1979c), which closely simulated the experimental results.

The simulated results of changes in the feed-flow ratio (A)
affecting the rate of substrate utilization (-rs) in the dialysis
fermentor are shown in Figure 4. At all cell-retention times (Tf),
the maximum.rate of substrate utilization occurred with no input of
cells from a prefermentor.

The extent of substrate conversion (E) increased with increased

A (Figure 5). Most of the increased conversion, however, occurred

because the feed substrate concentration (8:) was diluted by the input
from the prefermentor. The dilution effect of A on 5: (D5?) represents
the amount of feed substrate per total volume entering the fermentor
and this also is shown in Figure 5.

Figure 6 shows the simulated effects of changes in the cell-
retention time (Tf) on the concentration of cell mass in the
fermentor circuit (Xf) at various feed-flow ratios (A). Most
importantly, Xf decreased with increased A. The cell mass was
greatest in the dialysis fermentor without cell input (A = 0) from
the prefermentor.

DISCUSSION

Two systems for increasing the rate and extent of substrate
conversion for the continuous ammonium-lactate fermentation of depro-
teinized whey were modeled and simulated: a two-stage process in which
a first nondialysis stage (prefermentor) is optimized for cell mass,
the effluent of which goes into a second dialysis stage that is also
fed with concentrated substrate; and a single-stage dialysis process
with recycling of cells.

The outcomes predicted that the addition of a prefermentor to the
dialysis fermentor would only marginally increase the extent of
substrate conversion to product and would actually slow the rate of
substrate conversion in the dialysis fermentor. The cause of the latter
effect is that the effluent of the prefermentor contains lactic acid
as well as cells; as more cells from the prefermentor are fed into
the dialysis fermentor, more lactate is fed and so further decreases
the rate of the fermentation (equations 1, 2 and 3). The addition

also would complicate the process. Moreover, the cells from the

10

prefermentor, upon entering the dialysis fermentor, would be

affected by a large increase in substrate concentration and thus a
lag-growth phase may result. The use of a prefermentor, furthermore,
would lower the concentration of cell mass in the dialysis fermentor.
The cause is that with increased feed into the dialysis fermentor from
the prefermentor, the feed substrate concentration (S?) is increasingly
diluted (Figure 5). In a continuous culture, the cell mass is
directly proportional to the feed substrate concentration that can be
metabolized by the cells (Stieber and Gerhardt, 1979c). Altogether,
the addition of a nondialysis prefermentor to the dialysis fermentor
would not improve the process.

On the other hand, the recycling of cells would greatly improve
the dialysis continuous process for the ammonium lactate fermentation.
of deproteinized whey. The rate equations of the model for substrate
utilization, product formation and cell growth have been shown
previously to be valid (Stieber and Gerhardt, 1979a; Stieber and
Gerhardt, 1979b; Stieber and Gerhardt, 1979c). The lag-growth phase
of the recycled cells upon entering the fermentor should be minimal if
the process is managed at steady-state conditions and if a percentage
of the cells can be separated from the fermentor-circuit effluent,
concentrated and returned quickly to the fermentor circuit without
drastic changes. Accordingly, the simulated predictions of increased
rates and extents of substrate conversion to product should be
realized experimentally.

The recycling of cells also increases the cell mass in the fer-
mentor circuit, which results in greater rates of substrate utilization

and product formation. Practically, the recycling of cells would be

11

accomplished by separation (centrifugation, filtration or sedimentation),
and damaged and dead cells may result. Perhaps only 80% cell recycling
can be realized, which would still much improve the conversion
efficiency and rate of fermentation. 0n the other hand, the applica-
tion of cell recycling increases cost and complicates operation. The
advantages must compensate the disadvantages if cell recycling is to

be useful.

The limiting factor for the recycling of cells is the problem of
separating the cells from the effluent and returning the cells to the
fermentor without losing cell activity (Gledhill et al., 1973). An
alternative system, which may accomplish the same objectives as cell
recycling and yet not have the mechanistic problems, may be to use a
membrane to retain the cells in the fermentor. In the dialysis
continuous processes studied to date (Coulman et al., 1977; Stieber
and Gerhardt, 1979a), membranes have been used primarily for the
removal of inhibitory products and not for retaining cells. However,
the purpose might be accomplished by using a batch fermentor in
which the substrate is continuously fed to the fermentor culture by
way of membranes and the product is continuously removed from the
culture by way of membranes. Essentially 100% cell recycling
might be accomplished in this way, amounting essentially to an
immobilized cell system.

ACKNOWLEDGEMENT

This work was supported by grants ENG 76-17260 and DAR 79-10236
from the National Science Foundation.

This was assigned journal article no. from the

Michigan Agricultural Experiment Station.

12

REFERENCES

l. R. W. Stieber, G. A. Coulman, and P. Gerhardt, Appl. Environ.
Microbiol., 34, 733 (1977).

2. R. W. Stieber and P. Gerhardt, Appl. Environ. Microbiol.,
37, 487 (1979a).

3. R. W. Stieber and P. Gerhardt, J. Dairy Sci., (1979c)
submitted.

4. F. W. Juengst, Jr., J. Dairy Sci., 62, 106 (1979).

5. P. Gerhardt and C. A. Reddy, Dev. Ind. Microbiol., 19,
71 (1978).

6. C. A. Reddy, N. E. Henderson, and M. D. Erdman, Appl.
Environ. Microbiol., 32, 769 (1976).

7. A. K. Keller and P. Gerhardt, Biotechnol. Bioeng., 17, 997
(1975).

8. G. A. Coulman, R. W. Stieber, and P. Gerhardt, Appl. Environ.
Microbiol., 34, 733 (1977).

9. Z. Fencl, in Theoretical and Methodological Basis of
Continuous Culture of Microorganisms, I. Malek and Z. Fencl, Eds.,
Academic Press, New York, 1966, p. 67.

10. E. D. Schroeder, water Research, 5, 29 (1971).

11. S. J. Pirt, Principles of Microbe and Cell Cultivation,
John Wiley & Sons, New York, 1975, p. 45.

12. H. H. Topiwala and G. Hamer, Biotechnol. Bioeng., 13, 919
(1971).

13. T. G. Wilkinson and G. Hamer, Biotechnol. Bioeng., 16, 251
(1974) .

14. S. J. Pirt and W. M. Kurowski, J. Gen. Microbiol., 63, 357
(1970).

15. H. H. Topiwala and B. Khosrovi, Biotechnol. Bioeng., 20, 73
(1978).

16. J. C. Weissman and J. R. Benemann, Biotechnol. Bioeng., 21,
648 (1979).

17. M. Ramanathan and A. F. Gaudy, Jr., Biotechnol. Bioeng.
11, 207 (1969).

13

18. M. Ramanathan and A. F. Gaudy, Jr., Biotechnol. Bioeng.
13, 125 (1971).

19. A. F. Gaudy, Jr., and R. Srinivasaraghaven, Biotechnol.
Bioeng., 16, 723 (1971).

20. R. B. Grieves, W. O. Pipes, W. F. Milbury, and R. K. WOod,
J. Appl. Chem., 14, 478 (1964).

21. R. W. Stieber and P. Gerhardt, J. Dairy Sci., (1979b) in
press.

22. W. E. Gledhill, I. D. Hill, and P. H. Hodson, Biotechnol.
Bioeng., 15, 963 (1973).

14

TABLE I

Glossary of Mathematical Symbols

 

 

Symbol Description Units
Am Area of membrane available for dialysis cm
Fd Flow rate into and out of dialysate circuit ml/h
F: Flow rate into fermentor circuit ml/h
Ff Flow rate out of fermentor circuit ml/h
Fcf Flow rate of cell feed into fermentor circuit ml/h
KP Product inhibition constant mg/mg
Ks Substrate limitation constant mg/ml
Pd Product concentration in dialysate circuit mg/ml
P: Product concentration in fermentor feed mg/ml
Pf Product concentration in fermentor circuit mg/ml
PCf Product concentration in fermentor cell-feed mg/ml
Pmp Permeability of membrane to product mg/cm:-h
Pms Permeability of membrane to substrate mg/cm -h
rg Rate of cell growth mg/ml-h
rp Rate of product formation mg/ml-h
-rs Rate of substrate utilization mg/ml-h
Sd Substrate concentration in dialysate circuit mg/ml
s: Substrate concentration in fermentor feed mg/ml
Sf Substrate concentration in fermentor circuit mg/ml
Scf Substrate concentration in fermentor cell-feed mg/ml
t Time h
Vd Volume of liquid in dialysate circuit ml
Vf Volume of liquid in fermentor circuit m1
Xf Cell-mass concentration in fermentor circuit mg/ml
ch Cell-mass concentration in fermentor feed mg/ml
a Substrate/cell ratio mg/mg
8 Specific maintenance rate h-l
Y Product/substrate ratio mg/mg
E Extent (efficiency) of lactose conversion %
um Maximum specific growth rate of cells h-1
Tf Cell-retention time in fermentor circuit h

 

15

THUSLIIIII

Glossary of Dimensionless Parameters

 

 

Type Symbol and definition Description
Material parameters P; I d“752) Product factor in dialysate
circuit
5} I Pg/(ySZ) Product factor in fermentor
circuit
3: I P:/(yso) Product factor in fermentor
f
feed
5;: I Pct/(75:) Product factor in fermentor
cell feed
S' I s /S0 Substrate factor in dialysate
d d f
circuit
S. I S /S0 Substrate factor in fermentor
f f f
circuit
3': I scf/s: Substrate factor in fermentor
c cell feed
X; I ext/s: Cell factor in fermentor
circuit
i. - ex fls° Cell factor in fermentor feed
cf c f
Operational parameters R I P [Pn. Ratio of product/substrate
mp membrane permeabilities
X I chle Feed-cell ratio
A I Pct/Pf Feed-flow ratio
H I PIsAm/Pf Membrane permeability factor
o I Fd/Ff Flow-rate ratio
Kinetic parameters E; I Ki/S; Substrate-limitation factor
i'p - yrp Product-inhibition factor
9 I uan Time factor

 

16

 

uouceauououm
new: umeoon
mooscwucoo
mamxanao we

 

m. mm. II om. ~.H o.s mm.a o.vm m.ooa o.n o.mm a.m mo. ~.~ e .m .v mcoauuaeaua nos
mcwaoaoou Have
sum: nuoooum
mSODCH UGOU
mamsamae co
n. mm. N. am. ~.H 6.» II II m.ooe o.m II H.m mo. ~.~ n .u uneducaewua son
AsIe. “sexes. .msxms. .AIEC lmsxms. lasxms. lasxmss .stms. lasxosa lasxms. issxas. .62 me
E U U 0
= a < r m o u x N m Wm m u A we ox or nousmam on:

 

occuucucoauom no unawusasaam uousmsou now own: mman>

HHH mumda

l7

 
 

H 0 F
2 v d
SCf'Pcf'fo'FCf

 

 

 

 

 

 

 

Sf' Pf'xf' Ff 3:1:de Fa
FERMENTOR DIALYSATE
CIRCUIT CIRCUIT

Fig. 1. Schematic of dialysis continuous fermentation system
with a cell-feed flow in addition to the usual feed flow. Symbols

are described in Table I.

18

 

AX
|.0

 

 

 

 

 

Fig. 3. Simulated effects of changes in the time factor (8) on
the cell factor in the fermentor circuit (XE) at four fractions of cell-

recycling (AX).

20

IO

 

-rs (mg/ml-h)

 

 

 

 

Fig. 4. Simulated effects of changes in the feed-flow ratio (A)

on the rate of substrate utilization (-rs) at three cell-retention times

(Tf) for the dialysis continuous process with a prefermentor. -rs =

O
[S£F£(l ‘- A) +schcf -(SfE'f + SdFd)]/(TfFf).

21

 

 

 

 

 

 

 

200 I I l I00
'l'f-ZB
'60 \ 2° - 80
E '20 ‘ I0 - so
a
E 33
a? I!)
Q
8{>t- .. 4()
4()" ._ 2c,
() l l l (3
O 2 .4 6 8
A

Fig. 5. Simulated effects of changes in the feed-flow ratio
(A) on the extent of substrate conversion (E) and on diluting the

feed substrate concentration (D82) at three cell-retention times

(If) for the dialysis continuous process with a prefermentor. E

O O O
I- '- . DS 3 l " A S + A o
[1 (1'fo + rdsdvmcfscin £0319] x 100 f I ) f scf

22

 

 

 

 

 

X,(mglml)

 

 

 

 

Fig. 6. Simulated effects of changes in the cell-retention time
(Tf) on the cell-mass concentration in the fermentor circuit (xf)
at three feed-flow ratios (A) for the dialysis continuous process

with prefermentor.

23

174

ARTICLE VII

DIALYSIS CONTINUOUS PROCESS FOR AMMONIUM-LACTATE

FERMENTATION: SIMULATED AND EXPERIMENTAL

DIALYSATE-FEED SYSTEMS

BY

R. W. Stieber and Philipp Gerhardt

Dialysis Continuous Process for Ammonium-Lactate Fermentation:
Simulated and Experimental Dialysate-Feed Systems

R. W. STIEBER and PHILIPP GERHARDT

Department of Microbiology and Public Health,
Michigan State University,
East Lansing, Michigan 48824

SUMMARY

A generalized mathematical model, previously developed and
experimentally validated, was modified and used to computer simulate
a dialysate-feed system for operating a dialysis continuous process
for the ammonium-lactate fermentation of whey. Outcomes predicted
that the feeding of substrate into the fermentor via dialysis should
greatly improve the production of cell mass and cell-free metabolite.
Experiments were conducted to test selected predictions. Dried
cheese-whey ultrafiltrate was rehydrated to contain a normal
concentration of lactose (62 mg per m1) supplemented with yeast
extract, charged into a 5-liter fermentor without sterilization,
adjusted in pH (5.5) and temperature (44°C), and inoculated with an

adapted culture of Lactobacillus bulgaricus. A series of steady-

 

state conditions was managed without asepsis for 26 days. Results
showed that the dialysate-feed system is a useful way to obtain
immobilized cells for the purpose of producing a cell-free metabolite
at a high rate. The substrate consumed by the cells is converted

only via maintenance metabolism and is sterilized by dialysis.

INTRODUCTION

Previous studies of a dialysis continuous process for the
ammonium-lactate fermentation of whey (Coulman et al., 1977; Stieber
et a1., 1977; Stieber and Gerhardt, 1979a; Stieber and Gerhardt, 1979c)
were conducted in a system in which substrate is fed into a fermentor
and the fermentor contents are dialyzed against a stream of water to
remove an inhibitory product (lactic acid), thus maximizing the rate
and extent of substrate (lactose) conversion. Relative to nondialysis
continuous or batch processes, this dialysis continuous process enabled
the use of more concentrated substrate, had a higher rate of substrate
utilization, produced cells at a higher rate and concentration, and
additionally produced a dialysate effluent containing a cell-free
product.

An alternative way to operate a dialysis continuous process is
to dialyze the fermentor contents against a dialysate stream containing
substrate, primarily for the purpose of retaining cells in the
fermentor and thereby increasing the concentration of cells and
throughput of substrate. In this dialysate-feed system, the
substrate enters the fermentor by diffusion through membranes and
thereby is sterilized. Similarly, a metabolite product leaves the
fermentor by dialysis and thus is free of cells.

The dialysate-feed system was studied by use of the ammonium-

lactate fermentation of cheese whey by Lactobacillus bulgaricus as a

 

model. The fermentation advantageously can be conducted without
sterilization or asepsis because of the restrictive conditions of
low pH (5.5), high concentration of undissociated acid, high

temperature (44°C), and anaerobiosis. The disposal of whey is a

problem in the dairy industry, and the product of the fermentation
has commercial potential, including use as a nitrogen-enriched feed
supplement for ruminant animals (Gerhardt and Reddy, 1978; Juengst,
1979). Consequently the fermentation also has important economic
implications.

A mathematical model, previously developed and experimentally
validated (Stieber and Gerhardt, 1979a; Stieber and Gerhardt, 1979c),
was modified and used to computer-simulate the outcomes of operating
the dialysate-feed system for the fermentation. The simulations were
used to determine the optimum conditions for producing either
ammonium-lactate or lactobacillus cells. Laboratory experiments then
were conducted to test selected predictions. (This paper was
presented in part at the American Chemical Society Annual Meeting,

September 10-15, 1978, Miami Beach, Florida.)

THEORY
Design of Dialysate-Feed System

Figure 1 shows a schematic of the dialysate-feed system. The
symbols correspond to those used previously (Stieber and Gerhardt,
1979a) and are listed in Table I.

The feed into the dialysate circuit contains the substrate
(lactose) in normal concentration (52) and is maintained at a
relatively high rate of flow (Fd). A small concentration of product
(lactic acid) preexists in the whey feed (P3). The dialysate circuit
consists only of tubing, pump and the dialysate side of the dialyzer,
and contains a relatively small volume Iva) which is continuously
circulated. Substrate which is not transferred by the dialyzer to

the fermentor circuit remains in the dialysate effluent (Sd).

Product from the whey feed as well as that which is transferred from
the fermentor circuit by the dialyzer remains in the dialysate
effluent (Pd).

In the fermentor circuit, the liquid volume (Vf) and cell mass
(Xf) are maintained at constant levels. Substrate that is transferred
from the dialysate circuit by the dialyzer is converted to product

(P Unused substrate remains in the fermentor effluent (Sf) along

f)'
Wlth Pf and Xf.

Further rationale for the design of a dialysis continuous
fermentation system and the assumptions for purposes of mathematical

modeling are detailed by Coulman et al. (1977).

Generalized Model

 

From Figure l a set of material-balance equations (not shown)
was developed for substrate, product, and cell mass in the fermentor
circuit and for substrate and product in the dilaysate circuit.
Rate-relationship equations for substrate utilization, product
formation and cell growth were the same as those used previously
(Stieber and Gerhardt, 1979a). The rate equations were combined with
the material-balance equations and the variables were defined in
dimensionless parameters (Table II) to obtain a generalized model for
the system.

The resulting equations for the fermentor circuit are as follows:

 

d f _ sf BTf_ _
_—= ..(1 + ¢n)s - e + —X + w /T (1)
dt f - - -- f d f
K + s + KP a
s f p f

 

 

f _ _ __ +—-x +¢RIIF /Ir (2)
K f a
s + Sf + xppf a f
31$. 6[ S: J ..
K ... _ f f (3)
S Sf + Fpr

where Tf = Vf/Ff.

For the dialysate circuit, the corresponding equations are as

 

follows:
" " " - (4)
dsd I -(1 + msd + IISf . l + 1/4] I‘d/vd
" P " " ‘° '° (5)
de a L.” + RII)Pd + RHPf + Pd + Pd/¢]Fd/Vd

Generalized Steady-State Solution
The equations of the generalized model were rearranged and
combined, and the time derivatives were set at zero, to obtain a
generalized solution for the steady state.
The resulting equations for substrate, product and cell mass
for the fermentor circuit are as follows:

- - -o -o
KS(RH RHIP + l/o) + KPUIPPd + "p Pd/¢ + "s + nS/¢) (5)

 

S: I

(e - l)(RH - an"? + 1/¢) - rpInsn - n -_1/¢)

-o —o -—
(6 l)(1TPPd + ind/¢ + "s + ns/o) + Ks(nsH - H - l/o)

TI
u

"I

__ (7)
(e - 1 an - an - - -

)( "p + l/¢) KP(TSH H 1/¢)
¢H(sd - sf) - sf

(8)
f l + 86/(umc)

><
II

where "s = H/(¢ + H) and "p = RH/(¢ + RH).
For the dialysate circuit, the corresponding equations are

as follows:

sd = (H's-f + 1 + 1/¢)/(1 + H) <9)
_ _ - .0 .0 (10)
Pd - (RHPf + Pd + Pd/¢)/(l + RH)

Computer-Simulated Predictions
The generalized steady-state solution (equations 6 to 10) was used
with a digital computer to simulate operation of the system. Because
the fermentation was simulated before laboratory experiments were
conducted, approximate values of various parameters were used
(Table III). The values for um, Ks, KP, y, and R were obtained from
previous dialysis fermentations with deproteinized whey (Stieber

and Gerhardt, 1979a). The values for a and 8 were obtained

similarly (Stieber and Gerhardt, 1979c). The values for s: and

P2 were chosen arbitrarily.

The simulation outcomes in Figure 2 show the effects of changes
in the cell-retention time (Tf) affecting the residual substrate
(Sd) and the accumulation of product (Pd) in the dialysate circuit at

a constant flow-rate ratio (¢). As is characteristic of conventional

continuous fermentation processes, the substrate decreased and the
product increased with increasing retention time. The simulation
outcomes were almost identical at other values of ¢ (e.g., ¢ = 1.0 or
20.0).

In Figure 3 are shown the simulated effects of changes in the
membrane permeability factor (H) on the substrate concentration in the
fermentor circuit (Sf) and in the dialysate circuit (Sd). A high
value of H (at least 8.0) is required to obtain a concentration of
substrate in the fermentor circuit similar to that in the dialysate
circuit. The substrate concentration rather than the product con-
centration was simulated to demonstrate the membrane permeability
requirement, because it was thought that the fermenting culture would
be limited mostly in obtaining substrate for metabolic purposes.

Figure 4 shows the simulated effects of changes in the cell-
retention time (Tf) affecting the cell mass concentration (Xf) at
three flow-rate ratios (¢). At first, with increasing retention
time where growth metabolism (a) predominated, Xf increased to a
maximum. But beyond a critical retention time, maintenance
metabolism (8) superseded that of growth and then Xf decreased.

Xf also increased greatly with increased o. In fact, the results
predicted that the cell-mass productivity of the present process
(3.5 mg/ml-h at o = 20.0) would be about 20-fold greater than that
for an ordinary continuous process for the same fermentation
(Stieber and Gerhardt, 1979b) at a retention time of 10 h, where X

f

was greatest in both processes.

Design, Solution and Simulation
of Dialysate-Feed, Immobilized-Cell System

The system can be modified to maximize the production of the
metabolite product by eliminating the fermentor effluent stream
(Figure 5), which results essentially in an immobilized-cell system
with the metabolite product removed only in the dialysate effluent.
Schultz and Gerhardt (1969) termed this a system with a continuous
reservoir and a batch fermentor.

A generalized steady-state solution was obtained for this
dialysate-feed, immobilized-cell system by setting the fermentor flow
rate (Ff) equal to zero and reducing equations 6 to 10. For the
fermentor circuit, the resulting equations for substrate, product and

cell mass are as follows:

 

sf = 0 _ (11)
__ n P: + H
Pf = s (12)
RH - RHn
p
i = DH (13)
f (l + TUB/a

where the hydraulic dilution rate (D) equals Fd/Vf.
For the dialysate circuit, the corresponding equations are

as follows:

Sd

1/(1 + II) (14)

Pa = (R35; + 3§)/(1 + RH) ‘15)

This solution importantly indicates that the concentrations of
substrate (Sf and Sd) and product (Pf and Pd) in the fermentor and
dialysate circuits are independent of the hydraulic dilution rate
(D). The accumulation of cell mass (Xf) is linear with respect to D.
By use of values for the constants listed in Table III, a graphical
representation of the effects of changes in D on Sd, Pd' and Xf is
shown in Figure 6. The simulation outcomes predict that a satisfactory
low level of Sd (< 5 mg/ml) and a high level of Pd can be obtained no
matter what dilution rate is used. No change in the permeability
factor (H) is required with increased D as long as D is increased by
decreasing the volume of the fermentor circuit (Vf). However, if D
is increased by increasing the flow rate into the dialysate circuit
(Fd), there must be a concurrent increase in the permeability (Pmp and
Pms) or the area (Am) or the dialyzer membranes to maintain the
desired concentrations of 8d and Pd'

Figure 6 also shows that the accumulation of cell mass (Xf)
is directly proportional to the dilution rate (D). In other words,
given a constant flow rate of feed (Fd) into the system, no matter

what the fermentor-circuit volume (Vf) is, X remains unchanged. Thus

f
the conversion of substrate to product importantly would be accom-
plished only by maintenance metabolism of the cells.

The steady-state solution also predicts complete substrate
utilization in the fermentor circuit (see equation 11). The cell mass
(Xf) would increase in the fermentor circuit until substrate becomes

limiting, assuming that there is no accumulation of an inhibitory

product in the fermentor circuit.

Design of Dialysate-Feed, Fermentor-Throughput System

 

The system can be further modified by eliminating the dialysate
effluent stream (Figure 7), so that the metabolite product as well as
the cells are removed only in the fermentor effluent.

A generalized steady-state solution was obtained for this
dialysate-feed, fermentor-throughput system by setting the dialysate
effluent flow rate (Fd) sequal to zero, so that the flow rate of the
dialysate feed (F3) equals that of the fermentor effluent (Ff).
Equations 6 to 10 were then reduced as before.

EXPERIMENTAL MATERIALS AND METHODS
Inoculum

The inoculum culture was obtained from the fermentor effluent

(day 71) of a previous dialysis continuous fermentation (Stieber

and Gerhardt, 1979a) which originally was started with Lactobacillus

 

bulgaricus 2217 (Chris Hanson's Laboratory, Milwaukee, Wisconsin).

 

M

Dried deproteinized whey (prepared by ultrafiltration of whole
sweet-cheese wheys by Stauffer Chemical Co., Rochester, Minnesota)
was rehydrated to contain 62 mg of lactose per ml and was
supplemented with 2 mg of yeast extract per ml. The reconstituted
whey ultrafiltrate was made up in 7-1iter batches without steriliza-
tion, stored at 4°C and held in a stirred, heated (64°C) reservoir
during use.

Fermentation Systems

 

The experimental dialysis continuous fermentations were operated
at a temperature of 44°C and pH of 5.5 in the fermentor and with

essentially the same equipment as used previously (Stieber et al.,

10

1977; Stieber and Gerhardt, 1979a). However, membrane filters of

0.2 pm pore diameter (Versapore 200; Gelman Instrument Company, Ann
Arbor, Michigan) were used instead of the smaller pored dialysis
membrane used before. Eight membranes, providing 2,300 cm2 of

effective area, were used in the dialyzer. The pressures in the fer-
mentor and dialysate circuits varied throughout the fermentation and
were balanced by manually changing the circulation rate (2 to 4 liters
per min.) in the fermentor circuit. A series of steady-state conditions

was managed without asepsis and without interruption for 26 days.

Analytical Procedures

 

Samples were taken at 8 h intervals from the dialysate effluent
and from a glass "T" inserted in the tubing between the fermentor and
dialyzer. Lactose in the samples was determined by the colorimetric
method of Morris (1948). Lactic acid was determined by the colori-
metric method of Pryce (1969). Cell-mass measurements were made by
centrifuging the cells from 10 ml samples, washing the cells twice
with distilled water, and drying them at 105°C for at least 18 h.

EXPERIMENTAL RESULTS

Dialysate-Feed, Immobilized-Cell System

 

The dialysate-feed, immobilized-cell system (Figure 5) was
operated at steady-state conditions (D = 0.92 h-l) for 13 days
(Table IV). As predicted by the simulations, a low level of residual
lactose (Sf) and a high level of lactate (Pf) in the fermentor
circuit were maintained throughout the operation. Complete lactose
utilization did not occur in the fermentor circuit because lactate
accumulated to inhibitory levels. During operation of the system,

the pressures in the fermentor and dialysate circuits kept changing,

11

resulting in variation of the fermentor volume. The variation had
little effect on the lactose and lactate concentrations (Table IV),
but did affect the cell-mass concentration. Although the fermentor
volume (Vf) may change, the total cell mass (Xf) in the fermentor
should remain unchanged (Figure 6). The values for X listed in

f
Table IV are corrected values that are based on a mean Vf of 2,685 ml.
Examination of the data for cell mass (Xf) shown in Table IV
indicated that three different periods occurred during the 13 days of
operating the system. For the first few days Xf remained at an
average value of 1.02 mg/ml, suggesting that the energy resulting from
catabolism was used only for maintenance (nongrowth associated)

purposes. Next there was a four day period of increasing X during

f
which the extent (E) and the rate (-rs) of substrate utilization also
increased, though to a much lesser degree. In the period of the last

six days, X again remained stable and at a mean value (4.54 mg/ml)

f
which was closely predicted by the model (Figure 6).

At a similar dilution rate (D) and extent (E) and rate (-rs) of
substrate utilization, the cell mass (Xf) was 4.5 times greater in
period 3 than in period 1. Substrate was not limiting throughout
the fermentation except on day 16 where the steady-state condition of
the experiment was interrupted with dilute feeding. An explanation
for the increased cell mass in period 3 is that the interruption with
dilute feeding caused each cell to lower its metabolic rate because
of the lower ratio of substrate to cells. After the interruption,
with correct feed concentration and with cells adjusted to a lower

metabolic rate, the cells increased their metabolism to include

that for growth and then the cell mass increased to a new level.

12

 

This transition occurred in period 2. Growth metabolism ceased

at the start of period 3, and the cells remained at the low metabolic
rate. However, E and -r8 remained the same as in period 1 because

of the cell-mass increase. Microscopic examination showed that

there was no difference in the morphology of the cells during the
three periods: the cells appeared homogenous and healthy, and very
little cell debris was observed. This explanation postulates that

L, bulgaricus can maintain itself at different metabolic rates in a

 

given environment. The metabolic rate depends on the initial start-
up conditions (e.g., cell and substrate concentrations, temperature,
etc.). An analogous situation was observed with a continuous culture

of Bacillus linens growing exponentially (Finn and Wilson, 1954). The

 

culture was interrupted twice and, following each interruption, the
system achieved a new steady population of different size. Though the
metabolic rate of each B, linens cell did not change, the accumulation
of cell mass depended on the initial conditions and not on a limiting
concentration of substrate.

From material-balance data (Table IV), the equivalent lactose
concentration in the dialysate-feed stream (S3) was calculated by use
of a conservation-of-mass equation. The mean calculated value
(60.5 mg/ml) agreed well with the mean analytical value (62 mg/ml)
for $2, confirming that no significant portion of the substrate was
lost to products other than lactic acid.

The dialysate-feed, immobilized-cell system was also operated
on fermentation days 4 through 7 at steady state conditions. The

cell mass (Xf) remained stable (mean Xf = 2.66 mg/ml) at a dilution

13

rate (D) of 0.15 h-1 for the four days. The residual lactose was
8.7 mg/ml in the fermentor circuit (Sf) and 29.5 mg/ml in the dialysate
circuit (Sd). The lactate was 52.5 mg/ml and 36.0 mg/ml in the
fermentor and dialysate circuits (Pf and Pd), respectively.
Dialysate-Feed, Fermentor-Throughput System

The dialysate-feed, fermentor-throughput system (Figure 7) was
operated on fermentation days 10 and 11 to demonstrate the process.
At steady-state conditions and a retention time (Tf) of 7.6 h, the
mean fermentor effluent concentrations of cell mass (Xf) and lactate
(Pf) were 0.63 mg/ml and 34.3 mg/ml, respectively. The system may
have application for a continuous fermentation in which the substrate
contains vitamins or other labile molecules that must be continuously
sterilized by dialysis. Also, the energy requirements of the system
may be less than those which utilize the usual procedures of heating.

DISCUSSION

The experimental results indicated that dialysis provides a
novel and useful method for immobilizing microbial cells for the
purpose of producing a cell-free metabolite at a high rate without
appreciable expenditure of substrate for cell growth. A particular
advantage of this particular method is that each cell is not
physically immobilized, thus allowing maximum cell surface area for
fermentation activity. Further, since the cell population as a whole
is immobilized, the process can be operated at a hydraulic dilution
rate greater than the maximum specific growth rate of the bacterium.
The cell mass should increase in the fermentor circuit until substrate
becomes limiting or until an inhibitory concentration of product

accumulates. Thus, removal of the inhibitory concentration by dialysis

14

should result in complete substrate utilization in the fermentor
circuit. Other advantages are that dialysis serves also to sterilize
the feed and the process produces a cell-free fermentation metabolite.
The limiting factor in the process is the solute exchange capacity
of the dialyzer, as indicated by the large differences between S

f
and S and between P and P in Table IV. Further, the differences

d f d
increased as the fermentation progressed.

Improvement in the design of the present dialyzer (Schultz and
Gerhardt, 1969) seems attainable (Dawids and Boe, 1977a; Dawids and
Boe, 1977b). Membrane permeability also should be improvable; the
membrane used in the present study was relatively thick, resulting in
decreased performance. Improvement also should be obtainable by using
a greater membrane area.

Another potential for improvement of the process is to maintain
a constant pressure differential between the fermentor and dialysate
circuits. It was unknown why the pressure kept changing and gradually
increasing throughout the fermentation, as the chemical potential
was always lower in the fermentor circuit. A pressure monitoring
and regulating system for the process is being investigated.

The computer simulations of the model were useful for predicing
the characteristics of the systems. The simulations also increased
that the primary system (Figure 1) could be useful for the continuous
production of a high concentration of cells at a high rate. Though
the process is not simple, it may have potential for the continuous
production of a valuable enzyme or cell, or for producing starter

cultures for the dairy industry.

15

 

ACKNOWLEDGMENT
This work was supported by grants ENG 76-17260 and DAR 79-10236

from the National Science Foundation.

16

1.

2.

3.

37, 487

4.

5.
(1978).

6.

7.

REFERENCES

G. A. Coulman, R. W. Stieber, and P. Gerhardt, Appl. Environ.
Microbiol., 34, 725 (1977).

R. W. Stieber, G. A. Coulman, and P. Gerhardt, Appl. Environ.
Microbiol., 34, 733 (1977).

R. W. Stieber and P. Gerhardt, Appl. Environ. Microbiol.,
(1979a).

 

R. W. Stieber and P. Gerhardt, J, Dairy Sci, submitted (1979c).

P.

F.

C.

Microbiol.,

8.
(1975).

9.
(1979b).

10.
11.
12.

13.
(1954).

14.
(1977a).

15.
(1977b).

A.

Gerhardt and C. A. Reddy, Dev. Ind. Microbiol., 19, 71

W.

A.

 

Juengst, Jr., J, Dairy Sci., 62, 106 (1979).

Reddy, H. E. Henderson, and M. D. Erdman, Appl. Environ.

 

32, 769 (1976).

K.

Keller and P. Gerhardt, Biotechnol Bioeng., 17, 997

 

Stieber and P. Gerhardt, J, Daigy Sci., in press

Schultz and P. Gerhardt, Bacteriol Rev., 33, l (1969).

 

Morris, Science, 107, 254 (1948).
Pryce, Analyst, 94, 1151 (1969).

Finn and R. E. Wilson, J, Agri. Food Chem., 2, 66

 

Dawids and C. Boe, Acta. Med. Scand. Suppl., 607, 7,

 

Dawids and C. Boe, Acta. Med. Scand. Suppl., 607, 17,

 

l7

 

THUELEIII

Glossary of Mathematical Symbols

 

 

 

Symbol Description units
S. Area of membrane available for dialysis cm2
D Hydraulic dilution rate hfl
F: Flow rate into dialysate circuit ml/h
Pd Plow rate out of dialysate circuit Il/h
Pz Plow rate out of fermentor circuit Il/h
Kb Product inhibition constant Igyhg
t3 Substrate limitation constant Ig/ml
P: Product concentration in dialysate feed mg/ml
Pd Product concentration in dialysate circuit mg/ml
Pup Permeability of membrane to product mg/cmz-h
P.s Permeability of membrane to substrate mg/cmz-h
rg Rate of cell growth mg/ml-h
rp Rate of product formation Ig/ml-h
-r. Rate of substrate utilization mg/ml-h
s: Substrate concentration in dialysate feed mg/ml
Sd Substrate concentration in dialysate circuit mg/ml
S! Substrate concentration in fermentor circuit mg/ml
t Time h
V3 volume of liquid in dialysate circuit ml
v2 volume of liquid in fermentor circuit ml
x: Cell-mass concentration in fermentor circuit mg/ml
a Substrate/cell ratio mgflmg
8 Specific maintenance rate h'1
1 Product/substrate ratio mg/Ig
I Efficiency of lactose conversion \
u. Maximu- specific growth rate of cells )1.1
T! Cell-retention time in fer-enter circuit h

 

18

 

TABLE II

Glossary of Dimensionless Parameters

 

TYP‘

Symbol and Definition

Description

 

Material parameters

Operational parameters

Kinetic parameters

-O_O O
’a Pf/(st)

‘- o
’a ' ’a’ ”34’
- O
P, - III/(13¢)

Product factor in dialysate
feed

Product factor in dialysate
circuit

Product factor in fermentor
circuit

Substrate factor in
dialysate circuit

Substrate factor in
fermentor circuit

Cell factor in fermentor
circuit

Ratio of product/substrate
membrane permeabilities

Membrane permeability
factor

Plow-rate ratio

Substrate-limitation
factor

Product-inhibition factor

Time factor

 

19

TABLE III

Values Used in Computer Simulations

 

 

 

Symbola Value
‘1? 2.2
P

'1? 0.001
3

-°

R 3.0

0

5d (mg/m1) 68.0
c (mg/mg) 7.0

B (h-l) 1.2
Y (mg/mg) 0.96

-l
um (h ) 0.35

 

aSymbols are described in Tables I and II.

20

{PAQBIJE II\I

Data From Operation of the Dialysis Immobilized-Cell
§ystem at Steady:State Conditions

 

 

 

 

 

h .
-1 . x! . sf sd 9‘ Pd 9(9) 2. c £3: a

9.2169 Day 0(h ) (00/01) (00/01) (00/91) (90/91) (Ia/:1) (Ig/nJ-h) (Ia/Ill
.095 1.95 17.6 27.9

13 .096 1.11 5.2 12.6 19.5 19.5 57.4 1.64 29.6

.091 0.91 10.0 23.4 33.5 33.0 56.6 2.83 53.9

- 0.96 12.5 26.7 36.5 30.0 51.5 * 55.0

14 .099 0.92 12.4 26.2 41.5 35.0 55.4 2.90 59.7

out .092 1.02 14.9 29.2 43.0 34.5 53.2 2.95 60.2

.092 1.07 11.9 21.9 47.0 35.0 59.1 2.99 54.3

15 .091 1.09 13.9 27.3 45.9 35.9 55.0 3.02 60.6

.091 1.17 11.2 19.9 49.0 41.3 66.2 3.55 59.6

.094 0.93 6.7 14.0 33.0 29.3 64.0 2.42 39.0

16 .092 1.03 7.5 19.9 39.3 29.0 59.2 2.50 46.1

Mean. .091 1.02 12.4 24.0 42.4 35.1 56.7 3.04 57.3

.092 1.49 13.5 29.7 43.5 32.0 49.9 2.70 59.2

.090 1.60 13.9 23.1 47.5 35.5 50.0 2.96 56.1

17 .099 2.59 9.4 22.5 55.3 36.3 60.0 2.99 56.3

.091 3.23 10.7 22.7 46.0 36.9 60.2 3.13 57.0

.092 2.96 9.0 19.2 54.0 42.3 67.5 3.67 59.0

10 .092 3.04 9.9 24.7 59.0 41.9 61.4 3.61 64.0

THU .091 3.72 10.4 21.9 54.0 42.5 64.7 3.65 61.9

.099 3.95 7.0 16.5 59.3 45.6 73.1 3.79 61.9

19 .091 4.09 7.2 21.0 59.0 46.5 67.7 4.01 65.0

.092 4.17 6.5 22.3 59.5 41.3 63.5 3.57 61.1

.096 3.92 5.0 21.3 53.0 46.5 67.4 3.79 65.3

20 .095 3.94 5.0 19.3 54.5 43.3 67.9 3.96 60.1

H9an .091 .. 9.7 22.2 53.6 40.9 - . 60.6

.093 4.59 0.0 23.9 56.0 39.5 60.9 3.44 60.9

.094 4.45 5.3 20.9 51.5 39.9 63.6 3.40 57.1

21 .093 4.19 0.4 27.9 50.5 42.0 59.6 3.69 67.4

.091 4.43 9.4 25.3 49.0 39.3 59.3 3.34 62.1

.091 4.45 9.0 23.9 52.5 39.0 59.0 3.22 59.4

22 .093 4.63 9.0 23.9 49.0 37.5 59.4 3.25 50.9

.092 4.69 11.3 30.1 49.0 36.3 52.9 3.10 63.9

.093 0.55 7.1 26.0 53.3 17.8 56.0 3.26 62.1

runs: 23 .099 4.52 7.4 24.5 52.0 36.9 59.3 3.05 59.9

.095 4.79 4.2 23.6 50.3 34.3 57.4 3.03 55.4

.094 4.91 5.0 27.2 45.0 32.3 52.3 2.90 57.0

24 .091 4.77 6.3 25.7 47.0 34.9 55.7 2.94 50.0

.097 4.42 7.3 32.6 51.0 37.5 51.9 3.39 67.6

.095 4.37 7.6 32.0 49.0 37.3 52.1 3.31 66.9

25 .094 4.34 6.2 31.4 54.0 35.9 51.5 3.12 64.7

.094 4.12 5.9 29.9 53.0 34.9 52.0 3.02 62.1

.094 4.91 3.9 30.0 56.3 40.5 55.9 3.57 69.0

Moan .093 4.54 7.0 27.0 51.2 37.3 56.4 3.23 61.9

 

‘ 9.9.9 on tho 13 day ...: v: o: 2.696 .1.

b 2 - (1 - ‘9":’ x 100.

C O .

d o o o
sd-sd”d-Pdp m"".2osqmo

. 1119 9.99:: do“ not 131011309 day 13 (99929-09) and tho 199: 2/3 of day 16 (diluted feed).

21

 

 

 

 

 

 

 

 

sf,r>1.,xf,t=f - Sd,%,Fd

Fig. 1. Schematic of dailysate-feed system. Symbols are

described in Table I.

22

 

 

 

 

 

 

O 1 1 L
O 20 4O 60 80

1,00

 

Fig. 2. Simulated effects of changes in cell-retention time

(Tf) on residual substrate (3d) and accumulation of product (Pd) in

the dialysate circuit, when 0 = 10.0 and H = 10.0.

23

 

 

 

 

Sf 0r Sd(mg/ml)

 

 

 

 

30— "
Sf
20— ‘
IO - ‘
o_ 1 1 J
o 4 0 I2 I6

Fig. 3. Simulated effects of changes in the membrane permeability

factor (H) on substrate in the dialysate circuit (Sd) and the fermentor

circuit (Sf), when 0 = 10.0 and 0 = 4.0.

24

 

 

Xf(mg/ml)

 

 

 

 

 

/\
0 I L I
0 20 4O 60 80

mm

Fig. 4. Simulated effects of changes in cell-retention time
(Tf) on cell mass (Xf) in the fermentor circuit at three flow-rate

ratios (0), when H a 10.0.

25

 

 

 

 

 

 

 

Fig. 5. Schematic of dialysate-feed, immobilized-cell system.

Symbols are described in Table I.

26

 

 

 

 

 

50 r l l
‘3
50— -
'E 40* '—
‘\
I”
E
,2- 30- -
3“
01° x
0,6 20— f -
IO- 1
50
07 1 1 1

 

Fig. 6. Simulated effects of changes in hydraulic dilution
rate (D) on residual substrate (8d) and accumulation of product (Pd)
in the dialysate circuit and cell mass (Xf) in the fermentor circuit,

when H a 10.0.

27

 

 

 

 

 

 

 

Fig. 7. Schematic of dialysate-feed, fermentor-throughput system.

Symbols are described in Table I.

28

 

203

GENERAL DISCUSSION

The simulated and experimental results together with the review
in this thesis all indicate that membranes can be used in continuous
fermentation processes to enhance microbial growth, substrate
utilization, metabolite production and product recovery, whether the
product be metabolites or cells.

The potential for membrane technology lies principally with its
use to retain microbial cells in a fermentor, thereby increasing the
cell population and allowing increased throughput of substrate
(Article VII). Such a process can produce a cell-free metabolite
at a usefully high concentration and rate, and a concentrated cell
suspension by continuously or periodically harvesting a volume from
the fermentor.

The use of membranes to relieve inhibitory effects by metabolites
is obviously limited to fermentations which produce such metabolites.
In dialysis systems applied to these fermentations, ways must be
found to circumvent the production of large volumes of dilute
permeate, an environmental liability. For the ammonium-lactate
fermentation this was accomplished by feeding substrate at a high
concentration into the fermentor, so that the resulting permeate,
though diluted by water, still contained a useful concentration of
metabolite (Articles I through IV). For a cycloheximide fermentation,

Kominek (1975) used an extractor to eliminate the production of

 

204

large volumes of permeate. These continuous methods also eliminate
the need for a large reservoir vessel, which often is required in a
batch dialysis process (Schultz and Gerhardt, 1969).

For the potential of dialysis in fermentation to be realized, the
solute exchange capacity of the present dialyzer must be improved. The
best membranes for dialysis appear to be membrane filters (pore
diameter of between .05 and 4.0 um) which separate only cells from
solutes. The use of vast areas of membrane not only has physical
limitations, but is limited in maintaining turbulence on membrane
surfaces and in preventing membrane fouling. The available dialyzers
preclude use of mycelial and other organisms that cause rapid fouling.
These considerations along with scale-up problems indicate that the
potential for use of membranes in fermentation probably lies with
refinements of dialysis such as electrodialysis or microfiltration.
These processes can increase solute transfer across the membranes.

A schematic of a microfiltration fed-batch process which may
have application for many types of microbial cells in fermentation is

as follows:

F1 'sfo' Pf.

V

'9"- ,Sf,Pf

 

 

 

 

 

205

The process is designed for the continuous production of cell-
free metabolites and for intermittent harvesting of a concentrated
suspension of cells. Operationally, the process is started batchwise.
The fermentor is filled with a medium containing a relatively low
concentration of substrate and the fermentor is inoculated. As the
cells reach an exponential rate of growth, a feed of substrate into
the fermentor and an effluent flow from the microfilter are started
at the same exponential rate. Because of the relatively low substrate
concentration in the medium and in the feed, diffusible metabolites
will not accumulate to inhibitory levels. Thus for lactobacilli in
the ammonium-lactate fermentation, growth in a confined volume would
proceed at the exponential rate until the cell mass becomes unmanageable
or the microfilter becomes fouled. At such times, a volume (e.g., 90%)
of the concentrated cell culture is harvested from the fermentor, the
volume is replaced with the medimm, and operation is resumed as before.
For an aerobic fermentation, the cmmimal time for cell harvest may be
when the cell mass reaches a level at which oxygen becomes growth
limiting.

Computer control is required for this process in order to
maximize substrate utilization and cell mass productivity and, thus,
to increase the utility of the membranes. As applied to the ammonium-
lactate fermentation, the sensors needed for computer control are
several. Conductance in the microfilter effluent can be monitored by
a conductivity bridge and this information can give the lactate

concentration (P Article III). With the exponential flow rates,

f;

F1 and F2, equal to each other and knowledge of the lactose

concentration in the feed (5:) and P in the effluent, the computer

f

 

206

can maximize substrate utilization by changing the flow rates. The
lactate concentration can be checked by monitoring the rate of ammonia
addition (load cell or flowmeter) and by knowing the fermentor volume
(Vf; level probe). Similarly, F1 and F2 can be determined. Knowing

o and using the mathematical model (Article III),

F1, Sf, Pf, and Vf
the computer can instantly calculate the cell mass concentration, and
can predict and adjust the time for cell harvest. As the fermentation
progresses, the pressure in the conduits may have to be increased
automatically (e.g., by increasing the speed of a circulating pump

or by closing a valve) to compensate for membrane fouling, in order

to maintain the appropriate F A pressure transducer is needed in

2.
the conduits to alarm for excessive pressure, which also may indicate
harvest time.

This computer control scheme maximizes substrate utilization
and cell mass and metabolite production, thereby increasing the
potential for use of membranes in fermentation. The computer also is
necessary to acquire all the data provided by the sensors, to obtain
further data by calculations (e.g., cell mass) and, to make the data
readily available.

The microfiltration process is presented to highlight the
potential for application of membrane technology in fermentation.
Further, it emphasizes that the use of membranes requires complex
control systems. Presently, a microfilter needs to be designed
specifically for microbial purposes; the remaining equipment is
conventional. The capital investment in such a process is great, but

the fermentation potential for production and recovery of both cell

mass and metabolites appears greater.

 

207

SUMMARY-ABSTRACT

Dialysis continuous culture processes were mathematically modeled,
computer simulated and experimentally tested, and the literature was
reviewed, to assess the potential for application of membrane technology
to industrial fermentations. The ammonium-lactate fermentation of
whey was used as a model system for the experiments because of its
commercial potential. The simulated and experimental results both
showed ways that dialysis continuous processes can maximize substrate
utilization and produce high cell populations and cell-free metabolites
at high rates. The dialysis processes were a considerable improvement
over conventional continuous processes.

In one system for dialysis continuous fermentation, whey of high
concentration was fed into a fermentor while the fermentor contents
were dialyzed through a membrane against water to relieve the inhibi-
tory effect of metabolites. The fermentation was modeled and a
solution for the steady state was used to simulate the fermentation on
a digital computer. The results showed the effects of various para-
meters on the process and predicted that it could be operated efficiently.
To test the predictions, dried sweet-cheese whey was rehydrated to
contain 230 mg of lactose per m1, charged into a 5-1iter fermentor,
adjusted in pH (5.3) and temperature (44°C), and inoculated with

Lactobacillus bulgaricus. The fermentor and dialysate circuits were

 

connected, and a series of steady-state conditions was managed without

 

208

sterilization or asepsis for 94 days. As time progressed, the fermen-
tation remained homofermentative and increased in conversion efficiency,
although membrane fouling necessitated dialyzer cleaning about every 4
weeks. With a retention time,of 19 h, 97% of the substrate was
converted into products and cells. Relative to nondialysis continuous
or batch processes for the fermentation, the dialysis continuous
process enabled the use of more concentrated substrate, was more
efficient in the rate of substrate conversion, and additionally
produced a second effluent of less concentrated but purer ammonium
lactate.

The model was improved by incorporating separate terms for
substrate limitation and product inhibition into the equation describing
the rate of cell growth. The improved model was validated using the
results from previous experiments with whole whey and also in sub-
sequent experiments with deproteinized whey. The latter showed that
the cell accumulation, cell productivity, and molar growth yield were
greater than those obtained from a conventional continuous process for
the fermentation and that the dialysis process could produce a con-
tinuous effluent containing 50 mg of cell-free ammonium lactate per
ml. The whole whey proved to be a better substrate than the depro-
teinized whey.

In a second system for dialysis continuous fermentation with a
prefermentor, simulations predicted that the addition would not bring
improvement. In a third system with 100% recycling (feedback) of

cells, simulations predicted that the retention time would be reduced

 

209

9-fold and the conversion of substrate (deproteinized whey) to product
would be increased by 25%, accomplished almost entirely by maintenance
metabolism.

In a fourth system, the use of dialysis for substrate feed as
well as product removal also was modeled and simulated. The simulations
predicted that the feeding of substrate into the fermentor circuit via
the dialysate circuit has potential for production and recovery of
both cell mass and cell-free metabolite. The experimental tests
indicated that the fourth system would be a useful method for immobili-
zing microbial cells to produce a cell-free metabolite at a rate
greater than the maximum specific growth rate of the microbe, and

additionally would serve to filter-sterilize the substrate.

 

210

LITERATURE CITED

 

10.

11.

12.

 

211

LITERATURE CITED

Abbott, B. J. 1978. Immobilized cells, p. 91-123. In_
D. Perlman (ed.), Annual reports on fermentation processes,
Vol. 2, Academic Press, Inc., New York.

Abbott, B. J., and P. Gerhardt. 1970. Dialysis fermentation.
I. Enhanced production of salicylic acid from naphthalene by
Pseudomonas flourescens. Biotechnol. Bioeng. 12:577-589.

 

Abbott, B. J., and P. Gerhardt. 1970. Dialysis fermentation.
II. Kinetic analysis of salicylic acid production via cell
growth and maintenance. Biotechnol. Bioeng. 12:591-601.

Abbott, B. J., and P. Gerhardt. 1970. Dialysis fermentation.
III. Anomalous inhibition of threonine biosynthesis in an

auxotrophic mutant of Escherichia coli. Biotechnol. Bioeng.
12:603-613.

 

Aida, T., and K. Yamaguchi. 1969. Studies on the utilization of
hydrocarbons by yeasts. Part IV. On the dialysis culture of
Mycotorula iaponica and "growth-inhibitory factor" in the
dialyzable material. Agr. Biol. Chem. 33:1244—1250.

 

Alekseeva, G. V., and V. M. Yunker. 1969. A method of culti-
vating tissues in vivo in diffusion chambers. Bulletin of
Experimental Biol. Med. 67:217.

Algire, G. H. 1957. Diffusion chamber technique for studies of
cellular immunity. Ann. N.Y. Acad. Sci. 69 (Article 4):663-
667.

Ambrose, K. R., F. Chandler, and J. H. Coggin, Jr. 1969.
Characterization of tumor-specific transplantation immunity
reactions in immunodiffusion chambers in vivo. Proc. Soc.
Exptl. Biol. Med. 132:1013-1020.

Andersen, K., and K. T. Shanmugam. 1977. Nitrogen fixation:
determination of the ratio of formation of H2 to NH4+
catalyzed by nitrogenase of Klebsiella pneumonia in vivo.
J. Gen. Microbiol. 103:107-122.

 

Arko, R. J. 1972. Neisseria gonorrhoeae: experimental
infection of laboratory animals. Science 177:1200-1201.

 

Arko, R. J. 1973. Implantation and use of a subcutaneous
culture chamber in laboratory animals. Lab. Anim. Sci.
23:105-106.

Arko, R. J. 1974. An immunologic model in laboratory animals

for the study of Neisseria gonorrhoeae. J. Infect. Dis.
129:451-455.

 

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

 

212

Athens, J. W. 1970. Neutrophilic granulocyte kinetics and
granulocytopoiesis, p. 1143-1166. E£.A- S. Gordon (ed.),
Regulation of hemopoiesis, Appleton-Century-Crofts, New York.

Baman, S. I., and R. Haque. 1970. Production of multivalent
extracellular filtrates of Staphylococcus aureus. Can. J.
Microbiol. 16:1255-1261.

 

Barr, R. D., J. Whang-Peng, and S. Perry. 1975. Hemopoietic
stem cells in human peripheral blood. Science 190:284-285.

Baskett, R. C., and W. J. Lulves. 1974. A method of measuring
bacterial growth in aquatic environments using dialysis
culture. J. Fish. Res. Board Can. 31:372-374.

Batzdorf, U., R. S. Knox, S. M. Pokress, and J. C. Kennady.
1969. Membrane partitioning of the Rose-type chamber for the
study of metabolic interaction between different cultures.
Stain Technol. 44:71-74.

Bednarski, M. A., and M. Reporter. 1976. Nitrogenase activity
in rhizobia in vitro: evidence for auxiliary substances
elaborated by cultured soybean cells. Plant Physiol. 57:103.

Bednarski, M. A., and M. Reporter. 1978. Expression of
rhizobial nitrogenase: influence of plant cell-conditioned
medium. Appl. Environ. Microbiol. 36:115-120.

Belding, R. C., J. M. Quarles, T. C. Beaman, and P. Gerhardt.
1976. Exteriorized carotid-jugular shunt for hemodialysis of
the goat. Lab. Anim. Sci. 26:951-954.

Benestad, H. B. 1970. Formation of granulocytes and macro-
phages in diffusion chamber cultures of mouse blood leucocytes.
Scand. J. Haemat. 7:279-288.

Benestad, H. B. 1972. Cell kinetics in diffusion chambers:
survival, resumption of proliferation, and maturation rate of
murine haemopoietic cells. Cell Tissue Kinet. 5:421—431.

Benestad, H. B., J. Iversen, and B. Rolstad. 1971. Macrophage
proliferation in leucocyte populations from rat blood and
lymph during immune responses. Scand. J. Haemat. 8:44-52.

Benestad, H. B., and A. Reikvam. 1975. Diffusion chamber
culturing of haematopoietic cells: methodological investi-
gations and improvement of the technique. Exp. Hemat.
3:249—260.

Bergere, J. L., and J. Hermier. 1968. La production massive
de cellules de Streptocoques lactiques. LeLait 48:13-30.

Berman, I., and H. S. Kaplan. 1959. The cultivation of mouse
bone marrow in vivo. Blood 14:1040-1046.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

 

213

Berman, I., and H. 8. Kaplan. 1960. The functional capacity
of mouse bone marrow cells growing in diffusion chambers.
Exp. Cell Research 20:238-239.

Bhattacharyya, F. K. 1973. Dialysis-culture production of
Vibrio cholerae exotoxin and its precipitation with zinc
acetate. J. Med. Microbiol. 6:499-504.

 

Bissell, M. J., R. Tosi, and L. Gorini. 1971. Mechanism of
excretion of a bacterial proteinase: factors controlling
accumulation of the extracellular proteinase of a Sarcina
strain (coccus P). J. Bacteriol. 105:1099-1109.

Bissonnette, G. K., J. J. Jezeski, G. A. McFeters, and
D. G. Stuart. 1975. Influence of environmental stress on
enumeration of indicator bacteria from natural waters.
Appl. Microbiol. 29:186-194.

Boecker, W. R., A. Boyum, A. L. Carsten, and E. P. Cronkite.
1971. Human bone marrow (HBM) and blood (HPB) stem cell
kinetics. Blood 38:819.

Boecker, W. R., D. K. Hossfeld, W. M. Gallmeir, and
C. G. Schmidt. 1975. Clonal growth of Hodgkin cells.
Nature 258:235—236.

Boecker, W. R., S. Ohl., D. K. Hossfeld, and C. G. Schmidt.
1978. Differentiation of auer-rod positive leukaemic cells
in diffusion chamber culture. Lancet 1:267-269.

Borella, L. 1969. The suppressive effect of Rauscher
leukemia virus on the secondary antibody response of spleen
cells cultured in cell-impermeable diffusion chambers.

J. Immunol. 103:185-195.

Borella, L. 1971. Regulation of IgM memory expression by
spleen cells cultured in diffusion chambers. Immunol.
20:289-298.

Borella, L. 1972. The immunosuppressive effects of Rauscher
leukemia virus (RLV) upon spleen cells cultured in cell-
impermeable diffusion chambers. III. Inhibition of RLV-
induced cell pathways by antigenic stimulation with
hemocyanin. J. Immunol. 108:45-53.

Borichewski, R. M. 1967. Keto acids as growth-limiting
factors in autotrophic growth of Thiobacillus thiooxidans.
J. Bacteriol. 93:597-599.

 

Boyum, A., and R. Borgstrom. 1970. The concentration of
granulocytic stem cells in mouse bone marrow, determined
with diffusion chamber technique. Scand. J. Haemat.
7:294—303.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

214

Boyum, A., W. Boecker, A. L. Carsten, and E. P. Cronkite.
1972. Proliferation of human bone marrow cells in diffusion
chambers implanted into normal or irradiated mice.

Blood 40:163-173.

Boyum, A., A. L. Carsten, O. D. Laerum, and E. P. Cronkite.
1972. Kinetics of cell proliferation of murine bone marrow
cells cultured in diffusion chambers: effect of hypoxia,
bleeding, erythropoietin injections, polycythemia, and
irradiation of the host. Blood 40:174-188.

Boyum, A., D. Lovhaug, and W. R. Boecker. 1976. Regulation
of bone marrow cell growth in diffusion chambers: the
effect of adding normal and leukemic (CML) polymorphonuclear
granulocytes. Blood 45:373-383.

Breivik, H. 1971. Haematopoietic stem cell content of murine
bone marrow, spleen, and blood. Limiting dilution analysis of
diffusion chamber cultures. J. Cell. Physiol. 78:73-78.

Breivik, H., and H. Benestad. 1972. Regulation of granulocyte
and macrophage formation in diffusion chamber cultures of
mouse haematopoietic cells. Exp. Cell Research 70:340-348.

Breivik, H., H. B. Benestad, and A. Boyum. 1971. Diffusion
chamber and spleen colony assay of murine haematopoietic
stem cells. J. Cell. Physiol. 78:65-72.

Brooks, J. R., S. H. Sturgis, G. J. Hill. 1960. An evaluation
of endocrine tissue homotransplantation in the millipore
chamber: with a note on tissue adaptation to the host.

Ann. N.Y. Acad. Sci. 87:482-500.

Buchanan, T. M., and E. C. Gotschlich. 1973. Studies on
gonococcus infection. III. Correlation of gonococcal
colony morphology with infectivity for the chick embryo.
J. Exper. Med. 137:196-200.

Capalbo, E. E., and T. Makinodan. 1964. Doubling time of
mouse spleen cells during the latent and log phases of
primary antibody response. J. Immunol. 92:234-242.

Carsten, A. L., A. D. Chanana, G. Chikkappa, E. P. Cronkite,
and S. Ohl. 1975. An improved diffusion chamber (DC) for
culture of hemopoietic cells. Proc. Soc. Exp. Biol. Med.
150:107-109.

Chikkappa, G., A. L. Carsten, A. D. Chanana, and E. P. Cronkite.
1978. Culture of normal human blood cells in diffusion
chamber systems. I. Granulocyte survival and proliferation.
Exp. Hemat. 6:28-36.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

 

215

Ciccarelli, A. S., D. N. Whaley, L. M. McCroskey, D. F. Gimenez,
V. R. Dowell, Jr., and C. L. Hatheway. 1977. Cultural and
physiological characteristics of Clostridium botulinum type G
and the susceptibility of certain animals to its toxin.

Appl. Environ. Microbiol. 34:843-848.

 

Clive, D., and O. E. Landman. 1970. Reversion of Bacillus
subtilis protoplasts to the bacillary form induced by
exogenous cell wall, bacteria and by growth in membrane
filters. J. Gen. Microbiol. 61:233-243.

Collins, E. B., and A. W. Tillion. 1977. Detection of
antagonistic and symbiotic relationships among bacteria by
diffusion chamber procedure. J. Dairy Sci. 60:387-393.

Crook, R. H., L. V. Scott, and R. A. Patnode. 1969.
Characterization of antigens from Leishmania mexicana grown
in dialysate culture. J. Parasitol. 55:977-981.

 

Dawids, S., and C. Boe. 1977. A new parallel plate dialyzer.
I. Development and testing of membrane support. Acta Med.
Scand. Suppl. 607:7-16.

Dawids, S., and C. Boe. 1977. A new parallel plate dialyzer.
II. Description of design and in vitro performance. Acta
Med. Scand. Suppl. 607:17-24.

Dawids, S., and S. Vojnovic. 1977. A new parallel plate dia-
lyzer. III. Preliminary experience on clinical performance.
Acta Med. Scand. Suppl. 607:27-29.

Dent, P. B., D. Y. E. Perey, M. D. Cooper, and R. A. Good.
1968. Nonspecific stimulation of antibody production in
surgically bursectomized chickens by bursa-containing
diffusion chambers. J. Immunol. 101:799-805.

Dhople, A. M., and J. H. Hanks. 1972. Fate of Mycobacterium
lepraemurium in diffusion chambers incubated in vitro and
in vivo. Int. J. Lepr. 40:210.

 

 

Dhople, A. M., and J. H. Hanks. 1972. The energetics (ATP) of
Mycobacterium lepraemurium in diffusion chambers incubated
in vitro and in vivo. Int. J. Lepr. 40:465-466.

 

Dhople, A. M., and J. H. Hanks. 1973. Energetics (adenosine
5' - triphosphate) of Mycobacterium lepraemurium in
diffusion chambers incubated in vitro and in mice. Infect.
Immun. 8:907-910.

 

Diener, E., and W. E. Armstrong. 1967. Induction of antibody
formation and tolerance in vitro to a purified protein
antigen. Lancet ii:128l-1285.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

 

216

Dobry, D. D., and J. L. Jost. 1977. Computer applications
to fermentation operations, p. 95-114. IE.D° Perlman (ed.),
Annual reports on fermentation processes, Vol. 1, Academic
Press, Inc., New York.

Dor, I. 1975. High density, dialysis culture of algae on
sewage. Water Res. 9:251-254.

Dusanic, D. G. 1969. Cultivation of Trypanosoma lewisi in a
dialysate medium. I. Amino acid alterations during growth.
Comp. Biochem. Physiol. 30:895-901.

 

Duxbury, T. 1977. A microperfusion chamber for studying the
growth of bacterial cells. J. Appl. Bacteriol. 43:247-251.

Edwards, J. H. 1972. The double dialysis method of producing
farmer's lung antigens. J. Lab. Clin. Med. 79:683-688.

Eipert, E. P., L. Adorini, and J. Couderc. 1978. A
miniaturized in vitro diffusion culture system. J. Immunol.
Methods 22:283—292.

Evgenjeva, T. P. 1970. Heterotransplantation of human cancers
to animals by means of diffusion chambers. Europ. J. Cancer
6:533-535.

Fauerholdt, L., and N. Jacobsen. 1975. Cultivation of leukemic
human bone marrow cells in diffusion chambers implanted into
normal and irradiated mice. Blood 45:495-501.

Favero, M. S., N. J. Petersen, L. A. Carson, W. W. Bond, and
S. H. Hindman. 1975. Gram-negative water bacteria in
hemodialysis systems. Health Lab. Sci. 12:321-334.

Fensom, A. H., W. M. Kurowski, and S. J. Pirt. 1974. The use
of ferricyanide for the production of 3-keto sugars by non-
growing suspensions of Agrobacterium tumefaciens. J. Appl.
Chem. Biotechnol. 24:457-467.

 

Ficsor, G., and E. Muthiani. 1971. A microbial assay for
detecting chemical mutagens in tissue homogenates. Mut. Res.
12:335-337.

Fliermans, C. B., and R. W. Gorden. 1977. Modification of
membrane diffusion chambers for deep-water studies. Appl.
Environ. Microbiol. 33:207-210.

Fliermans, C. E., R. W. Gorden, T. C. Hazen, and G. W. Esch.
1977. Aeromonas distribution and survival in a thermally
altered lake. 33:114-122.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

217

Flynn, J., and S. A. Waitkins. 1973. Survival of Neisseria
gonorrhoeae in an artificial subcutaneous cavity of the mouse.
Brit. J. Vener. Dis. 49:432-434.

 

Fogarty, W. M., and P. J. Griffin. 1973. A device for
production of microbial extracellular enzymes in concentrated
form. J. Appl. Chem. Biotechnol. 23:401-406.

Friedman, M. R., and G. L. Gaden, Jr. 1970. Growth and acid
production by Lactobacillus (L,) delbrueckii in a dialysis
culture system. Biotechnol. Bioeng. 12:961-974.

 

 

Gallup, D. M., and P. Gerhardt. 1963. Dialysis fermentor
systems for concentrated culture of microorganisms. Appl.
Microbiol. 11:506-512.

Gardner, W. G., R. B. Prior, and R. L. Perkins. 1973. Fluid
and pharmacological dynamics in a subcutaneous chamber
implanted in rats. Antimicro. Agents Chemother. 4:196-197.

Gengozian, N. 1964. Heterotransplantation of human antibody-
forming cells in diffusion chambers. Ann. N.Y. Acad. Sci.
120 (Part l):9l-ll8.

Germain, G. S., H. Schneider, and E. E. Muirhead. 1966. Multiple
access diffusion chamber and the cultivation of liver cells.
Exp. Cell Res. 43:493-505.

Gerhardt, P., and D. M. Gallup. 1963. Dialysis flask for
concentrated culture of microorganisms. J. Bacteriol.
86:919-929.

Gerhardt, P., and D. M. Gallup. 1965. Process and apparatus for
dialysis fermentation. U.S. Patent 3,186,917.

Gerhardt, P., and C.-G. Hedén. 1960. Concentrated culture of
gonococci in clear liquid medium. Proc. Soc. Exp. Biol. Med.

Gerhardt, P., J. M. Quarles, T. C. Beaman, and R. C. Belding.
1977. Ex vivo hemodialysis culture of microbial and
mammalian cells. J. Infect. Dis. 135:42-50.

Globerson, A., and R. Auerbach. 1965. Primary immune reactions
in organ cultures. Science 149:991-993.

Golde, D. W., and M. J. Cline. 1973. Growth of human bone
marrow in liquid culture. Blood 41:45-57.

Gordon, M. Y., and N. M. Blackett. 1975. Stimulation of
granulocytic colony formation in agar diffusion chambers
implanted in cyclophosphamide pretreated mice. Br. J. Cancer
32:51-59.

89.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

218

Goto, S., S. Enomoto, Y. Takahashi, and R. Motomatsu. 1971.
Slime production by Pseudomonas aeruginosa. I. Conditions
for slime production by the cellophane plate method. Japan.
J. Microbiol. 15:317-324.

 

Green, I. 1966. Experiences with the use of microdiffusion
chamber techniques in man. Int. Arch. Allergy 29:434-446.

Grillo, R. S., and D. A. Spink. 1968. Experiments with tissue
homografts enclosed in diffusion chambers. Oncology 22:227-239.

Groves, D. L., W. E. Lever, and T. Makinodan. 1970. A model for
the interaction of cell types in the generation of hemolytic
plaque-forming cells. J. Immunol. 104:148-165.

Guthrie, R. K., and W. J. Nunez. 1970. Passive transfer of
hypersensitivity by cells enclosed in membrane chambers. Am.

Halsall, M. K., and T. Makinodan. 1974. Analysis of the
limiting-dilution assay used for estimating frequencies of
immunocompetent units. Cell. Immunol. 11:456-465.

Hafiz, S., and M. G. McEntegart. 1977. The survival of
gonococci and meningococci in subcutaneous diffusion chambers
in mice. J. Med. Microbiol. 10:37-42.

Hardt, F. W., J. C. Young, L. S. Clesceri, and D. R. Washington.
1971. Ultrafiltration for the control of recycled solids in a
biological system. Environ. Sci. Technol. 5:345-349.

Harrison, G. A., J. J. T. Owen, and M. A. Ritter. 1968.
Interaction of mouse spleen cells in diffusion chambers.
Nature 219:302—303.

Hendricks, C. W., and S. M. Morrison. 1967. Multiplication and
growth of selected enteric bacteria in clear mountain stream
water. Water Res. 1:567-576.

Hesslein, R. H. 1976. An in situ sampler for close interval pore
water studies. Limnol. Oceanogr. 21:912-914.

Hoelzer, D., E. G. Harriss, M. Slade, and E. Kurrle. 1976.
Growth of in vitro colony-forming cells from normal human
peripheral blood leukocytes cultured in diffusion chambers.
J. Cell. Physiol. 89:89-100.

Hoelzer, D., E. Kurrle, U. Ertl, and A. Milewski. 1974. Growth
of human leukaemic peripheral blood cells in diffusion
chambers. Europ. J. Cancer 10:579-589.

Hoelzer, D., E. Kurrle, H. Schmucker, and E. B. Harriss. 1977.
Evidence for differentiation of human leukemic blood cells in
diffusion chamber culture. Blood 49:729-744.

103.

104.

105.

106.

107.

108.

109.

110.

111.

112.

113.

114.

219

Horng, C. C. 1971. Dialysis culture of mammalian cells.
M.S. Thesis. Michigan State University, East Lansing, Michigan.

Horvath, I., R. J. Arko, and J. C. Bullard. 1975. Effects of
oxygen and nitrogen on the character of T, pallidum in
subcutaneous chambers in mice. Brit. J. Vener. Dis. 51:301-
304.

Huang, C. C., and M. Furukawa. 1978. Diffusion chamber cultures
for mutagenic and carcinogenic assays. Mutation Res.
53:201-202.

Huff, C. G., A. B. Weathersby, A. C. Pipkin, and G. H. Algire.
1960. The growth of exoerythrocytic stages of avian malaria
within diffusion chambers in different hosts. Exp. Parasitol.
9:98-104.

Humphrey, H. E. B. 1970. Dialysis culture of bacteria: dialyzer
design, nutrient transfer, growth, and dialysis aeration.
Ph.D. Thesis. Michigan State University, East Lansing, Michigan.

Hunter, K. M., and I. McVeigh. 1970. Development of a chemically
defined medium for growth of Neisseria gonorrhoeae. Antonie
van Leeuwenhoek 36:305-316.

 

Isaac, L., P. G. Leonard, and G. C. Ware. 1975. Modification of
the Powell culture chamber for the observation and micro-
manipulation of growing bacteria. Lab. Pract. 24:667-669.

Ito, T., and Y. Kishi. 1972. Application of diffusion chamber
technic to the cultivation of Mycobacterium lepraemurium.
I. In vivo studies. Int. J. Lepr. 40:118-129.

 

Jensen, A., and B. Rystad. 1973. Semi-continuous monitoring of
the capacity of sea water for supporting growth of phytoplank-
ton. J. Exp. Mar. Biol. Ecol. 11:275-285.

Jensen, A., and B. Rystad. 1976. Heavy metal tolerance of
marine phytoplankton. II. Capper tolerance of three species
in dialysis and batch cultures. J. Exp. Mar. Biol. Ecol.
22:249-256.

Jensen, A., B. Rystad, and S. Melsom. 1974. Heavy metal
tolerance of marine phytoplankton. I. The tolerance of
three algal species to zinc in costal sea water. J. Exp.
Mar. Biol. Ecol. 15:145-157.

Jensen, A., B. Rystad, and L. Skoglund. 1972. The use of
dialysis culture in phytoplankton studies. J. Exp. Mar.
Biol. Ecol. 8:241-248.

115.

116.

117.

118.

119.

120.

121.

122 O

123.

124.

125.

126.

127.

220

Johnson, K. T., C. G. Hill, Jr., and C. H. Amundson. 1976.
Electrodialysis of raw whey and whey fractionated by
reverse osmosis and ultrafiltration. J. Food Sci. 41:770-777.

Kan, J. K., and M. L. Shuler. 1978. Urocanic acid
production using whole cells immobilized in a hollow fiber
reactor. Biotechnol. Bioeng. 20:217-230.

Kearney, J. E., and P. C. Reade. 1974. The kinetics of
mouse thoracic duct lymphocyte activation by mitogens in
vitro. Aust. J. Exp. Biol. Med. Sci. 52:21-31.

Kitsukawa, H. 1969. In vitro cultivation of rabbit aortic
endothelium and its biologic nature. J. Kurume. Med. Assoc.
32:1439-1511.

Knazek, R. A., P. M. Gullino, and W. R. Kidwell. 1974. Cell
culture on semipermeable tubular membranes. U.S. Patent
3,821,087.

Knazek, R. A., P. M. Gullino, P. O. Kohler, R. L. Dedrick. 1972.
Cell culture on artifical capillaries: an approach to
tissue growth in vitro. Science 178:65-67.

Kominek, L. A. 1975. Cycloheximide production by Streptomyces
griseus: alleviation of end-product inhibition by
dialysis-extraction fermentation. Antimicrob. Agents
Chemother. 7:861-863.

 

Kominek, L. A. 1975. Cycloheximide production by Streptomyces
griseus: control mechanisms of cycloheximide biosynthesis.
Antimicrob. Agents Chemother. 7:856-860.

 

Kominek, L. A. 1975. Process for producing cycloheximide. U.S.
Patent 3,915,803.

Kominek, L. A. 1975. Dialysis process and apparatus. U.S.
Patent 3,915,802.

Kumegawa, M., M. Cattoni, and G. G. Rose. 1968. Electron
microscopy of oral cells in vitro. III. In situ embedding of
cultures in chambers of the circumfusion system. Texas Reports
on Biology and Medicine 26:205-214.

Kuralesova, A. I. 1968. Osteogenic potential of irradiated
bone marrow transplanted in diffusion chambers. Bull. Exp.
Biol. Med. 66:1279—1281.

Laerum, O. D., and A. Boyum. 1970. The separation and
cultivation of basal and differentiating cells from hairless
mouse epidermis. J. Investi. Dermatol. 54:279-287.

128.

129.

130.

131.

132.

133.

134.

135.

136.

137.

138.

139.

 

221

Laerum , O. D., A. Gruneisen, and M. F. Rajewsky. 1973.
Proliferative properties of malignant cell populations
cultured in intraperitoneal diffusion chambers. Europ. J.
Cancer 9:533-541.

Laissue, J., A. D. Chanana, E. P. Cronkite, D. D. Joel, and
M. Pavelec. 1974. Culture of autologous bone marrow (BM) in
diffusion chambers: effect of host irradiation. Federation
Proc. 33:598.

Laissue, J. A., A. D. Chanana, E. P. Cronkite, D. D. Joel, and
M. Pavelec. 1975. Culture of autologous bone marrow in
diffusion chambers: effect of host irradiation. Blood
45:417-425.

Landwall, P. 1978. Dialysis culture for the production of
extracellular protein A from Staphylococcus aureus A676.
J. Appl. Bacterial. 44:151-158.

 

Landwall, P., and T. Holme. 1977. Removal of inhibitors
of bacterial growth by dialysis culture. J. Gen. Microbiol.
103:345-352.

Landwall, P., and T. Holme. 1977. Influence of glucose and
dissolved oxygen concentrations on yields of Escherichia coli
B in dialysis culture. J. Gen. Microbiol. 103:353-358.

 

Landwall, P., and R. Mollby. 1978. Production of Escherichia
coli heat-labile enterotoxin in fermenter dialysis culture.
J. Appl. Bacterial. 44:141-149.

 

Lane, A. G. 1977. Production of food yeast from whey ultra-
filtrate by dialysis culture. J. Appl. Chem. Biotechnol.
27:165-169.

LaScolea, L. J., Jr., M. J. Dul, and F. E. Young. 1975.
Stability of pathogenic colony types of Neisseria gonorrhoeae
in liquid culture by using the parameters of colonial morphology
and deoxyribonucleic acid transformation. J. Clin. Microbiol.
1:165-170.

 

Levey, R. H., N. Trainin, and L. W. Law. 1963. Evidence for
function of thymic tissue in diffusion chambers implanted in
neonatally thymectomized mice. Preliminary report. J. Natl.
Cancer Inst. 31:199-217.

Luedeking, R., and E. L. Piret. 1959. A kinetic study of the
lactic acid fermentation: batch process at controlled pH.
J. Biochem. Microbial. Technol. Eng. 1:393-412.

Maizels, R. M., and D. W. Dresser. 1977. Conditions for the
development of IgM- and IgG-antibody-secreting cells from
primed mouse splenocyte in vitro. Immunol. 32:793-801.

140.

141.

142.

143.

144.

145.

146.

147.

148.

149.

150.

151.

152.

222

Makinodan, T., I. Hoppe, T. Sado, E. E. Capalbo, and M. R.
Leonard. 1965. The suppressive effect of supraoptimum doses
of antigen on the secondary antibody-forming response of
spleen cells cultured in cell-impermeable diffusion chambers.
J. Immunol. 95:466-479.

Makinodan, T., P. Nettesheim, T. Morita, and C. J. Chadwick.
1967. Synthesis of antibody by spleen cells after exposure to
kiloroentgen doses of ionizing radiation. J. Cell.

Physiol. 69:355-366.

Marbrook, J. 1967. Primary immune response in cultures of
spleen cells. Lancet ii:1279-1281.

Mariano, M., and W. G. Spector. 1974. The formation and proper-
ties of macrophage polykaryons (inflammatory giant cells).
J. Path. 113:1-19.

Marmor, J. B., J. L. Russell, A. M. Miller, and S. H. Robinson.
1975. Modulation of murine granulocyte proliferation in
diffusion chamber cultures. Blood 46:39-50.

Masover, G. K., and L. Hayflick. 1974. Dialysis culture of
T—strain mycoplasmas. J. Bacteriol. 118:46-52.

Mayer, L. M. 1976. Chemical water sampling in lakes and
sediments with dialysis bags. Limnol. Oceanogr. 21:909-912.

McFeters, G. A., G. K. Bissonnette, J. J. Jezeski, C. A. Thomson,
and D. G. Stuart. 1974. Comparative survival of indicator
bacteria and enteric pathogens in well water. Appl.

Microbiol. 27:823-829.

McFeters, G. A., and D. G. Stuart. 1972. Survival of coliform
bacteria in natural waters: field and laboratory studies
with membrane-filter chambers. Appl. Microbiol. 24:805-811.

McGarry, M. P., and A. M. Miller. 1974. Evidence for the
humoral stimulation of eosinophil granulocytopoiesis in
in vivo diffusion chambers. Exp. Hemat. 2:372-379.

Meck, R. A., A. L. Carsten, and J. J. Kelsch. 1976. Growth of
HeLa cells in diffusion chamber cultures in vivo. Cancer Res.
36:2317-2320.

Metcalf, T. G., and W. C. Stiles. 1967. Survival of enteric
viruses in estuary waters and shellfish, p. 439-448. I§_
G. Berg (ed.), Transmission of viruses by the water route,
Wiley Interscience, New York.

Metchnikoff, E., E. Roux, and T. (Aurelli) Salimbeni. 1896.
Toxine et antitoxine cholerique. Ann. Inst. Pasteur
10:257-282.

 

153.

154.

155.

156.

157.

158.

159.

160.

161.

162.

163.

164.

165.

223

Michaels, A. S. 1968. New separation technique for the CPI.
Chem. Eng. Prog. 64:31-43.

Millner, S. N. 1969. Apparatus for preparation of bacterial
extracellular enzymes. Appl. Micrdbiol. 17:639-640.

Mohan, K., R. C. Gordon, T. C. Beaman, R. C. Belding, D. Luecke,
C. Edmiston, and P. Gerhardt. 1977. Synergism of penicillin
and gentamicin against Listeria monocytogenes in ex vivo
hemodialysis culture. J. Infect. Dis. 135:51-54.

 

Molongoski, J. J. 1978. Sedimentation and anaerobic metabolism
of particulate organic matter in the sediments of a
hypereutrophic lake. Ph.D. Thesis. Michigan State University,
East Lansing, Michigan.

Naguib, M. 1975. Overall metabolic regulations in cultures of
the obligate methane-oxidizing strain M 102, p. 203-212. In_
Microbial growth on Cl-compounds, Tokyo, Japan.

Nakamura, F., and Y. Kurihara. 1978. Maintenance of a certain
rumen protozoal population in a continuous in vitro fermenta-
tion system. Appl. Environ. Microbiol. 35:500-506.

Nettesheim, P., T. Makinodan, and C. J. Chadwick. 1966.
Improved diffusion chamber cultures for cytokinetic analysis
of antibody response. Immunol. 11:427-439.

Niskanen, E., W. S. Tyler, M. Symann, F. Stohlman, Jr., and
D. Howard. 1974. The effect of neutropenia on the cell cycle
of granulocyte precursors in an in vivo culture system.
Blood 43:23—31.

O'Brien, R. T., and J. S. Newman. 1977. Inactivation of
polioviruses and coxsackieviruses in surface water. Appl.
Environ. Microbiol. 33:334-340.

Osborne, R. J. W. 1977. Production of frozen concentrated
cheese starters by diffusion culture. J. Soc. of Dairy
Technol. 30:40-44.

Osborne, R. J. W., A. T. Miles, and N. Tilley. 1975.
Preparation of concentrated starter cultures. U.S. Patent
3,911,140.

Osebold, J. W., and R. A. DiCapua. 1968. Cellular immunity of
mice infected with Listeria monocytogenes in diffusion
chambers. J. Bacteriol. 95:2158-2164.

 

Osebold, J. W., P. M. Outteridge, and L. D. Pearson. 1970.
Mutation of Listeria monocytogenes after prolonged in vivo
survival in diffusion chambers. Infect. Immun. 1:209-211.

 

 

AL. ”ML—1 __

 

 

166.

167.

168.

169.

170.

171.

172.

173.

174.

175.

176.

177.

224

Osebold, J. W., P. M. Outteridge, L. D. Pearson, and
R. A. DiCapua. 1970. Cellular responses of mice to diffusion
chambers. I. Reactions to intraperitoneal diffusion chambers
containing Listeria monocytogenes and to bacteria-free
chambers. Infect. Immun. 2:127-131.

 

Osebold, J. W., L. D. Pearson, and N. I. Medin. 1974.
Relationship of antimicrobial cellular immunity to delayed
hypersensitivity in listeriosis. Infect. Immun. 9:354—362.

Pan, P., and W. W. Umbreit. 1972. Growth of obligate
autotrophic bacteria on glucose in a continuous flow-through
apparatus. J. Bacteriol. 109:1149-1155.

Pan, P., and W. W. Umbreit. 1972. Growth of mixed cultures of
autotrophic and heterotrophic organisms. Can. J. Microbiol.
18:153-156.

Petersen, B. H., T. Meyer, H. Tjernshaugen. 1974. Kinetics of
rat bone marrow cells cultured in diffusion chambers: effect
of heterologous implantation and irradiation of the host.
Scand. J. Haemat. 13:39-47.

Petithory, J., and J. Rousset. 1969. Immunization of mice
against a homologous strain of virulent trypanosomes living
in diffusion chambers. Bull. Soc. Pathol. Exotique 58:1049-
1053.

Pirt, S. J., and W. M. Kurowski. 1970. An extension of the
theory of the chemostat with feedback of organisms. Its
experimental realization with a yeast culture. J. Gen.
Microbiol. 63:357-366.

Powers, C. D., R. G. Rowland, H. B. O'Connors, Jr., and
C. F. wurster. 1977. Response to polychlorinated biphenyls
of marine phytoplankton isolates cultured under natural
conditions. Appl. Environ. Microbiol. 34:760-764.

Powers, C. D., R. G. Rowland, and C. F. Wurster. 1976. Dialysis
membrane chambers as a device for evaluating impacts of
pollutants on plankton under natural conditions. Water Res.
10:991-994.

Payton, R. 0., and D. Branton. 1970. A multipurpose
microperfusion chamber. Exp. Cell Res. 60:109-114.

Prakash, A., L. Skoglund, B. Rystad, and A. Jensen. 1973.
Growth and cell-size distribution of marine planktonic algae
in batch and dialysis cultures. J. Fish. Res. Board Can.
30:143-155.

Quarles, J. M., Jr. 1973. Hemodialysis culture of bacteria.
Ph.D. Thesis. Michigan State University, East Lansing,
Michigan.

 

 

178.

179.

180.

181.

182.

183.

184.

185.

186.

187.

188.

189.

190.

225

Quarles, J. M., R. C. Belding, T. C. Beaman, and P. Gerhardt.
1974. Hemodialysis culture of Serratia marcescens in a goat-
artificial kidney-fermentor system. Infect. Immun. 9:550-558.

 

Quesenberry, P., E. Niskanen, M. Symann, D. Howard, M. Ryan,
J. Halpern, and F. Stohlman, Jr. 1974. Growth of stem cell
concentrates in diffusion chambers (DC). Cell Tissue Kinet.
7:337—348.

Quesnel, L. B. 1969. Chapter X. Methods of microculture,
p. 365-425. £2_J. R. Norris and D. W. Ribbons (eds.), Methods
in microbiology, Vol. 1, Academic Press, New York.

Rasmussen, P., and O. Hjortdal. 1969. In vivo culture of
blood cells. II. The origin of fibrobasts in blood and blood
buffy coat cultures studied by diffusion chamber implants in
the peritoneal cavity of rats. Acta Anat. 72:476-486.

Rathlev, T. 1973. Cultivation of pathogenic Treponema pallidum
in chambers surgically implanted in experimental animals.
Acta Path. Microbiol. Scand. Section B 81,2:269-271.

 

Reporter, M. 1976. Synergetic cultures of Glycine max root
cells and rhizobia separated by membrane filters. Plant
Physiol. 57:651-655.

 

Reporter, M., and N. Hermina. 1975. Acetylene reduction by
transfilter suspension cultures of Rhizobium japonicum.
Biochem. Biophys. Res. Commun. 64:1126-1133.

 

Rightsel, W. A., and M. F. S. Hall. 1976. Experiences on
attempts at cultivation of M, leprae in diffusion chambers
containing various tissue cultures and maintained for long
periods in mice. Int. J. Lepr. 44:535-536.

Rightsel, W. A., M. F. Sawyers, and J. H. Peters. 1978.
Comparative effects of sulfones and rifampin on growth of
Mycobacterium lepraemurium in macrophage diffusion chamber
cultures. Antimicrob. Agents and Chemother. 13:509-513.

 

Rightsel, W. A., and W. C. Wiygul. 1971. Growth of
Mycobacterium lepraemurium in cell-impermeable diffusion
chambers. Infect. Immun. 3:127-132.

Rose, G. 1954. A separable and multipurpose tissue culture
chamber. Texas Reports on Biology and Medicine 12:1074-1083.

Rose G. G. 1967. The circumfusion system for multipurpose
culture chambers. I. Introduction to the mechanics, techniques,
and basic results of a lZ-chamber (in vitro) closed circulatory
system. J. Cell. Biol. 32:89-112.

Rose, G. G., and M. Cattoni. 1974. Human gingiva cultivated in
circumfusion systems. Archs. Oral Biol. 19:113-123.

191.

192.

193.

194.

195.

196.

197.

198.

199.

200.

201.

202.

 

226

Rose, G. G., M. Kumegawa, and M. Cattoni. 1968. The
circumfusion system for multipurpose culture chambers.
II. The protracted maintenance of differentiation of fetal
and newborn mouse liver in vitro. J. Cell. Biol. 39:430-450.

Rose, G. G., M. Kumegawa, H. Nikai, M. Bracho, and M. Cattoni.
1970. The Dual-rotary circumfusion system for mark II culture
chambers. I. Design, control, and monitoring of the system
and the cultures. Microvasc. Res. 2:24-60.

Rose, G. G., M. Kumegawa, H. Nikai, M. Cattoni, and F. Hu. 1969.
The HFH-l8 mouse melanoma in roller tube, chamber, and circum-
fusion system cultures. Cancer Res. 29:2010—2033.

Rose, G. G., C. M. Pomerat, T. 0. Shindler, and J. B. Trunnell.
1958. A cellophane-strip technique for culturing tissue in
multipurpose culture chambers. J. Biophys. Biochem. Cytol.
4:761-764.

Rose, G. G., and T. Yajima. 1977. Fetal mouse lung in
circumfusion system cultures. In Vitro 13:749-768.

Rose, G. G., and T. Yajima. 1978. Terminal bronchiolar-alveolar
(TB-A) units in circumfusion system cultures. In Vitro
14:557-580.

Rosin, A., H. Freiberg, and G. Zajicek. 1963. The fate of rat
bone marrow, spleen and periosteum cultivated in vivo in the
diffusion chamber, with special reference to bone formation.
Exp. Cell Res. 29:176-187.

Rothstein, G., E. H. Hugl, C. R. Bishop, J. W. Athens, and
H. E. Ashenbrucker. 1971. Stimulation of granulocytopoiesis
by a diffusible factor in vivo. J. Clin. Investigation
50:2004—2007.

Ruezinsky, L. L. 1975. Dialysis culture of Neisseria
gonorrhoeae. M.S. Thesis, Michigan State University, East
Lansing, Michigan.

 

Sado, T. 1969. Functional and ultrastructural studies of
antibody-producing cells exposed to 10,000 R in millipore
diffusion chambers. Int. J. Radiat. Biol. 15:1-22.

Sado, T., E. H. Perkins, and T. Makinodan. 1970. Staircase
rise in the antibody-forming cell population in secondary
response. J. Immunol. 105:642-652.

Sakshaug, E. 1977. Limiting nutrients and maximum growth rates
for diatoms in Narragansett Bay. J. Exp. Mar. Biol. Ecol.
28:109-123.

 

 

203.

204.

205.

206.

207.

208.

209.

210.

211.

212.

213.

214.

227

Sarner, N. Z., M. J. Bissell, M. DiGirolamo, and L. Gorini. 1971.
Mechanism of excretion of a bacterial proteinase: demonstration
of two proteolytic enzymes produced by a Sarcina strain (Coccus
P). J. Bacteriol. 105:1090-1098.

 

Scales, R. W., and S. J. Kraus. 1974. Development and passive
transfer of immunity to gonococcal infection in guinea pigs.
Infect. Immun. 10:1040-1043.

Schieferstein, G., and O. D. Laerum. 1974. Diffusion chamber
culture of hamster melanoma cells in vivo. Culture
characteristics in iso- and heterologous host animals.

Arch. Derm. Forsch. 249:131-140.

Schneiberg, K., A. Jonecko, and W. Bartinikowa. 1968.
Thymus--a hematopoietic organ? III. Influence of thymus-
bearing chambers on the surviving and the hemopoietic system
of mice after a whole-body irradiation with x-rays. Folia
Haematol. 89:265-282. ‘

. 17_ '9. .123—a

Schultz, J. S., and P. Gerhardt. 1969. Dialysis culture of
microorganisms: design, theory, and results. Bacterial. Rev.
33:1-47.

Skoglund, L., and A. Jensen. 1976. Studies on N-limited
growth of diatoms in dialysis culture. J. Exp. Mar. Biol.
Ecol. 21:169-178.

Sortland, L. D., and C. R. Wilke. 1969. Growth of Streptococcus
faecalus in dense culture. Biotechnol. Bioeng. 11:805-841.

 

Spertzel, R. 0., and M. Pollard. 1970. Heterotransplanted spleen
tissues in diffusion chambers: therapeutic benefits to
irradiated mice. J. Reticuloendothelial Soc. 8:1-12.

Squires, D.J.P. 1975. The growth of human bone marrow in
diffusion chambers. Brit. J. Haematol. 29:89-97.

Stieber, R. W. 1976. Nonaseptic continuous dialysis fermenta-
tion of concentrated whey to produce ruminant feedstuff and
lactic acid. M.S. Thesis, Michigan State University, East
Lansing, Michigan.

Stohlman, F., Jr., P. J. Quesenberry, and W. S. Tyler. 1973.
The regulation of myelOpoiesis as approached with in vivo and
in vitro techniques. Prog. Hematol. 8:259-297.

Stutman, 0., E. J. Yunis, and R. A. Good. 1970. Studies on
thymus function. I. Cooperative effect of thymic function
and lymphohemopoietic cells in restoration of neonatally
thymectomized mice. J. Exp. Med. 132:583—600.

215.

216.

217.

218.

219.

220.

221.

222.

223.

2240

225.

226.

 

228

Suzuki, J. B., R. Booth, A. Benedik, and N. Grecz. 1971.
Pathogenesis of Clostridium botulinum type A: study of in
vivo toxin release by implantation of diffusion chambers

containing spores, vegetative cells, and free toxin. Infect.
Immun. 3:659—663.

 

Tight, R. R., and R. L. Perkins. 1976. Treponema pallidum
infection in subcutaneous polyethylene chambers in rabbits.
Infect. Immun. 13:1606-1612.

 

Tight, R. R., R. B. Prior, R. L. Perkins, and C. A. Rotilie.
1975. Fluid and penicillin G dynamics in polyethylene
chambers implanted subcultaneously in rabbits. Antimicrob.
Agents Chemother. 8:495-497.

Trench, C. A. H., J. W. Watson, F. C. Walker, P. S. Gardner, and
C. A. Green. 1966. Evidence for a humoral thymic factor in
rabbits. Immunol. 10:187-191.

Trowell, O. A. 1959. The culture of mature organs in a synthetic
medium. Exp. Cell Res. 16:118-147.

Trowell, O. A. 1963. The optimum concentration of sodium
chloride for the survival of lymphocytes i§_vitro. Exp.
Cell Res. 29:220-224.

Tucker, D. F., and J. J. T. Owen. 1969. Growth of tumour cells
in diffusion chambers on the chick chorioallantois. Europ.
J. Cancer 5:591-596.

Tyler, W. S, E. Niskanen, F. Stohlman, Jr., J. Keane, and
D. Howard. 1972. The effect of neutropenia on myeloid
growth and the stem cell in an in vivo culture system.
Blood 40:634-645.

Tyler, W. S., F. Stohlman, Jr., M. Chovaniec, and D. Howard.
1976. Effect of a congenital defect in hemopoiesis on
myeloid growth and the stem cell (CFU) in an in vivo culture
system. Blood 47:413-421.

Urso, P., and T. Makinodan. 1963. The roles of cellular
division and maturation in the formation of precipitating
antibody. J. Immunol. 90:897-907.

Vann, D. C. 1969. $2 vitro antibody synthesis by diffusion
chamber cultures of spleen cells. II. Effect of increased
levels of free antibody. J. Immunol. 102:451-456.

Vann, D. C., and T. Makinodan. 1969. In vitro antibody
synthesis by diffusion chamber cultures of spleen cells.
I. Methods and effect of 10,000 r an antibody synthesis.
J. Immunol. 102:442-450.

 

227.

228.

229.

230.

231.

232.

233.

234.

235.

236.

237.

238.

239.

229

Vargo, G. A., P. E. Hargraves, and P. Johnson. 1975. Scanning
electron microscopy of dialysis tubes incubated in flowing
seawater. Marine Biol. 31:113-120.

Vasconcelos, G. J., and R. G. Swartz. 1976. Survival of
bacteria in seawater using a diffusion chamber apparatus
in situ. Appl. Environ. Microbiol. 31:913-920.

Videman, T., L. Vilpo, and J. A. Vilpo. 1978. Inhibition of
osteogenesis by bone mass in diffusion chamber cultures.
Exp. Path. Bd. 15:182-184.

Vieth, W. R., and K. Venkatasubramanian. 1974. Part. IV.
Process engineering for immobilized enzyme systems. Chem.
Technol. 4:434-444.

Vilpo, J. A. 1972. Proliferation, survival and malignancy of
chloroleukaemia cells grown in diffusion chambers in vivo.
Annales Medicinae Experimentalis et Biologiae Fenniae 50:168-
174.

Vilpo, J. A., T. Videman, and L. Vilpo. 1978. An experimental
model of osteogenesis: a closed in vivo culture system using
diffusion chambers. Clin. Orthop. 135:291-294.

Waitkins, S. A. 1975. Effect of tissue culture cells in
promoting prolonged survival of N, gonorrhoeae in artificial
subcutaneous cavities of mice. Brit. J. Vener. Dis. 51:376-381.

 

Wang, D.I.C., A. J. Sinskey, and T. A. Butterworth. 1970.
Enzyme processing using ultrafiltration membranes, p. 98-119.
I2_J. E. Flinn (ed.), Membrane science and technology, Plenum
Press, New York.

Willardsen, R. R., F. F. Busta, and C. E. Allen. 1977. Dialysis
technique for containment of microbial populations inoculated
into food systems. Appl. Environ. Microbiol. 34:240-241.

Willemze, R., R. I. Walker, J. C. Herion, and J. G. Palmer. 1978.
Marrow culture in diffusion chambers in rabbits. I. Effect of
mature granulocytes on cell production. Blood 51:21—31.

Winfrey, M. R., and J. G. Zeikus. 1977. Effect of sulfate on
carbon and electron flow during microbial methanogenesis in
freshwater sediments. Appl. Environ. Microbiol. 33:275-281.

Wiygul, W. C., and W. A. Rightsel. 1971. Growth of
Mycobacterium lepraemurium in diffusion chambers containing
human embryonic skin cells and in cell-free chambers.

J. Gen. Microbiol. 68:375-380.

 

wyrick, P. B., and H. Gooder. 1971. Growth of streptococcal
protoplasts and L-colonies on membrane filters. J. Bacteriol.
105:646-656.

3“" .1 ‘" ‘m

240.

241.

230

Young, P. 1974. An introduction to electrodialysis. J. Soc.
Dairy Technol. 27:141-150.

Zabriskie, D. W., and A. E. Humphrey. 1978. Continuous dialysis
for the on-line analysis of diffusible components in a
fermentation broth. Biotechnol. Bioeng. 20:1295-1301.

 

malt-059511: ,

‘- I 5.9"

 

IES

”11111111111111"