Use of a Structured Kinetic Model of
Antibody Synthesis and Secretion for
Optimization of Antibody Production
Systems: I. Steadystate Analysis
Theodora A. Bibila*
Department of Chemical Engineering and Materials Science and Institute for
Advanced Studies, in Biological Process Technology, University of Minnesota,
St. Paul, Minnesota 55108
Michael C. Flickinger'
Department of Biochemistry and Institute for Advanced Studies in Biological
Process Technology, University of Minnesota, St. Paul, Minnesota 55108
Received January 21, 199IIAccepted June 26, 1991
Steady-state simulations using our previously developed
structured kinetic model of antibody synthesis and secre-
tion by hybridoma cells are used here in conjunction with
factorial design analysis to identify intracellular parameters
important in determining the specific antibody secretion
rate and predict the dependence of this rate on cell spe-
cific growth rate. Simulation results suggest that the spe-
cific growth rate, the assembly rate of the heavy and light
chains and the heavy- and light-chain gene dosage can sig-
nificantly affect the rate of antibody secretion. Based on
these results, environmentaland/or genetic manipulation ap-
proaches are proposed for maximizing the specific antibody
secretion rate and the antibody volumetric productivity in
large-scale antibody production systems.
Key words: monoclonal antibody secretory pathway
structured kinetic model factorial design steady-state
- otic and eucaryotic system^,^^,^^,^^ here we report the re-
simulations
INTRODUCTION
Optimization of the volumetric productivity of mono-
clonal antibody (MAb) secretion in large-scale hybri-
doma cultivation systems (micrograms of antibody per
unit volume of reactor per unit time) depends on (1) the
specific antibody secretion rate (micrograms of anti-
body per cell per unit time) and (2) the concentration of
viable cells in the culture that actively secrete antibody
(viable cells per unit volume of reactor). While the sec-
ond point can be directly addressed by appropriate
bioreactor and/or growth medium design, optimization
of the specific antibody secretion rate requires a thor-
ough understanding of the regulation of the pathway of
antibody secretion.
As a first step in better understanding the cellular
regulation of the biosynthesis and secretion of mono-
clonal antibody by hybridoma cells, we have previously
This work was presented at the 1990 American Institute of
Chemical Engineers (AIChE) Annual Meeting, Chicago Illinois.
* Current address: Merck Sharp & Dohme Research Labora-
tories, P.O. Box 2000, Rahway, NJ 07065-0900.
To whom all correspondence should be addressed.
Biotechnology and Bioengineering, Vol. 39, Pp. 251-261 (1992)
0 1992 John Wiley & Sons, Inc.
developed and experimentally validated a structured
kinetic model describing the main intracellular steps of
the monoclonal antibody secretory Tran-
scription of mRNAs coding for the heavy and light anti-
body chains, their translation, assembly of the heavy and
light chains to give the tetrameric antibody m~lecule,~
intraorganelle processing and interorganelle transport of
the antibody, and final secretion4 were considered.
Although attempts to describe individual steps of pro-
tein biosynthesis and the secretory pathway for secreted
proteins have been previously reported in both procary-
sults of steady-state model simulations of behavior of
the entire pathway of MAb synthesis, assembly, inter-
organelle transport, and secretion. A summary of the
model equations (i.e., the intracellular balances of path-
way intermediates) previously r e p ~ r t e dis
~ .included
~ in
the Appendix.
Here, the model is used as a tool to identify the pre-
dominant factors that affect the specific antibody secre-
tion rate, predict the growth rate dependence of the
specific antibody secretion rate, and gain a better in-
sight into possible environmental or genetic manipula-
tion approaches that would maximize this rate. As an
initial approach, steady-state model simulations are pre-
sented and the sensitivity of the secretion rate to
changes in different model parameters under these con-
ditions is analyzed by use of a factorial design al-
gorithm. The simulation results are used to propose
approaches to optimize the secretion rate and the over-
all volumetric productivity in large-scale antibody pro-
duction systems. These approaches may also be useful
for optimization of protein synthesis and secretion from
other lymphoid cells.24
Previous to this work, efforts to maximize antibody
secretion have focused on the effect of extracellular
nutrients or environmental parameters on hybridoma
growth and metabolism (pH, temperature, dissolved
CCC 0006-3592/92/030251-11$04.00
oxygen concentration, serum, carbon and energy The steady-state intracellular pool of free (unassembled)
sources, and inhibitory growth by-products such as light chains, [L],,, is given by solving the following third-
lactate and ammonia).22.29.32.34,35,40,46,55.56,62 The unstruc- order equation
tured and structured kinetic models previously devel-
oped by other investigators to describe hybridoma and A . [L]:, + B . [L]:, + C . [L],, + D = 0 (4)
mammalian cell systems in general are primarily growth where
models.',11361In contrast, our approach to maximizing
antibody secretion focuses on the intracellular pathway
for antibody synthesis to identify specific pathway steps
important in determining the overall secretion rate.
Once such a step(s) has been identified, the growth
medium composition or culture environment may be
changed or posttranscriptional steps in the secretory
pathway may be genetically manipulated to maximize
the rate of this step(s) and thereby the rate of antibody
secretion.

RESULTS AND DISCUSSION
Steady-State Model Solution
As an initial approach, a steady-state assumption was
used specifying that there is no intracellular accumula-
tion of any of the species involved in the antibody secre-
tory pathway, i.e., no intracellular accumulation of the
heavy- and light-chain coding mRNAs, the heavy- and
light-chain polypeptides, and the antibody assembly
and transport intermediates. Savinell et al.52have previ-
ously determined a range for the steady-state theoretical
maximum specific antibody secretion rate based on the
maximum theoretical capacity of the transcriptional
and translational machinery in the cell; however, in that
analysis only simplified intracellular balances of im-
munoglobulin mRNAs and polypeptides were consid-
ered with no detailed description of all the steps in the
intracellular secretory pathway.
For the steady-state solution of the system of model
equations [Equations (i)-(x) in the Appendix], the in-
tracellular accumulation terms (d/dt)were set equal to
zero and the resulting set of equations solved analyti-
cally. It should be noted that the model assumes that
there is no proteolytic or other degradation of the heavy-
and light-chain polypeptides in the cell and also that no
free heavy or light chains or assembly intermediates es-
cape from the endoplasmic reticulum during processing.
The steady-state intracellular pools of heavy and light
mRNAs ([mHIssand [mLlsS),and free (unassembled)
heavy chains ([HI,,) are given by
[~H]SS = NH' S H / ( K p)
[ ~ L =I NL
~ .~S L / ( K + p )
Equation (4)was solved using Graeffe's root squaring
method for numerical solution of nonlinear algebraic
equations." For the parameter values used in these simu-
lations, there was always one positive real root; the
other two roots of Equation (4) were either negative or
complex. Having determined [L],,, the steady-state in-
tracellular concentrations of all antibody assembly in-
termediates as well as the antibody concentrations in
the endoplasmic reticulum (ER) and the Golgi and the
steady-state specific antibody secretion rate can be cal-
culated from
Model Parameter Values
For the steady-state simulations, model parameter values
were varied based upon either experimental results ob-
tained with the 9.2.27 hybridoma cell or data
reported in the literature for immunoglobulin secre-
ting myeloma or hybridoma cells. The specific growth
rate was varied between 0 and 0.04 h-'; 0.04 h-' is the
(1)
maximum specific growth rate determined during expo-
(2) nential growth of 9.2.27 hybridoma cells." During cul-
tivation of 9.2.27 hybridoma cells in a total cell recycle
perfusion bioreactor, growth rates of populations of cells
at high viability very close to zero have been experimen-
tally attained.18
The number of heavy- and light-chain gene copies/
cell ( N Hand N t ) was set equal to 1. In both myeloma

252 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 39, NO. 3, FEBRUARY 5, 1992

and normal spleen cells (the two partners in hybri-
doma fusions) usually only one of the two heavy and
one of the two light-chain alleles are productively rear-
ranged and transcribed into functional heavy- and light-
chain mRNAs, respectively, a phenomenon known as
allelic e x c l ~ s i o n . ~The
' , ~ ~ remaining allele is either un-
rearranged (remains in the germline configuration) or
unproductively rearranged. Since myelomas used for
generating hybridoma fusions usually undergo a process
of selection for nonproductivity before fusion with the
spleen cells, the productively rearranged copies of the
heavy- and light-chain genes from the myeloma parent
are lost during this process. The resulting hybridoma
cells therefore inherit only one copy of each of the func-
tional heavy- and light-chain genes from their spleen
cell parent, suggesting that NH = Nt = 1. In order to
examine the effect of gene dosage on the specific anti-
body secretion rate, NH and NL were varied from 1 to 5
to 10 copies/cell. The number of copies of immunoglobu-
lin coding genes that can be stably integrated in the
genome of lymphoid cells following transfection with
appropriately designed plasmid vectors and successfully
expressed is between 1 and 10 for both heavy- and light-
chain genes.I3 The heavy- and light-chain gene copy
number can be further increased up to 200 copies/cell
by use of amplification systems such as the dihydrofolate
reductase (DHFR)/methotrexate (MTX) amplification
system.25x37 At very high gene copy numbers, transcrip-
tion and expression may be limited by the intracellular
concentration of factors necessary for t r a n ~ c r i p t i o nIt. ~ ~
has been suggested that the transcriptional and trans-
lational machinery of a typical lymphoid cell could
support the expression of up to 60 immunoglobulin
coding genes.37
The specific transcription rates of the heavy- and
light-chain coding mRNAs were taken as 3000 and
4500 mRNA molecules/gene/h, re~pectively.~~ For the
simulations at high heavy- and light-chain gene copy
numbers ( 5 and 10 copies/cell) it is assumed here that
the extra copies of the heavy- and light-chain genes are
transcribed as efficiently and at the same rate as the
endogenous heavy- and light-chain genes. However, the
transcription efficiency of transfected genes is often
dependent on the site of their integration in the chromo-
some and is usually lower than the transcription effi-
ciency at the endogenous gene
The half-lives of the 9.2.27 heavy- and light-chain
mRNAs have been previously determined as a function
of the growth state of 9.2.27 cells.3 Heavy- and light-
chain mRNA half-lives were found to be statistically
the same, with their values decreasing from about 12 h
in the exponential phase of growth to about 6 h in the
stationary growth phase.3 In the model simulations pre-
sented here, the heavy- and light-chain mRNA half-
lives were varied linearly with specific growth rate. The
specific translation rates of the heavy and light chains
were determined by the number of translating ribo-
BlBlLA AND FLICKINGER: STEADY-STATE ANALYSIS OF A STRUCTURED MODEL
somes associated with the heavy and light mRNAs
(ribosome loading) and the speed with which each ribo-
some is moving along the mRNA m ~ l e c u l eIt. ~has been
previously shown that in immunoglobulin secreting cells
heavy- and light-chain specific translation rates decrease
with entrance of cells from the exponential into the sta-
tionary growth phase and that this decrease is due to a
decrease in the ribosome loading on the heavy and light
~ R N A s . ~In' our own experiments, we have found that
the total cellular RNA content of 9.2.27 cells decreases
linearly with decreasing specific growth rate during the
course of batch c ~ l t i v a t i o nTotal
. ~ cellular RNA is pri-
marily ribosomal RNA ( 8 0 4 5 % ) and could thus be con-
sidered as a measure of the availability of ribosomes to
be used in translation. In the simulations presented here,
the ribosomal loading of the heavy and light mRNAs
(and thereby the heavy- and light-chain specific transla-
tion rates) was varied linearly with specific growth rate
proportional to the experimentally observed linear vari-
ation of total cellular RNA content.
The heavy- and light-chain assembly rate constant was
varied between lo-'' and ((molecules/cell) . h)-'. To
our knowledge no one has experimentally measured
in vivo assembly rate constants as a function of growth
rate or culture conditions in the antibody or any other
multimeric glycoprotein secretory pathway for that
matter. Instead, the assembly time for heavy and light
immunoglobulin chains, i.e., the time required for com-
plete assembly of a given intracellular pool of heavy
and light chains have been determined in vivo in expo-
nentially growing cells by pulse-chase experiments.
This time is between 10 and 20 min; however, its depen-
dence on cell growth rate and culture conditions has not
been studied.' Determination of the assembly rate con-
stant K A from such assembly time data is not straight-
forward since the experimentally determined assembly
time is a function of both K A and the intracellular
heavy- and light-chain concentrations. In the absence
of any literature data on assembly rate constant values,
the range of KA values used in our model simulations
was chosen so that the resulting characteristic assembly
times predicted by the model are between 10 and
20 min, close to the experimentally reported assembly
times. The half-times for ER-to-Golgi transport and
Golgi-to-extracellular-medium transport were varied
between 30 and 60 min and 5 and 20 min, respectively.
These ranges are typical for ER-to-Golgi transport and
Golgi-to-medium transport in constitutive secretory
pathways. 14,21,27,44,45,59,63,64
Factorial Design Analysis
In order to quantitate the relative effects of different
model parameters on the steady-state specific anti-
body secretion rate as well as the extent of parameter
interaction, simulation results were analyzed using a
2' factorial d e ~ i g n In
. ~ this analysis, a minimum and a
253
maximum value for each one of the five variable parame-
ters in the model (heavy- and light-chain gene dosage,
growth rate coupled with heavy- and light-chain spe-
cific translation rates, assembly rate constant, half-time
for ER-to-Golgi transport, and half-time for Golgi-
to-extracellular-medium transport) was considered, as
described above. The steady-state specific antibody se-
cretion rate was calculated for all possible combinations
(z5 = 32) of parameter values. Yates’s algorithm’ was
then used to calculate the relative effects of each one
of the parameters on the steady-state specific antibody
secretion rate as well as the extent of interaction be-
tween any two, three, four, or all five of the parameters
(Table I). The relative effect of each parameter on the
specific antibody secretion rate would be equivalent to
a normalized (dimensionless) “sensitivity coefficient”
showing the fractional change in the steady-state spe-
cific antibody secretion rate mediated by a fractional
change in the parameter of interest.
Table I. Estimated relative effects of model parameters on the
steady-state specific antibody secretion rate from a 2’ factorial
design.
~ ~
The results of the factorial design analysis (Table I)
suggest that the heavy- and light-chain gene dosage, the
assembly rate, and the growth ratehranslation rate are
the parameters that mostly affect the specific antibody
secretion rate. However, due to considerable interactions
between these parameters, no definite statement can be
made about the effect of each parameter alone. There is
a strong interaction between the specific growth rate
and the assembly rate constant (12,395). There is also a
strong interaction between the heavy- and light-chain
gene dosage and the assembly rate constant (10,145)and
a considerable interaction between the gene dosage and
the specific growth rate (3814). As far as higher order
interactions (interactions between more than two pa-
rameters), the three-parameter interaction between
growth rate, assembly rate constant, and gene dosage
appears to be important (10,145).
In order to visualize the quantitative effect of each
one of the model parameters on the specific antibody
secretion rate, the steady-state specific antibody secre-
tion rate was plotted as a function of the specific growth
rate and a range of values of the parameter of interest.
~~~~

The results of this analysis are discussed in the follow-

1 2 3 4 5 Identity Relative Effect ing sections.
Average -
+ - - 1 4 6 31
- + - 2 12,395 Effect of Heavy- and Light-Chain Gene Dosage on
+ + - 12 12,395 Specific MAb Secretion Rate
- - + 3 -173
+ - + 13 - 173 To predict the effect of the gene dosage (gene copies/
- + + 23 -171.4 cell) of the heavy- and light-chain genes on the specific
+ + + - 123 -171.4
antibody secretion rate, specific secretion rate vs. spe-
+ 4 - 89
+ 14 - 89 cific growth rate simulation profiles were generated for
+ 24 -88 1, 5, and 10 copies/cell of the heavy- and light-chain
+ + - + 124 - 88 genes. Figure 1 shows the simulation results for two val-
- - + + 34 1.2 ues of the assembly rate constant and lo-’’ ((mole-
+ - + + 134 1.2
cules/cell) * h)-’, resulting in positive or negative corre-
- + + + 234 1.2
+ + + + 1234 1.2 lation between secretion rate and growth rate, respec-
- - - - 5 16,709 tively (as discussed below). In both cases, the specific
+ - - - 15 3,814 secretion rate at a given specific growth rate increases
- + - 25 10,145 with increasing gene copy number. However, an in-
+ + - 125 10,145
- - + 35 - 141.9 crease in gene copy number does not result in a propor-
+ - + 135 -141.9 tional increase in the specific secretion rate except for
- + + - 235 -140.3 the case of specific growth rate equal to zero. For growth
+ + + - 1235 -140.3 rates higher than zero the predicted increase in specific
- - - + 45 - 73
secretion rate is higher than what would be predicted by
+ - - + 145 -73
a simple proportional relationship. This deviation from
- + - + 245 -72.3
+ + - + 1245 -72.3 a proportional reistionship increases with increasing
- - + + 345 1 growth rate. For example, for KA = ((molecules/
+ - + + 1345 1 cell) * h)-’ (Fig. la) at the maximum specific growth rate
- + + + 2345 1
of 0.04 h-’, the specific secretion rate increases by 6.5-
+ + + + 12345 1
and 14-fold with increases of the gene copy number to 5
Note: Parameters 1-5 correspond to the specific growth rate and 10 copies/cell, respectively. In the case where KA =
(coupled with the heavy- and light-chain translation rates), the as- lo-’’ ((molecules/cell) h)-’ (Fig. lb), at the maximum
sembly rate constant (&), the ER-to-Golgi transport half-time specific growth rate, the specific secretion rate increases
( t l , z , E R ) , the Golgi-to-extracellular-medium transport half-time
( t l i z . c )and , the heavy- and light-chain gene dosage ( N H= N L ) re- by 90- and 400-fold with increases of the gene copy num-
spectively. The pluses and minuses represent the maximum and ber to 5 and 10 copies/cell, respectively. It therefore
minimum values of each parameter, respectively. appears that heavy- and light-chain gene dosage has a

254 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 39, NO. 3, FEBRUARY 5, 1992

 bly rate until a plateau (saturation) level is reached. At

k=F
- I I

N=10 this point, an increase in the assembly rate constant

does not result in further increase of the specific secre-
tion rate. At very low assembly rates only a small per-
centage of the free heavy and light chains available for
assembly in the ER are actually assembled, resulting in
very low rates of secretion of assembled antibody. As
the assembly rate constant increases, the specific anti-
body secretion rate increases, suggesting that under
these conditions it is the value of the assembly rate con-
stant that determines secretion rate. Once the plateau is
reached, increases in the assembly rate constant do not
accelerate the assembly rate and consequently do not
accelerate secretion; at this point some other step in the
secretory pathway plays the dominant role in determin-
ing the specific antibody secretion rate.
As the growth rate increases, the value of the assem-
bly rate constant at which a plateau is reached increases
and so does the value of the specific antibody secretion
rate corresponding to the plateau. At a growth rate equal
Specific growth rate (l/hr) to 0.004 h-' (10% of the maximum growth rate deter-
Figure 1. Effect of heavy- and light-chain gene dosage on steady- mined for the 9.2.27 hybridoma), the plateau is reached
state specific antibody secretion rate for specific growth rates 0 at an assembly rate constant value of about 5 x lo-",
to pmax (0.04 h-I): (a) K A = molecule-' cell h; (b) K A = whereas at the maximum specific growth rate (0.04 h-')
lo-'' ((molecules/cell) . h)-'.
this assembly rate constant value is close to sug-
stronger effect on secretion rate at low assembly rate gesting that at low growth rates the specific antibody
constant values. secretion rate could only marginally be improved by in-
creases in the assembly rate constant. As the specific
growth rate decreases, the heavy- and light-chain trans-
Effect of Heavy- and Light-Chain Assembly Rate on lation rates also decrease and start to play an increas-
Growth Rate Dependence on Specific MAb
Secretion Rate ingly important role in the determination of the specific
antibody secretion rate. The results in Figure 2 would
The effects of the assembly rate constant on the specific therefore suggest that in fast-growing cells assembly is
antibody secretion rate are summarized in Figure 2, a very good candidate for the controlling step in anti-
where the steady-state specific antibody secretion rate is body secretion, whereas in slowly growing cells anti-
plotted for various assembly rate constant values as a body secretion is most likely controlled at the level of
function of specific growth rate. For each specific growth translation.
rate, very low assembly rate constant values result in It is also interesting to note the existence of an in-
very low specific antibody secretion rates, with the flection point at an assembly rate constant value of 2 x
specific secretion rate increasing with increasing assem- 1O-lO-3 x lo-'' ((molecules/cell) . h)-'. For assembly
rate constant values greater than that corresponding to
n
1.5E7 the inflection point (5 x lo-'' to 5 x molecule-'
cell h), there is positive correlation between secretion
\
2 rate and growth rate. For assembly rate constant values
/ I
lower than that corresponding to the inflection point
(lo-" to lo-'' ((molecule/cell) * h)-', there is a negative
correlation between secretion rate and growth rate.
p= 0.004 hr -1 5.OE6
Near the inflection point, the specific secretion rate is
essentially independent of growth rate. Secretion rates
negatively correlated with specific growth rate are lower
than about 5 x lo6 molecules/cell * h whereas secretion
. -.-
1E-12 1E-11 lEL.10 1E-Q 1E-8 1E-7 1E-6 rates positively correlated with specific growth rate are
Assembly rate constant between 5 x lo6 and 1.5 x lo7 molecules/cell . h. The
theoretical maximum antibody secretion rate, based on
Figure 2. Steady-state specific antibody secretion rate as a func-
tion of the assembly rate constant ((rnolecules/cell) . h)-' for spe- the limitations of the cell's translational machinery, has
cific growth rates 0.004 h-'to pmar(0.04 h-'); f l 1 2 . E R = 42 min and been calculated to be on the order of 1 x lo7-3 x
t 1 / 2 , ~= 20 min. lo7molecules/cell h." All the secretion rates predicted

BlBlLA AND FLICKINGER: STEADY-STATE ANALYSIS OF A STRUCTURED MODEL 255

by the model are close to or lower than this theoreti-
cal maximum.
The ability of the model to predict the correlation be- 1E9
tween specific antibody secretion rate and specific
growth rate depending on the assembly rate constant
i
value may provide for the first time a useful mecha-
nistic explanation of the variety of secretion rate trends
in experimental data reported in the literature. Experi-
mental reports on the dependence of specific antibody
secretion rate on cell growth rate suggest a posi-
C L ! 1E7
tive,6,8,12,15,17,19,35,41,46,48,51,53,57,60 negative,8,16.2x,47,49
or lack I
0.00 0.01 3 0.02 E
0.03 0.04 6
of c ~ r r e l a t i o n between
~ ~ ' ~ ~ ~the~ secretion rate and the
cell growth rate. Specific g r o w t h rate (l/hr)
In all these cases, the reported specific antibody se- Figure 4. Effect of the assembly rate constant on the intracellular
cretion rates range from 5 x lo5 to 3 x 10' molecules/ steady-state heavy-chain levels for specific growth rates 0 to pmax
cell . h. This range is essentially the same as the range (0.04 h-'); t1/2,ER = 42 min and t l l 2 . G = 20 min. Curves from top to
of secretion rates predicted by our model. For the 9.2.27 bottom correspond to assembly rate constant values of lo-'', lo-",
lo-", lo-', lo-', lo-', and respectively.
hybridoma cell line, during cultivation of the cells in
a 1-L total cell recycle perfusion bioreactor," specific
antibody secretion rates were found to be essentially
constant around 3 X 106-5 x lo6 molecules/cell . h for -- l E 9
specific growth rates ranging between zero and the
maximum specific growth rate (0.04 h-') (Fig. 3). The
steady-state solution of this model (developed indepen- .-IS8
dently from the recycle reactor experimental data) does
indeed predict that a specific antibody secretion rate
of approximately 4 x 106-5 x lo6 molecules/cell . h
-.lE7
should be independent of growth rate (Fig. 2).

0.00 0.01 0.02 0.03 0.04

Effect of Assembly Rate Constant on lntracellular
Heavy- and Light-Chain Levels
Figures 4-6 summarize the effect of the assembly rate
constant on the steady-state intracellular (intraER)
levels of free (unassembled) light and heavy chains
and their ratio for specific growth rates between 0 and
0.04 h-'. The results of Figures 4 and 5 suggest that at a
Specific growth rate (l/hr)
Figure 5. Effect of the assembly rate constant on the intracellular
steady-state light-chain levels for specific growth rates 0 to pmar
(0.04 h-'); t i / z , E R = 42 min and t l / Z , G = 20 min. Curves from top to
bottom correspond to assembly rate constant values of lo-", lo-",
lo-", lo-*, lo-', and respectively.
given specific growth rate, the intracellular heavy- and

-2 1 0 A
6E6
light-chain pools decrease as the assembly rate constant
increases. Depending on the assembly rate constant
value, the intracellular heavy-chain levels may decrease,
A
increase, or remain essentially constant with increasing
Ill specific growth rate (Fig. 4). Depending on the assembly
rate constant value, the intracellular light-chain levels
may continuously decrease with increasing specific
growth rate, show an initial rapid increase followed by
decrease at higher growth rates or show an initial rapid
increase followed by slight increases at higher growth
rates (Fig. 5).
W
The model-predicted intracellular ratio of light to
0.00 0.01 0.02 0.03 0.04 0.05 heavy chains is always 21.0 (Fig. 6). The presence of
Specific growth rate (l/hr) intracellular excess light chains has been experimentally
verified in various immunoglobulin secreting cell lines.54
Figure 3. Specific antibody secretion rate vs. specific growth rate At specific growth rate equal to zero, the predicted ratio
during cultivation of 9.2.27 hybridoma cells in a 1-L total cell re-
of intracellular light to heavy chains is equal to 1.0 and
cycle perfusion b i o r e a c t ~ r Symbols
.~~ represent reactor runs at dif-
ferent L-glutamine provision rates (GlnPR) and a dilution rate of is independent of the value of the assembly rate con-
0.05 h-': (0) GlnPR = 0.39 mg/min; (0)GlnPR = 0.74 mg/min; stant. For all values of the assembly rate constant there
(A) GInPR = 0.99 mg/min; (0)GlnPR = 1.44 mg/min. is an initial increase in the light-to-heavy-chain ratio

256 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 39, NO. 3, FEBRUARY 5, 1992

 X
3
X
3
-.
0.00 0.01 0.k 0.k 0.b
Specific growth rate (l/hr)
Figure 6. Effect of the assembly rate constant on the intracellular
steady-state light-to-heavy-chain ratio for specific growth rates 0 to
pmax (0.04 h-'); t 1 1 2 , ~=~ 42 rnin and t l i z , ~= 20 min. (a) From bot-
tom to top curves correspond to assembly rate constant values of
lo-'', lo-", lo-", lo-', and lo-', respectively. (b) From bottom to
top curves correspond to assembly rate constant values of 5 x lo-',
5 x 10-7,and respectively.
with increasing growth rate. This is followed by a con-
stant ratio with increasing growth rate for the lower as-
sembly rate constant values or a slight decrease of the
ratio with increasing growth rate for the higher assem-
bly rate constant values. At a constant growth rate, the
light-to-heavy-chain ratio increases with increasing as-
sembly rate constant. The dependence of the intracel-
lular light-to-heavy-chain ratio on growth rate has been
experimentally verified.5*In immunoglobulin secreting
myeloma cells this ratio was found to be close to 1.0 in
the stationary growth phase and higher than 1.0 (around
2.5) in the exponential growth phase; these experimen-
tal results are in good agreement with our model predic-
tions (Fig. 6).
Effect of ER-to-Gobi and Golgi-to-Extracellular-
Medium Transport Half-times on Specific M A 6
Secretion Rate
Figure 7a shows the specific antibody secretion rate as
a function of specific growth rate for values of ER-to-
Golgi transport half-time, t i / z , E R , between 30 and 60 min
for an assembly rate constant equal to ((molecules/
cell) . h)-'. Figure 7b shows the specific antibody secre-
tion rate as a function of specific growth rate for values
of the Golgi-to-extracellular-medium transport half-
time, tl,2,G,between 5 and 20 min for the same assembly
rate constant value. The results in Figure 7 suggest that
changes in the ER-to-Golgi or Golgi-to-extracellular-
medium transport rate constants do not significantly
BlBlLA AND FLICKINGER: STEADY-STATE ANALYSIS OF A STRUCTURED MODEL
0.00 0.01 0.k 0.63 0.04
Specific secretion rate (l/hr)
Figure 7. (a) Effect of the ER to Golgi transport half-time, ~ ~ / z , E R ,
on the specific antibody secretion rate for specific growth rates 0
to pmax (0.04 h-I); tlR,G = 20 min and Ka = lo-'. The value of t1/2,ER
was varied between 0.5 and 1 h with 0.1 h increments. The ob-
tained profiles overlap. (b) Effect of the Golgi-to-medium trans-
port half-time, tliz,G,on the specific antibody secretion rate for
specific growth rates 0 to p,,,,, (0.04 h-I); ~ ~ / z , E =
R 42 rnin and
Ka = The value of t , , 2 , ~was varied between 5 and 20 rnin
with 5-min increments. The obtained profiles overlap.
affect the specific antibody secretion rate, in agreement
with the factorial design analysis predictions.
Figure 8 shows the size of the ER and Golgi MAb
pools as a function of specific growth rate for values of
t112,ER between 30 and 60 min, tliZ,G= 20 min and KA =
lop9((molecules/cell) . h)-'. The ER antibody pool
increases with increasing specific growth rate for all
values of tlIZ,ER (Fig. 8a). As expected, at a constant
specific growth rate the size of the ER antibody pool
increases as the transport rate constant out of the ER
decreases, i.e., at t I / z , E R increases. The size of the Golgi
antibody pool remains essentially unchanged with vary-
ing t1/2,ER (Fig. 8b). This is due to the fact that although
the ER-to-Golgi transport rate constant varies, the size
of the ER antibody available for transport is adjusted so
that their product is essentially constant. According to
Equation (12), this would result in a constant Golgi anti-
body pool.
Figure 9 shows the size of the Golgi antibody pool as
a function of specific growth rate for KA = ((mole-
cules/cell) . h)-', t l / 2 , E R = 42 min, and tl12,G between 5
and 20 min. As expected, at a particular specific growth
rate, the size of the Golgi antibody pool increases as the
transport rate constant out of the Golgi decreases, i.e.,
as t1/2,G increases. The results of Figures 7b and 9 would
suggest that when tllZ,G changes under steady-state con-
ditions, the size of the Golgi antibody pool does change,
but it is regulated so that the Golgi-to-medium transport
257

0.00 0.01 0.02 0.09 0.04
Specific growth rate (l/hr)
Figure 8. Effect of the ER-to-Golgi transport half-time, t 1 / 2 , ~ ~ ,
on the ER and Golgi antibody pool sizes for specific growth rates 0
to pmal(0.04 h-') and an assembly rate constant value equal to
lo-' (molecule/cell . h); t l , 2 , ~= 20 min.
0.00 0.01 0.02 0.03 0.04
Specific growth r a t e ( l / h r )
Figure 9. Effect of the Golgi-to-medium transport half-time, f l , 2 , c ,
on the Golgi antibody pool size for specific growth rates 0 to pmar
(0.04 h-I); t l n , =
~~42 min and KA = lo-'.
rate (secretion rate) remains essentially constant. For
assembly rate constant values that result in a negative
correlation between the specific antibody secretion rate
and the specific growth rate, the size of the ER and
Golgi antibody pools decreases with increasing growth
rate, but again changes in I I I Z , E R or ~ ~ I do
z , not
G signifi-
cantly affect the specific antibody secretion rate.
Figure 10 shows the ratio of the ER MAb pool to the
Golgi MAb pool as a function of specific growth rate for
varying tl12,ER and tliZ,G,respectively. The simulation
profiles shown in Figure 10 are independent of the value
of the assembly rate constant. For a given I1/2,ER and
tliZ,G value, the predicted value of the ER-to-Golgi anti-
body ratio is constant with specific growth rate. The
results in Figure 10 suggest that despite the fact that the
sizes of the ER and Golgi antibody pools are strong
258 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 39, NO. 3, FEBRUARY 5, 1992
04 I
0.00 0.01 0.02 0.09 0.04
Specific growth rate (l/hr)
Figure 10. (a) Effect of the ER-to-Golgi transport half-time,
f l / z , E R , on the ratio of E R to Golgi antibody pools for specific growth
rates 0 to pmax (0.04 h-') and f l , 2 , ~= 20 min. The profiles shown are
independent of the assembly rate constant value. (b) Effect of the
Golgi-to-medium transport half-time, t l , ~ , on ~ , the ratio of ER to
Golgi antibody pools for specific growth rates 0 to F~~~ (0.04 h-')
and t l , z , ~=~42 min. The profiles shown are independent of the
assembly rate constant value.
functions of the growth rate, the value of the assembly
rate constant and the values of the ER-to-Golgi and
Golgi-to-medium transport half-times (Fig. 8 and 9),
their ratio only depends on the values of the interorgan-
elle transport half-times. In all cases, the predicted size
of the ER antibody pool is larger than the size of the
Golgi antibody pool, in agreement with experimental
observation^.'^ Doubling f l/Z ,E R from 30 to 60 min results
in doubling of the ER-to-Golgi MAb pool ratio from 1.5
to 3 (Fig. 10a),whereas a fourfold decrease of ~ I / Z , Gfrom
20 to 5 min results in a fourfold increase of the ratio
from 2 to about 8 (Fig. lob).
Implications for Optimization of Large-Scale
Production Systems
Having assessed the importance of secretory pathway
parameters in determining the specific antibody secre-
tion rate, how could these parameters be manipulated to
maximize antibody secretion?
Manipulation of Cell Growth Rate
The ability of our model to predict the dependence of
antibody secretion rate on cell growth rate may be
important for optimizing bioreactor performance in
large-scale antibody production systems. A positive cor-
relation between secretion rate and growth rate would
favor the maintenance of the cells at close to maximum
growth rate in a batch, fed-batch, or continuous biore-
actor operated at a high dilution rate. A negative corre-
lation between secretion rate and growth rate would
favor the maintenance of the cells at close to zero
growth rates in a system such as a perfusion or immobi-
lized cell bioreactor (hollow fiber, ceramic matrix).
However, the ideal growth medium and bioreactor
operation mode combination should not only maximize
the specific antibody secretion rate but also maximize
cell density in order to achieve maximum volumetric
antibody productivity [Equation (xi)]. The volumetric
antibody productivity can be significantly increased by
increasing the viable cell concentration via control of
growth rate. In a continuous culture the growth rate,
cell viability, and final cell concentration can be con-
trolled by controlling the dilution rate.” In a perfusion
culture with purge, cell viability and growth rate can be
controlled by controlling the purge rate26whereas the
final steady-state cell concentration can be controlled by
controlling the dilution rate.26The achievable final cell
concentration in perfusion and immobilized systems can
be as much as 10-fold higher than the maximum cell
concentration at the end of exponential phase in a batch
culture,’ leading to much higher antibody volumetric
productivitie~.’~~~However, manipulation of the bioreac-
tor operation mode to maximize cell viability and final
cell concentration should also take into account the ef-
fect of growth rate on specific antibody secretion rate,
which according to the model predictions presented
here can be negligible, positive, or negative.
Genetic Manipulation of M A 6 Secretory Pathway
Introduction of multiple copies of the heavy- and light-
chain genes in the cell could substantially increase
specific antibody secretion rates according to our model
predictions. However, the stability of the introduced
gene copies, their transcription and translation efficien-
cies, and their effect on cell growth rate may alter the
observed effect on antibody secretion rate from that
predicted by the steady-state solution of the model pre-
sented here.
Assembly of heavy and light chains in the ER was
found to be a very likely candidate for a controlling
step in the antibody secretory pathway under fast
growth conditions. Assembly is catalyzed by protein
disulfide isomerase (PDI), a multifunctional ER enzyme
which catalyzes formation of native disulfide bonds
in disulfide-bonded proteins.” Changes in the concen-
tration of intracellular PDI levels should result in
proportional changes in the assembly rate constant. In-
tracellular levels of PDI in immunoglobulin secreting
cells have been reported to correlate well with immuno-
globuiin secretion suggesting that the assembly
rate might determine the overall antibody secretion
rate, a conclusion similar to our model predictions. In-
tracellular PDI levels have also been reported to change
with the growth stage of cells, increasing up 5- to 10-fold
BlBlLA AND FLICKINGER: STEADY-STATE ANALYSIS OF A STRUCTURED MODEL
with entrance of cells from the exponential into the
stationary phase of as well as increase during
induction of stress proteins (e.g., heat shock, glucose
starvation, and glycosylation inhibitor additi~n).~’ Ge-
netic manipulations that would increase the levels of
PDI expression may be a useful approach to increasing
assembly rates and possibly antibody secretion rates in
fast-growing cells (see the companion article).
According to the steady-state simulation results pre-
sented here, antibody secretion by slowly growing or
stationary phase hybridoma cells (as is the case in
immobilized-type systems) is most likely controlled at
the level of heavy- and light-chain translation. Since
translation is a highly energy-intensive process, strate-
gies to increase specific antibody secretion rates in that
case should most likely focus on the design of appro-
priate nutrient feeding schemes. A more detailed dis-
cussion on the identity of rate-limiting steps in the
monoclonal antibody secretory pathway as well as a
demonstration of how our model could be used to de-
sign genetic manipulation or nutrient feeding strategies
to overcome pathway rate limitations are given in the
companion article conjunction with the study of the dy-
namic behavior of this pathway.
CONCLUSIONS
We have used our previously developed structured ki-
netic model of monoclonal antibody synthesis and se-
cretion in steady-state simulations to assess the role of
different model parameters in determining the final
specific antibody secretion rate. The assembly rate con-
stant for heavy- and light-chain assembly in the ER, the
specific growth rate (coupled with the heavy- and light-
chain translation rates), and the gene dosage of the
heavy and light chains are the parameters with the
greatest effect on the secretion rate. Strong interaction
was observed between the effects of these three parame-
ters as indicated by factorial design analysis. Our model
has the ability to predict a positive, negative, or no cor-
relation between the specific antibody secretion rate
and the specific growth rate. Optimization of the volu-
metric antibody productivity in large-scale production
systems should consider maximizing the specific anti-
body secretion rate via growth rate control, appropriate
nutrient feeding schemes, and/or genetic manipulations
of the MAb secretory pathway as well as maximizing
the viable cell denstiy by choice of an appropriate biore-
actor operation mode and growth medium.
APPENDIX: STRUCTURED KINETIC MODEL FOR
MAb SYNTHESIS AND
lntracellular Heavy- and Light-Chain
mRNA Balances
259
where mH,m,. = intacellular heavy- and light-chain mRNA KG = rate constant for Golgi-to-extracellular-medium
concent rat ions (mRNAs/cell) antibody transport (h-I)
NH, NL = heavy- and light-chain gene copy number tliZ,6= half-time for Golgi-to-extracellular-medium
(genes/cell) antibody transport (h)
S H ,SL = heavy- and light-chain gene specific transcription t112.c= In K KG
rates (mRNAs/gene/h)
KG . [ H z L ~=] q~M A b = specific antibody secretion rate
K = heavy- and light-chain mRNA decay rates (h-I) (molecules/cell/ h)
p = specific growth rate (h-I)
MAb Balance in Extracellular Medium
IntraER Heavy- and Light-Chain Balances d[MAb]/dt = qMAh ' xv
d[H]/dt = TH . mH - p [HI - RH
I
(iii) where [MAb] = antibody concentration in the extracellular
medium (molecules/mL)
d[L]/dt = TL . m L - p . [L]- RL (iv)
where H , L = free heavy- and light-chain concentrations in the Xv = viable cell concentration (viable cells/mL)
E R (chains/cell) The authors acknowledge the partial support of this work by
TH,TL = heavy- and light-chain specific translation rates Lilly Research Laboratories (Indianapolis, Indiana), R&D
Systems, Inc. (Minneapolis, Minnesota), Pfizer Central
(chains/mRNA h)
Research (Groton, Connecticut), and the Institute for Ad-
R H ,R L = rates of heavy- and light-chain consumption in vanced Studies in Biological Process Technology. The au-
assembly (chains/cell/h) thors are grateful to Drs. George Boder and Jon Schmidtke
(Lilly) for use of the 9.2.27 hybridoma cell line.

Heavy- and Light-Chain Assembly in ER: IntraER

Balances of Assembly Intermediates References
H+H=Hz 1. Batt, B. C., Kompala, D. S. 1989. Biotechnol. Bioeng. 34: 515.
=H2 + L H2L
2. Baumal, R., Scharff, M. D. 1973. Transplant. Rev. 14: 163.
3. Bibila, T. A., Flickinger, M. C . 1991. Biotechnol. Bioeng.
H2L + L H2Lz 37 210.
4. Bibila, T. A., Flickinger, M. C. 1991. Biotechnol. Bioeng. 38:
1 767.
R H = 2 ' - . KA . [HIz (v)
3 5. Birch, J. R., Lambert, K., Thompson, P.W., Kenney, A.C.,
Wood, L. A. 1987. p. 1 In: B. K. Lydersen (ed.), Large scale cell
RL = 2 KA ' [Hz] ' [ L ] + KA ' [HzL] ' [ L ] (vi) culture technology, Hanser Publishers, Munich.
1 6. Boraston, R., Thompson, P.W., Garland, S., Birch, J. R. 1984.
d[Hz]/dt= - . KA . [HI2 - 2 . KA . [ H z ]. [ L ] - p . [ H z ] (vii) Devel. Biol. Stand. 55: 103.
3
7. Box, G . E . P., Hunter, W. G., Hunter, J. S. 1978. Statistics for
d[HzL]/dt = 2 . KA . [Hz] . [ L ] - KA ' [HzL] ' [LI - p ' [ H 4 1 experimenters, Wiley.
(viii) 8. Castillo, F. J., Mullen, L., Thrift, J., Grant, B. 1990. Perfusion
culture of hybridomas for monoclonal antibody production,
where [H2],[H2L] = assembly intermediate concentrations in the paper presented at the American Chemical Society National
E R (molecules/cell) Meeting, Washington, DC, August.
KA = assembly rate constant ((molecule/cell) . h)-' 9. Choi, Y. S., Knopf, P.M., Lennox, E.S. 1971. Biochemistry
10:668.
10. Constantinides, A. 1987. Applied numerical methods with per-
IntraER M A b Balance sonal computers, McGraw-Hill, New York.
11. Dalili, M., Sayles, G. D., Ollis, D. E. 1990. Biotechnol. Bioeng.
d [ H z L z I ~ ~ /= d t KA . [HzLl . [ L ] - KER. [ H Z L Z I E-R p . [ H Z L Z ~ R 36: 74.
(ix) 12. Dean, R. C., Karkare, S. B., Phillips, P. G., Ray, N. G., Run-
standler, P.W. 1987. p. 145 In: B. K. Lydersen (ed.), Large scale
Where[HzLz]~~ = antibody concentration in the E R
cell culture technology, Hanser Publishers, Munich.
(molecules/cell) 13. Dorai, H., Moore, G . P. 1987. J. Immunol. 139: 4232.
K E R= rate constant for ER-to-Golgi antibody trans- 14. Drechou, A,, Rouzeau, J. D., Feger, J., Durand, G. 1989. Bio-
port (h-') chem. J. 263: 961.
15. Emery, N. 1986. Growth of hybridomas and secretion of mono-
t I / Z , E R = half-time for ER-to-Golgi antibody transport (h) clonal antibodies in vitro, paper presented at Society of
Chemical Industry Symposium on Large Scale Production of
tll?,ER = ln 2/KER
Monoclonal Antibodies, London, December.
16. Fazekas de St. Groth, S. 1983. J. Immunol. Meth. 57 121.
IntraGolgi M A b Balance 17. Feder, J. 1988. Adv. Biotechnol. Proc. 7: 125.
18. Flickinger, M.C., Goebel, N. K., Bohn, M.A. 1990. Bioproc.
d[H2Lz]c/dt = K E R. [ H Z L Z I E-R KG ' [ H ~ L z I - c p . [HzLzIG (X) Eng. 5: 155.
19. Frame, K. K., Hu, W. S. 1991. Biotechnol. Bioeng. 37:55.
where [ f f z L Z ]=~antibody concentration in the Golgi 20. Freedman, R. B., Brockway, B. E., Lambert, N. 1984. Biochem.
(molecules/cell) SOC.Trans. 12: 929.

260 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 39, NO. 3, FEBRUARY 5, 1992

21. Fries, E., Gustafson, L., Peterson, P. A. 1984. EMBO J. 3: 147.
22. Glacken, M. W., Fleischaker, R. J., Sinskey, A. J. 1986. Biotech-
nol. Bioeng. 28: 1376.
23. Heijne, G.V., Blomberg, C., Liljestrom., H. 1987. J. Theor. Biol.
125: 1 .
24. Hendricks, M. B., Banker, M. J., McLaughlin, M. 1988. Gene
64: 43.
25. Hendricks, M. B., Luchette, C.A., Banher, M. J. 1989. Biotech-
nology 7: 1271.
26. Karkare, S. B., Williams, L. R., Hettwer, D. J. 1990. Theoreti-
cal and practical aspects of a cell-recycle perfusion bioreactor
for c’ulturing recombinant mammalian cells, paper presented in
the American Chemical Society National Meeting, Washington,
DC, August.
27. Lodish, H., Kong, N., Snider, M., Strous, G. J. 1983. Nature
304: 80.
28. Low, K. S., Harbour, C., Barford, J. P. 1987. Biotechnol. Techn.
1: 239.
29. Luan, Y.T., Mutharasan, R., Magee, W.E. 1987. Biotechnol.
Lett. 9: 535.
30. Macmillan, J. D., Velez, D., Miller, L., Reuveny, S. 1987. p. 21
In: B. K. Lydersen (ed.), Large scale cell culture technology,
Hanser Publishers, New York.
31. Mader, A. 1988. J. Theor. Biol. 134: 135.
32. Meilhoc, E., Wittrup, K. D., Bailey, J. E. 1990. Bioproc. Eng.
5: 263.
33. Miller, W. M., Wilke, C . R., Blanch, H.W. 1988. Bioproc. Eng.
3: 103.
34. Miller, W. M., Wilke, C. R., Blanch, H.W. 1988. Bioproc. Eng.
3: 113.
35. Miller, W. M., Blanch, H.W., Wilke, C . R . 1988. Biotechnol.
Bioeng. 32: 947.
36. Morpurgo, G. 1985. J. Theor. Biol. 116: 395.
37. Morrison, S. L . , Oi, V.T. 1989. Adv. Immunol. 44: 65.
38. Myllyla, R., Koivu, J., Pihlajaniemi, T., Kivirikko, K. 1983.
Eur. J. Biochem. 134: 7.
39. Oster, G. F., Cheng, L.Y., Moore, H. P., Perelson, A . S. 1989.
J. Theor. Biol. 141: 463.
40. Ozturk, S. S., Palsson, B. 0. 1990. Hybridoma 9: 167.
41. Ozturk, S. S., Palsson, B. 0. 1991. Biotechnol. Bioeng. 3 7 35.
42. Perry, R. P. 1980. Proc. Nat. Acad. Sci. 7 7 1937.
43. Perry, R. P., Kelley, D. E., Cloeclough, C., Kearney, J. F. 1981.
Proc. Nat. Acad. Sci. 78: 247.
BlBlLA AND FLICKINGER: STEADY-STATE ANALYSIS OF A STRUCTURED MODEL
44. Peters, B. P., Krzesicki, R. J., Hartle, R.F ., Perini, R. J.,
Ruddon, R.W. 1984. J. Biol. Chem. 259: 15123.
45. Peters, B. P., Hartle, R. F., Krzesicki, R. J . , Kroll, T . G . ,
Perini, F., Balun, J.E., Goldstein, I. J., Ruddon, R.W. 1985.
J. Biol. Chem. 260: 14732.
46. Ramirez, O.T., Mutharasan, R. 1990. Biotechnol. Bioeng.
36: 839.
47. Ray, N.G., Tung, A.S., Ozturk, S . S . , Runstandler, P.W.,
Vorunakis, J. N. 1990. Cultivation of mammalian cells in
fluidized-bed bioreactors: effect of perfusion rate, paper
presented at the AIChE Annual Meeting, Chicago, Illinois,
November.
48. Reuveny, S . , Velez, D., Miller, L., MacMillan, J.D. 1986.
J. Immunol. Meth. 86: 61.
49. Robinson, D. K., Suemeghy, Z., Memmert, K . 1990. Effect of
specific cellular growth rates in continuous open and cell re-
tention animal cell bioreactors, paper presented at the AIChE
Annual Meeting, Chicago, IL.
50. Roth, R. A ,, Koshland, M. E. 1981. Biochemistry 20: 6594.
51. Runstandler, P.W., Cernak, S. R. 1988. p. 305 In: R. E. Spier
and J. B. Griffiths (eds.), Animal cell biotechnology, Vol. 3,
Academic, New York.
52. Savinell, J. M., Lee, G . M., Palsson, B. 0. 1989. Bioproc. Eng.
4: 231.
53. Savinell, J., Palsson, B.O. 1990. Estimation of metabolic re-
source distributions using linear optimization, paper presented
at the AIChE Annual Meeting, Chicago, IL.
54. Schibler, U., Marcu, K. B., Perry, R. P. 1978. Cell 15: 1495.
55. Schneider, Y. J. 1989. J. Immunol. Meth. 116: 65.
56. Schneider, Y. J., Lavoix, A. 1990. J. Immunol. Meth. 129: 251.
57. Seaver, S. S. 1987. p. 49 In: S. S. Seaver (ed.), Commercial pro-
duction of monoclonal antibodies, Marcel Dekker, New York.
58. Stevens, R. H. 1974. Eur. J. Biochem. 42: 553.
59. Strous, G . J., Lodish, H . F. 1980. Cell 22: 709.
60. Suzuki, E., Ollis, D. F. 1990. Biotechnol. Prog. 6: 231.
61. Truskey, G . A,, Nicolakis, D. P., DiMasi, D., Haberman, A,,
Swartz, R.W. 1990. Biotechnol. Bioeng. 36: 797.
62. Velez, D., Reuveny, S . , Miller, L., Macmillan, J. D. 1986.
J. Immunol. Meth. 86: 45.
63. Williams, D. B., Swiedler, J., Hart, G.W. 1985. J. Celk Biol.
101: 725.
64. Yeo, K.T., Parent, J. B., Yeo, T. K., Olden, K. 1985. J. Biol.
Chem. 260: 7896.
261