Chapter 17

CHAPTER 17
Bioprocess Engineering
and Technology
17.1 Introduction to Bioprocess Engineering and Technology

Bioprocess technology is the application of the principles of process technology (which are essentially derived from
chemical engineering) to biological systems. As such it introduces a much needed quantitative approach to our study of
biological systems. This quantitative approach is required not only because it provides a better and deeper understand-
ing of the behavior of biological systems, but also because the design and scale-up of bioprocesses to an industrial scale
necessitates quantitative analysis. Furthermore, optimization and regulatory requirements, especially of modern biotech
products, like recombinant therapeutics, have given an additional impetus to this need for quantification. The develop-
ment of theoretical tools coupled with advances in process instrumentation and control today allows us to identify and
monitor accurately the progress of a large number of key variables in any process. Thus, for example, in a fermentation
involving biomass and product formation the parameters usually measured were limited to pH, dissolved oxygen, agita-
tion rate, aeration rate level and pressure. Key variables like cell density, substrate and product concentration were all
measured offline. Modern fermentors today come equipped with instrumentation to measure these parameters online far
more accurately and continuously than was thought possible. Intracellular parameters like ATP, NADP/NADPH ratios,
intracellular pH are also monitored together with macroparameters like oxygen uptake rate (OUR), carbon dioxide evo-
lution rate (CER) and byproducts like acetic acid or ethanol. These measurements combined with sophisticated control
techniques guarantee not only an optimum fermentor running, but also allow a comprehensive analysis of the whole
process. This analysis is important not only for our understanding but also for complying with the extremely strict GMP
regulations that are in force today with respect to recombinant products or products intended for human use. Similar
rigid procedures have been adopted both upstream and downstream of the fermentation process thereby reducing the
variation that was once upon a time thought to be an intrinsic property of any biological system. This in turn ensures
compliance to the regulatory authorities and strict quality control. The following sections introduce the basic theo-
retical framework required to analyze these processes. The interested student may refer to the large body of literature
available in this subject. Some references are appended in the end as well.
17.2 The Component Parts of a Fermentation Process

A typical fermentation process can be divided into three parts: The upstream part, the fermentation itself and the
downstream part. The upstream part consists of: (a) the isolation, preservation and propagation of the microorganism
of interest. This includes strain improvement using a combination of classical and recombinant techniques in order to
enhance the productivity of the process. (b) Inoculum preparation from shake flask to large seed fermentors. (c) Media
design and preparation and (d) sterilization. The fermentation part consists of: (a) analysis of the stoichiometry and
energetics of the process. (b) Development of the ‘microkinetics’ of cell growth and product formation. (c) Integrating
with the macropicture by incorporating heat and mass transfer effects. (d) Developing a useful model of the fermenta-
tion process as a whole and hence process optimization. (e) Process instrumentation and control to keep the fermentor
and hence the cells in the optimum condition required for growth and product formation and (f) fermentor design. The
downstream processing part involves: (a) product recovery and purification, (b) packaging for transport and enhanced
shelf life of the product; (c) quality control requirements; (d) effluent treatment.
An integrated approach to the design of the complete process is essential to optimize its functional efficiency. Decisions
taken in one stage can have an impact on all other stages. Thus, for example, the nature of the product (whether it is
extracellular or intracellular) is determined by the microorganism used for expression and this in turn can completely
change the product recovery and purification equipment in a plant. Even a simple parameter like the substrate to be
used or temperature of growth can affect the heat and mass transfer calculations, thereby necessitating major changes in
fermentor design, which in turn determine overall process economics.
774 • Chapter 17/Bioprocess Engineering and Technology
17.3 Material Balance

Material and energy balances provide an elegant and powerful tool for the analysis of any process. Like an accountant
who provides us with picture of the cash flow coming in and going out of a company, these balances help us in quanti-
fying the various flows in and out of a system. The flows which cannot be measured directly can be computed using the
basic laws of conservation of matter and energy, thus providing us with a complete picture of the process.
The basic equation governing the change of any extrinsic variable in a system is given by
Accumulation = Input − Output ± Reaction (17.1)
This equation is best illustrated by a few examples given below:
Example 1 Consider a tangential flow filtration device which concentrates a cell culture suspension obtained from a
fermentor. The input concentration of cells is 5 g l−1 which is pumped in at 100 l h−1. The desired output concentration
should be 50 g 1−1. Calculate the rate of removal of the permeate supernatant?
Refer to Figure 17-1 which is a schematic diagram of the process. The system runs continuously, so there is no accu-
mulation. Such systems where the values of the variables do not change with time are referred to as ‘steady-state’ (SS)
systems. Also since there is no reaction, Eq. (17.1) becomes
Input = Output.
Two concepts are important in organizing any material balance problem (though they are trivial in this simple case).
One is the concept of a ‘basis’ for calculation and the second is the concept of a ‘tie’ substance which passes through
unchanged in the reaction.
In this case the ‘basis’ is 1 h. of operation when 100 l of cell culture enters the system. The ‘tie’ substance is the ‘cells’
which pass through unchanged through the assembly.
Calculations: We have 100 l of input (1 h operation).
Input cells = 5 g l−1 × 100 l = 500 g.
Input = Output.
Output cells = 500g.
If x liters is the output of concentrated cells, then
x × (50 g l−1) = 500 g
x = 10 l
From an overall balance,
Supernatant volume = 100 − 10 = 90 l.
Thus supernatant has to be removed at a rate of 90 l h−1 (Answer).
Suppose that it is now found out that the cells are concentrated only fivefold in a single-pass through the filtration
assembly. Therefore, in order to achieve the desired concentration of cells one should determine the fraction of the cells
that need to be recycled back.
Now since the concentration factor is fivefold we can calculate that the input cell concentration to be 50/5 = 10 g l−1.
Thus
100 × 5 + R × 50 = (100 + R) × 10,
which gives R = 12.5 l h .
−1
Thus from a total of (10 + 12.5) = 22.5 l, at the output 12.5 l needs to be recycled. (Note that the overall material
balance is not affected by the recycle.)
100 l h1
50 g l1
5 g l1
Supernatant Figure 17-1 Tangential flow filtration device.

775
17.3 Material Balance•
50 g l−1 R
5 g l−1 1 2 50 g l
−1
10 g l−1
Figure 17-2 Tangential flow with recycle. Supernatant
Example 2 Let us now consider a situation where a chemical reac- A AB

tion is taking place. Figure 17-3 shows a stirred tank reactor with a
continuous feed of component A. The outflow which is at the same
rate (to prevent accumulation) consists of a mixture of A and B. A
gets converted to B following first-order kinetics A→B,
rA = kCA,
where rA is the reaction rate (dCA/dt) and CA is the concentration of
A. Such a reactor is referred to as a continuous stirred tank reactor
(CSTR). If F is the feed rate and V is the volume of the reactor, the
dilution rate (D) is defined as D=F/V, which is the feed rate per unit Figure 17-3 Schematic of continuous stirred tank
volume and has the units of h−1. reactor (CSTR).
The material balance is:
Accumulation = Input − Output ± Reaction,
V (dCA/dt) = FCA0 − FCA − kCAV,
where CA0 is the input concentration of A and CA is the concentration of A in the reactor which is the same as the outlet
concentration since the reactor is well mixed.
Assuming that sufficient time has elapsed and the concentration of A and B does not change with time we can put
the accumulation term to zero. This situation is referred to as SS condition. Thus
D(CA 0 − CA 0 ) = kCA or CA / CA 0 = D / (k + D).
Knowing CA0, k and D, the value of CA can be determined. We will return to this system when studying the kinetics
of cell growth.
Example 3 Let us consider an unsteady state situation where the accumulation term is nonzero. We have the same
CSTR where the input concentration of A is increased at time t = 0 to CAf from CA0. The material balance is
dCA/dt = D(CAf − CA) − kCA
This is a differential equation which needs to be solved for the given boundary conditions. Here we solve for the
simple case where there is no reaction term. The above case is left as an exercise for the interested student. When there
is no reaction we have
dCA / dt = D(CAf − CA ) or dCA / (CAf − CA ) = D dt.
By applying appropriate boundary conditions (CA = CA0 at t = 0) and solving we get

ln(CAf − CA0)/(CAf − CA) = Dt
The result is plotted graphically in Figure 17-4. The student may try and solve this problem when two or more reac-
tors are connected in series.
17.3.1 Material Balance in Biological Systems

Consider the growth of microbial cells in a fermentor. During growth, substrate is consumed and biomass and product
generated. This substrate is usually a complex mixture of carbon, hydrogen, oxygen and nitrogen sources and typically
other elements like phosphorus, magnesium, potassium, sodium and chlorides are also required.
However for material balance purposes we need to consider the main elements which participate in a significant way
in the process. We will develop here a basic technique to solve the set of material balance equations, which if required
can be extended to more elements as well.
Consider a system of microbial cells at SS. This way of formulating the system allows us to set the accumulation term
to zero and reduce the material balance equations to simple algebraic equations. In this system substrate is consumed and
biomass produced and these become the input to and output from the system, respectively.
Figure 17-5 shows the various flows going in and out of the system which is at SS. As all flows are shown as coming
in, the output flows would turn out to have negative values.
There are a total of seven flows consisting of two substrate flows (representing the carbon and nitrogen sources),
biomass, product, oxygen, carbon dioxide and water. The chemical formula for the flows will be written on a C-mole
basis which helps in setting up the material balances. This means that the formula will have one carbon atom only and
the number of atoms of the other elements are adjusted accordingly. Thus glucose, which is a typical carbon source, has a
chemical formula of CH2O (and not C6H12O6) and biomass will be represented as CH1.8O0.5N0.2 (and not C10 H18 O5 N2)
which is a fairly good approximation, representative of the elemental composition of a range of microbial cells. For exam-
ple the difference between yeast cells and Escherichia coli in terms of the relative ratio of the different elements present
in these cells is less than 5%. However, more complex chemical formulas may be used, if required, to represent biomass.
In order to simplify the material balance further, elemental balances on carbon, hydrogen, oxygen and nitrogen will
be done. Since elements do not participate in the reactions taking place, we can set the reaction terms equal to zero.
Thus the material balance reduces to
Input = Output
The general formula for a flow Fi is designated as CHaiObiNci. Thus if F1 (i = 1) is the flow of glucose (having the formula
CH2O) then a1 = 2, b1 = 1 and c1 = 0 and if F2 is the flow of biomass having the formula CH1.8O0.5N0.2 then a2 = 1.8, b2 = 0.5
and c2 = 0.2. Combining the elemental balances with the overall material balance which is ∑Fi = 0 we get the following
four balances:
C balance: ∑Fi = 0, (17.2)
H balance: ∑aiFi = 0, (17.3)
O balance: ∑biFi = 0, (17.4)
N balance: ciFi = 0. (17.5)
Note that we have seven flows and hence seven unknowns and only four independent equations. Thus three flows
need to be determined experimentally for defining the complete system. In order to simplify the solving of the four
simultaneous equations we follow the standard practice of multiplying each equation by a factor and then adding the
equations to get a simple equation. This multiplication factor can then be appropriately chosen to reduce the number of
unknown flows in the final equation.
Thus multiplying the four balance equations [(17.2)–(17.5)] by λC′ , λH′ , λO and λN, respectively, and adding we get
ΣlC Fi + ΣlH a i Fi + ΣlO bi Fi + Σl N ci Fi = 0.
F7 (H2O) F6 (CO2) F5 (Oxygen)

CAf
Input profile
Output profile
CA
F1 SYSTEM F4
(Substrate (Substrate
CA0 C-source) N-source)
t0 Time F2 F3
(Biomass) (Product)
Figure 17-4 Unsteady state profile of concen- Figure 17-5 System of microbial cells at steady state with various
tration of A in a CSTR. flows.
777
Rearranging we get
ΣFi (lC + a i lH + bi lO + ci l N ) = 0.
The expression (λc + aiλH + biλO + ciλN) can be represented by γi. Thus the final material balance becomes
ΣFiγi = 0.
Now the values of λC′ , λH, λO and λN can be carefully chosen so that certain γi values become zero. Thus choosing
λH = 1 and λO = −2, we get
γH2O = 2 λH + λO = (2 − 2) = 0.
(Since water does not have any carbon so there is no λc term and also c7 = 0 because no nitrogen is present.) Similarly
choosing λC = 4 would set γ CO2 = (λC + 2λO) equal to zero.
If the nitrogen source is ammonia gas then choosing λN = −3 would set γN = 0 (γN = λN + 3λH). However, if the nitrogen
source is complex, for example glutamate (CH1.8O0.8N0.2), then to get λN equal to zero we can use the equation
g N = lC + 1.8lH + 0.8lO + 0.2l N = 0,
which requires choosing λN = −21.

Thus the final material balance ∑Fiγi = 0 which had seven terms (representing seven flows) now has only four terms.
Expanding we get:
F1g 1 + F2g 2 + F3g 3 + F5g 5 = 0,
or
Fsg s + Fx g x + Fpg p + Fog o = 0,
where subscripts s, x, p and o represent carbon source, biomass, product and oxygen, respectively. [Note that the γ values
of F4´ F6 and F7 (nitrogen source, carbon dioxide and water) are zero and hence they do not appear in the final material
balance equation.]
Of these four unknowns only three need to be determined experimentally in order to determine the complete stoi-
chiometry of the reaction. Let us illustrate this by simple examples.
Example 4 Consider the case when no product is being formed so that Fp = 0. Thus the material balance equation
becomes Fsγs + Fxγx + Foγo = 0.
We now define yield coefficient Yx/s which represents the amount of biomass formed per unit amount of substrate
consumed. Yield coefficient is thus a ratio having no units and can be represented as
Yx/s = C-mole of biomass formed/C-mole of substrate consumed.
Similarly other yield coefficients can be defined, for example Yx/o which is the ratio of biomass formed to oxygen
consumed. (This procedure is similar to choosing the ‘basis’ of the material balance as 1 C-mole of biomass formed and
calculating all the other flows with respect to this basis.)
The material balance equation thus becomes (on dividing by Fx throughout)
Fsg s / Fx + g x + F0g 0 / Fx = 0,
or
g x / Yx / s + g x + g o / Yx / o = 0.
Suppose cells are grown on glucose and ammonia as carbon and nitrogen sources, respectively, and the yield coef-
ficient is measured experimentally as 0.5 g/g (i.e. grams of biomass formed per gram of glucose consumed). Then the
oxygen yield coefficient Yx/o can be calculated from the above equation. We have
g s = 4 + 2 − 2 = 4(formula CH 2 O);
g x = 4 + 1.8 + 0.5 × 2 − 0.2 × 3 = 4.2(formula CH1.8 O 0.5 N0.2 );
g O = −4.
The molecular weights of biomass and glucose on a C-mole basis are (12 + 1.8 + 0.5 × 16 + 0.2 × 14) = 24.6 and
(12 + 2 + 16) = 30, respectively.
Thus the yield coefficient can be calculated on a C-mole/C-moles basis as Yx/s = (−0.5/24.6)/(1/30)
Yx/s (C-mole/C-mole) = (−)0.6 C-mole/C-mole,
where the negative sign indicates that biomass is formed while substrate is consumed. Substituting values in the above
equation we get
Yx/o = −1.7 C-mole/mole of O2.
A negative value signifies that while biomass is produced oxygen is consumed in the reaction. To determine the values
of the other flows in this reaction we can go back to the original balance equation. Thus a carbon balance gives
Fs + Fx + FCO2 = 0
Dividing by Fx we obtain
1 / Yx / s + 1 + 1 / Yx / CO2 = 0.
Since Yx/s = −0.6 C-mole/C-mole, substituting this we get

Yx / CO2 = 1.5 C-mole/C-mole,
a positive value indicating that both biomass and carbon dioxide are produced in the reaction. These values can be used
to calculate the ratio of oxygen consumed to carbon dioxide produced which is called the respiratory quotient (RQ)
given by YCO2/O:
YCO2/O = Yx/O/Yx/CO2 = 1.7/1.5 = 1.13 (with a negative sign which is being ignored since RQ is a ratio).
Similarly a nitrogen balance gives
F2C2 + F4C4 = 0
or
Fx × 0.2 + FN = 0,
Yx/NH3 = −5 C-mole/C-mole.
Thus all the flows can be determined with respect to one C-mole of biomass formed.
Can you now determine whether water is produced or consumed in this reaction?
Example 5 Consider another simple case where there is anaerobic growth as happens in yeast fermentations producing
ethanol. Then since Fo = 0, we have
Fsγs + Fxγx + Fpγp = 0.
If the sources of carbon and nitrogen are glucose and ammonia, respectively, then on dividing through by Fx we
obtain.
4/ Yx/s + 4.2 + 6 Yp/x = 0,
where Yp/x is the product yield per unit biomass on a C-mole basis and γp = (4 + 3 – 0.5 × 2) = 6 (since the formula of
ethanol is CH3O0.5). The above equation allows us to determine the product yield if the biomass yield is known. For
example, if Yx/s = 0.1 C-mole/C-mole then substituting this in the above equation, we get
Yp/x = 5.97 C-mole/C-mole
It often makes sense to determine Yp/s, the product yield per unit substrate. Dividing the above equation by Fs, we get
g s + Yx / sg x + Yp / sg p = 0.
The maximum product yield Yp/s is achieved when there is no cell growth and the biomass yield is zero (e.g. while
running an immobilized column of yeast cells which converts glucose to ethanol). This is given by
g s + Yp / cg p = 0.
substituting values of γs and γp, we obtain

779
Yp/s = −2/3 C-mole/C-mole.
Thus even when no biomass is produced one-third of the carbon flux goes to produce carbon dioxide.
The student may have perceived by now that the above choice of γ has in effect given us an electron balance of this
complex reaction of substrate consumption with biomass and product formation. In aerobic growth the ultimate electron
acceptor is oxygen where each molecule of oxygen can accept four electrons, while in anaerobic growth the product
is the final electron acceptor. The student is also expected to try out the material balances using different carbon and
nitrogen sources. For example, if glutamate is used as the nitrogen source, the value of λN = −21 and thus γx = 4 + 1.8 −
0.5 × 2 − 0.2 × 21 = 0.6. Thus, even though the equation remains
Fsg s + Fxg x + Fpg p + Fog o = 0,
the value of γx would change from 4.2 to 0.6.

However the above formulation is not limited to this choice of λ′s. We can, by an appropriate choice of λ′s, change
the form of the above equation, depending upon which flows can be measured easily (that in turn depends upon our
experimental setup). For example, we can have an exit gas analyzer to estimate oxygen consumption and carbon dioxide
production but having a substrate like cellulose whose consumption rate is difficult to estimate. Assuming the substrate
formula to be CH2O we can choose λc = 0. This makes γs = 0 + 2−2 = 0 and the final material balance equation becomes
Fxg x + Fpg p + Fog o + FCO2 g CO2 = 0.
If NH3 is the nitrogen source (λN = −3), γx = 0 + 1.8 + 0.5x − 2 + 0.2x−3 = 0.2 γ0 = −4 and γCO = 4.
2
For a case of no product formation we have
Fx (0.2) + Fo (−4) + FCO2 (−4) = 0.
Dividing by Fo we get
0.2 Yx/o − 4 − 4RQ = 0,
which allows us to relate the (RQ) with the biomass yield of oxygen.
Example 6 Consider a reactor where air is bubbled at the rate of 0.5 vvm (vvm represents air flow rate per unit volume of
culture per minute, thus 0.5 vvm mean 0.5 l of air per min per l of culture). If the inlet oxygen and carbon dioxide concen-
trations are 21 and 0% and outlet concentrations are 19.5 and 1.7%, respectively, determine, the rate of biomass formation.
Solution: Basis 1 l of culture
Since 22.4 l of gas at NTP contains 1 mole of the gas (ignoring temperature and pressure corrections). Oxygen
consumed per = 0.5(0.21 − 0.195)/22.4 = 3.35 × 10−4 mol.
CO2 produced per min = 0.017 × 0.5/22.4 = 3.795 × 10−4 mol.
RQ = 1.13.
Substituting in the previous equation:
0.2 Yx / o − 4 − 4 RQ = 0 (since RQ is -ve)
Yx / o = (−4 × 1.13 + 4) / 0.2 = −2.65 C-mole/mol −1 .
Therefore rate of biomass production per liter of culture in (gh−1) is

2.65 × 3.35 × 10−4 × 60 × 24.6 = 1.312 g h−1
A more accurate estimate would require correcting temperature, pressure and changes in humidity of the inlet and
outlet air, and also the small change in molar flow rates.
Can you now determine the rate of substrate consumption?
17.3.2 Energy Balance in a Biological System

Here we will briefly touch upon the basic theory of enthalpy and free energy balances in a biological system with specific
reference to microbial growth. We know from thermodynamics that the entropy of a system or more specifically the free
energy always increases. However this is true only for closed systems. We can, for open systems, write the energy balance
equation as:
Accumulation = Input − Output ± Reaction
For spontaneous reactions the free energy change (ΔG) is always negative, thus the reaction term in this equation
would always be negative. However the system free energy denoted by the accumulation term may change either way
(or remain constant) depending on the input and output values.
For enthalpy balance we can write
ΔH reactants − ΔHproducts = qmet
where qmet is the sensible heat produced during the reaction. During growth, substrate is consumed and biomass (and
products) formed and the difference in their enthalpies gets reflected in the metabolic heat qmet generated during the
process. We can safely approximate the enthalpy content of oxygen, carbon dioxide and water to zero and calculate ΔH
for substrate, biomass and product by the standard heats of formation.
In this regard an interesting observation which has been confirmed empirically for a large number of substrates is
that if we choose λC = 4, λH = 1,λO = −2 and λN = 0 then the γ values obtained are proportional to the standard heats of
formation, that is
∆H i ∝ g i′,
where the g i′ is used to denote that λN is chosen as zero.

(This choice of λN implies that g N′ is no longer zero, it would be 3 for NH3 and 4.2 for glutamate.) The proportionality
constant between ΔHi and g ′i has been empirically determined to be 115 kJ (C-mole)−1 with a 5–10% error
Thus using this correlation we can determine:
∆Hglucose = 115 × 4 = 460 kJ (C-mole)-1
∆H NH 3 = 115 × 3 = 345kJ mol −1 ,
which is a fairly good approximation of the real values.

We can now write the earlier material balance equation in the form
Fsg s′ + Fx g x′ + Fpg p′ + FN g N′ + Fog o′ = 0
(Note the g o′ remains equal to −4 and g N′ is no longer zero.)

Incorporating the direction of the flows into the equation so that only positive values are used, we have (since
substrate and nitrogen source are consumed and biomass and product are formed)
Fs′g s′ + FNg N′ − Fx g x′ − Fpg p′ = 4 Fo .
We can multiply this equation by 115 kJ (C-mole)−1 to get

Fs ∆H s + FN ∆H N − Fx ∆H x − Fp ∆H p = 460Fo
The LHS of this equation represents

ΔHreactant − ΔHproducts;
hence the metabolic heat is generated. qmet can be determined as
qmet = 460Fo
This is an important relationship which tells us that the heat produced in a fermentation is proportional to the
oxygen consumption. Since both heat removal and oxygen transfer are critical aspects of fermentor design their coupling
tells us that if one of the aspects (heat removal or oxygen transfer) is a problem then the other one will be a problem too.
The enthalpy efficiency ηH of the process can be defined as
hH = ∆H products / ∆Hreactants
= (115Fxg x′ + 115Fpg p′ ) / (115Fsg s′ + 1155FNg N′ ).
For no product formed and NH3 as nitrogen source, this becomes

hH = Fxg x′ / (Fsg s′ + FNg N′ ),
781
which can be approximated to

hH = Fxg x / Fsg s = Yx / s × g x / g s
Use of γ instead of γ′ allows use of previous material balance equations. Thus FsΔHs is the enthalpy content of the
substrate of which ηH(FSΔHS) is the enthalpy content of the product (in this case biomass) and (1−ηH)FSΔHS is the
metabolic heat generated in the process. The student can easily verify that
FsΔHs (1 − ηH) = 460Fo
We can now formulate two boundary conditions: first that ηH<1 which states that the enthalpy content of the
p roducts can never be higher than the reactants (enthalpy limitation) and second a carbon limitation, that is Yx/s < 1
which states that the carbon content of the product can never be higher than the reactants (if the nitrogen source
contains no carbon).
If we now plot these two boundary conditions by looking at how the yield coefficient (Yx/s) changes with increasing
values of γs we get Yx/s = ηHγs/γxThus as γs increases (for ηH<1), the upper bound of (Yx/s) increases till (Yx/s) reaches 1 after
which it enters the carbon limitation phase (Figure 17-6).
The figure simply demonstrates that when γs values are low, the substrate enthalpy content is also low. Thus since
the enthalpy content of the product has to be lower than reactants (enthalpy limited condition), the upper limit of Yx/s
increases linearly with increasing γ till it reaches 1 after which it does not increase further (carbon-limitation).
This switch from enthalpy-limited to carbon-limited would take place around a γs value of 4.2 (since γx = 4.2).
Let us now see how the plots of nH and qmet change with γs (Figure 17-7).
Figure 17-7 shows that the enthalpy efficiency declines as γs increases beyond 4.2 because once the upper limit of
Yx/s is reached the excess energy per C-mole of the substrate is simply lost as metabolic heat. This is obvious from the
equations
hH = Yx / sg x / g s and qmet = 115Fsg s (1 − hH ).
Experimental results from a range of fermentations show that the typical values of both nH and Yx/s never exceed 0.6,
much less than the upper limit of 1 set by the boundary conditions. As a matter of fact most fermentations which are
enthalpy-limited have a nH value between 0.5 and 0.6 and those which are carbon-limited a Yx/s value between 0.5 and
0.6. These values give us a very useful handle to predict within a fairly narrow range the expected performance of any
bioreactor with respect to its biomass and product yield, oxygen requirements, metabolic heat generation, etc., all of
which are useful parameters to know while designing a bioprocess.
We end this section by noting that, it is the free energy change (ΔG) and not ΔH which drives a reaction forward.
Thus ideally we should have dealt with free energy changes. However correlations for ΔG have much more scatter (and
hence error) and it is best to consult tables of the standard free energies of formation for accurately determining the
change in free energies. A typical correlation with around 10–15% error is
∆Gi = 94.4 × g i′ + 85.5.
Enthalpy Carbon
limitation limitation
H
qmet
Yx/s
4.2 s
s
H qmet
Figure 17-7 Effect of γs on enthalphy efficiency and

Figure 17-6 Effect of γs on yield coefficient. metabolic heat generation.
The student may use this correlation to determine the free energy change in a fermentation. In anaerobic fermentations
no oxygen is consumed, hence the enthalpy change (as predicted by the earlier correlation) is zero. This is indeed true
to a large extent because the metabolic heat generation is fairly low in an anaerobic system. (This is the reason why
anaerobic waste treatment plants do not require any heat removal devices and hence are easy to install and run).
However the free energy change is nonzero and it is the large entropy increase which drives the reaction forward.
Can you use the correlation given to determine the relationship between change in free energy and carbon dioxide
production in an anaerobic system?
17.4 Bioreactors
In any fermentation process the bioreactor plays a central role in determining the process efficiency. Even with recom-
binant products where stringent quality control implies that downstream processing is the major cost component, it is
the bioreactor performance which determines product yields.
Any vessel with facilities for aeration and agitation can be used as a bioreactor. Additionally it must meet the require-
ments of aseptic operation and provide the cells with a controlled environment conducive to growth and product for-
mation. Traditionally the vessel of choice has been the stirred-tank bioreactor which consists of a vessel with a vertical
rotating shaft with agitator blades (Figure 17-8).
Since the vessel must be sterilized, it must meet the requirements of pressure vessel design (since steam is used under
pressure to sterilize the bioreactor). Thus the material of construction is usually 4–5 mm thick stainless steel, which is
also resistant to the acids that are typically produced during fermentation. The height to diameter ratio (H/D) varies
from 1 (for small reactors) to 3 (for larger vessels). A high H/D ratio provides for a smaller footprint (space saving),
ease of construction and better mixing since the agitator blades need to be proportionally larger as the vessel diameter
increases. However tall tubular reactors with a high H/D ratio often lead to oxygen starvation in the gas phase which
affects oxygen transfer rates. The other factor which is critical for large reactors is heat transfer since the metabolic heat
generated during growth has to be removed. Cooling coils are used to carry cold water or a mixture of glycerol and water
if the temperature of the coolant has to be less than 0 °C.
The design of the agitator blade plays a crucial role in oxygen transfer and mixing. Traditionally the flat blade
impeller, the so-called Rushton stirrer (Figure 17-9) was used. This stirrer provides good radial mixing and does not get
‘flooded’ with air bubbles even with high air-flow rates. However it provides poor bulk mixing and this becomes a prob-
lem for fungal fermentations. This is because fungal broths shows non-Newtonian behavior where the viscosity changes
with the shear force. Thus the air bubbles get channeled through the center of the fermentor where the viscosity is
low (due to high mixing) effectively starving the media which are
Stirrer close to the walls of oxygen. To prevent this, agitators with better
shaft seal
bulk mixing characteristics have been designed, such as the Scaba
Aseptic
inoculation pipe
agitator and the Prochem Maxflow agitator (Figure 17-9). When
cells are shear sensitive the agitator is designed like a marine pro-
Working level
peller which gives good axial mixing.
Often the problems of heat and oxygen transfer cannot be
Impeller addressed with the conventional stirred-tank design. Airlift reac-
tors are then used to provide better oxygen transfer. These reactors
are modifications of bubble column reactors which have been used
Baffle in beer production. In bubble columns the mixing is provided by
the stream of bubbles entering the reactor at the bottom which
helps to reduce construction and operating costs. In airlift reactors
the liquid also rises with the bubbles through the ‘riser’, disengages
with the gas phase and comes down through the downcomer. This
Sampling ‘downcomer’ may be internal or external to the reactor. The cir-
P culation loop set up helps in improving heat and mass transfer
point
rates. Other specialized bioreactors are required for specific needs.
Thus photo-bioreactors need to have a large specific surface area
Sterile air line
(i.e. surface area per unit volume) since the growth of cells is
Air sparger dependent on the incident sunlight. Thus instead of a large tank,
Drain the cells are grown in tube banks (i.e. tubes arranged parallel to
point each other) and the material of construction is Plexiglas which is
Figure 17-8 Schematic of a stirred-tank bioreactor. transparent to sunlight. The media (usually seawater) enters with
783
17.5 Kinetics of Microbial Growth•
(A)
(B) (C)
Figure 17-9 Design of various agitators: (A) Scaba agitator, (B) Rushtor stirrer, (C) Prochem Maxflow agitator.
a small inoculum, flows through the tubes like in a plug glow reactor and provides a high cell density at the outlet. Often
a small fraction of the outlet cells is recycled to provide a continuous inoculum. Many novel bioreactors have been
designed, like the pulsed column bioreactor which combines a pneumatic or mechanical pulsing of the reactor medium
with a bubble column design. This helps in bubble break up thereby increasing the surface area and oxygen transfer rates.
Since the shear forces are low, this setup can be used for highly aerobic but shear sensitive organisms. However, most
of the designs remain at the lab bench as they have not been taken up by industry which still prefers conventional and
time-tested designs for large-scale operation.
17.5 Kinetics of Microbial Growth

Cell growth is an autocatalytic reaction since the rate of production of cells depends upon the number of cells present at
any given time. Thus each cell in the culture may divide into two, as is the case with binary fission or budding, to give
rise to many daughter cells. The important point to note is that the growth rate is a function of the cell amount. Thus
it makes much more sense to talk of a specific growth rate (i.e. the growth rate per unit biomass) rather than the growth
rate itself. The specific growth rate is a measure of how fast an individual cell is reproducing and this can be more mean-
ingfully correlated with the nutritional environment of the cell. Similarly we can talk of a specific product formation
rate rather than product formation rate, the former correlating much better with the nutritional status of an individual
cell. The specific growth rate is defined as μ≡(1/x)(dx/dt) where (≡ denotes equality by definition); similarly the specific
product formation rate is defined as qp ≡ (1/x)(dp/dt)
The specific growth rate (µ) and specific product formation rate (qp) are ‘intrinsic’ parameters depending on the
cell characteristics and their environment and do not depend on the scale of the culture, since they concern the indi-
vidual cell. Such scale independent kinetic parameters are useful in scaling up performance data from a lab-scale to
industrial scale. However constancy of µ or qp would also require that the environment which the cell ‘sees’ (i.e. the
microenvironment) remains unchanged during scale up.
From the definition of µ we note that the reaction rate for biomass production is given by
dx/dt = µx,
where µ is of course a complex function of many parameters like substrate concentration (both carbon and nitro-
gen sources and rarely the concentration of critical micronutrients as well), pH, temperature, by products, oxygen
availability, etc. Written in mathematical form we say
m = f (s, pH, temperature, O2 ,…).
Many interesting kinetic equations have been proposed to define this function, from the very simple to extremely
complex. However given the typical statistical errors associated with kinetic measurements especially in biological sys-
tems (things are changing a bit now!), there is little advantage gained in choosing complex kinetic equations to describe
cell growth. Thus unless the experimental data is sufficiently precise so as to distinguish between simple and complex
models of growth kinetics, the choice is fairly clear. One of the most popular (and simple) models which assumes that
the specific growth rate is a function only of the substrate concentration (usually assumed to be the carbon-source) is
the Monod equation
µ = µms/(Ks + s),
where µm is the maximum specific growth rate of the cells and Ks is a measure of the affinity of the cells towards the
substrate. The equation is similar to the Michelis Menten equation for enzyme kinetics and thus has the shape as shown
in Figure 17-10.
This similarity with enzyme kinetics is because cell growth is limited by the substrate uptake rate which in turn is
governed by the enzymes responsible for transporting the substrate inside the cells. However note that while Vmax in
Michelis Menten kinetics can be increased by increasing the enzyme concentration, µm is an intrinsic property of the
cell denoting the maximum rate of growth when it is not limited by substrate availability.
Example 7 Consider the situation when cells are growing at µm. [This assumption only requires that the substrate
concentration s is much larger than Ks (from Figure 17-10 you can see that µ ≈ µm when s >> Ks).]
If µm = 0.5 h−1determine the time required for cells to increase 10-fold in a bioreactor?
We have dx/dt = μmx or
xf t =t
∫ xi
dx / x = ∫
t=0
m m dt ,
ln(x f / x i ) = m m t,
where xi is the initial cell concentration at time t = 0 and xf is the final cell concentration at time ‘t’. The above equation
is referred to as the logarithmic form of the growth equation.
We can also write the above equation in the form
xf /xi = emt
which is referred to as the exponential form of the growth equation. Given that xf /xi = 10, we have
ln 10 = µt,
t = (ln 10)/0.5 = 4.6 h.
We can use the above equation to derive the relationship between the time required for the biomass to double, the
so-called ‘doubling time’ (td) and specific growth rate:
ln(xf /xi) = µt.
B A
m Given that in time td the value of xf = 2 xi, we get
ln 2 = µtd,
td = (ln 2)/µ.
Note that these two parameters are inversely correlated; the higher
the specific growth rate, the smaller is the doubling time.
The concept of doubling time is practically useful only in c ircumstances
when cell numbers can be counted and when they are undergoing binary
fission. Even then the cell number of a population increases continuously
and not in discrete jumps as is the picture obtained when one visualizes
one cell dividing into two and then into four and so on. This is because
there is a large population of cells that is dividing asynchronously in a bio-
KS
reactor. Such a number-based analysis also becomes tedious when we need
s
to know the cell number after time intervals that are not a whole multiple
Figure 17-10 Monod kinetics relating specific of the doubling time (say after one and a half doublings). Thus ‘doubling
growth rate (µ) to substrate concentration (s). time’ method of calculations finds little use in growth kinetics other than
785
when genetic stability is being studied as a function of the number

of generations elapsed (since every doubling corresponding to one
III IV
generation can be associated with a probability of emergence of a
revertant mutant). V
17.5.1 Growth Kinetics in Batch Culture

A batch culture is a system where either a simple shake flask or a X
II
bioreactor is used to grow cells. A small amount of growing cells
are inoculated in a sterile medium and allowed to grow till the
substrate is exhausted. Other than oxygen which is supplied by
agitation and/or bubbling air and CO2 which leaves the system, I
the system has no inputs or outputs.
Figure 17-11 shows a typical growth profile of cells in batch T
culture. Stage I represents the ‘lag’ phase where little or no
growth is observed while the ‘inoculated’ cells adjust to the new Figure 17-11 Growth profile of cells in batch culture.
environmental conditions. This internal adjustment of enzyme
levels, etc. allows the cellular machinery to respond optimally to the new culture conditions that the cell ‘sees’; thus if
healthy cells that have been growing in a similar medium are used for inoculation, they would have an extremely short
lag phase. Following this phase the cells enter into the exponential phase where the specific growth rate (µ) is constant.
This is because the substrate concentration ‘s’ is much higher than Ks and the value of µ is close to µm. We can visualize
this as a phase when the substrate concentration drops from the point A to point B (Figure 17-10) thus not seriously
affecting the value of µ. In other words, the availability of substrate is not ‘rate-limiting’ and hence the cells grow at their
maximum specific growth rate. In this stage we can write the kinetic equation for growth as
ln(x/x0) = µmt.
Thus a plot of ln x versus time would be a straight line with a slope of µm. Following this is the post-logarithmic phase
of growth when the substrate concentration declines leading to a fall in the specific growth rate. Note that because the
cells are still growing exponentially, only the exponent (i.e. µ) declines very rapidly. This phase of growth (Stage III) is
usually very short since Ks values are very small. However this stage, where µ values are low, is often a very critical phase
for the formation of certain products. A major drawback of batch culture is the inability to extend this phase of growth
for longer time periods allowing the production of specific metabolites. Also this short post-exponential phase means
that the data available from a batch culture run cannot be used to accurately determine Ks values. However batch culture
has the advantage of simplicity of operation. In the next section we will model this exponential and post-exponential
phase by setting up the balance equations for substrate and biomass. The stationary phase which follows this phase is a
situation of no growth because the culture is starved of nutrients. This ultimately leads to the death phase where the cells
lyse in the culture medium leading to a reduction in biomass densities.
It is important to remember that biomass can be measured using different techniques. The simplest and most com-
monly used method is measuring the absorbance at 600 nm. Measuring at lower wavelengths for example 550 nm not
only increases the sensitivity of the technique but also the interference by components of the culture medium. The cell
size, the relative refractive index and secondary scattering levels all of which change during the growth process need to
be corrected if an accurate estimate of biomass is required.
Techniques like measuring the wet weight or dry weight of cells are tedious but more accurate when high cell densi-
ties are involved. Typically for E. coli, the absorbance at 600 nm (O.D600) correlates with weight measurements by the
following equation:
Dry cell weight (DCW in g l−1) = 0.4 × O.D600 : Wet weight (g l−1) = 4 × DCW.
The variation in these correlations is around 10% and caution must be exercised while working with other cell types.
The above methods encounter difficulties when insolubles are present in the medium which interfere with absorbance
measurements and also get pelleted with the biomass during centrifugation. Indirect methods like estimating the DNA
or total protein content are then used to calculate biomass concentration. Total nitrogen can also be estimated using
techniques like the Kjeldahl method. However RNA is not used as a measure of biomass content since the fraction of
total RNA in a cell is a function of the specific growth rate of the cell. Recent advances in the development of probes
which can measure fluorescence, conductivity or capacitance of the culture medium are being used to estimate cell con-
centration on-line, in modern bioreactors.
As cells grow in a batch reactor the substrate concentration declines proportionally. This decline can be related to
cell growth by the following equation:
Yx/s = Δx/Δs.
Since the reactor volume is a constant, we can write
Yx/s(s0 − s)=x − x0, (17.6)
where Yx/s is the yield coefficient which is the stoichiometric ratio between biomass production and substrate consumption,
s0 is the initial substrate concentration, x0 the initial biomass concentration and s and x are substrate and biomass con-
centration at any given time. The increase in biomass concentration with time can be found using the biomass balance:
Accumulation = Input − Output ± Reaction.
Since the input and output terms are zero we have accumulation = reaction or
dx/dt = µx. (17.7)
Note that this is not a definition but a material balance on biomass which assumes no volume changes and no inputs
or outputs into the reactor.
The third equation is the kinetic equation governing the relationship between specific growth rate and substrate
concentration
µ = µms/(Ks + s). (17.8)
This is a simplification since in a reactor with cell growth the specific growth rate will be affected by more than one
parameter (e.g. changes in pH by product accumulation reduction in the concentration of other nutrients in the culture
medium, etc.)
Solving the three equations above, we have
s = s0−1/Yx/s (x − x0).
Therefore
dx / dt = { m m [ s0 − 1 / Yx / s (x − x 0 )]x } / { K s + [ s0 − 1 / Yx / s (x − x 0 )]}.
This equation can be integrated using the approach of partial fractions:
∫ {K Ys x/s + s0 Yx / s − x + x 0 )dx } / { x(s0 Yx / s − x + x 0 )} = ∫m m dt.
or
∫A dx / x + ∫B dx / (s Y 0 x/s − x + x 0 ) = m m ∫ dt,
where
A = (s0 Yx / s + x 0 + K s Yx / s ) / (s0 Yx / s + x 0 ),
B = K s Yx / s / (s0 Yx / s + x 0 ).
Solving and putting appropriate boundary conditions (x = x0 at t = 0) we get

A ln x / x 0 − B ln(s0 Yx / s − x + x 0 ) / s0 Yx / s = m m t. (17.9)
This equation bears further investigation.

First we note that the maximum value of x, that is, xmax = (Ys0 + x0).
This is the situation where s approaches zero and no further biomass formation takes place. Second if we plot this
equation to get x vs. time, we get a sigmoidal plot as shown in Figure 17-12. Thus this kinetic model is able to simulate
both log phase and declining log phase growth fairly well.
Let us illustrate by an example.
Example 8 Let the values of the various parameters be: x0 = 0.1 gl−1; s0 = 10 gl−1; Yx/s = 0.5 gl−1; µm = 0.5 h−1; and Ks = 0.1 gl−1;
We have
A = (S0 Y + x 0 + K s Y ) / (S0 Y + x 0 ) = 1.01 and B = 0.01.
787
Thus the growth equation becomes

1.01 ln(x / 0.1) − 0.01 ln{(5.1 − x) / 5.1} = 0.5t.
We can substitute various values of x in order to get ‘t’. The

student can plot the graph of x vs. t using the above equation and
also note that if we determine t using the much simpler equation
ln x/x0 = µt where µ = µms0/(Ks + s0) (a constant), the error is less
than 5% till values of x ≤ 5.0 gl−1. Since xmax = 5.1 gl−1 the above x
exercise demonstrates that the declining log phase lasts for a very
short period. This is primarily because typical Ks values for sub-
strates like glucose are very low and hence µ values remain close
to maximum for a large part of the fermentation.
17.5.2 The Ideal Plug Flow Reactor

t
Consider a reactor where the culture fluid flows like a plug in a
pipe (Figure 17-13). Figure 17-12 Sigmoidal growth kinetics.
Medium continuously enters the pipe along with a small inoc-
ulum. A small fraction of the outgoing cells may be recycled back to provide this inoculum. It is assumed that the
medium and the cells flow through this reactor without any significant mixing (like a plug of liquid hence ‘plug flow
reactor’). As this plug of liquid moves ahead the cells continue to grow and consume substrate so that by the time they
reach the outlet the inlet substrate has been depleted and the outlet biomass had increased proportionally. The profiles
of the substrate and biomass concentrations along with reactor length are shown in Figure 17-14.
We see that this profile is similar to the batch culture profile where instead of time on the x-axis we have reactor
length. We can see that each ‘plug’ of liquid which moves along the reactor length is like a small batch reactor progress-
ing in time. Thus the further this plug moves down the reactor the more time it gets to grow. The distance moved is
related to the time by the equation t = L//V where ‘V’ is the velocity of flow. The ratio of the final to the initial biomass is
related at steady state by the recycle fraction. We can, therefore, relate the minimum fraction that needs to be recycled,
to the reactor length and velocity of flow by the equation
ln(xf /xi) = mm(L/V), (17.10)
where xi is the recycled cells which serves as the inoculum and xf is the concentration of the cells at the outlet of the
reactor.
When the recycle fraction is larger the substrate gets exhausted and the specific growth rate falls appropriately.
17.5.3 The Ideal Continuous Stirred Tank Reactor

Laboratory-scale studies on fermentation are done essentially to determine the kinetic parameters of growth and prod-
uct formation. A lab-scale bioreactor is primarily an analytical, not preparative tool, because the kinetic parameters
determined can be used to predict bioreactor performance on an industrial scale. In this regard the continuous stirred
tank reactor (CSTR) is an ideal system since it can hold the cells at the desired specific growth rate (µ) for long periods.
The CSTR has also found use in an industrial setting where improvements in process control and the problems of con-
tamination have been adequately addressed, because these are much more critical while running continuous cultures.
The advantage gained is the enhanced productivities achieved because the down-time associated with batch systems is
removed.
Figure 17-3 shows the schematic of a CSTR. Cells are grown in a stirred tank with aeration. Sterile feed enters the
system and the outflow consisting of spent medium, cells and product is taken out at the same flow rate thereby keeping
the reactor volume constant. At steady state the rate of production of cells in the reactor matches the outflow, because
the input is sterile. Thus a CSTR needs to be ‘started’ as a batch reactor and the input and output are started only when
sufficient growth has taken place. The material balance on biomass and substrate is as follows:
Recycled cells
Input Output
Figure 17-13 Schematic of a plug flow reactor. (substrate) (cells)
Accumulation = Input − Output ± Reaction,
S Biomass balance V(dx/dt) = F × 0 − Fx + µxV,
X where F is the flow rate of both input and output and x
is the biomass concentration in the reactor, which is also
equal to the biomass concentration in the output (since the
reactor is well mixed). At steady state, if we denote the
biomass concentration by x , we get
Fx = m x V or m = D [ because D = F / V ]. (17.11)
X The units of D are in time−1, usually h−1, which is the

S same for specific growth rate. Thus by simply adjusting
L the dilution rate (and waiting for steady state) we can get
the cells to grow at the desired specific growth rate.
Figure 17-14 Substrate and biomass profile in a plug flow The substrate balance gives
reactor.
V(ds / dt) = Fs0 − Fs − mxV / YX / S .
S0
At steady state, substituting µ = D, we get

Yx / s (s0 − s ) = x . (17.12)
X
S X
This expression can also be derived directly if we use
S
an overall material balance on substrate. Since the ratio
of substrate consumed to biomass produced is Yx/s, we have
Yx/s = Δx/Δs = (x − 0)/(s0 − s).
This overall balance is useful since it tells us that even
with recycle streams attached to the reactor the above
D m
equation will still be valid.
Figure 17-15 Plot of steady-state substrate and biomass Finally the kinetic equation for growth completes the
concentrations versus dilution rate in a CSTR. description of the system, that is
m = m m s / (K s + s ).
Using the above equations we can predict the biomass and substrate profiles at steady state at different dilution rates:
D = m = m m s / (K s + s ) or s = DK s / (m m − D)
and similarly
x = Yx / s [ s0 − DK s / (m m − D)].
Figure 17-15 show a plot of s and x against D. Note that the value of x does not change significantly when the values
of D are low. This is because of the typically low values of Ks which lead to low values of s . Thus for a wide range of D
values, the substrate gets almost completely consumed, leading to a constant biomass concentration in the reactor. (The
student may check out as to how this pattern changes when the Ks values change.)
As the dilution rate is increased, µ rises till it reaches the upper limit given by µm. At this point the rate of production
of biomass cannot keep pace with the output flow and the cells start getting washed out of the reactor. This rate of ‘wash
out’ can be determined by the following equation:
dx/dt = mmx − Dx. (17.13)
which is a negative term given that D > µm. Slow washout is often used in lab-scale reactors to determine the value of µm
by running the CSTR at values of D greater than µm and plotting ln x vs. time. This gives a slope of (μm−D) from the
equation ln x = (μm−D)t.
The CSTR can also be used to determine the values of YX/S, µm and Ks by measuring outlet biomass and substrate
concentrations for different values of D.
789
Since D = m = mms/(Ks + s), we have

1 / D = K s / mm s + 1 / mm .
A plot of 1/D vs. 1/ s (double reciprocal plot) has a slope of Ks/µm and an intercept of 1/µm. Yx/s can be determined
from a plot of x vs. (s0 − s).
17.5.4 The Problem of ‘Maintenance’ Requirements

It turns out that when a plot of x vs. (s0 − s) is made, very often a straight line is not obtained signifying that the yield coef-
ficient Yx/s is not a constant, but changes with changing specific growth rate of the cell. This aspect has been recognized
long ago and a simple model to explain this variation is based on the concept of maintenance requirements for cell growth.
We assume that for a given amount of substrate consumed by the cell, only a part is used for growth purposes and the
remaining is consumed in order to ‘maintain’ the cell. Thus
ΔsT = ΔsG + Δsm.
where ΔsT = total substrate consumed, ΔsG = substrate consumed for growth and Δsm = substrate consumed for maintenance.
Dividing by Δx (the biomass produced) we obtain
ΔsT/Δx = ΔsG/Δx + Δsm/Δx
The term Δx/ΔsT is called the Y , that is the observed yield, while Δx/ΔsG is defined as the true yield Yxtrue
obs
x/s / s because
this is the yield if all the substrate were to be utilized for growth. We also define a maintenance coefficient ms which is
given by the rate of substrate consumption per unit biomass per unit time. Thus,
ΔsM = m × x × Δt.
Given that Δx = μxΔt, we have
/ s = 1 / Yx / s + m / m .
1 / Yxobs true
(17.14)
/ s vs. 1/µ would thus give a straight line with a slope ‘m’ and intercept 1 / Yx / s .
A plot of 1 / Yxobs true
Such a graph can easily be plotted using a CSTR where µ = D and Yx / s = x /(s0 − s).
obs
[Note that if we introduce the concept of m (and hence a varying Yx/s) in a batch reactor, solving the integral in order
to get x vs. time is no longer possible algebraically and numerical techniques will have to be used.]
The plot of x vs. D would also change if the maintenance requirements are significant (Figure 17-16) while the plot
of s vs. D would remain unchanged.
The concept of maintenance is also useful in describing stationary phase cultures where substrate is consumed but no
biomass is produced.
In order to link up with the stoichiometric relationships which were developed in the previous section we need to
understand that the substrate consumed for maintenance purposes is essentially converted to carbon dioxide and thus
oxygen is required for this process. We can thus write analogously for oxygen consumption,
/ o = 1 / Yx / o + m o / m ,
1 / Yxobs true
(17.15)
where mo is the maintenance requirement with respect to oxygen.

In order to relate mo to ms we can use the balance equations
g s / Yxobs
/ s + g x = 4 / Yx / o ,
obs
(17.16) S
X S
X
g s / Yxtrue
/ s + g x = 4 / Yx / o
true
(17.17)
because the balance equations are correct for both situations. Substracting
Eq. (17.17) from Eq. (17.16) and substituting in Eqs. (17.14) and (17.15)
we get
gs (ms/µ) = 4(mo/m) D
or Figure 17-16 Plot of steady-state biomass

concentration vs. dilution rate with siginificant
msgs = 4mo. maintenance requirements.
How would you interpret this equation?

In the case of anaerobic growth
gs + Yx/sgx + Yp/sgp = 0.
Thus Yx/s and Yp/s are inversely related, and a fall in Yxobs
/ s (because of maintenance) translates into a higher product yield.
Example 9: Calculate the product yield (YP/S) for ethanol using an anaerobic culture of yeast growing on glucose as a
carbon source at two specific growth rates, µ = 0.5 h−1 and 0.05 h−1 given Yxtrue
/ s = 0.4 C-mole/C-mole and ms = 0.2 h .
−1
Solution: Since
/ s = 1 / Yx / s + m s / m ,
1 / Yxobs true
therefore at µ = 0.5 h−1
/ s = 1 / 0.4 + 0.2 / 0.5

1 / Yxobs
/ s = 0.345 C-mole/C-mole
Yxobs
Similarly at m = 0.05 h−1; Yxobs

/ s = 0.154 C-mole/C-mole.
g
Since s + Y g
x/s x + Y g
p/s p = 0, therefore at m = 0.5 h−1,
4 + (−0.345 × 4.2) + Yp / s × 6 = 0
Yp / s = −0.425 ( C-mole / C-mole)
and at m = 0.05 h−1,

Yp/s = −0.56 C-mole/C-mole.
The product yield thus increases by lowering the specific growth rate because of maintenance. A high maintenance
requirement is not a bad thing after all.
17.5.5 Product Formation Kinetics

Till now we have not talked about the kinetics of product formation. However, just as it is for biomass formation, the
study of product formation is best done in a CSTR. Also product formation kinetics should be understood in terms of
the production capacity of a ‘single’ cell. We thus talk of specific product formation rate (qp) which is defined as the
product formation rate per unit biomas [qp ≡ (1/x)(dp/dt)]. Product formation is said to be ‘growth associated’ when qp is
proportional to µ, that is qp = αμ, where α is a constant. In such cases we have
(1/x)(dp/dt) = a(1/x)(dx/dt) or (dp/dx) = a.
Since we define product yield Yp/x as Δp/Δx, we observe that for growth-associated product formation YP/ X is a constant
which is equal to α.
All products which are part of the growth process and hence form an integral fraction of the cell itself are
growth-associated products. Examples include: NADP, NADPH, TCA cycle intermediates, etc.
‘Non-growth-associated’ product formation is when qp is a complex function of µ (or not a function). Thus the
simplest kinetic model for non-growth-associated product formation would be
qp = β (a constant).
This simple model is a fairly good approximation for the product formation kinetics of various ‘secondary metabolites’
like penicillin G and many other antibiotics. However, this equation is only valid for a range of specific growth rates
because high specific growth rates suppress product formation, that is
qp = 0; µ > µcritical,
qp = β; µ < µcritical.
This model would explain why cells have to be held in the post-exponential (and stationary) phase for them to
p roduce the desired secondary metabolite. We can postulate more complex kinetic models for product formation (For
example, in the above model we can add the fact the when µ falls to very low values, qp also declines, that is qp → 0 as
µ → 0.)
791
However one model which combines the two models mentioned earlier has found applicability to a wide range of
products. The model is defined by the equation
qp = αm + β.
First proposed by Luedking and Piret in 1959, it has been used by many authors to explain the kinetics of product
formation of lactic acid and many other primary metabolites which are not completely growth dependent.
The model has a growth-associated and a non-growth-associated term both of which can be determined using a
CSTR. We have a product balance:
V dp/dt = (αm + β)x − Dp.
At steady state
(am + b)x = Dp or p / x = a + b / D.
A plot of p / x vs. 1/D would give the values of α and β from the intercept and slope, respectively. [The student may
try and determine the product profiles in a batch reactor for different values of α and β. Thus when β = 0 and the product
is completely growth-associated, the product profile would be similar to the biomass profile (since YP/X is a constant).]
17.5.6 Modifications of the CSTR

Steady state is achieved in a CSTR by keeping the dilution rate (or flow rate) constant. This kind of CSTR is also
referred to as a chemostat. However, instead of holding the dilution rate constant, we may hold the biomass concentra-
tion in the reactor a constant by measuring it spectrophotometrically and adjusting the flow rate accordingly (increasing
the flow rate if the O.D. (biomass concentration) increases and reducing it if it decreases). This way of running a CSTR
is called a turbidostat and a steady-state operation can be achieved by this method as well. Similarly we can think of a pH
stat or a D.O. stat (which keeps the dissolved oxygen in the culture medium a constant). The difference between these
various techniques of achieving steady state in a CSTR is to do with the control efficacy (in other words how effective
they are in holding the CSTR at the desired set point of operation). To illustrate this let us consider a CSTR containing
cells which have a very low Ks value. The plot of x vs. D is given in Figure 17-17.
If we want to run the CSTR at the desired biomass concentration x D then the dilution rate has to be D*. A minor
variation in the flow rate (due to fluctuations in the pump) would lead to major changes in the value of x D , even leading
to washout since D* is very close to µm. On the other hand, a turbidostat would do this job effectively because using an
appropriate feed back control loop it can fine tune the value of D to get the desired value of biomass concentration in
the reactor.
Note that the turbidostat becomes ineffective at low dilution rates (and hence high biomass concentration) since
under these operating conditions the biomass density (which is the measured variable) is completely unresponsive to
changes in dilution rate. Similarly a pH stat would be effective in a cell culture which generates a lot of acids (which, in
turn, should be dependent on the specific growth rate of the cells). The D.O. stat keeps the oxygen uptake rate a con-
stant and hence can be used to control the growth rate (given by µx) rather than specific growth rate.
17.5.7 Chemostats with ‘Wall Growth’

Consider a nonideal situation where cells adhere to the wall of the reactor
vessel. To model such a system we assume that: (a) the amount of cells X
sticking to the walls is a constant Xw, that is the rate of increase of cells
adhering to the wall is cancelled out by the scouring action of the agita-
tor blades which remove cells from the wall by turbulence and (b) the
cells sticking to the wall ‘see’ the same substrate concentration as the cells
in suspension and hence grow at the same specific growth rate. Defining XD*
xw = X w / V , that is the amount of cells sticking to the wall per unit reac-
tor volume, we have at steady state,
Biomass balance, m x + m xw = Dx .
D D*
Overall substrate balance is, YX / S (s0 − s ) = x . (The student can verify
that we arrive at the same equation if we do the standard substrate balance.)
Figure 17-17 CSTR run at very high dilution
The kinetic equation: m = m m s / (K s + s ). rate (D*) close to wash out.
Solving the above three equations we get m = Dx / (x + xw ) which tells that µ < D in this system. Substituting this
in the kinetic equation, we obtain
DYX / S (s0 − s ) / [ YX / S (s0 − s ) + x w ] = m m s / (K S + s ).
This is a quadratic equation in s , which can be solved to get
Increasing s and hence x .
xw
x Example 10: Given the steady-state wall growth is 1 g l−1.
Determine outlet biomass concentration (x ) at a dilution rate
of 0.5 h−1, given YX/S = 0.4 g/g, µm = 0.5 h−1 and Ks = 0.1 gl−1 and
s0 = 10 gl−1
Solution: Substituting the values in the above equation, we have
0.5 × 0.4 (10 − s ) 0.5 s x
= .
0.4 (10 − s ) + 1 0.1 + s
D
⇒ s = 0.38gl −1 ,
Figure 17-18 Plot of biomass vs. dilution rate for CSTR
with wall growth. which gives x = 3.8 gl −1.
Note that if there was no wall growth we would have observed
xout ‘wash out’ since D = μm (in this case). But wash out never takes
place since the cells adhering to the wall keep growing and pro-
ducing cells. Figure 17-18 shows the plot of x vs. D for different
levels of wall growth.
(The reader may try out how maintenance requirements
would change the above figure.)
17.5.8 Chemostat with Recycle

Internal recycle: Figure 17-19 shows a CSTR with internal
recycle. The outlet flow is connected to a filtration membrane
Figure 17-19 CSTR with internal recycle. set up inside the reactor which partially blocks the entry of cells.
Consequently the outlet biomass concentration (xout ) is lower
than the biomass concentration in the reactor (x ).
At steady state,
xout m x = Dxout .
Here too we have μ < D since x > xout . The ratio of x to xout
Increasing
recycle depends on how efficiently the filter assembly prevents the cells
from passing through the filter and leaving via the outflow. If it
is 100% efficient then the outflow contains no cells ( xout = 0 ).
This is referred to as ‘total cell recycle.’ A steady state can only
be achieved in these circumstances if μ = 0, which is not practi-
D cally attainable for long periods, given that cells enter the death
Figure 17-20 Plot of biomass vs. dilution rate for CSTR
phase after some time. For filter cut-off efficiencies less than
with recycle. 100% we can define a ratio (A) which is equal to xout / x ; where
A ≤ 1. We then have
(F R+F )
F m = AD,
F
xout = Yx / s (s0 − s ),
FR
m = m m s / (K s + s),
which gives
FR
s = ADK s / m m − AD.
A plot of xout vs. D is shown in Figure 17-20. for reducing
values of A. Note that wash out is delayed as A declines, since
we have
Figure 17-21 Schematic of CSTR with external recycle. µm = AD (for wash out).
793
17.5.9 CSTR with External Recycle

Figure 17-21 shows an external recycle reactor. The outlet stream passes through a settling tank (or any other concen-
trating device) and a concentrated stream of biomass is recycled back into the reactor. A dilute stream of cells thus,
finally, leaves the reactor having a concentration xout, where xout < x , which is the concentration of cells inside the
reactor. We can understand the similarity with internal recycle if we look at the overall balance where x in the concen-
tration of the cells inside the reactor, xout the concentration of cells leaving the system and xout < x .
If the cells are concentrated by a factor of ‘g’, the inlet and outlet flow to the system is F and FR is the recycle flow
rate, we can define a recycle ratio (R) given by R = FR/F
Doing a biomass balance around the settling tank, we obtain
(F + FR ) x = F xout + FR g x
⇒ (1 + R) x = xout + Rg x
⇒ xout = (1 + R − Rg) x .
An overall balance gives

V m x = Fxout
⇒ m x = Dxout
⇒ m = (1 + R − Rg)D,
which is similar to the earlier formulation for internal recycle, with A = (1 + R − Rg); this value is less than 1 since g ≥ 1.
(The student may check that a biomass balance around the CSTR also gives the same result.)
We note that all the above methods, viz. CSTR with wall growth, CSTR with internal or external recycle, have a
mathematically similar analysis because in all cases the biomass concentration in the outflow is lower than the biomass
concentration in the reactor thus leading to a situation where μ < D. These reactors can thus run at higher dilution rates
without having a wash-out situation. However, lower µ values lead to low values of YX/S (if we were to consider mainte-
nance requirements) and also enhanced oxygen requirements per unit biomass formed.
Example 11: Waste treatment plants often prefer low concentrations of biomass at the outlet since they cause problems
of disposal. Consider a CSTR with external recycle requiring the outlet biomass concentration to be ≤ 10 g 1−1. Given
that s0 = 50 g l −1 , YX / S = 0.4 g/g , m s = 0.3 h −1, D = 0.2 h−1, g(concentration factor) = 10, what is the recycle ratio at which
the reactor should be run?
Solution: Assuming that Ks is very small, xout = YX / S (s0 ) since s values are around Ks and thus {(s0 − s ) ≈ s0 }
/ s = 10 / 50 = 0.2 g/ g.
Yxobs
Now
/ s = 1 / Yx / s + m / m .
1 / Yxobs true
Substituting the values given we obtain

1 / 0.2 = 1 / 0.4 + 0.3 / m ⇒ m = 0.12 h −1 .
Since D = 0.2 h−1, we have

m = (1 + R − Rg)D
⇒ 0.12 = (1 + R − 10 R)0.2
⇒ R = 0.044.
Thus 44% of the cells need to be recycled in order to achieve the desired concentration of cells in the outflow. Note
that if Ks values were large then µm needs to be known and the solution to this problem becomes quite complex involving
a trial and error method.
17.5.10 Fed-Batch Culture

One of the most favored cultivation techniques used in the industry is the use of fed-batch culture. Typically feed is
added at a slow rate to a batch culture (Figure 17-22) but unlike a CSTR no culture medium is removed.
FSF The reactor slowly fills up and the process is stopped when the vessel
is filled up or the desired product/biomass concentration is achieved. The
reactor may be partially emptied so that the residual culture serves as a
starting inoculum and the feed started once again in the case of ‘repeated
fed-batch’ culture. The slow feed has many advantages, chief among
them being the ability to control the residual substrate concentration in
the reactor and hence the specific growth rate of the cells. Excess sub-
strate concentration often leads to poor product formation (the ‘crabtree’
effect). To avoid this, the initial substrate concentrations are kept low
in the batch phase. However, if only this substrate were available for
Figure 17-22 Schematic of fed-batch culture. growth this would lead to low final biomass concentrations. Feeding sub-
strate slowly into the reactor overcomes this problem thereby helping in
the achievement of high biomass densities. Also the rate of feed can be
adjusted so as to get the desired specific growth rate which is best suited for product formation.
17.5.10.1 Feed of Concentrated Media

In this case the feed consists of a very high concentration of substrate and the feed rate is comparatively low. The
volume change is thus insignificant, but the biomass concentrations rise significantly because of the availability of excess
substrate. A material balance on the substrate gives
V ds / dt = FSF − mX / Yx / s − mX
where F is the feed rate, SF is the feed concentration and X = xV. If the specific growth rate is low then the residual
substrate concentrations are also low (given low Ks values).
Thus all the substrate added is completely consumed by the growing cells giving
FSF = mxV / Yx / s + mX.
Consider a situation when the feed rate is a constant and maintenance requirements are negligible. We get
mx = FSF Yx / s / V,
which is a constant. This is a situation of linear growth where ‘µx’ is a constant and hence µ declines as x increases
throughout the cultivation period.
If we consider maintenance effects to be significant then with increasing biomass a larger proportion of substrate
would be utilized for maintenance and the growth rate would decline slowly till it reaches zero and all the substrate is
utilized for maintenance. In such a case
FSF = mXmax.
Note that this is also the situation when the biomass levels are the highest. If µ is to be held constant, the feed rate
needs to be increased exponentially so as to match the desired specific growth rate.
17.5.10.2 Feed of Dilute Substrate

Here the volume changes cannot be ignored and we need to incorporate this in our material balance equations. Given
x = X/V. Differentiating this we get
dx/dt = (V dx/dt − X dV/dt)/V2.
Since dx/dt = μx and dV/dt = F, we get
dx / dt = V mX / V 2 − XF / V 2
⇒ dx / dt = mx − Dx,
where D = F/V.
If the substrate concentration of the feed were to be the same as the initial batch concentration (and given that the
residual substrate concentration in the reactor is very low), we can define a pseudo-steady-state condition where dx/dt = 0
or m = D. However unlike a CSTR the dilution rate is not a constant since D = F / V = F / (Vo + Ft) for a constant feed
795
17.6 Heat Transfer•
rate. D thus continuously declines leading to a decline in µ. We can consider this situation analogous to the linear growth
described previously, where instead of a linear increase in biomass concentration the volume increases linearly. Note that
in both cases since µ declines the residual substrate concentration declines as well. However, assuming low values of Ks, this
decline does not add significantly to the biomass produced which is given by (since s0  s )
x = YX / S (s0 − s) ≈ Ys0 .
Fed-batch techniques are extremely useful while doing high cell density cultivation (HCDC) and producing second-
ary metabolites, where substrate starvation is an important trigger for product formation.
Example 12: Determine total product formed in a fed-batch culture with constant feed given F = 100 l h −1 , SF = 200 gl −1 ; V = 10, 000 l, Yx / s =
−1 −1
h , SF = 200 gl ; V = 10, 000 l, Yx / s = 0.4 g/g, s0 (initial substrate concentration in the batch phase) = 20 g1−1. The production formation
kinetics is given by qp = 0 if ; m ≥ 0.1h−1; qp = β = 0.05 when 0.02 ≤ m ≤ 0.1 h−1; qp = 0 when m ≤ 0.02 h−1. (Assume Ks to be
very small.)
Solution: Let the feed begin only after cells have grown and the initial substrate is almost exhausted, that is
x = Ys0 = 8 gl −1 .
Considering volume changes to be negligible we have a substrate balance,

FSF = m x V / YX / S .
substituting the values we get

100 × 200 = m × 8 × 10, 000 / 0.4
⇒ m initial = 0.1 h −1 .
Thus product formation would start with the start of feed and continue till µ falls to 0.02 h−1. This fall in µ is because
of increase in x. Using the substrate balance again and substituting µ = 0.02 h−1, we get x = 40 g l−1.
At this point the value of µ falls below 0.02 h−1 and no further product will be formed. The rate of product formation
is given by
q p = 1 / x(dp / dt) = b
p t
dp / dt = bx or ∫ dp = ∫ bxdt
p0 0
where po is the product concentration at the end of the batch phase (which can be considered to be zero. The rate of
increase of x is linear and is given by
dx / dt = mx = FSF YX / S / V = 0.8gl −1 h −1x = x 0 + 0.8t.
Substituting for x in the integration given above and putting appropriate boundary conditions, we get
32
p − p0 = b ∫ (8 + 0.8t)dt = b [8 × 32 + 0.8 × 322 / 2] = 665.6b ,

0
which is the amount of product formed. The problem would become complex if maintenance requirements are introduce
in the material balance, because Yx/s would no longer be a constant.
17.6 Heat Transfer

The metabolic heat generated by the cells needs to be removed because it is important to maintain a constant temperature
in the reactor. The overall heat balance on a reactor would include: the metabolic heat generated, the heat generated
due to stirring and agitation, the heat lost due to evaporation, the sensible heat content of the input and output flows
and the heat exchanged with the surroundings. In a large reactor only the first two terms are important, with the
metabolic heat often contributing more than 80% of the heat load on the system. It is fairly easy to calculate this load if
we know the rates of biomass formation (which are determined by growth

∆X kinetics) and the enthalpy efficiency of the process(ηH):
T1
Qmet = Fx ∆H x (1 − hH ) / hH .
Since ηH represents the fraction of substrate energy going for biomass for-
mation and (1−ηH) is the fraction which is lost as metabolic heat. We need
to look into the basic principles of heat transfer, to be able to design the heat
transfer equipment (cooling coils, etc.) to take care of this heat load.
Heat transfer rate is determine by Newtons Law of cooling:
T2
q = kA dT/dX,
where q is the heat transfer rate ((Js−1), k is the conductivity [J s−1(°Cm)],
q A is the area of heat transfer (m2) and dT/dX is the temperature gradcent
(°C/m). Therefore q = kA (T1 − T2)/Δx (Figure 17-23).
Figure 17-23 Temperature gradient for The heat flow at steady state is proportional to the temperature gradi-
steady-state heat flow through one slab.
ent and conductivity. As such the analogy with Ohm’s Law for the flow of
electricity is obvious:
i = kV,
where i is the current, V the voltage and k the conductivity. This is true for many transport processes at steady state and
the general equation can be written as
flux = driving force/resistance,
where resistance is the inverse of conductivity. We can thus use the principle of resistance in series (R = R1 + R2 + R3 + ⋅ ⋅ ⋅)
as given by Ohm’s Law to calculate heat fluxes and temperature changes when more than one resistance to heat flow
is involved. We have to substitute the voltage by temperature
difference (driving force), current by heat flux and resistance by
X1 X2 the thickness of the slab divided by its conductivity.
T1 Figure 17-24 shows steady-state heat transfer through two
K1 slabs of thickness Δx1 and Δx2 with conductivities K1 and K2. Now
we determine the heat flux and the interfacial temperature. We
K
have R1 = ΔX1/K1 and R2 = ΔX2/K2. Therefore
2
R = ∆X1 / K 1 + ∆X2 / K 2 ,
(T1 − T2 ) = (q / A)(∆X1 / K 1 + ∆X2 / K 2 );
T2
which is similar to V = iR.
The numerical values can be substituted to determine q/A.
Similarly, since voltage drop is proportional to resistance, we
Figure 17-24 Temperature gradient for steady-state have
heat flow through two slabs.
T1 − Tinterface /(T1 − T2 ) = (∆X1 / K 1) / (∆X1 / K 1 + ∆X2 / K 2 ),
Wall Cooling which can be solved to determine Tinterface.

of coil liquid Let us now look at the various resistances to heat flow when
Fermentor a cooling coil is used to remove heat. Figure 17-25 shows an
Tb expanded view of the wall of a cooling coil, whose outside surface
Two is in contact with the fermentor medium (of temperature Tb) and
inside has cooling liquid (usually water) flowing at temperature
Tc. Note that the temperature of the outside wall surface is Two
which is less than Tb and similarly the temperature of the inside
Twi surface of the wall is Twi which is greater than Tc. This is because
Tc heat flows from the outside to the inside and there is a resistance
to heat flow which shows up in this temperature gradient. We
can postulate the existence of a thin film of liquid of thickness δ
Figure 17-25 Temperature gradient for steady-state across which the temperature drop takes place (from the wall to
heat flow between fermentor and cooling liquid. the bulk liquid) and thus have
797
17.6 Heat Transfer•
q = ko A(Tb − Two ) / d o ,
q = k i A (Twi − Tc )d i ,
where ko and ki are the conductivities of these films (of liquid) and δo and δi are the thicknesses. Since it is impossible to
experimentally measure this film thickness we can couple this with the conductivity of the film and define a film heat
transfer coefficient h given by
ho = ko /d o and h i = k i / d i
there by getting
q = ho A(Tb − Two ),
q = h i A(Twi − Tc ),
q = k A(Two − Twi ) / ∆X,
where k is the conductivity of the wall (typically stainless steel which is the material of construction of the cooling tube)
and ΔX is the thickness of the cooling tube. Adding up the resistances we have
1 / U = 1 / ho + ∆X / k + h i ,
where U is defined as the overall conductivity and h0 and hi are the external and internal ‘film heat transfer coefficients,’
respectively. We can now write the heat balance as
q = U A ∆T = U A(Tb − Tc ),
where q is the heat removed by the cooling coil, A is the surface area of the coil and ΔT is the temperature difference
between the fermentor culture medium and the cooling liquid. The student can try and derive the above expression
using basic algebraic techniques which involve using the three equations given earlier to get rid of the two unknowns,
namely, Two and Twi.
However this temperature difference (Tb − Tc = ΔT) is not a constant since the temperature of the cooling liquid rises
as it flows inside the fermentor (here it picks up the heat generated by the growing culture). This change in the cooling
coil temperature (ΔTc) changes ΔT between the coil and fermentor medium (Figure 17-26). Thus an average value of
ΔT needs to be used while using the above equation.
This average ΔT can be determined if we know the ΔT at the inlet and the outlet of the cooling coil. We have
q = m Cp ∆Tc ,
where m is the mass flow rate of the cooling liquid and Cp is its heat capacity.
If we consider the heat exchange in the small shaded portion in the Figure 17-26, we have
dq = U dA ΔT,
where dA is the area of the shaded part. Also from the temperature increase of the cooling liquid we have
dq = mCp d(ΔT)
since the change in ΔT is because of the rise in temperature of
the cooling liquid. Thus
mCp d(ΔT) = U dA ΔT.
Rearranging and integrating across the whole length of the Tb
coil, we get Temp
Tc
mCp ∫ d(∆T)/ ∆T = U ∫ dA
T
⇒ mCp ln(∆T1 / ∆T2 ) = UA, Tc
where ΔT1 and ΔT2 are the temperature differences at the L

inlet and the outlet of the cooling coil. We also have
Figure 17-26 Temperature gradient showing rise in
ΔTC = ΔT1 − ΔT2, temperature of cooling water and hence change in ΔT.
therefore
q = mCp (ΔT1 − ΔT2).
Substituting for mCp in the above equation we get
q = U A[(ΔT1 − ΔT2)/ ln ΔT1/ΔT2].
The term within brackets is referred to as the log mean temperature difference (LMTD) and the use of this logarith-
mic mean instead of the arithmetic mean of the temperature differences (ΔTavg) gives a more accurate estimate of the
heat transfer rate. Thus
q = UALMTD.
Example 13: Estimate surface area of cooling required for a reactor growing yeast cells at 30 °C, given that
m = 0.2 h −1 , x = 20 g l −1 ; V = 1000 l; hH = 0.6; U = 100 W/m −2 K −1, cooling coil liquid is water which enters at 15 °C and
leaves at 20 °C. Also estimate the flow rate of cooling water in the coil?
Solution: Rate of biomass generation is (0.2 × 20 × 1000)/24.6 = 162.6 C-mole/h.
Now
qmet = 162.6 ∆H X (1 − n H ) / hH = 6 × 104 kJ h −1 .
The total heat generated (qT) can be assumed to be 1.25 times this value (if 80% of the heat generated is due to
metabolic heat):
qT = 1.25 × 6 × 104 kJ h −1 = 20.83 kJs −1 .
LMTD = (∆T1 − ∆T2 ) / ln ∆T1 / ∆T2 ,
where ΔT1 = (30 − 15) = 15 °C and ΔT2 = (30 − 20) = 10 °C (since the reactor temperature is 30 °C and water enters at 15 °C
and leaves at 20 °C).
LMTD = (15 − 10)/ ln(15/10) = 12.33 °C.
(Note that this value is slightly lower than the arithmetic mean of temperature difference 12.5 °C)
We have seen that
q = UALMTD
⇒ 20.83 × 10 (Js ) = 100 W m −2 K × A × 12.33
3 −1
⇒ A = 16.9 m 2 .
The cooling water flow rate can be determined by

q = m Cp ∆T
⇒ 7.5 × 10 (kJ h ) = m (kg h −1)(4.2 (kJ kg −1 ° C −1) × 5(°C)
4 −1
⇒ m = 3571 kg h −1 .
A more accurate analysis would require independently estimating the other heat flows in and out of the fermentor.
17.7 Dimensional Analysis

While data for typical ranges of values of overall heat transfer coefficients (U) are available, often precise design
calculations require estimating values of the individual film heat transfer coefficients (h0 and hi) and then calculating U
from the equation given earlier.
The values of the film heat transfer coefficients are dependent on a large number of parameters. Thus ‘hi’ is a function
of the conductivity of the cooling liquid k, the pipe diameter (D), the liquid viscosity (µ), the heat capacity (Cp), the
liquid velocity (v) and the density (ρ), that is
h = f (K , D, m , Cp , v, r).
799
17.7 Dimensional Analysis•
These functional relationships need to be determined empirically. However, since there is a physical basis to these
relationships we postulate that the form of these relationships require that they be dimensionally homogenous. This
means that the LHS and RHS of this equation have the same units in terms of the four fundamental units, viz. mass
(M), length (L), time (t) and temperature (T). The requirement of dimensional homogeneity imposes constraints on the
nature of the equation that can be formulated and thus reduces the number of constants that need empirical determina-
tion. The strategy outlined below is called the Buckingham Pi method and though we are applying it now to determine
the relationship between h and other variables, it has applicability in a wide range of other cases. We define a power law
relationship
h = Ck a D b m c Cdp v e r f ,
where C is a dimensionless constant and a, b, c, … are the exponents. Note that if there were no constraints on this e quation
we would need to empirically determine the value of the seven constants. The dimensional units of the v ariables are
h = [ Mt −3T −1 ]; k = [ MLt −3 T −1 ]; D = [ L ];
m = [ ML−1t −1 ]; Cp = [ L2t −2 T −1 ]; v = [ Lt −1 ];
r = [ ML−3 ].
The equation in terms of the dimensional units is (C being a constant has no units)
[ Mt −3 T −1 ] = [ MLt −3 T −1 ]a [ L ]b [ ML−1t −1 ]c [ L2t −2 T −1 ]d [ Lt −1 ]e [ ML−3 ]f .
Equating the units of LHS and RHS, we get

for M: 1 = a + c + f; (17.18)
for L : 0 = a + b − c + 2d + e − 3 f ; (17.19)
for t : − 3 = −3a − c − 2d − e; (17.20)
for T : − 1 = − a − d. (17.21)
We have four equation but six unknowns and thus we can express four of the above exponents in terms of the other
two. Retaining the variables ‘d’ and ‘f’ (this makes the job easier), we get
from Eq. (17.21), a = (1 − d);
from Eq. (17.18) c = (1 − a − f ) = (d − f );
from Eq. (17.19) e = f ;
from Eq. 17.20) b = −1+f.

Substituting we can write in terms of d and f only
h = Ck(1− d) D(−1+ f ) m (d − f ) Cp du f r f .
Rearranging the terms we get

(hD / k) = C(Cp m / k)d (ruD / m)f .
Note that this equation is dimensionally homogenous whatever values we choose for d and f. This is possible because
each of the groups within brackets represents a combination of variables which taken together have no units. They
are thus referred to as dimensionless numbers. Thus (hD/k) is called the Nusselt number (Nu), (Cpμ / k) is the Prandtl
number (Pr) and (ρνD/μ) is the Reynolds number (Re). The Buckningham Pi method thus becomes a technique for
generating these dimensionless numbers. Here we have
Nu = C (Pr)α(Re)β,
where the constant C and the values of α and β need to be determined empirically. We have thus reduced the number
of constants that need empirical determination from seven to three. Typical values of C, α and β are available in liter-
ature helping us to correlate the above dimensional numbers. Thus the Dittus Boelter equation for liquid flowing inside
pipes is:
(Nu) = 0.0265 (Re)0.8 (Pr)0.3 .
For (Re) > 10,000 and 0.7 < (Pr) < 170.

A lot of literature is available on such correlations linking up different dimensionless numbers which are applicable
within a range of operating conditions. These correlations can be used to determine the values of different parameters
like film heat transfer coefficients.
17.8 Mass Transfer

Mass transfer deals with the problem of diffusive transfer. Diffusion is controlled by concentration gradients and given
by Fick’s first law:
N A = D(dCA / dx),
where NA is the flux of component A per unit area, D is the diffusivity and (dCA/dx) is the concentration gradient. This
is similar in form to the equation governing heat transfer. We are concerned with mass transfer primarily because cells
growing inside a fermentor need oxygen which has to be supplied by bubbling air/oxygen into the fermentor.
For the transfer of oxygen to the cells, oxygen has to first diffuse through the bubbles into the fermentor medium
and then into the cells from the liquid medium. The equilibrium relationship governing gas solubilities in a liquid is
governed by Henry’s Law:
pA = HcA
where pA is the partial pressure in the gas phase, cA is the solubility (referred to as the dissolved oxygen concentration)
in the liquid phase and H is Henry’s constant. The value of H is a function of pH, temperature, ionic strength and pres-
ence of dissolved solutes and hence varies from medium to medium. When air is bubbled inside liquid (either water or
fermentor medium) so that equilibrium is reached, the dissolved oxygen concentration reaches its saturation value c*
given by p A = Hc A ∗ which is typically in the range of 1.0 ± 0.2 mmoles 1−1.
Let us now consider the situation where oxygen is transferred from bubbles to medium containing growing cells and
simultaneously taken up from the medium by these cells for respiration. A material balance on the liquid phase for
oxygen gives
V (dcL/dt) = OTR − OUR,
where V(dcL/dt) represents the rate of accumulation of oxygen in the liquid,
OTR is the oxygen transfer rate from the bubbles to the liquid and OUR is the
Gas Liquid interface
oxygen uptake rate by the growing cells given by
OUR = µx/YX/O.
Pb
At steady state dcL/dt = 0 and we have OTR = OUR
Note that the dissolved oxygen concentration at steady state given by cL is
lower than c* (the saturation value), thus allowing continuous oxygen transfer
Pi
from the gas to the liquid phase.
Ci To determine OTR we consider the resistance to diffusive flow of oxygen due
to two films, one on the gas side causing the partial pressure of oxygen (which is
CL the driving force) to drop from pb (the bulk gas phase oxygen partial pressure) to
pi (the partial pressure at the interface) (Figure 17-27).
Similarly the film on the liquid side causes the dissolved oxygen concentration
Figure 17-27 Oxygen gradient due to to fall from ci and cL, where ci and cL are the dissolved oxygen concentrations at
diffusion of oxygen from bubbles to cul- the interface and bulk liquid, respectively. (This formulation is similar to the for-
ture medium containing growing cells. mulation for heat transfer.) We can now write the equations for mass transfer as
801
17.8 Mass Transfer•
N = K G A(pb − pi ),
where N is the flux of oxygen, KG is the gas side film mass transfer coefficient, A is the surface area of the bubbles. Also
N = K L A(ci − c L ),
where kL is the liquid-side mass transfer coefficient.

Since the interface has zero thickness it cannot provide any resistance to flow. We consider that pi and ci are in equi-
librium with each other and hence governed by Henry’s law, that is
pi = Hci.
Solving these three equations to get rid of pi and ci, we have N = K L A(pb / H − c L ) , where 1 / K L = 1 / K L + 1 / K G H .
1/KL=1/KL+1/KGH. KL is the overall mass transfer coefficient incorporating both the gas-side and liquid-side mass transfer
coefficients.
We can substitute c* for pb/H where c* is the saturated dissolved oxygen concentration in the liquid phase, in equilib-
rium with oxygen in the gas phase. We need to assume that the bulk gas-phase concentration does not change signifi-
cantly between entry and exit (even though some oxygen gets transferred to the liquid phase). This assumption is not
valid for tall tubular reactors where air enters at the bottom and leaves at the top.
Also since the oxygen flux is measured per liter of reactor volume we define a specific surface area of the gas bubbles
as a = A/V where A is the total surface area of the bubbles and V is the reactor volume. We have
N / V = K L a(c∗ − c L ).
The specific surface area (a) can be estimated if we know the volume of gas held up in the reactor (Vg) and the radius
of the bubbles (r). We therefore have
ε = Vg/VT,
where ε is called the gas hold up ratio. Also the specific surface area of a bubble is given by its area per unit volume, that
is
a = (4pr 2 ) / (4pr 3 / 3) = 3 / r.
Thus the specific surface area per unit reactor volume is given by Vg × a / VT = 3e / r .
[Since we have a bubble size distribution we need to use ravg, a mean radius of the bubbles where the concept of ‘Sauter
mean radius’, rsm, is used (rsm = Σri 3 / Σri 2 ). The student can check that the Sauter mean radius gives an accurate measure
of the specific surface area.]
Correlations are available for KL in terms of the diffusivity D and other variables in the form of dimensionless num-
bers. Similarly correlations for a are also available in literature. However, for practical purposes since both KL and a affect
the OTR, we use correlations which estimate KLa as a single parameter. A typical correlation is
K L a = C (P / V)a (u s )b ,
where C is a constant, P is the power consumed due to agitation, V is the reactor volume and vs is the superficial gas
velocity (given by the gas flow rate divided by the cross-sectional area of the reactor).
The power consumption for agitation can be related in turn to other variables using dimensionless numbers. Thus
P = f (N, D, ρ, g, µ), Laminar Transient Turbulent
where N is the rpm of the agitator, D the diameter of the agitator, ρ
the liquid density, g is gravity and µ is the viscosity of the liquid. The Po
dimensionless numbers relating the above variables are (the students
can derive this from the Buckingham Pi method described earlier)
(P / N 3 D 5 r) = c(rND 2 / m)a (N 2 D / g)b .
The LHS of the equation is the Power number (P0), the term
(ρND2/μ) is the modified Reynolds number for agitated liquids (Re) Re
and (N2D/g) is the Froude number.
Typically ‘baffles’ are used in reactors to break up vortex f ormation. Figure 17-28 Plot of Power number vs. Reynolds
These are thin strips of metal stuck to the inside wall of the reactor number.
(in the perpendicular direction), which break up the circular motion of the liquid and thus cause turbulence. Under
these circumstances the Froude number has little effect on the power consumption. A plot of Power number vs. Reynolds
number is given in Figure 17-28, where three zones can be identified during agitation. These are:
1. Laminar zone: It is the zone where po declines with increasing Re.
2. Transient zone: It is the zone where a complex relationship exist between po and Re.
3. Turbulent zone: It is the zone where po is a constant (and hence independent of Re).
Most fermentors are run in the turbulent zone and therefore the Power number is a constant which is independent of
the Reynolds number. We thus have
P = Po (N 3 D 5 r).
This relationship can also be derived from basic principles if we note that the power consumption is proportional to
the kinetic energy (KE) it imparts to the liquid while mixing. Thus the mass of liquid an agitator blade displaces while
rotating is proportional to ‘ρND3’ (where D3 is proportional to the volume displaced by the agitator blades, so ND3 is the
volume displaced per unit time) and the velocity it imparts to this liquid is ‘ND’ (which is the tip velocity of the agitator
blade). Thus the KE imparted is
1
2
mv2 ∝ rND 3 × (ND)2 = rN 3 D 5 .
Therefore P ∞ ρN3D5 and the proportionality constant is the power number.

The Power number depends on the number and shape of the agitator blades and whether the system is aerated or not
(since the effective density of the liquid culture would fall with aeration and gas hold up).
The above correlations allow us to identify the effect of operating variables like agitator rpm on the power consumption
and OTR.
Scale-up: In the earlier sections we have developed stoichiometric and kinetic relationships which were scale
independent and could thus be applied to large-scale systems after they had been determined in the small scale. In
a similar fashion, parameters like KLa (which determine the OTR), power consumption per unit volume (P/V) and
maximum shear rate (given by the tip velocity of the impeller blades, ND) need to be determined upon scale-up. Since
it is not possible to keep all the above operational variables constant as scale-up is done, only the critical parameters are
considered and kept constant. Thus, if oxygen supply is the critical parameter then KLa would be kept constant but if the
cells are shear sensitive then shear has to be kept constant. Note that geometric similarity is maintained during scale-up
and thus parameters which are length dependent increase as L, those which are area dependent increase as L2 and those
which are volume dependent increase as L3.
Example 14 A reactor has to be scaled up (keeping the OTR constant) from 1 l to 1000 l. How would the power
consumption, rpm and shear rate change? Given are the initial rpm = 800 and aeration rate = 1vvm (volume of air flow
per reactor volume per min).
Solution: Since V2/ V1 = 1000/1 = 1000.
Given the condition of geometric similarity
D2 3 / D13 = 1000 ⇒ D2 / D1 = 10,
if the aeration rate is kept constant per unit volume of reactor (i.e. 1 vvm), then the superficial gas velocity is given by
v = vvm × D3/D2 = D × vvm.
Thus v2/v1=10. This increase is very high and could lead to problems of foaming and flooding of the agitator. Thus the
aeration rate in vvm is typically reduced to 0.5 vvm as the scale increases. Let us assume the following correlation for KLa:
K L a ∝ (P / V)0.7 (v s )0.3 .
Also let the system be running in the turbulent zone and hence
Po = constant and P ∝ r N 3 D 5 .
Thus
K L a ∝ (r N 3 D 5 / D 3 )0.7 (vvmXD)0.3 = N 2.1 D1.7 vvm 0.3 .
803
17.8 Mass Transfer•
If KLa needs to be kept constant on scale-up then

N12.1 D11.7 vvm10.3 = N 22.1 D21.7 vvm 20.3 .
We have D2/D1 = 10; vvm2/vvm1 = 0.5, therefore

N2/N1 = 0.171 and N2 = 137 rpm.
The power consumed per unit vol (P/V) is ∞ N3 D2. Thus N 23 D22 / N13 D12 = 0.5 and the (P/V) value falls to 50% of
its original value while the maximum shear rate (ND) increases by a factor of 1.7. The changes in the various parame-
ters presuppose that the correlations used for KLa and power consumption are valid since different correlations will give
different results.
17.8.1 Experimental Determination of KLa

KLa can be experimentally determined by a variety of methods. The best technique is to do an overall oxygen balance
on the gas phase. If we have mass flow meters to accurately measure the air flow rate and paramagnetic gas analyzers
to estimate the fraction of oxygen entering and leaving the bioreactor, we can do an oxygen balance to estimate OTR
and hence KLa. For this the volumetric air flow rate needs to be converted to NTP in order to determine the molar
flow rate of O2. The only problem with the technique is the need to measure the outlet concentration of O2 precisely
since the concentration difference of O2 between the inlet and outlet is often very low and it is the difference in the O2
concentrations which gives the value of OTR.
Another technique that is used frequently is the ‘dynamic gassing out’ method where the growing cells of the
bioreactor are used to reduce the dissolved oxygen concentration to low levels and the rate of recovery to steady state
is measured. Figure 17-29 shows the profile of the dissolved oxygen concentration CL in a bioreactor with growing cells,
where CL is the steady-state concentration.
Air is switched off at time t = 0 and thus the OTR becomes zero. The dissolved oxygen concentration (CL) falls
according to the following equation:
dCL/dt = −OUR.
The rate of fall is thus equal to the oxygen uptake rate. Air is switched ‘ON’ after some time (to prevent the CL values
becoming so low that the growing cells go into oxygen shock) and the CL value recovers slowly back to its original
steady-state value according to the equation.
dC L /dt=OTR − OUR
⇒ dC L /dt=K L a(C* − C L ) − OUR.
Since at steady state we have

K L a(C* − CL ) = OUR,
we can write
dCL / dt = K L a(C* − CL ) − K L a(C* − CL )
⇒ dCL / dt = K L a(CL − CL ).
Integrating and putting appropriate boundary condi- Air ‘off’

CL
tions we obtain
∫ dC L / (CL − CL ) = ∫ K L a dt. CL
Since at t = 0, air is switched ON and CL = CLminimum,

that is
Air ‘on’
− ln(CL − CL min) / (CL − CL ) = K L a × t,
Time
a plot of ln(CL − CL min) / (CL − CL ) vs. time would
thus gives a straight line of slope KLa. Figure 17-29 Dynamic gasing out method for determination of KLa.
17.9 Measurement and Control of Bioprocess Parameters

Measurement and control of various parameters in a fermentor helps in understanding and optimizing the process.
Significant advances in biophysics, electronics and computer-based data analysis have revolutionized this area. Today
the range, sophistication and wealth of data that is routinely available from fermentation would be unimaginable only
a couple of decades ago. The parameters being measured routinely fall into three main classes: physical, chemical and
biological. Physical parameters include temperature, pressure, reactor weight, liquid level, foam, agitator rpm, power
consumption, air flow rate, feed rate, viscosity and gas hold up. The chemical parameters are dissolved oxygen (DO),
pH, redox potential, dissolved carbon dioxide, exit gas composition, conductivity and culture medium composition like,
glucose, ethanol, acetate, product, etc. The biological parameters include biomass, enzymes, cell composition (DNA,
RNA, protein, ATP/ADP, NAD/NADH levels, etc.), cell viability morphology, etc.
In general the physical parameters are the easiest to measure and these are estimated ‘on-line.’ The chemical
parameters are estimated both ‘on-line’ such as pH, DO exit gas, etc., as well as ‘off-line’ such as substrate and product
concentrations. However, good fermentor control requires that key parameters like glucose, acetate, etc. be measured
‘on-line’ so that corrective action can be taken if the values of these parameters deviate from the optimum required for
the process. Two approaches have been taken in this regard. First, there have been attempts to make sterlizable probes
which can be inserted into the fermentor to measure these variables ‘in-situ’. Second is the design of fast sampling
devices which withdraw small samples at reasonably short intervals to estimate parameters off-line. Discrete analyz-
ers using flow injection techniques like FIA (flow injection analyzer) are now available commercially. These can be
worked on-to off-line probes to generate data quickly enough so as to make them appear almost on-line. Similarly, most
biological parameters are estimated off-line using biosensors which are coupled with FIAs in order to generate quick
results (since they cannot be sterilized and used in situ). Combined with other analytical equipments like HPLC, image
analyzers, NMR, flow cytometry, IR mass spectroscopy and fluorometry today the modern bioreactor can be used to com-
prehensively determine the biochemical and physiological state of the cell.
17.9.1 Feedback Control

In any bioprocess the values of the various parameters need to be maintained at their optimal value. The job of a
controller is to take corrective action in case these values deviate from the optimum value or set point. External dis-
turbances as well as the process of cell growth can bring about changes in the values of parameters; for example, the
temperature would tend to increase because of the metabolic heat generated, the pH may rise or fall depending upon
acid or base production and the dissolved oxygen concentration may change with a change in the oxygen uptake rate.
These changes are measured by appropriate sensors and (for the purpose of control analysis) can be considered as the
output of the system, namely, the bioreactor. The measured values are then compared with the ‘set point’ values and the
deviations, which can be called the ‘error,’ form the input to a controller. The controller response (which is a function
of this error) is then fed back to the system through an actuator so that the deviation is corrected. The actuator may be
the acid/base pump in the case of pH correction or a valve which controls the flow of cooling water in the case of tem-
perature correction or the stirrer motor speed controller when the dissolved oxygen concentration needs to be corrected.
This way of correcting deviations from set point in a system is referred to as feedback control. A schematic is shown in
Figure 17-30.
The controller response (i.e, how much corrective action needs to taken for a certain level of error) is a critical factor
which needs to be formulated with care. The system (in this case the bioreactor) characteristics play the most important
role in designing this response. Take for example a system which responds very slowly to acid/base addition due to poor
mixing. Thus a slight change from set point in pH might trigger the controller to actuate the acid/base pumps. However,
since the system output does not change quickly, too much of acid or base may get added by the time the output reaches
Set point Actuator Disturbance
E
Controller System Output

Measuring
device
Figure 17-30 Schematic of feedback control.
805
17.9 Measurement and Control of Bioprocess Parameters•
the set point value. This over correction (which is called the
‘overshoot’) will then require further corrective action and a Final set
situation may arise where the cycles of overshoots continue point
indefinitely and the pH oscillate continuously around the set C
point. On the other hand, a very slow response of the con-
troller may lead to an unacceptable delay in the correction of B
the error. The possible responses to a step change in the set
A
point are given in Figure 17-31. Initial set
It should be noted that a slight ‘overshoot’ while correct- point
ing the error is usually acceptable and forms a basic feature of
the response curve. However, the oscillations should lessen
in amplitude quickly enough (a quarter rate decay is usu- Time
ally acceptable, i.e. the amplitude of the oscillations should
decline by a quarter in each cycle). Figure 17-31 Response of a system with feedback control
to a step change in the set point.
17.9.2 Controller Characteristics
As stated earlier the controller response is a function of the error (ε). Typically the response is given by
Response = K c e + 1/t I ∫ edt + t D de / dt.
The first term Kcε represents the proportional component of the response, that is the response is proportional to the
magnitude of the error. When only this term is present the controller is called a proportional controller. It should be
noted that this proportional response cannot increase indefinitely with increasing ε since it would soon reach its upper
limit (of 100% response). For higher values of ε the response thus saturates at its upper limit. Therefore the response is
proportional only for a range of ε values. This range is called the proportional band (PB). We can relate Kε to PB by the
following equation.
Kc = 100%/PB.
Thus a higher value of Kc (which is also referred to as the controller gain) implies a narrow proportional band.
The problems associated with having only a proportional control is that the controller response declines as the ε
value falls thereby leading to ‘offset’ error which is the permanent deviation from set point. To understand this consider
a simple case of a growing culture producing acids which lower the pH. The proportional control would correct for this
fall in pH by adding base at a rate proportional to the deviation from the set point. However, when this deviation is low,
the rate of base addition would fall till it matches the acid production rate leading to a permanent offset error in the set
point pH. A solution to this lies in increasing the proportional ‘gain’ which can, however, lead to an oscillatory response.
To counter this, integral response is introduced where the response is dependent on the integrated sum of errors over
time. This removes both offset error and reduces the possibility of oscillatory response. The third component of the
response is the derivative response which allows a faster response time. However random noise or any error having sharp
fluctuations can cause an inappropriate response and hence derivative control needs to be used with care.
Taken together they form the proportional-integral-derivative (PID) response which is a standard feature of most
controllers today. The values of Kc, τI and τD need to be finetuned so as to get the best response and these are dependent
on the characteristics of the system being controlled. Setting these values is referred to as controller tuning.
17.9.2.1 Controller Tuning

Most of the time the control loops are tuned by the manufacturer on the basis of experience. However, one may
attempt tuning if the process in radically different from standard processes or if customization leads to better response
characteristics. Two of the simplest methods for tuning are given below.
Ultimate gain method: In this method the system is run using only the proportional control and the gain is increased
slowly till steady oscillations are observed. This value of the gain is called the ultimate gain Ku and the time period of
the oscillation is called τu. The controller setting are chosen from these values as
K c = 0.6 K u ; t I = 0.5 t u and t D = 0.125 t u .
These values give the standard quarter decay response; however, lower values may be chosen if a more conservative
response is required.
Slope = S
Step
Variable input Output
t=0 TD
t Figure 17-32 Process reaction curve of a system.
Process reaction curve: In this method the response of the system to a small step input ‘m’ is measured in the absence
of any control. A typical response is shown in Figure 17-32.
A tangent is drawn at the point of inflexion of the response curve. The point where this tangent cuts the x-axis is
used to determine the value of TD, the delay time in the response. The normalized slope is given by S, where S is the
slope of the tangent/m.
The values of TD and S are used to determine the controller characteristics where
K c = 1.2 / TD S; t I = 2 TD ; t D = 0.5 TD .
The process reaction curve may also be used to model the system response by considering it equivalent to a delay time
response plus a first order response (see Box 17-1)in which case the system response to various controller settings can be
predicted for different kinds of errors in a quantitative fashion. From these an appropriate choice of controller settings
can easily be made.
Major advances have taken in recent years in process control techniques. Thus ‘adaptive’ control is now used where
the controller settings are changed as the process changes. For example, as the culture grows from low to high cell den-
sity the system characteristics with respect to fluctuations in the dissolved oxygen concentration change. Hence the
controller settings are tuned in real time.
Feed-forward control is another method often used to counter the effect of large delay times. If good models of the
process are available then the possible effect of input errors can be estimated and corrective action initiated even before
the system output shows the error. Controller training can also be done using complex computer techniques like artificial
neural networks or support vector machines.
Box 17-1: System Response

Consider an ideal plug flow reaction (PFR) where liquid moves in a long tube without axial mixing. A step change in the
input of any variable, say concentration, would show up in the output as a step change only after a certain delay time which
is equivalent to the residence time of the liquid inside the reactor. A PFR is thus an ideal representation of a delayed time
response. System with high delay or ‘dead’ time are usually the most difficult to control since the controller gets the error
message very late and corrective action is initiated well after the initial error. Now consider a CSTR where the input feed
concentration undergoes a step change. The output here increases in an inverse exponential fashion given by the equation
(assuming no reaction term)
C = C0 (1 − e − Dt ),
where the initial concentration at time t = 0 is zero and the step input is C0. Such a response is called a first-order response.
If two CSTRs are connected in series, the output would be a second-order response. The student can check that the typical
process reaction curve given in the earlier section can be simulated by the combination of a time delay response followed
by a first-order response.
807
17.10 Sterilization•
17.10 Sterilization
Sterlization is the process of getting rid of contaminating agents like bacteria, viruses, etc. from the system so that we
can use it for growing the desired microorganisms. Thus the bioreactor vessel and medium need to be sterilized. Also
all inputs to the system like air, media feed, acid/base need to be sterile and aseptic operating conditions need to be
maintained. The inoculum needs to be free of contaminating agents aswell.
Air sterilization is a fairly straightforward process nowadays because of the availability of filters which retain all the
contaminating agents with 100% efficiency. This is possible because polymers can now be designed (usually PTFE-
based) where the pore size can be controlled very accurately. Similarly the reactor vessel can be sterilized with steam at
high pressure (121 °C for 20 min). Media sterilization, however, poses problems because the constituents undergo degra-
dation during the sterilization process and hence the trade off between nutrient quality and degree of sterilization needs
to be well-understood and optimized.
17.10.1 Kinetics of Cell Death

The rate of ‘kill’ for any bacteria is a first-order reaction given by
dN/dt = kN,
where N is the number of viable bacteria and k is a constant dependent on bacterial type and temperature. Design of
sterilization is done assuming that all the bacteria present have a kill rate equal to Bacillus stearothermophilus spores (since
these are the most heat resistant) and this provides the necessary safety margin in the design. For B. stearothermophilus
spores we can write
k = A e−ΔE/RT,
which is the Arnehenius form of dependence of the first-order rate constant on temperature. The values of A and ΔE are
1 × 1036.2 s−1 and 67 kcal g−1mole−1 respectively. However, these rate constants are dependent on medium composition.
Given an initial cell count in the medium as N0 and the final desired cell count Nf , we have
∫ dN / N = ∫ −kdt.
⇒ ln(N 0 / N f ) = −kt.
The term ln N0/Nf is called the ‘del’ factor denoted by the symbol ‘∇.’ The ∇ value represents the required
degree of sterilization. For example, if the initial cell count is 103 cells ml−1 and the media volume is 1000 l then
N0 = 103 × 103 × 103 = 109 cells. Nf is chosen typically as 0.001 which implies that the probability of 1 cell surviving
post-sterilization is 1/1000 (or that 1 fermentor in a 1000 would not be completely sterile because 1 cell would survive
the sterilization process). Then ∇ = ln 109 / 10 −3 = ln 1012 = 27.6 . To achieve this desired ∇ value a time–temperature
regimen is required, which is given by
∇ = Ae−ΔE/RTt.
If the temperature is a constant, the time can be easily determined (since a constant temperature implies a constant
value of k). However, media which is usually sterilized in holding tanks (or the fermentor vessel itself) usually has a
heating phase (where the temperature rises), a holding phase (where the temperature is kept constant) and a cooling
phase (where the temperature is brought down to normal). Typically since the fermentor or the holding vessel cannot
withstand high pressures (> latm gage) the holding temperature is 121 °C. The overall ∇ is thus
∇total = ∇heating+ ∇holding + ∇cooling.
If the temperature profiles for heating and cooling phase are available, the above equation can be integrated to obtain
the heating and cooling values. Numerical integration can also be done by calculating the ‘k’ values for short time
intervals where the temperature can be assumed to be a constant and then adding the values. Thus
Σ∇ = ΣkΔt.
A simplified calculation protocol can be used if one assumes linear heating and cooling rates and also that significant
contributions to the ∇ value are obtained only above 100 °C.
Table 17-1 gives the ‘k’ values for B. stearothermophilus spores for different temperatures above 100 °C. Assuming a
1 °C/min−1 rise/fall in temperature, the ∇ values for different temperatures are also given. Thus at T = 115 °C, the ∇
value has been calculated by assuming that the temperature rose from 100 °C to 115 °C in 15 min.
Table 17-1 The ‘k’ values for B. Stearothermophilus spores for different temperatures
T(°C) k (min−1) ∇heating/cooling
100 0.019 —
101 0.025 0.044
102 0.032 0.076
103 0.040 0.116
104 0.051 0.168
105 0.065 0.233
106 0.083 0.316
107 0.105 0.420
108 0.133 0.553
109 0.168 0.720
110 0.212 0.932
111 0.267 1.199
112 0.336 1.535
113 0.423 1.957
114 0.531 2.488
115 0.666 3.154
116 0.835 3.989
117 1.045 5.034
118 1.307 6.341
119 1.633 7.973
120 2.037 10.010
121 2.538 12.549
122 3.160 15.708
123 3.929 19.638
124 4.881 24.518
125 6.056 30.574
126 7.506 38.080
127 9.293 47.373
128 11.494 58.867
129 14.200 73.067
130 17.524 90.591
An example of how to calculate sterilization time by the simple method is given below.
Example 15: The initial cell count presterilization is 104 cells ml−1 and the media volume is 10,000 l. The heating from
100 °C to 121 °C (the holding temperature) takes 15 min and the cooling down to 100 °C takes 20 min. Calculate the
holding time.
Solution: We have
∇total = ln 104 × 104 × 10 3 / 10 −3 = ln 1014 = 32.2.
If the heating was at a rate of 1 °C min−1, then from 100°C to 121 °C, ∇heating = 12.549 (Table 17-1).
However, since the heating took only 15 min we have
∇heating = 12.549 × 15/21 = 8.96.
Similarly
∇cooling = 12.549 × 20 / 21 = 11.95.
We know that
∇total = 32.2 = ∇heating + ∇holding + ∇cooling.
Therefore
∇holding = 11.29.
Since k at 121 °C is equal to 2.538 (Table 17-1) the holding time for adequate sterilization is given by ∇holding = kt
or
809
17.11 Media Design•
t = ∇holding/k = 4.44min.
Note that the heating and cooling phase contribute significantly to the overall sterilization process. Thus, in order to
avoid oversterlization (and hence excess nutrient degradation) we need to include the effects of heating and cooling as
well. Continuous sterilizers are used where the culture medium flows through a heat exchangers which rapidly increases
its temperature to high values for a short period. This high-temperature–short-time-period regimen reduces nutrient
degradability. To understand this we need to look at the basic sterilization equation:
∇ = Ae − ∆E / RT t
⇒ ln ∇ = ln A − ∆E / RT + ln t.
If we consider a constant degree of sterilization (∇ = constant), a plot of ln t vs. I/T gives a straight line of slope ΔE/R
(Figure 17-33).
Points on this line represent different time–temperature regimens which give the same degree of sterility. Lines drawn
parallel and above this line would represent larger values of ∇.
Now consider nutrient degradation (especially of thermolabile components in the media) which is also a first-order
relationship
ln (x0/xf) = kt,
where x0 is the concentration of these nutrients in the initial medium and xf is the final concentration after sterilization.
The first-order rate constant ‘k’ also has an Arrhenius dependence on temperature but the ΔE values in this case are
much lower (20–30 kcal g−1 mole−1). Thus a plot of ln t vs. I/T for constant degree of nutrient degradation would have
a lesser slope than that for constant ∇ (Figure 17-15). Lines parallel and below represent lesser degree of nutrient deg-
radation. From the figure we can deduce that choosing short time intervals and high temperature allows to reduce the
degradation and hence improve nutrient quality while keeping ∇ values constant.
17.11 Media Design

All microorganisms require water, energy source, carbon, nitrogen, minerals and possibly vitamins for growth.
While defined media containing pure compounds are useful for research so as to calculate yield coefficients, com-
plex media are usually used for large-scale industrial fermentations. A well-designed media has to meet many
important criteria like high product and biomass yields, high concentrations of product, high rates of product
formation, low levels of u ndesirable by products, ease of sterilization, waste disposal and most importantly product
purification.
Depending on process economics, the relative importance of the above criteria changes drastically. Thus if media
cost is the major manufacturing cost then high yields are important; however, if the capital cost (in terms of fermentor
size) is critical then rates of product formation become important. If purification constitutes the major expense then
the concentration of the product is critical. Also for an extracellular product, a clean media may help downstream
processing.
Complex natural materials have traditionally been Increasing
used for fermentation since they are of low cost. However
the batch-to-batch variations often cause variations in
product quality which may be critical depending upon the
intended end-use of the product. Thus, while the tradi-
tional fermentation industry like brewing still uses natural 1/ T
materials, many modern biotech firms especially those in Increasing
manufacture of human therapeutics prefer to use defined deactivation
or semi-defined medium.
17.11.1 Medium Formulation

The formulated media must satisfy the elemental require-
ments of biomass and product formation. In addition, it
should supply the energy for growth and maintenance.
Quantitative analysis of the above requirements allows In t
stoichiometric design of media. Thus waste product for- Figure 17-33 Time and temperature profiles for constant
mation should be minimized as much as possible. This is degree of sterilization and deactivation.
not always possible with complex media since each ingredient contains a range of elements. An approximate analysis of
commonly used media ingredients like molasses (which is used as a carbon source) or corn steep liquor (used as a nitro-
gen source) is available in literature. Since the carbon source usually serves the dual purpose of supplying carbon as well
as energy to the cell, the material balance needs to be formulated with care.
Typical carbon source are: cerelose (commercial glucose), glycerol, cane or beet molasses, oils (soyabean, corn and
cottonseed), corn starch, dextrins, whey (65% lactose) and alcohols (e.g. methanol).
Typical nitrogen sources are: Corn steep liquor, soyabean meal or flour, dried distillers solubles, yeast extract, fish
meal, peanut meal, etc.
Other than carbon and nitrogen, cells need minerals, trace elements and vitamins for growth. Additionally in
fermentors we need to add buffers (for pH control) precursors, inhibitors or inducers to regulate the metabolism of the
cells and antifoams to reduce foaming. The design of the final media thus becomes an exercise of juggling with many
variables each of which may act independently or in concert to effect the fermentor performance. The traditional
method of optimizing media, where one variable is varied and the others kept constant, is not very useful in these
circumstances. These have been replaced by statistical techniques which allows the simultaneous changing of many
parameters (usually the concentration of media ingredients) in order to get the best media design.
The testing of all possible combinations of media concentrations is called the complete factorial design. Since this
would lead to an unacceptably large number of experiments, only certain critical combinations are tested in what is
called a fractional factorial design. Two such fractional factorial designs are described below.
17.11.1.1 Packett–Burman Design

Here we choose the number of variables to be studied as (N−1), where N is a multiple of 4, and perform N experiments
(each media ingredient is a variable). These experiments helps us to determine the key variables, that is the key ingre-
dients whose concentration has the maximum role in product formation. If we do not have (N−1) variables then we
introduce dummy variables which help us to calculate the statistical variation in the data. Thus if we have 9 variables
we introduce 2 dummy variables to get a total of 11. Each variable is studied at two concentrations, namely, a high (H)
or a low (L) value. The H and L values are chosen so that they cause a significant change in the output, something
which needs to be decided by previous experience. For the dummy variables the H and L values are kept the same, thus
ideally the output should not be affected by these dummy variables. The design of the N experiments for N = 12 is given
in Table 17-2.
The principle for the design is that each variable is chosen as H for half the experiments and L for the other half.
Also, for each H value of one variable the next variable is chosen as H in one case and L in the other. Thus if we have
H values of A in six cases then B would be H in three cases and L in the other three. This effect would cancel out any
effect of B on A. Similar for the six experiments, where A is low B would be H in three cases and L in the other three.
This logic is followed for all the variables.
Table 17-2 Plackett-Burman design for 12 experimental trials

Assigned variable Dummy variables
Trial Random ordera
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11
1 H H L H H H L L L H L
2 L H H L H H H L L L H
3 H L H H L H H H L L L
4 L H L H H L H H H L L
5 L L H L H H L H H H L
6 L L L H L H H L H H H
7 H L L L H L H H L H H
8 H H L L L H L H H L H
9 H H H L L L H L H H L
10 L H H H L L L H L H H
11 H L H H H L L L H L H
12 L L L L L L L L L L L
a
Use random number table.
Note: H, high level; L, low level.
811
17.12 Isolation and Preservation of Industrial Microorganisms•
The results of the experiments for which A was low (L) are averaged together and subtracted from the average of the
results of the experiment when A was high (H) to get the effect of A:
Effect of A = (ΣH A − ΣL A ) / 6.
Similarly the effect of B is

Effect of B = (ΣH B − ΣL B ) / 6.
If the statistical variation were to be negligible, the effect of the dummy variables should be zero. Therefore, the
average effect of the dummy variables is taken as a measure of the statistical error. An ‘F’ test which compares the square
of the effect of the concerned variable with that of the dummy variables gives us a relative index which defines the
importance of each variable. The variables with the highest F values can now be defined as the key variables and chosen
for further study.
17.11.1.2 Box Wilson Design

This is also a fractional factorial design where each variable is tested at five levels (i.e. five concentrations). Thus if we
label the minimum and maximum concentrations as −k and +k, respectively, we can label the midpoint concentration
as zero. The two intermediate concentrations between +k and 0 and 0 and −k are labeled +1 and −1, respectively, and
called the factorial points. The value of the factorial points depends on the number of variables being studied and is
given by +k / n and −k / n , when n is the number of variables being studied. Consider the situation where three
variables are being tested, thus the factorial points would be +k / 3 and −k / 3 . The experiments are performed in
the following combinations:
Center point: (0, 0, 0) usually performed in duplicate or triplicate to get a sense of the statistical variation.
Axial points: (+k, 0, 0); (−k, 0, 0); (0, +k, 0); (0, −k, 0); (0, 0, +k); (0, 0, −k).
Factorial points: (1, 1, 1); (1, 1, −1); (1, −1, 1); (1, −1, −1); (−1, 1, 1); (−1, 1, −1); (−1, −1, 1); (−1, −1, −1).
Thus we have a total of 15 points and hence 15 experiments to be performed instead than 53(125) points which
would have represented the full factorial design.
If experiments are to be performed one at a time then pattern search methods can be adopted to determine the next
point at which the experiment needs to be done. This is usually a reflection of the point which gave the poorest results
(e.g. the simplex search algorithm).
17.12 Isolation and Preservation of Industrial Microorganisms

The existence of a large number of type culture collections usually precludes the need to search for new organisms. These
include the ATCC (American Type Culture Collection), various British and European Type Culture Collections as well
as the MTCC (Microbial Type Culture Collection) at IMTECH Chandigarh and the NTCC (National Type Culture
Collection) at Pune.
Ordering microorganisms from collection centers saves the time and expense involved in isolating and characterizing
a new organism from the environment. It is also very likely, given the fairly extensive nature of preexisting collections,
that one may end up isolating an organism which already exists in these type collections which defeats the entire
purpose of the exercise. On the other hand, a ‘new’ organism may provide novel advantages including the intellectual
property rights which may be critical in the process of commercialization. Isolation is a three-step process involving
collection of a pool of microorganisms usually from an ecologically favorable niche, followed by enrichment of the pool
for the desired type, followed by final screening of individual microbes.
Isolation from harsh environmental conditions has resulted in a fairly good collection of extremeophiles, microor-
ganisms which can withstand extremes of pH, temperature or pressure. These find potential use in the chemical industry
which often uses ‘harsh’ environmental conditions not suitable for the growth of normal microorganisms. However
‘new’ and ‘novel’ organisms may well be found in the normal ecological niches since microenvironments often vary
significantly from each other. The enrichment protocol is a critical step because it involves retaining and enriching
the isolated pool thereby simplifying the extremely tedious process of individual screening. Conditions which favor the
growth of the desired type of microorganism and inhibit the growth of undesirable types are often used. Knowledge of
the desired microbial taxa is often useful in designing media to provide selection pressure. To maintain this selection
pressure repeated subculturing is done. The time of subculturing (whether early or late) may lead to different kinds of
enrichment since nutrient starvation may lead to the dominance of a different kind of microorganism. Continuous
B culture techniques have found favor since the dilution rate

can be adjusted to get enrichment of different types. To illus-
trate consider two organisms A and B having growth kinetics
A as shown in Figure 17-34.
At low dilution rates, organism A would be selected because
of its higher specific growth rate at low substrate concentrations
while at high dilution rates organism B would be selected since
it has a higher µm. ‘A’ has low Ks values and would be useful
when the substrate needs to be utilized efficiently, while ‘B’ has
a higher µm and could give better productivities. Continuous
cultures also help in selecting mixed cultures of microorgan-
ism which because of their interaction may work better than
pure culture systems. This is especially true for waste treatment
S
where a combination of methanogens and acidogens is required
Figure 17-34 Growth kinetics of two organisms having to degrade the various constituents of waste to methane (which
different mm and Ks values. is a useful by-product). Useful properties like resistance to con-
tamination may be obtained by periodically adding ‘sludge’ to
the running CSTR. Simple techniques like pasteurization or filtration using glass wool would help in selectively enriching
spore-forming bacteria or fungi.
Developments in robotics as well as in modern biology techniques like ELISA and/or the development of genetic
probes have simplified and automated the screening process allowing increased sample handling capacities. This, in turn,
has significantly enhanced the probabilities of isolating and identifying novel microorganisms. The creative design of
screens is possibly the most important development in this regard. Designs to look for inhibitors of key enzymes involved
in human metabolism have helped in the development of new therapeutic drugs.
17.12.1 Preservation of Microorganisms

Preservation is important not only for storage and later use but also because the process of repeated subculturing carries
with it the risk of contamination and genetic alterations (genetic instability). The more the cycles of growth and division,
the higher is the probability of emergence of mutations. These are often revertant mutants which revert the desirable
mutations or genetic modification which have been carried out to enhance the productivity of the microorganism.
Preservation techniques follow two basic principles: storage at low temperature and storage by removing water from
the system.
1. Storage at low temperature: Cells can routinely be stored for 2–4 weeks at 4°C (in a refrigerator) on slants or p etriplates.
For longer time periods lower temperatures like −20 °C or −70 °C are required which necessitate the addition of
cryopreservatives like glycerol. A tradeoff exists between the percentage viability of the culture (due to temperature
shock) and the period of storage. The lower the temperature, the longer the cells can be stored without subsculturing
but the difficulty in reviving them in correspondingly higher (due to poor viability). Storage under liquid nitrogen
(−193 °C) requires special facilities but cells can be stored indefinitely in this stage. The procedure involves quick
freezing of the cells to the desired temperature of storage but slow thawing to room temperature. Master cell banks
which rarely need to be used are stored using such techniques.
2. Storage using moisture removal: Sporulating cells can be easily stored once spores are formed. However the most
common method of desiccating cells is lypholization which involves first lowering the temperature to form ice and
then sublimating this ice to vapour at low pressure. This process of sublimation leaves the cell structure intact and
cell viability is not lost. Care must be taken to not allow air or moisture to leak back into the system. Though this
process is more tedious, once lypholized, cells can be kept indefinitely without requiring any special facilities for
storage.
17.13 Downstream Processing

One of the most important components in any biotechnology setting is the capacity for separation of biological materials
from one another. Whenever desired biological products are produced, be it in a small-scale laboratory setting or large-
scale industrial fermentation, they have to be isolated and purified. In industrial terminology, production methods
are characterized as upstream or downstream processes. Upstream processes, like fermentation or cell culture, produce
the starting material of interest. Downstream processes are separation procedures that result in a particular product.
813
17.13 Downstream Processing•
Therefore, for commercial purposes, as well as for further study and analysis, a series of separation techniques are exe-
cuted in succession to purify a product from the myriad of possible impurities. A fermentation broth typically consists of
a complex aqueous mixture of cells, soluble extracellular products, soluble/insoluble intracellular products and uncon-
verted substrates to name a few. When solutions carrying particulate substances, having densities greater than the
solutions, are allowed to stand for long, the particles settle down under the action of gravity with a velocity which is a
complex function of its size and density.
There are many different types of separation methods used for biological materials, based on their molecular p roperties,
differences in size, charge, relative solubility, affinity for a particular chemical, etc. This section introduces the basic fea-
tures of downstream processing, the methods of bioseparation following an industrial bioprocess. The strategy used for
any particular case depends on a lot of factors like (i) broth characteristics (viscosity, product concentration, impurities,
etc.); (ii) final product purity and concentration and (iii) the desired form of the product (crystallized product, dried
powder, concentrated liquid, etc.)
17.13.1 Removal of Microbial Cells and Solid Matter

The first step is to separate the liquid broth containing soluble components from the solid matter which includes living
cells, dead cells and cell debris along with insoluble particles (if any). The dimensions of a typical cell (bacteria, etc.) can
vary from 0.3 to 3µm, whereas eukaryotic cells (yeast, fungi, mammalian cells, etc.) can have dimensions upto 20 µm.
Consequently, the mass of a single prokaryotic cell could be as little as 10−12g while a eukaryotic cell can have a mass which
is 1000 times greater. Depending on the nature of the microbial culture (size, density, etc.), different strategies are adopted
to enable solid–liquid separation. Sometimes because of the physical characteristics of broth (high viscosity, gelatinous
broth material, particles with small density difference with water, etc.) such separations become difficult. Nevertheless,
methods like sedimentation, filtration and centrifugation are routinely used for all separations. Sometimes a broth pre-
treatment like heat treatment, pH treatment and addition of chemicals which aid cell flocculation (poly-electrolytes,
calcium chloride, colloidal clay, etc.) is given to render the broth amenable to separation.
17.13.2 Characterization of Fermentation Broth

The rheological behavior of fermentation broths often affect the performance of filters and centrifuges. When cells
are separated from the broth, the supernatant behaves generally like water. However, the presence of cells suspended
in it can alter the viscosity and the extent of alteration is influenced by cell shape and cell adhesion. In order to
understand the alteration we need to note that on the basis of flow behavior liquids are classified as (i) Newtonian and
(ii) non-Newtonian.
17.13.2.1 Newtonian Liquids

Liquids which have a constant viscosity (independent of the shear force acting on them) are classified as Newtonian
fluids. The flow behavior of such liquids (e.g. water, oil) can be described by the Newtonian equation as
τ = η dv/dx,
where τ is shear stress (N m−2), η is the dynamic viscosity (N s m−2 or Pa), dv/dx = shear rate (s−1). Thus, the shear stress
is directly proportional to the shear rate and the viscosity (η) is the constant of proportionality.
For cells (or rigid particles) suspended in liquid, Einstein formulated (for dilute solutions) the following equation to
determine the change in the viscosity:
hrelative = (hsuspension / hliquid ) = 1 + K L Φ,
where KL is a constant and ϕ is the volume fraction of particles. KL depends on particle shape, thus for spherical particles
KL = 2.5 and for rod-shaped particles KL varies from 20–1200.
For more concentrated solutions, Eilers derived the following relationship
hrelative = [1 + K L Φ × 0.5 (1 − Φ / Φ max )−1 ]2 ,
where Φmax is the maximum packing capacity (0.6–0.7) for the type of cells involved.
Example 16: Determine the shear stress for a suspension of spherical cells in water with a cell volume of 0.2 and a max-
imum packing density of 0.6 when the shear rate is 60 s−1.
Solution: Since the cell volume is 0.2, we use Eilers equation to determine relative viscosity (ηrelative), in which, KL = 2.5
as the cells are spherical. With Φ = 0.2 and Φmax = 0.6, we get ηrelative = 1.89. Since the viscosity of water is 0.001 (Nm−2s)
at 20 °C we have ηsuspended = 0.00189. Substituting in the earlier equation, at dv/dx = 60s−1, we get t = 0.1134 Nm −2 .
17.13.2.2 Non-Newtonian Liquids

In this case the viscosity of the liquid varies with the shear stress; thus the shear stress is a nonlinear function of shear
rate:
t = K f (dv / dx).
This sometimes takes a power law form:

t = K (dv / dx)n
where K is the consistency index (Nm −2 s n ) and n is the power law index. Both these constants can be measured by a
rheometer typically a Brookfields viscometer which involves measuring the torque required to rotate two concentric
cylinders at varying rates when the annulus is filled with the liquid whose viscosity needs to be measured. When η < 1
the liquid is referred to as a pseudo-plastic while dilatant fluids have η > 1.
Many fermentation broths follow the Bingham plastic model given by
t = t 0 + h(dv / dx),
where τ0 is the minimum stress required to cause liquid flow.

A combination of the above equations gives
t = t 0 + K (dv / dx)n .
This is called the Herschel–Buckley model (Figure 17-35).

Example 17: Which of the following is most likely to deviate from Newtonian rheology: A culture of unicellular bacte-
ria, filamentous fungi, unicellular algae, protozoan?
Solution: Filamentous fungi. The filamentous nature of the culture will influence the flow behavior of the broth.
Generally such cultures follow a power law where n is less than one.
17.13.3 Sedimentation
Sedimentation is the settling of substances in a simple gravitational field. (The settling achieved is enhanced by
centrifugal force, and in these cases faster rates of settling are obtained.) At a larger-scale sedimentation can be done
when cells have a tendency to adhere closely (coagulate) or to form multicelled aggregates (sometimes aided by pre-
treatment). Many flocculent yeast strains, for example, are in use in breweries and single-cell-protein production. In cell
stic
pla
am
Bingh
tic
p las
do
eu
Ps ian
ton
N ew
t
la tan
Di
Figure 17-35 Herschel–Buckley model
relating shear stress to shear rate for
dv/dx Newtonian and non-Newtonian liquids.
815
suspension settlers, the rate of setting declines with increasing solid concentration. The interface between the cell-free
and the cell-bearing liquid moves downward according to the equation:
U s = kc − m ,
where Us is the setting velocity, c is the cell concentration, k is a constant and m takes a value between 1.7 and 2.6. This
is called hindered settling. (The principles governing free settling are given in the next section.)
Sedimentation finds use in sludge waste treatment for removal of water. Solid wastes are converted to cell-sludge by
biological treatment. This suspension can be treated by sedimentation to separate the free water and get concentrated
sludge which can be used as fertilizer.
17.13.4 Centrifugation
A centrifuge is a piece of equipment that accelerates the rate of sedimentation by rapidly spinning the samples, thus
creating a centrifugal force many times that of gravity. The force F, acting on a particle of volume Vp and density rp
located at a distance r from a point about which it is revolving with angular velocity ω is the difference between the
centrifugal force, mω2r, and the buoyant force (Vp ρω2r) exerted by the solution. That is F = mw 2 r − Vp rw 2 r, where Vp is
the particle volume and ρ the density of the solution. Since m = VPρp and the drag force can be expressed by Stokes Law:
Drag force = 6phRVr .
From the above calculation we get that

6phRVr = Vpw 2 r(rp − r),
where η is the visocity of the fluid medium and Vr is the terminal particle velocity in the r direction and R is the radius
of the particle.
Integrating the above equation allows us to determine the time (T) required for a particle to move from a distance r1
from the center of rotation to r2. Thus
T = 9 / 2h / w 2 R 2 (rp − r) ln r2 / r1 .
Different particles will therefore take different times to settle through the same distance. A particle may be
c haracterized by its sedimentation rate. Every particle can have a sedimentation coefficient, s = V/ω2r, where V is the
terminal settling velocity and ω2r is the centrifugal force. This coefficient depends on the shape, size and density of the
particle and also on the viscosity of the liquid. This is the basis for differential centrifugation, separating different sized
particles using different centrifugal forces and different times.
Centrifuges can be of many types (serving different purposes) depending on their internal features (solids bowl, solids
effecting, nozzle, scroll, etc.) Preparative ultracentrifuges can be used in two distinct ways:
1. Zonal centrifugation in which particles are separated according to their sedimentation coefficients.
2. Equilibrium density gradient centrifugation where particles are separated according to their densities.
17.13.5 Filtration
This is the most common method of separating solid particles in suspension (like whole cells) from liquid media.
Separation is accomplished by forcing the liquid through a porous membrane where the solid particles are trapped. These
particles build up as a layer on the surface of this membrane called a ‘cake’ while the clear liquid goes through. To attain
high throughput rates the pressure drop for flow may be increased or the resistance to flow may be decreased. Most indus-
trial equipment decreases the resistance by increasing the filtration area. This is because increasing the pressure of liquid
tends to compact the ‘filter cake’ especially if it is made up of compressible material like cells. This in turn increases the
cake resistance and hence reduces the flow. In batch filters as more liquid flows through, the solids deposited on the filter
membrane increase thereby increasing the cake thickness. This in turn reduces the flow due to increased resistance of
the filter cake and the filtration has to be stopped after some time so that this cake can be removed from the filter. The
length of the batch cycle depends on the liquid flow rate, the pressure used, the amount of solids in the feed, etc. Filter
aids are often used to prevent blocking of the filter membrane by the solids, and also help in the formation of a porous
filter cake, which lowers the cake resistance and increases the length of the batch cycle.
17.13.5.1 Filtration Calculations

The flow rate of liquid through a filter decreases with time as the resistance to flow increases.
We define Vs = 1 / A (dV / dt) When Vs is the flow velocity, A is the surface area of filtration and dV/dt is the flow
rate. We have
∆ PahVs L
where ΔP is the pressure drop, η is the viscosity and L is the cake thickness. The proportionality constant would depend
upon the porosity and particle size of the cake. The thickness of the cake (L) can be related to the volume of filtrate by
a material balance
LA(1−€)ρs = wV,
where € is the porosity, ρs is the density of solids in the cake, w is weight fraction of solids in the feed and V is the volume
of filtrate. Substituting we get
∆P = Kη(1/A dV/dt) wV/[A(1−€)ρs].
Rearranging and putting α = K/(1 − €)ρs, we get
1 / A dV / dt = ∆P / (ha w V / A),
where α is called the specific cake resistance.
Often the resistance of the filter medium (the membrane) needs to be incorporated into the calculations. For conve-
nience, this resistance is expressed in terms of an equivalent volume of filtrate:
(1 / A)dV / dt = ∆P / [h a w (V + Ve ) / A ].
Here Ve is the volume of filtrate necessary to build up a fictitious filter cake the resistance of which is equal to that of
the membrane.
Rearranging we get
(V + Ve )dV = ∆PA 2 dt / h a w.
Integrating this equation we get

t = h a w(V 2 / 2 + VVe ) / A 2 ∆P,
which allows us to calculate the time required to collect a certain volume of filtrate. Typically, however, we need to
calculate α and Ve from experimental runs. For this we write
dt / dV = h a w (V + Ve )/ A 2 ∆P.
This gives a straight line plot between (dt/dV) and V if ΔP is kept constant. From the slope and intercept the values
of α and Ve can be calculated, which in turn can be used to predict filtration performance.
For small volume fermentations, a plate and frame filter can be used, which can be opened and the filter cake removed
after each batch operator. For larger volumes, continuous filters are required.
Various kinds of continuous filters are used, one of the most common being a rotary vacuum filter, where strings are
used to lift off the rotating filter cake which accumulates on the filter surface. The clear liquid passes through the surface
to the inside due to the vacuum applied.
17.13.6 Precipitation
Precipitation is usually done to separate and concentrate intracellular proteins after cell lysis. The two major methods
used are: (a) salting-out by adding inorganic salts such as ammonium sulphate at high ionic strength; (b) solubility
reduction at low temperatures by adding organic solvents such as acetone.
In the first process the ions added interact strongly with water, causing protein molecules to precipitate. The solubility
of the solution is given by
ln S / S0 = K s I,
where S is the solubility of the protein in solution, S0 is the solubility at I = 0, I is the ionic strength of solution and Ks is
constant depending on pH and temperature.
Alternatively, addition of organic solvents at low temperatures can precipitate proteins by reducing the dielectric
constant of the solution. The solubility of a protein as a function of the dielectric constant of a solution is given by
817
ln S / S0 = K ′ (1 / D 2s ),
where Ds is the dielectric constant of the buffer–solvent solution.
17.13.7 Liquid–Liquid Extraction

Inhibitory fermentation products such as ethanol or acetone–butanol are separated from the fermentation broth by
liquid extraction. Antibiotics too are recovered by this method. The extractant used is usually a nontoxic liquid which
is immiscible with the fermentation broth having a high affinity for the product. Any liquid mixture (in this case the
fermentation broth) can be separated by this method by contacting it with a second solvent liquid. The components
of the mixture are soluble to varying extents in the solvent liquid. Ideally, only the component to be extracted is sol-
uble in the solvent and the other components are insoluble. Then this solute is the only component transferred from
the initial mixture to the solvent phase. The initial mixture is called the ‘raffinate’ as it is stripped of solute, while the
solvent phase is called the ‘extract’ as it picks up solute. Thus separation of one of the components from a homogenous
solution is accomplished by adding another insoluble constituent (the solvent) in which this component (the solute) is
preferentially soluble. The solute diffuses into the solvent at a characteristic rate till equilibrium concentrations of the
solute have been reached in each phase.
Mechanical mixers are often used to disperse the two phases so as to increase the rate of diffusion of the solute
following which the phases are separated in a settling tank by gravity. These units are called mixer-settlers. Mixer-settlers
can be used in series to give multistage separation.
Calculations: We define an equilibrium stage when the two phases which leave the stage (settler in this case) contain
solute concentrations that in equilibrium are with each other.
Figure 17-36 shows an equilibrium stage where two solvents enter, carrying different concentrations of the solute. The
light phase is called the ‘V’ phase and the heavy phase the ‘L’ phase by convention. The concentration of the solute in the
V phase is y and in the L phase is x. The subscripts denote the stage from which the stream is flowing. This c onvention is
useful when multiple equilibrium stages are involved. The basic material balance on the equilibrium stage gives
L0 + V2 = L1 + V1 .
If we calculate the flows on a solute-free basis and also assume that the L and V phases are essentially immiscible then
we can write
L0 = L1 and V1 = V2 .
Thus the subscripts of L and V can be dropped. This also requires that the solute concentrations (x and y) be reported
as grams per liter of pure solvent (and not grams per liter of solution) since the mass flow rates of the L and V phases
would change as the solute is transferred from one phase to the other.
The component balance gives
Lx0 + Vy2 = Lx1 + Vy1.
The equilibrium relationship between the solute concentrations leaving the equilibrium stage can be written as
V1Y1 V2Y2
(1)
L 0X 0 L1X1
Figure 17-36 Schematic of single equilibrium
operation.
Lx0 X1 X2 Xn
Lxn1
1 2 n n1
Vyn2
Figure 17-37 Schematic of multistage equilib- Y1 Y2 Y3 Yn+1
rium operation with counter current flow.
y1 = f (x1).
In many cases a proportional relationship applies: y1 =

Equilibrium
curve kx1, where k is called the partition coefficient. The above
equations can be used to determine the outlet solute
concentrations.
x1
y1 For multistage equilibrium calculations (Figure 17-37),
x2
we write a component balance over the first to the nth
y2
stage (dotted line). This gives
x3 Lx0 + Vyn+1 = Lxn + Vy1
y3
x4 Operating ⇒yn+1 = (L/V)xn + (Vy1 − Lx0)/V.
line slope
y4 = L /V This equation relates yn+1 to xn, that is the concentrations
of the solute flowing past each other between the nth and
(n + 1)th stage. Plotted graphically this is called the ‘oper-
yn+1 xn x0 ating line’ which has a slope of L/V and an intercept given
Figure 17-38 Graphical technique for doing multistage equi- by the inlet and outlet concentrations of the solute. If we
librium calculations. also plot the equilibrium relationship graphically (Figure
17-38) then a simple graphical technique can be devel-
oped to do multistage equilibrium calculations.
For example if we know the values of x0 and y1 and the slope (L/V) we can plot the operating line. From y1 we can
determine x1 (the concentration of solute in the L phase leaving the first equilibrium stage) by drawing a horizontal line
at y1 and intersecting the equilibrium curve at x1 (as shown in the figure). Next the composition of y2 may be determined
from x1 by using the equation of the operating line. Since the operating line is a general equation relating yn+1 with xn
it also relates y2 with x1 (n = 1). Thus a vertical line drawn from x1 would intersect the operating line at y2. Now x2 can
be determined by drawing a horizontal line as before. This stepwise procedure is continued till (xn, yn+1) is reached. As
shown in the figure, slightly over four stages would be required to achieve this desired concentration. Note that since the
operating line is below the equilibrium curve the solute is transferred from the L to the V phase
17.13.8 Chromatography
Chromatography is a process of separation of mixtures into its components by passing the fluid mixture through a bed of
adsorbant material. In the commonly practiced elution-type chromatography, a column is packed with adsorbed gel (or
porous solid/liquid phase immobilized on a solid). A fluid mixture is injected, followed by an eluent. Different solvents
interact differently with the adsorbent material. Solutes interacting weakly with the matrix pass out rapidly while those
interacting strongly exit slowly. The differential migration rates are used to separate different components. Some of the
important chromatographic methods are:
1. Adsorption chromatography (ADC) is based on the different adsorption of solute particles onto solid particles like
alumina and silica gel by van der Waal’s and steric interactions.
2. Liquid–liquid partition chromatography (LLC) is based on the different partition coefficients of solute molecules between
an adsorbed liquid phase and passing solution.
3. Ion-exchange chromatography (IEC) is based on the adsorption of ions on ion exchange resins by electrostatic forces.
4. Gel filtration chromatography (GFC) is based on the penetration of solute molecules into small pores of packing m aterial
on the basis of molecules shape and size.
5. Affinity chromatography (AFC) is based on specific chemical interactions between the solute molecules and bound
ligands.
6. Hydrophobic interaction chromatography (HIC) is based on hydrophobic interactions between proteins (or other solute
molecules) and the functional groups (e.g. alkyl residues) on the support matrix.
7. High pressure liquid chromatography (HPLC) is based on general chromatography except that it is done under high
liquid pressure which provides fast and high resolution.
17.13.9 Membrane Processes

Membranes are frequently used to separate proteins or to concentrate them. The important membrane-based methods
are as follows.
819
17.13.9.1 Dialysis
Dialysis is a membrane-based process that separates low molecular weight solutes from a solution. The dialysis membrane
is selective and has a cut off value allowing solutes with molecular weight lower than the cut off to move through the
membrane from a high to a low concentration region. At equilibrium the chemical potentials of a diffusing solute on
both sides of the membrane are equal, that is
ma = m b
⇒ RT ln Ca Y a = RT ln Cb Y b
⇒ Ca Y a = Cb Y b ,
where Cα, Cβ are concentrations of the solute in the two phases and Yα, Yβ are the activity coefficients. For ideal
solutions
Ca = Cb .
Hence by varying the volume of one of the phases, the concentration of the solute in the other phase can be controlled.
17.13.9.2 Reverse Osmosis (RO)

RO is another membrane-based process in which a pressure is applied to the dissolved solute containing phase of a
solution which drives water molecules from lower concentration to a higher one across the membrane, thus concentrat-
ing the solute molecules on one side of the membrane. Sometimes membranes allow passage of some solutes along with
the solvent. A retention coefficient (σ) for each solute is defined as the fraction of solute molecules retained on one side
of the membrane in the presence of a solvent flux. If σ = 0, all the solute passes through and if σ = 1, none passes.
The osmotic pressure for a multi component system is defined as
p = Σ p i = ΣCi RT(1 + B2i C i + B3i Ci2 + ),
where Σp i denotes the osmotic pressure due to ith component, Ci is concentration of the ith component, T is the
temperature and R is the gas constant.
In ideal cases B2i = B3i = 0, and the equation reduces to
p = Σp i = ΣCi RT.
The magnitude of pressure varies according to the concentrations of the solutes. As an example, a pressure of
30–40 atm is required for a 0.6 M salt solution. A major problem with RO membranes is due to increased concentration
of solutes close to the membrane surface, reducing solvent flow. This phenomenon is called concentration polarization.
17.13.9.3 Ultrafiltration and Microfiltration

Membranes are sometimes used as molecular sieves to separate proteins of different molecular size. Different membranes
have different cutoff values and different molecular weight proteins. Microfiltration is a similar method used for sepa-
rating different species, like bacteria and yeast. Most of these membranes are made of polymeric materials like celluloses
acetate, nylon, PVDF, etc.
Ultra- and microfiltration operations are pressure-driven processes where low molecular weight solutes pass through
the membrane and high molecular weight solutes are retained. This results in the build up of a concentration gradient and
consequently concentration polarization. The rate of convective transfers of the solute toward the membrane becomes
equal to the rate of diffusion of solute in the opposite direction because of concentration polarization. At steady state
De (dc / dx) = JC,
where De is the effective diffusivity of the solute in liquid film (cm2), J is the volumetric filtration flux of the liquid (cm3/
cm2s) and C is the solute concentration (mol cm −3 ) .
This equation can be solved if the specific boundary conditions are known.
17.13.10 Drying and Crystallization

The final stage of downstream processing consists of crystallization of the product and drying. Specific proteins and other
micromolecules have to be crystallized as a part of their formulation strategy as a therapeutic drug, or for agricultural,
chemical and other purposes. The usual strategy to bring about crystallization is to direct a macromolecular solution
toward a supersaturated state by modifying the properties of the solvent or the solute. This can be achieved by as simple
a physical process as the change of temperature or pH. Alternatively, a change of salt concentration or the addition of a
precipitating agent can also bring about crystallization. Most processes use a combination of various factors like altering
pH and salt concentrations and addition of precipitating polymers. Out of the many approaches attempted so far, the
most successful has been the manipulation of salt concentration. The reason behind this is that, as the salt ions compete
with each other for hydrogen bonding with the solvent molecules, the increase in salt concentration forces the protein
molecules to self-associate to satisfy their electrostatic requirements. Polymers that dehydrate macromolecules (e.g.
polyethylene glycol) and organic solvents like ethanol and acetone also help crystallization in similar ways. However,
optimization of crystallization conditions is a matter of empirical experimentation.
Techniques used for drying also depend on the physical properties of the product, properties of the solid–liquid
system, the drying environment and heat-transfer parameters. The following processes are usually applied for drying.
1. A vacuum-tray drier consisting of heated shelves is mainly used for pharmaceutical products.
2. Freeze drying (lyophilization), a common method used for antibiotics, enzymes and bacterial suspensions, uses
sublimation as a method of removal of water from a frozen solution.
3. Rotary drum driers remove water by evaporation from a thin film of solution on the steam-heated surface of a rotating
drum and the dried product is scraped from the drum with a knife.
4. Spray dryers, used for heat sensitive materials, use an atomization technique where the product solution is sprayed
into a heating chamber through a nozzle where the solvent evaporates and the dry particles are separated using
cyclones.
5. Pneumatic conveyor dryers use a hot air stream to suspend and transport particles. They are also well suited for
heat-sensitive and easily oxidized materials.
17.13.11 Effluent Treatment

Effluents are wastes which are produced in any of the following ways.
1. Industrial waste whose characteristics vary greatly from case to case. These wastes usually contain hydrocarbons,
carbohydrates, alcohols, lipids and aromatic amino acids and can be potentially toxic.
2. Somatic wastes due to daily activities. This includes ground garbage, laundry water, excrement, etc.
3. Agricultural wastes produced from farm animals and plants.
To treat these wastes so as to remove their toxicity and restore environmental balance, many kinds of waste t reatments
are available. The major ones are:
1. physical treatment including flocculation, sedimentation filtration and flotation which are used for removing insoluble
materials;
2. chemical treatment including oxidations or chemical precipitation;
3. biological treatment, which is done by the aerobic and anaerobic treatment of waste water by a mixed culture of
microorganisms.
The carbon content of any wastewater can be expressed in terms of its biological oxygen demand (BOD) or chemical
oxygen demand (COD). A 5-day BOD value is usually taken. The BOD5 is defined as the amount of dissolved oxygen
consumed when a wastewater sample is sealed with active bacteria and incubated at 20 °C for 5 days. This measure is
applicable, however, only to biodegradable soluble organics and requires a high concentration of active bacteria.
COD, on the other hand, is a measure of the concentration of chemically oxidizable organic compounds. Since most
compounds in wastewater are oxidized by strong oxidants, the COD value is usually higher than the BOD value.
Biological wastewater treatment is usually done in the following ways:
1. Activated sludge are grown in well aerated and stirred, continuous flow reactors with a mixed culture of microorganisms.
A common bacterium used is Zoogloea ramigera. They sometimes secrete a polymeric gel which causes the organisms
to flocculate. These floes are called activated sludge.
2. Trickling biological filters use a packed bed of inert support particles covered with a film of microorganisms. Wastewater
is fed to the top. As it descends, the organic compounds diffuse through the microbial film and are utilized.
3. Rotating biological contactors have rotating disks that come periodically in contact with wastewater. The biological
film is on the surface of the rotating disk. As the disk comes in contact with the wastewater, nutrients, diffuse through
the biofilm and are utilized simultaneously.
821
17.14 Whole Cell Immobilization and its Industrial Application•
4. Oxidation pounds are shallow reactors where mixed cultures grow. They require large land areas and can have
environmental side-effects.
17.14 Whole Cell Immobilization and its Industrial Application

Nelson and Briffin, in 1916, through a fortituous event found that yeast invertase adsorbed on activated charcoal retained
its catalytic ability in the hydrolysis of sucrose. This was arguably the first demonstration of a biocatalyst. Immobilization
of whole cells was also attempted by various groups after this result. Finally in 1973, Chibata and co-workers demonstrated
the first application of whole cell immobilization in industry. They used polyacrylamide gel strapped E. coli cells having
a high aspartase activity to produce L-aspartate from ammonium fumarate. Since then immobilized cells have been rou-
tinely used in industrial applications.
The central feature of immobilized cell systems is the use of confining or binding structures to constrain cells in some
section of a bioreactor. This is usually achieved by entrapment or attachment to small particles or by involving barri-
ers to cell transport. However, these cells, though immobilized, retain all their catalytic activities, which can be used
repeatedly and continuously. Owing to a variety of reasons such as low environmental pollution along with economic
utilization of natural resources and energy, immobilized cells have found application in a variety of fields. They have
been found to catalyze complex multistep reactions, which are difficult to achieve with other forms of biocatalysis.
17.14.1 Advantages of Whole Cell Immobilization

One of the major motivations for cell immobilization is the attainment of high cell densities that cannot be obtained
in suspended cell systems. However this is not the only reason. In case of several mammalian cell lines, attachment
to surface is a precondition for growth and division. For such cell lines, immobilization is more of a rule than option.
Additionally, by confining microbial growth to the surface and interstices of support particles, the rheological prop-
erties of the broth become defined. These properties do not, therefore, change much during batch growth as in other
cases. Immobilization also offers continuous processing without any organism washout. It also makes continuous sepa-
ration of products simpler. Sometimes immobilized cells have been used as the basis for specific electrodes to measure
concentration of substrate, metabolic drugs and toxic chemicals.
Immobilized cells can be used at various levels of complexity. At the simplest level of catalysis, immobilized cells are
used for the activity of one of several enzymes. Here, they serve essentially as an immobilized enzyme catalyst, with the
entire cell being used to minimize treatment and processing costs in formulating an immobilized enzyme catalyst.
At a higher level, a nongrowing immobilized cell system can provide catalysis through a multistep pathway. This
will essentially involve a co-factor which will be utilized by the cells and regenerated. Usually such use is derived from
organisms which already possess a co-factor using pathway.
At an even higher level, it is possible to have biosynthesis and maintenance requirements without significant cell
growth or cell division. Finally, the most complex system is one in which the full metabolic activity goes on and the cells
actively grow and divide. These systems can be used for batch operations as well as in a continuous reactor to generate
the biocatalyst. This is done in situations where organisms do not exhibit biocatalysis in nongrowth environments.
There are various considerations for each stage of complexity. If a substrate can be converted stoichiometrically to a
certain product by the action of a particular metabolic pathway without concomitant cell growth or division, it is obvi-
ous that the yield of such a process would be high. This is however not always possible. A second consideration is that of
overall reaction rate. Reaction time increases greatly as we move from the simple to the complex stage. This is because
though the enzyme catalysis has a small time scale in minutes, the time taken by the organism to grow and divide, so as
to regenerate the enzymes and substrate, etc., is of the order of hours to days. It is this time of growth and division that
then controls the overall reaction time. Hence, it is extremely important to design a reactor residence time so as to use
the enzyme and co-factors optimally.
17.14.2 Methods of Immobilizing Cells

There are two distinct types of immobilization. They are called active immobilization and passive immobilization. Active
immobilization is the entrapment or binding of cells by physical or chemical forces, while passive immobilization is the
multilayer growth of cells on solid support surfaces (also called biofilms). The most widely used entrapment methods
employ the immobilization of cells within porous matrices.
Active immobilization can be further classified into four categories: carrier binding, crosslinking, entrapment and
combined methods. Each method has its own advantages and drawbacks and selection is usually, dependent on the pur-
pose of immobilization and characteristics of the cell and the reactor.
17.14.2.1 Carrier-binding Method

This method is based on the binding of biocatalysts to water-insoluble carriers through covalent or ionic bonds, physical
adsorption or biospecific binding. The carriers, along with the property of binding cells, etc., should also have sufficient
mechanical strength as well as physical, chemical and biological stability and they should be nontoxic. The various
substances routinely used as carriers are water insoluble polysaccharides (cellulose, dextran, agarose derivatives, etc.),
proteins (albumin, gelatin, etc.) synthetic polymers (polystyrene derivatives, ion exchange resins, polyurethane, etc.)
and inorganic materials (brick, sand, glass, ceramics, magnetites, etc.). A popular method to immobilize yeast cells
is by covalent binding to porous silica beads that had been treated with aminopropyltriethoxy silane and activated
with glutaraldehyde. Micrococcus luteus cells have been immobilized to the carboxyl group of agarose beads. E. coli and
Azotobacter cells have been reported to have been immobilized to Dowex-1.
17.14.2.2 Crosslinking Method

In this case, cells are linked to each other by the reagents and the crosslinked cells form pellets. Glutaraldehyde, for
example, combines amino groups through a Schiff base linkage. Many kinds of disocyanates (e.g. hexamethylene diisocy-
anate, toluene disocyanate, etc.) are also used as linkage agents. E. coli has been crosslinked with both glutaradehyde and
diisocyanate. The industrial method used for isomerization of glucose uses Bacillus coagulans immobilized by crosslinking.
17.14.2.3 Entrapment Method

Various kinds of entrapment are possible, of which the following are important. (1) Lattice-type entrapment in which, bio-
catalysts are trapped in gel matrices of various types of polymers. (2) Microcapsule-type entrapment which is done in micro-
capsules of semipermeable synthetic polymers. (3) Liposome-type entrapment which is carried out within a liquid membrane
prepared from amphiphatic compounds. In membrane-based entrapment, biocatalysts are separated from the environment
by ultrafiltration membranes or hollow filters. The lattice-type method is the most widely used. Polyacrylamide gel entrap-
ment, for example, was used for the first application of immobilized growing cells for the production of L-glutamic acid
by Corynebacterium glutamicum. Several natural polysaccharides such as alginate, κ-carrageenan and agar are excellent gel
materials and have been extensively used for entrapment.
17.14.3 Procedures for Making Polymer Beads with Immobilized Cells

Polymer beads are usually formed in the presence of cells by one of the following methods.
1. Gelation of polymers: Gelatin or agar beads are prepared by mixing the liquid form of these polymers with cell
suspensions and using a template to form beads. Temperature reduction in the templates causes solidification of the
polymers with the consequent entrapment of the cells.
2. Precipitation of polymers: Cells are dispersed in a polymer solution (polystyrene, cellulose triacetate, etc.) in an organic
or water-miscible solvent like ethanol or acetone. With a change in pH, the polymer is precipitated. Direct contact
of the solvent with the cells may lead to inactivation or cell-death.
3. Ion-exchange gelation: When a water-soluble poly electrolyte is mixed with a salt solution, ion-exchange gelation
occurs as the polyelectrolyte reacts with the salt solution to form a solid gel. Na-aliginate gel is prepared in this way
by mixing Na-alginate solution with a CaCl2 solution (Figure 17-39).
4. Poly condensation: Epoxy, hydroxy, amino, polyurethane, silica gel, etc. are prepared by condensation of liquid
precursors with a multifunctional component.
5. Polymerization: Polymers like acrylamide are prepared by crosslinking co-polymers of a vinyl-group-containing
monomers. It is usually initiated by co-polymerization with a divinyl compound like methylene bis-acrylamide.
17.14.4 Biological Films

Biofilms or layered growth on solid surfaces, are common in natural and industrial fermentation systems, such as biological
wastewater treatment and mold fermentations. In mixed culture biofilms, the presence of some polymer-producing organisms
facilitates biofilm formation and stability.
In a stagnant biological film, nutrients diffuse into the film and products diffuse out into liquid nutrient medium without
growth of cells. The thickness of the biofilm becomes an important factor affecting the performance of the biotic phase.
Thin films with low biomass concentrations will have low rates of conversion which thicker films may be diffusionally
limited. In many cases, an optimal biofilm thickness resulting in maximum rate of bioconversion can be determined.
823
17.15 Industrial Production of Chemicals and Biomolecules•
Magnetic bead
for agitation
8% (w/v) alginate
solution containing cells
Drops
Gelation
(15 min)
2% w/v CaCl2
solution
H2O wash
30 min Drying 20 h
30 min
Biocatalyst
beads
0.1 M Al2(SO4)3 solution
Figure 17-39 Protocol for entrapping cells in alginate beads.
17.15 Industrial Production of Chemicals and Biomolecules

Microorganisms have been routinely used for a long time for production of a large number of chemicals with varied
applications. In this section we take a look at the industrial production of a few such chemicals.
17.15.1 Production of Ethanol

Though alcohols had been brewed for thousands of years by traditional means, it was industrially synthesized by the
catalytic hydration of Ethylene. In recent years, production of ethanol is once again being done by fermentation. In
many countries efforts are being made to produce ethanol from sucrose and starch, to be used for automobile fuels among
other possibilities. Ethanol is also being produced by fermentation from ethylene and other hydrocarbons.
Both yeast and bacteria have been used to produce ethanol. Zymomonas mobilis is the most commonly used bacteria
while Saccharomyces cerevisiae is the most commonly used yeast. Sometimes the yeast Kluyveromyces fragilis has also been
employed. Glucose is metabolized in aerobic conditions by Saccharomyces but no alcohol is produced. Under anaerobic
conditions however, pyruvate from the Glycolytic pathway is split with pyruvate decarboxylase into acetaldehyde and
CO2. Ethanol is then produced from acetaldehyde by reduction with alcohol dehydrogenase.
Glucose
L
So
Glycolysis
Pos
Pyruvate
decarboxylase
Pyruvate
Mg2+, Thiamine pyrophosphate
Soi
Acetaldehyde + CO2
Alcohol dehydrogenase
NADH2
Ethanol
Poi
Sos 17.15.2 Production of Organic Solvents
17.15.2.1 Acetone/Butanol
Chain Weizman performed fundamental research on the
Support fermentation of Clostridium acetobutilicum for the production of
surface
Po acetone, butanol and ethanol. During World War I, acetone was
the product of interest, used for the production of the e xplosive
tri-
nitrotoluene. After the war, butanol became important
Biofilm
primarily for the production of nitrocellulose lacquers. Butyric acid,
Liquid film butanol, acetone and isopropanol are all obtained through clos-
Figure 17-40 Schematic representation of a biofilm tridial fermentation of starch, molasses, sucrose, wood hydrolysates
showing substrate and product profiles in the liquid film and pentoses. The relative proportions of each of these products
and biofilm. S0: subtrate concentration in bulk medium; depends on the strain used and the fermentation conditions.
S0i: substrate concentration in the biofilm liquid inter- Of the many fermentation products, only the acetone-butanol
face; S0s: substrate concentration on the support surface. part is of current economic interest. During the bioprocess, one
has to remember that butanol at less than 0.5% has no influence
on the cells, while at higher concentrations it causes damage to
the cell membrane and at concentrations above 1.3% b utanol production ceases. Autolysins are also produced some-
times which causes the cells to lyse.
17.15.3 Production of Organic Acids

17.15.3.1 Citric Acid
Citric acid is used as an acidulant in food, confectionary and beverages. Its pharmaceutical and industrial applications
are also significant. Production of citric acid from sugar solutions was first realized by Penicillium. Nowadays, however,
Asperigillus niger is the most widely used organism for production. The yeast Candida can also be used to produce citric
acid from carbohydrates or alkanes with yields as high as 225 g/l. Because of its low price and nutrient rich nature, molas-
ses is usually preferred over pure sugar solutions.
Citric acid production follows non-growth associated, kinetics mainly taking place under nitrogen and phosphate
limitations after growth has ceased. It is the result of primary metabolism, hence it is produced in high concentrations
only under very specific conditions. These include restricted growth, medium deficient in one or more essential elements,
high sugar concentrations, high dissolved oxygen concentration, low pH (pH<2) and the absence of trace metals. Such
conditions provoke an overflow in metabolism that results in an overproduction of citric acid. Potassium ferrocyanide is
added to reduce the concentration of metals and is a growth inhibitor that promotes the production of citric acid. In the
process the fungi morphology also change. The hyphae become short, stubby and forked in small pellets.
Fermentation is carried out in either batch or fed batch mode in deep stainless steel vessels of 100 m3 or more volume.
Aeration is done by air sparging (0.1–0.4 vvm) andagitation is usually gentle (50–100 rpm) to avoid shear damage to the
mycelia. The initial pH is adjusted to 2.5–3. Spores of A. niger are allowed to germinate in an inoculum medium before
transferring them to the fermentation media. Dissolved oxygen concentration is critical for citric acid production and
oxygen enriched air is often used. Typically 80% of supplied carbon is converted to citric acid. Batch operation results
in productivities of about 0.5–1 kg/m3 h, while fed batch operation can be used to prolong production and reach product
concentrations of 130 kg/m3. When citric acid production stops, usually after 4–5 days, the fermentors are emptied and
825
the biomass is separated by filtration. Precipitation is usually accomplished by addition of Calcium hydroxide (lime) to
the fermentation broth. The precipitate is washed and treated with dilute sulphuric acid yielding an aqueous solution of
citric acid and CaSO4. After bleaching and crystallization, either anhydrous or monohydrate citric acid is obtained. The
demand for citric acid was 400,000 tons in 1999 ($ 1,400 million), which is expected to increase 3% annually.
17.15.3.2 Gluconic Acid

Gluconic acid is used in the manufacture of metal, leather and food. Its derivatives like Sodium gluconate is used as
a sequestering agent in many detergents. Calcium gluconate is used in medicine while d- gluconolactone functions as
baking powder. The production of gluconic acid from glucose can be carried out by many organisms. Gluconolactone
formed in the first step hydrolyses either spontaneously or enzymatically to gluconic acid.
In 1911, Alsberg described the production of gluconic acid with Pseudomonas. A surface process using the fungus
Penicillium luteum-purpurogenum was begun in 1928 with yields amounting to 80–87% of the theoretical yields. Nowadays
the fungus Aspergillus niger and the bacterium Acetobacter suboxydans are used for production. The fermentation is carried
out at a pH between 4.5–6.5 in a growth medium which is both nitrogen and phosphorus limiting. The fermentation
runs 20 hours at 28–30 °C with a high aeration rate. By raising the pressure in the system, gluconic acid yields can be
increased. Commercial yields are around 90–95% of theoretical values.
17.15.4 Production of Amino Acids

17.15.4.1 Lysine
Lysine is an essential amino acid occurring in low proportion in plant proteins. Hence plant based food is often enriched
with lysine. Currently it is only produced by microbial processes. Mutant strains of Corynebacterium and Brevibacterium
which are homoserine auxotrophs (or Methionine-threonine auxotrophs) are found to produce L-lysine. A lysine analog
S-(β-aminoethyl)-L-cysteine (AEC) inhibits lysine synthesis by mimicing its control properties. Strains mutated for
analog resistance (B. Lactofermentum) have been found to give high yields since they do not have a pathway which is
feedback regulated by lysine.
17.15.4.2 Glutamic Acid

Though glutamic acid can be manufactured chemically, it is predominantly done through microbial means. The i ndustrially
important organisms used for glutamic acid production are Corynebacterium, Brevibacterium and Microbacterium. They
are all Gram-positive, non sporulating, nonmotile bacteria. They usually require biotin for growth. Glutamic acid is syn-
thesized from α-keto-glutarate, which is formed in the TCA cycle, through reductive amination with free NH4+ ions.
This step is catalyzed by the NADP dependent glutamate dehydrogenase.
17.15.5 Production of Antibiotics

17.15.5.1 Penicillin
Antibiotics are products of secondary metabolism that inhibit the growth processes of other organisms. They are
produced by different species of bacteria, actinomycetes and fungi. Penicillin, first described by Fleming in 1929, was
isolated from the fungus Penicillium notatum. Many species of fungi, particularly Penicillium and Aspergillus produce pen-
icillin. Penicillin possesses a β-lactam ring which acts as a preventor of bacterial cell wall biosynthesis. Penicillins are
thus potent inhibitors against gram positive bacteria. Nowadays a variable acyl moiety in the penicillin structure is
replaced by adding different side chain precursors to create semi synthetic penicillins. Penicillin G, penicillin V and
penicillin O have been produced in this way.
Submerged culture production of penicillin in 40,000 to 200,000 litre fermentors is nowadays done by using an initial
inoculum of lyophilized spores of the fungus Penicillium chrysogenum. A mutagenic strain is commonly used. Strain devel-
opment for penicillin production is an active area of research. The medium used is usually a mixture of corn steep liquor,
additional nitrogen sources like soya meal or yeast extract and a carbon source like lactose.
The fermentation process requires an oxygen uptake rate of 0.4–0.8 mML−1 min−1, therefore aeration is done with
turbine impellers operating at 120–150 rpm. The growth phase is about 40 h with a doubling time of 6 h. When
production occurs, the specific growth rate drops to about 0.01 h−1.
The production phase can be extended to 120–180 h by feeding and final productive concentrations over 50 gl−1 have
been acheived.
17.15.5.2 Streptomycin
Aminoglycoside antibiotics (streptomycin, kanamycin, gentamycin) are produced in fermentors of 150,000 l capacity,
with optimal aeration, a temperature of 28–30 °C and a pH in the neutral range. The process is of 4–7 days duration.
Glucose with starch is used as carbon source. Soya meal acts as nitrogen source and NaCl (1–3 g l−1) is added to the
fermentation growth. Strains have now been isolated which give product concentrations of 15 g l−1.
17.15.5.3 Tetracycline
Tetracyclines are broad spectrum antibiotics effective against gram-positive and gram-negative bacteria as well as
ricketsias, mycoplasmas, Leptospiras, Spirochetes and Chlamydias. They have a naphthalene ring system and act as inhibi-
tors of protein synthesis by preventing the binding of the amino acyl tRNAs to the ribosomal A-site. Chlortetracycline
was the first antibiotic of this group to be isolated. Tetracycline was initially synthesized from chlortetracycline but later
was found to be produced directly by Streptomyces viridifaciens. At least 20 different Streptomycetes are now known to
produce a mixture of different tetracyclines. Many mutant strains have now been developed secreting high amounts of
tetracycline. The tetracycline biosynthetic pathway requires 72 intermediate products, all of which have not yet been
characterized.
17.15.6 Production of Single-cell Protein

The term single-cell protein means dried cells of organisms, used for feed or feed additives (which are primarily of
proteinaceous nature). Species like algae, actinomycetes, bacteria, yeasts, molds, higher fungi, etc. have been used to
generate protein-based food for human consumption. Examples which have been used for a long time include yeast in
leavened bread, lactic acid bacteria in dairy products (milk, cheese, etc.) and algae (spirulina) harvested from ponds
and lakes. The use of biomass produced by microbial cells, which have high protein content, is very contextual in view
of the insufficient world food supply. There are several bottlenecks, however, which include: (a) the high nucleic acid
content (4–6% in algae, 10–16% in fungi) which can be hazardous to health, (b) toxic or carcinogenic substances may
be present and (c) it may cause indigestion or allergic reactions.
17.15.7 Production of Recombinant Proteins/Products

A large number of industrially important molecules are now made by recombinant techniques where foreign genes from
another host are inserted into a bacterial host such as E. coli (or a more complex host depending on the need) which is
then used as a production platform. It is nowadays a very powerful method for imparting novel attributes to the host cell.
We discuss briefly some of the advancements in this area from a bioprocess perspective.
17.15.7.1 Recombinant Protein Production in E. coli

The choice of an expression system for the high-level production of recombinant proteins depends on many factors.
These include cell-growth characteristics, expression levels, intracellular and extracellular expression, and biological
activity of the target protein. In addition, the selection of a particular expression system requires a cost down in terms of
process, design and other economic considerations. The combination of recombinant DNA technology and large-scale
culture process has enabled recombinant proteins to be produced in massive quantities. E. coli was the first host that
was used to produce a recombinant DNA (r-DNA) pharmaceutical, enabling the approval of Eli Lilly’s r-DNA human
insulin as early as in 1982. It still remains the primary host for the production of recombinant and therapeutic proteins
in laboratory research as well as in industries. Though a wealth of expression systems designed for various applications
associated with recombinant protein production exists, E. coli still remains the ‘first choice’ as an expression system.
E. coli facilitates protein expression by its relative simplicity, inexpensive growth requirements and fast high-cell-density
cultivation. Additionally, the well-understood genetics and metabolism of E. coli, the availability of a large number of
compatible tools for biotechnology, specially the variety of available plasmids, recombinant fusion partners and the
mutant strains, all make E. coli based expression system as an attractive proposition for recombinant protein expression.
To present a statistic, approximately 80% of the proteins used to solve the three-dimensional (3-D) structures submitted
to the protein data bank (PDB) in the year 2003 were prepared in an E. coli expression system. Recent progress in the
fundamental understanding of transcription, translation and protein folding in E. coli, together with serendipitous dis-
coveries and the availability of improved genetic tools are making this bacterium more valuable than ever for the expres-
sion of complex eukaryotic proteins. Although there is no guarantee that a recombinant gene product will accumulate
in E. coli at high levels in a full length and biologically active form, a considerable amount of effort has been directed
at improving the performance and versatility of this useful microorganism. E. coli contains two membranes (inner and
outer membrane) and can be divided into three compartments by these two membranes: cytoplasmic, periplasmic and
extracellular space. The production of recombinant proteins is targeted to one of these three compartments and the
mode of gene expression affects the location of the protein produced.
827
In the 1980s, however, it had become apparent that E. coli could not be used to produce some proteins particularly
requiring post-translational modification for biological activity and some other complex proteins containing multiple
disulfide bonds. In addition, many proteins expressed in E. coli were accumulated intracellularly in the form of insol-
uble, inactive inclusion bodies, from which biological activity must be recovered by complicated and costly denatur-
ation and refolding process. During the mid-1980s, these drawbacks of E. coli led to the development of animal/plant
cell expression systems. However, although such systems have proved capable of producing authentic active proteins,
they have often failed to develop the simple and efficient, high-yield, low-cost methods for the production of proteins
in large amount. In contrast, recent advances in protein refolding, translocation and roles of molecular chaperons and
foldases have made possible the design of recombinant E. coli strains that accumulate proteins in soluble form, secrete
proteins into periplasm, export proteins into culture medium and direct proteins to the outer membrane of the cell for
surface display.
17.15.7.2 E. coli Host Strains

The strain or the genetic background for recombinant protein expression is naturally important. Expression strains
should be deficient in the most harmful native proteases, maintain the plasmid stably and contain the genetic elements
relevant to the expression system. The most common strategy to enhance the in vivo stability of recombinant proteins
(especially protease sensitive ones) involves the use of a host that lacks one or more proteases or the other regulatory pro-
teins involved with degradation. Using a ‘lon’-strain, deficient in La protease (an E. coli cytoplasmic protease responsible
for degradation of proteins with non-native structures), stabilization of the recombinant protein was obtained. Strains
that do not produce the alternative sigma factor σ32, which controls the transcription of heat shock genes, have been
shown to have a decreased capacity to degrade unstable or abnormal proteins. Moreover, these are also defective in heat
shock response. Mutant strains deficient in several chaperone proteins involved with protein degradation (dnaK, dnaJ,
grpE or groEL) have been shown as potential alternatives for expression of foreign unstable proteins in E. coli. However,
there are no general solutions, as over-expression of some of these chaperones in some cases has helped to increase the
half-life of other recombinant proteins. The development of secretion systems is an attractive option for large-scale
production of recombinant proteins in E. coli, thus creating and using strains lacking proteases in the cell envelope
for the production of recombinant proteins. The use of a mutant E. coli strain lacking the DegP proteolytic activity in
the periplasm, resulting in complete stabilization of recombinant proteins, has also been reported. Similarly, reports of
improved secretion of a number of recombinant proteins with OmpT-strain have been published in literature. Lin and
co-workers emphasized the importance of selecting specific strains for the production of recombinant penicillin acylase
(PAC) as a target protein in high-cell-density cultures of E. coli. Reports on expression studies with double and triple
mutants of the cell envelope proteases for secretory expression have also been published; however, development of these
combined mutant strains impairs their viability and thus hampers good growth. Meerman and Georgiou created a set of
20 different E. coli strains deficient in all known loci affecting the proteolytic stability of secreted recombinant proteins,
in an attempt to generate E. coli host strains that exhibit the optimal combinations of growth and protein stability.
An alternative expression system, the ‘E. coli’ quiescent cell expression system’, has been developed where the
quiescent state (non-growing but metabolically active cells) is established by co-overexpressing an ‘Rcd’ transcript
(a regulatory transcript encoded by E. coli plasmid ColE1), in a hns-205 mutant host. The purpose of using this system was
to shift the balance towards more and better target protein yields rather than biomass by redirecting the metabolic flux
from biomass production to recombinant protein synthesis. These quiescent cells have indeed proved to be more efficient
in resource utilization than conventional E. coli expression systems. Specific strains for a number of individual applica-
tions are available. The E. coli BL21 (DE3) strain has proven to be outstanding in standard recombinant expression appli-
cations. (λDE3) is a lysogen integrated into the genome of this host strain where the T7 RNA polymerase gene is under
the control of the lacUV5 promoter. BL21 is a robust E. coli B strain that is able to grow vigorously in minimal media
while being non-pathogenic and unlikely to survive in host tissues and cause disease. BL21 is deficient in OmpT and lon,
the two major proteases that may interfere with the isolation of recombinant proteins in their intact form. Derivatives of
BL21 include recA negative strains for the stabilization of the target plasmids containing repetitive sequences [BLR (DE3)
strain, Novagen], trxB/gor negative mutants for the enhancement of cytoplasmic disulphide bond formation (Origami
and AD494 strains, Novagen), lacY mutants enabling adjustable levels of protein expression (Tuner series, Novagen)
and mutants for the soluble expression of inclusion body prone and membrane proteins [CD41 (DE3) and CD43 (DE3)
strains, Avidis]. These mutant strains can be screened for the selection of the most suitable host strain/vector strategy for
production of a specific target recombinant protein which would help in the development of an ‘ideal cell factory’ for that
protein. Techniques for improving the heterologous protein production in E. coli are summarized in Table 17-3.
Table 17-3 Techniques for improving the heterologous protein production in E. coli
Host strain Choice of host strain impacts expression
Plasmid copy number Gene dosage, as manipulated through plasmid copy number, affects expression.
Selection antibiotic Choice of antibiotic resistance on the expression plasmid can influence heterologous protein
expression.
Promoter Strong/weak, inducible/constitutive; promoter regulation is a major influence on protein expression,
which is also affected by relative orientation and strength of promoters on the plasmid.
Transcription termination Effectiveness and spacing of transcription terminators affect expression.
mRNA stability The stability of the mRNA impacts yield. Secondary structure, especially at the 5′ end of the message,
often plays a critical role.
Translation signal The ribosome-binding site affects the level of ribosome loading and clearance, and hence expression.
Secondary structure at the 5′ end of the message can affect the accessibility of the ribosome-binding site.
Codon usage Utilizing the optimal E. coli codons often improves yield.
Temperature Temperature has a pronounced effect on protein folding and stability. Lower temperature frequently
improves soluble yields.
Growth conditions/media Growth conditions, oxygen levels, growth rate, carbon source and fermenter configuration affect yield.
Fusion proteins Fusing heterologous proteins to peptide that are typically highly soluble in E. coli often improve yield
of soluble product.
17.15.7.3 Alternative Expression Systems

Apart from E. coli, other bacterial systems which are successfully used in recombinant protein production include Bacillus
subtilis, Bacillus stearothermophilis, Corynebacterium and Streptomyces species.
Bacillus: It is the most preferred prokaryote for the industrial production of a range of proteins like amylase, protease,
xylanase, etc. This organism is favored due to the fact that its cultivation in large-scale production systems at high cell
densities is easy and inexpensive. At present, about 60% of the commercially available enzymes are produced by Bacillus
species, mostly being homologous proteins that are naturally secreted in the growth medium, such as alkaline proteases
as washing agents or amylases for the starch industry.
B. subtilis is an attractive host because it has a naturally high secretory capacity and exports proteins directly into the
extracellular medium. In contrast to E. coli, the Gram-positive bacterium B. subtilis is considered a GRAS (Generally
Recognized As Safe) organism. For this reason, the use of B. subtilis for the production of food products is highly favored
over E. coli. B. subtilis, like other gram-positive bacteria, have a rather simple cell envelope comprising only a single layer
of cell membrane, which makes it a suitable candidate for expression of traditionally secretory proteins such as amylases
and proteases directly into the culture medium. This process obviates the necessity of disrupting the cells and also makes
downstream processing simpler. Additionally, the product does not contain membrane lipopolysaccharides (LPS), the
pyrogenic components that commonly constitute the ‘endotoxins’ in E. coli. However, there remain some bottlenecks
in the expression and secretion of heterologous proteins in B. subtilis. The major reservations about using Bacillus have
been lack of suitable expression vectors, plasmid instability, presence of proteases, and occurrence of misfolded proteins.
B. stearothermophilus is a typical thermophile which is used for production of enzymes such as thermostable β-galactosidase
and D-hydantoinase. Also because of the high cultivation temperatures (> 55 °C), the number of possible contaminating
organisms is greatly reduced.
B. megaterium is another bacterium which has generated lot of interest lately. In contrast to B. subtilis, B. megaterium
does not possess alkaline proteases and is known for better and stable replication and maintenance of plasmids. The high
stability of these plasmids during growth allows stable gene expression in long-term cultivations in bioreactors. A large
number of heterologous proteins such as dextransucrase, levansucrase, Clostridium difficile toxinA and penicillin amidase
have been expressed and secreted by designing suitable plasmid vector systems compatible with the B. megaterium host.
Production of heterologous proteins has been further improved by the use of an exoprotease NprM-deficient B. megaterium
strain MS941 and by co-expression of the signal peptidase gene sipM. High yields of functional scFv antibody fragments
have been produced and secreted into the culture medium, making it a reasonable alternative to the E. coli system.
Streptomyces: Streptomyces species are well-known producers of several antibiotics. Many of these are constantly used
in human and veterinary medicine and in agriculture. It has been used as a potential host for recombinant production
of Mycobacterium tuberculosis proteins. Application of recombinant DNA techniques to Streptomyces plays an important
role in strain improvement aimed at increasing the antibiotic yields (metabolic engineering, gene dosage, etc.) and the
generation of novel antibiotics (chimeras) by incorporating parts of different antibiotic biosynthetic pathways into the
829
strain. Streptomyces are also considered to be potentially suitable for the development of efficient secretion systems. The
reported yields are still below cost-effective ranges in most cases. The comparatively low yields demonstrate that a lot of
fundamental research is still necessary to render Streptomyces systems competitive. However a recent study comparing
recombinant alpha-amylase yields demonstrated that both the final yields and enzyme activity were considerably higher
in S. lividans in comparison to the periplasmic production in E. coli. High-level heterologous expression and secretion of
two major antigenic proteins of M. tuberculosis has been achieved in S. lividans.
Corynebacterium and related bacteria: These are a group of pleomorphic asporogenous gram-positive bacteria
found in a wide variety of ecological niches. Their major industrial importance is in the production of various amino
acids like lysine, glutamic acid, threonine, etc. Gene manipulations in these industrially important bacteria were ini-
tiated by the discovery of their indigenous small cryptic plasmids used for vector construction. Mukherjee et al. con-
structed a series of cloning vectors based on pBL1 replicon. Cloning vectors harboring hygromycin and chloramphenicol
resistance were developed by Cadenas et al. Reports on the efficient expression of foreign genes are also common. Ovine
gamma interferon was expressed as a GST fusion protein in two bacterial hosts, E. coli and C. glutamicum. Other foreign
genes successfully expressed in C. glutamicum include subtilisin of B. subtilis and basic protease of Dichelobacter nodosus.
The gene encoding fibronectin-binding protein 85A of M. tuberculosis has also been expressed in C. glutamicum using
pBL1 replicon-based vectors containing tac promoter of E. coli or endogenous cspB gene promoter. Expression of GFPUV
under the control of positively regulated araBAD promoter has been reported by Ben-Samoun et al. This study estab-
lished that C.glutamicum RNA polymerase is activated by the E. coli positive regulator of transcription AraC. Cloning
and expression of xylanase (xys1>) and the cellulase gene (celA1) from Streptomyces halstedii JM8 has been reported in B.
lactofermentum. The broad host range plasmid pEP2 was used to express the protective protease gene from D. nodosus,
causing ovine footrot in C. pseudotuberculosis. Recently, a series of gene fusion vectors have been constructed for expres-
sion of recombinant proteins in corynebacteria.
Yeast expression systems: These are unique as they offer the advantages of both being a microorganism and a
eukaryote. Its basic advantage is that it can be cultivated like bacteria on defined media, while it has the protein pro-
cessing machinery of a eukaryotic cell. Unlike E. coli, yeast provides advanced protein folding pathways for heterologous
proteins, and with yeast, signal sequences secrete the correctly folded and processed proteins. Therefore functional and
fully folded heterologous proteins can be secreted into the culture media. Unlike mammalian expression systems, yeast
can be rapidly grown on simple growth media. For the expression of clinically and industrially important proteins, yeast
is an attractive option as industrial scale fermentation technology is now widely used. Whole antibodies and antibody
fragments have been expressed using this system. The binding activity of whole antibody and Fab secreted from yeast was
found to be similar to that of their counterparts derived from lymphoid cells. Single-chain antibodies have also been suc-
cessfully expressed in yeast systems, for example, an anti-fluorescein scFv has been produced in Schizosaccharomyces pombe
and anti-recombinant human leukemia inhibitory factor scFv has been expressed in Pichia pastoris. scFv proteins which
are produced as insoluble inclusion bodies in E. coli are often soluble when expressed in yeast. In addition, the degrada-
tion of heterologous proteins, often a problem in E. coli, is usually reduced in yeast. Yeast systems include Saccharomyces
cerevisiae, P. pastoris, Hansenula polymorpha and Pichia methanolica. S. cerevisiae is a genetically well-characterized yeast
whose complete gene sequence is known. It is a commercially important strain, because it can be cultivated like bacteria
on defined media. The cultivation time for S. cerevisiae is also shorter compared to other eukaryotic cells. Recent studies
with S. cerevisiae have explored both the strong cell cycle dependence and metabolic burden of heterologous protein
secretion in these cultures. These studies provide interesting points for comparison between microbial and mammalian
systems, as well as an improved fundamental understanding of the use of yeasts for recombinant protein production.
Pichia pastoris, Hansenula polymorpha, Candida boidinii and Pichia methanolica are methylotrophic yeasts. Pichia pastoris
has developed into a highly successful system for the production of a variety of heterologous proteins. The increasing
popularity of this particular expression system can be attributed to several factors, most importantly: (1) to the presence
of strong inducible promoters (AOX1) as well as constitutive promoters (like GAP) which allow over-expression of
recombinant proteins; (2) the simplicity of techniques needed for the molecular genetic manipulation of P. pastoris and
their similarity to those of S. cerevisiae; (3) the ability of P. pastoris to produce foreign proteins at high levels, either
intracellularly or extracellularly; (4) the capability of performing many eukaryotic post-translational modifications, such
as glycosylation, disulfide bond formation and proteolytic processing and (5) the availability of the expression system as
a commercially available kit.
The optimal process conditions for human chymotrypsin B production in P. pastoris resulted in product levels of 480
mg L−1 in the broth with a biomass concentration of 150 g L−1. Alternatively, a continuous culture technique increased
the volumetric productivity by fivefold compared to fed batch culture. An insulin precursor was also expressed with
yields of 1.5 g L−1. By using more complex feeding strategies for high-cell-density cultures of P. pastoris, recombinant scFv
antibody fragments have been produced in large amounts. Very high yields in the range of few grams per liter have been
reported for a range of proteins.
Insect cells/Baculovirus system: Baculovirus gene expression is a popular method for producing large quantities of
recombinant proteins in insect host cells. In most cases, post-translational processing of eukaryotic proteins expressed
in insect cells is similar to protein processing in mammalian cells. As a result, insect-cell-processed proteins have com-
parable biological activities and immunological reactivity to proteins expressed in mammalian cells. Protein yields from
baculovirus systems are higher, and costs are significantly lower than in mammalian expression systems. The baculovirus
expression system can express genes from bacteria, viruses, plants and mammals at levels from 1–500 mg L−1; most pro-
teins are expressed in the 10–100 mg L−1 range, although making predictions is difficult. The baculovirus most commonly
used to express foreign proteins is Autographa californica nuclear polyhedrosis virus. AcMNPV can be propagated in
certain insect cell lines; the virus enters the cells and replication begins approximately 6 hours post-infection (h.p.i.).
At approximately 20–48 h.p.i., transcription of nearly all genes ceases. The viral polyhedrin and p10 genes, however, are
transcribed at high rates. The polyhedrin protein is essential for propagation of the virus in its natural habitat; however,
in cell culture, polyhedrin is not needed, and its coding sequence can be replaced with a sequence for a target protein.
Hence, the powerful polyhedrin promoter can drive high-level transcription of the insert, resulting in expression of a
recombinant protein that can account for over 30% of total cellular protein (TCP).
Mammalian cell culture expression systems: Over the past 9–10 years, there has been a considerable progress in
fine tuning mammalian expression systems for high-level recombinant gene expression. Chinese hamster ovary (CHO)
cells and murine myeloma cells (NS0) have established themselves as predominant systems of choice for mamma-
lian expression. Mammalian cell cultures are the most acceptable way of producing eukaryotic proteins because the
authenticity of the protein is guaranteed, given the similarity of the host protein processing machinery. Thus proteins
like murine erythopoietin (EPO) and TPA are still produced for pharmaceutical purposes in mammalian cell cultures.
However, the cost of mammalian cell culture remains prohibitive because of high cost of media as well as extremely high
level of sterility to prevent contamination. Thus, scale up of mammalian cell cultures remains a big challenge not only
because of sterility requirements but also because of the shear sensitive and fragile nature of mammalian cells. Genetic
engineering approaches too have rationally modified the specific features of mammalian host cells and improved their
utility in recombinant protein expression. Recent developments in the use of serum-free media for cell cultivation
have helped cut down operational costs. Simultaneously, fine tuning of bioprocess parameters has helped increase the
productivity and yields of mammalian cell expression systems.
Cell-free expression system: Cell-free methods introduce a new degree of complexity because here the cells must
be first grown and then cell extracts prepared for further expression studies. However, the cell-free approach offers
many potential advantages; it allows the synthesis of proteins toxic to the host cells, the channelization of the met-
abolic resources towards product synthesis and also provides incredible flexibility in manipulating protein synthesis
and folding. The efficiency of ‘cell-free’ expression systems also continues to improve, both for the traditional systems
using cell extracts and for systems reconstituted from purified components. Kigawa and co-workers synthesized up to 6
mg L−1 of chloramphenicol acetyl transferase (CAT) using a continuous exchange cell-free system. Recently a report
on co-translational expression and folding of firefly luciferase using an E. coli cell-free translation system’ has been
described. Although currently useful for high throughput screening applications in this era of ‘proteomics’, it will be
interesting to see in coming years, to what extent these can be useful for large-scale protein production.
Heterologous expression has also been dealt with in Chapter 14, Section 14.13.
17.16 Mineral Beneficiation and Oil Reparation

17.16.1 Bioleaching
The process by which metals are dissolved from ore-bearing rocks by the use of microorganisms is called bioleaching.
Conventional chemical methods of leaching turn out to be nonprofitable in many cases where the mineral content of
the ore is very low. For many metals nowadays there are bioleaching methods which permit extraction from metal sul-
phides and other ores. Microorganisms are used to convert them to water soluble metal sulphates which can be easily
extracted. Metals like copper and uranium have been widely produced with the use of microorganisms.
The microorganism Thiobacillus ferroxidans is one of the most widely used in bioleaching. This organism was first
identified in 1950 and was found to be able to oxidize elemental sulphur and ferrous ions at much higher rates that could
be achieved by inorganic chemistry. Indeed, it is the catalysis of the oxidation of ferrous ions that makes T.ferroxidans
and other iron and sulphur oxidizing microorganisms such important catalysts in the extraction of metals by bioleaching.
831
17.16 Mineral Beneficiation and Oil Reparation•
17.16.1.1 Organisms Used in Bioleaching

The isolation of thermophilic sulfur-oxidizing microorganisms from hot springs in the late 1960s provided the opportunity
of operating reactions at high temperatures. Carl Woese proposed that these organisms belonged to an entirely new
kingdom of life called Archaea. The two most commonly used organisms for bioleaching are Thiobacillus thiooxidans
and Thiobacillus ferroxidans. Some other organisms that may be used include Thiobacillus concretivorus, Pseudomonas
fluorescens, P. putida, Achromobacter, Bacillus licheniformis, B. cereus, B. luteus, B. polymyxa and B. megaterium. Some
other thermophilic bacteria that have been used in bioleaching are Thermothrix thioparus, Thiobacillus thermophilica and
Sulpholobus acidocaldarius.
17.16.1.2 The Chemistry of Bioleaching

The most extensively studied organism in bioleaching is Thiobacillus ferrooxidans. It is a gram-negative rod-shaped
bacterium, which is 0.5–0.8 µm by 1.0–2.0 µm in size. As an autotrophic aerobe, it obtains carbon for biosynthesis
through CO2 fixation. The energy required to drive these biosynthetic reactions is obtained from the oxidation of Fe2+ to
Fe3+ or from the oxidation of elemental sulfur and from the conversion of reduced sulfur compounds to sulfates:
4FeSO 4 + 2H 2SO 4 + O 2 → 2Fe2 (SO 4 )3 + 2H 2 O
2S + 3O 2 + 2H 2 O → 2H 2SO 4
2FeS2 + 7O 2 + 2H 2 O → 2FeSO 4 + 2H 2SO 4
This oxidation occurs in the periplasmic space of the organism
17.16.1.3 Copper Leaching

Chalcocite, chalcopyrite or covellite are typical ores that are used for the production of copper. Chalcopyrite is oxidized
as follows:
1
2CuFeS2 + 8 O 2 + H 2SO 4 → 2CuSO 4 + Fe2 (SO 4 )3 + H 2 O
2
Covellite is oxidized to copper sulfate by the following reaction:
CuS + 2O 2 → CuSO 4
Copper leaching plants use a leaching solution (sulfate/Fe3+ solution) which carries the microbial nutrients in and
the dissolved copper out. It is sprinkled over the heap and percolates through the rock pile below where the copper-rich
liquid is collected. Using such methods large installations have been known to produce up to 200 tons of copper per day.
17.16.1.4 Uranium Leaching

Though the material turnover for uranium is less than that for copper, uranium leaching is more significant economically.
This is because almost a thousand tons of uranium ore must be handled to obtain about one ton of pure uranium. In
this process, insoluble tetravalent uranium is oxidized using a hot H2SO4/Fe3+ solution to soluble hexavalent uranium
sulfate:
UO 2 + Fe2 (SO 4 )3 → UO 2SO 4 + 2FeSO 4
This process is called an indirect leaching process because the microbial action is not directly on the uranium ore but
on the iron oxidant. In the commercial process, the dissolved uranium is extracted from the leach liquor with organic
solvents like tributylphosphate.
17.16.1.5 Commercial Processes

Three general methods of commercial bioleaching have found practical applications.
1. Slope leaching: Finely ground ores are dumped in large piles down a mountainside and continuously sprinkled with
water containing Thiobacillus. The metal-rich water is collected at the bottom.
2. Heap leaching: The ore is arranged in large heaps and treated as in slope leaching.
3. In situ leaching: Water-containing Thiobacillus is pumped through drilled passages to unextracted ore. The acidic water
seeps through the rock and collects in the bottommost cavity from where the metal is extracted.
17.16.1.6 Commercial Bioleaching Developments

The South African company Gencar pioneered commercial scale bioleaching of refractory gold bearing sulphide
concentrates using the microorganism Thiobacillus auroxidaus at the Fair View gold mine. After about 10 years of in-house
research a 750 kg/day pilot plant was established in 1986. This ensured not only a far superior recovery of gold but also
a clean environment.
Recently an overall process, termed BIONICTM, for nickel extraction was tested at a demonstration facility at BPR in
South Africa. In late 1998, a technical and economic evaluation was conducted on the use of thermophilic microorgan-
isms for extraction of zinc from sulphide concentrates. This process was named BioZINCTM and played a major role in
the Lanping project in China. Overall therefore, bioleaching satisfies the current industry requirements regarding ease
of construction, use, expandability and maintenance, environmental safety as well as competitive economics.
17.16.2 Oil Reparations

Large-scale oil spills, particularly in marine environments, occur occasionally and cause substantive damage to the
ecosystem in general and to the marine organisms in particular. Bioremediation is a technique that may be useful in
removing spilled oil under certain geographical and climatic conditions. The process of bioremediation is defined in
a broad way to include the use of nutrients to enhance the activity of indigenous microorganisms, or the addition of
naturally occurring nonindigenous microorganisms to respond to an oil spill. There is considerable controversy on the
use of genetically modified microorganisms in such situations. Most studies of bioremediation have involved natural
microorganisms such as the Psuedomonas species.
17.16.2.1 Methods
The success of oil spill bioremediation depends on the establishment of favorable conditions in the contaminated
environment. The first condition is that bacteria with appropriate metabolic capabilities must be present. If that exists
then their rates of growth and hydrocarbon metabolism need to be maximized by supplying adequate nutrients and
maintaining appropriate pH conditions. One has to remember that heavy crude oils contain large amounts of resin and
asphaltene compounds that are less amenable to bioremediation than light weight oils which are rich in aliphatic com-
ponents. The oil surface area is also important as the growth of oil degrading microbes occur at the oil–water interface.
17.16.2.2 Bioremediation in Marine Environment

In marine surroundings, contamination of coastal areas by oil from offshore spills usually occur in intertidal zones where
dissolved nutrients can be washed out easily. Therefore, design of effective oil bioremediation strategies and nutrient
delivery systems require an understanding of the transport of dissolved nutrients in the intertidal zone. Tracer studies
conducted on the shorelines of Mains, in 1994–1995, suggested that surface application of oleophilic (oil loving) nutri-
ents was ineffective on high-energy beaches as all nutrients were lost to dilution at high tide. Subsurface application was
a better approach. This nutrient should reach the oil-contaminated surface, where microbial growth could occur.
17.16.2.3 Bioremediation in Freshwater Environment

Oil spills have the greatest impact on wetlands and marshes. Studies are currently being conducted by the US EPA
and fisheries in Quebec to examine bioremediation with nitrate and ammonium in the presence and absence of wet
land plant species (Scirpis americanus). Technology for increasing oxygen concentration in these environments are still
undeveloped, and rely on wetland plants themselves to pump oxygen down to the rhizosphere through the root system.
17.16.2.4 Bioremediation in Soil Environment

A demonstration of an increased rate of oil degradation using biostimulation of soil was provided from a field study con-
ducted on the shoreline of Delaware Bay in 1994. However no significant difference was observed between plots treated
with nutrients alone and plots treated with nutrients plus indigenous inoculum of oil degrading microbes from the site.
Tilling, forced aeration and addition of peroxides are sometimes used for providing favorable conditions and enhancing
the rate of the bioremediation.
17.16.2.5 Conclusion
Research is currently going on to evaluate bioremediation and phytoremediation (plant assisted enhancement of oil
biodegradation) for their applicability to clean up oil spills contaminating salt marshes and freshwater wastelands. If
successful, it can be extremely cost effective and has the advantage that toxic hydrocarbons are degraded rather than
removed to another site. The main challenge is to maintain sufficient nitrogen and phosphorus levels for bioremediation
to occur. Studies show that application of dry granular fertilizer to the impact zone is probably the most cost-effective
833
17.17 Food Technology•
way to control nutrient concentrations. The Exxon Valdez spill that occurred in Alaska in 1987 formed the basis for a
major study on bioremediation through fertilizer application. Inipol (an oleophilic micro-emulsion with urea as a nitro-
gen source, lauryl phosphate as a phosphate source and oleic acid as a carbon source) and Customblen (a slow-release
fertilizer composed of calcium phosphate, ammonium phosphate, and ammonium nitrate within a polymerized vegetable
oil coating) were used. Within 2–3 weeks, oil on the surfaces of cobbled shorelines treated with Inipol and Customblen
was degraded so that these shorelines were visibly cleaner than nonbioremediated shorelines. The success of this field
demonstration programme has introduced bioremediation as a key technique in any clean up strategy which needs to be
developed for future oil spills.
17.17 Food Technology

Food processing is done in order to achieve the following objectives:
1. increased safety and shelf life;
2. improve the ease and reliability of preparation by consumers;
3. protect from contamination and thus help storage and transportation.
Various engineering principles are involved in these processes which need to take into account the biological activity
and chemical instability of foods and also the ability of microorganisms and pests to invade and grow on foods. Some of
the typical processes involve preservation of foods, changing the size, shape and texture of foods and the separation of
different food components.
17.17.1 Heat Transfer

Most food processing operations involve heat transfer, the major problem being the temperature sensitivity of foods.
Food components rapidly decompose in undesirable ways at high temperatures. The highest temperatures used are for
roasting coffee (∼250 °C) though the media used for heating (typically air) may reach up to 500 °C. The maximum
tolerance temperature for most foods is typically much lower. The other problem associated with heat transfer is the
temperature gradient that exists between the surface and the inside of foods. Foods are often solid or complex composites
(e.g. multicellular tissue, porous materials, stiff doughs) that act like solids. Heat transfer especially unsteady-state heat
transfer requires solving complex differential equations to get an idea of the temperature profile (see Box 17-2).
Similar equations need to be developed for mass transfer rates (since moisture removal usually accompanies heat
transfer). One also needs to account for heat generated by ongoing biological processes, surface evaporation, anisotropic
conduction, etc.
Box: 17-2
Consider a slab though which heat flow is taking place (Figure 17-41).
Consider a thin section of thickness Δx at a distance x from one end (shaded portion). The heat flux entering at the LHS
is given by
qin = kA(dT / dx)x .
Similarly the heat flux leaving from the RHS is

qout = kA (dT / dx)x + ∆x ,
where the temperature gradient is a function of x.

Using the overall balance equation, accumulation = input − output, we get
mc p dT / dt = kA (dT / dx)x + ∆x − kA(dT / dx)x .
Since m = ρA dx, substituting we have

dT / dt = (k / rC p )d 2 T / dx 2
This is the Fourier equation for unsteady-state heat transfer in one dimension, which becomes dT/dt = a∇2T for three-
dimensions, where α = k/ρCp. This equation can be solved for different boundary conditions.
x The boundary conditions are often nonuniform, thus we need

to take recourse to numerical methods which have computer-based
solutions. There are techniques available for calculating temperature
qin qout changes during microwave and dielectric heating of foods as well.
17.17.2 Food Preservation

Temp. This is possibly the most important part of food processing. The tech-
profile Direction of
heat flow nique used are drying, chilling, freezing, heat-induced inactivation of
dT microorganisms and enzymes, application of bacteriocides, storage in
dx x controlled atmospheres, solute induced water activity and pH reduc-
dT tion. A combination of techniques may also be used.
dx x+x
17.17.2.1 Drying
Drying, the evaporative removal of water from moist solids or solidifi-
able solutions, is one of the earliest method of preserving foods. This
Figure 17-41 Temperature profile for unsteady- preservation takes place because biological activity (typically micro-
state heat transfer through a slab. bial growth) slows down drastically or stops when moisture levels fall
below 0.15 kg water/kg of dry solids. Hot air is the most common
medium used to remove moisture. Immediately after contact between the moist food and hot air the food surface tem-
perature adjusts till it reaches steady state which is given by the wet bulb temperature of the air. The drying rate also
becomes constant and this is referred to as the ‘constant rate drying period’. This period ends when the moisture content
of the solids falls below a certain level (called the critical moisture content). The drying rates fall rapidly and the surface
temperature of the food rises. For large food particles this period called the falling rate period takes a far longer time than
the constant rate period.
During the constant rate period the entire solid surface is saturated with water. Drying thus proceeds as if from a pool
of water and the solid has little role to play in the drying rate. As drying proceeds the rate of liquid movement to the
surface from the inside (by capillary action and diffusion) becomes slower than the rate of mass transfer from the surface.
Consequently the surface starts to dry and the drying rate starts getting controlled by the diffusion rate. Since this rate
of diffusion is concentration dependent, the drying rates fall in the falling rate period. Drying also cases shrinkage in
foods making a theoretical analysis of drying rates difficult. The drying air temperature and humidity affect food quality
in addition to affecting drying rates and thermal efficiency. Therefore air conditions are adjusted to minimize quality loss
and prevent in - process microbial growth.
Foods are air dried in many different types of equipment. These include – shallow-bed dryers where the food is moved
in perforated conveyor beds; deep-bed dryers where drying takes place in bins (mostly used for grains), spray driers which
are used to dry liquid foods and food slurries, freeze dryers used to dry beverage extracts, shrimp soup ingredients, etc.
17.17.2.2 Chilling
Chilling, where food is not frozen, is also widely used to extend the shelf life of fresh produce like meat fish or diary prod-
ucts. Temperatures close to 0 °C and humidities between 85–95% are usually recommended. Data for heat capacities
& heats of respiration are available for most fresh foods which allow computation of the refrigeration loads. Shelf life
can vary widely from 2 days to 12 months, though typically for a majority of products the shelf life is less than 1 month.
Chilling is combined with the use of controlled atmospheres with low oxygen (2–5%) & up to 5% CO2 which
increases the shelf life of apples, pears & cabbage. This technique is also used for long distance shipping of meat.
17.17.2.3 Freezing
Freezing preserves food by reducing the rate constants for chemical reactions, reducing reactant mobility and reducing
the water activity (aw). The water activity of a frozen food depends only on its temperature as long as the food is moist:
aw = 0.908 at −10 °C; 0.824 at −20 °C, and 0.748 at −30 °C.
Foods are frozen using a cold stream of refrigerated air, direct contacting with brine-ice mixtures, contact with inert
evaporating refrigerant or cryogenic gas or by compressing them between refrigerated plates. Most water in moist foods
will freeze but water bound to food solutes or solids in nonsolvent form will not freeze even at very low temperatures.
Food solutes also depress food freezing points. Equilibrium initial freezing points usually lie between −1 and −2 °C and
supercooling of 5 to 6 °C is required before ice nucleates and freezing starts. As freezing proceeds solutes concentrate in
the residual nonfrozen solution further reducing the freezing temperature. Freezing of water in foods thus takes place over
a range of temperatures. If freezing is slow, water gets transferred from cell interiors (especially in fruits and vegetables)
835
17.18 Enzyme Engineering•
and forms extracellular ice before nucleation occurs inside cells. This adversely affects the texture of foods. Rapid
freezing allows intracellular nucleation to take place and most of the water freezes inside the cells minimizing adverse
textural affects.
17.17.2.4 Thermal processing

Commercial sterilization is designed to kill substantially all microorganism that can grow or spores that can g erminate
at storage conditions. Because the sterilized food is packed in sealed containers that contain no air, anaerobic organisms
are of greatest concern. Destruction of Clostridium botulinum spores is often used to evaluate the efficacy of sterilization
for low-acid foods (pH > 4.5). The fractional survival must be less than 1 × 10−12 for such spores. For foods with pH < 4.5,
facultative anaerobes like Bacillus coagulans and B. mascerans are used as marker organisms. Pasteurization where
temperatures less that 100 °C is used can help inactivate enzymes, most pathogenic microbes and many food spoilage
organisms. It is usually used in combination with protective packaging and other treatments (like chilling) to greatly
retard subsequent microbial growth.
17.18 Enzyme Engineering

(Please read the chapter on Basic Enzymology (Chapter 5) before studying this section)
The most common expression for enzyme kinetics is the Michaelis Menten kinetics given by;
Vmax S
V=
Km + S
Where ‘V’ is the rate of product formation or reaction velocity. Note that the shape of this function is asymptotic,
i.e. velocity increases with increase in substrate concentration when substrate concentrations are low, (like a 1st order
reaction) and velocity saturates to a maximum value Vm when substrate concentrations are high (zero order reaction)
Vmax
V
Zero order
1st order
Km S
Vmax thus represents the maximum reaction velocity for a given amount of enzyme ET
i.e. Vmax = k 2 E T when substrate is not rate-limiting.
Exercise : Prove that the reaction velocity is half of Vmax when substrate concentration = Km and v = ¾ Vm when S = 3 Km
The constant k2 is often called kcat and represents the rate of product formed per unit amount of enzyme, when
substrate is not rate limiting. It is also called the turnover number which is the number of substrate molecules converted
to product per second by a single catalytic site, with the units of s−1. Catalase for example has a turnover number of 4 ×
107 s−1. The Km terms can be understood as the affinity of the enzyme towards its substrate, with low Km values imply-
ing high affinity and vice-versa. Clearly if Km is small then ‘V’ would remain close to Vmax and not fall as the substrate
concentration declines as long as S>>Km.
The ideal enzyme would thus have a low Km and high kcat, attaching to substrate with high affinity and converting
them efficiently to product. Thus the catalytic efficiency of an enzyme is given by kcat/ Km. Remember that the substrate
has to reach the ‘active site’ of the enzyme by diffusion. Therefore the “perfect” enzyme which ‘immediately’ converts
the substrate to product would have a Kcat/Km value limited only by diffusion control. This sets the upper limit of its
value between 108 – 109 M−1s−1 depending on the micro environment of the enzyme. The most commonly used method
to estimate the kinetic parameters Vmax and Km is the Lineweaver Burke plot of 1/V vs 1/S. Since the equation
Vmax S
V=
Km + S
can be written as
1  K m   1  1 
=  +
V  Vmax   S   Vmax 
the double reciprocal plot has a slope of (Km/Vmax) and an intercept of (1/Vmax)
The method of generating such a plot is to first plot the product concentration versus time for a given amount of
enzyme using different starting concentration of the substrate. We then get a plot as shown below.
S4
S3
P
S2
S1
Here S4>S3>S2>S1 and therefore the rate of product formation rate is maximum with S4. This rate of product formation
is calculated by taking tangents and calculating its slope in the initial (early time points) part of the graph. This is not
a simple exercise especially when the Km values are small. Choosing low concentrations of the starting substrate can
cause it to decline quickly so that the velocity also declines sharply with time. This introduces errors in estimating the
slope of the tangent since we need at least a few time point samples for plotting the graph. On the other hand a high
starting value of substrate concentration leads to saturation (zero order) kinetics and thus no change is observed in ‘V’
at different substrate concentrations (as long as S>>Km). For example, if Km = 1mM then all So values of 5mM and above
would result in similar values of V (which will be close to Vmax). Conversely if the velocities (reaction rates) are around
1mM min−1 then choosing low So values would exhaust the substrate within minutes before sufficient time point samples
can be collected. A partial solution to this is choosing very dilute enzyme concentration to lower reaction velocities.
A second problem is uneven spacing of the data points while plotting the Lineweaver Burk plot. Taking reciprocals of
the ‘V’ and ‘S’ values cause the data points to cluster near the origin. Thus a straight line drawn to estimate Km and
Vmax, using linear regression would be very sensitive to errors in the smaller V and S estimates. (Since the errors in the
reciprocals would be large). We therefore use the Eadie Hofstee plots of S/V versus S (see below figure) and since;
S Km + S Km S
= = +
V Vmax Vmax Vmax
this line has a slope of 1/Vmax and an intercept of Km/Vmax.
Lineweaver Burke plot Eadie Hofstee Plot

0.021 0.12
0.019 0.1
0.017
0.015 0.08
0.013 0.06
1/V
S/V
0.011
0.01
0.009
0.007 0.02
0.005 0
0 0.5 1 1.5 2 2.5 0 2 4 6 8 10 12
1/S S
An even better solution is to utilize the integral version of the Michaelis Menten, kinetics where the equation
dS V S
V=− = max
dt K m + S
837
can be integrated as
Km + S
−∫ dS = Vmax ∫dt
S
S0
K m ln + (S0 − S) = Vmax t
S
Given the stoichiometry between substrate and product i.e., (S0 −S) = P, we can use this equation to simulate the
product profile w.r.t time for different Km and Vmax values using either excel or MATLAB software. One useful strategy is
to independently assess Vmax by calculating ‘V’ value with increasing substrate concentrations till it saturates. Then we
can fix Vmax and simulate the product profiles with different Km values till is best matches the experimental data.
17.18.1 Substrate and Product Inhibition and Enzyme Deactivation

Additional complications are introduced if the enzyme instead of following Michaelis Menten kinetics gets inhibited
by substrate or product. A comprehensive analysis is beyond the scope of this text, however simply put most substrate
inhibition kinetics have the form
Vmax S
V=
K m + S + S2 / K I
Thus with increasing substrate concentrations the plot looks like figure shown below
Here the velocity reaches a maximum which is lower than Vmax and then declines when the substrate concentration
is increased further.
Exercise: Determine the value of the maximum velocity possible in a substrate inhibited system and the substrate
concentration at which it occurs: Hint:- Differentiate the above equation to find its maxima].
Product inhibition kinetics can also be represented by a similar equation, i.e
Vm S
V=
K m + S + P2 / K I
However determining whether product inhibition is taking place requires a careful design of experiments. This is
because now the decline in ‘V’ is observed, not in the initial stage (as with substrate inhibition) but in the later stages
when product buildup takes place (see figure below)
t
From the figure it is clear that the product formation rate declines with time and this product formation rate can be
calculated by taking tangents(to calculate dP/dt) at different time points. However this decline in dP/dt) or ‘V’ can be
due to many reasons.
1) ‘V’ declines because the product concentration builds up and inhibits the enzyme
2) ‘V’ declines because substrate levels fall and this fall is sharper when [S] is close to Km
3) ‘V’ declines because of enzyme inactivation that takes place over time.
To distinguish between these 3 possibilities we do the following experiments
P P
t t
(a) (b)
If the product profiles with ‘high’ substrate and different enzyme concentrations is like ‘a’ above then clearly it
is product inhibited, since after reaching a certain product concentration (at different times due to different initial
velocities) the product does not increase further. However in ‘b’ above product concentrations do not increase after a
certain time has elapsed pointing to enzyme deactivation as the cause for ‘V’ going to zero. To ensure that substrate is
not rate limiting in the above experiments we need to perform the experiment alluded to earlier in the text, where ‘low
concentrations of enzyme are used and substrate concentrations are increased continuously till no significant increase in
velocity is observed (implying zero order kinetics w.r.t substrate).
Biochemists often report specific activities of enzymes as moles of product formed per unit time. This implicitly
assumes a constant velocity during the time period of measurement and zeroth order dependence on substrate concentra-
tion. As is clear from the above discussion, any change in starting substrate concentration or even the amount of enzyme
used can change these assumptions and introduce non-linear rates of product formation. One of the reasons why specific
activity measurements of enzymes like cellulases and xylanases have significant variations is precisely because of these
non-linear affects. Thus with a crystalline insoluble substrate like cellulose, Km values tend to be high and the velocity
is critically dependent of initial substrate concentrations.
For industrial uses therefore enzymes need to be characterized much more thoroughly than simple specific activity
measurements.
17.18.2 Immobilization of Enzymes and Industrial Applications

In an industrial setting, enzymes need to be reused to lower the cost of production. This is achieved simply by i mmobilizing
the enzyme on a matrix and packing it in a column where substrate enters from the bottom and product is collected
from the top. This ensures continuous operation and hence requires a much smaller volume of the enzyme reactor. The
reactor is run continuously until the enzyme undergoes de-activation due to a variety of reasons including leaching from
the matrix. The operational stability of the enzyme in a realistic industrial situation can be very high, allowing such
reactors to be run for more than 100 days. This effectively ensures that the enzyme catalyst is reused multiple times, thus
reducing enzyme costs in the total production process. Scaling up such systems requires a careful analysis of the kinetics
of this process.
The substrate enters at a fairly high concentration in the bottom of the reactor and this concentration declines as it
reaches the top. For enzymes with low Km values, it is expected that the substrate concentration would decline almost
to zero and we would obtain close to 100% conversion. When the substrate concentrations decline to values close to
Km, the enzymatic reaction rate shifts from zero order to first order with respect to substrate concentration (as explained
earlier). Therefore the effectiveness of the enzyme reaction is higher at the bottom of the reactor than at the top. This is
the reason why tall tubular columns are used for conversion, rather than a CSTR where the outlet substrate and product
concentrations are the same as the contents of the whole reactor. This becomes even more critical when the reaction is
product inhibited. As product concentrations increase along the length of reactor the reaction rates drops faster leading
to a mix of fast and slow reaction rates within the reactor. Finally axial dispersion needs to be taken into consideration.
The inlet liquid does not flow like a perfect plug, we need to take into account that axial mixing causes some of the mol-
ecules to travel faster and hence stay in the reactor for shorter time periods. All these factors tend to reduce the overall
conversion efficiency.
839
Diffusion processes play an important role in the design of such reactors. In the simplest case we consider enzymes
immobilized on the surface of solid matrices. If the bulk concentrations of substrate is So we can model this diffusion
process by proposing a thin film on the surface of the matrix through which the substrate diffuses to reach the enzyme,
which ‘sees’ a lower substrate concentration ‘S’. The rate of diffusion is given by
N s = k s (So − S)
where S and So are substrate concentrations at the interface and in bulk fluid respectively, and Ks is mass-transfer coeffi-
cient. At Steady State it matches the enzyme reaction rate. Thus
Vmax S
k s (So − S) = V =
Km + S
Introducing dimensionless variables

S V K
x= , Da = max , = m
So K s So So
This effectively reduces the number of parameters (or variables) from 4 to 2. Solving we get
(1 − x ) = x
Da k +x
where x lies between 0 and 1.
Da is also called the Damkohler number and represents the ratio of the maximum enzymatic reaction rate to the
maximum diffusion rate. Thus low values of Da imply that the system is reaction rate limited (since reaction rate is low
compared to the rate of diffusion). Conversely a diffusion rate limited system would have high Da values. Note that the
velocity of the reaction no longer exhibits a Michaelis Menten dependence on bulk substrate concentration. We rather
define an effectiveness factor η as
observed reaction rate
h=
reaction rate if there was no diffusion ratee limitation
x / (k + x), which is less than one

η=
1 / (k + 1)
On the other hand when Da is close to zero, η = 1.
The process of immobilization (especially when covalent linkages are used) can change the protein structure in subtle
ways so that the Vm and Km values are no longer same as the free enzyme. These new values of Vm and Km need to be
determined by experiments where Da is close to zero so that the effect of diffusion does not cause errors in the estimation
of these parameters.
For a completely diffusion limited system where Da is very large we have
(1 + k )
η= and V = K s So
Da
which implies that the reaction is first order with respect to bulk substrate concentration, and not dependent on enzyme
kinetics.
In many cases we wish to load a large amount of enzyme on the matrix support which requires a large specific surface area
of the matrix. Reducing the size of the matrix beads is not viable option since pressure drops in tall columns can become
very large with small beads. Thus porous matrices are often used which provided a much larger surface area for enzyme
immobilization. The effective diffusion now has to incorporate the effect of small pore diameters, tortuosity of the pores and
blockage of various pores by solids. Thus diffusion rates are effectively much slower and substrate concentration declines
drastically as it diffuses inside the pores. Under certain situation this substrate concentration may go to zero even before
it reaches the central core of the matrix because of fast reaction rates coupled with slow diffusion. In effect much of the
enzyme immobilized in the central part of the core then remains unutilized. The effectiveness factor η is again defined as
observed rate
η=
rate in the absence of concentration gradients in the pellet
Two dimensionless numbers are defined; the Thiele modulus φ which is given by
R Vmax / K m S
f= and b = o
3 Des Km
where Des is effective diffusion coefficient for substrate and β is the magnitude of saturation. It is observed that the
effectiveness factor η is a function of these two dimensionless numbers, where values of φ < 0.3 gives η values close to 1
implying that the system is controlled by the enzymatic reaction rate while values of φ > 3 makes η inversely proportional
to ϕ implying that the system is controlled by diffusion.
Estimation of which factor is rate controlling can be done by choosing porous particles of different sizes. Since large
particles will give large φ the system will get diffusion controlled while small particles will be controlled by the enzymatic
reaction rate. Simultaneously we can vary the bulk concentration So and measure the corresponding reaction velocities.
Given that for large So the reaction is zero order with respect to substrate we can calculate Vm when there is no diffusion
control (for small particles). This data can then be used for measuring the change in V when the particles are larger or
when So is small to determine the effect of diffusional limitation
The above analysis ignores the effect of film diffusion and in a realistic situation we need to assess the relative values
of the film mass transfer, diffusional rates within the pore and reaction rate to determine which of these is the rate lim-
iting factor in enzyme kinetics.
Keywords
Bioreactors and the fermentation process Downstream processing

Material balances Whole cell immobilization
Heat and mass transfer Bioproduction of chemicals
Bioprocess parameters Mineral beneficiation and oil reparation
Media design and sterilization Food technology
Suggested Additional Readings
Aiba. S., A.E. Humphrey and N.F. Millis (1973) Biochemical Stanbury, P.F. and A. Whitaker (1989) Principles of
Engineering, University of Tokyo Press, Tokyo. Fermentation Technology, Pergamon Press, Oxford.
Atkinson, B. (1982) Biochemical Reactors, Pion Ltd., Nielsen, J. and J. Villadsen (Eds.) (1994) Bioreaction
London. Engineering Principles, Plenum Press, New York.
Baily, J.E. and D.F. Ollis (1986) Biochemical Engineering Schuler, M.L. (Ed.) (1989) Chemical Engineering Problems
Fundamentals, McGraw-Hill Book, New York. in Biotechnology, AICHE, New York.
Enfors-Haggström (2000) Bioprocess Technology: Lee, J.M. (1991) Biochemical Engineering, Prentice Hall,
Fundamentals and Applications, KTH, Stockholm. Englewood Cliffs, NJ.
Jackson, A.T. (1991) Process Engineering in Biotechnology, Vieth, W.F. (1994) Bioprocess Engineering–Kinetics, Mass
Prentice Hall, Englewood Cliffs, NJ. Transport, Reactors and Gene Expression, John Wiley &
Shuler, M.L. and F. Kargi (2002) Bioprocess Engineering: Sons, New York.
Basic Concepts, Prentice Hall, Upper Saddle River, NJ.
Model Questions
For answers see the sections mentioned following the questions.

1. A CSTR is run and the following results areobtained Given that sF = 3 g l−1, determine the values of µm, Ks,
Yxtrue
/ s , ms. Hint: Plot 1/D vs. 1/s to determine µm and Ks;
D(h−1) S(gl −1) X (gl −1) p(gl−1)
plot 1/Yx/s vs. 1/µ to get Yx/s and ms. Use x = Yx / s (sF − s )
0.1 .08 .648 1.361 to get Yx/s.
0.2 0.2 0.8 0.88 2. A two-stage reactor is run as shown in the figure.
0.3 0.4 0.82 0.629 Given that mm = 0.6 h−1, Ks = 1 gl−1, Yx/s = 0.5, g/g (no
0.4 0.8 0.733 0.44 maintenance) and a non-growth product formation
0.5 2.0 0.345 0.1725
with β = 0.1, determine outlet product concentration
841
Model Questions•
from second reactor. Would the answer change if we If the same broth is now to be filtered in an industry
used only one reactor of (2+2) = 4l. using a plate and frame filter of 10 m2 surface area till
the cake is 3 cm thick. Determine volume processed
F0.5 lh1
sF 20 gl1 and time required per batch. [Hint: Plot reciprocal of
F1lh1 filtration rate vs filtrate volume to determine filtration
sF 10 gl1 constants.]
7. Distillation of a two-component mixture (A and B) is
carried out in a batch mode. If α = 1.5 and the initial
mixture has a 50:50 ratio of both components, deter-
mine the composition of the distillate when 60% of
the distillate has been collected. For the same mix-
V2 l V2 l ture determine graphically the minimum number of
plates required (at total reflux) to get a distillate con-
[Hint: Use the material balance equations to derive the centration of 95% A, 5% B and a bottoms product
outlet biomass and substrate concentration from the of 90% B and 10% A in a distillation column. [Hint:
first reactor; Use biomass balance on second reactor, Plot the equilibrium curve, since the operating line
overall balance on substrate and the kinetic equation at total reflux corresponds to the 45 °C diagonal the
to get biomass and substrate at the outlet of the second minimum number of plates can be easily determined.]
reactor. Use product balance to determine product 8. A 100 l reactor gets sterilized (as per design c alculations)
concentration.] by holding it for 5 min at 121 °C. If this sterilization is
to be scaled up to 50,000 l what should be the hold-
3. A waste treatment external recycle reactor has to run ing time. (State your assumption clearly). Given are
so that the outlet biomass concentration xout should not K121 °C = 2.54 min−1 ∇121 °C = 12.5 (Richards). [Hint:
/ s = 0.5 g/g, ms = 0.1
exceed 5 gl−1. Feed, SF = 20 gl−1, Yxtrue Note that the ∇ factor increases since the initial cell
h−1, µm = 1.0 h−1, Ks = 20 mg l−1 and D = 0.5 h−1. Determine count increases by a factor of 500.]
the fraction of cells to be recycled if the cells are con- 9. An acid-producing fermentation is controlled at pH 7
centrated 10-fold in the sedimentation tank. (Section by a proportional controller with a proportional band
17.5.10) (PB) = 2. The pH meter reads 6.7 continuously. Why?
4. Explain how an enrichment strategy using continuous Would this reading change if PB = 1? Can the PB be
culture techniques is better than serial inoculations. reduced further? Discuss. (Section 17.9.2)
How would this help in screening for desirable charac- 10. In a dynamic gasing out method air is shut off and the
teris other than the yield of product. (Figure 17-34 and DO reading recorded at 20 intervals. These are given
accompanying text) below.
5. How would the choice of media differ when industrial
alcohol is being produced in comparison to production DO (%) 80, 59, 37, 17 (after 10 s), 9 (air put on).
of recombinant therapeutic protein. Discuss the sim- DO (%) 39, 53, 63, 70, 76, 78.
plex algorithm for determining optimum media con- Determine KLa. If air is bubbled at 1vvm what outlet
centrations. (Section 17.11 and also references at the oxygen concentration would you get.
end of the chapter) Hint: Plot ln/ (C / L − CL )min / (C / L − CL ) vs. time to
6. Filtration carried out using a plate and frame filter of cul- get KLa.]
ture fluid gave the following data. Filter area = 250 cm2, 11. Given the KLa, α(P/V)0.8 (Rs)0.3, do a scale-up on
cell density = 7 g l−1 in culture fluid, dried cake den- the basis of constant KLa from a 5 l to 10,000 l
sity = 1.5 g cm−3, ΔP = 2atm (constant). fermentor given that the air flow rate is reduced
from 1 vvm to 0.5 vvm. By what ratio would the rpm
Volume of filtrate(l) Time (s)
change?(Section 17.8)
0.2 1.8 12. Glucose and yeast extract (YE) are to be optimized
0.4 4.2 using Box–Wilson design. The range of concentration
0.6 7.5 to be used is 0.5 to 2% and 0.1 to 1% for glucose and
0.8 11.2 YE respectively. What are the concentration to be
1.0 15.4 used in the individual experiments. IF high glucose
1.2 20.5
concentration repress product synthesis while high YE
1.4 26.7
1.6 33.4
concentration favour product formation how would
1.8 41.0 the contour plot look. (Section 17.11.1.2)
2.0 48.8 13. Design a sterilization protocol for a 10,000 l
2.2 57.7 fermentation media containing 103 cells per ml ini-
tially. To prevent nutrient loss at least 50% of the
sterilization should be during the holding period 20. A recycle reactor is run where the outgoing cells from
at 121 °C. Given at T = 121 °C, k = 2.538 and V121 the reactor are fed back in a highly concentrated form
°C = 12.549 (Richards). [Hint: Assume equal rates of to the reactor. Given mm = 1.0 h−1 of KS = 20 mg l−1,
heating and cooling and therefore equal contribution D = 0.5 h−1 of So = 10 g l−1Yx/s = 0.5 g/g the following
to the ∇ factor.] results are obtained when different percentages of the
14. A fermentation running at 37 °C is to be maintained outgoing cells from the reactor are fed back.
at the temperature using cooling water at 12 °C. If
80% of the heat generated is due to metabolic energy % fed back Product concentration (gl−1)
(@ 5kcal/l/h). Determine the cooling surface area 10% 2.25
requirement per m3 of fermentor volume. Given: Rise 20% 2.51
in cooling water temperature = 5 °C, Uoverall = 1000 w/ 30% 2.86
m2 K. [Hint: See text for worked example.] 40% 3.33
15. Yeast cells. (CH1.8 O0.5 NO2) are grown anaerobically
to produce ethanol. The carbon source is glucose Determine product formation kinetics (α and β).
and the nitrogen source is glutamate (C5H9O4N). If [Hint: Given Ks is small the outlet biomass
the true yield ( Yxtrue concentration would not change (overall balance)

/ s ) on glucose under these circum-
stances is 0.1 g/g and ms = 0.1 g/gh. Calculate the prod- while the µ would change since D is constant and
uct yield Yp/x at a specific growth rate of 0.1 h−1. [Hint: recycle is changing. Use the product balance equation
Calculate Yxobs (αμ + β)x = Dp to graphically determine the result.]
/ s and then use stoichiometry.]
16. E. coli cells (mm = 1.0 h−1, Ks = 10 mg l−1, Yx/s = 0.5 g/g) 21. A plug flow reactor has to be designed to treat liquid
are grown in fed-batch culture using constant feed @ waste comprising 5% starch at the rate of 106 l/year.
200 ml h−1 Sfeed = 10 gl−1 after growing the cells in batch Given outlet starch concentration has to be ≤0.5%,
culture Vo = 2l, So = 10/g/l). The fed batch is done for determine the reactor size if the organism has a
10 h. Assuming negligible product concentration at mmax = 0.5 h−1 and Ks = 100 mgl−1. Assume that 5% of
the beginning of feed calculate final product concen- the outlet biomass is recycled with the inlet feed.
tration given α = 0.2 and β = 0.1 (state your assump- If a CSTR were used would the size change? Given
tions clearly). [Hint: Since Ks is small assume all the Yx/s = 0.5 g/g.(Section 17.5.3)
feed is consumed leading to linear growth profile. 22. Pichia pastoris is grown on a mixture of methanol
Hence set up the integration to determine product YE (which comprises 50% protein having an approx-
concentration.] imate formula C5H9O4N and the remaining as min-
17. Cooling water enters at 15 °C and leaves at 25 °C erals and vitamins). If only biomass is formed what
thereby removing the metabolic heat from a CSTR yield would you expect in terms of gms of biomass per
being run at 37 °C. If m = 0.2 h−1 and x = 25 gl −1 gram of methanol. Also what would be the stoichio-
calculate cooling coil surface area per m3 of fermen- metric balance between methanol and YE. What is
tor volume. Given Ui = 1000 w m−2 k−1 the carbon the O2 consumption per gm of biomass formed. (Hint:
source is glucose, nitrogen source is ammonia and the Assume hH = 0.6 or Yx / s = 0.6 .
enthalpy efficiency = 0.6. (Section 17.6) 23. An anaerobic culture of Zymomanas mobilis is run on
18. Derive the expression for KLa using the two- film glucose and NH3. If Yxtrue / s = 0.2 C-mole/C-mole and
theory of mass transfer and Henry’s law. Given dis- ms = 0.2 C-mole/C-mole h, determine outlet biomass,
solved oxygen (DO) concentration at saturation substrate and ethanol concentration at a dilution rate
(100%) = 8 mg l−1 and the fermentor is to run at 40% of 0 h−1. Given S0 = 10 g l−1, m m = 0.5 h −1 K s = 200 mgl −1
DO. Calculate the maximum biomass concentration (Hint: Determine Yxobs / s and use the stoichiometric rela-
it can attain at a specific growth rate of 0.3 h−1 if tionship to determine Yp/s)
KLa = 100 h−1. Given 24. What kind of yield coefficient (Yx/s) would you predict
for cells growing on (a) glucose; (b) methanol. State
(a) carbon source is glucose; clearly the principle you would use in each case. What
(b) carbon source is methanol. values of Yx/o would you get. Assume that no product
(State your assumptions. Hint: Note that Yx/o is is formed and NH3 is the nitrogen source. (Hint:
d ifferent for different carbon sources.) Determine whether the system is enthalpy limited or
19. Scale up is done keeping KLa constant from 1 l to carbon limited.)
106 l using the correlation KLa α(P/V)0.7 (Vs)0.2. If the 25. Determine the reactor volume required to produce
air flow rate/unit vol. (vvm) is kept constant during a product X @ 1 ton/yr. Given that the product is
scale-up, how does the power consumed/unit volume growth-associated with Yp / x = 0.2g/g and the batch
change. (Section 17.8) reactor is run with S0 = 50 gl −1 to a point where
843
Model Questions•
99% of the substrate is consumed; also given that concentration and the value of Yx/s to get inlet sub-
Yx / s = 0.5g/g and 1 year has 300 working days; the strate concentration. Determine OUR and hence the
specific growth rate is 0.5 h −1 and K s is 20 mgl −1. State outlet oxygen concentration.)
your assumptions clearly. 31. Yeast cells grown anaerobically with NH3 as the
nitrogen source produce 0.1 g cells/g of glucose.

(Determine the product concentration at the end of What values of Yp/s do you expect. (Hint: Use the
the run. Also determine the time required for one run stoichiometric equation.)
by noting that the substrate concentration at the end 32. If Yxtrue
/s is 0.5 g/g and ms = 0.2, design an external
of the run is still higher than Ks.) feedback CSTR which gives a Yx/s of 0.3 g/g at a
26. A CSTR is run and the following values are obtained dilution rate of 0.5 h −1 . Given the concentration

by the researcher assuming that it is an ideal well- factor in the external sedimentation tank is equal to
mixed reactor, m m = 1.0 h −1 , K s = 20 mgl −1 Yx / s = 0.5g/g 10, determine the recycle ratio (R). Also determine
ms = 0. However it turns out that due to poor mixing Yo/x and heat dissipated per gm of cells formed. (Hint:
the concentration of cells in the bottom half of the See worked example in text.)
reactor is twice the top half from where the outlet is 33. Cells growing on glucose and glutamic acid as nitrogen
connected. How will the above parameters be reevalu- source has a Yx / s = 0.7 g/g . No product is formed.
ated taking this nonideality into consideration. (Hint: Determine Yo/x and ηH. (Hint: Use the material and
Set up the material balance equations as in the case of energy balance equation.)
internal recycle.) 34. Determine KLa if the outlet air contains O2 at a partial
pressure of 0.16 atm. Air flow rate = 0.5 vvm, C∗ = 8 mgl −1
at 100% saturation and C (steady state) = 50%. (Hint:
Use the gas balance equation to determine OTR.)
35. Scale up a process from 1 l to 10,000 l with constant
KLa a(P / V)0.8 Vs0.3 . Given that the initial air flow
rate = 1vvm and N = 600 rpm and final air flow
X rate = 0.5 vvm, determine final rpm. What would
be the final rpm if scale-up is done on the basis of
constant (P/V). (Hint: See worked example in text.)
2X 36. A 5l CSTR is to be operated at a dilution rate (D)
of 0.3h−1 using inlet glucose concentration (S0) = 10
27. What is meant by quasi-steady-state in fed-batch reac- gl−1 and ammonia as nitrogen source mm = 1 h−1 and
tors. Derive the expression μ = D for fed-batch and Ks = 50 mgl−1. Cooling water is available at 15 °C with
explain. (Section 17.5.11) a flow rate of 100 ml min−1. Calculate the length of
28. A CSTR is run at a dilution rate of 0.2 h −1 where the tubing required to maintain the fermentor at its opti-
outlet biomass concentration is 20 gl −1. The substrate mum temperature of 37°C by removing the heat of
is glucose, ammonia is the nitrogen source. The true fermentation. Clearly state your assumption. Given
−1
/ s = 0.5 g/g, while m s = 0.2 h . Determine
yield is Yxtrue that tube is thin copper of diameter = 5 mm, m = 1
the surface area of cooling required per liter of reac- cp (0.001kg/ms) of water, k = 0.15 cal/sm2 (°C/m),
tor volume given that reactor operating temperature Cp = 1 cal/g °C(1000 cal/kg°C), Nu = 0.023 Re0.8(Pr)1/3
= 37 ° C , cooling water inlet = 10 ° C, outlet = 15 ° C, (for internal heat transfer coefficient) and
U = 100 wm −2 k −1 and the heat removed is essentially h0 = 500 cal / sm 2 °C . (Hint: Determine hi from Nu,
the heat produced due to fermentation. (Section 17.6) hence determine U calculate qmet and therefore the
29. Determine Yo/x for cells growing on (a) glucose and rise in temperature of cooling water, calculate LMTD
(b) methanol if Yx / s = 0.5 g/g. What is the ηH in each and hence area of cooling required.)
case. What is the heat generated (D)? (Hint: Use the 37. A two-stage CSTR as shown in figure is used to grow
material and energy balance equation.) recombinant E. coli. The cells are grown in the first
30. In the previous question if the cells were growing in a stage and induced in the second with addition of
CSTR (D = 0.5 h −1) and the outlet biomass concentra- IPTG (small amount). Recombinant protein pro-
tion x = 1.5 mgl −1. Determine outlet substrate concen- duction is growth associated with α = 0.3. The μm
tration ( s ) and outlet oxygen concentration. Given of induced cells is half the μm of noninduced cells.
m m = 0.8 h −1 K s = 50 mgl −1, aeration rate = 1vvm , Calculate outlet product concentration given that it
what is the inlet substrate concentration (S0). (Hint: is efficiently secreted out into the medium. (Hint: See
Use the kinetic equation to get the outlet substrate question 2.)
So = 20 g l−1 15 °C and leaves at 25 °C. IF the fermentor operates

F = 0.5 ml h−1 at 30 °C and U = 500 W/m2 °C, calculate the surface
lPTG area of cooling. (Hint: Use the heat balance equation
F = 200 ml h−1 q = UALMTD.)
SF
43. KLa is estimated using the Dynamic gassing out
SF = 10 g l−1 method the following data is obtained.
µm (uninduced) = 0.5 h−1
Ks = 100 mg l−1 Time(s) (air shut off) 0 100 200 300 400 Air put on
Yx/s = 0.5 g/g
DO (%) 58.5 49.4 39.7 32 25
Time (s) 450 500 550 600 650
DO (%) 36.3 46.9 52.6 55.2 56.5
V = 1l V = 1l
Calculate KLa (Hint: See question 10).
44. A CSTR is run with wall growth where 50% of the
38. Cells are growing anaerobically and producing cells in the reactor are stuck to the wall. Determine
product. Would you choose a system having high or outlet biomass, product and substrate concentration
low-maintenance coefficient. Explain using the equa- given that D = 0.3 h−1, µm = 0.5 h−1, Ks = 0.2 gl−1, S0 = 1
tions for product formation. (Assume that true yield gl−1, Yx/s = 0.3 g/g, α = 0.2 and β = 0.1 (for growth-asso-
of biomass is similar and specific growth rate is also ciated and non-growth-associated production forma-
the same. Thus calculate the product yield for the two tion). (Sections 17.5.5–17.5.7)
systems.) 45. A continous sterilizer is used to sterilize media by
39. An yield coefficient (Yx/s) of 0.5 g/g is observed for instantaneously heating to 131 °C and quick cooling.
cells growing on glucose and NH3. Predict how a stoi- To prevent loss of nutrient quality it is decided to oper-
chiometric amount of glutamate (C5H9NO4) is to be ate the sterilizer at 140 °C. What is the fractional gain
used as a nitrogen source instead of ammonia. (Hint: in nutrient quality (given ΔEbacillus = 67.7 kcal/mole and
Assume the ηH remains constant.) ΔEnutrient degradation = 20 kcal/mole). (Hint: Determine the
40. Cells growing on glucose and NH3 produce 1.92 reduction in time since∇ values are the same. Hence
C-moles of ethanol/C-mole biomass and consume determine change in nutrient quality.)
4.26 C-moles of glucose/C-mole of biomass. Calculate 46. Penicillin is extracted from a fermentation broth using
the respiratory (RQ), that is the ratio of CO2 produced isoamyl acetate as the organic solvent in a continuous
to O2 consumed. (Hint: Use the material balance counter-current cascade extraction unit. The flow rates
equation.) of organic (L) and aqueous (H) phases are L = 10 l min−1
41. An anaerobic digester (waste treatment plant) gets
and H = 100 l min−1, respectively. The distribution coef-
a feed of 1001h−1 consisting of 10% carbohydrates
ficient of penicillin between the organic and aqueous
(CnH2nOn) and converts 99% to biomass and prod-
uct (ethanol). Calculate the heat produced by the phases at pH = 3 is KD = YL/XH = 50. If the penicillin con-
digestor per hr (assume nitrogen source is ammonia). centration in the feed stream is 20 gl−1, determine the
(Hint: Do anaerobic systems produce heat?) number of equilibrium stages required to reduce the pen-
42. A fermentor produces 10,000 kJh−1 of heat which is icillin concentration to 0.1 gl−1 in the effluent. (Section
removed by cooling coils where cold water enters at 17.13.7)

Chapter 17

Uploaded by

Copyright:

Available Formats

Chapter 17

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 17

Uploaded by

Copyright:

Available Formats

CHAPTER 17

17.1 Introduction to Bioprocess Engineering and Technology

17.2 The Component Parts of a Fermentation Process

17.3 Material Balance

Supernatant Figure 17-1 Tangential flow filtration device.

Figure 17-2 Tangential flow with recycle. Supernatant

Example 2 Let us now consider a situation where a chemical reac- A AB

By applying appropriate boundary conditions (CA = CA0 at t = 0) and solving we get

17.3.1 Material Balance in Biological Systems

F7 (H2O) F6 (CO2) F5 (Oxygen)

which requires choosing λN = −21.

Since Yx/s = −0.6 C-mole/C-mole, substituting this we get

substituting values of γs and γp, we obtain

the value of γx would change from 4.2 to 0.6.

Therefore rate of biomass production per liter of culture in (gh−1) is

17.3.2 Energy Balance in a Biological System

where the g i′ is used to denote that λN is chosen as zero.

which is a fairly good approximation of the real values.

(Note the g o′ remains equal to −4 and g N′ is no longer zero.)

We can multiply this equation by 115 kJ (C-mole)−1 to get

The LHS of this equation represents

For no product formed and NH3 as nitrogen source, this becomes

which can be approximated to

Figure 17-7 Effect of γs on enthalphy efficiency and

17.5 Kinetics of Microbial Growth

m = f (s, pH, temperature, O2 ,…).

when genetic stability is being studied as a function of the number

17.5.1 Growth Kinetics in Batch Culture

This equation can be integrated using the approach of partial fractions:

∫ {K Ys x/s + s0 Yx / s − x + x 0 )dx } / { x(s0 Yx / s − x + x 0 )} = ∫m m dt.

Solving and putting appropriate boundary conditions (x = x0 at t = 0) we get

This equation bears further investigation.

Thus the growth equation becomes

We can substitute various values of x in order to get ‘t’. The

17.5.2 The Ideal Plug Flow Reactor

17.5.3 The Ideal Continuous Stirred Tank Reactor

X The units of D are in time−1, usually h−1, which is the

At steady state, substituting µ = D, we get

Since D = m = mms/(Ks + s), we have

17.5.4 The Problem of ‘Maintenance’ Requirements

where mo is the maintenance requirement with respect to oxygen.

or Figure 17-16 Plot of steady-state biomass

How would you interpret this equation?

therefore at µ = 0.5 h−1

/ s = 1 / 0.4 + 0.2 / 0.5

Similarly at m = 0.05 h−1; Yxobs

and at m = 0.05 h−1,

17.5.5 Product Formation Kinetics

17.5.6 Modifications of the CSTR

17.5.7 Chemostats with ‘Wall Growth’

17.5.8 Chemostat with Recycle

17.5.9 CSTR with External Recycle

An overall balance gives

Substituting the values given we obtain

Since D = 0.2 h−1, we have

17.5.10 Fed-Batch Culture

17.5.10.1 Feed of Concentrated Media

17.5.10.2 Feed of Dilute Substrate

Considering volume changes to be negligible we have a substrate balance,

substituting the values we get

Example 2 Let us now consider a situation where a chemical reac- A AB

⇒ mCp ln(∆T1 / ∆T2 ) = UA, Tc