Você está na página 1de 15

1

1
8
4
Review
Received: 30 September 2009 Revised: 12 February 2010 Accepted: 12 February 2010 Published online in Wiley Interscience: 7 April 2010
(www.interscience.wiley.com) DOI 10.1002/jctb.2387
Bioprocess scale-up: quest for the parameters
to be used as criterion to move
frommicroreactors to lab-scale
Marco P. C. Marques,

JoaquimM. S. Cabral and Pedro Fernandes


Abstract
Advancesinhigh-throughput processdevelopment andoptimizationinvolvetherational useof miniaturizedstirredbioreactors,
instrumented shaken asks and microtiter plates. As expected, each one provides different levels of control and monitoring,
requiring a compromise between data quantity and quality. Despite recent advances, traditional shaken asks with nominal
volumes below250 mL and microtiter plates are still widely used to assemble wide arrays of biotransformation/bioconversion
data, because of their simplicity and low cost. These tools are key assets for faster process development and optimization,
provided data are representative. Nonetheless, the design, development and implementation of bioprocesses can present
variations depending on intrinsic characteristics of the overall process. For each particular process, an adequate and
comprehensive approach has to be established, which includes pinpointing key issues required to ensure proper scale-
up. Recently, focus has been given to engineering characterization of systems in terms of mass transfer and hydrodynamics
(through gaining insight into parameters such as k
L
a and P/V at shaken and microreactor scale), due to the widespread use
of small-scale reactors in the early developmental stages of bioconversion/biotransfomation processes. Within this review,
engineering parameters used as criteria for scaling-up fermentation/bioconversion processes are discussed. Particular focus is
on the feasibility of the application of such parameters to small-scale devices and concomitant use for scale-up. Illustrative case
studies are presented.
c 2010 Society of Chemical Industry
Keywords: scale-up; small-scale reactors; k
L
a; volumetric power consumption; fermentation; bioconversion
NOTATION
a Specic interfacial area (m
1
)
a
i
Initial specic surface area (m
1
)
a
f
Final specic surface area (m
1
)
Bo Bond number, D
2
gWt
1
()
d Maximum inside shaking ask diameter (m)
D Well or vessel diameter (m)
d
i
Diameter of drops in size class i (m)
D
i
Diffusivity (m
2
s
1
)
d
max
Maximum drop diameter (m)
d
n
Nozzle diameter (m)
d
o
Shaking diameter (m)
d
32
Sauter mean diameter (m)
Fr Froude number, d
o
(2N)
2
(2g)
1
()
g Gravitational constant (m s
2
)
h Liquid height (m)
k Number of size classes ()
k
L
Mass transfer coefcient (m s
1
)
k
L
a Volumetric oxygen mass transfer coefcient (s
1
)
N Shaking frequency, stirring speed (s
1
)
n
i
Number of drops ()
N
f
Pumping number ()
N
P
Power number ()
P Gassed power input (W)
P
o
Ungassed power input (W)
P/V Volumetric power consumption (W m
3
)
Q Volumetric gas ow rate (m
3
s
1
)
Re Reynolds number, Nd
o
2

1
, NT
2

1
()
Sc Schmidt number, (D
i
)
1
()
T Stirrer diameter (m)
u
o
Nozzle velocity (m s
1
)
V Filling volume (m
3
)
v
g
Supercial gas velocity (m s
1
)
V
o
Flask volume (m
3
)
v
tip
Impeller tip speed (m s
1
)
W Width of turbine blades (m)
We Weber number, N
2
T
3

1
()
Wt Wetting tension (N m
1
)
Subscripts
c continuous
T Tank
Greek symbols
Viscosity (Pa s)

Correspondence to: Marco P. C. Marques, IBB-Institute for Biotechnology and


Bioengineering, Centre for Biological and Chemical Engineering, Instituto
Superior T ecnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal.
E-mail: mpc.marques@ist.utl.pt
IBB-Institute for Biotechnology and Bioengineering, Centre for Biological and
Chemical Engineering, Instituto Superior T ecnico, Av. Rovisco Pais, 1049-001
Lisboa, Portugal
J ChemTechnol Biotechnol 2010; 85: 11841198 www.soci.org c 2010 Society of Chemical Industry
1
1
8
5
Bioprocess scale-up: parameters to be used www.soci.org
Density (kg m
3
)
Local energy dissipation rate (W kg
1
)
Volume fraction of the dispersed phase ()
Interfacial tension (N m
1
)

V
Viscous dissipation term (m
2
s
2
)
INTRODUCTION
Despite sharing a common pattern, the design, development and
implementation of microbial processes present subtle variations,
depending on the product, microbial strain, growth conditions,
bioconversion/biotransformationconditions, amongothers. Thus,
for a given product, process or facility, an adequate and
comprehensive approach has to be established that encompasses
the detailed characterization of the process and the timely
identication of key process parameters likely to affect product
yield, quality and consistency. These are to be kept as constant
as possible throughout the scale-up process, in order to ensure
success of the later task.
1,2
The development of microbial processes is strongly anchored
in small-scale reactors, typically Erlenmeyer type asks and bench-
scale reactors, which are used for strain screening, media design
and optimization and strategies of operation, ultimately aiming
for the highest attainable productivity.
3
In recent years, the
range of small-scale vessels has increased to include multi-well
plates (MWPs) with different levels of complexity and built-in
devices, and miniature reactors that clearly emulate the larger
vessels, but with a volume in the milliliter (or lower) range
4,5
Along with the technological developments that allowed for such
hardware, efforts have been made to ensure the data gathered
from experiments performed at these scales is reproducible
throughout scales. Such efforts have relied on gaining further
insight into mass (and heat) transfer and uid dynamics at
microliter scale, identication and validation of key parameters
for scale-up, preferably from MWPs to bench-scale bioreactor,
and predictive modelling.
6,7,8
Knowledge gathered allows full
advantage to be taken of the high level of parallelization
provided by most miniaturized devices (in particular MWPs),
and to speed up bioprocess development in a cost-effective
manner.
9,10,11
The task at hand is quite complex since there
are several parameters inuencing transport phenomena and
chemical dynamics within a bioreactor. These parameters relate to
mass transfer, mixing, partitioning, power input, shear induced by
agitation, dilution rates, substrate and products concentration,
nutrients, micronutrients and stabilizing agents, temperature
and pH. Parameters of (bio)chemical nature are screened and
optimized for bioconversion/growth kinetics. Physical parameters
are, on the other hand, conditioned by process design and
operational conditions. The parameters for scale-up are selected
within the range of physical parameters. Naturally favored
candidates to fulll such role are the process parameters
or coefcients that are known to have some effect on the
biological agent (enzyme or cell), particularly in their physiology.
These include those affecting oxygen supply, heat transfer and
mixing, namely aeration, agitation, mixing time, power input
and oxygen mass transfer coefcient.
2,12
In many cases, these
physical parameters have to be combined with each other,
or with other variables, in dimensionless numbers that are
kept constant throughout scaling, therefore establishing scale-
up criteria. In any case, the environmental conditions have to
remain constant.
1,2
Again, and despite the relevance of the
scale-up issue in biotechnology, there is no straightforward
and uniform strategy to tackle this matter. A suitable scale-up
approach has therefore to be established again on a casuistic
basis for a given product, process or facility.
1,2,12
Most of these
strategies, when bioconversion, fermentation or cell culture
processes are involved, rely on the use of k
L
a or volumetric
power consumption as criteria for scale-up, although constant Re,
constant impeller tip speed and equal mixing and recirculation
time are occasionally used.
2,13
Given the relevance of the
two former, they will be addressed in detail in this present
work.
SCALE-UP BASEDONK
L
A
Oxygen mass transfer coefcient
Oxygen is a key substrate in most microbial processes of industrial
relevance, where it can be required for growth, maintenance
or production of metabolites.
14,15
These processes are typically
performed in an aqueous environment, but oxygen is sparingly
soluble in water, roughly 0.272 mmol L
1
, at 25

C and 101 kPa


air pressure, and thus often becomes the limiting substrate.
16
A
suitable supply of oxygen to the liquid media, typically from air,
is mandatory, but the process of mass transfer is inuenced by
several variables, such as physical properties of the uids involved,
operational conditions and geometry of the reactor. The oxygen
transfer rate (OTR) can be increased by altering stirring speed and
gas ow, which concomitantly alters the power input. Increasing
the OTR is required to cope with the microbial oxygen demand,
the oxygen uptake rate (OUR). It is possible to assess the rate
limiting step of a microbial process, i.e. mass transfer or reaction
limited, through the use of a modied Damk ohler number (Da),
which is calculated as the ratio between the maximum oxygen
uptake and transfer rates.
17
Da =
OUR
max
OTR
max
(1)
A large q
O2
or low diffusivity leads to Da > 1, hence the
process is mass transfer limited; oppositely, small q
O2
or high
diffusivity results in Da 1, thus the process is limited by the
biochemical reaction rate.
18
Oxygen transfer rate is thus a critical
feature for the characterization of a given process and likewise
an engineering parameter suitable for the design, selection and
scale-up of bioreactors is to address OTR. One such parameter
is the volumetric mass transfer coefcient k
L
a, which relates the
oxygen mass transfer rate to the oxygen concentration gradient,
according to (2). In a bioprocess the oxygen mass transfer rate can
be described by
12,16
OTR
_
C

C
L
_ = k
L
a (2)
where OTR is expressed as the molar ux of oxygen through the
gasliquid interface; C

is the dissolved oxygen concentration


which would be in equilibrium with the gas phase, C
L
is the
dissolved oxygen concentration in the bulk liquid, k
L
is the
local mass transfer coefcient in the liquid phase, and a is the
specic interfacial area. Although oxygen is transferred from the
bulk gas phase to the bulk liquid phase, it is assumed that
the gas phase resistance to mass transfer is negligible. Oxygen
consumption through the process due to biochemical reactions
can be considered using a biological enhancement factor, E,
13,18
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
8
6
www.soci.org MPC Marques, JMS Cabral, P Fernandes
leadingtoanoverall volumetric mass transfer coefcient K
L
a given
by
K
L
a = E.k
L
a (3)
E incorporates the transport enhancement due to the oxygen
uptake by microorganisms, alongside with opposing mass
transfer resistances caused by layers of materials placed between
the gas bubble and the bulk liquid phase, namely adsorbed
surfactant and cells, and stagnant liquid lm.
18,19
E has been
shown experimentally to change with the concentration of
biomass, typically increasing with increased cell concentration,
but with varying patterns, according to different strains and
incubation media. In these dedicated experiments, E was within
0.81.3, although values as high as 5 were also reported.
13,18,19
Nonetheless, and for most cases, the biochemical rate is not
signicantly larger than the mass transfer rate, hence it is usually
assumed that E = 1. Details on this matter can be found in the
recent review by Garcia-Ochoa and Gomez.
13
The volumetric mass transfer coefcient in bioreactors can be
obtained experimentally or predicted using empirical correlations
for k
L
a, or for k
L
and a, which are thoroughly described
elsewhere,
13
coupled to a model for the estimation of E. The
introduction of the latter parameter in predictive determinations
has nevertheless been seldom reported, which may account for
shifts between predicted and experimental data at high biomass
concentrations. The use of constant k
L
a as scaling criterion is
widely disseminated in conventional scales, from laboratory to
production scales, encompassing volumes ranging from 1 L to
1000 m
3
, and ultimately has also been found to be suitable down
to the milli/micro-liter scale.
68,2025
Experimental determination of the volumetric mass transfer
coefcient (k
L
a)
Several methods have been developed to determine k
L
a
experimentally.
13,26,27
Within the scope of bioreactor design,
several items are considered: stoichiometry, thermodynamics,
microbial kinetics, transport phenomena (heat and mass transfer)
and economics. While the rst three items are scale-independent,
transport phenomena and economics are extremely dependent
on scale. The relevance of the transport phenomena in the design
and scale-up of the bioreactor is particularly noticeable, since the
overall rate of aerobic bioprocesses is generally controlledby mass
transfer rates. The mass balance for the dissolved oxygen in the
well-mixed liquid phase can be established as
2,12,13,16
dC
dt
= OTR OUR (4)
where dC/dt is the oxygen accumulation rate in the liquid phase.
The determination of OTR can be performed either when oxygen
is being depleted by growing biomass (direct methods) or when
nooxygenuptake takes place (indirect methods). Inthe latter case,
equation (4) reduces to
dC
dt
= OTR = k
L
a
_
C

C
_
(5)
The direct methods rely on oxygen probes, which allow for
the determination of OTR through gas phase analysis or through
the dynamic method. Until recently, only relatively bulky probes
were available, which prevented oxygen monitoring in miniature
vessels (MWPs, miniature/micro bioreactors), limiting its use to
bench scale and larger bioreactors. Recently Erlenmeyer-type
shaken asks were adapted in order to be equipped with oxygen
gas sensors, allowing for on-line determination of OTR in several
parallel experiments in shaken vessels.
28,29,30
Developments in
uorescence methods, (micro)fabrication techniques and optic
ber, have allowed for the implementation of sensitive dyes that
can either be inserted into a patch and adhered inside a vessel,
including individual wells from multiwall plates, yielding the so-
called sensor spots, or incorporated onto the tip of ber optic
probes.
3,4,11,25,3137
The indirect approach for the determination of OTR relies on
chemical or physical methods. The most commonly used among
the former is the cobalt-catalyzed oxidation of sulte, which was
optimized for application in miniature bioreactors (MWPs and
shaken asks),
38
and is routinely used in such formats.
33,39
A
methodbasedonCO
2
absorptionis more rarely used. The physical
methods, again relying on probes, allow for the dynamic method
of OTR determination.
13
Chemical methods, and in particular
the sulte method, may be biased due to modications in uid
dynamics, uid properties and surface tension, as a result of the
addition of chemicals. They may therefore lead to misleading
information and data gathered.
13,27,40,41
The use of fast enzymatic
methods has also recently been introduced within the scope of
the indirect approach. These are clearly designed for application
in miniature systems, such as MWPs.
42,43
These methods are
based on the use of glucose oxidase and, preferably, of catechol
2,3-dioxygenase. The latter method relies on a single-step well
dened stoichiometric reaction, whereas the former requires
calibration with the sulte method (or a similar one) but on
the other hand, all reagents are easily available. Among physical
methods, the dynamic method is the most commonly used to
evaluate k
L
a, because of its simplicity and relative accuracy. Both
the absorption and desorption measurements give equal values
of k
L
a under identical hydrodynamics conditions.
44
When the
characteristic time for the oxygen electrode and the characteristic
time for the oxygen transfer process (1/k
L
a) are of the same
magnitude, the dynamic response of the electrode has to be
considered in the determination of k
L
a. A detailed description
on the nature and limitations of the methods used for the
experimental determination of k
L
a is given by Garcia-Ochoa and
Gomez.
13
Empirical correlations for the determination of k
L
a in stirred
tank reactors
Both dimensional and dimensionless equations for the volumetric
mass transfer coefcient as functions of different variables have
been proposed.
45
There are, however, considerable problems
concerningthe accuracy of k
L
a estimation. Discrepancies between
experimental data and those estimated from these equations are
often found. The discrepancies are mainly found when k
L
a for
real broths are estimated from equations proposed for aqueous
solutions. This can be due to the strong inuence of the type and
sizeof thebioreactor, thedifferent rangeof operational conditions,
the system considered (solutions or real broths), or even the
measuring method used.
46,47
The addition of ions, hydrocarbons
or temperature increase k
L
a, whereas the addition of surfactants
or antifoams or increases in media viscosity decrease k
L
a, when
compared with data obtained fromwater.
26,4852
Vant Riet
26
proposed an overall correlation of k
L
a with
volumetric power consumption and supercial gas velocity:
k
L
a = C
1
_
P
V
_
C
2
_
v
g
_
C
3
(6)
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198
1
1
8
7
Bioprocess scale-up: parameters to be used www.soci.org
where P is the gassed power input, V is the liquid volume, v
g
the gas supercial velocity and C
1
, C
2
and C
3
are constants that
may vary considerably. Weuster-Botz et al.
53
used correlation (6)
topredict k
L
a ina magnetically drivenstirredcolumn, designedfor
parallel operation in an incubator chamber, operated as shaking
asks. Constants varied from 0.11 to 0.14, 0.06 to 0.37 and 0.73 to
0.22, for C
1
, C
2
andC
3
, respectively, accordingtodifferent turbines.
Mass transfer coefcients of up to 0.34 0.05 s
1
were obtained
in the reactors, using dened medium for growing E. coli, using
the dynamic gassing-out method.
Montes et al.
54
determined values of k
L
a in yeast broths
(Trigonopsis variabilis) over wide ranges of both impeller speeds
and supercial gas velocities, in three different mechanically-
stirred, bafed reactors (2, 5 and 15 L). Experimental data were
tted using Equation (6) and the values for the parameters C
2
, C
3
and C
1
were 0.35, 0.41 and 3.2 10
3
, respectively. Since most
of the yeast broths behave as non-coalescent uid, according to
the authors, the correlation improved the prediction of k
L
a values
with respect to other generic correlations usually developed for
strong coalescent and non-coalescent uids. Additionally, Shin
et al.
55
veried that in high cell density cultures of fast-growing
aerobes, such as recombinant E. coli, where the biomass may
increase to more than 70 gL
1
, the oxygen availability can be the
rate-limiting step of the fermentation process. Accordingly, the
followingcorrelationfor k
L
a incorporatingthe effect of cell density
(X) in oxygen transfer has been proposed:
k
L
a = 0.0192
_
P
V
_
0.55
_
v
g
_
0.64
_
1 +2.12X +0.2X
2
_
0.25
(7)
Extensive details on this matter can be found elsewhere.
13
Another approach for the estimation of k
L
a relies on the use
of empirical correlations incorporating dimensionless groups.
13,45
This approach has certain advantages because the correlations
obtained for a known system can be used to estimate k
L
a in other
systems with different dimensions.
5658
Althoughseveral correlations havebeendevelopedfor different
systems, most of them are not specic to fermentation broths or,
when developed with such a purpose, they do not take into
consideration all the variations of parameters (surface tension,
viscosity) throughout the time course of cultivation, which may
hamper its effectiveness.
Two-phase partitioning bioreactors have demonstrated signif-
icant potential for enhancing the productivity of many biopro-
cesses by overcoming issues of poor substrate solubility and
toxicity. The oxygen mass transfer coefcient can also be evalu-
ated in these systems. In order to take into account the effect of
the organic phase and the organic phase volume fraction on k
L
a
in aerated liquidliquid dispersions, empirical correlations have
been proposed, assuming that the two liquid phases behave as a
single homogeneous phase:
59
k
L
a =
_
P
V
_

_
v
g
_

(1 )

(8)
where is the volume fraction of the dispersed liquid phase, and
, , and are numerical constants.
Gomes et al.
60
applied the correlation to the biotransformation
of methyl ricinoleate into -decalactone by the yeast Yarrowia
lipolytica. They showed that k
L
a had an inuence on the
aroma production; however, for the low hydrophobic substrate
concentration used (1.08% v/v) and cellular density of 2.0 10
7
cells mL
1
, a minimal k
L
a value of 70 h
1
was necessary to attain
the maximum aroma production, 141 21 mg L
1
(obtained at
agitation and aeration rates of 400 rpmand 0.6 vvm, respectively).
The numerical constants used were 650, 0.3, 0.7 and 1.7, for , ,
and , respectively.
Hydrodynamic studies in two-phase partitioning bioreactors
have focused on gaining further insight into understanding the
mechanisms relatedwithformationof theinterfacial area available
for mass transfer, so that substrate supply (normally fromthe non-
aqueous phase) does not become the rate limiting step of the
process. The interfacial area has previously been correlated with
the dispersed phase hold-up fraction and the Weber number.
61
The interfacial area available for mass transfer (a) is given by
a =
6
d
32
(9)
where d
32
is the Sauter mean drop diameter. Accurate knowledge
of the effect of bioreactor operating conditions on d
32
is therefore
very important. Knowledge of d
32
can also give an early indication
of the stability of the liquidliquid dispersion created. The
physicochemical properties of the media can inuence both
mean drop size and drop size distribution, as observed by Torres-
Martinez, in the characterization of a multiphase system involving
ionic liquids.
62
k
L
a determination in shaken devices
One of the key challenges for shaken fermentation technology
is to provide sufcient oxygen for the optimal growth of aerobic
microorganisms.
63
Adequate oxygen supply is crucial not only
for industrial production, but also for meaningful screening and
process development.
64
Under oxygen-limitingconditions aerobic
microbes grow slowly, production of the intended metabolites is
scarce, if any, andfurthermore, unwantedsynthesis of metabolites
typical of anoxic conditions is prone to occur. Results obtained
under such conditions are likely to be misleading, particularly for
scale-up purposes.
63
To study the effect of organism properties,
medium composition or cultivation strategy on growth and
production, incubation in a non-limiting oxygen environment
is absolutely necessary. Otherwise wrong information about the
variables under study might be obtained.
65,66
Several operational parameters affect the OTR in shaking
devices, namely the shape and size of the vessel; the shaking
frequency; the shaking amplitude; the shaking angle; the type
of agitation (orbital or linear); and the lling volume. OTR is
also inuenced by the surface properties of the ask material,
which may be either hydrophobic or hydrophilic; and by the
physical chemical properties of the liquid (viscosity, oxygen
solubility, diffusivity and surface tension, the latter being more
noticeable in MWP).
3,27,67
Some particular setbacks are likely to occur when operating
shaken vessels, specically:
1. The reproducibility of microbial growth might be poor.
68
2. When bafed vessels are used, small differences in depth
and positioning of the bafes lead to signicant differences
in oxygen supply, growth and product formation of parallel
cultivations.
3. Out-of-phase phenomena might occur.
67
4. Shaking frequency in these asks has to be reduced to avoid
splashing of the liquid. Were droplets to reach the plug of
the asks, gas transfer limitations or contaminations might
occur.
63
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
8
8
www.soci.org MPC Marques, JMS Cabral, P Fernandes
Engineering features and correlations for k
L
a in shaken asks
and MWP
Since bioprocess development (strain selection, strain enhance-
ment, process optimization) is widely carriedout inshakenvessels,
and efforts towards engineering characterization of the oxygen
transfer mechanisms in such devices have been undertaken.
Maier and B uchs determined the maximum gasliquid mass
transfer capacity in 250 mL shaking asks on orbital shaking
machines, using the sulte oxidation method, by variation of
the shaking frequency and diameter, lling volume and viscosity
of the medium. The distribution of the liquid within the ask
was modeled, taking as reference the intersection between the
rotational hyperboloid of the liquid and the inner wall of the
shaking ask.
69
The mass transfer within the shake ask is conditioned by
two resistances: one due to the sterile closure of the vessel, the
other due to the gasliquid interface. Owing to the resistance of
the closure, the oxygen partial pressure within the ask is lower
than the oxygen partial pressure of the surroundings. Mrotzek
et al.
70
concluded that, under normal conditions, the mass transfer
resistance of sterile closures is far smaller than the resistance of
the gasliquid interface.
The gas exchange through the sterile closure was described
by the extended model of Henzler and Schedel,
71
where besides
Fick diffusion ow, other parameters and issues are taken into
consideration, such as: (i) gas transfer by combined action of
diffusion and convection due to non-equimolar mass exchange
(Stefanow); (ii) diffusioncoefcients arenot regardedas constant
but are instead calculated as a function of the respective local gas
concentration; and (iii) consideration of the water vapor ow.
Operating with bafed asks, and using suitable operating
conditions, such as large shaken amplitudes, high OTRs can be
obtained. In 250 mL shaking asks with a lling volume of 50 mL, a
k
L
a of 400 h
1
was obtained with a shaking amplitude of 2.5 cm. If
the amplitude was increased to 5 cm, similar values were obtained
at lower shaking frequencies. In these cases, the resistance of the
sterile closure can become the limiting factor.
Experimental data from given sets of experiments were tted
by least-squares after dimensional analysis,
67
and the following
proportionality for the maximum oxygen transfer capacity
(OTR
max
) was found
OTR
max
N
0.84
V
0.84
d
0
0.27
d
1.25
(10)
Consequently, an increase of the maximum oxygen transfer
capacity of a given system can be achieved by increasing the
shaking frequency, reducing the lling volume, increasing the
shaking diameter (d
0
) or reducing the maximumask diameter (d)
(this is only valid when the V
1/3
/d ratio remains constant).
Taking this into account, Maier et al.
72
modeled the gasliquid
mass transfer in shake asks at water-like liquid viscosity, in ask
sizes between 50 and 1000 mL. Relative lling volumes of 416%,
shaking diameters of 1.25, 2.5, 5, 7, 10 cmand shaking frequencies
of 50500 rpm were tested. Furthermore, the previous model of
the gasliquid mass transfer Equation (10) was extended to a two
sub-reactor model, to account for different mechanisms of mass
transfer in the liquid lm on the ask wall, and the bulk of the
liquid rotating within the ask. The two-reactor system approach
consists of a stirred tank reactor (bulk liquid) and a lm reactor
(lmon ask wall and base). The mass transfer into the lmon the
ask wall and base at in-phase operating conditions, is described
by Higbie penetration theory. Two different mass transfer theories
were applied to successfully describe the mass transfer into the
bulk liquid: a model by Kawase and Moo-Young
73
and a model by
Gnielinski.
74
Extensive details can be found elsewhere.
72
The agreement between the new modeling approach and the
experimental data was within 30%. The applicability of the
latter models to a biological system was shown using a Pichia
pastoris culture. The OTR
max
was determined using the sodium
sulte method which displayed a correlation factor of 2.8, when
compared with the OTR determined using an oxygen limited
P. pastoris culture, under variation of the operating conditions
(250 mL shake ask with lling volumes of 15, 25 and 40 mL,
under shaking frequencies of 50500 rpm with 5 cm of shaking
diameter).
72
Moreover, the volumetric mass transfer coefcient
models allowed a signicant agreement between experimental
dataandmodel predictions, as reectedbyacorrelationcoefcient
of 0.88. The maximum volumetric mass transfer coefcient of the
experimental investigation was found to be 0.157 s
1
in the 50 mL
ask, at a relative lling volume of 4%, a shaking frequency of
450 rpm and a shaking diameter of 7 cm.
Liu et al. correlated the experimental data for the determination
of OTR with a multivariable power correlation in carotenoid
(astaxanthin) production by the red yeast Phafa rhodozyma.
15
The constant parameters were derived by linear regression of the
data, resulting in the following equation:
k
L
a = 0.141N
0.88
_
V
V
0
_
0.80
(11)
where V
o
is the ask volume. Results showed a direct linear
correlation between carotenoid yield and OTR (OTR from 0 to
690 mg L
1
h
1
), indicating that carotenoid production is limited
by oxygen transfer. Mantzouridou et al.
75
also investigated the
effect of oxygen transfer rate on -carotene production by
Blakelsea trispora in shake asks. The results indicated that the
concentration of -carotene (704.1 mg L
1
) was highest in culture
grown at maximum OTR of 20.5 mmol L
1
h
1
. Moreover, OTR
levels higher than20.5 mmol.L
1
h
1
werefoundtobedetrimental
to cell growth and pigment formation.
Besides increasing the diameter or frequency of shaking or
decreasing the lling volume, one possibility to achieve high
maximum oxygen transfer capacities (OTR
max
) is to modify the
usual round geometry of the shaken vessel. As shown by many
groups, changing the geometry of the vessel (i.e. from round-
bottomed base to square-bottomed base) and/or introducing
bafes into shake asks result in a signicant increase in maximum
oxygen transfer capacity.
76
The OTR
max
can be increased up to
510-fold even at low shaking frequencies.
Alternatives for the well bottom design have been established,
among them the square shaped well. This type of well geometry
has been investigated by Duetz and Witholt
42,77
and Duetz et al.
78
The effects found for square wells are comparable with those in
bafedshakeasks. Eventhoughsquarewells allowfor anincrease
in OTR
max
of roughly 100% when compared with round wells, the
aforementioned problems described for shake asks (splashing
and out-of phase phenomena) may limit the utilization of square
deep well plates as cultivation vessels.
More recently, Funke et al.
79
presented well congurations with
a geometry that aimedat avoidingsplashingwhile simultaneously
allowing for a high lling volume. Different plate formats were
studied, where variation were achieved by increasing the number
of edges and rounding the edges in square geometry plates,
rounding edges in pentagonal shape wells, star and ower
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198
1
1
8
9
Bioprocess scale-up: parameters to be used www.soci.org
shape wells, among others. Dissolved oxygen tension (DOT) was
measured in an experimental setup described by Samorski et al.
80
The well geometry proved inuential for OTR
max
. By introducing
such bafes, an OTR
max
in excess of 100 mmol L
1
h
1
(k
L
a over
600 h
1
) was obtained, roughly double that of round 48-well
MWPs. In conventional MWPs, values for OTR
max
reported so far
have not generally exceeded 50 mmol L
1
h
1
.
27
Values of the
OTR in stirred tank reactors can reach 500 mmol L
1
h
1
, but
overall, in standard batch fermentations, the OTR does not exceed
100 mmol L
1
h
1
.
26
The nal well design, a six-petaled format,
provided the best compromise among different levels of bafing,
allowing simultaneously for a stable liquid height at the well
center during shaking (strong), high OTR
max
independently of
lling volume and shaking frequency (moderate), and high lling
volume without splashing.
79
Nonetheless, there are two common problems associated with
the use of microtiter plates for carrying out fermentations. These
are evaporation and cross-contamination due to spillover. In order
to avoid these drawbacks, MWPs can be sealed with adhesive
tape. The use of sealing tape leads to airtight closure of the wells
decreasingoxygentransfer tothereactor well andconsequentially
to the reaction medium.
11
Doiget al.
81
modeledthe volumetric oxygentransfer coefcient
in MWPs using dimensionless groups. The basis for the modeling
was the experimental measurements of airliquid specic surface
area, determined both by the rate of evaporation and by high-
speed video photography.
Flowbehavior was modeledseparately accordingtothe specic
microplate conguration (24, 96 and 384-well), and the following
correlations, respectively, were obtained:
a
f
a
i
= 2.895 Fr
0.86
Bo
0.03
(12)
a
f
a
i
= 1.092 Fr
0.64
Bo
0.15
(13)
a
f
a
i
= 0.607 Fr
0.51
Bo
0.18
(14)
The different models for predicting specic airliquid surface
area (af/ai) converged, with most data points lying within 25%
of the predicted values.
The k
L
a values ranged from about 0.005 to 0.055 s
1
(shaking
amplitude of 3 mm and shaking frequencies ranging from 200
to 800 rpm) and are within the range reported by other authors
for these devices.
5,35,78,82
Since the specic surface airliquid area
increased by a maximum factor of 4, and k
L
a increased by up to
10, it is clear that the liquid side oxygen transfer coefcient k
L
was
also affected by shaking conditions. Hermann et al.
82
observed
an increase in k
L
from about 5.5 10
5
to 1.4 10
4
ms
1
as
shaking frequency was increased from 200 to 800 rpm at 25 mm
amplitude. Values calculated by Doig et al.
81
for k
L
varied from
2.8 10
5
ms
1
to 8.3 10
5
ms
1
(shaking amplitudes of 3 mm
to 8 mm, in 24, 96 and 384 well microtiter plates).
Correlations for k
L
typically encompass Reynolds and Schmidt
numbers, andareusuallyintheformSh = aRe
b
Sc
c
, wherebranges
from 0.7 to 0.8 and c is usually 0.33.
81
The optimized correlation
for all three microplate geometries is
Sh = 0.19 Re
0.68
Sc
0.36
(15)
All the experimental data were modeled within 30% de-
viation, using correlation (15). Combining equation (15) with
Equations (12)(14), an overall correlation for predicting the vol-
umetric oxygen transfer coefcient k
L
a, in round bottom plates
becomes
k
L
a = 31.35D
i
a
i
Re
0.68
Sc
0.36
Fr
x
Bo
y
(16)
where D
i
is the diffusion coefcient, a
i
is the initial specic surface
area, and x and y are constants depending on the microplate
geometry (according to Equations (12)(14).
Islam et al.
83
modied Equation (16) in order to predict k
L
a in
square well microtiter plates. The correlation obtained was:
k
L
a = 3.94 10
4
_
T
D
_
a
i
Re
1.91
e
aFr
b
(17)
where the values for a and b were 1.66 and 2.47 for 48 rectangular,
at wells; 0.70 and 1.51 for 24 square wells, round bases; and 0.88
and 1.24 for square wells, pyramidal bases. The predicted values
were in good agreement with the experimental data for k
L
a.
Case studies: scale-up from shaken devices and miniature
stirred reactors based in k
L
a similarity
In previous studies
84
k
L
a was identied as the key engineering
parameter for characterization of an E. coli based process for
heterologous protein expression, in microtiter plate format. A
scale-up to 7.5 and 75 L stirred tanks was performed, using as
criterion k
L
a xed at either 0.069 s
1
or at 0.015 s
1
. At 0.069 s
1
boththe fermentationprole (biomass concentrationandglycerol
consumption) and product yield were identical in all scales,
enabling a 15 000-fold quantitative scale-up, representative of
fermentation performance. At 0.015 s
1
, due to poor gasliquid
distributions observed within the larger stirred tanks at matched
k
L
a, the overall fermentation prole was not reproduced.
Micheletti et al.
6
studied the scale-up of aerobic fermentation
of E. coli JM107:pQR706 for overexpressing transketolase (TK),
from microtiter plates to stirred reactors. Using the correlation
displayed in Equation (17), a k
L
a of 0.079 s
1
at 1000 rpm for
a 96 deep well microtiter plate was obtained. Maintaining the
volumetric oxygen mass transfer coefcient, it was possible to
scale-up the process directly from the microwell reactor to a lab-
scale reactor (1.4 L). Apart fromsome differences in duration of the
lag phase observed when the two scales were compared, similar
values of
max
and nal biomass concentration were obtained.
Likewise, the rates of L-erythrulose formation when the cells
from the respective fermentations were used for the subsequent
HPA(-hydroxypyruvate) andGA(glycolaldehyde) bioconversion,
were also similar.
Zhang et al.
8
used computational uid dynamics (CFD) to
provide a detailed characterization of uid mixing, energy
dissipation rate and mass transfer in single well bioreactors,
from deep square 24-well and 96-well microtiter plates. The CFD
simulations showed that liquid mixing is more intensive in 96-well
than in 24-well bioreactors, due to the vertical movement of the
bulk uid, in addition to the rotational movement. Liquid motion
was strongly dependent on the orbital shaking amplitude which
generally has a greater impact than the shaking frequency.
Predicted k
L
a values were compared with experimental results
obtained from dynamic gassing out experiments using a bre-
optic dissolved-oxygen probe. The resulting k
L
a values measured
were 0.036 s
1
and 0.023 s
1
, at 1000 rpm and 500 rpm, respec-
tively, in 96-well MWPs, at an orbital shaking amplitude of 3 mm.
Thecorrespondingpredictedvalues, 0.065 s
1
and0.056 s
1
, were
higher. The discrepancy was ascribed to some features of the dy-
namicmethod, namelytotheprobesensitivitytowardstheshaking
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
9
0
www.soci.org MPC Marques, JMS Cabral, P Fernandes
movement of the microtiter plates, as well as to the positioning of
the probe and to the mixing time. CFDstudies could give more ac-
curate values if mediumlosses due to evaporation were taken into
account, and if the actual physical properties of the fermentation
media (viscosity, density) were used instead of water, as input for
performing the calculation. Batch cultures of E. coli DH5 showed
similar maximum specic growth rates and nal biomass yields
in shaken 24-well MTP and Erlenmeyer bioreactors, and in stirred
miniature and 20 L bioreactors at matched k
L
a values.
The biotransformation of benzaldehyde to l-phenyl acetyl
carbinol was scaled-up from a 100 mL shake ask to a 5 L reactor,
using resting cells of Saccharomyces cerevisiae.
85
The following
correlations were used for the prediction of k
L
a in the growth
medium and in the biotransformation medium, respectively,
k
L
a = 0.024
_
P
V
_
0.725
v
0.892
g
(18)
k
L
a = 6.99 10
6
_
P
V
_
1.14
v
0.365
g
(19)
The volumetric oxygen mass transfer coefcient varied from
0.010.07 s
1
in the growth medium to 0.0050.02 s
1
in the
biotransformation medium. Maintaining the k
L
a constant as
scale-up criteria, both for the cell growth process and for the
biotransformation, a 50-fold scale-up was achieved.
Gill et al.
7,86
studied the inuence of k
L
a on the fermentation
of E. coli TOP10 pQR239 in a 100 mL reactor. A correlation was
developed for the miniature reactor
k
L
a = 0.224
_
P
V
_
0.35
v
0.52
g
(20)
which was validated using the dynamic gassing out technique.
For scale-up to a 2 L bioreactor, different values of k
L
a were
used, varying from 0.06 to 0.11 s
1
. The results showed that there
was good agreement between cell growth and dissolved oxygen
tension proles across the range of k
L
a values studied, compared
with experiments at matched P/V values. The trend in oxygen
depletion during the growth phase and the time taken for both
systems to become oxygen limited were similar and reproducible
in both cases, thus resulting in a satisfactory 20-fold scale-up.
Using an E. coli JM107:pQR706 overexpressing transketolase
system, Micheletti et al.
6
showed that k
L
a is the most suitable
scale-up parameter from a 24-well microtiter plate (1 mL lling
volume) to a 2L stirred tank reactor. On the other hand, the
prediction for the power consumption in the mechanically stirred
bioreactor (3.64 W m
3
) working under operational conditions
that allowed for growth patterns that matched those in shaken
vessels was roughly 10-fold lower.
More recently, Marques et al.
5
achieved an 8000-fold scale-up
for the side-chain cleavage of -sitosterol performed by whole
cells of Mycobacterium sp. NRRL B-3805 in 24-well microtiter
plates to a 5 L reactor. Scale-up was performed based on k
L
a
similarity at 0.058 s
1
(at a shaking frequency of 250 rpm and
lling volume of 0.5 mL). Similar proles (glycerol consumption
and 4-androstene-3,17-dione (AD) production) were achieved in
both scales validating multi-well plates as a small-scale reactor
for performing complex bioconversions. Nonetheless, there was
a consistent gap between values obtained in the multi-well
plate system and the bench-scale reactor for AD production and
biomass formation visible also in the levels of oxygen depletion.
The behavior was ascribed to diverse intrinsic hydrodynamic
conditions in the different reactor congurations, since mass
transfer plays an important role, both for oxygen and substrate
uptake.
Both MWP as well as miniature bioreactors proved effec-
tive as starting platforms for scaling up to bench/pilot scale
biotransformation/fermentation, but further studies in different
environments would be welcome that would hopefully further
validate this approach. Case studies are relatively scarce, where
miniature bioreactors are concerned. This can be partly ascribed
to the wider dissemination and lower cost of MWP platforms,
when compared with miniature bioreactors, whose availability in
the market is clearly lower.
11
The studies referred to in this paper
related to the use of miniature bioreactors are based on in-house
developed prototypes. The use of the simpler MWP platforms may
be of limited use when pH shifts take place during the process,
given the lack of mechanisms for pH control, only available in
more complex platforms.
11
Processes involving signicant shifts
in broth rheology (i.e. production of polymers) or viscous envi-
ronments may be more adequately dealt with using miniature
bioreactors, since these may provide more efcient mixing.
SCALE-UP BASEDONVOLUMETRIC POWER
CONSUMPTION
Power consumption is a key parameter in (bio)chemical engineer-
ing, i.e. anengineeringcharacterizationparameter. Concomitantly,
it is a strong candidate for use as a criterion for (bio)reactor de-
sign and process scale-up. Often referred to as volumetric power
consumption (P/V), it is dened as the amount of energy required
to generate movement of a uid within a vessel in a given pe-
riod of time. Along with the power actually drawn by the uid,
relevant to the outcome of a given process, further power is re-
quired to account for energy losses, mostly due to friction, and
power consumption by motors and gearboxes. This excess power,
although often relevant in terms of overall power consumption,
is not considered for design or scale-up of the process. The vol-
umetric power consumption is representative of the turbulence
degree and media circulation in vessels, and inuences heat and
mass transfer, mixing and circulation times.
87
Constant volumetric
power input was appliedsuccessfully as scale-upparameter for the
early industrial penicillin fermentations (1 hp gallon
1
, equivalent
to1.8 kW m
3
), andinfermentations withlowenergy inputs,
88
but
it is limited in fermentations requiring high energy inputs, such
as recombinant E. coli cultures,
1
possibly due to high associated
costs and to high shear stress in the larger-scale stirred vessels.
AccordingtoRushtonet al.,
89,90
thepower input for theagitation
of a non-aerated mixture, P
o
is characterized by the dimensionless
variable power number (N
P
):
P
0
= N
p
N
3
T
5
(21)
where N is the stirrer speed and T the stirrer diameter. The
power number depends on other dimensionless groups such as
Reynolds number and Froude number, as well as on the number
of agitator turbines. The power consumption in ungassed systems
is always higher than the power consumption in gassed systems,
since aeration signicantly inuences the power drawn from the
impeller by the uid.
7
The effect of aeration has been studied
extensively by Nienow et al.,
91
Oosterhuis and Kossen,
92
Yawalkar
et al.
93
and Gogate et al.
46
It has been shown that the gassed
power input is usually 3040% of the ungassed power input,
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198
1
1
9
1
Bioprocess scale-up: parameters to be used www.soci.org
depending on the type of impeller and aeration rates used.
79
Cavities (gas pockets) are formed behind the impeller blades in
gassed systems, resulting in different densities of the uid under
gassed and ungassed conditions.
94,95
Too high gas ow is to be
avoided since this may lead to cavities extending between blades,
which result in ooding and mechanical instability. Cavities may
beminimized, hencegas handlingcapacity improved, withtheuse
of adequate impellers, such as the Rushton turbine. Alternative
blade impeller congurations (i.e. concave blades) have also been
introduced.
96,97
Hughmark
98
suggests the following relationship for estimating
the volumetric power consumption:
P
P
o
= 0.1
_
_
N
2
T
4
gWV
2
3
_
_
1
6 _
Q
NV
_
1
4
(22)
where W is the width of turbine blades and Q is the volumetric
gas ow rate. Over the years several modication were proposed
in order to improve this model. Such modications included
taking into account the number and type of impellers and reactor
dimensions, among others.
99,100,101
Along with the use of predictive correlations, power con-
sumption may be determined experimentally through the use
of electrical or calorimetric measurements, or through the use of
dynamometers, torquemeters and strain gauges.
87
In bench- and pilot-scale bioreactors typically used for the
fermentation of bacteria and fungal micro-organisms, the power
consumption ranges from 1 to 3 kW m
3
.
102,103
Shaken asks
Although well established for bench-scale and above reactors,
only recently have dedicated efforts been made to provide a
suitable characterization of the volumetric power consumption in
miniaturized devices, and to take advantage of this parameter for
the engineering characterization of said devices. Most reports on
the experimental determinationof power consumptioninshaking
systems are by the group of Jochen B uchs at RWTHAachen. B uchs
et al.
104,105
measured the power consumption in shaken asks at
high and low-mediumviscosities. The experimental assembly was
a simple rotary shaking machine xed to a frame, combined with a
torque sensor attachedto the powering drive. Torque andshaking
speed were monitored and correlated by the following relation:
P
V
= Ne

N
3
d
4
V
2
3
= C
3

N
3
d
4
V
2
3
Re
0.2
(23)
where Ne

is the modied Newton number for shake asks.


The correlation implies that the specic power consumption is
dependent on the shaking frequency according to P N
2.8
. Such
a correlation is typical of that found in unbafed agitated tank
reactors. The model includes C
3
as the only tting parameter
using least-squares non-linear tting for the description of all the
experimental results obtained in their study, having a value of
1.94.
The values for volumetric power consumptioncalculatedby this
empirical correlation and the experimentally measured values t
within a deviation range that does not exceed 30%. In the range
0.01 to 0.2 kW m
3
, with shaking frequency from80 to 380 min
1
,
lling volumes of 4% to 20% of nominal ask volume and
shaking diameter from 2.5 to 5 cm, larger discrepancies are found
between calculated and measured values. The corresponding
data was gathered at low shaking frequency (80 to 120 min
1
)
and, therefore, at low Reynolds numbers (Re 500 to 5000),
possibly within the transition from laminar to turbulent ow, a
feature that could account for the increased deviation between
predicted and experimental values.
The discrepancy between experimental and predicted data at
low power consumption was overcome in liquids with viscosities
between 0.8 and 200 mPa s,
105
where a correlation was found that
took into consideration all the ow regimes:
Ne

= 70Re
1
+25Re
0.6
+1.5Re
0.2
(24)
This correlation consists of a laminar (Re
1
), a transition (Re
0.6
),
and a turbulent term (Re
0.2
). From the power number variation
with the ask Reynolds number, two ow conditions were
identied: in-phase conditions, where the bulk of the liquid
in the ask circulates in phase with the shaking platform; and
out-of-phase conditions, where only a minor fraction of the
liquid is actually moving along the ask wall.
67
The out-of-phase
conditions lead to a decrease in volumetric power consumption,
mixinggas/liquidmass transfer. Inorder tosystematically describe
the in-phase and out-of-phase conditions, a newnon-dimensional
number, called the phase number (Ph), was derived based on an
analogy to a partially lled, rotating horizontal drum, and can be
expressed as
Ph =
d
0
d
_

_
1 +3 log
10
_
_
_
(2N)

d
2
4
_
_
_1

_
1
4

_
V
0.33
d
_
2
_
_
_
2
_

_
_

_
(25)
B uchs et al.
98
established that all operating conditions where
Ph > 1.26arein-phasewhileout-of-phaseconditions areobserved
for Ph < 1.26, which are prone to occur when large asks or high
viscosity uids are used.
11
Nonetheless, out-of-phase conditions in
largeasks at water-likeviscosities arepossibleif theaskis quickly
accelerated.
106
This may occur, if (i) the shaking machine has a
strong engine, (ii) the load of the machine is low (small number of
asks placed on the shaker table or (iii) the shaking asks which
wereremovedfromtheshakingtable(e.g. for sampling) areplaced
back on the running machine.
This phenomenonof out-of-phase operatingconditions may be
of great practical relevance, since working under such conditions
will signicantly reduce the oxygen transfer and mixing intensity,
having a strong impact on strain and medium development. The
ow in a shaking ask tends to be in-phase at large shaking
diameter, low viscosity, large lling volume, higher shaking
frequency and small number and size of bafes.
Kato et al.
107
used a calorimetric method, based on the method
developed by Sumino et al.,
108
for assessing power consumption
in larger shaken asks (nominal volume up to 20 L). Despite the
increase in volume, the order of magnitude of P/V is the same as
for smaller scale shaking asks, up to a nominal volume of 2 L and
having different geometries.
104
Similar studies have been conducted reporting a stronger
dependency between power consumption and shaking speed,
P N.
5.75
Nonetheless, a narrower range of shaking frequencies,
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
9
2
www.soci.org MPC Marques, JMS Cabral, P Fernandes
between 100 and 200 rpm was used, compared with the
100300 rpm range of Kato et al.
1
This may also account for
the volumetric power consumption reported in either work, from
less than 2 kW m
3
to 5 kW m
3
, respectively. Therefore, it can be
concluded that the increase in power consumption is much more
pronounced in the region of low shaking frequencies, between
100 and 200 rpm.
Raval et al.
109
compared the torque method and temperature
methodfor thedeterminationof power consumptionindisposable
shaken bioreactors of 2 L polycarbonate (PC) bottles and 20 L
polypropylene (PP). Data points collected by both methods were
within30%tolerancerange. Values of 3.5 kW m
3
wereobtained
in 2 L asks with lling volume of 250 mL at 300 rpm (shaking
diameter 50 mm).
Another approach for estimation of the volumetric power
consumption was proposed by Zhang et al.
103
Using CFD
techniques, theseauthorswereabletopredict power consumption
when 250 mL shake asks were operated between 100 and
300 rpm, with shaking diameters between 20 and 60 mm and
lling volumes between 25 and 100 mL. CFD models suggest that
P/V can be correlated with the lling volume or with the shaking
frequency according to P/V V
0.7
and P/V N
2.7
, respectively.
Both correlations are in agreement with those suggestedby B uchs
et al.
104
Nonetheless, for lower shaking diameters the exponent
in the shaking frequency correlation decreased to 1.6, suggesting
that the applicability of these correlations are limited to a range of
amplitudes. Values predicted were between 40 and 1200 W m
3
.
Thepower consumptionestimateswerefoundtobemoresensitive
to changes in shaking amplitude than to frequency
Multi-well plates
Despite the widespread application of MWPs, currently there is
no experimental report on power consumption in these devices,
mainlyduetolackof commercial torquemeters or other measuring
methods withsufcient sensitivitytoperformsuchmeasurements.
In order to overcome this situation, Zhang et al.
8
proposed the use
of CFD to obtain estimates of power consumption. These authors
applied the methodology followed by Zhang et al.
103
where the
local energy dissipation rate was obtained using
=

v

(26)
where
V
is the viscous dissipation term. The power consumption
based on the uid friction is calculated from
P
V
=
_
V
0

v
dV
V
(27)
where
V
, can be expressed in terms of shear rates:
103

V
= 2
_
_
u
x
_
2
+
_
v
y
_
2
+
_
w
z
_
2
_
+
_
u
y
+
v
x
_
2
+
_
v
z
+
w
y
_
2
+
_
u
z
+
w
x
_
2
(28)
where U(u,v,w) is the velocity vector (velocity component) and x(x,
y, z) is the moving grid spatial vector (spatial component).
The results showed that power consumption and energy
dissipation rates in the shaken microwells are strongly affected
by the size of the well and the lling volume, as well as by the
shaking frequency and amplitude. These effects are higher in 96-
well microtiter plates than in 24-well microtiter plates under the
same operating conditions.
Moreover, two situations can be distinguished:
1. The power consumption in the 24-well reactor does not
increase linearly as the shaking frequency is increasedat 3 mm
shaking diameter. Up to roughly 800 rpm, there is a decrease
in power consumption, presumably due to out-of-phase
conditions. The estimated volumetric power consumption
is more sensitive to changes in shaking amplitude than to
frequency.
2. The volumetric power consumption at orbital shaking ampli-
tudes of both 3 mmand 6 mmincreased linearly with increase
in shaking frequency. Increasing the shaking amplitude from
3 mm to 6 mm, led to an increase in the volumetric power
consumption of 10002000 W m
3
using shaking frequencies
in the range of 8001000 rpm.
The volumetric power consumptionwas comparedfor a 24-well
microtiter plate, 96-well microtiter plate, 20 L stirred reactor, 6 mL
miniature stirred reactor
110
and a 250 mL shake ask
102
at feasible
operating conditions.
The miniature stirred reactor closely matched the conguration
of conventional stirred reactors, only scaled down to 6 mL
scale.
11
The P/V was similar in all reactors with the exception
of shaken asks and the 24-well microtiter plates. The values used
for microtiter plates were obtained by CFD studies in specic
conditions; nonetheless similar values of P/V can be reached in
shaken systems.
Case studies: scale-up from shaken devices and miniature
stirredreactorsbasedinspecicpower consumptionsimilarity
Aqueous systems
Despite the large number of studies on scale-up of fermentation
processes and some on bioconversion processes, examples on the
use of volumetric power consumptionas key parameter are scarce.
Gill et al.
7
developed a prototype of a microreactor that enables
parallel operation of 416 independently controlled experiments.
Each microreactor has a maximumworking volume of 100 mL and
is equipped with a magnetically driven six-blade Rushton turbine,
with a stirring range of 1002000 rpm. Growth of E. coli TOP10
pQR239 was scaled-upfromthis microreactor toa 2 L reactor tted
with two six-blade Rushton turbine impellers
111
using as constant
the power consumption. P/V values used in this study were 657,
1487 and 2960 W m
3
.
At the lowest P/V value of 657 W m
3
, the 2 L reactor signif-
icantly underperforms compared with the miniature bioreactor,
achieving a nal biomass concentration of almost 3 g L
1
less
(X
nal
in the miniature bioreactor was 5.6 g L
1
). This is likely to be
the result of operating at a reduced agitation rate, and the poor
gasliquid dispersion that was observed at the 2 L scale under
these operating conditions. Given the poor oxygen transfer at the
2 L scale, the dissolved oxygen tension reached zero much earlier,
and cell growth was clearly seen to become oxygen limited. The
performance of the 2 L reactor was improved at higher P/V values
(>1000 W m
3
), with very similar
max
and nal biomass con-
centration values obtained at both scales. The trends for oxygen
depletion during the exponential growth phase and for the time
takenfor bothsystems tobecome oxygenlimitedwere similar and
reproducible at the twohigher P/V values (1487 and2960 W m
3
).
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198
1
1
9
3
Bioprocess scale-up: parameters to be used www.soci.org
The volumetric power consumption was also used to scale-up
the alginate production using Azotobacter vinelandii cells, from
500 mL shake asks to 14 L reactor.
112
The study showed that
when an initial power drawof 0.27 kW m
3
was applied, a specic
growth rate of 0.16 h
1
was obtained in a stirred reactor, which
roughly doubled the specic growth rate obtained in shake asks
(0.09 h
1
). Moreover, differences in the broth viscosity, concentra-
tion proles and molecular weight of the alginate were observed.
In order to overcome these, the initial P/V in the reactor or along
the time course of the cultivation was reduced, which ultimately
allowed one to match the molecular characteristics of the alginate
obtained in shake asks. This approach further highlights the
potential limitations of shaken vessels as starting platforms for
scaling-up when signicant modications in medium rheology
take place throughout the process. This study was based on the
theoretical analysis of power consumption in shake asks. The
power consumption in shake asks was estimated from extrapo-
lation of data reported by B uchs et al.
104,105
This is a key feature, as
thechanges inpower input affect theOTR, whichinturnaffects the
molecular characteristics of alginate. Further work was performed
in order to gain insight into the power consumption along cell
growth for the same alginate-producing system. Pe na et al.
113
showed that the power consumption increased exponentially
during fermentation, achieving a maximum value of 1.4 kW m
3
after 40 h cultivation. This increase was due to the increased
viscosity of the culture broth, which resulted from the increase
in the molecular mass of alginate and polymer concentration. In
this period, a maximal alginate concentration of 5 kg m
3
, with a
maximum molecular weight of 550 kDa was obtained.
Although the viscosity increased in the period from40 to 70 h, a
slight drop in the power consumption was observed, leading to a
value of 1.2 kW m
3
at 70 h cultivation. This behavior was possibly
due to out-of-phase conditions, which could also account for the
results obtained previously by Reyes et al.
112
Pen a et al.
114
triedtoreproduce the meanmolecular mass of the
alginates obtained in shake asks, in a stirred reactor maintaining
P/V constant. A 500 mL Erlenmeyer ask was used, containing
1/5 lling volume. The power consumption during alginate pro-
duction increased exponentially from 0.18 to 1.4 kW m
3
during
the rst 40 h of culture and remained practically constant during
the rest of the fermentation. The exponential prole of the power
consumption(from0.2to1.2 kW m
3
) was simulatedalongthefer-
mentationina14 Lbenchreactor, containing10 Lof Burkmedium,
by adequately controlling the agitation rate from 250 to 515 rpm.
Further dissemination of the evaluation of the validity of
constant P/V as criterion for scaling-up is dependent on the
availability of methodologies and tools for easily assessing this
parameter in miniaturized devices.
Studies in two-phase systems
All the above examples are for aqueous systems. When substrates
that have low water solubility are involved, two-liquid phase
systems can be used. In these cases, P/V is particularly important
as drop diameter (and hence the interfacial area available for
mass transfer between phases) depends on the maximum energy
dissipation rate or volumetric power consumption. Moreover, for
efcient extraction it determines also the mean drop size and the
dispersion of the immiscible organic phase.
115
In a stirred vessel d
max
(maximum drop diameter) will be given
by
d
max
T
We
0.6
T
(29)
and
We
T
=

c
N
2
T
3

(30)
whereWe
T
is thestirred-tankWeber number and is theinterfacial
tension.
116
To account for the effect of volume fraction on d
max
, a
linear concentrationcorrectionfunctionwiththefollowinggeneral
form was used:
d
max
T
= c
1
(1 +c
2
)We
0.6
T
(31)
where c
1
and c
2
are constants and is the volume fraction of
the dispersed phase.
61
The constant c
2
is considered equal to 3
when it accounts for turbulence damping at low dispersed-phase
concentrations, or higher than 3 for coalescing systems.
117
The maximum drop diameter is also a key parameter since it is
generally considered proportional to the Sauter mean diameter
d
32
, although this has been questioned.
118
The Sauter mean
diameter, commonly used in processes depending on interfacial
area, is dened as the ratio of the third to the second moment of
the drop size distribution:
d
32
=
k

i=1
n
i
d
3
i
k

i=1
n
i
d
2
i
(32)
where k is the number of size classes, n
i
the number of drops and
d
i
the diameter of drops in size class i.
117119
These aforementioned equations predict an increase in drop
size with increasing dispersed phase volume fraction, up to
about 40%. At higher fractions (usually above 50%), a further
increaseinthedispersedphaseconcentrationresults indecreasing
drop size. This behavior is attributed to a change in the drop
breakage mechanism, fromturbulent eddy at lowconcentrations,
to boundary layer at high concentrations.
117,120
B uchsandZoels
120
showedthat shakeasksledtolower levelsof
hydro-mechanical stress, as power consumption was much more
evenly distributed than in stirred tanks. This is due to different
mixing mechanism present in either reactor. Thus, in stirred tank
reactors, the power drawn from the region close to the stirrer is
quitehighcomparedwiththeaverageP/V, sincea small impeller is
usedtopromote mixingina large vessel. a relatively small impeller
agitates a relatively bulky tank, leading to higher power drawn
in the region adjacent to the stirrer compared to the measured
average or specic P/V. On the other hand, when shaken asks are
used, power is more homogeneously distributed throughout the
vessel, since energy is introduced through a relatively large wall
area.
121
These effects must be taken also into account if cultures
are used where shear stress inuences physiological behavior,
such as lamentous organisms.
Cull et al.
116
exploited the use of two-phase systems for
the whole-cell bioconversion of 1,3-dicyanobenzene (1,3-DCB)
to 3-cyanobenzamide.with resting cells of Rhodococcus R312.
Volumetric power consumption (or N
3
D
i
2
constant) and constant
tip speed (or ND
i
constant) were the two criteria tested for scale-
up, from a 3 L to a 75 L reactor. The system was composed
of a phosphate buffer phase, where the cells were dispersed,
and toluene, which contained 1,3-DCB and, throughout the
bioconversion, the 3-cyanobenzamide formed.
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
9
4
www.soci.org MPC Marques, JMS Cabral, P Fernandes
Sauter mean drop diameters and drop size distributions were
verysimilar for scale-uponthebasis of P/V, ingeometricallysimilar
reactors. In all cases the drop size distributions obtained were log-
normal. The results presented in that work were obtained in
a specic phase system comprising 20% v/v toluene dispersed
in an aqueous buffer containing up to 10 g
cell wet weight
L
1
of Rhodococcus R312 cells. Despite being system specic, this
work contributed to establish specic power consumption as a
suitable tool for scale-up of two-phase bioconversion systems.
Moreover, it is demonstrated that a scale increase of 25-fold is
possiblemaintainingsystemspecicity. Theuseof smaller reactors
leads to cost reduction with reagents and overall process power
consumption, among others. Moreover, problems that might be
encountered at the large scale can be rapidly and efciently
identied in smaller scale.
Another example of scale-up of a two-phase system based on
the volumetric power consumption is the production 6-pentyl-
-pyrone (6PP) performed by cells of Trichoderma harzianum.
Rocha-Valadez et al.
121
reported a 20-fold scale-up, from 500 mL
shake asks to a 10 L stirred tank reactor. 6PP production
followed a sigmoid-shaped relationship with P/V, regardless of
the production scale or impeller diameter. Synthesis of 6PP was
triggered earlier in the stirred reactor, suggesting a physiological
response mechanism of T. harzianum towards hydrodynamic
stress. Overall, at low P/V values (from 0.08 to 0.4 kW m
3
) a
gradual increase in P/V improved 6PP production; however, for
P/V > 0.6 kW m
3
, 6PP concentration was signicantly reduced
due to higher hydrodynamic stress.
P/V shift inuencedmicrobial growthinthe stirredreactor since
the specic growth rate decreased from 0.052 h
1
at 0.08 kWm
3
to 0.033 h
1
at 1.6 kWm
3
. On the other hand the specic growth
rate was practically unaffected by P/V in the shake ask system.
Higher shear rates involve drastic physiological changes including
cellular differentiation and conidiospore production. This work
showedthat physiological changes observedduringprocess scale-
up (i.e. microbial growth, production rates and sporulation) were
triggered as result of differences in the shear conditions prevailing
in the two systems employed.
P/V provides a suitable scale-up criterion from miniaturized
devices when two-phase systems are involved. Further studies
would be welcome addressing particular cases where MWP are
used as the starting platforms.
OTHER SCALE-UP PARAMETERS
Examplesof theapplicationof other scale-upparametersarescarce
due to the fact that both k
L
a and P/V incorporate information
contained in the Reynolds number and mixing time, among
others. Nonetheless, examples of such criteria are given.
Constant impeller tip speed
The impeller tip speed, v
tip
, is expressed as
v
tip
= NT (33)
Tip speed is used as a rule for scale-up when the relationship
between shear and morphology is far from well understood, as
happens in mycelial cultures.
122
A rough rule of thumb suggests
that cell damage can occur at tip speeds above 3.2 m s
1
, but
the exact value is inuenced by many factors such as broth
rheology. Calculated tip speeds are usually greater than 3 m s
1
for production scale reactors.
122
Although useful for estimating
the potential for hyphae breakage and thus alteration of broth
morphology when branched yeast, lamentous bacterial and
fungal fermentations are involved, tip speed is less useful for
single cell bacterial or yeast fermentations. If scale-up is carried
out using constant tip speed (with geometric similarity), then the
value of P/V is often lowered, which can adversely affect aeration
efciency. It is possible to overcome this drawback by using more
impellers in the larger vessel in such a way that both tip speed and
P/V are kept constant. Tip speed inuences impeller shear, which
is proportional to the product of impeller tip speed and impeller
diameter, ND
i
2
, for turbulent ow conditions.
122
Hiruta et al.
123
demonstrated that maintaining impeller tip
speed of 270 m min
1
allowed scaling-up the production of -
linolenic acid by Mortierella ramanniana mutant MM 15-1 from a
30 L to a 1 m
3
reactor.
More recently, Dubey et al.
124
scaled-up the demethylation
of colchicine and their derivatives using Bacillus megaterium
ACBT03 cells, from a 5 L to a 70 L reactor. Under optimum culture
conditions the key monitoring factors to scale-up the process of
demethylation were aeration rate of 2.5 vvm and impeller tip
velocity of 4710 cm min
1
.
Similar Reynolds number
Reynolds number is expressed as:
Re =
TN
2

(34)
The use of constant Reynolds number is hardly ever used
for fermentation scale-up, since the effect of aeration on the
process is not incorporated,
125
andtheReynolds number generally
increases for successful scale-up designs. Other dimensionless
groups have also been examined for scale-up with limited
success, often resulting in technically unrealistic equipment and
operatingparameters. As it is difcult tomaintainall dimensionless
parameters constant upon scale-up, those most important to the
process must be identied accurately.
122
Constant mixing time
The mixing time t
m
denotes the time required for the reactor
compositiontoachieve a speciedlevel of homogeneity following
additionof a tracer pulse at a single point inthe vessel. Mixingtime
contains informationabout owandmixingwithinthereactor and
can be useful for biosynthesis processes scale-up. The mixing time
t
m
is dened as1
t
m
=
V
N
f
NT
3
(35)
where N
f
is the pumping number. The mixing time is typically
measured in stirred vessels. Nonetheless, there is an increased
interest in determining this parameter in shaken reactors. Gerson
et al.
126
used a mixing probe to determine uid mixing in a 1 L
shaken ask with lling volume of 540 mL at different shaking
frequencies. The results demonstrated that it is possible to
make such measurements, and that the mixing intensity rises
monotonically with shaking frequency in both stirred reactors and
shake asks. Also, the range of mixing intensities measured by the
device is similar in both systems over the range studied.
Recently, Nealon et al.
127
developed a high-speed video
technique for the accurate quantication of jet macro-mixing
times in static microwell plates, which also enables visualisation
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198
1
1
9
5
Bioprocess scale-up: parameters to be used www.soci.org
of jet formation and liquid ow patterns within the wells. Three
microwell geometries were investigated: a single well from a
standard 96-round well plate and dimension modied 96-well
plate. A general correlation was found for the time to reach 95%
homogeneity (s):
t
95
=
2.60D
1.5
h
0.5
u
0
d
n
(36)
where d
n
is the nozzle diameter, D is the well or vessel diameter, h
is the liquid height and u
0
is the nozzle velocity.
CONCLUSIONS
Despite the relevance of the scale-upissue inbiotechnology, there
is no straightforward and uniform strategy to tackle this matter.
A suitable scale-up criterion is elaborated in accordance with the
individual product, process and the facility.
An overall scale-up strategy consists of (i) a comprehensive
and detailed process characterization, and (ii) an appropriate
process control and process design. Owing to the complex
nature of bioprocesses (biotransformation, bioconversion and
fermentation) successful scale-up in most cases will not be based
on a straightforward strategy, but rather is the outcome of an
independent optimization on each process scale. Nonetheless,
bioconversions and fermentations are tentatively scaled-up on
the basis of k
L
a or volumetric power consumption from lab- to
pilot-scale. With the increasing ability to incorporate in MWPs and
in miniature and microreactors devices that allow the monitoring
of key process variables, such as dissolved oxygen tension,
pH, optical density or protein production, the application of
such small scale reactors is gaining widespread use in process
optimization, rather than just in the early stages of screening.
This pattern, that places small scale reactors in later stages of
process development, puts more focus on the need for increasing
the accuracy of scaling criteria. This has clearly been taken into
consideration, as shown by the number of papers dedicated
to this matter that have been published recently. Accordingly
the number of case studies reporting the successful scaling
up from MWPs and similar devices to bench scale reactor
or above is increasing. The insight required to obtain more
knowledge on this matter has received signicant contributions
from the developments in image analysis, data processing and
development of predictive models, namely through the use
of CFD. In spite of the many existing studies, no examples of
direct scale translation to industrial scale exist. There is a strong
likelihood that scaling from shaken vessels to plant scale will
not become a reality, given distinct hydrodynamic environments,
preventingfull reproducibility, hence compromisingthe outcome.
The translation from miniaturized to production environment is
nevertheless possible by scaling-out/numbering-up rather than
scaling-up. In this approach, large numbers of microreactors,
operating in continuous mode, are assembled so that production
up to tons/year basis can be achieved.
ACKNOWLEDGEMENTS
MPC Marques and P Fernandes thank Fundac ao para a Ci encia
e Tecnologia (Portugal) for nancial support in the form of a
PhD grant SFRH/BD/24433/2005 and programme Ci encia 2007,
respectively. This work was partially funded by research project
POCI/SAU-MMO/59370/2004 from Fundac ao para a Ci encia e a
Tecnologia (Portugal).
REFERENCES
1 Schmidt FR, Optimization and scale up of industrial fermentation
processes. Appl Microbiol Biotechnol 68:425435 (2005).
2 Najafpour GD, Biochemical Engineering and Biotechnology. Elsevier
Science, Amsterdam, The Netherlands (2007).
3 Fernandes P and Cabral JMS, Microlitre/millilitre shaken bioreactors
in fermentative and biotransformation processes a review.
Biocatal Biotransform24:237252 (2006).
4 Betts JI and Baganz F, Miniature bioreactors: current practices and
future opportunities. Microbial Cell Factories 5:21 (2006).
5 Marques MPC, Magalh aes S, Cabral JMS and Fernandes P,
Characterization of 24-well microtiter plate reactors for a complex
multi-step bioconversion: from sitosterol to androstenedione.
J Biotechnol 141:556561 (2009).
6 Micheletti M, Barrett T, Doig SD, Baganz F, Levy MS, Woodley JM,
et al, Fluid mixing in shaken bioreactors: implications for scale-up
predictions from microlitre-scale microbial and mammalian cell
cultures. ChemEng Sci 61:29392949 (2006).
7 Gill NK, Appleton M, Baganz F and Lye GJ, Quantication of
power consumption and oxygen transfer characteristics of a
stirred miniature bioreactor for predictive fermentation scale-up.
Biotechnol Bioeng 100:11441155 (2008).
8 Zhang H, Lamping SR, Pickering SCR, Lye GJ and Shamlou PA,
Engineering characterisation of a single well from 24-well and
96-well microtiter plates. BiochemEng J 40:138149 (2008).
9 Micheletti M and Lye GJ, Microscale bioprocess optimization. Curr
Opin Biotechnol 17:611618 (2006).
10 Weuster-Botz D, Puskeiler R, Kusterer A, Kaufmann K, John GT and
Arnold M, Methods andmilliliter scaledevices for high-throughput
bioprocess design. Bioprocess Biosyst Eng 28:109119 (2005).
11 Marques MPC, Cabral JMS and Fernandes P, High throughput
in biotechnology: from shake-asks to fully instrumented
microfermentors. Recent Patents Biotechnol 3:124140 (2009).
12 Burke F, Scale up and scale down of fermentation processes. in
Practical FermentationTechnology, ed. by McNeil B and Harvey LM.
John Wiley & Sons, pp. 231269 (2008).
13 Garcia-Ochoa F and Gomez E, Bioreactor scale-up and oxygen
transfer rate in microbial processes: an overview. Biotechnol Adv
27:153176 (2009).
14 B uchs J, Introductiontoadvantages andproblems of shakencultures.
BiochemEng J 7:198 (2001).
15 Liu YS, Wu JY and Ho KP, Characterization of oxygen transfer
conditions and their effects on Phafa rhodozyma growth and
carotenoid production in shake-ask cultures. Biochem Eng J
27:331335 (2006).
16 Doran P, Bioprocess Engineering Principles. Academic Press, London
(1995).
17 Calik P, Yilg or P, Ayhan P and Demir AS, Oxygen transfer effects
on recombinant benzaldehyde lyase production. Chem Eng Sci
59:50755083 (2004).
18 Gomez E, Santos VE, Alcon A and Garcia-Ochoa F, Oxygen transport
rate on Rhodococcus erythropolis cultures: effect on growth and
BDS capability. ChemEng Sci 61:45954604 (2006).
19 Ju L-KandSundarajan A, Model analysis of biological oxygentransfer
enhancement in surface-aerated bioreactors. Biotechnol Bioeng
40:13431352 (1992).
20 IslamRS, Tisi D, Levy MS and Lye GJ, Scale-up of Escherichia coli
growth and recombinant protein expression conditions from
microwell to laboratory and pilot scale based on matched k
L
a.
Biotechnol Bioeng 99:1281139 (2008).
21 Kostov Y, Harms P, Randers-Eichhorn L and Rao G, Low-cost
microbioreactor for high-throughput bioprocessing. Biotechnol
Bioeng 72:346352 (2001).
22 Zanzotto A, Szita N, Boccazzi P, Lessard P, Sinskey AJ and
Jensen KF, Membrane-aerated microbioreactor for high-
throughput bioprocessing. Biotechnol Bioeng 87:243254 (2004).
23 Betts J, Doig S and Baganz F, Characterization and application of
a miniature 10 mL stirred-tank bioreactor, showing scale-down
equivalence with a conventional 7 L reactor. Biotechnol Prog
22:681688 (2006).
24 Kensy F, Engelbrecht CandB uchs J, Scale-upfrommicrotiter plate to
laboratory fermenter: evaluation by online monitoringtechniques
of growthandproteinexpressioninEscherichiacoli andHansenula
polymorpha fermentations. Microb Cell Factories 8:68 (2009).
25 Doig SD, Ortiz-Ochoa K, Ward JM and Baganz F, Characterization
of oxygen transfer in miniature and lab-scale bubble column
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
9
6
www.soci.org MPC Marques, JMS Cabral, P Fernandes
bioreactors and comparison of microbial growth performance
based on constant k(L)a. Biotechnol Prog 21:11751182 (2005).
26 Vant Riet K, Review of measuring methods and nonviscous
gasliquid mass transfer in stirred vessels. Ind Eng Chem Process
Design Dev 18:357364 (1979).
27 Duetz WA, Microtiter plates as mini-bioreactors: miniaturization of
fermentation methods. Trends Microbiol 15:469475 (2007).
28 Anderlei T and B uchs J, Device for sterile online measurement of the
oxygen transfer rate in shaking asks. Biochem Eng J 7:157162
(2001).
29 Anderlei T, Zang W, Papaspyrou M and B uchs J, Online respiration
activity measurement (OTR, CTR, RQ) in shake asks. Biochem Eng
J 17:187194 (2001).
30 Guez JS, M uller CH, Danze PM, B uchs J and Jacques P, Respiration
activity monitoring system (RAMOS), an efcient tool to study
the inuence of the oxygen transfer rate on the synthesis
of lipopeptide by Bacillus subtilis ATCC6633. J Biotechnol
134:121126 (2008).
31 Wittmann C, Kim HM, John G and Heinzle E, Characterization and
application of an optical sensor for quantication of dissolved O
2
in shake-asks. Biotechnol Lett 25:377380 (2003).
32 Zhang Z, Perozziello G, Boccazzi P, Sinskey AJ, Geschke O and
Jensen KF, Microbioreactors for Bioprocess Development. JALA
12:143151. (2007).
33 Kensy F, Zimmermann HF, Knabben I, Anderlei T, Trauthwein H,
Dingerdissen U, et al, Oxygen transfer phenomena in 48-well
microtiter plates: determination by optical monitoring of sulte
oxidation and verication by real-time measurement during
microbial growth. Biotechnol Bioeng 89:698708. (2005).
34 Puskeiler R, Kaufmann K and Weuster-Botz D, Development,
parallelization, and automation of a gas-inducing milliliter-
scale bioreactor for high-throughput bioprocess design (HTBD).
Biotechnol Bioeng 89:512523 (2005).
35 John GT, Klimant I, Wittmann C and Heinzle E, Integrated optical
sensing of dissolved oxygen in microtiter plates: A novel tool for
microbial cultivation. Biotechnol Bioeng 81:829836 (2003).
36 Kensy F, Zang E, Faulhammer C, Tan R-K and B uchs J, Validation
of a high-throughput fermentation system based on online
monitoring of biomass and uorescence in continuously shaken
microtiter plates. Microb Cell Fact 8:31 (2009).
37 Schneider K, Sch utz V, John GT and Heinzle E, Optical device for
parallel online measurement of dissolved oxygen and pHin shake
ask cultures. Bioprocess Biosyst Eng. DOI 10.1007/s00449-009-
0367-0 (E-pub ahead of print). Latest info required.
38 Hermann R, Walther N, Maier U and B uchs J, Optical method for the
determinationof the oxygen-transfer capacity of small bioreactors
basedonsulphiteoxidation. Biotechnol Bioeng74:355363(2001).
39 Hermann R, Lehmann M and B uchs J, Characterization of gasliquid
mass transfer phenomena in microtiter plates. Biotechnol Bioeng
81:178186 (2003).
40 Thibault J, Leduy A and Denis A, Chemical enhancement in the
determination of k
L
a by the sulte oxidation method. Can J Chem
Eng 68:324326 (1990).
41 Linek V, Kordac M and Moucha T, Evaluation of the optical sulte
oxidation method for the determination ofthe interfacial mass
transfer area in small-scale bioreactors. BiochemEngJ 27:264268
(2006).
42 Duetz WAandWitholt B, Oxygentransfer byorbital shakingof square
vessels and deepwell microtiter plates of various dimensions.
BiochemEng J 17:181185 (2004).
43 Ortiz-Ochoa K, Doig SD, Ward JM and Baganz F, A novel method for
the measurement of oxygen mass transfer rates in small-scale
vessels. BiochemEng J 25:6368 (2005).
44 Sanchez A, Garcia F, Contreras A, Molina E and Chisti Y, Bubble-
column and airlift photobioreactors for algal culture. AIChE J
46:18721887 (2005).
45 Garcia-Ochoa F and Gomez E, Prediction of gasliquid mass transfer
in sparged stirred tank bioreactors. Biotechnol Bioeng 92:761772
(2005).
46 Gogate PR, Beenackers AACM and Pandit AB, Multiple-impeller
systems with a special emphasis on bioreactors: a critical review.
BiochemEng J 6:109144 (2000).
47 Garcia-Ochoa F and Gomez E, Mass transfer coefcient in stirrer tank
reactors for xanthan solutions. BiochemEng J 1:110 (1998).
48 Garcia-Ochoa F, Gomez E andSantos VE, Oxygentransfer anduptake
rates during xanthan gum production. Enzyme Microb Technol
27:680690 (2000).
49 Kilonzo PM and Margaritis A, The effects of non-Newtonian
fermentation broth viscosity and small bubble segregation on
oxygen mass transfer in gas-lift bioreactors: a critical review.
BiochemEng J 17:2740 (2004).
50 Hiraoka S, Kato Y, Tada Y, Kai S, Inoue N and Ukai Y, Power
consumption and gasliquid mass transfer volumetric coefcient
in a mechanically agitated vessel with wiregauze impeller. J Chem
Eng Jpn 34:600605 (2001).
51 Dumon E and Delmas H, Mass transfer enhancement of gas
absorption in oil-in-water systems: a review. Chem Eng Process
42:419438 (2003).
52 Clarke KG, Williams PC, Smit MS and Harrison STL, Enhancement and
repression of the volumetric oxygen transfer coefcient through
hydrocarbon addition and its inuence on oxygen transfer rate in
stirred tank bioreactors. BiochemEng J 28:237242 (2006).
53 Weuster-Botz D, Stevens S and Hawrylenko A, Parallel-operated
stirred-columns for microbial process development. Biochem Eng
J 11:6972 (2002).
54 Montes FJ, Catal an J and Gal an MA, Prediction of k
L
a in yeast broths.
Process Biochem 34:549555 (1999).
55 Shin CS, Hong MS and Lee J, Oxygen transfer correlation in high
cell density culture of recombinant E. coli. Biotechnol Technol
10:679682 (1996).
56 Badino Jr. AC, Facciotti MCR and Schmidell W, Volumetric oxygen
transfer coefcients (k
L
a) in batch cultivations involving non-
Newtonian broths. BiochemEng J 8:111119 (2001).
57 Yagi H and Yoshida F, Gas Absorption by Newtonian and non-
Newtonianuids inspargedagitatedvessels. IndEngChemProcess
Design Dev 14:488493 (1975).
58 Li GQ, Qiu HW, Zheng ZM, Cai ZL and Yang SZ, Effect of uid
rheological properties on mass transfer in a bioreactor. J Chem
Technol Biotechnol 62:385391 (1959). Please conrm or correct.
59 Nielsen DR, Daugulis AJ and McLellan PJ, A novel method of
simulating oxygen mass transfer in two-phase partioning
bioreactors. Biotechnol Bioeng 83:735742 (2003).
60 Gomes N, Aguedo M, Teixeira J and Belo I, Oxygen mass transfer in
a biphasic medium: inuence on the biotransformation of methyl
ricinoleate into -decalactone by the yeast Yarrowia lipolytica.
BiochemEng J 35:380386 (2007).
61 Quadros PA and Baptista CMSG, Effective interfacial area in agitated
liquid-liquid continuous reactors. Chem Eng Sci 58:39353945
(2003).
62 Torres-Martinez D, Melgarejo-Torres R, Gutierrez-Rojas M, Aguilera-
Vazquez M, Micheletti M, Lye GJ, et al, Hydrodynamic and oxygen
mass transfer studies in a three-phase (air-water-ionic liquid)
stirred tank bioreactor. BiochemEng J 45:209217 (2008).
63 B uchs J, Introductiontoadvantages andproblems of shakencultures.
BiochemEng J 7:9198 (2001).
64 Freyer SA, Konig M and Kunkel A, Validating shaking asks as
representative screening systems. Biochem Eng J 17:169173
(2004).
65 Peter CP, Lotter S, Maier U and B uchs J, Impact of out-of-phase
conditions on screening results in shaking ask experiments.
BiochemEng J 17:205215 (2004).
66 Zimmermann HF, Anderlei T, B uchs J and Binder M, Oxygen
limitation is a pitfall during screening for industrial strains. Appl
Microbiol Biotechnol 72:11571160 (2006).
67 B uchs J, Lotter S and Milbradt C, Out-of-phase operating conditions,
a hithertounknownphenomenoninshakingbioreactors. Biochem
Eng J 7:135141 (2001).
68 Delgado G, Topete MandGalindo E, Interactionof cultural conditions
and end-product distribution in Bacillus subtilis grown in shake
asks. Appl Microbiol Biotechnol 31:288292 (1989).
69 Maier U and B uchs J, Characterisation of the gas-liquid mass transfer
in shaking bioreactors. BiochemEng J 7:99110 (2001).
70 Mrotzek C, Anderlei T, Henzler HJ and B uchs J, Mass transfer
resistance of sterile plugs in shaking bioreactors. Biochem Eng
J 7:107112 (2001).
71 Henzler HJ and Schedel M, Suitability of the shaking ask for oxygen
supply to microbiological cultures. Bioprocess Eng 7:123131
(1991).
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198
1
1
9
7
Bioprocess scale-up: parameters to be used www.soci.org
72 Maier U, Losen M and B uchs J, Advances in understanding and
modeling the gasliquid mass transfer in shake asks. Biochem
Eng J 17:155167 (2004).
73 Kawase Y and Moo-Young M, Mathematical models for design of
bioreactors: applications of Kolmogoroffs theory of isotropic
turbulence. ChemEng J 43:B1941 (1990).
74 Gnielinski V, Berechnung mittlerer W arme- und Stoff uberg-
angskoefzienten an laminar und turbulent uberstr omten
Einzelk orpern mit Hilfe einer einheitlichen. Gleichung Forsch Ing
Wes 41:145153 (1975).
75 Mantzouridou F, Roukas T and Achatz B, Effect of oxygen transfer
rate on -carotene production from synthetic medium by
Blakeslea trispora in shake ask culture. Enzyme Microbiol Technol
37:687694 (2005).
76 Gupta A and Rao G, Astudy of oxygen transfer in shake asks using a
non-invasiveoxygensensor. Biotechnol Bioeng84:351358(2003).
77 Duetz WAandWitholt B, Oxygentransfer byorbital shakingof square
vessels and deepwell microtiter plates of various dimensions.
BiochemEng J 17:181185 (2004).
78 Duetz WA, R uedi L, Hermann R, OConnor K, B uchs J and Witholt B,
Methods for intense aeration, growth, storage, and replication
of bacterial strains in microtiter plates. Appl Environ Microbiol
66:26412646 (2000).
79 Funke M, Diederichs S, Kensy F, M uller C and B uchs J, The bafed
microtiter plate: increased oxygen transfer and improved online
monitoring in small scale fermentations. Biotechnol Bioeng
103:11181128 (2009).
80 Samorski M, M uller-Newen G and B uchs J, Quasi-continuous
combinedscatteredlight anduorescencemeasurements: anovel
measurement technique for shaken microtiter plates. Biotechnol
Bioeng 92:6168 (2005).
81 Doig SD, Pickering SCR, Lye GJ and Baganz F, Modelling surface
aeration rates in shaken microtitre plates using dimensionless
groups. ChemEng Sci 60:27412750 (2005).
82 Hermann R, Lehmann M and B uchs J, Characterisation of gasliquid
mass transfer phenomena in microtiter plates. Biotechnol Bioeng
81:178186 (2002).
83 Islam RS, Tisi D, Levy MS and Lye GJ, Scale-up of Escherichia coli
growth and recombinant protein expression conditions from
microwell to laboratory and pilot scale based on matched k
L
a.
Biotechnol Bioeng 99:11281139 (2008).
84 Islam RS, Tisi D, Levy MS and Lye GJ, Framework for the rapid
optimization of soluble protein expression in Escherichia coli
combining microscale experiments and statistical experimental
design. Biotechnol Prog 23:785793 (2007).
85 Shukla VB, Parasu Veera U, Kulkarni PR and Pandit AB, Scale-up of
biotransformation process in stirred tank reactor using dual
impeller bioreactor. BiochemEng J 8:1929 (2001).
86 Gill NK, Appleton M, Baganz F and Lye Gj, Design and
characterisation of a miniature stirred bioreactor system for
parallel microbial fermentations. BiochemEng J 9:164176 (2008).
87 Ascanio G, Castro B and Galindo E, Measurement of power
consumption in stirred vessels a review. Chem Eng Res Design
82:12821290 (2004).
88 Kim CH, Rao KJ, Youn DJ and Rhee SK, Scale-up of recombinant
hirudin production from Saccharomyces cerevisiae. Biotechnol
Bioprocess Eng 8:303305 (2003).
89 Rushton JH, Costich EW and Everett HJ, Power characteristics of
mixing impellers: part I. ChemEng Prog 46:395404 (1950).
90 Rushton JH, Costich EW and Everett HJ, Power characteristics of
mixing impellers: part II. ChemEng Prog 46:467476 (1950).
91 Nienow AE, Wisdom DJ and Middleton JC, Effect of scale and
geometry on ooding, recirculation and power in stirred vessels.
Proc Eur Conf Mix 2:116 (1977).
92 Oosterhuis NMG and Kossen NWF, Power input measurements in a
production scale bioreactor. Biotechnol Lett 3:645650 (1981).
93 Yawalkar A, Heesink ABM, Versteeg GF and Pangarkar VG, Gas-liquid
mass transfer coefcient in stirred tank reactors. Canadian J Chem
Eng 80:840848 (2002).
94 vant Riet K and Smith JM, The behaviour of gas-liquid mixtures near
Rushton turbine blades. ChemEng Sci 28:10311037 (1973).
95 Paglianti A, Takenaka K and Bujalski W, Simple model for power
consumption in aerated vessels stirred by Rushton disc turbines.
AIChE J 47:26732683 (2001).
96 Junker BH, Stanik M, Barna C, Salmon PandBuckland BC, Inuenceof
impeller type on mass transfer in fermentation vessels. Bioprocess
Biosyst Eng 19:403413 (1998).
97 Albaek MO, Gernaey KV and Stocks SM, Gassed and ungassed power
draw in a pilot scale 550 litre fermentor retrotted with up-
pumping hydrofoil B2 impellers in media of different viscosity and
with very high power draw. ChemEng Sci 63:58135820 (2008).
98 Hughmark GA, Power requirements andinterfacial areaingasliquid
turbine agitated systems. Ind Eng Chem Process Design Dev
19:638641 (1980).
99 Cui YQ, van der Lans RGJM and Luyben ChK, Local power uptake in
gasliquid systems with single and multiple Rushton turbines.
ChemEng Sci 51:26312636 (1996).
100 Mockel HO, Weissgarber H, Drewas EandRahner HJ, Modellingof the
calculation of the power input for aerated single and multistage
impellers with special respect to scale-up. Acta Biotechnol
10:215224 (1990).
101 Luong HT and Volesky B, Mechanical power requirements of
gasliquid agitated systems. AIChE J 25:893895 (1979).
102 Nielsen J, Villadsen J and Lid` en G, Scale-up of bioprocess. in
BioreactionEngineeringPrinciples, 2ndedn, ed. by Nielsen J. Kluwer
Academic/PlenumPublishers, New York, pp. 477517 (2003).
103 Zhang H, Williams-Dalson W, Keshavarz-Moore E and Shamlou PA,
Computational-uid-dynamics (CFD) analysis of mixing and
gasliquid mass transfer in shake asks. Biotechnol Appl Biochem
41:18 (2005).
104 B uchs J, Maier U, Milbradt C and Zoels B, Power consumption in
shaking asks on rotary shaking machines: I. power consumption
measurement in unbafed asks at lowliquid viscosity. Biotechnol
Bioeng 68:589593 (2000).
105 B uchs J, Maier U, Milbradt C and Zoels B, Power consumption in
shaking asks on rotary shaking machines: II. non-dimensional
description of specic power consumption and ow regimes in
unbafed asks at elevated liquid viscosity. Biotechnol Bioeng
68:594601 (2000).
106 Lotter S and B uchs J, Utilization of specic power input
measurements for optimization of culture conditions in shaking
asks. BiochemEng J 17:195203 (2004).
107 Kato Y, Peter CP, Akg un A and B uchs J, Power consumption and heat
transfer resistance in large rotary shaking vessels. Biochem Eng J
21:8391 (2004).
108 Sumino Y, Akiyama SandFukuda H, Performance of the shakingask
(I) power consumption. J Ferment Technol 50:203208 (1972).
109 Raval K, Kato Y and B uchs J, Comparison of torque method and
temperature method for determination of power consumption in
disposable shaken bioreactors. BiochemEng J 34:224227 (2007).
110 Lamping SR, Zhang H, Allen B and Shamlou PA, Design of a
prototype miniature bioreactor for high throughput automated
bioprocessing. ChemEng Sci 58:747758 (2003).
111 Gill NK, Appleton M, Baganz F and Lye GJ, Design and
characterization of a miniature stirred bioreactor system for
parallel microbial fermentations. BiochemEngJ 39:164176(2008).
112 Reyes C, Pe na C and Galindo E, Reproducing shake asks
performance in stirred fermentors: production of alginates by
Azotobacter vinelandii. J Biotechnol 105:189198 (2003).
113 Pe na C, Peter CP, B uchs J and Galindo E, Evolution of the specic
power consumption and oxygen transfer rate in alginate-
producing cultures of Azotobacter vinelandii conducted in shake
asks. BiochemEng J 36:7380 (2007).
114 Pe na C, Mill an MandGalindo E, Productionof alginatebyAzotobacter
vinelandii in a stirred fermentor simulating the evolution of power
input observedinshakeasks. ProcessBiochem43:775778(2008).
115 Oku S, Perez de Ortiz ES and Sawitowski H, Scale-up of liquid-liquid
dispersions in stirred tanks. Can J ChemEng 68:400406 (1990).
116 Cull SG, Lovick JW, Lye GJ and Angeli P, Scale-down studies on
the hydrodynamics of two-liquid phase biocatalytic reactors.
Bioprocess Biosyst Eng 25:143153 (2002).
117 Lovick J, Mouza AA, Paras SV, Lye GJ and Angeli P, Drop size
distribution in highly concentrated liquidliquid dispersions
using a light back scattering method. J Chem Technol Biotechnol
80:545552 (2005).
118 Pacek AW, Man CC and NienowAW, On the Sauter mean diameter
and size distributions in turbulent liquid/liquid dispersions in a
stirred vessel. ChemEng Sci 53:20052011 (1998).
119 Leng D and Calabrese R, Immiscible liquidliquid systems, in
Handbook of Industrial Mixing: Science and Practice, ed. by Paul EL,
J ChemTechnol Biotechnol 2010; 85: 11841198 c 2010 Society of Chemical Industry www.interscience.wiley.com/jctb
1
1
9
8
www.soci.org MPC Marques, JMS Cabral, P Fernandes
Atiemo-Obeng V and Kresta SMJ. Wiley & Sons, New Jersey,
pp. 639736 (2003).
120 B uchs J and Zoels B, Evaluation of maximum to specic power
consumption ratio in shaking bioreactors. J Chem Eng Jpn
34:647653 (2001).
121 Rocha-Valadez JA, Estrada M, Galindo E andSerrano-Carre on L, From
shake asks to stirred fermentors: scale-up of an extractive
fermentation process for 6-pentyl--pyrone production by
Trichoderma harzianum using volumetric power input. Process
Biochem41:13471352 (2006).
122 Junker BH, Scale-up methodologies for Escherichia coli and yeast
fermentation processes. J Biosci Bioeng 7:347364 (2004).
123 Hiruta O, Futamura T, Takebe H, Satoh A, Kamisaka Y, Yokochi T, et al,
Optimization and scale-up of -linolenic acid production by
MortierellaramannianaMM15-1, ahigh -linolenic acidproducing
mutant. J Ferment Bioeng 82:366370 (1996).
124 Dubey KK, Ray AR and Behera BR, Production of demethylated
colchicine through microbial transformation and scale-upprocess
development. Process Biochem43:251257 (2008).
125 Ju LK and Chase GG, Improved scale-up strategies of bioreactors.
Bioprocess Eng 8:4953 (1992).
126 Gerson DF and Kole MM, Quantitative measurements of mixing
intensity in shake-asks and stirred tank reactors: use of the
Mixmeter, a mixing process analyzer. Biochem Eng J 7:153156
(2001).
127 Nealon AJ, OKennedy RD, Titchener-Hooker NJ and Lye GJ,
Quantication and prediction of jet macro-mixing times in static
microwell plates. ChemEng Sci 61:48604870 (2006).
www.interscience.wiley.com/jctb c 2010 Society of Chemical Industry J ChemTechnol Biotechnol 2010; 85: 11841198

Você também pode gostar