SCALEDOWN OF BIOPHARMACEUTICAL
PURIFICATION OPERATIONS
ANURAG S. RATHORE and
VARSHA S. JOSHI
Chemical Engineering
Department, Indian
Institute of Engineering,
Hauz Khas, New Delhi,
India
INTRODUCTION
Biotech-based therapeutics have emerged as the growth
engine for the pharmaceutical industry (1). Successful
development and commercialization of biotech processes
require us to understand the relationships between the
various attributes of the product and the product’s clinical
safety and efficacy, as well as the effect of the differ-
ent process parameters on the product’s attributes (2).
Considering the complexity of the biotech products, under-
standing the latter requires a lot of experimentation.
Working at small scales offers a potential solution to this
problem.
There are many drivers for working at small scales
(3–5):
• Smaller the scale, lesser the requirement of feed
material. This is often an issue during early stages
of product development.
• Parallelization and automation offer an enormous
increase in productivity when working at small scale
when compared to manufacturing scale operations.
This allows us to examine a lot more parameters
than what we can otherwise.
• Higher productivity may also result in accelerated
timelines, something that can be very critical in the
pharmaceutical industry. In case of a blockbuster
drug (sales greater than $1 billion/year), each extra
day delay in development can result in excess of $1
million in lost sales revenue (3).
• Newer initiatives such as Quality by Design (QbD)
(6–8) and Process Analytical Technology (PAT)
(9–11) demand a deeper understanding of the
process and the product. This in turn requires more
extensive experimentation than what has been
traditionally done.
When using small-scale models for process development, it
is important that the small-scale model performs similar
to the corresponding manufacturing scale unit operation.
This ensures that the data from experimentation and
the conclusions reached from the data are applicable to
the manufacturing scale (12–14). Such small-scale models
may be used for a variety of purposes:
• process development and optimization;
Encyclopedia of Industrial Biotechnology: Bioprocess, Bioseparation, and Cell Technology, edited by Michael C. Flickinger
© 2013 John Wiley & Sons, Inc. Published 2013 by John Wiley & Sons, Inc.
• process characterization to support process valida-
tion (15–18). These include range finding studies to
set the acceptable ranges for the operating param-
eters, spiking studies to establish clearance of host
cell impurities [host cell proteins (HCPs) and DNA]
and other product-related impurities, and establish-
ing viral clearance capability (19–23). The latter is a
requirement for regulatory approval;
• establishing lifetime for reusable media for
chromatography and membrane separation steps
(19,24,25);
• manufacturing support for licensed commercial
processes to address manufacturing investigations,
postapproval process change for comparability, and
evaluate raw material changes postlicensure.
GENERAL CONSIDERATIONS
Use of Real Feed Streams and Raw Materials
When performing scale-down experiments, it is recom-
mended that we use feed streams that are representative
of manufacturing scale (24). It is best to use process inter-
mediates produced at the manufacturing scale itself. If
this is not possible then feed streams should be produced
at pilot scale that is representative of the manufactur-
ing scale. Furthermore, the raw materials used for the
experiments (buffer, resin, etc.) should be taken from the
cGMP-released inventory. This ensures that any addi-
tional variability in the experimentation originating from
use of noncomparable raw materials or feed materials is
avoided. Knowledge of the storage and stability of the feed-
stream material is acquired beforehand. Storage-related
degradation of the inventory could affect the outcome of
the small-scale runs (24).
Qualification of a Scale-Down Model
The qualification of the scale-down model must be carried
out as per current good manufacturing practice (cGMP)
validation guidelines that require a written qualification
protocol describing qualification studies to be conducted
and should include preapproved acceptance criteria for
key output parameters. For qualification, the runs are
performed in triplicate (or more) to assess the repro-
ducibility of the scale-down model and for meaningful
comparison to full-scale. In most cases, multiple perfor-
mance parameters need to be compared across scales to
establish comparability of performance and qualify the
scale-down model. Compared to the univariate compari-
son that is usually done, the application of multivariate
analysis (MVA) tools in performing this comparison has
been recently proposed (26). This allows for a more com-
prehensive comparison of the data and an examination
of all important inputs and output variables at the same
time. This allows for correlations between variables to
1
2 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS

be more easily identified. In addition, MVA projection value is kept constant. The cell culture solution is continu-
tools condense complex multivariate data sets into easily ously fed to the outside of the centrifuge and flows through
interpretable graphical projections (26,27). the spaces between the stacked discs toward the center of
In the following sections, we discuss major unit the centrifuge. Cells and particles are thrown radially out-
operations currently used in biopharmaceutical industry ward by the centrifugal force toward the underside of the
for downstream processing (Fig. 1) with respect to the upper disc and slide down the disc surface to the sediment
scale-down methodology and the review literature. holding space. The collected solids are either continuously
or discontinuously discharged from the centrifuge. The
liquid flows continuously to the top of the centrifuge and
CENTRIFUGATION is discharged under pressure. 
Unique expressions for exist for each centrifuge
With the recent developments in mammalian and micro- design. For a disc stack centrifuge (28),
bial cell culture, solids concentrations of up to 30% (w/v)  2π nd ω2 (r32 − r31 )
are readily achievable (28). For such cases, centrifugation = f1 (1)
is often the method of choice for liquid–solid separation 3g tan θ
because of its cost effectiveness, robust operation, and where, nd is the number of active discs, r1 and r2 are the
applicability to continuous operation (29,30). Harvesting inner and outer disc radius (in meters), g is acceleration
of bacterial and mammalian cells is commonly done using due to gravity (in meters per square second), and θ is
stacked disc centrifuges (31). the half disc angle. Furthermore, ω = 2π N, where ω is the
In this unit operation, the centrifugal force is used to angular velocity and N the rotational speed (in revolutions
separate mixtures of particles of varying masses or densi- per second). Also, f1 is the correction factor for spacer
ties suspended in a liquid. When a vessel is rotated at high caulks determined using the following expression:
speeds, the angular momentum yields an outward force to ⎛
2 ⎞
each particle that is proportional to its mass. This force   r1
3ZL BL ⎜ 1 − r2 ⎟
is countered by the so-called ‘‘drag’’ force that the liquid f1 = 1 − ⎝
3 ⎠ (2)
exerts on the particle. The net effect of ‘‘spinning’’ the 4π r2 r
1 − r1
2
sample in a centrifuge is that larger/denser particles move
outward faster than smaller/lighter particles. The rate of where ZL is the number of caulks on a disc and BL the
centrifugation is determined by the angular acceleration caulk width (25).
applied to the sample, typically measured in comparison For the laboratory centrifuge,
with the gravitational force (g). The sedimentation rate of
a particle is proportional to the liquid flow rate  through Vω2 (3 − 2x − 2y)
=
(3)
the centrifuge (Q) divided by the quantity sigma ( ). The 6g ln R2R o
o +R i
latter is the area of a simple gravity settling tank of equiva- 
lent sedimentation characteristics to that of the centrifuge. where is the equivalent settling area of a laboratory cen-
Theoretically, the performance of two centrifuges process-  trifuge, V is the sample volume, ω is the angular velocity,
ing the same feed material will be equivalent if the Q/ x and y are fractional times required for acceleration and

Mammalian or microbial cell culture

Microfiltration

Cell removal and concentration Centrifugation

Coagulation/flocculation
Ultrasonic vibrations
Cell disruption
Homogenizer (e.g. in case of intracellular expression)
Bead or ball mill
Microfiltration
Removal of cell debris
Centrifugation

Solubilization and refolding of IB proteins

Traditionally done by dilution
(e.g. in E.coil fermentation)

Protein precipitation/extraction Figure 1. Process flow diagram for

a typical biotech process used for
manufacturing of a biotech-based
therapeutics. The diagram is meant
Combination of UF/DF and to be illustration and not extensive,
chromatographic purification step as it only shows some of the options.
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
deceleration, respectively (measured for different centrifu-
gation speeds using a stopwatch), Ri is the inner radius
(the distance between the center of rotation and the top
of the liquid), and Ro is the outer radius (the distance
between the center of rotation and the bottom of the tube)
(32).
The performance of a continuous full-scale centrifuge
is often quantified by the amount of the solids remain-
ing after centrifugation and can be described using the
following relationship (4):
 
Qc
S=f  (4)
c design region where near complete particle recovery has
the solids remaining, Qc the volumetric flow
where S is
rate, and c the equivalent settling area.
Scale-Down Modeling Approach
Scale-downof the centrifugation step can be achieved by
keeping Q/ constant (33,34). Sigma theory modeling can
be generically applied to a wide number of centrifuge types
using calibration factors to account for changes in flow
patterns and centrifuge geometry (35). The feed flow rate,
centrifuge angular velocity, and system back pressure can
be suitably controlled. The temperature of the feed and
centrate is kept identical to that used in the manufacturing
scale. The smallest
 disc stack centrifuges available for
scale-down have values of 1800 m2 and can process
feed volumes of 2–20 L at flow rates of 10–60 L/h with
a 0.5 L solid holding volume. Production-scale disc stack
centrifuges can have sigma values as high as 1.3 × 105 m2 .
Thus, scale-down factors of 70-fold are achievable (12).
Scale-Down Qualification
The performance of the centrifugation step can be mea-
sured in terms of step yield, the clarity of the centrate,
filterability of the centrate, particle size distribution, and
extent of cell lysis. The clarity of the centrate can be
measured using particle counters, ultraviolet (UV) spec-
troscopy, or turbidimeter. Yield can be determined by
analyzing the feed material and the output for method by
high pressure liquid chromatography (HPLC).
Literature Review
Researchers have proposed the use of a benchtop
centrifuge as a scale-down model for pilot-scale equipment
(3). The focus of the investigation was on properties of the
protein precipitate, the fraction of solids recovered, and
the extent of dewatering achieved. Good agreement was
obtained at bench scale (a 1000-fold scale-down factor)
for all these parameters when compared to pilot-scale,
batch-feed operation. In addition, the methodology
developed here allowed for identification of the extent of
breakup that occurs in continuous-feed centrifuges when
processing shear-sensitive materials such as protein
precipitates. Other researchers have used the laboratory
scale-down model to predict the performance of disc stack
centrifuge for the recovery of yeast cell debris particles
(34). The scale-down model was also used to investigate
the extent of shear-sensitive protein aggregates during
3
recovery in continuous high speed centrifuges. The
authors concluded that such breakup can lead to over
10-fold reduction in separator capacity. Mannweiler and
Hoare (36) have described the methodology for 10-fold
scale-down of a disc stack centrifuge. Scale-down was
achieved by the reduction of the number of discs available
for separation purposes and by careful positioning of
these discs in the overall disc stack. The authors used a
combination of dye tracer and particle separation studies
to optimize the disc stack configuration. The resulting
grade efficiency curve was an accurate reflection of that
for the full-scale centrifuge, especially in the critical
been proposed. To date, several models that make good
predictions at the 10–100 mL scale have been proposed
(37–39). However, in recent years, there has been an
increased interest in models that are based on microtiter
plate technology (40–45). The use of a high speed rotating
shear device to reproduce shear conditions in combination
with benchtop centrifugation has successfully been used to
demonstrate that an ultra-scale-down (USD) model using
15-mL centrifuge tubes can give an accurate prediction
of the expected clarification of a mammalian cell culture
in industrial disc stack centrifuges (46). Tustian et al.
developed and verified a methodology in which USD
curves for centrifugal clarification were corrected by
diluting the feed material by 2%. After dilution, they
found that the USD curves accurately predicted pilot-scale
clarification performance of high cell density broths of
Saccharomyces cerevisiae and Escherichia coli cells (28).
Zaman et al. (47) have developed a USD-based approach
accounts for both gravitational and shear effects on
clarification performance.
In summary, despite advancements in scale-down of
centrifugation, the models that exist at present are suit-
able for knowledge generation and preliminary examina-
tion of the effects of the various process parameters. Final
determination of the acceptable ranges or the process
validation acceptance criteria still requires experimenta-
tion with the manufacturing scale centrifuge. This is due
to the complexity of the fluid dynamics in an industrial
centrifuge.
HOMOGENIZATION
E. coli is among the most frequently used microbial hosts
to produce recombinant therapeutic products. An intrinsic
characteristic of E. coli and some other similar expres-
sion systems is that they retain the expressed proteins
within the cell (48). Protein release in such cases requires
cell lysis induced chemically (49,50), enzymatically (51),
or mechanically. A robust and rapid way to achieve the
latter is by mechanical cell disruption via homogeniza-
tion by performing mechanical cell disruption followed by
cell debris removal by centrifugation (52,53). Options for
industrial-scale mechanical cell disruption include high
speed bead mills (54–57), high pressure homogenization
(HPH) (58–61), and ultrasonication. Most popular of these
in manufacturing of biotech-based therapeutics is HPH.
In this unit operation, cell suspension is forced by
pressures of several hundred bars to pass through an
4 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
adjustable discharge valve, which has a restricted orifice.
The combination of high shear, the high pressure drop,
and the impact forces in the discharge valve results in cell
disruption (62). Homogenizers may use discrete multiple
passes through a recycle loop to effect the multiple passes
needed for continuous processing. Disruption occurs via a
first-order process with respect to the number of passes
with the rate constant being a strong function of pressure
(61) and the valve and valve seat design (63). In addition,
the rate constant is affected by the conditioning of the cell
suspension (e.g. chemical pretreatment) and the nature of
the fermentation growth media. Cells grown in complex
media are more difficult to disrupt compared with those
grown in defined media (64).
Scale-Down Modeling Approach
Scale-down of an HPH-based process is complicated by
the fact that varying the flow rate by orders of magni-
tude results in significant variation in the gap opening of
the homogenization valve, with severe consequences on
the fluid dynamics and mechanical interaction between
the valve and cells. A procedure for calculating the pres-
sure drop as a function of gap opening has been reported.
This procedure has been validated for Stansted high pres-
sure homogenizers, which exhibit a quadratic dependence
on process medium flow rate (65). A typical industrial
scale homogenizer needs at least 30 L of the feed volume,
while a typical scale-down model may need up to 40 mL
of the feed volume. Thus, scale-down factors of 750-fold
are achievable. Inactivation depends not only on applied
pressure of homogenization and number of passes but also
on the microbial strain. For instance, in a single homoge-
nization pass at 250 MPa, six log cycles of inactivation can
be achieved for E. coli, five log cycles for S. cerevisiae but
only one log cycle for Lactobacillus delbrueckii. The latter
is characterized by a thicker and more resistant cell mem-
brane. Lethality can be increased not only by increasing
the pressure but also through multiple-pass treatments
(66). Since temperature rises during homogenization, one
should aim at keeping the temperature rise similar so as
to get similar step performance.
Scale-Down Qualification
The desired product recovery and purity after homoge-
nization can be calculated using HPLC. Generation of
aggregates can be monitored using size-exclusion high
pressure liquid chromatography (SE-HPLC). The log cycle
reduction in the feed should be determined at the end of the
homogenization cycle along with particle size distribution
of the cell debris and viscosity of the product. Temperature
and pressure should be monitored throughout the process.
Literature Review
Literature is scarce on the topic of scale-down model of high
pressure homogenizers. Donsi et al. (66) have compared
the homogenization efficiency at two scales (laboratory
scale and pilot scale) and demonstrated that the scale
of the equipment does not affect the level of inactivation
attained.
In summary, like centrifugation, homogenization too
is a challenging unit operation to scale-down. While the
existing models can be used to generate some fundamental
knowledge about the behavior of cells under shear stress,
the flow dynamics in most scale-down systems is quite
different from that of their large scale counterparts. This
necessitates that the final experimentation be done on
the manufacturing equipment itself for setting process
validation criteria and acceptable ranges.
REFOLDING
Protein refolding includes solubilization and reduction
of inclusion bodies (IBs) followed by renaturation (slow
removal of denaturant). In most cases, performance of the
refolding step is critical to overall recovery of the process.
There are many factors, both chemical (67–69) and phys-
ical (70–73), that can affect protein refolding. Refolding
by dilution is still the preferred technology for large-scale
refolding. The first step is solubilization of the IBs in a
mixing vessel. After any remaining undissolved IBs are
removed by centrifugation, the dissolved IBs are further
reduced and renatured in an agitated tank. Aggregation
is the major issue confronted during refolding and can
be caused by improper mixing and/or high local protein
concentration. Rapid and uniform dilution is necessary to
avoid aggregation.
Scale-Down Modeling Approach
It has been observed that the refolding yield is inversely
proportional to the protein concentration (67,69,74). Fur-
thermore, slower addition of the denatured protein to
refolding buffer has been known to result in higher yields
as the influence of macroscopic concentration gradients
is minimized even under poor mixing conditions (75,76).
It is recommended that a temperature-controlled and
agitator-driven closed vessel (such as a bioreactor) be used
as scale-down model. This way, engineering considerations
can be applied to mimic conditions at the manufacturing
scale. Chemical conditions such as amounts and concen-
tration of the various additives, sequence of addition,
temperature, pH, conductivity, and protein concentration
should be kept identical (Table 1). Physical aspects such
as mixing and distribution of oxygen in the tank could
be maintained similar by keeping the mass transfer coeffi-
cient (kL a) same across scales. Similar quenching protocols
followed by proper storage of the refolded protein should
be followed at both the scales.
Scale-Down Qualification
Product quality can be monitored during and after refold-
ing by a variety of generic (e.g. solubility and folding)
or specific (e.g. activity) methods (77). Functional assays
(bioassays, immunoassays, etc.) can be very informative.
Analytical techniques such as intrinsic fluorescence, cir-
cular dichroism (CD, secondary structure), dynamic light
scattering (DLS), turbidity measurements, or chromato-
graphic analysis (HPLC) can be used as indicators to mon-
itor performance of the refolding step (77–80). SE-HPLC
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
can be used to determine the amount of dimers or aggre-
gates generated in the refolding process. CD can be used
to measure protein secondary structure and DLS to mea-
sure protein aggregation so as to verify protein structure
(77,78). Crystallogenesis offers another approach for prod-
uct quality measurement as only properly folded proteins
with a uniform aggregation state (monodispersed sample)
yield well-ordered crystals. While the method of choice
will depend on the target protein or project, it is generally
recommended that more than one method be employed
(80).
Literature Review
Mannall et al. (81) have studied the USD approach to pro-
tein refolding in microwell plates and demonstrated the
applicability and scalability of this microwell screening
for the development of protein refolding processes. Qoron-
fleh et al. (82) have suggested the use of high throughput,
automated refolding screens. Ordidge et al. (83) have used
an automated robotic platform to develop a dilution refold
process-screening platform on which a hierarchical set of
assays rapidly determine optimal refolding conditions at
the microscale.
In summary, the typical bioreactor offers a platform
for performing refolding under conditions that mimic
manufacturing conditions. Furthermore, high throughput
process development (HTPD) platforms for performing
refolding have also been successfully demonstrated. We
expect these to be adopted by the biotech industry to
expedite process development, optimization, and charac-
terization of protein refolding step.
PRECIPITATION
Precipitation offers a low cost, high yield unit oper-
ation that can be used for separation of the various
host-cell-related and process-related contaminants (84).
Although the use of this step is somewhat limited in
production of therapeutic proteins because of the concerns
related to product stability, it is the step of choice for
certain applications (85). A common application is as
a primary isolation procedure in which the amounts of
contaminating proteins (and other species) are decreased
along with a significant reduction of process volume (86).
The basis of separation in this unit operation is reduc-
tion in the solubility of the product to induce precipitation.
This can be achieved via several methods: ionic precip-
itation (e.g., ammonium sulfate and sodium chloride),
temperature, pH, metal ions (e.g. Cu2+ , Zn2+ , and Fe2+ ),
nonionic polymers (e.g. polyethylene glycol), organic sol-
vents (e.g. ethanol and acetone), tannic acid, heparin,
dextran sulfate, cationic polyelectrolytes (e.g. protamines),
short-chain fatty acids (e.g. caprylic acid), trichloroacetic
acid (TCA), lectins (e.g. concanavalin A), group-specific
dyes (e.g. Procion Blue), and ligand–antibody interac-
tion (87). Of these methods, ionic precipitation utilizing
inorganic salts is the most commonly used method.
The precipitation step is generally used at an early
stage of the purification procedure, often on the clarified
extract, to obtain initial purification and concentration.
5
It entails manipulation of the feed solution to favor pro-
tein precipitation, mixing of the vessel contents as the
precipitation takes place, and subsequent solid–liquid
separation to separate the two phases. Mixing is a critical
aspect of this unit operation and impacts apparent protein
solubility, protein structure in precipitates, particle size,
morphology, and recoverable activity (22). Aggregate lev-
els and characteristics in the precipitation output impact
the performance of the subsequent centrifugal separation.
Protein solubility is affected primarily by the type and
concentration of precipitant, pH, and ionic strength (67,
88–91). The physical properties of the aggregates depend
on several factors: the type of precipitation reactor (90),
choice, concentration, and rate of addition of precipitat-
ing agent (89); initial protein concentration; and nature
and extent of mixing and residence time in the reactor
(68,69,71).
Batch reactors are the simplest type of precipitation
reactor and have been widely used in the industry. The
precipitating agent is slowly added to the protein solution
under mixing. Since the particles are exposed to a wide
range of shear stresses for a long period of time, the
aggregated protein particles tend to be compact, dense,
and mechanically stable.
Scale-Down Modeling Approach
Parameters that are likely to impact the performance of
the precipitation step need to be held constant (Table 1).
These include pH, conductivity, final precipitant concen-
tration, and temperature. Design parameters that need to
be maintained same across scales include tank geometry
(height-to-diameter ratio), impeller type, diameter of the
impeller relative to that of the tank, placement of the
impeller, baffle geometry, mean velocity gradient, aging
parameter (Camp number), and residence time (4). In
cases where the small-scale system does not mimic the
manufacturing design, dimensionless parameters such as
the mass transfer coefficient (kL a) can be used to target
identical mixing conditions across scales (Table 1) (92).
Qualification of Scale-Down Model
Product purity (measured by specific activity or chro-
matographic methods), product and total protein yield,
and impurity levels [assessed by sodium dodecyl sulfate
polyacrylamide gel electrophoresis (SDS-PAGE)] can be
used to compare step performance across scales. Addi-
tional measurements could include measurement of ion
concentrations (sodium content), turbidity of the super-
natant to gauge the extent of precipitate removal, and
the phase ratio and/or fraction of solids in the precipitate.
The kinetics of precipitation can be measured to establish
equivalence across scales (12).
Literature Search
Chhatre et al. have investigated the integrated appli-
cation of USD techniques and economic modeling as a
means for identifying optimal bioprocess operating con-
ditions. An economic model of this process was used to
identify where greatest product loss occurs and hence
6
Table 1. Summary of Scale-Down Approach for Different Purification Unit Operations—Input Parameters That are Held Constant During Scale-Down, Input
Parameters That are Altered During Scale-Down, and Performance Parameters That are Monitored to Ensure Qualification of the Scale-Down Model
Purification Step Operational (Input) Parameters That are Kept
Identical
Centrifugation Linear flow rate, temperature, sigma factor, feed
characteristics
Homogenization Linear flow rate, temperature, solution pH,
conductivity, protein concentration, feed
characteristics
Refolding Mass transfer coefficient (kL a), temperature,
chemical stock concentration, feed characteristics
Microfiltration Protein loading (g product/m2 membrane area),
number of diavolumes, volumetric concentration
factor, feed characteristic, temperature, linear
flow velocity, cross flow rate per channel,
transmembrane pressure
Precipitation Mass transfer coefficient (kL a), precipitant
concentration, precipitant-to-feed-volume ratio,
feed characteristics
Chromatography Bed height, buffers (pH, conductivity, etc.), resin,
linear flow rate, protein loading (g product/L
resin), pooling criteria, temperature, feed
characteristics
Ultrafiltration/Diafiltration Protein loading (g product/m2 membrane area),
number of diavolumes, volumetric concentration
factor, feed characteristics, temperature, linear
flow velocity, cross flow rate per channel,
transmembrane pressure
Viral inactivation Mass transfer coefficient (kL a), temperature,
contact time, chemical stock concentration, feed
characteristics
Virus removal filtration Protein loading (g product/m2 membrane area), feed
characteristics, temperature, linear flow velocity
Membrane adsorption Bed height, buffers (pH, conductivity, etc.), resin,
linear flow rate, protein loading (g product/L
resin), temperature, feed characteristics
Operational (Input)
Parameters That are
Altered
Process volumes, volumetric
flow rate, centrifugal speed
Process volumes, volumetric
flow rate, homogenizer
pressure, number of passes
Agitation rate
Process volumes, volumetric
flow rate, membrane area
Mixing speed, precipitant
feed rate, process volume
Column diameter, column
volume, volumetric flow
rate
Process volumes, volumetric
flow rate, membrane area
Agitation rate
Process volumes, volumetric
flow rate, membrane area
Column diameter, column
volume, volumetric flow
rate
Performance Parameters for Model Qualification
Product concentration and recovery, turbidity, cell
lysis
Product concentration and recovery, turbidity, cell
lysis, product-related impurities (aggregates)
Product concentration and recovery, conductivity,
pH, product-related impurities (aggregation,
deamidation, methylation, etc.)
Product concentration and recovery, turbidity, filter
capacity
Product concentration and recovery,
product-related impurities, process-related
impurities, host-cell-related impurities, turbidity
Product concentration and recovery,
chromatographic profiles, HETP, asymmetry,
conductivity, pH, product-related impurities
(aggregation, deamidation, methylation etc.),
process-related impurities (protein A, antifoam,
etc.), host-cell-related impurities (host cell
impurities, DNA, etc.)
Product concentration and recovery, turbidity, filter
capacity, pH, conductivity, excipient
concentration, product-related impurities,
host-cell-related impurities
Product concentration and recovery,
product-related impurities
Product concentration and recovery,
product-related impurities (aggregates)
Product concentration and recovery, conductivity,
pH, product-related impurities (aggregation,
deamidation, methylation, etc.), process-related
impurities (protein A, antifoam, etc.),
host-cell-related impurities (host cell impurities,
DNA, etc.)
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
where the largest opportunity cost was being incurred.
Scale-down analysis was used to characterize the
influence of the critical steps, identified as precipitation
and centrifugation, on recovery of lactoperoxidase
(LPO). The results demonstrate the power of combin-
ing USD data with models of economic and process
performance in order to establish the best overall
operating strategies for biopharmaceutical manufacture
(93). Boychyn et al. (4) have used a pilot-scale batch
stirred-tank reactor and a multichamber-bowl centrifuge
to predict the performance of larger reactors and
centrifuges.
In summary, the lack of literature on the scale-down
reflects the sparse usage of precipitation in production
of therapeutic proteins. This may gain more interest in
the future, as researchers develop new approaches toward
mainstream application of precipitation for achieving cost
effective way of protein purification.
CHROMATOGRAPHY
Chromatography is the primary workhorse of the biotech
industry for product purification (94). In the commonly
used bind and elute mode of chromatography, the first
step often involves binding of the product to the adsor-
bent. Impurities that do not bind to the column will just
flow through and go to waste. Then, a wash is conducted
to wash off remaining unbound and/or weakly bound
impurities. Finally, elution is performed by altering the
conditions of the mobile phase in the manner that differen-
tial binding of the protein of interest and other impurities
is exploited to achieve purification of the product. Alterna-
tively, the column can be operated in flow through mode,
whereby the impurities bind to the column and the product
flows through. A variety of chromatographic techniques
are employed in purification processes, mainly differ-
ing in the nature and structure of the stationary phase
(resin particles). These techniques exploit differences in
properties of proteins such as (i) size and shape (gel
filtration/size-exclusion chromatography); (ii) net charge
and distribution of charged groups (ion exchange chro-
matography); (iii) isoelectric point (chromatofocussing);
(iv) hydrophobicity (hydrophobic interaction chromatogra-
phy, reversed phase chromatography); (v) metal binding
(immobilized metal ion affinity chromatography); (vi) con-
tent of exposed thiol groups (covalent chromatography);
and (vii) biospecific affinities for ligands, inhibitors, recep-
tors, and antibodies (affinity chromatography) (95). In
recent years, the concept of multimodal, or mixed-mode
adsorption chromatography, is receiving an increasing
amount of attention. These stationary phases allow for
more than one ‘‘mode’’ of chromatography to be conducted
in a single step. In all cases, separation can be influenced
by factors such as the nature of ligand and matrix, solvent
pH, temperature, and size of the beads and pores (96). In
the following sections, we discuss some of the key consid-
erations when performing scale-down of chromatography
steps.
7
Size of Scale-Down Model
Chromatography is one of the more straightforward unit
operations to scale-down (13). The size of the scale-down
system varies with the intended use of the system. For
performing binding studies, a 96-well format HTPD plat-
form for chromatography may suffice and is likely to be the
most efficient (5). For performing resin screening studies,
the above mentioned HTPD platform or small columns
(∼ 2.5 cm height, 1 mL) can be used to screen for var-
ious matrices and mobile phase compositions in order
to perform preliminary optimization (97). Viral clearance
studies are often done with shorter columns (5–10 cm bed
height) because of the difficulties and expenses associated
with creating virus spikes (98,99). Final process character-
ization studies that form the basis for setting the process
validation acceptance criteria and the acceptable ranges
for operating parameters are often performed using a col-
umn with diameter of at least 1.6 cm and same bed height
as at the manufacturing scale (97). Column diameters of
less than 1.0 cm are generally not used in these studies
since the ratio of system dead volume to column volume
can result in pool concentrations significantly lower than
those seen at large scale (13). It should be noted that when
compared to the higher end of column size in manufac-
turing scale (2 m), the scale-down factors may range from
1:100 to over 1:1000000 (12).
Chromatographic Equipment
Automated small-scale chromatography systems, such as
the äKTA Explorer, excel at these scale-down studies.
They offer precise, accurate, and reproducible control of
flow rates and gradients. However, accuracy and precision
of in-line detectors (pH, conductivity, temperature) should
be calibrated at reasonable frequency by using off-line
probes (13).
Scale-Down Modeling Approach
Column scale-down is normally performed by maintaining
bed height while reducing column diameter, which main-
tains the residence time (Table 1). In this way, the linear
flow rate is maintained, while the volumetric flow rate is
reduced (100). The International Conference on Harmo-
nization (ICH) guideline on viral safety states that the
validity of the scaling down should be demonstrated and
that the column bed height should be shown to be repre-
sentative of commercial scale manufacturing (101). Thus,
if the bed height is different at small scale, as in case of
viral clearance studies, the linear flow velocity should be
adjusted appropriately to keep the residence time constant
(102).
The operating conditions of the experiments should be
identical to those in manufacturing. This is so as subtle
changes in temperature, pH, or conductivity could affect
the retention time of proteins on chromatographic resins.
Heat exchangers on the inlet tubing and column jack-
eting maybe necessary for accurate, consistent, and pre-
cise temperature control during scale-down. For example,
hydrophobic interaction chromatography has been report-
edly sensitive to temperature fluctuations and could give
8 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
rise to product retention and selectivity with variation
in temperature (12). Some of the significant operational
parameters include the following:
• Protein loading (g protein/L resin) should be
identical. In size-exclusion chromatography, loading
should be based on volume percentage of the column
bed volume.
• Gradient lengths (in number of column volumes) and
slopes (e.g. increase in percent limit buffer per unit
time) should be kept identical.
• Temperature, pH, and conductivity should be main-
tained identical. Differences in temperature across
scales can result in significant changes in conductiv-
ity and thus impact step performance.
• Bed height of the chromatography column and the
linear flow velocity are generally kept constant across
scales.
• Pooling criteria should be the same as in manufactur-
ing. They can be online (conductivity or absorbance)
or off-line (gels or HPLC). If the pooling criteria are
UV based, it is a good practice to ensure that the UV
detectors of the small-scale systems correlate well
with those of the large-scale system. UV detectors
may be inaccurate or nonlinear over the optical den-
sity (OD) ranges required. If the absorbance range is
nonlinear at 280 nm, it may be prudent to measure
the online OD at 300 nm or some other wavelength
that has a lower extinction coefficient (13).
Qualification of Scale-Down Model
Attributes that need to be compared would depend on the
objective of the step and could include chromatogram,
product concentration, product recovery, quality
attributes (concentration of the various product-related,
process-related, and host-cell-related impurities), and
other feed stream characteristics (such as pH and
conductivity).
Factors at the large scale such as column-header design
and packing can be difficult to control and may result in
reduced accuracy of the scale-down model (13). Differences
in column packing at the two scales may have an impact
on separation and result in differences in chromatograms.
Height equivalent to theoretical plate (HETP) and asym-
metry factor (As ) are typically used to evaluate quality of
column packing (103,104). While these values are likely
to be different across scales, they should be monitored to
ensure consistency of packing quality during small-scale
experimentation. Furthermore, differences in equipment
such as tubing length and diameter, monitors, and pumps
should also be carefully examined to ensure comparability
of performance across scales (12).
Table 2 summarizes the literature on scale-down or
process chromatography. Researchers have used micro-
liter resin wells (as in 96-well plates) for performing
chromatography. A lot of effort has also undergone into
modeling of chromatography so that the data generated
from scale-down models can be applied toward prediction
of performance at manufacturing scale. The optimal bal-
ance between the high throughput that the microdevices
offer and the higher level of comparability that comes in
the laboratory-scale columns is likely to be different for
different applications. For most cases, however, the best
would be to use the HTPD platforms for examining a large
number of process parameters in wide ranges followed
by use of the traditional laboratory-scale chromatography
columns for performing the final process characterization
experiments that support process validation acceptance
criteria. In the future, the HTPD platforms are likely to
continue to evolve and gradually with increased rigor they
may be able to replace the laboratory-scale experimenta-
tion completely.
MICROFILTRATION AND
ULTRAFILTRATION/DIAFILTRATION (UF/DF)
Microfiltration
Microfiltration (MF) membranes can be used to harvest
cells from mammalian and bacterial cell culture to provide
cell-free conditioned medium (containing the product) to
the downstream purification steps. The pore sizes of typical
MF membranes range from 0.1 to 0.8 μm. MF membranes
retain cells and cellular debris by a sieving mechanism
at the membrane surface. Pressure is applied to the feed
stream to force permeate through the membrane. Pro-
teins and small-molecular-weight species (media salts)
pass through the membrane and are collected in the per-
meate stream (12).
Ultrafiltration (UF) and diafiltration (DF) are com-
monly used in downstream processing for product con-
centration and buffer exchange, respectively. In some
cases, it is also used for fractionation or purification.
Membranes are rated by molecular weight cutoffs, called
nominal molecular weight cutoffs (NMWC). Membrane
structure can influence the quality and robustness of sep-
aration. Minimizing macrovoids enables the use of higher
pressures and higher temperatures and gives better sep-
arations because of sharp NMWC. Cassettes are the first
choice for most protein processing. Hollow fibers are more
appropriate when the product is sensitive to shear or
a high viscosity solution needs to be processed such as
adenoviral vectors and influenza vaccines (113).
Size is the basis of separation in filtration in most
cases. The solution is contacted with the membrane under
an applied pressure (114). The applied pressure forces
the solvents or buffer salts and the smaller molecules to
migrate through the membrane. The membrane retains
larger molecules such as proteins. MF and UF operations
can be performed in a tangential flow filtration (TFF)
mode or normal flow filtration (NFF) mode. An example of
NFF would be dead-end or cartridge filters where cells are
retained at the membrane surface or within the membrane
structure. As cells build up at the membrane surface, the
filtration rate, or flux, decreases as the resistance to perme-
ate flow through the membrane increases. In TFF, the feed
solution is passed tangentially over the membrane surface.
Cells build up on the membrane surface as the permeate
flows through the membrane and are swept off the mem-
brane and recirculated back to the feed tank (51). UF is
typically conducted in a TFF mode where the feed solution
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS 9

Table 2. Review of Literature for Development of Scale-Down Models for Process Chromatography

Topic Summary References

General
Affinity chromatography
Cation exchange
chromatography
Expanded Bed chromatography
Hydrophobic interaction
chromatography (HIC)
Annular chromatography
Different formats of microscale devices for scale-down of chromatographic separations. (94)
Advice on how best to operate these devices, examples of use, comparison with
respect to metrics such as cost, usability, and role in QbD
Scale-down experimentation integrated with general rate modeling, a 2.5 cm height, (98)
1 mL column model calibrated against 20 cm height columns from 40 mL to 160 L
volume. Simulations in agreement with experimental results. Overall objective to
reduce the number of experiments, shorten development time, and reduce costs
Novel ultra-scale-down approach that uses empirical correlations derived from (105)
conductivity changes during operation to correct chromatographic profiles for
differences in dispersion and retention. Tested successfully with 1 mL column to
predict elution profiles of a chimeric monoclonal antibody obtained from protein A
chromatography columns of 3 mL and 18.3 L scale
Development of software that uses data obtained from scale-down column studies to (106)
relate yield to protein load and the number of matrix reuses to enable decision
making for chromatographic separations of antibodies
Mass balance of protein and Fab fragments, measurement of capacity from batch (107)
uptake kinetics and use of an imaging technique to investigate the impact of varying
column loading and resin reuse on matrix performance. Using a 20-fold scale-down,
number of reuses before recovery starts decreasing was established, as in current
large-scale manufacturing operations
Influence of pH for optimized binding and removal of HCP. Initial screening was done (108)
by batch-binding capacities, followed by scale-down to 96-well plate format and
proved that assay miniaturization still provided reliable data
Use of scale-down and scouting studies to assist in scaling up a method for isolating (109)
potentially clinically relevant monoclonal antibody C595 from heterogeneous
feedstocks
The hydrodynamic model was in agreement with experimental results obtained with 1- (110)
and 5-cm column diameters with buffer solutions of different viscosities. Supported
by visual observations and scale-down model predictions, it was concluded that
operating conditions (liquid viscosity and superficial velocity) resulting in a bed-void
fraction between 0.7 and 0.75 would provide the optimal separation efficiency in
terms of plate number
Suitability of expanded bed chromatography as a scale-down technique to predict the (111)
behavior at large-scale was examined. The performance of three columns of diameters
5.0, 1.0, and 0.5 cm were compared in terms of the bed expansion, hydrodynamics,
and breakthrough for lysozyme adsorption onto STREAMLINE-SPTM. Performance
was similar across a scale-down factor of a 100-fold
Development and scale-up of a preparative HIC step for a therapeutic protein, removal (97)
of product- and process-related impurities, and direct transferability of the
scale-down optimization experiments via the pilot scale to the final production scale
Mathematical model to estimate the minimum radius of an annular chromatograph. (112)
The minimal geometry of an annular chromatograph was calculated for a model
system immunoglobulin and bovine serum albumin.

is recirculated over the membrane and then returned to
the feed vessel. UF membranes are available in nominal
molecular weight limits of 1000–1,00,000 Da (115–118).
A typical system consists of a feed pump, strainer, rack
equipped with filtration modules, filtrate/backwash tank,
and connecting piping and valves. A separate pump is
used to add DF buffer to the recirculation vessel. Control
parameters may include feed and retentate pressure (the
permeate pressure is usually atmospheric for UF appli-
cations), temperature, and the feed flow rate per feed
channel.
Scale-Down Modeling Approach
TFF has traditionally been scaled up by maintaining the
following parameters constant: filtrate volume to mem-
brane surface area ratio, transmembrane pressure, cross
flow velocity, membrane material and pore size, channel
height, flow path geometry, and retentate and filtrate pres-
sures (Table 1). The membrane area is linearly decreased
as per the desired feed volume such that the filtrate volume
to membrane surface area ratio stays constant (119,120).
In the case of hollow fibers, linear scaling with hollow
fibers is readily accomplished by producing scale-down
modules in which fiber length is equal to that of industrial
scale systems. Equal flow distribution between fibers and
cartridges is easily accomplished through proper piping
manifold and cartridge housing design. Flat sheet designs,
however, provide a suitable geometric format for a wide
range of linear-scale UF.
The challenge with using scale-down systems for
performing MF or UF/DF is that the small-scale devices
that are commercially available may differ from their
10 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
large-scale counterparts with respect to their design.
While cassette designs are based on parallel addition of
membrane channels, the feed-port channels are added
in series. This directly impacts the fluid dynamics of
the manufacturing scale cassettes and the scale-down
devices, resulting in differences in flow distribution,
channel height compression, and entrance/exit effects. All
these factors impact the performance of the step. Equal
flow distribution throughout the system is important to
maintain both the system average and system limit values
of feed channel pressure drop (PL ), transmembrane
pressure (TMP; PTM ), filtrate flux (Jf ), and protein
concentration at the membrane wall (Cw ) constant at
all scales of operation. Channel height compression is
dependent on the physical deformation characteristics
of membrane, spacer screen, encapsulant, and gaskets.
Variations in channel height will have a significant impact
on both pressure losses and mass transfer coefficients.
Hence, process performance will vary at different scales of
operation if consistent compression is not achieved (119).
Qualification of Scale-Down Model
Since MF and UF/DF steps do not alter the product purity
(in most cases) and are generally used for clarification
or concentration or buffer exchange, step recovery may
be determined using UV absorbance at 280 nm. In cases
where the sample matrix interferes with the UV signal,
use of HPLC or product-specific assays may be required.
Mass balance is required for determining if the product
is being lost into the permeate stream, left behind in
the holdup volume, or irreversibly bound to membrane.
A comparison of the flux versus TMP curve across scales
provides a good comparison of performance. If the UF sys-
tem is being used for purification, analysis of the impurity
concentration is a key measure of step effectiveness. It is
important that appropriate cleaning is performed on the
small-scale system so as to achieve consistent performance
over long-term use (12).
Literature Review
A comprehensive methodology for implementation of lin-
ear scale-up and scale-down of TFF process has been devel-
oped by Reis and coworkers (119). The researchers modi-
fied the cassette manufacturing procedures and tolerances
to achieve linear scale-up of UF processes for recovery for
a ribosomal deoxyribonucleic acid (rDNA)-derived human
therapeutic. A 400-fold linear scale-up was achieved with-
out intermediate pilot-scale tests. In another publication,
novel pleated scale-down devices (Am = 1.51 to 15.1 ×
10−3 m2 ) have been designed and fabricated. It was
shown that these can more accurately predict the per-
formance of industrial scale single-use pleated membrane
cartridges (Am = 1.06 m2 ) that are commonly used within
biopharmaceutical manufacture (121). The authors found
that single-use scale-down cartridges retain the same
pleat characteristics of the larger cartridges but require
a reduced feed volume by virtue of a substantially dimin-
ished number of active membrane pleats. A 1000-fold
scale-down was achieved (121). A method to mimic flux
and transmission performance in a laboratory scale cross
flow operation by an USD rotating disc filter (RDF) has also
been reported (122). The RDF has been modified by build-
ing in inserts to allow the flexibility of the chamber volume,
so that only 1.5 mL of processing material is required for
each DF experiment. Wall shear rate correlations have
been established for at both the laboratory-scale cas-
sette and the USD device, and identical performance was
achieved by operating both scales under conditions with
equivalent-averaged shear rates. Predicted flux and trans-
mission data agreed well with the experimental results of
a laboratory-scale DF where the cassette resistance was
considered (122). In another publication, the authors have
described operation of a continuous mode laboratory fil-
ter to prepare filter cakes and filtrate similar to what
may be achieved at the industrial scale (123). Analysis of
the filtration rate profile indicated the filter cake to have
changing properties (compressibility) with time. Results
from the USD filter were used to predict how the cake
compressibility changes with time and thus allowed for
accurate predictions about the flux profile in large-scale
filtration. Furthermore, the USD filter allowed the pro-
duction of materials (cake and filtrate) representative of a
large-scale continuous filter.
In summary, there has been significant progress in
creating scale-down devices for UF/DF unit operation in
the past two decades. Two aspects remain that require
some consideration from the users. First is to ensure that
the design of the scale-down device is comparable to that
of the large-scale device so that the flow pattern is similar
if not identical. Second, the dead volume of the system
(volume of the connected tubings, pump, and other parts
of the setup besides the cassette) is always significant
when compared to the cassette volume at small scale.
Very high dead volume will result in underperformance of
the UF/DF step and limit the usefulness of the scale-down
model. The user should carefully select the equipment that
has a low dead volume.
VIRAL CLEARANCE VIA CHROMATOGRAPHY AND
FILTRATION
Assuring the viral safety of plasma-derived biologicals and
biopharmaceuticals is critical for safety of health care con-
sumers. Incidences of contamination of products derived
from human plasma in the past have adversely impacted
the health of hundreds of patients and tainted the image of
certain segments of the health care industry (124). Owing
to stringent measures taken by the industry and regula-
tors to mitigate viral safety risks, similar incidents have
not occurred in the history of biotech-based therapeutics in
the recent decades (124). Viral clearance can occur either
via process chromatography (the virus not binding to the
resin while the product binds) or nanofiltration (viruses
cannot pass through the nanofilters). The aspects related
to process chromatography are already covered in the
above section. In this section, we focus on viral clearance
via nanofiltration.
Nanofiltration is a pressure-driven process where per-
meate is forced through a semipermeable membrane with
typical pore size in the range of 15–50 nm (newer products
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
have pores in the narrower range of 15–20 nm). This pore
size distribution allows for passage of the product and
buffer ingredients while retaining viral particles. Viral
clearance devices are used once and then discarded. Viral
clearance and design considerations have been reviewed
in several publications (125–127).
Virus clearance studies are always performed with
scale-down models of the manufacturing process steps.
It is not feasible to perform viral clearance studies at the
manufacturing scale because it would be inappropriate
to introduce infectious virus into a cGMP manufactur-
ing facility. Also, the volumes of virus needed to achieve
a satisfactory spiking level at the manufacturing scale
would be impractical and prohibitively expensive. Thus,
in order for the virus clearance studies to be extrapo-
lated to the manufacturing scale, it is imperative that the
scale-down model is a true representation of the full-scale
manufacturing process. Typically, the virus is spiked into
the relevant intermediate, and the spiked material is pro-
cessed across the scaled down unit operation (both for
process chromatography and nanofiltration). The reduc-
tion in the virus load by the unit operation demonstrates
the effectiveness of the process step for virus removal or
inactivation. The virus spike used in viral clearance stud-
ies should be representative of a potential contaminant to
the extent achievable. It is important not only to select
appropriate relevant or model viruses but also to consider
the properties of the virus spike (124). For example, the
presence of serum in a virus spike may be problematic for
a validation study of a serum-free manufacturing scheme.
It is important that contaminants in the virus spike itself
do not impact key or critical performance parameters in
a way that makes the scale-down model unrepresentative
of the large-scale process (116). The product fraction post
filter is then assayed for residual virus infectivity. This
enables a reduction factor to be calculated as follows:
Viral reduction
Total virus measured in spiked load
= log10
Total virus recovered in the product fraction
(5)
Scale-Down Modeling Approach
The loading, as measured in volume of load material fil-
tered per unit surface area of the filter, should be kept
constant across the two scales. A typical load material
from an earlier processing step obtained from manufac-
turing should be used for the scale-down studies. Since
this step is usually carried out under constant pressure,
the inlet pressure should be matched for the two scales and
similar flush volumes in terms of liters of buffer flushed to
recover any protein held up in the pores per unit surface
area should be used. An important output parameter is
the average volumetric flux as a function of volumetric
capacity.
The membranes used for viral clearance can be oper-
ated in direct flow or TFF mode. For viral clearance
applications, it is required that viral clearance devices
have an integrity test that is correlated to virus removal
performance. The manufacturer generally provides this
11
correlation. The operation of scale-down viral clearance
systems can present complications. When using multi-
layer membrane disks for evaluation, specifically designed
holders may have to be used to provide the required seal-
ing between the membrane sheets to prevent feed material
from bypassing layers. Scale-down viral clearance devices
may have the same membrane in the production scale
devices, but the feed-side flow geometries of the scale-down
devices may also not be consistent with the production
modules. Feed-side path lengths may not be consistent
and feed flows may have to be adjusted to provide equiva-
lent shear forces at the membrane surface. Similar to UF
application, air–liquid interfaces should be minimized to
prevent protein denaturation and aggregation (12).
Scale-Down Qualification
The step recovery and purity for viral clearance can be
determined using HPLC or product-specific assays. Aggre-
gate generation can be monitored using SE-HPLC. When
pure proteins are used, UV absorbance at 280 nm can be
used for yield measurements. Creating a mass balance
is useful to estimate losses coming from dead volume in
the system or irreversible binding to the membrane. A
comparison of the process flux as a function of time and
pressure over the whole process is also an indicator of
comparability of performance.
Literature Review
Zhou et al. (128,129) have evaluated different scale-down
models and proposed a new model to mimic the liquid flow
path found in the large-scale capsule. They were able to
achieve same capacity at small scale as at larger scale.
Results from a four-model virus study with a redesigned
Sartobind Q absorber scale-down model at the new pro-
cess capacity were presented. The authors have discussed
problems faced with the Q-membrane scale-down model
as well as the proposal of the new model. The proposed
Q-membrane scale-down model had a process capacity
greater than 3000 g/m2 or 10.7 kg/L with a log reduction
in viability (LRV) over 5 for four-model viruses.
In summary, scale-down modeling for viral clearance
through process chromatography and nanofiltration is
widely practiced as every biotech company does these
studies as per the regulatory requirements. The existing
models work reasonably well, but as discussed in the ref-
erences mentioned in this section, there is scope of further
improvements with respect to the scale of operation. If the
scale can be further reduced, the cost of the studies will be
significantly lowered as well.
VIRAL INACTIVATION
Viral inactivation, along with other steps for viral clear-
ance discussed above, plays a key role in biotech processes
in demonstrating virus clearance capability. It is a part
of the monoclonal antibody platform in most companies
(130). Most commonly used approach for viral inactivation
is acidic pH/low pH inactivation, and this is the focus of
this section.
12 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
The inactivation of enveloped viruses by low pH is a
very common step in a monoclonal antibody purification
process. The working principle is that some viruses, when
exposed to a low pH, denature spontaneously. Typically,
low pH inactivation step is performed after the protein A
affinity chromatography step. The product from the pro-
tein A affinity column is titrated to a pH of 3.8 or lower
and incubated for a period of 15–60 min, depending on
the stability of the monoclonal antibody under considera-
tion. This brief exposure to low pH effectively inactivates
most lipid-enveloped viruses. Following this incubation,
the protein solution is then titrated upward to a pH of
5.0 or higher to prepare for the next step. While the
inactivation of viruses is favored by low pH, there is an
inherent risk of aggregating the target protein at these
low pH conditions. Moreover, a strongly acidic solution
that is used as a titrating solution for the low pH step can
cause localized low pH conditions, if not adequately mixed.
This could potentially cause aggregation of the protein
solution.
Scale-Down Modeling Approach
Scale-down of a low pH inactivation step is relatively easy.
Care must be taken to ensure adequate mixing at this
step during the addition of the inactivating agent such
that no undue aggregation of the target protein results
across this step. As mentioned before, the mass transfer
coefficient (kL a) can be used to define mixing conditions at
small scale that will mimic conditions used in the man-
ufacturing scale. The scale-down factors for these types
of virus inactivation steps may be large, as there are few
limitations to designing a temperature-controlled contact-
ing vessel, which simply needs to maintain well-mixed
solution. Test tubes or small vessels may be used, if
they are immersed in a temperature-controlled heating
block, and either mixed before or during contact. The
level of the heating solution should be above the liq-
uid level in the test vessel to ensure all the test solu-
tion contents have reached the target temperature. A
high temperature short-time treatment can be employed
using microwave assistance, but this system needs careful
scale-down using miniature model of the process equip-
ment, which could limit the scale-down to more modest
ratios. The mixing of the vessel contents plays impor-
tant role in low pH assisted inactivation. Standard mixing
studies must be employed and the kinetics of inactivation
should be measured, as it is a regulatory requirement.
The addition rate of inactivation chemicals should match
the manufacturing process, and the final product con-
centration of acid should be same as that of full scale.
Scale-down variables that should be controlled through-
out the inactivation process are pH, temperature, con-
tact time, and the final inactivation chemical concentra-
tions.
Scale-Down Qualification
The aggregate content of the protein during scale-down
studies is an important quality attribute to match across
scales for qualification of the scale-down model. The chem-
icals used for inactivation should show concentration
profile similar to that used in the manufacturing scale
(12).
Literature Review
Not much has been written on the topic of scale-down of
viral inactivation, perhaps because of the simple design of
such a system.
In summary, scale-down of viral inactivation is quite a
simple exercise. However, care should be taken that the
mixing is equivalent so that the end results are comparable
across scales.
MEMBRANE ADSORBERS
Membrane chromatography is increasingly becoming pop-
ular as an alternative to certain chromatographic steps,
offering higher productivity (131–138). In the membranes,
binding sites are located along the through pores rather
than nestled with long diffusive pores. Even in the newer
chromatography resins with ‘‘through’’ pores or convective
pores (139), the liquid has the choice to flow through the
particle or around it. Unlike this, in a stacked membrane,
there is no choice but to flow through the pores. This
reduces the mass transport resistance for the solute to the
matrix by eliminating pore diffusion, leaving film diffusion
from the core of the liquid to the membrane surface in the
interior of a through pore as the only transport resistance.
As film diffusion is usually orders of magnitude faster
than pore diffusion, mass transport limitations are drasti-
cally reduced in membrane chromatography, shifting the
limitation of the process more to the properties of the
matrix–solute interactions.
Scale-Down Modeling Approach
Most of the scale-down systems that are available in the
market do not have a liquid flow path similar to that
found in the large-scale capsule and thus tend to generate
extremely high operational backpressure when a high flow
rate is applied (≥ 450 cm/h). The axial velocity of a mobile
phase is faster in the center of the membrane adsorber
(MA) unit than near the edges of adsorptive bed (106).
The decrease in apparent porosity from the center of the
bed to the gasket area or the edges of the adsorptive bed
may account for the decrease in velocity observed adjacent
to the gasket or edges (106,140). In contrast, radial flow
adsorbers, prepared by spirally winding a flat sheet mem-
brane over a porous cylindrical core, are used in a capsule
for large-scale process (114). Distorted or poor inlet flow
distribution has less effect in a large-scale capsule, thus
leading to a smaller operational backpressure drop for
large-scale operation (17,18). Despite being very complex
to model in a radially outward direction, radial flow adsor-
bers are claimed to be suitable for large-scale application
(114). The scalability and operating backpressure, there-
fore, are main issues for the MA scale-down model. An
extremely high operational backpressure observed with
the scale-down model can lead to considerably less capacity
and an oversized membrane at full scale. This oversizing
may lead to an erroneous economic calculation. Larger
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS 13

scale membranes do not have backpressure issues. How- LIST OF ABBREVIATIONS

ever, the use of a large-scale MA in a scale-down model
is also not practical, especially for a viral clearance study CD circular dichroism
because of the high cost of viruses and feedstock. cGMP current good manufacturing practice
DLS dynamic light scattering
DNA deoxyribonucleic acid
Scale-Down Model Qualification HCP host cell proteins
These aspects are same as in process chromatography. HETP height equivalent to theoretical plate
HPH high pressure homogenization
HPLC high pressure liquid chromatography
Literature Review HTPD high throughput process development
IBs inclusion bodies
Zhou et al. have redesigned scale-down models for
ICH International Conference on
Q-membrane, Q125 and Q40 (128,129). These scale-down
Harmonization
models mimic the liquid flow path found in the large-scale
LPO lactoperoxidase
modules and achieve performance parameters comparable
MA membrane adsorber
to those of the larger Q units. However, even with
MF microfiltration
improved scale-down devices, the operational backpres-
MVA multivariate analysis
sure for Q-membrane chromatography still influences
NFF normal flow filtration
process capacity. Several researchers have focused on
NMWC nominal molecular weight cutoffs
creating an improved scale-down model for membrane
OD optical density
chromatography (128,129).
PAT process analytical technology
In summary, the newer scale-down models seem to
RDF rotating disc filter
have addressed the most critical of the problems that one
rDNA ribosomal deoxyribonucleic acid
faces. However, not all manufacturers offer such devices.
SDS-PAGE sodium dodecyl sulfate polyacrylamide gel
Careful comparison of the performance of the device should
electrophoresis
be done in order to understand the differences between
SE-HPLC size-exclusion high pressure liquid
the two scales, and the resulting data should be suitably
chromatography
interpreted.
TCA trichloroacetic acid
TFF tangential flow filtration
UF ultrafiltration
SUMMARY USD ultra-scale-down
UV ultraviolet
In the era of QbD and PAT, use of scale-down models
has become an integral part of characterizing and vali-
dating industrial processes. Characterization studies form
the backbone of experimentation intended to provide pro- NOMENCLATURE
cess knowledge. These studies are used to identify critical
parameters and to establish acceptable ranges, process 
equivalent settling area of a laboratory centrifuge
validation acceptance criteria, and specifications. Process 2
 (m )
validation studies such as evaluation of viral clearance and 2
c equivalent settling area (m )
impurity clearance require such scale-down models, as it As asymmetry factor
is not feasible from a safety perspective to perform them at BL caulk width (m)
manufacturing scale. This article presents general guide- Ca camp number
lines and relevant parameters affecting scale-down of the Cw protein concentration at the membrane wall
various unit operations involved in purification of proteins. (kg/m3 )
The key parameters, assessment methods, assessment PL feed channel pressure drop (kg/ms2 )
techniques, and challenges involved in scale-down are PTM transmembrane pressure (Pa)
discussed. The quality of the existing scale-down mod- f1 correction factor for spacer caulks
els varies from unit operation to unit operation. Factors g acceleration due to gravity (m/s−2 )
that complicate the scale-down process include difficulties G average value of velocity gradient
in mimicking the hardware design resulting in differ- Jf filtrate flux (m3 /s)
ences in hydrodynamics across scales (mixing, shear, kL a mass transfer coefficient
etc.), higher holdup volumes at smaller scales, and dif- LRV log reduction in viability
ferences in materials used for making the equipment and N rotational speed (rps)
the separation media. A lot of progress has been made nd number of active discs
in each of these areas but with the growing importance ω angular velocity (radians/s)
of scale-down models, scope for more work remains. We P power dissipated (W)
hope the chapter would be valuable to those engaged in Po power number
small-scale development, optimization, and characteriza- Q liquid flow rate through the centrifuge (m3 /s)
tion of biotech processes. Qc volumetric flow rate (m3 /s)
14 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
r1 inner disc radius (m)
r2 outer disc radius (m)
Re Reynolds number
Ri inner radius (the distance between the center of
rotation and the top of the liquid) (m)
Ro outer radius (the distance between the center of
rotation and the bottom of the tube) (m)
S solids remaining (kg)
θ half disc angle
TMP transmembrane pressure (Pa)
μ dynamic viscosity (Pa·s)
V sample volume (L)
x fractional time required for acceleration (s)
y fractional time required for deceleration (s)
ZL number of caulks on a disc
REFERENCES
1. Wheelwright SM. J Biotechnol 1989; 11: 89–102.
2. Rathore AS. Trends Biotechnol 2009; 27: 698–705.
3. Willoughby N, Martin P, Titchener-Hooker N. Biotechnol
Bioeng 2004; 87: 641–647.
4. Boychyn M, Doyle W, Bulmer M, More J, Hoare M. Biotech-
nol Bioeng 2000; 69: 1–10.
5. Bhambure R, Kumar K, Rathore AS. Trends Biotechnol
2010; 29: 127–135.
6. Rathore AS, Winkle H. Nat Biotechnol 2009; 27: 26–34.
7. Rathore AS. Trends Biotechnol 2009; 27: 546–553.
8. Xu Z, Li J, Zhou JX. Prep Biochem Biotechnol 2012; 42:
183.
9. Read EK, Park JT, Shah RB, Riley BS, Brorson KA, Rathore
AS. Biotechnol Bioeng 2010; 105: 276–284.
10. Read EK, Park JT, Shah RB, Riley BS, Brorson KA, Rathore
AS. Biotechnol Bioeng 2010; 105: 285–295.
11. Rathore AS, Bhambure R, Ghare V. Anal Bioanal Chem
2010; 398: 137–154.
12. Godavarti R, Petrone J, Robinson J, Wright R, Kelley BD,
Bolton GR. In: Rathore AS, Sofer G, editors. Process valida-
tion in manufacturing of biopharmaceuticals. Boca Raton
(FL): Taylor & Francis Group LLC; 2005. pp. 69–142.
13. Rathore AS, Krishnan R, Tozer S, Smiley D, Rausch S,
Seely J. BioPharm Int 2005; 18: 58–60.
14. Tait AS, Aucamp JP, Bugeon A, Hoare M. Biotechnol Bioeng
2009; 104: 321–331.
15. Cecchini DJ. In: Rathore AS, Mhatre R, editors. Quality by
design for biopharmaceuticals. Hoboken, New Jersey John
Wiley and Sons; 2008. pp. 127–142.
16. Kozlowski S, Swann P. In: Rathore AS, Mhatre R, editors.
Quality by design for biopharmaceuticals. Hoboken, New
Jersey John Wiley and Sons, 2008, pp. 9–30.
17. Points to Consider in the Manufacture and Testing of
Monoclonal Antibody Products for Human Use. Center for
Biologics Evaluation and Research, F.D.A.; 1997.
18. Banerjee A. Biopharm Int 2010; 23: 26.
19. Adner N, Sofer G. BioPharm 1994; 7: 44–48.
20. Leonard MW, Sefton L, Costigan R. In: Kelley BD,
Ramelmier RA, editors. Validation of biopharmaceuti-
cal manufacturing processes. Washington (DC): American
Chemical Society; 1998. pp. 55–68.
21. Breece T, Gilkerson E, Schmelzer C. BioPharm 2002; 15:
16–20.
22. Kelley BD. In: Sofer G, Zabriskie D, editors. Biopharmaceu-
tical process validation. New York: Marcel Dekker; 2000.
pp. 29–59.
23. Box GEP, Hunter WG, Hunter JS. Statistics of experiments.
New York: John Wiley and Sons; 1978.
24. Seely R, Wright H, Fry H, Rudge S, Slaff G. BioPharm 1994;
7: 41–48.
25. Rathore AS, Sofer G. In: Rathore AS, Sofer G, editors.
Process validation in manufacturing of biopharmaceuticals.
Boca Raton (FL): Taylor & Francis Group LLC; 2005. pp.
169–204.
26. Pieracci J, Yusuf-Makagiansar H. In: Rathore AS, Sofer
G, editors. Process validation in manufacturing of biophar-
maceuticals. Boca Raton, FL, USA CRC press, Taylor and
Francis Group; 2012.
27. Wold S, Geladi P, Esbensen K, Öhman J. J Chemom 1987;
1: 41–56.
28. Tustian AD, Salte H, Willoughby NA, Hassan I, Rose MH,
Baganz F, Hoare M, Titchener-Hooker NJ. Biotechnol Prog
2007; 23: 1404–1410.
29. Axelsson H. In: Flickinger MC, Drew SW, editors. Encyclo-
pedia of bioprocess technology. New York: John Wiley and
Sons; 1999. pp. 513–531.
30. Hanle DD. In: Flickinger MC, Drew SW, editors. Encyclo-
pedia of bioprocess technology. New York: John Wiley and
Sons; 1999. pp. 553–559.
31. Kempken R, Preissmann A, Berthold W. J Ind Microbiol
1995; 14: 52–57.
32. Maybury JP, Mannweiler K, Titchener-Hooker NJ, Hoare
M, Dunnill P. Bioprocess Biosyst Eng 1998; 18: 191–199.
33. Leung WWF. Industrial centrifugation technology. New
York: McGraw Hill; 1998.
34. Maybury JP, Hoare M, Dunnill P. Biotechnol Bioeng 2000;
67: 265–273.
35. Mosqueira FG, Higgins JJ, Dunnill P, Lilly MD. Biotechnol
Bioeng 1981; 23: 335–343.
36. Mannweiler K, Hoare M. Bioprocess Eng 1992; 8: 18–25.
37. Betts JI, Baganz F. Microb Cell Fact 2006; 5: 21.
38. Gill NK, Appleton M, Baganz F, Lye GJ. Biochem Eng J
2008; 39: 164–176.
39. Pampel L, Mcnerney T, Wypych J. Pap Am Chem Soc 2004;
227: 264–BIOT.
40. Duetz WA. Trends Microbiol 2007; 15: 469–475.
41. Fernandes P, Cabral JMS. Biocatal Biotransformation
2006; 24: 237–252.
42. Islam RS, Tisi D, Levy MS, Lye GJ. Biotechnol Prog 2007;
23: 785–793.
43. Islam RS, Tisi D, Levy MS, Lye GJ. Biotechnol Bioeng 2008;
99: 1128–1139.
44. Micheletti M, Lye GJ. Curr Opin Biotechnol 2006; 17:
611–618.
45. Micheletti M, Barrett T, Doig SD, Baganz F, Levy MS,
Woodley JM, Lye GJ. Chem Eng Sci 2006; 61: 2939–2949.
46. Hutchinson N, Bingham N, Murrell N, Farid S, Hoare M.
Biotechnol Bioeng 2006; 95: 483–491.
47. Zaman F, Allan CM, Ho SV. Biotechnol Prog 2009; 25:
1709–1716.
48. Sommer B, Friehs K, Flaschel E, Reck M, Stahl F, Scheper
T. J. Biotechnol 2009; 140: 194–202.
 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
49. Ariga O, Watari T, Andoh Y, Fujishita Y, Sano Y. J Ferm
Bioeng 1989; 68: 243–246.
50. Zhao FS, Yu JY. Biotechnol Prog 2001; 17: 490–494.
51. Malamy MH, Horecker BL. Biochemistry 1964; 3:
1889–1893.
52. Clarkson AI, Lefevre P, Titchener-Hooker NJ. Biotechnol
Prog 1993; 9: 462–467.
53. van Hee P, Middelberg APJ, van der Lans RGJM, van der
Wielen LAM. Biotechnol Bioeng 2004; 88: 100–110.
54. Kloosterman J, Vanwassenaar PD, Slater NKH, Vander-
padt A. Chem Eng J 1988; 37: B47–B54.
55. Limon-Lason J, Hoare M, Orsborn CB, Doyle DJ, Dunnill
P. Biotechnol Bioeng 1979; 21: 745–774.
56. Ricci-Silva ME, Vitolo M, Abrahao-Neto J. Process Biochem
2000; 35: 831–835.
57. Schutte H, Kula MR. Biotechnol Appl Biochem 1990; 12:
599–620.
58. Engler CR. Bioprocess Technol 1990; 9: 95–105.
59. Hetherington PJ, Follows M, Dunnill P, Lilly MD. Trans
Inst Chem Eng 1971; 49: 142–148.
60. Middelberg APJ. Trans IChemE 1993; 71C: 215–219.
61. Sauer T, Robinson CW, Glick BR. Biotechnol Bioeng 1989;
33: 1330–1342.
62. Keshavarz-Moore E, Hoare M, Dunnill P. Enzyme Microb
Technol 1990; 12: 764–770.
63. Kelly WJ, Muske KR. Bioprocess Biosyst Eng 2004; 27:
25–37.
64. Gray PP, Dunnill P, Lilly MD. In: Terui G, editors. Fermen-
tation technology today. Japan: Society of Fermentation
Technology; 1972. pp. 347–351.
65. Floury J, Bellettre J, Legrand J, Desrumaux A. Chem Eng
Sci 2004; 59: 843–853.
66. Donsı̀ F, Ferrari G, Lenza E, Maresca P. Chem Eng Sci
2009; 64: 520–532.
67. Clark EDB. Curr Opin Biotechnol 1998; 9: 157–163.
68. Hevehan DL, De Bernardez Clark E. Biotechnol Bioeng
1997; 54: 221–230.
69. Yasuda M, Murakami Y, Sowa A, Ogino H, Ishikawa H.
Biotechnol Prog 1998; 14: 601–606.
70. Boyle DM, Buckley JJ, Johnson GV, Rathore AS, Gustafson
MG. Biotechnol Appl Biochem 2009; 54: 85–92.
71. Buswell AM, Middelberg AP. Biotechnol Bioeng 2003; 83:
567–577.
72. Lee C, Buswell A, Middelberg AP. Chem Eng Sci 2002; 57:
1679–1684.
73. Mannall GJ, Titchener-Hooker NJ, Chase HA, Dalby PA.
Biotechnol Bioeng 2006; 93: 955–963.
74. Goldberg ME, Rudolph R, Jaenicke R. Biochemistry 1991;
30: 2790–2797.
75. Katoh S, Katoh Y. Process Biochem 2000; 35: 1119–1124.
76. Katoh S, Sezai Y, Yamaguchi T, Katoh Y, Yagi H, Nohara
D. Process Biochem 1999; 35: 297–300.
77. Cowieson NP, Wensley B, Listwan P, Hume DA, Kobe B,
Martin JL. Proteomics 2006; 6: 1750–1757.
78. Vincentelli R, Canaan S, Campanacci V, Valencia C, Maurin
D, Frassinetti F, Scappucini-Calvo L, Bourne Y, Cambillau
C, Bignon C. Protein Sci 2004; 13: 2782–2792.
79. Willis MS, Hogan JK, Prabhakar P, Liu X, Tsai K, Wei Y,
Fox T. Protein Sci 2005; 14: 1818–1826.
80. Rathore AS, Sharma A, Chillin D. BioPharm 2006; 19:
48–57.
15
81. Mannall GJ, Myers JP, Liddell J, Titchener-Hooker NJ,
Dalby PA. Biotechnol Bioeng 2009; 103: 329–340.
82. Qoronfleh MW, Hesterberg LK, Seefeldt MB. Protein Expr
Purif 2007; 55: 209–224.
83. Ordidge GC, Mannall GJ, Liddell J, Dalby PA, Micheletti
M. Biotechnol Prog 2012; 28: 435444
84. Fisher RR, Glatz CE. Biotechnol Bioeng 1988; 32: 777–785.
85. Glatz CE, Hoare M, Landa-Vertiz J. AIChE J 1986; 32:
1196–1204.
86. Rothstein F. Protein process engineering. New York: Marcel
Dekker; 1994.
87. Bell DJ, Dunnill P. Biotechnol Bioeng 1982; 24: 1271–1285.
88. Smoluchowski M. Z Phys Chem 1917; 92: 129–168.
89. Rushton JH, Costich EW, Everett HJ. Chem Eng Prog 1950;
46: 467–476.
90. Bell DJ, Brunner K-H. Filt Sep 1983; 20: 274–286.
91. Mannweiler K, Titchener-Hooker NJ, Hoare M. IChemE
Symp, Adv Biochem Eng/Biotechnol 1989; 1: 105–117.
92. Garcia-Ochoa F, Gomez E. Biotechnol Adv 2009; 27:
153–176.
93. Chhatre S, Pampel L, Titchener-Hooker NJ. Biotechnol
Prog 2011; 27: 998–1008.
94. Chhatre S, Titchener-Hooker NJ. J Chem Technol Biotech-
nol 2009; 84: 927–940.
95. Watanabe E, Tsoka S, Asenjo JA. Ann N Y Acad Sci 1994;
721: 348–364.
96. Errson B, Ryden L, Janson JC. In: Janson JC, editor.
Protein purification: principles, high resolution methods,
and applications. New Jersey: John Wiley and Sons; 2011.
pp. 8–9.
97. Graumann K, Ebenbichler A. Chem Eng Technol 2008; 28:
1398–1407.
98. Gerontas S, Asplund M, Hjorth R, Bracewell DG. J Chro-
matogr A 2010; 1217: 6917–6926.
99. Rubino M, Bailey M, Baker JC, Boose JA, Metzka L, Moore
V, Quertinmont M, Wiler W. In: Rathore AS, Sofer G,
editors. Process validation in manufacturing of biopharma-
ceuticals. Boca Raton (FL): CRC Press; 2005. pp. 545–564.
100. Smith TM, Wilson E, Scott RG, Misczak JW, Bodek JM,
Zabriskie DW. In: Kelley BD, Ramelmier RA, editors.
Validation of biopharmaceutical manufacturing process.
Washington (DC): American Chemical Society; 1998. pp.
80–92.
101. Viral safety of biotechnology products derived from cell lines
of human or animal origin, ICH Viral Safety Document.
International Conference on Harmonization; 1998.
102. Yamamoto S, Nomura M, Sano Y. J Chromatogr 1987; 409:
101–110.
103. Barry A, Chojnacki R. BioPharm 1994; 7: 43–47.
104. Sofer G, Hagel L. Handbook of process chromatography: a
guide to optimization, scale-up, and validation. New York:
Academic Press; 1997.
105. Hutchinson N, Chhatre S, Baldascini H, Davies JL,
Bracewell DG, Hoare M. Biotechnol Prog 2009; 25:
1103–1110.
106. Chhatre S, Thillaivinayagalingam P, Francis R,
Titchener-Hooker NJ, Newcombe AR, Keshavarz-Moore E.
Biotechnol Prog 2007; 23: 888–894.
107. Thillaivinayagalingam P, O’Donovan K, Newcombe AR,
Keshavarz-Moore E. J Chromatogr B Analyt Technol
Biomed Life Sci 2007; 848: 88–96.
16 SCALEDOWN OF BIOPHARMACEUTICAL PURIFICATION OPERATIONS
108. Stein A, Kiesewetter A. J Chromatogr B Analyt Technol
Biomed Life Sci 2007; 848: 151–158.
109. Denton G, Murray A, Price MR, Levison PR. J Chromatogr
A 2001; 908: 223–234.
110. Fenneteau F, Aomari H, Chahal P, Legros R. Biotechnol
Bioeng 2003; 81: 790–799.
111. Ghose S, Chase H. Bioseparation 2000; 9: 21–28.
112. Uretschlager A, Jungbauer A. J Chromatogr A 2000; 890:
53–59.
113. Hagel L, Jagschies G, Sofer G. Handbook of process chro-
matography: development, manufacturing, validation and
economics. Netherland: Academic Press; 2008.
114. Mulder M. Basic principles of membrane technology. Dor-
drecht, The Netherlands: Kluwer Academic Publishers;
1996.
115. Ma G, Aucamp J, Gerontas S, Eardley-Patel R, Craig A,
Hoare M, Zhou Y. Biotechnol Prog 2010; 26: 466–476.
116. Reynolds T, Boychyn M, Sanderson T, Bulmer M, More J,
Hoare M. Biotechnol Bioeng 2003; 83: 454–464.
117. Englard S, Seifter S. Methods Enzymol 1990; 182:
285–3006.
118. Paul EL, Rosas CB. Chem Eng Prog 1990; 8: 17.
119. Stavrinides S, Ayazi Shamlou P, Hoare M. In: Ayazi
Shamlou P, editor. Processing of solid–liquid suspensions.
London: Butterworth Heinman; 1993. pp. 118–158.
120. Harrison RG. Protein purification process engineering. New
York: Marcel Dekker; 1993.
121. Foster PR, Dunnill P, Lilly MD. Biotechnol Bioeng 1976;
18: 545–580.
122. Iyer HV, Przybycien TM. AIChE J 1994; 40: 349–359.
123. Iyer HV, Przybycien TM. Biotechnol Bioeng 1995; 48:
324–332.
124. Asher D, Brorson K, Hotta J, Hughes J, Martin J, Ruppach
H, Sofer G, Wisher M, Willkommen H, Yang B Preparation
of virus spikes used for virus clearance studies. Technical
Report No. 47. University of Iowa: Parenteral Drug Associ-
ation, Inc.; 2010.
125. Levy RV, Philips MW, Lutz H. In: Meltzer TH, Jornitz
MW, editors. Filtration in biopharmaceutical industry. New
York: Marcel Dekker; 1998. pp. 619–646.
126. Kelley BD, Petrone J. In: Wang WK, editor. Membrane
separations in biotechnology. New York: Marcel Dekker;
2001. pp. 351–396.
127. Huang PY, Peterson J. In: Wang WK, editor. Membrane
separations in biotechnology. New York: Marcel Dekker;
2001.
128. Zhou JX, Tressel T, Gottschalk U, Solamo F, Pastor A,
Dermawan S, Hong T, Reif O, Mora J, Hutchison F, Murphy
M. J Chromatogr A 2006; 1134: 66–73.
129. Zhou JX, Tressel T. Biotechnol Prog 2006; 22: 341–349.
130. Shukla AA, Hubbard B, Tressel T, Guhan S, Low D. J
Chromatogr B Analyt Technol Biomed Life Sci 2007; 848:
28–39.
131. Rathore AS, Shirke A. Prep Biochem Biotechnol 2011; 41:
307–315.
132. Thommes J, Kula MR. Biotechnol Prog 1995; 11: 357–367.
133. Brandt S, Goffe RA, Kessler SB, O’Connor JL, Zale SE.
Bio/Technology 1988; 6: 779–782.
134. Briefs KG, Kula MR. Chem Eng Sci 1992; 47: 141–149.
135. Charcosset C. J Chem Technol Biotechnol 1998; 71: 95–110.
136. Lutkemeyer D, Bretschneider M, Buntemeyer H, Lehmann
J. J Chromatogr A 1993; 639: 57–66.
137. Reif O-W, Freitag R. J Chromatogr A 1993; 654:29–41.
138. Riese U, Lutkemeyer D, Heidemann R, Buntemeyer H,
Lehmann J. J Biotechnol 1994; 34: 247–257.
139. Afeyan NB, Fulton SP, Gordon NF, Maszaroff I, Varady L,
Regnier F. Bio/Technology 1990; 8: 203–206.
140. Fisher RR, Glatz CE, Murphy PA. Biotechnol Bioeng 1986;
28: 1053–1063.
Please note that the abstract and keywords will not be included in the printed book, but are required for the
online presentation of this book which will be published on Wiley’s own online publishing platform.
If the abstract and keywords are not present below, please take this opportunity to add them now.
The abstract should be a short paragraph upto 200 words in length and keywords between 5 to 10 words.

Abstract: Biotech-based therapeutics have emerged as the growth engine for the pharmaceutical industry. Successful
development and commercialization of biotech processes require the understanding of the relationships between the
various attributes of the product and the product’s clinical safety and efficacy, as well as, the effect of the different
process parameters on the product’s attributes. The complexity of the biotech products and the requirement for rigorous
experimentation justify working at small scales.
This article presents general guidelines and relevant parameters affecting scale-down of the various unit operations
involved in purification of proteins. The key parameters, assessment methods, assessment techniques, and challenges
involved in scale-down are discussed.

Keywords: scale-down; biopharmaceutical; purification; quality by design (QbD);