ARTICLE

A Scale-Down Mimic for Mapping the Process

Performance of Centrifugation, Depth, and Sterile
Filtration
Adrian Joseph,1 Brian Kenty,2 Michael Mollet,2 Kenneth Hwang,2 Steven Rose,2
Stephen Goldrick,1 Jean Bender,2 Suzanne S. Farid,1 Nigel Titchener-Hooker1
1
The Advanced Centre of Biochemical Engineering, Department of Biochemical
Engineering, University College London, Bernard Katz Building, London, WC1E 6BT,
United Kingdom; telephone: þ44207 679 9586; fax: þ44207 916 3943;
e-mail: nigelth@ucl.ac.uk
2
MedImmune LLC Gaithersburg Headquarters, One MedImmune Way, Gaithersburg,
Maryland

Introduction
ABSTRACT: In the production of biopharmaceuticals disk-stack
centrifugation is widely used as a harvest step for the removal of Quality by Design (QbD) regulatory initiatives over the last decade
cells and cellular debris. Depth filters followed by sterile filters are have required that biopharmaceutical manufacturers develop a
often then employed to remove residual solids remaining in the thorough understanding of a product’s quality attributes and
centrate. Process development of centrifugation is usually manufacturing process through the generation of design spaces
conducted at pilot-scale so as to mimic the commercial scale (Rathore, 2009). High throughput scale-down techniques now enable
equipment but this method requires large quantities of cell culture
and significant levels of effort for successful characterization. A the rapid generation of extensive experimental data representative of
scale-down approach based upon the use of a shear device and a large-scale performance, both in the upstream and downstream
bench-top centrifuge has been extended in this work towards a manufacturing process. Such large experimental data sets generated
preparative methodology that successfully predicts the perfor- through these techniques allow for better identification of the effects
mance of the continuous centrifuge and polishing filters. The use of and interactions of the input parameters on the process performance
this methodology allows the effects of cell culture conditions and
large-scale centrifugal process parameters on subsequent filtration and product quality (Titchener-Hooker et al., 2008).
performance to be assessed at an early stage of process development Continuous flow disk-stack centrifugation is often used to
where material availability is limited. harvest commercial-scale cell culture processes because of its
Biotechnol. Bioeng. 2016;9999: 1–8. robustness and relatively low running costs (Axelsson, 2002). The
ß 2016 The Authors. Biotechnology and Bioengineering Published intermittent discharge provided by the disk-stack centrifuge
by Wiley Periodicals, Inc. enables removal of a considerable quantity of cells and large
KEYWORDS: centrifugation; continuous centrifugation; scale- cellular debris in a semi continuous fashion. However, one of the
down; primary recovery; mammalian cell; disk-stack centrifuge; disadvantages of disk-stack centrifugation is that in many designs
depth filter; capillary shear; filter capacity the cells enter the centrifuge through a feed zone in which high
levels of shear are present. This shear can damage shear-sensitive
mammalian cells resulting in the generation of submicron particles
which are carried over to the centrate (Jain et al., 2005). These fine
particles can cause subsequent fouling in later chromatographic
This is an open access article under the terms of the Creative Commons Attribution processes, resulting in high column pressures and accompanied
License, which permits use, distribution and reproduction in any medium, provided the reductions in column lifetime and efficiency (Kempken et al., 1995).
original work is properly cited. In order to avoid chromatographic column fouling, depth filtration
Abbreviations: CSD, capillary shear device; RSD, rotating shear device; USD, ultra
scale down; LDH, lactate dehydrogenase; mAb, monoclonal antibody; NADH, is often used immediately after centrifugation to remove these
nicotinamide adenine dinucleotide; PEEK, polyetheretherketone; QbD, quality by submicron particles (Yigzaw et al., 2006). Hence, a typical process
design. sequence for a mammalian cell culture process might begin with the
Correspondence to: N. Titchener-Hooker
Received 20 October 2015; Revision received 19 February 2016; Accepted 24 February removal of cells and cell debris achieved through a combination of a
2016 centrifugal step followed by a depth filtration step (Shukla and
Accepted manuscript online xx Month 2016; Kandula, 2008).
Article first published online in Wiley Online Library
(wileyonlinelibrary.com). In order to scale among different types of centrifuges,
DOI 10.1002/bit.25967 correction factors are used to account for the deviations from

ß 2016 The Authors. Biotechnology and Bioengineering Published by Wiley Periodicals, Inc. Biotechnology and Bioengineering, Vol. 9999, No. xxx, 2016 1
ideal conditions such as those caused by differences in flow Materials and Methods
patterns (Mosqueira et al., 1981). Typically, Sigma theory is used
to scale centrifuges irrespective of size, geometry, and type Theoretical Considerations
(Ambler, 1959). However, Sigma theory does not take into
A detailed discussion of scale-down centrifugation is given in prior
account the generation of small particles through cell damage in
studies (Boychyn et al., 2001; Hutchinson et al., 2006). In essence by
the high shear regions of centrifuge feed zones. In order to
maintaining the ratio of flow rate to equivalent settling area and
capture accurately these effects at laboratory-scale, the shear
accounting for deviations from ideal flow conditions, Sigma theory
generated in the feed zone needs to be mimicked (Boychyn et al.,
enables the direct comparison of clarification efficiencies between
2001). A Rotating Shear Device (RSD) has been developed to
centrifuges of different sizes and geometries (Ambler, 1959). Hence,
reproduce the prevailing shear conditions in such feed zones
this methodology can be applied to predict the clarification
(Boychyn et al., 2001). Operated in combination with a bench-top
performance of a disk-stack centrifuge using a laboratory bench-top
centrifuge, this has been shown to successfully predict the
centrifuge:
clarification efficiency of mammalian cell culture in a pilot-scale
Qds V lab
disk-stack centrifuge (Hutchinson et al., 2006). The RSD has also ¼ ; ð1Þ
cds Sds t lab clab Slab
been used in conjunction with micro well plates using sub-
millilitre volumes to successfully model pilot-scale centrifugation where Qds is the volumetric flow rate into the disk-stack centrifuge,
performance (Tait et al., 2009). Vlab is the volume of material used in the laboratory centrifuge, tlab
In the routine development of primary recovery operations, the is the centrifugation time, and cds and clab are correction factors that
performances of both centrifugal and filtration processes need to account for the non-ideal flow properties in the disk-stack and
be characterized. The RSD was developed as an ultra scale-down laboratory scale centrifuges, respectively. The correction factor for
(USD) tool to investigate under well-controlled and -defined the disk-stack centrifuge is quoted to be 0.4 by (Pinheiro and
conditions using small quantities of feedstock. Extensive Cabral, 1993) while the correction factor for the laboratory scale
publications have shown the utility of the RSD to understand centrifuge is 1. Here, the relationship determined by Maybury et al.
the impact of exposure to various levels of shear as might be (2000) is used to describe the equivalent settling area of a
experienced in the centrifugal step at pilot to manufacturing scale laboratory-scale centrifuge (Slab):
(Boychyn et al., 2001; Hutchinson et al., 2006; Tait et al., 2009).
However, in order to test secondary depth filter capacity, large V lab v2 ð3  2x  2yÞ
Slab ¼ ; ð2Þ
quantity (1 L) of centrate is required. The RSD is limited to 6g ln ð2R2 =ðR2 þ R1 ÞÞ
shearing a maximum of 20 mL of material per run hence
restricting its ability to provide the quantity of feedstock needed v is the angular velocity of the rotor; x and y are the fractional
to characterize subsequent filtration performance. The Capillary acceleration and deceleration times. R1 is the distance from the top
Shear Device (CSD) has also shown to be a preparative device of the liquid level of the centrifuge tube to the center of the
with the ability to mimic feed zone shear. Flow through the laboratory-scale centrifuge’s axis of rotation, while R2 is the distance
capillary enables the generation of energy dissipation rates (EDR) from the bottom of the centrifuge tube to the center of the axis of
equivalent to those found in disk-stack centrifuges (Westoby rotation. The equivalent settling area for a disk-stack centrifuge
et al., 2011). Furthermore, this methodology has shown that it (Sds) is described by:
can generate centrates with a particle size distribution equivalent 2

to that from a large-scale centrifuge (Westoby et al., 2011). In Sds ¼ p z v2 cos u r2 2  r1 2 ; ð3Þ
3g
theory, such a device can be used to produce unlimited quantities
of sheared material. To date no comparison between the where z is the total number of disks, u is the half disk angle, r1 and r2
characteristics of the material prepared by the CSD and the are the inner and outer disk radii, respectively.
established RSD has been published.
In this paper, a combination of scale-down devices was used to
Cell Culture
explore the impact of centrifugal separation on filter capacity so
as to determine the best integration between the steps of Cell culture used in the experiments were generated using CHO cell
centrifugal solids removal and submicron particulate elimination lines expressing monoclonal antibody products. The cultures
by depth filtration ahead of packed bed high resolution steps of produced had a range of cell densities and viabilities measured at
purification. It explores the utility of the CSD to create a pool of the day of harvest as summarized in Table I. Bench-scale cell culture
sheared feed material for centrifugal separation and the creation was conducted in 3 L fed batch bioreactors while pilot-scale culture
of a realistic centrate pool for the subsequent characterization of was performed in 50, 100, and 2000 L stainless steel bioreactors. All
the following depth filtration step. In this sense, it is proposed to cultures are harvested between days 11 and 14 during the decline
use the CSD as a preparatory method for filtration studies greater phase of growth.
than 1 L or more of feed material. Critical to the success of this
approach is the quality of the CSD material. The process
Rotating Shear Device (RSD)
characteristics and properties of the CSD preparatory scale
material are compared against the RSD as a proven method for The design, theory, and application of the rotating shear device to
scale-down studies. mimic levels of shear found in centrifuge feed zones has been

2 Biotechnology and Bioengineering, Vol. 9999, No. xxx, 2016

Table I. Cell culture properties.

Bio reactor size Cell density 106 Cell viability

Material (L) (cells/mL) (%)
Culture-A 100 24.5 68
Culture-B 100 26.8 58
Culture-C 2000 14.9 74
Culture-D 2000 16 70
Culture-E 2000 12.2 45
Culture-F 3 30.8 31
Culture-G 50 11.6 72

thoroughly covered in previous studies (Boychyn et al., 2004;

Hutchinson et al., 2006; Maybury et al., 2000). The material to be
sheared is held in a cylindrical chamber with a diameter of 50 mm
and a height of 10 mm. A rotating disk located within the chamber
creates the shear and has a diameter of 40 mm and a thickness of
1 mm. The correlation of the rotating speed of the disk to the levels
of energy dissipation has been developed in previous studies
(Boychyn et al., 2004). In the experiments conducted for this study,
the rotating shear device was run at a range of speeds in order to Figure 1. Preparative CDS apparatus schematic and parts list.
identify the equivalent shear generated in both the Capillary shear
device (CSD) and the pilot-scale centrifuge. Material was subjected
to shear in the chamber for 20 s to ensure that the total content was Identification of Pilot-Scale Centrifuge Separation
equally exposed to shear generated by the rotating disk (Hutchinson Efficiency
et al., 2006).
Correction factors of 0.4 have been used to describe industrial
disk-stack centrifuges; however, not all centrifuges behave
Capillary Shear Device (CSD) identically. To estimate the correction factor of the pilot-scale
The use of capillary based shear systems to mimic levels of centrifuges examined in this study, material was generated using
energy dissipation experienced in disk-stack centrifugal processes the preparative CSD under conditions which corresponded to the
have been extensively described in literature (Aucamp et al., 2014; level of shear observed in the centrifuge as verified by the RSD data.
Chan et al., 2006; Westoby et al., 2011). The CSD described in this Subsequently, the sheared material was centrifuged in a bench top
publication consisted of a Harvard syringe pump (Cambridge, centrifuge at a range of V/tS. The supernatants from the range of
UK), 0.0100 ID 10 cm long Polyetheretherketone (PEEK) tubing, centrifuged samples were analyzed for the extent of solids
10 mL glass syringe, and check valves as set out in Figure 1. The remaining. This data set was used to identify the best value for
cell culture was delivered from a vessel to the glass syringe via a the effective centrifuge correction factor.
check valve (ID-2). Subsequently, the flow through the capillary
was achieved through another check valve (ID-3) with the syringe Centrifuge Experiments and Protocols
pump in infuse mode. Varying the level of shear was achieved by
adjusting the flow rate through the capillary. This relationship has Floor Swing-Out Rotor Centrifuge
been developed in more detail in previous studies (Ma et al., In order to generate large quantities of centrate required for depth
2002; Mollet et al., 2004, 2007). The culture was kept well mixed filtration experiments, it was necessary to explore the preparative
using a stir plate and stir bar. The sheared material was collected characteristics of the CSD. Immediately following exposure to
in a sealed vessel. capillary shear, the sheared cell culture material was placed into a
series of 50 mL polypropylene tubes and centrifuged at 2600g with
Identification of Pilot-Scale Centrifuge Shear varying liquid height and for varying durations in order to generate
the required range of V/tS. Centrifugation was conducted in a
In order to identify the levels of shear at a range of different flow Beckman J-HC centrifuge with a JS 4.2A rotor (Brea, CA). The
rates through the CSD, the lactate dehydrogenase (LDH) Optical Density (OD) of the supernatant was measured following
concentration of the cell culture was measured before and after each run at a wavelength of 600 nm.
shearing. The relationship between the level of LDH increase as a
function of flow rate (Q) was established and used to calculate the
Pilot-Scale Centrifuge
capillary flow rate that resulted in an equivalent level of cell lysis in
the preparative CSD to that of the standard ultra scale-down RSD. Continuous flow Alfa Laval LAPX-404 and BTPX-305 disk-stack
This provided the necessary bridging study between the RSD, centrifuges (Lund, Sweden) were used to clarify the cell culture at
preparative CSD, and the pilot-scale centrifuge. large-scale. In the experiments conducted with the LAPX-404 flow

Joseph et al.: Harvest Characterisation Using Scale-Down Methodology 3

Biotechnology and Bioengineering
rates were set at 90–120 L/h and the rotational speed of the bowl centrifugation for 30 min at 16,000g (Tait et al., 2009)) ODf refers
was varied to generate between 6000 and 10,000 g while the larger to the clarity of the feed stream prior to depth filtration. All optical
BTPX-305’s flow rate and bowl speed were maintained at 480 L/h densities were measured at 600 nm.
and 12,500 g, respectively.
Characterization of Cell Lysis
Depth and Sterile Filtration Protocols
Lactate dehydrogenase (LDH) is an intracellular enzyme that is only
The centrate clarification experiments were conducted using a released during cell rupture. LDH increase was used to characterize
Millistakþ X0HC depth filter (EMD Millipore, Billerica, MA) with a the extent of cell lysis occurring during the centrifugation process
nominal pore size ranging from 0.1 to 2 mm. Sterilizing grade SHC (Ma et al., 2002; Petersen et al., 1988). It was also used to establish
filters (EMD Millipore) were used to filter the depth filtered comparability between the established scale-down method for
material. The SHC has a bilayer structure with a pore size rating of generating defined levels of shear (RSD) and the preparative CSD
0.5/0.2 mm. method explored in this study for the generation of filter feed
Depth filtration capacity (L/m2) was identified by observing volumes. Following assay protocols provided in the BioVision Kit
the change in filter pressure at a constant flow rate. This approach (Milpitas, CA) samples that had been exposed to shear in either
is known as the Pmax methodology (Yavorsky et al., 2003). In this device were immediately centrifuged at 10,000g for 15 min for the
study, Pmax was defined as the capacity at which a pressure drop separation of cell debris from the sample. The supernatant was
of 10 psi was reached. The experimental runs were conducted removed and combined with Nicotinamide adenine dinucleotide
using a 23 cm2 filter capsule at 200 LMH. Pressure drops and (NADH) and pyruvate solutions. The sample mixture was then
filtrate turbidities (NTU) were recorded for the duration of the aliquoted into a microwell plate and a microplate reader used to
experiment. In cases where the feed pool of centrate had been measure the absorbance change. The LDH increase (LDHINC) was
exhausted and a pressure of 10 psi had not been attained an used calculated by dividing the change in LDH activity in the non-
intermediate pore blockage model was linearized (Eq. 4) to sheared sample (LDHNS) and sheared sample (LDHNS) by that of
predict theoretically the volume of filtrate per filter area (V) the non-sheared sample.
required to reach a pressure (DP) of 10 psi (Hlavacek and
 
Bouchet, 1993). In the linearized equation, a00 and b00 were used LDHSS  LDHNS
as dimensionless coefficients. LDHINC ¼  100% ð7Þ
LDHNS
lnDP ¼ a00 þ b00 ; ð4Þ
Results and Discussion
The capacity of SHC sterile filters required to filter the X0HC
Extensive work has been published on the use of the Rotating shear
filtrate was established through the Vmax methodology (Badming-
device (RSD) to understand the consequence of shear stress placed
ton et al., 1995). Experiments were conducted with a 3.5 cm2 at a
on cell culture materials and thus predicting the impact of similar
constant pressure of 10 psi and the flux decline was monitored over
levels of exposure prevailing in the feed zones of pilot-scale
time. The Vmax method is a linearized form (Eq. 5) of the pore
centrifuges. By the use of such a scale-down device it has been
constriction model used to compute theoretically the maximum
possible to approximate the levels of energy dissipation in both
volume of filtrate per unit filter area (Vmax) that can be obtained
hydro and non-hydro hermetically sealed centrifuges (Boychyn
before complete fouling through plugging occurs (Kong et al., 2010;
et al., 2004). In this study, the need to prepare large quantities of
Lau et al., 2013; Zydney and Ho, 2002).
sheared material so as to satisfy the feed requirements for
  subsequent filtration experiments was met by use of a Capillary
t 1 1
¼ þ t; ð5Þ shear device (CSD). Integral to this was the need to compare the
V Q0 V max
product of this device to that of the RSD to demonstrate
where V is the total volume of filtrate volume per unit filter area comparability in the amount of shear stress generated and of the
collected over time (t) and Q0 is the initial specific volumetric resultant material properties. Figure 2 shows the LDH increase for a
filtrate flow rate per unit filter area. typical cell culture broth sheared using the CSD at flowrates from 12
to 20 mL/min and the RSD at rotational speeds between 6000 and
9000 RPM. A linear increase in LDH release is observed in both
Solids Remaining cases and the data from both systems overlap each other suggesting
The performance of a centrifuge in removing solids at any given set similar levels of energy dissipation can be generated using the two
of operating conditions may be described by calculating the devices. These results encompass the flow rates reported in
percentage of solids remaining in the supernatant (S). literature to mimic the shear in a hydro-hermetically sealed
centrifuge for operation of a CSD (Westoby et al., 2011) and the
 
ODs  ODo range of rotational speeds used in the RSD to achieve the same
S¼  100%; ð6Þ
ODf  ODo mimic (Boychyn et al., 2004).
Having established the comparability of the two devices, the next
where ODs is the filtrate clarity post depth filtration, ODo is the base step was to mimic the process performances of the pilot-scale
line for a well clarified centrate (obtained by extended centrifuges; Alfa Laval LAPX-404 and BTPX-305. The CSD was used

4 Biotechnology and Bioengineering, Vol. 9999, No. xxx, 2016

 relationship between solids remaining (S) and V/tS was
empirically determined (S ¼ 7.1  ln (V/tS)  130.3, R2 ¼ 0.99)
and the correction faction of the Alfa Laval LAPX-404 was found to be
approximately 0.27 while a correction factor of 0.55 was established
to describe the BTPX-305 (modeling data set: S ¼ 2.5  ln
(V/tS)  48.0, R2 ¼ 0.99). These results and other independent
studies using multiple cell culture samples determined the correction
factor of the LAPX-404 to be in the range of 0.27–0.31, confirming
the correction factor values indicated were representative of the
centrifuge’s attributes. The preparative methodology based on the
CSD was then used to mimic a set of Q/cS in the LAPX-404 and
BTPX-305 machines (Table II). The solids remaining in centrates
generated using the CSD closely matched the solids remaining in
centrates generated by the pilot-scale centrifuges (Fig. 4). This
confirmed the validity of the approach to provide quantities of
Figure 2. Comparison of LDH increase for Culture-A (Table I) sourced centrates representative material for process integration studies.
generated using the preparative scale capillary shear device (CSD, &) and rotating
shear device (RSD, ). The ranges tested for the rotational speed and flow rate
In a typical mAb harvest process, the unit operation subsequent
incorporate literature values that have been shown previously to mimic the shear to centrifugation is depth filtration. If an accurate centrate mimic
damage generated in a centrifuge equipped with a hyrdo-hermetic feed-zone. was to be developed, it had to show equivalent filtration properties
to that of the centrate obtained at scale. Figure 5 shows the pressure
and turbidity profiles generated during the operation of the X0HC
as a proxy to approximate the levels of shear to which material is depth filter provided with two feeds: the BTPX-305 centrate and the
exposed inside the centrifuge at 6000, 7900, and 10,000g for the centrate from the preparative CSD approach calibrated against the
LAPX-404 and at 12,500g for the BTPX-305. Data shown in Figure 3 established RSD scale-down tool. The CSD centrate showed pressure
provided a correlation connecting the operating flow rate within the and turbidity profiles that closely matched those of the BTPX-305.
CSD to the conditions of operation in the LAPX-404 centrifuge This indicated that the CSD centrate had similar filtration properties
(Table II) via levels of LDH increase (LDHINC). Regression analysis to the BTPX-305 centrate, confirming utility of the preparative CSD.
of the modeling dataset and matching of the LDH increase showed Having established that the CSD can be used to prepare large
that the levels of shear in the feed zone of the LAPX-404 and BTPX- quantities of sheared material for integrated process studies, the
305 were equivalent to the levels of shear created at a flow rate range impact of the combination of centrifugal and depth filtration
of 16.6–18.0 mL/min in the CSD. This result is consistent with operating conditions on the performance of the primary recovery
earlier published data (Westoby et al., 2011). train was examined. Figure 6 shows the solids remaining and the
The next step of predicting the pilot-scale centrifuges’ process X0HC and SHC capacities for centrates mimicking the LAPX-404
performance was to identify the centrifuge correction factors. The centrifuge at a range of conditions (Table II). Unsurprisingly, the
increase in capacities of both X0HC and SHC are inversely related to
the levels of solids remaining.
When the impact of flux on the X0HC depth filter and subsequent
SHC sterile filtration operations was examined, it was found that
with the given centrate an increase in flux from 50 to 200 LMH had
no significant effect on the capacity of the depth filter (50 LMH:
407 L/m2, 200 LMH: 435 L/m2). The flux decline profiles (Fig. 7) for
the SHC suggested that X0HC filtrates created at the two fluxes
examined (50 and 200 LMH) had similar fouling propensities.
Subsequent studies focused on the impact of cell culture
conditions on the performance of a harvest operation comprising
centrifugation, depth filtration, and sterile filtration. The cell
cultures examined represented differing extremes of harvest
challenge; a difficult to harvest material and a relatively facile to
recover broth. The former was from a bioreactor producing a high
cell density, low viability cell culture whereas the latter was
characterized by relatively low cell density and high cell viability.
It was shown that processing difficult to harvest cultures
Figure 3. LDH increase for the modeling datasets generated by shearing material negatively impacted the ability of the centrifuge to remove solids
(Culture-B(—)) (Table I) using the preparatory CSD to identify the shear in the LAPX- (Fig. 8A). This form of material is challenging for harvest
404 centrate at conditions 1 (&), 2 (*), and 3 (~) (Table II). Regression analysis was
operations as there is a high solid content, leading to the centrifuge
used to determine the following relationship between LDH increase (LDHINC) and
flowrate (Q) for Culture-B: LDHINC ¼ 1.60 Q  14.49, R2 ¼ 0.98. bowl filling up rapidly and requiring a high frequency of discharge.
Another challenge was the low viability of the material. Such broths

Joseph et al.: Harvest Characterisation Using Scale-Down Methodology 5

Biotechnology and Bioengineering
Table II. Operational settings for pilot-scale centrifuge and corresponding CSD settings.

Pilot-scale centrifuge Capillary shear device (CSD)

Condition Centrifuge Flow rate (L/h) Relative centrifugal force (g) Q/cS, V/tS  108 (m/s) CSD flowrate (mL/min)
1 LAPX-404 90 6000 4.9 16.63
2 LAPX-404 105 7900 2.92 17.57
3 LAPX-404 120 10,000 1.65 18.03
4 BTPX-305 480 12,500 2.2 17.08

will contain large populations of cell debris which are small in size particles carried through from the high cell density culture which in
and have a lower density than viable cells. The removal of such cell turn lead to a more rapid rate of pore blockage and flux decline
debris by centrifugation is difficult and these particles often end up (Fig. 8C).
in the centrate resulting in a poor separation. However, the easy to
harvest broth has a lower solid content and a greater population of
Conclusion
large viable cells, which facilitates separation. Furthermore, it may
be expected that high cell viability may lead to healthy cells with This work examines the impact of primary recovery process
intact cell walls and consequently, more resistant to shear damage parameters and cell culture conditions on the performance of
within the centrifuge feed zone. This would be expected to yield a subsequent centrifugal harvest and both depth and sterile filtration
centrate of high clarity and fewer cell debris particles being passed as a process sequence. In order to characterize the process
on to subsequent filtration processes hence requiring much lower sequence, the ability to prepare adequate volumes of sheared
capacities when processing easy to harvest cultures (Fig. 8A). samples was necessary. This study commenced by establishing the
Figure 8B shows the X0HC pressures and turbidity profiles for conditions of operation for a preparative device based on capillary
centrates derived from both culture materials. Upon filtration, the shear to create levels of shear damage similar to the industrially
centrate material from difficult harvest conditions showed accepted standard RSD used routinely for scale-down studies.
significant increases in both the pressure and turbidity profiles Experiments were conducted to determine the flow rates through
(D15.5 psi, D8.7 NTU) compared to that of the easy to harvest the preparative CSD to generate equivalent levels to the shear
material (D6.0 psi, D2.8 NTU). This can be directly attributed to the developed in the standard RSD which has already been shown to
higher levels of smaller debris in the centrate of the high cell density match shear levels found in the feed zone of industrial disk-stack
culture causing a more rapid rate of pore blockage. centrifuges. The centrates generated were then filtered. In the
As noted in Figure 8A, processing of the difficult to harvest process of filtration, equivalent turbidity and pressure profiles were
culture yielded lower SHC capacities compared to the easy to observed to those obtained when filtering centrate generated from a
harvest culture. This was attributed to the larger amount of small large-scale disk stack centrifuge.

Figure 5. Comparison of pressure and turbidity profiles for 0.1–2.0 mm X0HC depth
Figure 4. Comparison of solids remaining in centrate generated using the filter when filtering centrate from the BTPX-305 machine ( ) and the mimic centrate
preparative CSD mimicking LAPX-404 and BTPX-305 at a range of conditions (Table II) (&) generated applying the preparative methodology presented in the paper. BTPX-
marked in and solids remaining from pilot, scale centrifuge runs marked in &. 305 centrate for this study was generated at a relative centrifuge force of 12,500g.
Centrates generated for LAPX-404 were sourced from Culture-B (Table I) while the Preparatory CSD centrate was processed to mimic the large-scale centrifuge. The
BTPX-305 centrate was sourced from Culture-C (Table I). The values plotted are shown material for this experiment was sourced from Culture-C (Table I) and filtered at
as mean  SD (n ¼ 3). 200 LMH.

6 Biotechnology and Bioengineering, Vol. 9999, No. xxx, 2016

Figure 6. Comparison of solids remaining post centrifugation (&), and filter
capacities of 0.1–2.0 mm X0HC depth filter ( ) and 0.2 mm SHC sterile filter ( ). Data
obtained from preparatory CSD centrate mimicking the LAPX-404 at a range of
centrifugal conditions (Table II). All centrates and filtrates were generated from
processing Culture-D (Table I).

The preparative methodology was then applied to explore how

culture conditions impact primary recovery. One feed was from a
culture producing high cell densities and low viabilities. This feed
had a negative effect on the performance of a typical bioprocess
sequence reducing the centrifuge capability to clear solids and also
reducing the capacity of subsequent depth and sterile filtration
processes. Altering the centrifugal operating conditions to produce
centrates with lower solids remaining yielded higher capacities of
both depth and sterile filters.
The proposed preparative methodology enables the user to
calibrate and confirm the operating conditions of the CSD and
bench centrifuge that would allow for predictive small-scale

Figure 8. Comparison of process performance at different stages of the

primary recovery sequence. Feed streams were sourced from two cell cultures
representing difficult to harvest material (Culture-F) and easy to harvest material
(Culture-G). The culture properties for this are listed in Table I. (A) (&) Post
centrifugation mimicking LAPX-404 at a V/tS of 2.92  108 (m/s) (Table II); ( )
capacity of 0.1–2.0 mm depth filter (X0HC) at 200 LMH; ( ) capacity of 0.2 mm
sterile filter (SHC) at 10 psi. (B) Comparison of pressure and turbidity profiles
generated operating 0.1–2.0 mm depth filter (X0HC) processing centrate obtained via
the preparative methodology using Culture-F (&) and Culture-G (4). Centrate
generated to mimic the LAPX-404 operated at V/tS of 2.92  108 (m/s).
(C) Comparison of 0.2 mm sterile filter (SHC) flux decline for Culture-F (&) and
Figure 7. Comparison of 0.2 mm SHC sterile filter flux decline at 10 psi. Feed was Culture-G (4). The filtrate was generated through conditioning both cultures
from a 0.1 to 2.0 mm depth filter (X0HC) operated at 200 LMH (&) and 50 LMH ( ). Both using the preparative methodology to mimic the LAPX-404 at V/tS of 2.92 
filtrates generated from processing Culture-E (Table I) using the preparatory CSD 108 (m/s) and processing the centrate through a 0.1–2.0 mm depth filter (X0HC) at
mimicking LAPX-404 at a V/tS of 2.92  108 (m/s) (Table II). 200 LMH.

Joseph et al.: Harvest Characterisation Using Scale-Down Methodology 7

Biotechnology and Bioengineering
centrifuge and polishing filter performance. It was demonstrated Boychyn M, Yim S, Bulmer M, More J, Bracewell D, Hoare M. 2004. Performance
that the scale-down methodology can successfully predict large- prediction of industrial centrifuges using scale-down models. Bioprocess
Biosyst Eng 26(6):385–391.
scale performance. This adds to the toolbox of available small-scale
Boychyn M, Yim S, Shamlou PA, Bulmer M, More J, Hoare M. 2001. Characterization
models for high throughput characterization of mammalian cell of flow intensity in continuous centrifuges for the development of laboratory
culture processes. Future work should leverage the methodology mimics. Chem Eng Sci 56(16):4759–4770.
developed in this publication to incorporate a wider range of cell Chan G, Booth A, Mannweiler K, Hoare M. 2006. Ultra scale-down studies of the
culture broths and other models of disk-stack centrifuges. effect of flow and impact conditions during E. coli cell processing. Biotechnol
Bioeng 95(4):671–683.
Hlavacek M, Bouchet F. 1993. Constant flowrate blocking laws and an example of
Nomenclature their application to dead-end microfiltration of protein solutions. J Membr Sci
82(3):285–295.
a00 , b00 dimensionless coefficients for the linearized intermedi- Hutchinson N, Bingham N, Murrell N, Farid S, Hoare M. 2006. Shear stress analysis
ate pore blockage model of mammalian cell suspensions for prediction of industrial centrifugation and
its verification. Biotechnol Bioeng 95(3):483–491.
c correction factor to allow for flow conditions not Jain M, Paranandi M, Roush D, G€oklen K, Kelly WJ. 2005. Using cfd to understand
considered in the theoretical derivation of S for large- how flow patterns affect retention of cell-sized particles in a tubular bowl
scale centrifuges centrifuge. Ind Eng Chem Res 44(20):7876–7884.
LDHNS absorbance of no shear sample at 450 nm Kempken R, Preissmann A, Berthold W. 1995. Assessment of a disc stack centrifuge
for use in mammalian cell separation. Biotechnol Bioeng 46(2):132–138.
LDHSS absorbance of sheared sample at 450 nm
Kong S, Aucamp J, Titchener-Hooker NJ. 2010. Studies on membrane sterile filtration
ODf optical density of feed at 600 nm of plasmid dna using an automated multiwell technique. J Membr Sci 353(1–
ODo optical density of well spun sample at 600 nm 2):144–150.
ODs optical density of supernatant at 600 nm Lau EC, Kong S, McNulty S, Entwisle C, Mcilgorm A, Dalton KA, Hoare M. 2013. An
Pmax predicted capacity at 10 psi for constant flow operations ultra scale-down characterization of low shear stress primary recovery stages to
enhance selectivity of fusion protein recovery from its molecular variants.
(L/m2)
Biotechnol Bioeng 110(7):1973–1983.
Q0 initial volumetric filtrate flow rate (L/h) Ma N, Koelling KW, Chalmers JJ. 2002. Fabrication and use of a transient
Qds flow rate of pilot scale disk-stack centrifuge (L/h) contractional flow device to quantify the sensitivity of mammalian and insect
r1 inner radius of centrifuge discs (m) cells to hydrodynamic forces. Biotechnol Bioeng 80(4):428–437.
r2 outer radius of centrifuge discs (m) Maybury JP, Hoare M, Dunnill P. 2000. The use of laboratory centrifugation studies to
R1 radius to sample surface in an individual tube (m) predict performance of industrial machines: Studies of shear-insensitive and
shear-sensitive materials. Biotechnol Bioeng 67(3):265–273.
R2 radius to sample base in an individual (m) Mollet M, Godoy-Silva R, Berdugo C, Chalmers JJ. 2007. Acute hydrodynamic forces
S solids remaining in processed samples (%) and apoptosis: a complex question. Biotechnol Bioeng 98(4):772–788.
t filtration time (h) Mollet M, Ma N, Zhao Y, Brodkey R, Taticek R, Chalmers JJ. 2004. Bioprocess
tlab laboratory centrifugation time (h) equipment: Characterization of energy dissipation rate and its potential to
damage cells. Biotechnol Prog 20(5):1437–1448.
V total filtrate volume per filter area (L/m2)
Mosqueira FG, Higgins JJ, Dunnill P, Lilly MD. 1981. Characteristics of mechanically
Vlab volume of suspension in individual tubes (L) disrupted bakers’ yeast in relation to its separation in industrial centrifuges.
Vmax predicted maximum volume filtered per unit area for Biotechnol Bioeng 23(2):335–343.
constant pressure operations (L/m2) Petersen JF, McIntire LV, Papoutsakis E. 1988. Shear sensitivity of cultured
x fraction of overall centrifugation time for deceleration hybridoma cells (crl-8018) depends on mode of growth, culture age and
metabolite concentration. J Biotechnol 7(3):229–246.
phase
Pinheiro H, Cabral JMS. 1993. Centrifugation. In: Kennedy JF, Cabral JMS, editors.
y fraction of overall centrifugation time for acceleration Recovery processes for biological materials. Chichester: John Wiley & Sons Ltd.
phase p 145.
zP number of discs in disc-stack centrifuge Rathore AS. 2009. Roadmap for implementation of quality by design (qbd) for

Pds equivalent disk stack centrifuge settling area (m2) biotechnology products. Trends Biotechnol 27(9):546–553.
equivalent laboratory centrifuge settling area (m2) Shukla AA, Kandula JR. 2008. Harvest and recovery of monoclonal antibodies: Cell
lab
removal and clarification. In: Gottschalk U, editor. Process scale purification of
v angular velocity (s1) antibodies. Hoboken, New Jersey: John Wiley & Sons, Inc. p 53–78.
Tait A, Aucamp J, Bugeon A, Hoare M. 2009. Ultra scale-down prediction using
microwell technology of the industrial scale clarification characteristics by
References centrifugation of mammalian cell broths. Biotechnol Bioeng 104(2):321–331.
Titchener-Hooker N, Dunnill P, Hoare M. 2008. Micro biochemical engineering to
Ambler CM. 1959. The theory of scaling up laboratory data for the sedimentation accelerate the design of industrial-scale downstream processes for biopharma-
type centrifuge. J Biochem Microbiol Technol Eng 1(2):185–205. ceutical proteins. Biotechnol Bioeng 100(3):473–487.
Aucamp JP, Davies R, Hallet D, Weiss A, Titchener-Hooker NJ. 2014. Integration of Westoby M, Rogers JK, Haverstock R, Romero J, Pieracci J. 2011. Modeling industrial
host strain bioengineering and bioprocess development using ultra-scale down centrifugation of mammalian cell culture using a capillary based scale-down
studies to select the optimum combination: An antibody fragment primary system. Biotechnol Bioeng 108(5):989–998.
recovery case study. Biotechnol Bioeng 111(10):1971–1981. Yavorsky D, Blanck R, Lambalot C, Brunkow R. 2003. The clarification of bioreactor
Axelsson H. 2002. Cell separation, centrifugation. In: Flickinger MC, Drew cell cultures for biopharmaceuticals. Pharm Technol 27(3):62–67.
SW, editors. Encyclopedia of bioprocess technology. New York: Wiley. Yigzaw Y, Piper R, Tran M, Shukla AA. 2006. Exploitation of the adsorptive properties
p 513–531. of depth filters for host cell protein removal during monoclonal antibody
Badmington F, Wilkins R, Payne M, Honig ES. 1995. Vmax testing for practical purification. Biotechnol Prog 22(1):288–296.
microfiltration train scale-up in biopharmaceutical processing. Pharm Technol Zydney AL, Ho CC. 2002. Scale-up of microfiltration systems: Fouling phenomena
19(9):64–76. and Vmax analysis. Desalination 146(1–3):75–81.

8 Biotechnology and Bioengineering, Vol. 9999, No. xxx, 2016