ARTICLE

Caprylic Acid Precipitation Method for

Impurity Reduction: An Alternative to
Conventional Chromatography for
Monoclonal Antibody Purification
Yan Brodsky, Cheng Zhang, Yinges Yigzaw, Ganesh Vedantham
Purification Process Development, Amgen Inc., 1201 Amgen Court West, Seattle,
Washington 98119; telephone: þ1-206-265-7443; fax: (206) 217-0491;
e-mail: brodskyy@amgen.com

ABSTRACT: We report the use of caprylic acid based impu-
rity precipitation as (1) an alternative method to polishing
chromatography techniques commonly used for monoclo-
nal antibody purification and (2) an impurity reduction step
prior to harvesting the bioreactor. This impurity reduction
method was tested with protein A purified antibodies and
with cell culture fluid. First, the operational parameters
influencing precipitation of host cell proteins and high
molecular weight aggregate in protein A pools were investi-
gated. When used as a polishing step, the primary factor
affecting purification and yield was determined to be pH.
Caprylic acid precipitation was comparable to polishing IEX
chromatography in reducing host cell protein and aggregate
levels. A virus reduction study showed complete clearance of
a model retrovirus during caprylic acid precipitation of
protein A purified antibody. Caprylic acid mediated impu-
rity precipitation in cell culture showed that the impurity
clearance was generally insensitive to pH and caprylic acid
concentration whereas yield was a function of caprylic acid
concentration. Protein A purification of caprylic acid pre-
cipitated cell culture fluid generated less turbid product pool
with reduced levels of host cell proteins and high molecular
weight aggregate. The results of this study show caprylic acid
precipitation to be an effective purification method that can
be incorporated into a production facility with minimal cost
as it utilizes existing tanks and process flow. Eliminating
flow through chromatography polishing step can provide
process intensification by avoiding the process tank volume
constraints for high titer processes.
Biotechnol. Bioeng. 2012;109: 2589–2598.
ß 2012 Wiley Periodicals, Inc.
KEYWORDS: caprylic acid; purification; impurity precipi-
tation; mAb process development
Cheng Zhang’s present address is Genzyme Corporation, Framingham, MA 01701.
Correspondence to: Y. Brodsky
Received 13 January 2012; Revision received 3 April 2012; Accepted 20 April 2012
Accepted manuscript online 30 April 2012;
Article first published online 16 May 2012 in Wiley Online Library
(http://onlinelibrary.wiley.com/doi/10.1002/bit.24539/abstract)
DOI 10.1002/bit.24539
Introduction
A number of monoclonal antibodies (mAbs) have become
established therapeutic products with more molecules
of this class being developed and tested in clinical trials
(Kelley, 2009). Production scale for clinically successful
antibodies can reach several hundred kilograms of drug
substance per year (Aldington and Bonnerjea, 2007; Kelley,
2009). Consequently, the production of mAbs requires
considerable investment in process development and
manufacturing capability. At such scale, productivity of the
manufacturing process becomes increasingly important for
minimizing the cost of goods. Until recently, the productivity
of the mammalian cell culture titers was recognized as
limiting the mass throughput of mAb manufacturing
processes in existing facilities (Birch and Onakunle, 2005).
However, recent advances in cell culture technology and
growing demand for mAbs are increasingly focusing on the
limitations in purifying large quantities of product in a single
batch (Kelley, 2009; Thommes and Etzel, 2007).
Though purification process technology continues to
adapt to meet new challenges, chromatographic separations
still remain the workhorse of mAb purification (Kelley,
2007; Przybycien et al., 2004). A conventional purification
process includes a protein A capture column followed by
two polishing chromatography steps (Shukla et al., 2007).
Occasionally, an anion exchange flow-through column or a
bind-and-elute cation exchange is used as a sole polishing
step (Kelley et al., 2008; Moscariello, 2008). Typically, the
polishing steps consist of a combination of ion exchange and
hydrophobic interaction chromatography. The key chal-
lenges in existing manufacturing facilities, imposed by
higher pool volumes due to improved productivity of
high titer cell culture processes, are (1) turning around
purification unit operation within the bioreactor turn-
around time and (2) meeting the in-process pool tank
constraints (Gottschalk, 2008). Increasing column loading

ß 2012 Wiley Periodicals, Inc. Biotechnology and Bioengineering, Vol. 109, No. 10, October, 2012 2589
does provide some relief but use of larger columns has its
limitations. With higher column loading, viscosity at
peak protein concentrations can limit the operating
flow rate (due to pressure limitations) and extend the
processing times. Furthermore, introducing columns
larger than 2.0 m in diameter, a typical upper limit for
industrial scale, is expensive and often not supported by
current manufacturing facilities (Gottschalk, 2008).
Alternatively, increased cycles can be performed on
the columns until processing times, buffer and process
intermediate volumes become rate limiting. Limitations in
scale up of conventional chromatography operations bring
about a search for alternative technologies for mAb
purification (Low et al., 2007). In addition to high
throughput, a new purification method needs to provide
separation power equivalent to a chromatography step, fit
within the current design of manufacturing facilities, and be
at a competitive cost.
In this article, caprylic acid precipitation of impurities has
been evaluated as a possible alternative to conventional
polishing chromatography. Also called octanoic acid,
caprylic acid is an eight-carbon, saturated fatty acid. Its
molecular formula is CH3(CH2)6COOH. This method
involves selective precipitation of impurities while main-
taining the product in solution. Plasma product manu-
facturers have effectively precipitated non-immunoglobulin
proteins from serum, plasma, and ascites fluid by using
caprylic acid (McKinney and Parkinson, 1987; Parkkinen
et al., 2006; Russo et al., 1983; Steinbuch and Audran, 1969).
It has also been reported to be an effective agent for
inactivating enveloped viruses (Johnston et al., 2003;
Korneyeva et al., 2002; Lundblad and Seng, 1991).
Caprylic acid precipitation can be implemented in a
bioreactor, prior to cell separation, or post Protein A
capture step. In both cases, the purpose is to eliminate a
polishing chromatography step and/or to increase overall
process robustness.
The objective of this study was to identify operational
conditions for caprylic acid precipitation of impurities
in mAb process as a means for providing alternatives to
the conventional polishing chromatography step. Two
possible process scenarios for integration of caprylic
acid precipitation step were investigated. Operational
conditions including pH, caprylic acid concentration,
reaction time, and buffer components were investigated
with respect to yield and purification factors, removal
of host cell proteins (HCP) and high molecular weight
(HMW) aggregates (includes higher order aggregates
and dimers), for six protein A purified IgG1 and IgG2
molecules. Process robustness was evaluated with three
mAbs using a design of experiment (DoE) approach. A
study was also performed to evaluate clearance of
Xenotropic Murine Leukemia Virus (xMuLV) with caprylic
acid precipitation of protein A purified product intermedi-
ate. Finally, precipitation conditions were tested with cell
culture fluid, prior to cell separation, for three different
mAbs.
2590 Biotechnology and Bioengineering, Vol. 109, No. 10, October, 2012
Materials and Methods
Materials
The mAbs used in this study (Table I), were expressed in
CHO cells, and produced at Amgen (Seattle, WA).
MAbSelect SuRe1 Protein A media was obtained from
GE Healthcare (Uppsala, Sweden). The media was packed in
Vantage columns (i.d. 1.1 cm) obtained from Millipore
Corporation (Bedford, MA). 4–25% Tris–glycine gel,
SDS 2
sample buffer, SDS 10
running buffer, and
standard marker were purchased from Invitrogen (Carlsbad,
CA). Silver staining kit was obtained from GE Healthcare.
Slide-A-Lyzer dialysis cassettes (3,500 MWCO) were
purchased from Pierce (Rockford, IL). TRIS–base, TRIS–
HCl, sodium chloride, sodium citrate, sodium sulfate,
citric acid, sodium acetate, acetic acid, sodium phosphate
dibasic, and phosphoric acid were purchased from JT Baker
(Phillipsburg, NJ). Caprylic acid (Octanoic acid 99%,
C2875) was purchased from Sigma (St. Louis, MO).
Equipment
Protein A chromatography was carried out on an AKTA
Explorer chromatographic system from GE Healthcare with
built-in UV, pH and conductivity detectors to monitor
column effluent. HPLC analysis was carried out using a
Waters (Milford, MA) HPLC 2690 Separation module,
coupled with Waters 2487 Dual l absorbance detector.
Absorbance of protein samples was measured using 8453
UV-Visible spectrophotometer with 10 mm path length flow
cell from Agilent (Santa Clara, CA). Centrifugation of the
cell culture supernatant was performed on the Allegra X-15R
bench top centrifuge (Beckman Coulter, Fullerton, CA).
Turbidity of protein solutions was measured with 2100P
Turbidimeter (Hach, Loveland, CO).
Methods
Protein A Chromatography
Protein A chromatography was carried out on a 1.1 cm
D
10 cm H column. The column was equilibrated with
25 mM tris, 75 mM sodium sulfate, pH 7.4 buffer, loaded to
Table I. A list of mAbs used in this study along with their IgG subtype and
isoelectric point (pI).
mAb Subtype pI
A IgG 1 9.1
B IgG 1 8.9
C IgG 2 8.0
D IgG 2 6.9
E IgG 2 8.5
F IgG 2 8.8
G IgG 2 7.0
H IgG 1 9.0
35 g/L of resin, washed with equilibration buffer, and
eluted with 50 mM acetate, pH 3.6. Following elution, the
column was stripped with 100 mM phosphoric acid and
regenerated with 0.3 M sodium hydroxide. The eluate was
collected, neutralized with 2 M tris base and analyzed with
HCP ELISA and analytical SEC assays. Turbidity and protein
concentration were determined following neutralization of
the elution pool.
Caprylic Acid Precipitation
Precipitation of protein A purified mAb. Samples of protein
A purified mAbs were dialyzed into 50 mM acetate or
50 mM citrate at different pHs between 4.5 and 5.5. Two
percent caprylic acid emulsion was prepared at the same pHs
by dispersing neat caprylic acid in either 50 mM acetate or
50 mM citrate buffers using a vortex followed by a pH
adjustment with 10 N NaOH. Precipitation was initiated by
adding 10 mL of caprylic acid emulsion to 10 mL of dialyzed
mAb solution at the same pH. Unless stated otherwise,
starting caprylic acid concentration in the mixtures was 1%
(v/v; 60 mM). The mixtures were stirred with a magnetic bar
at 170 rpm for 1 h at ambient temperature. Precipitate was
removed by filtration through a 0.2 mm filter and discarded.
The optimal pH conditions were chosen based on IgG yield
and product quality.
Precipitation of cell culture fluid. Cell culture fluid (with
cells) was titrated to pH 4.9 with 1 M phosphoric acid.
A 20% caprylic acid emulsion was prepared by dispersing
neat caprylic acid in Milli-Q water by using a vortex.
Precipitation was initiated by addition of 20% caprylic
acid emulsion to titrated cell culture fluid. Unless stated
otherwise, starting caprylic acid concentration and pH in the
precipitation mixture was 1% caprylic acid at pH 4.7. The
mixture was stirred with a magnetic bar at 170 rpm for
1 h at ambient temperature. Precipitate was removed by
centrifugation at 2,000g for 10 min, followed by filtration
through a 0.2 mm filter. The supernatant was then titrated
to pH 7.0 with 2 M tris base and protein A purified.
Virus Clearance Study
The study was conducted at the Biosafety Development
laboratory at Amgen (Seattle, WA). Xenotropic Murine
Leukemia Virus stock was prepared by culturing chronically
infected mink lung cells (ATCC CCL-64). The virus was
concentrated by centrifugation (Romanowski et al., 2008).
The titer of the virus containing solutions was determined
using a standard 50% Tissue Culture Infectious Dose
(TCID50) end-point titration assay with the indicator cell
line PG4 (feline astrocyte, ATCC CRL-2032). The titer of the
virus was calculated using a modified Spearman–Karber
method and is expressed as a Log10 TCID50/mL value.
Purified mAb samples were prepared at 5, 15, and 25 g/L
in 100 mM Acetate, pH 4.3. Combination of 1 M sodium
caprylate and 2 M tris base were used to titrate samples to
desired experimental pH (4.9, 5.1, or 5.3) and final caprylate
Brodsky et al.: Monoclonal Antibody Purification Using Caprylic Acid
concentration of 0.5% or 1.0% (v/v). In the negative control,
the samples were titrated with 2 M tris base only. Following
the titration, virus stock was added to the samples to the
concentration of 5% (v/v). Solutions were gently stirred at
ambient temperature with time samples taken at 10, 30, 60,
and 120 min. At each time point samples were filtered
with a 0.45 mm filter, neutralized to pH 6–8 with 2 M tris
base and assayed for infectious virus using the TCID50
method. Samples were also frozen to be analyzed by reverse
transcriptase quantitative polymerase chain reaction (RT-
qPCR) at a later date.
Reverse transcriptase quantitative polymerase chain reaction
(RT-qPCR) assay. Total RNA was extracted from each
sample using the QIAamp Viral RNA kit (QIAgen Inc.,
Chatsworth, CA) and transcribed into cDNA using a high
capacity reverse transcription kit (Applied Biosystems, Foster
City, CA) according to the manufacturers’ protocols. Each
25 mL qPCR reaction contained 5 mL of cDNA, two primers
which amplify a 76 base pair fragment of the xMuLV env
gene, and a probe as described previously (Shi et al., 2004).
The probe was labeled with a fluorescent reporter dye FAM
(6-carboxyfluorescein) at the 50 end and a quencher dye
TAMRA (6-carboxytetramethylrhodamine) at the 30 end.
qPCR reactions were performed and analyzed on an ABI
Prism 7900 Sequence Detection System (Applied Biosystems,
Foster City, CA). A 6 log dilution range of the standard was
assayed along with the samples. The amplification curve of
the xMuLV standard’s dilution range was used as standard
curve to determine the xMuLV particle number in each
sample. The assay variability is approximately 0.5 logs.
Analysis
Host cell protein analysis. Host cell protein (Chinese
hamster ovary cell protein, or HCP) levels were analyzed
by an enzyme-linked immunosorbent (ELISA) assay
developed at Amgen.
Size exclusion chromatography. HMW (high molecular
weight) was quantified by analytical SEC with a G3000
SWXL column (4.6 mm D
300 mm L) from Tosoh
Biosciences (Montgomeryville, PA).
Antibody concentration determination. The mAb concentra-
tion in harvested cell culture fluid was determined with a
POROS protein A column (2.1 mm D
30 mm L) from
Applied Biosystems, while antibody concentrations in
protein A purified samples were determined by absorbance
at 280 nm using 8453 UV-Visible spectrophotometer with
10 mm path length flow cell from Agilent.
Quantitation of residual caprylic acid. The residual caprylic
acid in the supernatants or eluate pools was quantified
by RP-HPLC with a Waters Symmetry1 C18 column
(150 mm
4.6 mm, 3.5 mm, WAT 200632). The running
protocol, a linear gradient of 30–80% acetonitrile in 0.1%
phosphoric acid over 25 min, is a modified Fliszar method
(Fliszar et al., 2006). Injection volume was 20 mL and the
flow rate was 1 mL/min. The absorbance was monitored at
220 nm.
2591
Biotechnology and Bioengineering
Results
105%
Caprylic Acid Precipitation of Protein A Purified mAb

Introduction of caprylic acid precipitation step into mAb 100%

purification process requires detailed knowledge of the

operating space in order to understand the effects of input

Yield
95%
process parameters. Of the mAbs outlined in Table I, six
were selected to test precipitation conditions, including two
90%
IgG1 mAbs (A and B) and four IgG2 mAbs (C, D, E, and F).
It was previously reported that precipitation of non-
immunoglobulin proteins is best obtained at a slightly 85%
4.4 4.6 4.8 5.0 5.2 5.4 5.6
acidic pH (McKinney and Parkinson, 1987; Russo et al., pH
1983). Preliminary experiments showed that precipitation
2000
pH is a key factor for consistent purification performance. In
this study, pH < 4.5 was not effective in precipitating HCP
while pH > 5.5 was found to have adverse effect on product 1600

yield. Secondary variables such as caprylic acid concentra-

Host Cell Protein (ppm)

tion, reaction time, and mAb concentration were also 1200

established. Protein A elution pool in 50 mM sodium acetate

was used to study process variables. Caprylic acid 800

concentrations of 1–3% (w/v) were cited in the literature

for purification of IgG from serum and ascites fluid
(McKinney and Parkinson, 1987; Russo et al., 1983). An
experiment investigating the effects of using caprylic acid at
1–4% concentrations in one percent increments was carried
out with mAb E to determine the optimum concentration
needed to achieve the highest product purity and yield. At all
four caprylic acid concentrations comparable clearance of
HCP (from 1,179 to 29–41 ppm) and HMW aggregate (from
1.9% to 1.0–1.2%) was achieved. Product yield was between
99% and 102% for all tested concentrations of caprylic acid.
Based on this data, 1% caprylic acid was used for future
experiments. Reaction kinetics measurement from prelimi-
nary experiments suggested that precipitation approached
completion in 15 min, which was consistent with previously
published results (McKinney and Parkinson, 1987). The
initial concentration of mAbs, prior to caprylic acid
addition, was kept in a range consistent with protein A
elution pool concentration of 10–16 g/L.
To evaluate purification capability of a precipitation step
and to establish operational pH range for the process, a
broad screening experiment was designed to study pH range
of 4.5–5.4 at a fixed caprylic acid concentration of 1% and a
set reaction time of 60 min. Impact of pH on HCP clearance
and precipitation yield on all six mAbs is depicted in
Figure 1. Optimal precipitation pH for all antibodies fell
400
0
4.4 4.6 4.8 5.0 5.2 5.4 5.6
pH
Figure 1. Effect of pH on HCP impurity reduction and step yield - broad
screening study with six different protein A purified mAbs. Legends: mAb A (^),
mAb B (*), mAb C (), mAb D (~), mAB F (&), and mAb H (&).
within a pH range of 5.0–5.4. For mAbs A, C, and F the
operating pH was confined to a smaller range due to
incomplete HCP clearance at pH  5.1 and possibility of
>10% product loss at pH  5.3. On the other hand, mAbs B,
D, and H demonstrated comparable impurity clearance and
yield across a wide pH range of 4.5–5.4. Table IIa shows HCP
concentration in mAbs A–F prior to caprylic acid addition.
Product loss observed at higher pH conditions deter-
mined the upper limit for step operation. Minor variability
was noted in the pH at which mAbs started to precipitate
(pH 5.3–5.6). As suggested in the literature, it is the
non-ionized form of a fatty acid that binds proteins, which
in turn leads to the conformational changes (Bull and
Breese, 1967). Extensive binding of CA could result in

Table IIa. Effect of caprylic acid on protein A pool impurity reduction and yield for six mAbs.

mAb Opt pH Feed HCP (ppm) Pool HCP (ppm) HCP removal (%) Feed HMW (%) Pool HMW (%) HMW removal (%) Yield (%)
A 5.4 1,100 71 94 1.1 0.8 27 86
B 5.2 29,000 120 100 2.4 1.3 46 99
C 5.1 1,676 106 94 3.3 1.1 67 90
D 5.3 672 22 97 2.9 2.1 28 96
E 5.0 1,850 85 95 1.7 1.5 12 92
F 5.3 2,632 181 93 2.2 1.0 55 100

2592 Biotechnology and Bioengineering, Vol. 109, No. 10, October, 2012
protein-protein association and subsequent protein precip-
itation. At lower pH non-ionized form becomes dominant
(Caprylic acid pKa ¼ 4.89) but at the same time leads to a
decrease in the solubility of CA. Consequently, concentra-
tion of CA available at lower pH is below the critical
concentration required for extensive CA binding and
precipitation of mAbs. Although the pH dependent binding
of fatty acids to protein has been linked to protein
hydrophobicity (Bull and Breese, 1967), we were unable
to find a correlation between the starting pH and mAb
properties such as hydrophobicity (8-anilino-1-naphthalene
sulfonate binding), isoelectric point, and melting tempera-
ture in this study. A relatively narrow pH window at which
mAbs begin to precipitate suggests that critical CA
concentration is similar between the antibodies. This is
not surprising given the highly homologous structure of
mAbs. Conversely, HCPs generally have lower pI values than
an antibody and hence could be precipitated with a lower
critical CA concentration and at lower pH. Therefore, pH
constraint on the lower side of the range was driven by
variability in precipitation of contaminants. Results from
the screening experiment showed considerable differences
in pH at which HCP starts to precipitate for six mAbs tested
(pH 4.5–5.1). Previous work (Shukla and Hinckley, 2008)
had suggested that the HCPs were mostly associated with
mAbs, which was the likely reason for the difference in pH at
which the complex starts to precipitate. Differences in host
cell protein species due to varied production cell lines for the
mAbs were also a likely contributing factor.
Table IIa shows the reduction in HCP achieved with six
mAbs at the optimized pH. For all cases, levels of HCP were
reduced below 200 ppm, regardless of the starting concen-
tration. These results are at least as good as the clearance
achieved by mAb specific IEX chromatography platform
purification process (Table IIb). Besides considerable
clearance of HCP, caprylic acid precipitation was effective
in reducing aggregate levels for all six antibodies (Table IIa).
Extent of HMW aggregate precipitation varied from
molecule to molecule; reduction was observed in dimer
as well as higher order multimers. Overall HMW aggregates
levels achieved with caprylic acid precipitation step were
comparable to the clearance obtained with the polishing
chromatography step (Table IIb). Higher order aggregates
were more susceptible to CA precipitation compared to
dimers, likely due to its lower solubility. In all the cases,
product yield at the optimal pH condition remained
above 90% with the exception of mAb A that was at 86%
Table IIb. Impurity reduction and yield for six mAbs obtained on mAb specific IEX chromatography platform purification process.
mAb Polishing step Feed HCP (ppm) Pool HCP (ppm) HCP removal (%) Feed HMW (%) Pool HMW (%) HMW removal (%) Yield (%)
A AEX 811 87
B CEX 34,000 5,000
C CEX 1,600 150
D CEX 529 39
E CEX 2,559 92
F CEX 2,800 325
(Table IIa). Impurity clearance and product yield was
comparable for IgG1 and IgG2 molecules. As stated earlier,
the data discussed previously was investigated in 50 mM
sodium acetate buffer conditions. The impact of process
variables on caprylic acid precipitation of impurities was the
same in another commonly used buffer system for protein A
purification, 50 mM sodium citrate, as compared to 50 mM
sodium acetate with selected molecules. No differences in
impurity reduction or yield were observed between the two
buffer systems (data not shown).
A modified Fliszar method (Fliszar et al., 2006) was used
to measure residual caprylic acid in the supernatant protein
solution. Levels of caprylic acid were found to be highly
dependent on the precipitation pH regardless of mAb
concentration or the starting caprylic acid concentration
(investigated for two mAbs). The data presented in Table III
show residual caprylic acid is about 0.2% at pH 5.0 and
about 0.3% when precipitation is done at pH 5.2. This is
consistent with the solubility profile of caprylic acid in an
aqueous solution. Moreover, CEX purification of precipi-
tated mAb solution reduces residual caprylic acid to
undetectable levels (LOD ¼ 0.025%). Historically, sodium
caprylate (salt form of CA) has been used in preparation of
therapeutic human serum albumin for over 50 years
(Korneyeva et al., 2002). From a safety standpoint, the
intravenous administration of large volumes of albumin
formulated in 4 mM (0.06%) caprylic acid over an
extended period of time shows low risk for implementation
of caprylic acid containing step in the manufacturing
process.
Robustness Study
Design of experiment (DoE) approach was used to set up a
2
2 full factorial experiment with two center points to
investigate the interaction effect between pH and concen-
tration of caprylic acid. Three mAbs were selected for this
study, mAb B (IgG1) and mAbs C and D (IgG2). Conditions
chosen for this experiment were intended to mimic a
representative range for a scale-up process. Caprylic acid
concentrations studied were 0.5% and 1.5% with a center
point at 1.0%. Based on previous screening experiments, for
mAbs B and D, pH points were set at pH 4.9 and pH 5.3 with
a center point at pH 5.1. For mAb C, pH 5.0 and pH 5.3 with
a center point at pH 5.15 were chosen based on a narrow
operating window suggested by the screening experiment.
89 1.3 1.2 8 99
85 2.2 1.3 41 93
91 4.5 3.5 22 90
93 1.8 1.5 17 96
96 1.3 1.1 15 94
88 3.2 1.4 56 90
Brodsky et al.: Monoclonal Antibody Purification Using Caprylic Acid 2593
Biotechnology and Bioengineering
Table III. Residual caprylic acid (CA) levels in filtrates of mAb C at yield conditions are satisfied. The caprylic acid precipitation
different precipitation pH and mAb concentration. step proved to be robust for a wide range of operating
parameters with mAbs B and D. However, for mAb C, the
Precipitation pH mAb (g/L) Feed (CA) (%) Pool (CA) (%)
applicability may be limited due to the narrow operating
5.0 12 1.0 0.22 window.
4 1.0 0.21
5.2 6 0.5 0.30
1.0 0.29
2.0 0.30 Virus Clearance
Caprylate treatment was previously shown to be an effective
method for inactivation of enveloped viruses in plasma
derived intermediates (Johnston et al., 2003; Korneyeva
HCP and HMW aggregate levels as well as the product yields
et al., 2002; Lundblad and Seng, 1991). A study was
from these experiments are shown in Table IV. The data
conducted to evaluate the operating space for the viral
does not reveal any statistically significant interaction effects
clearance of xMuLV with a caprylate precipitation step. Two
between caprylic acid concentration and pH. Caprylic acid
purified mAbs were selected for this study, mAb B and mAb
concentration was statistically significant only for HMW
F. Impact of pH on viral clearance was evaluated with mAb B
aggregate clearance with mAb C but even the worst
at fixed caprylate concentration of 1%. Similar to the
conditions resulted in over two fold reduction in aggregate.
robustness study, pH points were set at 4.9, 5.1, and 5.3.
In all other cases, caprylic acid concentration had an
Caprylate concentration of 0.5% was also tested with mAb B
insignificant effect indicating that within a range of 0.5–
at pH 5.1. Concentration of mAb B for all conditions was
1.5% it hardly influences performance parameters: yield,
kept constant at 15 g/L. For mAb F, the pH and caprylate
HCP and HMW aggregate clearance. Even though pH had
concentration were fixed at pH 5.1 and 1%, respectively,
statistically significant effects in all three cases, for mAbs B
while mAb F concentration was varied between 5, 15, and
and D, the variation in performance parameters were
25 g/L. Samples tested for viral infectivity were previously
practically small. For both of these mAbs, the impurity
filtered with a 0.45 mm filter to remove potential precipitate
precipitation method can be operated at pH 5.1  0.2 with
that could interfere with the assay. In order to evaluate
1.0  0.5% caprylic acid to achieve HCP and HMW
possible virus reduction on the 0.45 mm filter, for conditions
aggregate clearance at least equivalent to that of a polishing
tested with mAb B, pre- and post-filter samples were assayed
chromatography step with yields >90% (Table IIb). Results
by RT-qPCR. Infectivity assay results (Table V) show
for mAb C show strong dependence on pH of precipitation.
complete clearance of xMuLV for all conditions tested
While at pH 5.0 there is insufficient HCP clearance
starting with the first time point at t ¼ 10 min. However,
(>1,000 ppm), at pH 5.3, the product loss is on the order
RT-qPCR results suggest that the virus was precipitated by
of 20%. This data show that mAb C has a narrow operating
caprylate and consequently removed by the 0.45 mm
pH window (<0.2 pH units) where both product purity and
filtration. Also, RT-qPCR results show a trend of increasing
xMuLV reduction at higher pH. In contrast, varying
caprylate or mAb concentrations did not affect virus
Table IV. DoE experiment exploring the effect of pH and caprylic acid
concentration (CA) on HCP reduction, HMW reduction, and yield. The clearance. This data suggest that incorporation of caprylic
caprylic acid concentration range was the same for all mAbs (0.5–1.5%). acid precipitation method into mAb purification process
would add a robust and effective viral clearance step that
HCP HMW Yield would be complementary to the commonly used low pH
DoE pH/(CA) (ppm) (%) (%)
inactivation, thus providing an additional safety margin to
mAb B pH 4.9–5.3 Load 29,600 2.4 – the antibody products.
/þ 149 1.4 100
/ 173 1.4 100
0/0 117/121 1.3/1.3 99/99
þ/þ 84 1.2 97 Caprylic Acid Precipitation of Cell Culture Fluid
þ/ 117 1.3 97
mAb C pH 5.0–5.3 Load 1,676 3.3 –
Advances in cell culture technology have brought about
/þ 1,065 1.3 90 higher production titers by increasing cell densities,
/ 1,244 1.6 91 production duration, and specific productivity, which
0/0 106/107 1.1/1.1 90/87 also typically lead to higher levels of contaminants (Abu-
þ/þ 74 0.4 78 Absi et al., 2006; Glynn et al., 2009). Conventional cell
þ/ 144 0.9 81
mAb D pH 4.9–5.3 Load 565 3.6 –
separation techniques are not sufficient in clearing increased
/þ 15 2.5 99 levels of cellular debris resulting in high turbidity in the
/ 20 2.6 94 centrate. This would require the use of large depth filter
0/0 11/13 2.3/2.2 93/90 areas to prevent plugging and fouling of the depth filters and
þ/þ 9 2.0 90 the subsequent sterilizing grade absolute filters (Yigzaw
þ/ 12 2.0 92
et al., 2008). Effects of insufficient contaminant clarification

2594 Biotechnology and Bioengineering, Vol. 109, No. 10, October, 2012
Table V. Clearance of xMuLV in purified solutions of mAb B and F.

mAb Viral clearance condition mAb (g/L) Infectivity (Log10) RT-qPCR (Log10)
B pH 4.9, 1% CA 15 3.4 a
4.9
pH 5.1, 0.5% CA 15 4.7a 5.5
pH 5.1, 1% CA 15 4.2a 5.3
pH 5.3, 1% CA 15 3.4a 6.4
F pH 5.1, 1% CA 5 2.8 –
pH 5.1, 1% CA 15 3.0 –
pH 5.1, 1% CA 25 3.0 –
a
Assayed a large volume sample.

can be felt in the downstream unit operations as well of cell culture fluid (titration to pH 4.7 with phosphoric
(Yigzaw et al., 2006). Caprylic acid can be used to induce acid) versus lowering the pH and caprylic acid addition.
flocculation of impurities and cellular debris in the cell Data in Table VII show yields as well as HCP and HMW
culture fluid, while maintaining the mAb in the supernatant, aggregate levels in the protein A pool resulting from
in order to enhance removal of precipitated solids by disc- precipitation of impurities in the cell culture fluid. As can be
stack centrifugation. This method is more efficient and less gathered from the table, a higher fold reduction in HCP and
expensive than super sizing depth filters (Yigzaw et al., HMW aggregates is achieved with caprylic acid precipitation
2008). Precipitation of cell culture fluid was evaluated with when compared to acid precipitation or un-treated control.
mAbs C, F, and G to investigate how it affects the An important factor for consideration of caprylic acid
purification process. Precipitated cell culture fluid was precipitation step is a modest loss of product due to co-
protein A purified and the resultant elution pools were precipitation in cell culture fluid (yield 95%). However, it
analyzed for select product quality. Preliminary experiments is desirable to decrease the impurity burden over the
with mAb C, similar to the screening experiments with protein A step with a modest yield loss (5%). This will
protein A purified mAb solutions, were performed using result in superior filter throughput and a longer protein A
select concentration of caprylic acid and precipitation pH resin lifetime, both of which would contribute to a decrease
for cell culture fluid. Precipitation with 1% caprylic acid was in cost of goods. A direct observation of impurity reduction
sufficient to achieve HCP and HMW aggregate clearance due to flocculation of cell culture fluid can be made in the
with an acceptable yield (Table VI). However, unlike the case of mAb C and mAb F. When the clarified cell culture
protein A purified solutions, the output product quality fluid is flocculated with caprylic acid and the resultant
parameters of cell culture fluid precipitation had minimal supernatant samples analyzed by SDS–PAGE (Fig. 2), a
dependence on pH at caprylic acid concentration of 1%. noticeable reduction in high molecular weight species, along
Consequently, pH 4.7 was selected to be used in subsequent with a multitude of other impurities, is observed. This is also
experiments because of slightly higher yields. Similar pH is substantiated by direct measurement of HCP levels in
reported in the literature to be favorable for caprylic acid harvest cell culture fluid (see Fig. 2 caption). For both mAbs,
precipitation of serum and ascites fluid (McKinney and C and F, HCP levels in the protein A elution pools are
Parkinson, 1987). To further examine the effect of caprylic significantly lower when caprylic acid is used (Table VII).
acid flocculation of cell culture fluid on impurity reduction For mAb F, it appears that the precipitation of protein A
and process yield, three mAbs (C, F, and G) were used. This purified sample (Table IIa) is more effective in reducing
experiment compared the impact of simply lowering the pH HCP level than precipitation of cell culture fluid—181 ppm
versus 569 ppm.
Table VI. Effect of caprylic acid (CA) concentration (at a fixed pH) and
flocculation pH (at a fixed CA concentration) on yield and impurity
reduction for mAb C. Table VII. HCP and HMW aggregate levels in the protein A elution pool
after purification of un-treated feed (control), acid precipitated (AP), and
CA (%) (at pH 4.7) Yield (%) HCP (ppm) HMW (%) caprylic acid (CA) treated cell culture fluid for mAbs C, F, and G. AP and
CA precipitation performed at pH 4.7.
Feed – 6,948 1.6
0.5 98 470 1.0 mAb Condition Floc Yield (%) HCP (ppm) HMW (%)
1.0 93 490 0.6
1.5 79 600 0.7 C Feed – 756 3.1
AP 98 2,197 4.9
1.0% CA 90 137 0.7
Floc pH (at 1% CA) Yield (%) HCP (ppm) HMW (%) F Feed – 1,372 3.8
Feed – 6,948 1.6 AP 94 1,293 3.3
4.3 93 510 0.5 1.0% CA 92 569 1.9
4.7 94 470 0.6 G Feed – 2,729 2.0
5.1 97 780 1.0 AP 100 1,073 2.0
5.5 96 570 1.1 1.0% CA 101 111 1.2

Brodsky et al.: Monoclonal Antibody Purification Using Caprylic Acid 2595

Biotechnology and Bioengineering
 As already demonstrated with precipitation of protein A
purified mAb solutions, turbidity in the neutralized protein
A eluate is greatly reduced by caprylic acid (Fig. 3). In the
case of mAb C, severe turbidity that results in product loss is
observed when protein A elution pool is neutralized to pH 7
(Fig. 3). However, turbidity is drastically reduced from
1,000 NTU to 20 NTU by using caprylic acid treated cell
culture fluid and to 150 NTU in the case of acid
precipitation.

Process Integration and Scale Up Considerations

Figure 2. Precipitation of impurities in the cell culture fluids for mAb C and mAb
F. Lanes: 1 and 6, protein Std marker; Lanes: 2 and 7, control cell culture fluid; Lanes: 3
and 8, supernatant from AP treated cell culture fluid; and Lanes 4 and 9, caprylic acid
treated cell culture fluid. MAb C HCP levels in control, acid-precipitated, and caprylic
acid treated cell culture fluid is 375,471, 353,251, and 127,148, respectively. MAb F HCP
levels in control, AP treated, and CA treated cell culture fluid is 379,155, 374,584, and
91,301, respectively. [Color figure can be seen in the online version of this article,
available at http://wileyonlinelibrary.com/bit]
Process considerations with respect to ease of scale up were
considered for caprylic acid addition. The procedure of
caprylic acid addition to mAb solutions was evaluated in
order to minimize dilution and provide better control over
reaction pH. Precipitation reaction begins when introducing
caprylic acid at a concentration above its solubility limit
(Johnston et al., 2003). A homogeneous reaction mixture for
consistent precipitation results was obtained by introducing
caprylic acid emulsion to the process solution. Preparation
of caprylic acid emulsion required vigorous agitation with a
stock solution of caprylic acid and the buffer solution. This
procedure, though successful at lab scale, presents challenges
for scaling up, including, the size of dilution and
inconveniences associated with emulsion preparation.
Alternatively, sodium caprylate, a water soluble salt form
of caprylic acid, can be used to start the precipitation.
Implementation of a precipitation step after a protein A

1000

100
Turbidity (NTU)

10

1
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5

Neutralization pH

Figure 3. Turbidity of mAb C protein A pool neutralized to different pH after caprylic acid treatment of (A) protein A pool or (B) cell culture fluid. Legend: Untreated cell
culture/protein A pool (*), acid precipitated cell culture fluid (^), caprylic acid treated cell culture fluid (&), and caprylic acid treated protein A pool (~).

2596 Biotechnology and Bioengineering, Vol. 109, No. 10, October, 2012
capture column requires neutralization of the elution pool
to a desirable precipitation pH. The final caprylate
concentration was obtained by using the appropriate ratio
of tris base to sodium caprylate. The dilution that occurred
during the titration was minimal due to high concentration
of sodium caprylate and tris base (concentration > 2 M).
Moderate mixing of protein A elution pool during addition
of sodium caprylate solution was sufficient to form the
requisite emulsion. This method also provided a tight
control over pH of precipitation reaction. Sodium caprylate
was also used for precipitation of cell culture fluid. In this
case, cell culture fluid was first titrated to pH 4.2 with 1.0 M
citric acid and then 2.0 M sodium caprylate solution was
used to titrate up to pH 4.7. Two methods of precipitant
addition were evaluated using cell culture fluid of mAbs F
and G with 1% caprylic acid or sodium caprylate. For both
mAbs precipitation yield and protein A pool product quality
were comparable between the two methods. For mAb F,
HCP concentration in the protein A pool were reduced from
1,372 ppm in the non-precipitated control sample to 569
and 680 ppm for caprylic acid and sodium caprylate
precipitation method, respectively. For the corresponding
samples, the HMW aggregate was reduced from 3.8% to
1.9% and 2.3%, respectively. For mAb G, HCP concentra-
tion in the protein A pool were reduced from 2,729 to 111
and 128 ppm with caprylic acid and sodium caprylate
precipitation method, respectively. For the corresponding
samples, the HMW aggregate level was reduced from 2% to
1.2% for both the precipitation methods. For all conditions,
product yield was between 92% and 102%.
The data presented in the previous sections demonstrate
potential of caprylic acid precipitation method to become an
effective mAb purification step. Implementation of protein
A pool impurity precipitation with caprylic acid, however,
would require a dedicated depth filtration step to remove the
precipitants. In conventional processing, a substantial
increase in turbidity is often observed upon neutralization
of protein A elution pool with most mAbs (Yigzaw et al.,
2006), which requires a depth filtration step as a guard to
sterilizing grade absolute filters. As such, the mAb processes
would not require additional depth filtration steps as the
caprylic acid precipitated pools could be processed using the
existing bank of depth filters. Turbidity is significantly
reduced when caprylic acid precipitation is performed as
shown in Figure 3, due to reduction in impurities.
Minimizing turbidity would also allow for greater in-
process pool stability by minimizing effects of hold duration
on absolute filter throughput.
Further understanding of the effects of caprylic acid
precipitation of cell culture fluid should be pursued as
having a cleaner feed stream for a mAb purification process
has several benefits. The goal of precipitation step before the
cell separation is to minimize the burden on the harvest
depth filter train and the following capture column by
increasing efficiency of the disc-stack centrifuge operation.
This has a potential to (1) reduce the cost of goods by
decreasing required depth filter and/or sterilizing grade filter
Brodsky et al.: Monoclonal Antibody Purification Using Caprylic Acid
area and (2) enabling better facility fit for high cell density/
titer process into existing manufacturing footprints. The
current study did not evaluate centrifuge performance but
instead concentrated on demonstrating the effect caprylic
acid precipitation has on product purity. Implementation of
caprylic acid precipitation step in the bioreactor is suitable
from the operational stand point since it does not require an
additional step for precipitate removal. In addition, analysis
of protein A elution pools revealed no measurable caprylic
acid when precipitated CCF was used (LOD ¼ 0.025%).
Evaluation of centrifuge and depth filter performance for
harvest clarification is an important next step toward
incorporation of the precipitation step in the purification
process.
Conclusion
Improvements in the production titers of the mAb cell
culture process raise concerns about purification process
bottlenecks in the existing facility resulting in low plant
throughput and/or poor facility fit using conventional
purification methods. At commercial scale with multiple
bioreactor configurations needing swift turnaround, large
process volumes associated with flow-through polishing
chromatography causes constraints due to dilution of pool
volume resulting in process pool tank limitations. At clinical
scale with the desire to operate at smallest column scale for
minimizing resin costs, long processing times due to process
titers and multiple cycles of operations can lead to
throughput limitations. Precipitation of impurities with
caprylic acid provides an alternative to polishing chroma-
tography step for mAb purification. It can be implemented
in a bioreactor prior to cell separation or used to precipitate
impurities after protein A capture column. Since precipita-
tion can be carried out in existing tanks in manufacturing
operations, there is no need for substantial capital
investments. This method makes it possible to operate at
a very large scale, requiring short processing time and
minimal buffer preparation. The process volumes required
for the addition of sodium caprylate can be managed in a
relatively small (<500 L) disposable bio-bag at commercial
scale. In the case of precipitating impurities in cell culture
bioreactor, caprylic acid precipitation method provides a
higher quality feed stream for purification process,
minimizing the burden on the clarifying depth filter train
and the following capture step. Precipitation of impurities
after the protein A column provides clearance of host cell
proteins and high molecular weight aggregates that is
equivalent or superior to a polishing chromatography step.
Caprylic acid precipitation of the protein A pool leads to
complete clearance of xMuLV virus, which can be claimed as
an additional orthogonal and robust step for virus
reduction. Productivity is increased by a short process
development time and a simple scale-up procedure relative
to a chromatography step. Notably, caprylic acid precipita-
tion is effective for multiple mAb molecules demonstrating
2597
Biotechnology and Bioengineering
the capability of this technique to become a part of a generic
purification process.
We acknowledge the following groups and individuals for their
contributions to this work: (1) the Analytical Testing and Compliance
group at Amgen, WA for the analysis of multiple HCP ELISA and
analytical SEC HPLC samples; (2) the Biosafety Laboratory at Amgen,
WA for execution of the viral clearance study; (3) Joanne Reeder, Lisa
Connell-Crowley, Clayton Brooks, Suresh Vunnum, and Brian Hub-
bard for their continued support during this technology development
work.
References
Abu-Absi SF, Martinez A, Liu M, Mays C, Zhang C, Konstantinov KB. 2006.
Shear in high cell density perfusion cultures: Effects on host cell
impurity release and aggregation. San Francisco, CA: American
Chemical Society National Meeting.
Aldington S, Bonnerjea J. 2007. Scale-up of monoclonal antibody purifica-
tion processes. J Chromatogr B Analyt Technol Biomed Life Sci
848(1):64–78.
Birch JR, Onakunle Y. 2005. Biopharmaceutical proteins: Opportunities
and challenges. Methods Mol Biol 308:1–16.
Bull HB, Breese K. 1967. Binding of fatty acids by proteins. Arch Biochem
Biophys 120(2):303–308.
Fliszar KA, Wuelfing WP, Li Z, Reed RA. 2006. Profiling of medium chain
glycerides used in pharmaceutical formulation development by
reversed-phase HPLC. J Pharm Biomed Anal 40(4):896–900.
Glynn J, Hagerty T, Pabst T, Annathur G, Thomas K, Johnson P, Ramasu-
bramanyan N, Mensah P. 2009. The development and application of
a monoclonal antibody purification platform. BioPharm Int Suppl 8:
15–19.
Gottschalk U. 2008. Bioseparation in antibody manufacturing: The good,
the bad and the ugly. Biotechnol Prog 24(3):496–503.
Johnston A, Uren E, Johnstone D, Wu J. 2003. Low pH, caprylate incuba-
tion as a second viral inactivation step in the manufacture of albumin.
Parametric and validation studies. Biologicals 31(3):213–221.
Kelley B. 2007. Very large scale monoclonal antibody purification: The case
for conventional unit operations. Biotechnol Prog 23(5):995–1008.
Kelley B. 2009. Industrialization of mAb production technology: The
bioprocessing industry at a crossroads. mAbs 1(5):1–10.
Kelley BD, Tobler SA, Brown P, Coffman JL, Godavarti R, Iskra T, Switzer
M, Vunnum S. 2008. Weak partitioning chromatography for anion
exchange purification of monoclonal antibodies. Biotechnol Bioeng
101(3):553–566.
Korneyeva M, Hotta J, Lebing W, Rosenthal RS, Franks L, Petteway SR, Jr.
2002. Enveloped virus inactivation by caprylate: A robust alternative to
2598 Biotechnology and Bioengineering, Vol. 109, No. 10, October, 2012
solvent-detergent treatment in plasma derived intermediates. Biologi-
cals 30(2):153–162.
Low D, O’Leary R, Pujar NS. 2007. Future of antibody purification.
J Chromatogr B Analyt Technol Biomed Life Sci 848(1):48–63.
Lundblad JL, Seng RL. 1991. Inactivation of lipid-enveloped viruses in
proteins by caprylate. Vox Sang 60(2):75–81.
McKinney MM, Parkinson A. 1987. A simple, non-chromatographic
procedure to purify immunoglobulins from serum and ascites fluid.
J Immunol Methods 96(2):271–278.
Moscariello J. 2008. Comparison of Potential Monoclonal Antibody
Purification Processes with Two Chromatography Steps. The IBC
BioProcess International conference. San Diego, CA.
Parkkinen J, Rahola A, von Bonsdorff L, Tolo H, Torma E. 2006. A modified
caprylic acid method for manufacturing immunoglobulin G from
human plasma with high yield and efficient virus clearance. Vox
Sang 90(2):97–104.
Przybycien TM, Pujar NS, Steele LM. 2004. Alternative bioseparation
operations: Life beyond packed-bed chromatography. Curr Opin
Biotechnol 15(5):469–478.
Romanowski P, Schleh M, Rajurs V, Chinniah S, Dehghani H. 2008.
Variables affecting titer and long-term stability of virus stocks.
BioProcess Int 6(2):44–52.
Russo C, Callegaro L, Lanza E, Ferrone S. 1983. Re.: Purification of IgG
monoclonal antibody by caprylic acid precipitation. J Immunol
Methods 65(1–2):269–271.
Shi L, Chen Q, Norling LA, Lau AS, Krejci S, Xu Y. 2004. Real time
quantitative PCR as a method to evaluate xenotropic murine leukemia
virus removal during pharmaceutical protein purification. Biotechnol
Bioeng 87(7):884–896.
Shukla AA, Hinckley P. 2008. Host cell protein clearance during protein A
chromatography: Development of an improved column wash step.
Biotechnol Prog 24(5):1115–1121.
Shukla AA, Hubbard B, Tressel T, Guhan S, Low D. 2007. Downstream
processing of monoclonal antibodies–application of platform
approaches. J Chromatogr B Analyt Technol Biomed Life Sci
848(1):28–39.
Steinbuch M, Audran R. 1969. The isolation of IgG from mammalian sera
with the aid of caprylic acid. Arch Biochem Biophys 134(2):279–284.
Thommes J, Etzel M. 2007. Alternatives to chromatographic separations.
Biotechnol Prog 23(1):42–45.
Yigzaw Y, Piper R, Tran M, Shukla AA. 2006. Exploitation of the
adsorptive properties of depth filters for host cell protein removal
during monoclonal antibody purification. Biotechnol Prog 22(1):288–
296.
Yigzaw Y, Piper R, Thomas A, Zhao X, Hershberg R, Hewig A, Vedantham
G. 2008. Improving upstream feed stock to downstream operations
Recovery of Biologics Conference XIII. Quebec, Canada.