Article

pubs.acs.org/ac

A Novel Sample Preparation for Shotgun Proteomics

Characterization of HCPs in Antibodies
Lihua Huang,* Ning Wang, Charles E. Mitchell, Tammy Brownlee, Steven R. Maple,
and Michael R. De Felippis
Bioproduct Research and Development, Lilly Research Laboratories, Eli Lilly and Company, Indianapolis, Indiana 46285, United
States
*
S Supporting Information

ABSTRACT: Residual host cell proteins (HCPs) in bio-

pharmaceuticals derived from recombinant DNA technology
can present potential safety risks to patients or compromise
product stability. Thus, the downstream purification process is
designed to demonstrate robust removal of these impurities.
ELISA using polyclonal anti-HCP antibodies as reagents for
capture, detection, and quantitation purposes is most commonly
used to monitor HCP removal during process development, but
this technique has limitations. More recently, LC-MS for
residual HCP characterization has emerged as a powerful tool to
support purification process development. However, mass spectrometry needs to overcome the enormous dynamic range to
detect low ppm levels of residual HCPs in biopharmaceutical samples. We describe a simple and powerful methodology to
characterize residual HCPs in (monoclonal) antibodies by combining a novel sample preparation procedure using trypsin
digestion and a shotgun proteomics approach. Differing from the traditional methodology, the sample preparation approach
maintains nearly intact antibody while HCPs are digested. Thus, the dynamic range for HCP detection by MS is 1 to 2 orders of
magnitude less than the traditional trypsin digestion sample preparation procedure. HCP spiking experiments demonstrated that
our method could detect 0.5 ppm of HCP with molecular weight >60 kDa, such as rPLBL2. Application of our method to
analyze a high-purity NIST monoclonal antibody standard RM 8670 derived from a murine cell line expression system resulted in
detection of 60 mouse HCPs; twice as many as previously reported with 2D-UPLC/IM/MSE method. A control monoclonal
antibody used for 70 analyses over 450 days demonstrated that our method is robust.

H ost cell proteins (HCP) are process-related impurities of

biopharmaceuticals derived from cell culture and may
cause immunogenic reactions, adjuvant activity, product
instability, and result in protein-specific biological activity1−9
after injection. Thus, the downstream purification process must
be designed to achieve robust removal of these impurities. An
enzyme-linked immunosorbent assay (ELISA) with polyclonal
anti-HCP antibodies used for capture, detection, and
quantitation purposes is most commonly used to monitor
HCP removal during process development.10,11 Accurate
monitoring and measuring of residual HCPs in biopharma-
ceuticals by ELISA is highly dependent on the procedures used
to generate the polyclonal antibodies and analytical standards.
Generally, the antigens used for producing polyclonal antibody
reagents and analytical HCP standards are obtained from a
blank-run fermentation (or null strain) imitating the production
run but lacking the specific coding gene for the protein product
of interest. There are thousands of HCPs present in a null
strain sample, but not every host protein will produce
antibodies for ELISA detection. HCP profiles in a null strain
are very different from process samples, which may cause
inaccurate quantitation by ELISA. Furthermore, ELISA
provides no capability to identify specific host proteins. This
information is essential to assess HCP immunogenicity risks of
the biopharmaceutical7 and potential effects on stability.8,9
Application of mass spectrometry for protein identification
using shotgun proteomics is now routine. However, if host
proteins coelute with the biopharmaceutical, a mass spec-
trometer with up to 6 orders of magnitude dynamic range is
required to directly detect 1 to 100 ppm of HCPs in the
sample, which is out of range of current mass spectrometers.
There are several ways to overcome the dynamic range issues
for HCP characterization. One way is to resolve coeluting
peptides before MS/MS analysis with data-dependent acquis-
ition (DDA) or data-independent acquisition (DIA) by
increasing the separation time,12−17 adding another dimension
separation,6,18−23 such as 2D-HPLC, or using ion mobility.24
The other way is to enrich residual HCPs either by removing
the therapeutic protein from the sample with affinity
purification, such as using Protein A or G for an antibody,25,26
or by capturing residual HCPs with polyclonal antibodies.27 In
one study, near-single-digit ppm detection sensitivity28,29 was
Received: January 24, 2017
Accepted: April 17, 2017
Published: April 17, 2017

© 2017 American Chemical Society 5436 DOI: 10.1021/acs.analchem.7b00304

Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry
reported for HCPs in biopharmaceutical samples using LC-
MS/MS with DIA and a pre-established library comprising
accurate masses, retention times, and fragment ions for each
peptide derived from HCPs in the null strain runs.
The existing methods are powerful tools for detecting HCPs
in biopharmaceutical samples but they also have some
limitations. Generally, one-dimensional or even two-dimen-
sional HPLC MS/MS is only capable of detecting 10 to 50 ppm
of HCP, and the cycle times are very long, especially for 2D-
HPLC (e.g., 1 to 2 samples/day). These techniques are not
sensitive enough to detect very low level HCPs, which may be
present in sufficient amount to cause product instability.9 LC-
MS/MS with DIA and a pre-established library is sensitive and
has rapid throughput; however, the method may miss detection
of HCPs which are only coexpressed with specific products.
Furthermore, the technique is only suitable for a specific mass
spectrometer and still suffers from the dynamic range issue for
ion trap instruments when peptides are coeluted.
In this paper, we describe a simple and powerful method to
characterize HCPs in monoclonal antibodies (mAbs) or related
products by combining a novel tryptic peptide sample
preparation with traditional shotgun proteomics. Antibody
samples are directly treated with trypsin, pH adjusted, and
incubated overnight. Undigested antibody is precipitated with
heat treatment, membrane filtered, or kept in the solution
before shotgun proteomics analysis. Following this procedure,
the antibody is not digested or only minimally digested while
residual HCPs in the sample are digested. The dynamic range
of the mass spectrometer required for detection of HCPs is
found to be 1 or 2 orders of magnitude less than the same LC/
MS method but with a traditional (i.e., denatured first) sample
preparation for the tryptic digest.
■ EXPERIMENTAL SECTION
Materials. All chemicals were reagent grade or higher purity
and obtained from commercial sources. Chromatographic
solvents were LC-MS grade. CHO null strains, monoclonal
IgG1, IgG4, and five CHO recombinant proteins, rLPLA2,
rPLBL2, rLAL, rPPT1, rIAH1, as well as recombinant human
PCSK9 and bovine trypsin were produced at Eli Lilly and
Company (Indianapolis, IN). Dithiothreitol (DTT) was
purchased from Thermo Fisher Scientific (Rockford, IL,
U.S.A.). Alcohol dehydrogenase from Bakers yeast was
obtained from Sigma-Aldrich Co. (St. Louis, MO) and
monoclonal antibody standard RM 8670 was from National
Institute of Standards and Technology (NIST).
LC/MS Analysis for IgG1 and IgG4 Antibodies without
or with Treatment of Trypsin. An aliquot containing 1 mg
of antibody of IgG4 or IgG1 solution was mixed with 5 μL of 1
M tris-HCl buffer, pH 8 and water to 195 μL and then treated
with 5 μL of 0.5 mg/mL recombinant bovine trypsin, 0.05 mg/
mL dehydrogenase (ADH) from Bakers yeast, and 0.02 mg/mL
recombinant human PCSK9 (hPCSK9) solution at 37 °C for
overnight. Each digest was mixed with 2 μL of 50 mg/mL DTT
solution, heated at 90 °C for 5−10 min, and then centrifuged at
15 000g for 2 min. The supernatant was transferred for LC/MS
analysis. At each step, a 5 μL aliquot of each solution and 20 μL
of pure water were transferred into HPLC vials for LC/MS
analysis.
A 0.5 μL injection of each sample was analyzed by LC-MS. A
Waters Acquity UPLC (Milford, MA U.S.A.) coupled to Waters
SYNAPT G2-S mass spectrometer (Manchester, U.K.) was
applied for intact analysis. Separations were performed on a
5437
Article
Varian PLRP-S reversed-phase column (1 × 50 mm, 1000 Å, 5
μm) at 80 °C using 0.05% TFA in water as mobile phase A and
0.04% TFA in acetonitrile as mobile phase B. Each peak eluted
from the column was analyzed using an electrospray ionization
(ESI) source operating at positive, resolution model, scan range
of 400 to 4000 amu, spray voltage of 3.0 kV, cone voltage of
100 V, source offset of 120 V, source temperature of 120 °C,
desolvation temperature of 400 °C, and desolvation gas of 900
(L/h).
Tryptic Digests of CHO Null Strain Solutions without
or with Denaturation. Aliquots containing 200 μg of CHO
proteins from each of five CHO null strain solutions were
mixed with 5 μL of 1 M tris-HCl buffer, pH 8, and Barnstead
purified water to 198.4 μL or dried using a speed-vacuum.
Dried samples were reconstituted with 15 μL of 7 M guanidine·
HCl, 0.625 mM tris-HCl buffer, pH 8, and 2 μL of 50 mg/mL
DTT solution, incubated at 37 °C for 30 min, and then mixed
with 3 μL of 100 mg/mL iodoacetamide solution and purified
water to 198.4 μL. Each solution was then treated with 1.6 μL
of 2.5 mg/mL r-bovine trypsin solution and incubated at 37 °C
overnight (>16 h). Each tryptic digest without denaturation was
mixed with 2 μL of 50 mg/mL DTT solution and then
incubated at 90 °C for 10 min. The digests were centrifuged at
13 000g for 2 min, and each supernatant was mixed with 5 μL
of 10% formic acid in water before LC/MS/MS analysis.
Tryptic Digest of NIST Monoclonal Antibody Stand-
ard RM 8670 and Control Antibody. A100 μL aliquot of 10
mg/mL NIST standard or control antibody solution was mixed
with 5 μL of 1 M tris-HCl buffer, pH 8, and 90 μL of pure
water. To these solutions was added 5 μL of 0.5 mg/mL
recombinant bovine trypsin, 0.05 mg/mL ADH1, and 0.02 mg/
mL hPCSK9 solution and the preparations were incubated at
37 °C for overnight. Each sample was mixed with 2 μL of 50
mg/mL DTT solution and then heated at 90 °C for 10 min.
Each sample was centrifuged at 13 000g for 2 min and the
supernatant was transferred into HPLC vials and acidified with
5 μL of 10% FA in H2O before LC/MS analysis.
Tryptic Digest of IgG1 and IgG4 Samples Spiked with
CHO Null Strain or CHO Proteins. IgG1 and IgG4 solutions
were mixed with CHO null strain to produce 0, 0.1, 0.2, 0.5, 1,
2, 5, or 10 k ppm spiked samples or spiked with 0, 0.1, 0.2, 0.3,
0.5, 1, 2, 5, 10, 20, 50, and 100 ppm each of five recombinant
CHO proteins: lyophospholipase A2 (rLPLA2), phospholipase
B-like 2 (rPLBL2), lysosomal acid lipase (rLAL), palmitoyl-
protein thioesterase 1 (rPPT1), and isoamyl acetate-hydro-
lyzing esterase 1 homologue (rIAH1). Each spiked sample
containing 1 or 2.5 mg IgG1 or IgG4 antibody was mixed with
5 μL of 1 M tris-HCl buffer, pH 8, and water to 195 μL. The
samples were then subjected to the trypsin digestion procedure
described above.
UPLC-MS/MS Analysis of Tryptic Digests. The prepared
tryptic peptides were analyzed using UPLC-MS/MS. Samples
were directly injected onto a Waters Acquity UPLC CSH C18
(Milford, MA, U.S.A.) (2.1 × 50 mm, 1.7 μm particle size) at a
volume of 50 μL. The column was heated to 60 °C during
analysis. Separation was performed on a Waters Acquity UPLC
system with mobile phase A consisting of 0.1% formic acid in
water and mobile phase B consisting of 0.1% formic acid in
acetonitrile with equilibrating at 0% mobile phase B for 2 min
at 200 μL/min, linearly increasing from 0% to 10% over 23
min, to 20% B over 57 min, to 30% over 30 min at a flow rate
of 50 μL/min, followed with multiple zigzag wash cycles at a
flow rate of 400 μL/min. Mass spectrometric analysis was
DOI: 10.1021/acs.analchem.7b00304
Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry Article

Figure 1. LC-UV chromatograms of LC/MS analysis of IgG1 or IgG4 samples with or without treatment with trypsin at 37 °C overnight prior to or
after heating at 90 °C. (a−c) LC-UV chromatograms of IgG4 without trypsin treatment, with trypsin treatment, and with heating at 90 °C; (d−f)
LC-UV chromatograms of IgG1 without trypsin treatment, with trypsin treatment, and with heating at 90 °C.

performed on a Thermo Scientific Q Exactive Plus mass
spectrometer (Bremen, Germany). Data-dependent MS/MS
was performed as follows: the first event was the survey positive
mass scan (m/z range of 230−1500) followed by 10 HCD
events (28% NCE) on the 10 most abundant ions from the first
event. Ions were generated using a sheath gas flow rate of 15, an
auxiliary gas flow rate of 5, a spray voltage of 4 kV, a capillary
temperature of 320 °C, and an S-Lens RF level of 50.
Resolution was set at 35 000 (AGC target of 5E6) and 17 500
(AGC target of 5E4) for survey scans and MS/MS events,
respectively. The maximum ion injection time was 250 ms for
survey scan, 300 ms for the other scans. The dynamic exclusion
duration of 60 s was used with a single repeat count.
Protein Identification and Quantification. A customized
protein database composed of sequences obtained from the
CHO-K1_refseq_2014.Protein.fasta database (downloaded 08/
23/2014 from http://www.chogenome.org) was developed to
5438
predict the identities of HCPs from the MS/MS data. The MS/
MS data was searched with a mass tolerance of 10 ppm and
0.02 Da, and a strict false discovery rate (FDR) ≤ 1% against
this database using the Proteome Discoverer software package,
version 1.4 (Thermo Scientific, Bremen, Germany) with
Sequest HT searching. Further peptide/protein filtering was
performed by eliminating proteins that had scored 0 and single
spectrum hit, or single spectrum hit and ≥10 ppm and
contaminated human proteins. Protein area from the top 3
peptides (if possible) for each HCP and the areas for the three
spiked proteins, r-trypsin, PCSK9, and ADH1 were used to
calculate individual HCP concentration (ppm or ng HCP/mg
mAb).
DOI: 10.1021/acs.analchem.7b00304
Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry Article

■ RESULTS AND DISCUSSION

Host cell proteins (HCPs) are process-related impurities in
Table 1. Number of HCPs Detected by Shotgun Proteomics
for the Tryptic Digests of CHO Null Strains (NS) with the
biopharmaceutical products derived from cell culture. Applica- Traditional or Novel Method and Number of Unique HCPs
tion of sensitive analytical methods to monitor HCPs Associated with Sample Preparations Are Reported for the
Top 500 HCPs Detected
throughout product development is expected by worldwide
regulatory authorities. Although ELISA remains the industry for top 500 HCPs,
standard method for HCP detection, a shotgun proteomics HCPs detected unique HCPs with
approach using mass spectrometry is becoming more generally null strain traditional novel traditional novel
applied as an additional characterization tool. One problem for NS 1 1159 1199 18 7
mass spectrometry detection of HCPs in biopharmaceutical NS 2 1179 1165 19 3
products is the dynamic range, which requires 5 to 6 orders of NS 3 1147 1211 20 7
magnitude to detect 10 to 1 ppm of individual HCPs if their NS 4 1113 1049 18 4
peptides are coeluted. This dynamic range is outside the NS 5 1077 1134 24 13
capability of current mass spectrometers. NS 5, 10% injected 871 959 33 18
In general, antibodies are very stable and resistant to trypsin NS 5, 20% injected 1084 1074 26 11
digestion in their native state. Data for an IgG1 and an IgG4 NS 5, 50% injected 1175 1176 15 11
monoclonal antibodies treated with trypsin overnight is shown
in Figure 1. For IgG4, the UV profiles are very similar without
(Figure 1a) or with (Figure 1b) trypsin treatment followed by secreted host cell protein that is an antigen of the expressed
overnight incubation. For IgG1, the UV profile after treatment monoclonal antibody (mAb), and its epitope is conformational.
with trypsin (Figure 1e) is very different from that of the Although a rare case, such a situation may exist. Regardless, less
than 2% of HCPs for the top 500 HCPs detected in the CHO
untreated sample (Figure 1d). The two new peaks eluting
null strain could not be identified with our method. This small
earlier than the IgG1 intact peak (8.1 min) are Fab (7.6 min)
difference is less than typical run to run variability.
and (Fc)2 (7.4 min) due to the cleavage in the hinge region (···
Other differences between the two sample preparations were
CDK/THT···), which is a sensitive site for trypsin and Lys-C.
observed in the protein sequence recovery, score, and unique
However, for either IgG4 or IgG1 antibody samples, addition of
peptides for each identified protein. For example, the large,
the reducing reagent, DTT, followed by heating results in
highly abundant basement-membrane-specific heparan sulfate
precipitation of undigested antibody or generated Fab and Fc
proteoglycan core protein, yielded the following results: score
domains, which can be removed by centrifugation. Even of 414.2, protein sequence recovery of 35.3, and number of
without the precipitation and removal of the IgG proteins, they unique peptides of 102, by the traditional sample preparation,
will be eluted later than most of the tryptic peptides of HCPs whereas values of 228.4, 19.5 and 55, respectively, were
and will not complicate HCP detection. If CHO HCPs can be obtained for our method. In general, proteins identified with
digested or partially digested with trypsin under the same the traditional sample preparation produced slightly higher
conditions, the required dynamic range to detect HCPs is only scores and protein sequence recovery. The average score,
3 to 4 orders of magnitude, which is the range of current mass protein sequence recovery, and unique peptides for top 500
spectrometers. Additional experiments were conducted to proteins identified in the five null strain samples are 38.7 ± 2.6,
confirm whether this type of sample preparation is sufficient 30.8 ± 2.0, and 10.0 ± 0.9 with the traditional sample
to enable detection and quantitation of trace amounts of preparation and 32.8 ± 4.1, 25.4 ± 1.8, and 7.8 ± 0.4 with our
residual HCPs in monoclonal antibody preparations. method.
CHO Protein Identification for Null Strain Samples. The ability to identify CHO proteins in the null strain
We first wanted to determine whether our sample preparation samples and data for native antibody demonstrate that direct
method is as effective at digesting host cell proteins compared treatment of the sample with trypsin followed by overnight
to the traditional method, which involves tryptic digestion incubation causes minimal degradation of the antibody while
following denaturing and reduction/alkylation. A sample most CHO HCPs are digested. The procedure we devised
containing the protein pool produced from a null CHO cell achieves a major improvement in the dynamic range required
line fermentation (i.e., cells do not contain the gene for for mass spectrometry detection which is 1 to 2 orders of
antibody production) was used as an example to evaluate our magnitude lower than what is required using the traditional
method against the traditional procedure. Typically, we sample preparation method. Another advantage with having
observed greater than 1000 HCPs in null strain samples less digested antibody is the RP-HPLC column has more
treated by both sample preparation methods (Table 1). For the capacity to separate HCP peptides from antibody peptides
top 500 HCPs, 18 to 24 unique proteins were identified with thereby increasing the HCP identification capability.
the traditional method, whereas 3 to 13 unique HCPs were HCP Detection with Spiked Samples. A series of
identified with our procedure. Careful analysis of the unique experiments were conducted to evaluate whether our method
HCPs related to each sample preparation revealed a total of 10 is effective for detecting HCPs spiked into samples. In one
HCPs, detected with the traditional sample preparation experiment, preparations containing IgG1 or IgG4 antibodies
procedure but not found in any of the five null strain samples were spiked with varying amounts (0 to 10 000 ppm) of a
examined using our method for sample preparation. These CHO null strain sample. The results from this experiment are
HCPs generally contain a high content of Cys residues. For presented in Table 2. At a low level of spiking, more HCPs
example, granulins, an HCP identified only with the traditional were detected in the spiked IgG samples than the null strain
sample preparation, contains 94 Cys residues out of a total of sample, whereas the opposite was observed at the very high
610 amino acid residues. Another possibility for missed HCP level of spiking. This result may be related to there being more
detection using our method would involve a specific and small HCPs injected for the spiked IgG samples as compared to the
5439 DOI: 10.1021/acs.analchem.7b00304
Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry Article

Table 2. Number of HCPs Detected and HCP CHO protein spiking experiments, which were all linear
Concentration (ppm) Quantified by the Novel Method for between 2−100 ppm spiking range (Table 3).
IgG1 and IgG4 Preparations Spiked with Various Amounts One advantage of our method is that much higher protein
of the Null Strain concentration can be used for tryptic digestion without
experiencing precipitation. With the traditional tryptic digestion
null strain without IgG1 spiked with IgG4 spiked with
antibody null strain null strain approach, it is difficult to keep antibodies in solution when they
are denatured at high protein concentration without including a
null strain HCP HCP HCP
spiked no. concn no. concn no. concn high concentration of denaturing agents, such as urea or
(ppm)a HCP (ppm) HCP (ppm) HCP (ppm) guanidine. Hence, in order to increase the detection sensitivity
0 0 0 11 31 7 9 of HCP, concentrations of antibody greater than 5 mg/mL can
100 6 1 34 101 33 56 be used. For example, using our method to analyze a sample
200 21 28 56 209 75 208 containing 12.5 mg/mL antibody, the five CHO proteins could
500 36 92 103 636 135 559 be detected at a spiked level of 0.5 ppm.
1000 84 239 156 1289 221 1278 The applicable antibody concentration for our method is not
2000 207 735 223 1969 343 2773 limited to 12.5 mg/mL. We have analyzed samples with
5000 438 3643 388 7728 481 6524 concentrations ≥150 mg/mL when instrumentation with lower
10000 594 9589 502 16872 575 13169 sensitivity than the Thermo Q-Exactive plus mass spectrometry
a
ppm (or ng HCPs/mg antibody) based on antibody concentration. was used, or to detect very low (<0.1 ppm) residual HCPs,
such as lipases or proteases. We selected a 5 mg/mL antibody
concentration for our studies on the basis of the following
null strain sample due to the trace amount of residual HCPs in reasons: (1) The concentration provides sufficient material for
the nonspiked antibody samples. In contrast, at the high spiked detecting 10 ppm or lower residual HCP. (2) Antibody
level, the minimally digested antibody peptides diminished concentrations of most in-process samples are generally >5 and
HCP detection due to the selected peptide competition for <20 mg/mL. For higher concentration samples, a simple
MS/MS analysis. Regardless of the spiking level, application of dilution works well.
our sample preparation procedure achieved ≥85% detection of HCP Quantitation. Protein sequence recovery with our
spiked null strain proteins. method is generally lower than that obtained with the
In another experiment, five recombinant CHO proteins were traditional sample preparation procedure. Consequently,
spiked into samples to evaluate the detection sensitivity of our protein quantitation with Hi3 peptide intensities30,31 may be
method. These proteins were spiked into preparations different for our method compared with the traditional sample
containing 5 mg/mL antibody and subjected to our sample preparation approaches. To achieve more consistent quantita-
preparation procedure followed by mass spectrometry analysis. tion, instead of using one internal standard as in the original
As shown in Table 3, all five proteins spiked at 1 ppm each Hi3 method, three standards were chosen here, including two
were detected in the IgG4 preparations, whereas three of the spiked proteins, recombinant human PCSK9 (hPCSK9) and
proteins were detected in the IgG1-containing samples. When alcohol dehydrogenase (ADH1) from Baker’s yeast, as well as
the proteins were spiked at ≥2 ppm, all five were identified for recombinant bovine trypsin (i.e., the enzyme used for
both the IgG1 and IgG4 containing preparations. digestion). Individual HCP titers can be calculated on the
A similar number (Table 2) of HCPs was detected for the basis of the top three peptides of each spiked protein to give
spiking experiments with a CHO null strain sample three HCP titers. These values are then averaged to yield the
demonstrating that antibody precipitation did not coprecipitate final reported number. The recombinant bovine trypsin and
HCP peptides. The fact that HCP peptides were preserved human PCSK9 (hPCSK9) were selected because no trypsin or
during the precipitation was also confirmed by the purified PCSK9 was detected in the CHO null strains. Users of our

Table 3. Recovery of Recombinant Proteins Spiked into IgG1 or IgG4 Samples by Shotgun Proteomics with the Novel Methoda
recovery of spiked protein concentration, ppm (no. unique peptides)
IgG1 spiked with proteins (ppm)
spiked protein 0b 0.3b 0.5b 0.75b 1.0 2.0 5.0 10 20 50 100
rLPLA2 - ND ND ND 2.1 (3) 3.8 (2) 7.2 (3) 14 (5) 40 (7) 113 (10) 227 (12)
rLAL - ND ND ND - 2.6 (2) 4.4 (1) 15 (5) 33 (6) 76 (7) 171 (9)
rPPT1 - ND ND ND - 2.2 (2) 6.1 (4) 18 (6) 39 (6) 112 (9) 211 (12)
rPLBL2 - ND ND ND 2.6 (2) 3.1 (2) 6.9 (8) 12 (7) 27 (11) 70 (13) 129 (14)
rIAH1 - ND ND ND 0.6 (1) 1.4 (2) 2.3 (4) 8.4 (8) 20 (10) 56 (13) 95 (13)
IgG4 spiked with proteins (ppm)
spiked protein 0b 0.3b 0.5b 0.75b 1.0 2.0 5.0 10 20 50 100
rLPLA2 - - 0.8 (1) 1.8 (2) 1.5 (3) 3.1 (4) 9.7 (8) 21 (10) 44 (13) 105 (16) 226 (18)
rLAL - - 1.2 (1) 1.4 (2) 1.4 (1) 2.4 (4) 9.8 (7) 18 (8) 37 (7) 91 (9) 191 (10)
rPPT1 - 0.4 (1) 0.6 (2) 0.7 (3) 0.6 (1) 1.3 (4) 8.4 (7) 16 (7) 43 (10) 102 (12) 214 (13)
rPLBL2 - 0.0 (1) 0.5 (4) 0.8 (5) 0.5 (2) 1.9 (4) 3.4 (4) 10 (7) 25 (9) 62 (14) 124 (22)
rIAH1 - 0.3 (1) 0.5 (2) 0.6 (3) 0.4 (1) 1.2 (5) 4.0 (8) 7.4 (8) 16 (10) 40 (11) 95 (14)

a
Note: ND = not determined; - = not detected. bmAb concentration of the final tryptic digest is 12.5 mg/mL, and all the other are 5 mg/mL.

5440 DOI: 10.1021/acs.analchem.7b00304

Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry
method can apply other purified proteins as standards for
quantitation.
Using this approach, the calculated recoveries for IgG1 and
IgG4 preparations spiked with CHO null strain and the five
recombinant proteins are shown in Tables 2 and 3. The
recovery of CHO null strain is between 98 to 169% or 112 to
139% for IgG1 or IgG4, respectively, when spiked at ≥500
ppm. For the spiked recombinant proteins, the recovery of each
protein is different. rLPLA2, rLAL, and rPPT1 showed ≥100%
recovery; rIAH1 < 100%; and rPLBL2 approximately 100% for
levels between 2 to 100 ppm spiked into preparations
containing 5 mg/mL digest concentration of IgG1 or IgG4.
Although an HCP standard is certainly required for accurate
measurement of each HCP, quantitation using the Hi3 method
versus the three protein standards is sufficient to support
purification process development.
Method Robustness. A separate monoclonal antibody
control sample is simultaneously analyzed in each experiment
to ensure run-to-run consistency. During a 450-day time
interval, the control was analyzed 70 times. The calculated
concentrations (ppm) for the predicted HCPs in the control
sample are presented in Figure 2a. Average values for HCPs are
919 and 788 ppm with relative standard deviation of
approximately16 and 17% accounting for every HCP detected
or only accounting for those HCPs ≥ 10 ppm per HCP,
respectively. The control antibody also contains more than 10
Figure 2. (a) Total HCP concentration (ppm) for all HCPs (◆) or
each HCP ≥ 10 ppm (■) detected in the control monoclonal
antibody and (b) concentrations of five HCPs detected in the control
monoclonal antibody for 70 runs over a period of greater than 450
days.
5441
Article
HCPs at levels ≥10 ppm per HCP and detected in every run.
The relative standard deviation is between 20 to 40% for the 70
runs over the 450-day period of time. Figure 2b shows the
estimated concentrations for five HCPs which are present at
levels from ∼10 to >200 ppm. These data demonstrate that our
method is robust and capable of consistently detecting HCPs at
levels ≥10 ppm.
HCPs in NIST Monoclonal Antibody Standard RM
8670. Many powerful methods have been published for
characterization of HCPs in biopharmaceutical samples but
most have been only applied for specific products that are not
available for analysis by other researchers. For this reason, it is
not possible to directly compare results across the different
methods. Doneanu et al. applied their 2D-HPLC/IM/MSE
method in three independent laboratories24 to analyze the
NIST antibody standard RM 8670 and identified a combined
34 HCPs. Fourteen of the 34 HCPs were detected by all three
laboratories and one, mouse peroxiredoxin 5, was present at a
level of about 1 ppm or 1 ng/mg antibody.
For comparison, we applied our method to characterize the
HCPs in the NIST RM 8670 standard. More than 100 hundred
mouse proteins were identified for each of three LC/MS/MS
injection replicates processed using Thermo Scientific
Proteome Discoverer 1.4 against the Uniprot mouse protein
database and a false positive detection rate (FDR) ≤ 1%. If we
only consider those HCPs with at least two unique peptides per
protein and detected in all three injections, 59 mouse HCPs
were identified (see Table 4). In total, 13 of the 14 HCPs
detected with 2D-HPLC/IM/MSE procedure were identified
by our method. The missing protein, beta-2-microglobulin, was
actually detected but only with one unique peptide for each of
the three injections. The tandem mass spectrum of the unique
peptides identified demonstrated excellent peptide sequence
coverage as shown in Figure 1S in the Supporting Information.
While similar trends were observed, the quantitation data for
HCPs obtained with our method was not in good agreement
with reported data for several HCPs.
Forty-six mouse HCPs in the NIST antibody standard RM
8670 identified with our method were not reported in the study
using 2D-HPLC/IM/MSE. In total, 14 of the 46 mouse HCPs
have tandem mass spectra for five or more unique peptides, 17
for ≥4 unique peptides, and 28 for ≥3 unique peptides. Most of
the 46 HCPs were at levels less than 10 ppm or ng HCP/mg
antibody, and only three HCPs, protein disulfide-isomerase A6
(132 ppm), hepatocyte growth factor-like protein (14 ppm),
and protein ABHD11 (10 ppm), were ≥10 ppm. Generally, the
quality of the tandem mass spectrum data obtained for the
identified peptides was very good even for HCPs present at ≤1
ppm (Figure 2S in the Supporting Information).
Why did our method identify many more HCPs than 2D-
HPLC/IM/MSE? Under native conditions, the NIST antibody
standard treated with trypsin overnight only showed five major
peaks, and the other peaks related to antibody are relatively low
(Figure 3). For a 250 μg antibody (calculated) injection, the
intensities of the five major peaks are much lower and peak
widths much narrower than the tryptic peptide peaks obtained
with the traditional method. Comparing 1D-UPLC (our
method run time: 2 h) to 2D-UPLC (total run time: ∼10 h),
the resolution of separation was increased 5 times assuming
that the correction of the resolution and HPLC run time is
linear. However, our method only digests a few percent of the
antibody, effectively increasing the resolution of separation
more than 10 times. Furthermore, with our procedure, the
DOI: 10.1021/acs.analchem.7b00304
Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry Article

Table 4. HCPs Detected in the National Institute of Standards and Technology (NIST) Monoclonal Antibody Standard, RM
8670, by the Novel Methoda
measured HCP (ppm)
no. accession description unique pep inj. no. 1 inj. no. 2 inj. no. 3 ave
1 P05064 fructose-bisphosphate aldolase A 22 222 222 221 222
2 P05063 fructose-bisphosphate aldolase C 19 61 54 50 55
3 P06745 glucose-6-phosphate isomerase 15 19 20 20 20
4 Q91YR9 prostaglandin reductase 1 8 5 5 4 5
5 P40142 transketolase 5 2 1 1 1
6 Q99KN9 clathrin interactor 1 4 10 12 11 11
7 Q9CZ44 NSFL1 cofactor p47 4 3 3 4 3
8 P08101 low affinity immunoglobulin gamma Fc region receptor II 4 3 4 3 3
9 Q923D2 flavin reductase (NADPH) 4 3 2 3 3
10 P99029 peroxiredoxin-5, mitochondrial 4 1 1 1 1
11 Q8BL97 serine/arginine-rich splicing factor 7 3 2 2 2 2
12 Q9WTP6 adenylate kinase 2, mitochondrial 3 2 2 2 2
13 Q60864 stress-induced-phosphoprotein 1 2 7 7 7 7
14 P01887 beta-2-microglobulin 1 12 12 12 12
15 Q9Z0 × 1 apoptosis-inducing factor 1, mitochondrial 11 4 6 4 5
16 Q8C7U7 polypeptide N-acetylgalactosaminyltransferase 6 10 5 7 6 6
17 Q8K4F5 protein ABHD11 7 10 10 9 10
18 Q62179 semaphorin-4B 7 5 10 5 6
19 O88569 heterogeneous nuclear ribonucleoproteins A2/B1 7 6 6 6 6
20 Q9EPX2 papilin 7 6 7 5 6
21 Q9ER00 syntaxin-12 7 4 5 3 4
22 Q6PB93 polypeptide N-acetylgalactosaminyltransferase 2 7 3 3 3 3
23 P49312 heterogeneous nuclear ribonucleoprotein A1 6 3 4 4 4
24 Q03173 protein enabled homologue 6 3 3 3 3
25 Q922R8 protein disulfide-isomerase A6 5 132 131 132 132
26 P26928 hepatocyte growth factor-like protein 5 14 11 16 14
27 Q9D2M8 ubiquitin-conjugating enzyme E2 variant 2 5 4 4 4 4
28 P10126 elongation factor 1-alpha 1 5 2 2 2 2
29 Q3UEB3 poly(U)-binding-splicing factor PUF60 4 3 3 3 3
30 P40124 adenylyl cyclase-associated protein 1 4 0.8 0.8 0.8 0.8
31 Q9D8B3 charged multivesicular body protein 4b 4 1 1 1 1
32 Q68FL6 methionine–tRNA ligase, cytoplasmic 3 6 6 8 7
33 Q8BGD9 eukaryotic translation initiation factor 4B 3 5 5 4 5
34 Q6KCD5 nipped-B-like protein 3 3 4 4 4
35 Q68FF6 ARF GTPase-activating protein GIT1 3 3 3 2 3
36 P45878 peptidyl-prolyl cis−trans isomerase FKBP2 3 2 3 2 2
37 P11680 properdin 3 1 2 1 2
38 Q8BND5 sulfhydryl oxidase 1 3 2 2 2 2
39 Q62165 dystroglycan 3 2 1 1 1
40 Q9DBP5 UMP-CMP kinase 3 1 1 1 1
41 P35700 peroxiredoxin-1 3 1 1 1 1
42 P21460 cystatin-C 3 <0.5 1 <0.5 <0.5
43 Q3TLH4 protein PRRC2C 2 7 10 7 8
44 Q8CGC7 bifunctional glutamate/proline–tRNA ligase 2 3 3 4 3
45 Q80WJ7 protein LYRIC 2 3 2 1 2
46 O70305 ataxin-2 2 2 2 2 2
47 Q8K4Z5 splicing factor 3A subunit 1 2 1 1 1 1
48 P32020 nonspecific lipid-transfer protein 2 1 1 1 1
49 Q9QUR8 semaphorin-7A 2 1 1 1 1
50 O55131 septin-7 2 1 1 1 1
51 P18242 cathepsin D 2 1 1 1 1
52 P09041 phosphoglycerate kinase 2 2 1 1 1 1
53 P03975 IgE-binding protein 2 1 1 1 1
54 Q9CQF3 cleavage and polyadenylation specificity factor subunit 5 2 1 1 1 1
55 P34902 cytokine receptor common subunit gamma 2 1 1 <0.5 1
56 P53996 cellular nucleic acid-binding protein 2 1 <0.5 1 1
57 P08249 malate dehydrogenase, mitochondrial 2 <0.5 <0.5 <0.5 <0.5
58 Q6PDM2 serine/arginine-rich splicing factor 1 2 <0.5 <0.5 1 <0.5

5442 DOI: 10.1021/acs.analchem.7b00304

Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry Article

Table 4. continued
measured HCP (ppm)
no. accession description unique pep inj. no. 1 inj. no. 2 inj. no. 3 ave
59 Q6PGH2 hematological and neurological expressed 1-like protein 2 <0.5 <0.5 <0.5 <0.5
60 P19157 glutathione S-transferase P 1 2 <0.5 <0.5 <0.5 <0.5
a
Note: the first 14 HCPs were reported by Doneanu et al.24 The concentrations (ppm) for those 14 HCP reported are 116, 97, 12, 7, 3, 16, 7, 14, 2,
1, 5, 4, 13, and 7 ppm. All human proteins contaminated were removed.

Figure 3. LC-UV chromatogram obtained from LC/MS analysis with 250 μg of injection of NIST monoclonal antibody standard RM 8670 after
direct treatment with tryspin in the ratio of antibody/trypsin = 400/1 at 37 °C overnight.

dynamic range of mass spectral detection is 1 to 2 orders of
magnitude lower than the traditional sample preparation.
Interestingly, the NIST antibody standard RM 8670 is the
only sample that did not precipitate after the trypsin-treated
sample was heated with DTT at 90 °C for 10 min out of more
than 60 molecules of different types including: IgG1, IgG1-
effector null, IgG2, IgG4 antibodies, bispecific antibodies, Fc-
fusion peptides and proteins, Fab, and PEG-Fab analyzed with
our method. However, HCP detection is still possible even if
when its level is ≥10 ppm in preparations having antibody
concentrations at 5 mg/mL. Our method also detected a
greater number of HCPs in the NIST monoclonal antibody
standard RM 8670 compared to results obtained using 2D-
HPLC to analyze samples prepared with the traditional tryptic
digest procedure. More importantly, our methodology has an
overall shorter cycle time compared to the 2D-HPLC method
making it possible to achieve a higher-throughput sample
analysis to support purification process development.

■
the antibody does not precipitate because the reduced antibody
is eluted later than most of the HCP tryptic peptides and does ASSOCIATED CONTENT
not interfere with detection. In fact, removal of undigested
antibody is not necessary for detection of HCPs using our *
S Supporting Information

method. The only benefit to removing undigested antibody is
to increase the column lifetime or shorten the HPLC run cycle.
Without removal, 250 μg of digested or undigested (major
form) antibody will be injected on the column. The column will
lose resolution after several injections without using a
prolonged column wash step. With the precipitation step, the
column resolution will be maintained after several hundred
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.anal-
chem.7b00304.
Additional information as noted in the text, including
experimental workflow, tandem mass spectrum of a beta-
2-microglobulin peptide, and tandem mass spectrum of
an adenylyl cyclase-associated protein 1 peptide (PDF)

■
injections with moderate column washing.

■ CONCLUSIONS
We developed a simple and powerful methodology for HCP
AUTHOR INFORMATION
Corresponding Author
identification and quantitation in antibody preparations by *E-mail: huang_l@lilly.com.
combining a shotgun proteomics approach with a novel sample ORCID
preparation procedure utilizing trypsin digestion under non- Lihua Huang: 0000-0001-9037-3668
denaturing conditions. Under these conditions, the therapeutic Notes
derived from the CHO cell line is maintained largely intact The authors declare no competing financial interest.

■
while HCPs are digested. Spike recovery experiments
demonstrate that the method is very sensitive and robust.
The method consistently detected recombinant proteins spiked ACKNOWLEDGMENTS
as low as 0.5 ppm in preparations having antibody We thank Troii Hall and Stephanie Sandefur for preparing
concentrations at 12.5 mg/mL, and it always detected HCP purified proteins used in this work.
5443 DOI: 10.1021/acs.analchem.7b00304
Anal. Chem. 2017, 89, 5436−5444
Analytical Chemistry Article

■ REFERENCES
(1) Bierich, J. R. Acta Paediatr. 1986, 325, 13−18.
(31) Silva, J. C.; Denny, R.; Dorschel, C.; Gorenstein, M.; Li, G. J.;
Richardson, K.; Wahl, D.; Geromanos, S. J. Mol. Cell. Proteomics 2006,
5, 589.
(2) Rosenberg, A. S.; Worobec, A. S. BioPharm. Int. 2004, 17, 34−42.
(3) Romer, T.; Peter, F.; Saenger, P.; Starzyk, J.; Koehler, B.;
Korman, E.; Walczak, M.; Wasik, R.; Ginalska-Malinowska, M.;
Solyom, E.; Berghout, A. J. Endocrinol. Invest. 2007, 30, 578−589.
(4) Zuch de Zafra, C. L.; Quarmby, V.; Francissen, K.; Vanderlaan,
M.; Zhu-Shimoni, J. Biotechnol. Bioeng. 2015, 112, 2284−2291.
(5) Bracewell, D. G.; Francis, R.; Smales, C. M. Biotechnol. Bioeng.
2015, 112, 1727−37.
(6) Jawa, V.; Joubert, M. K.; Zhang, Q.; Deshpande, M.;
Hapuarachchi, S.; Hall, M. P.; Flynn, G. C. AAPS J. 2016, 18,
1439−1452.
(7) Fischer, S. K.; Cheu, M.; Peng, K.; Lowe, J.; Araujo, J.; Murray,
E.; McClintock, D.; Matthews, J.; Siguenza, P.; Song, A. AAPS J. 2017,
19, 254−263.
(8) Dixit, N.; Salamat-Miller, N.; Salinas, P. A.; Taylor, K. D.; Basu, S.
K. J. Pharm. Sci. 2016, 105, 1657−1666.
(9) Hall, T.; Sandefur, S.; Frye, C. C.; Tuley, T.; Huang, L. J. Pharm.
Sci. 2016, 105, 1633−1642.
(10) Krawitz, D. C.; Forrest, W.; Moreno, G. T.; Kittleson, J.;
Champion, K. M. Proteomics 2006, 6, 94−110.
(11) Hogwood, C. E. M.; Bracewell, D. G.; Smales, C. M.
Bioengineered 2013, 4, 288−291.
(12) Rogers, R.; Bell, A.; Kowski, T.; Bailey, R. Presented at 61st
ASMS Conference on Mass Spectrometry and Allied Topics,
Minneapolis, MN, June 9−13, 2013.
(13) Yu, C. Presented at 10th Symposium on the Practical
Applications of Mass Spectrometry in the Biotechnology Industry,
Boston, MA, September 23−26, 2013.
(14) Joucla, G.; Le Senechal, C.; Begorre, M.; Garbay, B.; Santarelli,
X.; Cabanne, C. J. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci.
2013, 942−943, 126.
(15) Reisinger, V.; Toll, H.; Mayer, R. E.; Visser, J.; Wolschin, F.
Anal. Biochem. 2014, 463, 1.
(16) Xiao, G.; Ren, D.; Bondarenko, P. Presented at 62nd ASMS
Conference on Mass Spectrometry and Allied Topics, Baltimore, MD,
June 15−19, 2014.
(17) Sperry, J. B. Presented at First Annual BEBPA Workshop on
Host Cell Proteins Assays, Dubrovnik, Croatia, May 16, 2014.
(18) Doneanu, C.; Xenopoulos, A.; Fadgen, K.; Murphy, J.; Skilton,
S. J.; Prentice, H.; Stapels, M.; Chen, W. mAbs 2012, 4, 24−44.
(19) Schenauer, M. R.; Flynn, G. C.; Goetze, A. M. Anal. Biochem.
2012, 428, 150.
(20) Schenauer, M. R.; Flynn, G. C.; Goetze, A. M. Biotechnol. Prog.
2013, 29, 951.
(21) Tscheliessnig, A. L.; Konrath, J.; Bates, R.; Jungbauer, A.
Biotechnol. J. 2013, 8, 655.
(22) Zhang, Q.; Goetze, A. M.; Cui, H.; Wylie, J.; Trimble, S.; Hewig,
A.; Flynn, G. C. mAbs 2014, 6, 659−670.
(23) Farrell, A.; Mittermayr, S.; Morrissey, B.; Mc Loughlin, N.;
Iglesias, N. N.; Marison, I. W.; Bones, J. Anal. Chem. 2015, 87, 9186−
9193.
(24) Doneanu, C. E.; Anderson, M.; Williams, B. J.; Lauber, M. A.;
Chakraborty, A.; Chen, W. Anal. Chem. 2015, 87, 10283−10291.
(25) Thompson, J. H.; Chung, W. K.; Zhu, M.; Tie, L.; Lu, Y.;
Aboulaich, N.; Strouse, R.; Mo, W. Rapid Commun. Mass Spectrom.
2014, 28, 855−860.
(26) Madsen, J. A.; Farutin, V.; Carbeau, T.; Wudyka, S.; Yin, Y.;
Smith, S.; Anderson, J.; Capila, I. mAbs 2015, 7, 1128−1137.
(27) Bishop, E. Presented at BEBPA HCP Conference, Lisbon,
Portugal, May 17−19, 2016.
(28) Johansen, E. Presented at BEBPA HCP Conference, Lisbon,
Portugal, May 17−19, 2016.
(29) Ecker, D. M.; Jones, S. D.; Levine, H. L. mAbs 2015, 7, 9−14.
(30) Silva, J. C.; Gorenstein, M. V.; Li, G. J.; Vissers, J. P. C.;
Geromanos, S. J. Mol. Cell. Proteomics 2006, 1, 144.

5444 DOI: 10.1021/acs.analchem.7b00304

Anal. Chem. 2017, 89, 5436−5444