Anda di halaman 1dari 7

Profiling the Effects of Process Changes on Residual Host Cell Proteins in

Biotherapeutics by Mass Spectrometry


Matthew R. Schenauer, Gregory C. Flynn, and Andrew M. Goetze
Dept. of Process and Product Development, Amgen Inc., Thousand Oaks, CA
DOI 10.1002/btpr.1748
Published online May 21, 2013 in Wiley Online Library (wileyonlinelibrary.com)

An advanced liquid chromatography=mass spectrometry (MS) platform was used to identify and quantify residual Escherichia coli host cell proteins (HCPs) in the drug substance
(DS) of several peptibodies (Pbs). Significantly different HCP impurity profiles were
observed among different biotherapeutic Pbs as well as one Pb purified via multiple processes. The results can be rationally interpreted in terms of differences among the purification
processes, and demonstrate the power of this technique to sensitively monitor both the quantity and composition of residual HCPs in DS, where these may represent a safety risk to
patients. The breadth of information obtained using MS is compared to traditional multiproduct enzyme-linked immunosorbent assay (ELISA) values for total HCP in the same samples and shows that, in this case, the ELISA failed to detect multiple HCPs. The HCP
composition of two upstream samples was also analyzed and used to demonstrate that HCPs
that carry through purification processes to be detectable in DS are not always among those
that are the most abundant upstream. Compared to ELISA, we demonstrate that MS can provide a more comprehensive, and accurate, characterization of DS HCPs, thereby facilitating
process development as well as more rationally assessing potential safety risks posed by
C 2013 American Institute of Chemical Engineers Biotechnol.
individual, identified HCPs. V
Prog., 29:951957, 2013
Keywords: mass spectrometry, host cell proteins, biotherapeutics

Introduction
Modern recombinant biotherapeutics are typically produced in non-human cell lines. Despite rigorous purification,
low levels of host cell protein (HCP) impurities can remain
in the final drug product (DP). Residual HCPs represent
potential safety risks for patients, including immunogenicity,1 adjuvant activity,2,3 decreased product stability due to
enzymatic activity,4 or, more theoretically, direct biological
activity.5 To reduce such concerns, clearance of HCPs to
levels deemed safe is required by regulatory agencies.6 This
could be considered even more important for todays highdose therapeutics, such as antibody products, which are often
dosed at 100 mg, in contrast to the lower doses used for
first-generation biotherapeutics such as insulin and growth
hormone (10 mg=dose). With high-dose products, a patient
might receive impurity HCPs at levels comparable to the
active protein in first-generation biotherapeutics. Truly meaningful a priori evaluation of potential risks associated with
residual HCPs in DPs requires both identification and quantification of the individual HCPs present, as individual proteins can vary widely with respect to attributes that might
generate safety concerns (e.g., immunogenicity). In addition,
Current Address of Matthew R. Schenauer: 1 DNA Way, South San
Francisco, CA 94080.
Additional Supporting Information may be found in the online version
of this article.
Correspondence concerning this article should be addressed to A.M.
Goetze at goetzeam@amgen.com.
C 2013 American Institute of Chemical Engineers
V

total HCP levels may be less relevant than the amount of


specific, high-risk protein(s). With HCP identification and
individual quantification, modern tools such as in silico or in
vitro prediction of immunogenicity could be used to help
assess safety risks; in addition, over time, clinical experience
with common HCPs may become correlatable with levels of
specific HCPs in DPs.
To date, HCP levels have been most commonly monitored
by multianalyte enzyme-linked immunosorbent assay
(ELISA), using polyclonal antisera raised against large numbers of HCPs present during an upstream process step.7,8
This type of assay provides a single numerical result representing the totality of immunoreactive HCPs and is often
used as a lot release specification test. However, it provides
no HCP identification, and the accuracy of total HCP quantification is questionable as: (a) it is difficult to obtain, and
demonstrate, proportional antibody coverage against all
potential HCPs and (b) the assay standard is unlikely to
match the HCP composition (analyte) of the sample being
analyzed, which is a fundamental requirement for quantitative analytical assays. Consequently, numerical results are
antisera (and cell-line) specific, and, because most major biotherapeutic companies develop their own proprietary antisera, not interchangeable between sponsors. These
uncertainties likely contribute to the lack of a universal HCP
specification target, although with each sponsors unique
ELISA, values between 1 and 100 ppm (w=w) HCP are
often reported for approved products.7 However, in this paradigm, informed risk assessment, based on identification and
951

952

Biotechnol. Prog., 2013, Vol. 29, No. 4

Table 1. Summary of Significant Differences Among Unit Operations


for Pb1 Purification
Process:
1
2
3
4
Relative
Unit Op
Unit Op
Unit Op
Unit Op

Refold Conc:
1:
2:
3:
4:

13
Precipitation
A
B
C

13
A
C

13
A
C
E
D

53
A
C
E
D

AE 5 unique separation modality, e.g., ion-exchange or hydrophobic


interaction chromatography.

reliable quantification of individual HCP components in DPs,


is not possible. Without sharing proprietary HCP ELISA
reagents or manufacturing=purification procedures, companies cannot meaningfully compare the HCP profiles of products, including those of biosimilar and innovator products.
Another significant drawback to lack of HCP identification is
that process improvements to reduce HCP levels must proceed by trial and error, as rational process development that
exploits known differences in physiochemical properties
between identified HCPs and product is not possible.
Mass spectrometry (MS) is now an indispensable tool for
protein characterization as well as for the characterization of
complex protein mixtures such as those used in proteomics
or biomarker discovery. Detecting ppm levels of HCPs in a
biotherapeutic background presents major challenges to the
dynamic range of such methods.9,10 One approach to mitigate this challenge is to couple a high-resolution two-dimensional (2D) liquid chromatography (LC) separation with
high-resolution MS. In one specific embodiment of this
approach, 2D-LC=MSE, hereafter abbreviated as MSE, trypsin-digested samples are chromatographically separated by
2D reversed-phase chromatography using high and low pH
mobile phases in the first and second dimension, respectively, and the peptides are analyzed by MS.11,12 Data-independent acquisition methods such as MSE may offer higher
duty cycles, improved chromatographic peak sampling, and
more reproducible mass spectra compared to traditional datadependent acquisition methods.11,13,14 This approach has
demonstrated the ability to identify, and quantify, individual
HCPs present in biotherapeutic monoclonal antibodies
(mAbs)9,10 and, importantly, to do so in an objective, antiserum-independent manner.
Peptibodies (Pbs) are therapeutics in which bioactive peptides are fused to a human IgG1 Fc for improved stability
and circulatory half-life.15 Pbs are typically expressed in
Escherichia coli as inclusions bodies, which are subsequently
solubilized (including complete reduction of all cysteines),
refolded (with accurate disulfide bond pairing), and purified.
We describe here a retrospective study using 2D-LC=MSE of
the HCP content of one purified peptibody (Pb1) as a function of process changes during development. Our focus was
on the analysis of drug substance (DS), as this contains the
HCPs at the stage at which they could pose a safety risk to
patients. The breadth of information obtained by MSE is contrasted to the single HCP ELISA number and is used to illustrate the significant advantages provided by the MSE
information. The HCP profiles of two other purified Pbs,
each with unique purification schemes, were also determined
and compared with those of Pb1. These results provide
examples of how both major and minor process changes can
affect the HCP profile in Pb DP. In addition, MSE was used
to compare the DS HCPs following extensive purification
with their levels in two upstream samples. The type of

detailed information obtained retrospectively by MSE in this


study, could, in future applications, be used to more efficiently guide process development and, in principle, be used
as a starting point to better evaluate potential HCP-associated safety risks.

Materials and Methods


The therapeutic peptibodies Pb1, Pb2, and Pb3 were produced at Amgen (Thousand Oaks). Pb2 was produced using
an older, nonrepresentative process. Total E. coli HCP by
ELISA was determined using a multianalyte (and multiproduct) assay that was developed from commercial strain-appropriate antisera raised against E. coli lysates and uses
in-house E. coli lysate as assay standard and which was validated according to ICH guidelines. DnaK was specifically
quantified using a polyclonal DnaK-specific ELISA developed in house using commercial DnaK as both immunogen
and assay standard and also fully validated. Identification
and quantification of HCPs by 2D-LC=MSE was performed
essentially as previously described.10 Briefly, reduced, alkylated tryptic protein digests were prepared for all samples.
Chromatography was carried out on a Waters nanoAcquity
ultra-performance LC (UPLC) instrument with 2D technology. XBridge BEH 130 C18 (5 mm, 300 mm 3 50 mm)
Nanoease columns, Symmetry C18 (5 mm, 180 mm 3 20
mm) trap columns, and HSS T3 (1.8 mm, 75 mm 3 150 mm)
analytical columns (Waters Corporation, Milford, MA) were
used in all analyses. The first-dimension chromatography
buffers were 20 mM NH4HCO2, pH 5 10, and acetonitrile,
while second-dimension buffers were H2O and acetonitrile,
respectively, each with 0.1% formic acid. MSE analyses
were carried out on a Synapt (G1) Q-IMS-TOF mass spectrometer (Waters Corporation) operating in TOF V-mode.
The eluate from the D2 column was sampled into the mass
spectrometer via a Z-spray nanosource (with lock mass)
incorporating a universal sprayer and using PicoTip Emitters.
Data were processed using ProteinLynx Global Server Version 2.4 (PLGS, Waters Corporation), and Microsoft Excel.
As described previously, only high confidence (pass 1) peptides were used for quantitative DS analysis, and when high
confidence data on DS HCPs were available from additional
analyses, these data were used to further restrict the quantity-indicating HCP peptides in DS to those ranking tenth or
better in the higher confidence acquisitions.10
Optimal loading targets the HCPs to fall into the instruments dynamic range, often necessitating that the therapeutic be loaded at levels at least 1,000 times above that range.
Changes in total protein loading across products and sample
types were accommodated to facilitate optimal DS HCP
detection. For 10-fraction 2D-LC=MSE DS runs, 716 mg
digested protein were injected per run, while 2430 mg were
loaded in 20-fraction analyses. In all cases, restricted top 3
peptide signals for identified HCPs were externally calibrated
against the average response for 12 standard proteins spiked
at various levels into therapeutic DS in both the 10- and 20fraction modes.10 Pb1, Process 2, Unit Op A (see Table 1)
eluate pool was analyzed with a 10-mg injection in a single
10-fraction run. Null E. coli lysate (containing no therapeutics) was quantitatively analyzed with 390-ng injections in
three replicate 10-fraction runs to assess the 250 most abundant HCPs. One additional, nonquantitative 20-fraction run
was performed on the null material loaded at 7.8 mg to aid
in identification of low-abundance proteins.

Biotechnol. Prog., 2013, Vol. 29, No. 4

953

Figure 1. Venn diagram showing distribution of identified DS HCPs among different Pb products and processes. Superscripts on the
Pb process numbers refer to the number of lots of DS analyzed for each process followed by the total number of MSE
acquisitions of those lots. For brevity, only the top 10 (previously identified) HCPs in Pb 2 are shown.10

Criteria for identification of DS HCPs in this study varied


from our previous study, and the stringency of criteria for
confident identification was somewhat relaxed to reflect that
the method performance had largely been previously characterized10 and the substantial time requirements for excessive
replicate two-dimensional LC=MS analyses of many samples. Criteria for DS HCP identification in this study
required that the HCP be identified in >50% of data acquisitions for a particular process. In cases in which only one set
of data was obtained, the identification threshold was set
with a PLGS score of 500, a value we have previously
found to be reproducible in DS HCP analyses using PLGS
2.4. Since only one analysis was performed on both the Pb1
column pool 1 material and the high-sensitivity null lysate
analysis, the PLGS score 500 criteria were also applied.
For the quantitative MSE null lysate analysis, HCPs must
have been identified in 2 of 3 runs to be considered
identified.

Results and Discussion


Pb 1 HCP profiles
Table 1 provides a summary of the key differences among
four distinct purification processes used during the development of Pb1. Process 1 was the initial purification process
that enabled first-in-human clinical studies; Process 2 was
subsequently developed to facilitate commercial-scale purification. A multiproduct E. coli-specific HCP ELISA showed
little difference in total DS HCP levels between these processes, yielding average values of 3 and 5 ppm (w=w, total
HCP=product) for Process 1 and Process 2 DS, respectively.
In contrast, 2D-LC=MSE (MSE) identified significantly different residual HCP impurity profiles between the same
materials. Two HCPs were identified in Process 1 DS, and
11 HCPs were identified in Process 2 DS (Figure 1); only
one of these (DnaK) was identified in common between both
processes. MSE quantification showed that this process

Table 2. Average Quantification Results for DnaK and Total HCP


by MSE and ELISA for Pb1 Processes
MSE
ELISA
Pb1
Process
DnaK
HCP
DnaK
HCP
1
2
3
4

20
96
<LOD
<LOD

26
391
195
166

8
73
5
2

3
5
1
2

Average quantification results for DnaK and total HCP by MSE and
Total HCP by MSE is estimated as the sum of all identified individual
HCPs. All numbers are in ppm (w=w). LOD 5 limit of detection.

change had resulted in an increase in total detectable HCPs


from 26 to 391 ppm (Table 2).
Levels of one specific HCP, DnaK, increased from 20 to
96 ppm between the lots of Process 1 and 2 DS, as measured by MSE. Identification of DnaK as a specific impurity
was confirmed by in-gel digestion and peptide map fingerprinting of Process 2 DS. DnaK shares 53% sequence identity with human HSP70, and its presence in a biotherapeutic
at measurable levels was deemed a potential risk to induce
anti-human HSP70 antibodies. MSE estimates for levels of
this specific HCP were independently confirmed by DnaK
ELISA. With this ELISA, average values of 8 and 73 ppm
DnaK were obtained for the lots of Process 1 and 2 DS
tested, respectively (Table 2).
Interestingly, the changes from Process 1 to Process 2, initially monitored via the HCP ELISA to preserve a similar
HCP impurity level, resulted in a highly distinct HCP profile
by MSE. DnaK was the only common HCP across the two
processes, and its levels were significantly increased from
Process 1 to Process 2. In this example, the MSE HCP profile of a DS sensitively reflected major process changes,
whereas the HCP ELISA could not. The HCP ELISA values
not only did not reflect the increased amounts of DnaK in
Process 2, but also provided no indication that the HCP

954

Figure 2. Comparison of the MSE-determined concentrations


of the eight most abundant secondary DS HCPs in
Pb1, Process 2 lots with the primary HCP, DnaK.

composition and total HCP level had changed. Such an impurity profile difference could represent an increased safety
risk for patients, although no effort was made to assess this.
In this case, MSE made this information available, whereas
the use of HCP ELISA did not.
Four Process 2 DS lots were analyzed quantitatively by
MSE; they were chosen to represent the extremes in DnaK
levels as measured by ELISA. Nine of the eleven HCPs identified for Pb1, Process 2 in Figure 1 were identified in all
four of these lots, while two were identified in three lots. Figure 2 plots the MSE-determined concentrations of the nine
most abundant Process 2 HCPs against the DnaK ELISAdetermined DnaK concentration, which itself shows very
good quantitative correlation with MSE.10 The other eight
HCPs appear to fall into two classes: those whose concentrations scale in proportion to that of DnaK (MiaB, AsnA, and
NarP) and those whose concentrations are approximately constant, independent of the DnaK levels (HsIU, YhbS, YdhR,
and NfuA). These two groups of proteins did not exhibit
obvious differences in pI or amino acid composition. It is
conceivable that the proteins that scale similarly to DnaK
might be removed from the product with similar efficiency at
the same key chromatographic step(s) as DnaK, whereas the
remainder might be cleared at step(s) that have less impact
on DnaK concentration.
Process 3 was the result of adding a new purification step
to Process 2 (Table 1), with the specific goal of reducing residual DnaK. Following identification of DnaK, differences
in physiochemical properties between DnaK and Pb1 were
readily exploited for DnaK removal with the addition of a
tailored chromatography step. As a result, DnaK levels were
reduced from 73 ppm to 5 ppm by DnaK ELISA and from
96 ppm to below detection by MSE. In contrast, the HCP
ELISA was unable to detect a significant HCP level difference in Process 3 compared to Processes 1 or 2 (Table 2). In
addition to greatly decreasing residual DnaK in Process 3,
the spectrum of identified, non-DnaK HCPs was reduced to
a smaller subset of those identified in Process 2 DP (Figure
1). The latter observation is reasonable since Process 3 differed from Process 2 only by the addition of a single unit
operation. With such a change, the quantity and spectrum of
HCPs might be expected to decrease, but no new HCP(s)

Biotechnol. Prog., 2013, Vol. 29, No. 4

would be likely to be identified, as observed here. MSE


therefore not only confirmed that the targeted reduction in
DnaK levels had been achieved in the augmented purification scheme, but also demonstrated that multiple other HCPs
had been reduced in parallel. The latter is information that
the HCP ELISA failed to provide and that a singly targeted
ELISA is fundamentally incapable of providing.
In a final change to Process 4, the protein concentration
during the refolding step was increased 5-fold over that of
Process 3. As measured by their respective ELISAs, neither
total HCP nor DnaK levels changed significantly as a result.
However, MSE provided more detailed information, demonstrating that the number of detectable HCPs had increased.
Three more HCPs were detected in Process 4 DS compared
to Process 3 DS; two of these were also present in process 2
DS but another (HinT) had not been previously observed in
any Pb1 DS (Figure 1). The fact that Pb1, Process 4 DS was
only analyzed once notwithstanding, one could easily conceive that a five fold higher protein concentration at a key
process step might result in a wider spectrum of residual
HCPs carried though to DS, as observed here.
Pb2 and Pb3
The purified DS of two other peptibodies, Pb2 and Pb3,
were also analyzed by MSE. Forty-three HCPs were confidently identified in Pb2 DS.10 Six HCPs were identified in
Pb3 DS. These results are summarized in Figure 1, which, for
simplicity, shows only the 10 most abundant Pb2 HCPs. In
these Pbs, the presence of DnaK was nearly ubiquitous, being
identified (by MSE) in the Pb1 DS resulting from two of four
purification processes as well as in Pb2 and Pb3 DS, but also
suggested to be present at lower concentrations in the other
two Pb1 processes based on the DnaK ELISA. The protein
PhoP was the only other HCP found in more than one product, being detected in Pb1, Processes 2, 3, and 4, as well as
in Pb3. However, the large majority of HCPs identified in
Pb2 and Pb3 DS were unique to their respective processes,
which suggests that, for the most part, each therapeutic
Pb presents unique HCP clearance challenges, a finding
perhaps not unexpected given the large number of HCPs
initially present, their tremendous diversity with respect
to physicochemical properties, each Pbs distinct physicochemical properties, and the unique purification schemes
used.
Comparison of DS HCPs to those in upstream samples
To gain some quantitative insights into the carry-through
of E. coli HCPs into DS, two upstream samples were analyzed for HCPs by MSE for comparison with HCPs in DS.
The first sample was a null cell lysate of the cell line used
to produce Pb1, Pb2, and Pb3. In the absence of product
expression, the HCP expression profile has been shown to be
highly similar to what would be observed for the production
cell line when producing product.16 Overall, 274 E. coli proteins were identified in at least 2 of 3 10-fraction runs of the
null E. coli lysate, while 1,539 proteins were identified in a
single 20-fraction run loaded with 20-fold more protein. A
second sample consisted of the first column pool of Pb1,
Process 2 (Table 1, Process 2, Unit Op 1 eluate pool). By
HCP ELISA, this pool contained 2,000 ppm total HCP.
MSE identified 154 HCPs in this sample, the 50 most
abundant of which are listed in Supporting Information
Table 1.

Biotechnol. Prog., 2013, Vol. 29, No. 4

955

Table 3. Compilation of all HCPs Identified in Pb1, Pb2 or Pb3 DS Along with Status of Each Proteins Identification in Upstream Samples
All Null Lysate
250 Most Abundant
Unit Op 1
Entry
Identified E coli Protein
MW
ECPs (1539 Total)
Null Lysate ECPs
Pool of Pb1, Pr2
ASNA
CH60
CLPB
DNAK
ERPA
FLMA
FNR
GRCA
HINT
HNS
HSLU
IDH
MIAB
MPRA
NARP
NFUA
NIFU
PFLB
PHOP
PTKB
RIMM
RL3
SUCC
YBEL
YDHR
YHBS

Aspartate ammonia ligase=Asparagine synthetase A


60 kDa chaperonin=groL
Chaperone protein ClpB
Chaperone protein dnaK Heat shock protein 70
Iron sulfur cluster insertion protein erpA
Stable plasmid inheritance protein flmA
Fumarate and nitrate reduction regulatory protein
Autonomous glycyl radical cofactor
HIT like protein hinT
DNA binding protein H NS
ATP dependent protease ATPase subunit HslU
Isocitrate dehydrogenase NADP
Dimethylallyl adenosine tRNA methylthiotransferase miaB
Transcriptional repressor mprA
Nitrate nitrite response regulator protein narP
Fe S biogenesis protein nfuA
NifU like protein
Formate acetyltransferase 1
Transcriptional regulatory protein phoP
Galactitol specific phosphotransferase enzyme IIB
Ribosome maturation factor rimM
50S ribosomal protein L3
Succinyl CoA ligase ADP forming subunit beta
Uncharacterized protein ybeL
Putative monooxygenase ydhR
Uncharacterized N acetyltransferase YhbS

36651
57329
95585
68984
12101
6108
27967
14284
13241
15540
49594
45757
53663
20564
23575
20998
13849
85357
25535
10222
20605
22244
41393
18797
11288
18534

Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y

N
Y23
Y51
Y8
N
N
N
Y240
Y234
Y60
N
Y5
N
N
N
Y171
Y227
Y49
Y224
Y145
Y45
Y95
Y24
N
Y226
N

Y
Y
N
Y
Y
N
N
N
N
N
Y
N
Y
N
Y
N
Y
Y
Y
Y
Y
Y
N
Y
Y
Y

Superscript in column 250 Most Abundant Null gives rank among top 250 identified ECPs.

Table 3 lists a composite of all HCPs identified in Pb1


and Pb3 DS, as well as the top 10 HCPs identified in the significantly less pure Pb2 DS. Each of the identified DS HCPs
was then mapped back as to whether it was identified in either of the two analyzed upstream samples. Overall, 25 of
the 26 DS HCPs were identified in the null cell lysate. However, only 16 of these (62%) ranked among the 250 most
abundant proteins in the null cell lysate, indicating that proteins of relatively low abundance in upstream material may
also persist into the DS. This finding has implications for the
development of multicomponent HCP ELISA antisera, since
a proteins abundance in the null material could impact
whether the protein elicits a sufficient antibody response in
the immunized animal. Nevertheless, these initially lowabundance proteins, as shown here, can persist through purification (and potentially even become concentrated in DS via
copurification). It is worth noting that MS quantifies based
on moles of peptide(s) detected, whereas ELISA results
expressed in ppm are mass-based. With MS, larger HCP proteins may have an inherently higher probability of being
detected, based on generating a larger number of highly ionizable, and detectable, peptides. However, for an equivalent
weight-based HCP level (expressed in ppm), smaller proteins
are expected to be quantified with greater sensitivity, since a
larger molar quantity of peptide(s) will be involved. How
these competing factors play out and potentially influence
the comparison between ELISA and MS-based HCP quantification remains to be determined.
All DS HCPs identified in Pb1, Process 1, 2, or 3 were
also identified in the upstream Pb1 Unit Op 1 eluate pool,
while only one HCP in the Process 4 DS was not identified
in that same column pool. In contrast, six of the ten most
abundant HCPs in Pb2 DS and three of the six DS HCPs in
Pb3 were not detected in the Pb1 SM1 eluate pool, indicating that divergence of HCP profiles may occur very early in
the purification process.

Because lower-abundance upstream HCPs were also


detected in DS it is likely that for many DS HCPs, copurification can be attributed to sharing one or more physiochemical properties with the product, thereby providing relatively
little basis for chromatographic separation during purification. Ion-exchange forms a common modality for the purification of these Pbs and, in this regard, it is striking how, for
the most part, the isoelectric points (pI) of residual DS HCPs
cluster close to, or slightly below, that of their respective
product (Figure 3). Only Pb3, with a pI of 7.91, contained
DS HCPs with pI > 7. One HCP that may copurify due to a
different mechanism is the common residual HCP DnaK, a
molecular chaperone known to bind segments of unfolded
proteins,17 which could be copurifying with the Pb through
direct binding to an unfolded peptide portion of the drug, as
suggested by native gel western blotting (data not shown).
When HCP copurifies due to sharing physiochemical properties with product, increased resolution or greater orthogonality among purification modalities may be most effective in
lowering HCP levels, whereas when copurification occurs
due to binding to product, more stringent wash steps of
resin-bound product could be considered.
Multiproduct HCP ELISAs most commonly use cell
lysates or supernatants as the immunogen, in an effort to
generate the broadest possible antibody coverage.7,18 However, more process-specific strategies, in which the HCP
pool originating from null cells has gone through one or
more mock purification steps, have also been less frequently
used.8,18 Process-specific assays offer the potential for
greater accuracy, as the composition of the assay standard is
likely to more closely match that of downstream samples.
The present results show that HCPs partition to a significant
extent at each stage of the purification process, as early as
the first chromatography step, the conditions for which will
typically vary among Pbs. Even mildly process-specific ELISAs, using HCPs carried through a first (mock) purification

956

Biotechnol. Prog., 2013, Vol. 29, No. 4

Figure 3. pI comparisons of residual DS HCPs among three


Pbs. Red triangles, product; green triangles, DnaK;
black circles, other HCPs. Theoretical pIs were calculated using Protein Calculator 3.3 (www.scripps.
edu=cdputnam=protcalc.html).

step for one Pb, will not generally be applicable to others


that utilize altered conditions for the first purification step
since, at this step, a major portion of potential DS impurity
HCPs may already have been removed. This observation is
made possible by the fact that analytical technologies have
now advanced to the levels of resolution and sensitivity to
allow detection DS HCPs for comparison with upstream
samples, thus facilitating a significantly enhanced understanding of not only our purification processes, but also the
methods we have traditionally relied upon to characterize
them.

Conclusions
This study represents, to our knowledge, only the second
detailed MS characterization of the HCP content of a biotherapeutic DS=DP (following10), and the first that correlates
residual HCP content with changes in the purification process steps used. It demonstrates not only that MS has sufficient sensitivity to identify and quantify multiple residual
HCPs in highly purified Pb biotherapeutics from E. coli, but
also that the composition of residual HCPs in a single therapeutics DS may sensitively reflect purification process
changes. Significant differences in both composition and
quantity of individual HCPs as a result of process changes
were monitored by MSE. The technique sheds light on overall changes in HCP impurity profiles, while quantification of
a specific HCP, DnaK, using this technology was corroborated by a quantitative DnaK-specific ELISA. In contrast, a
multiproduct E. coli-specific HCP ELISA failed to detect
significant changes in DnaK levels as well as major changes

in overall HCP composition and quantities. These observations likely reflect the fact that a single analyte ELISA is capable of good accuracy, whereas quantification with a
multianalyte ELISA, like that for HCP in the present study,
can be problematic and even fail to detect key components
depending on the initial immunoreactivity of HCPs in the
immunogen and complement of HCPs that eventually persist
into DS. If residual HCPs present safety risks, our study
indicates that patient safety may not always be adequately
assured with sole reliance on a traditional HCP ELISA. One
potential lesson is that whereas the biotechnology industry
typically uses multiple high-resolution bioanalytical methods
to analyze the DP itself, this rigor is conventionally not
matched by the assay methodologies used to monitor residual HCP. In our study, a validated HCP ELISA was clearly
unable to detect DnaK, a HCP that presented possible safety
concerns, as well as other HCPs. As demonstrated in the
present study, high-resolution LC=MS methodologies are
now capable of providing more comprehensive, and accurate,
DS HCP characterization, thereby facilitating rational assessment of potential safety risks posed by individual, identified
HCPs. This information can also be used to accelerate and
improve process development by intelligently removing
newly identified HCPs of concern by exploiting each components hitherto unknown physiochemical properties. In conclusion, the deeper understanding of product quality with
regard to HCPs provided by a method such as MSE, not only
addresses the expectations of the Quality by Design initiative,19,20 but also provides a viable path forward in addressing HCP comparability for biosimilars as well as in other
manufacturing changes.

Acknowledgments
The authors would like to acknowledge Ken Chen, Hai
Pan, and Gang Huang for the original identification of DnaK
as an impurity in Pbs, Susan Callahan for DnaK ELISA
development and support, and Amy Hu, Brian Williamson,
and Oliver Kaltenbrunner for process development.

Notation
E

2D-LC=MS = the specific high-resolution two-dimensional


LC separation with high-resolution MS
technique used in this study
DP = drug product
DS = drug substance
ELISA = enzyme-linked immunosorbent assay
HCP = host cell protein
LC = liquid chromatography
mAb = monoclonal antibody
MS = mass spectrometry
MSE = the specific high-resolution 2D-LC
separation with high-resolution MS technique
used in this study
Pb = peptibody
pI = isoelectric point
UPLC = ultra-performance LC

Literature Cited
1. Wadhwa M, Skog AL, Bird C, Ragnhammar P, Lilljefors M,
Gaines-das R, Mellstedt H, Thorpe R. Immunogenicity of

Biotechnol. Prog., 2013, Vol. 29, No. 4

2.
3.

4.

5.
6.
7.
8.
9.

10.
11.

granulocyte-macrophage colony-stimulating factor (GM-CSF)


products in patients undergoing combination therapy with GMCSF. Clin Cancer Res. 1999;5:1353-1361.
Bierich JR. Treatment by hGH of constitutional delay of growth
and adolescence. Acta Paediatr Scand Suppl. 1986;325:7175.
Pavlovic M, Girardin E, Kapetanovic L, Ho K, Trouvin JH.
Similar biological medicinal products containing recombinant
human growth hormone: European regulation. Horm Res. 2008;
69:1421.
Gao SX, Zhang Y, Stansberry-Perkins K, Buko A, Bai S,
Nguyen V, Brader ML. Fragmentation of a highly purified
monoclonal antibody attributed to residual CHO cell protease
activity. Biotechnol Bioeng. 2011;108:977982.
Briggs J, Panfili PR. Quantitation of DNA and protein impurities in biopharmaceuticals. Anal Chem. 1991;63:850859.
ICH. ICH Harmonized Tripartite Guideline. Specifications: Test
Procedures and Acceptance Criteria for Biotechnological=Biological Products Q6B; 1999.
Champion K, Madden H, Dougherty J, Shacter E. Defining your
product profile and maintaining control over it, Part 2. BioProcess Int. 2005;3:5257.
Eaton LC. Host cell contaminant protein assay development for
recombinant biopharmaceuticals. J Chromatogr A. 1995;705:
105114.
Doneanu CE, Xenopoulos A, Fadgen K, Murphy J, Skilton SJ,
Prentice H, Stapels M, Chen W. Analysis of host-cell proteins
in biotherapeutic proteins by comprehensive online two-dimensional liquid chromatography=mass spectrometry. MAbs. 2012;4:
2444.
Schenauer MR, Flynn GC, Goetze AM. Identification and quantification of host cell protein impurities in biotherapeutics using
mass spectrometry. Anal Biochem. 2012;428:150157.
Geromanos SJ, Vissers JP, Silva JC, Dorschel CA, Li GZ, Gorenstein MV, Bateman RH, Langridge JI. The detection, correlation, and comparison of peptide precursor and product ions from

957

12.

13.

14.

15.
16.

17.
18.

19.
20.

data independent LC-MS with data dependant LC-MS=MS. Proteomics. 2009;9:16831695.


Silva JC, Denny R, Dorschel CA, Gorenstein M, Kass IJ, Li
GZ, McKenna T, Nold MJ, Richardson K, Young P, Geromanos
S. Quantitative proteomic analysis by accurate mass retention
time pairs. Anal Chem. 1 2005;77:21872200.
Blackburn K, Mbeunkui F, Mitra SK, Mentzel T, Goshe MB.
Improving protein and proteome coverage through data-independent multiplexed peptide fragmentation. J Proteome Res.
2010;9:36213637.
Plumb RS, Johnson KA, Rainville P, Smith BW, Wilson ID,
Castro-Perez JM, Nicholson JK. UPLC=MS(E); a new approach
for generating molecular fragment information for biomarker
structure elucidation. Rapid Commun Mass Spectrom. 2006;20:
19891994.
Shimamoto G, Gegg C, Boone T, Queva C. Peptibodies: a flexible alternative format to antibodies. MAbs. 2012;4:586591.
Champion KM, Nishihara JC, Aldor IS, Moreno GT, Andersen
D, Stults KL, Vanderlaan M. Comparison of the Escherichia
coli proteomes for recombinant human growth hormone producing and nonproducing fermentations. Proteomics. 2003;3:
13651373.
Bukau B, Horwich AL. The Hsp70 and Hsp60 chaperone
machines. Cell. 1998;92:351366.
Wang X, Hunter AK, Mozier NM. Host cell proteins in biologics development: identification, quantitation and risk assessment.
Biotechnol Bioeng. 2009;103:446458.
FDA. Pharmaceutical cGMPs for the 21st century: a risk-based
approach. Rockville, MD: FDA, September 2004; 2004.
Rathore AS, Winkle H. Quality by design for biopharmaceuticals. Nat Biotechnol. 2009;27:2634.

Manuscript received Mar. 12, 2013, and revision received Apr. 1,


2013.

Anda mungkin juga menyukai