Version of Record: https://www.sciencedirect.

com/science/article/pii/S1570023217314599
Manuscript_5611f1bc476d3832a90379041464bb55

Characterization of 30 therapeutic antibodies and related products by size exclusion

chromatography: feasibility assessment for future mass spectrometry hyphenation

AUTHORS: Alexandre GOYON1, Valentina D’ATRI1, Olivier COLAS2, Szabolcs FEKETE1,

Alain BECK2, Davy GUILLARME1*

1
School of Pharmaceutical Sciences, University of Geneva, University of Lausanne, Centre
Médical Universitaire (CMU), Rue Michel-Servet 1, 1206, Geneva, Switzerland
2
IRPF, Center of Immunology Pierre Fabre, 5 Avenue Napoléon III, BP 60497, 74160 Saint-
Julien-en-Genevois, France

CORRESPONDENCE: Davy GUILLARME

Phone: +41 22 37 934 63
E-mail: davy.guillarme@unige.ch

Highlights
 30 therapeutic mAbs, ADCs, Fc-fusion-proteins and a bsAb were analyzed by
SEC
 Aggregates amounts were comprised between 0.1% and 13.1%
 The level of fragments ranged from 0.1% to 0.9%
 Only acidic mAb products (pI <7) can be analyzed with ammonium acetate
mobile phase

ABSTRACT
Despite the popularity of targeted and immune therapies, the number of studies dealing with
the quantitation of aggregates for Food and Drug Administration (FDA) and European
Medicines Agency (EMA) approved mAb and related products are still very scarce in literature.
In this work, 30 therapeutic proteins including monoclonal antibodies (mAbs), antibody-drug
conjugates (ADCs), Fc-fusion proteins and a bi-specific antibody (bsAb) were investigated
using size exclusion chromatography (SEC). Their levels of high molecular weight species
(HMWS) were experimentally estimated between 0.1% and 13.1%. Except for blinatumomab,
etanercept and pembrolizumab, the HMWS amount for the other antibodies was well below
the limit of 5% usually set a specification for therapeutic mAbs in the biopharmaceutical
industry. The main chromatographic peak shape of 24 therapeutic antibodies and the NIST
mAb [1] was found suitable (0.8 < As < 1.5) with a generic SEC method involving potassium-
based salts mobile phase. Conversely, only acidic therapeutic proteins (pI < 7) could be
successfully analyzed with a mass spectrometry (MS) compatible mobile phase containing 100

1
© 2018 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
mM ammonium acetate. This study aimed to provide HMWS data for 30 therapeutic proteins
covering a wide range of physico-chemical properties with molecular weights between 54 and
153 kDa, pI values comprised between 6.1 and 9.4 and hydrophobic interaction
chromatography (HIC) retention factors ranging from 1.2 to 6.0 for the mAbs.

Keywords:
monoclonal antibodies, antibody-drug conjugates, Fc-fusion proteins, bi-specific antibodies,
size exclusion chromatography, aggregates, SEC-MS

1. INTRODUCTION

Monoclonal antibodies (mAbs) became a therapeutic opportunity thanks to the hybridoma

technology developed by Kohler and Milstein in 1975 [2]. Since the successful
commercialization of rituximab as anti-CD20 treatment for B-non Hodgkin’s lymphoma [3], the
development of therapeutic mAbs experienced a tremendous progress and the FDA has
already approved 75 therapeutic mAbs to date [4]. MAbs targeting cancer cells can induce
immune mediated effects by either antibody-dependent cellular cytotoxicity (ADCC) or
complement-dependent cytotoxicity (CDC), or direct effects by blocking the binding of an
activating ligand that is responsible for the survival of the cancer cell [5]. IgG1 mAb sub-class
stimulates ADCC and/or CDC, whereas IgG2 or IgG4 mAb sub-classes are less effective or
unable to do so [6]. A new class of mAbs called checkpoint inhibitors (CKI) showed spectacular
results in a wide range of cancers, being able of activating immune responses in a non-antigen-
specific manner [7]. Therefore, sales of IgG4 CKI blockbusters mAbs, namely pembrolizumab
and nivolumab, exponentially increased from $83M and $40M in first quarter 2015 to $881M
and $1,195M in second quarter 2017, respectively [8].
Next to the fast growing of mAbs applied in clinical oncology, biopharmaceutical companies
diversified their portfolio and now focus on the development of ADCs and bsAbs with more
than 60 and 30 candidates in clinical trials in 2017, respectively [7,9]. Canonical mAbs such as
trastuzumab or rituximab cannot kill a cancer cell if the targeted antigen are expressed with
less than 100 000 molecules on the tumor cell surface [10]. ADCs were subsequently
developed to kill tumor cells, when associated genes were expressed down to 10 000
molecules, to extend and/or increase the benefit of targeted therapy [10]. Furthermore,
bispecific antibodies were introduced for binding T cells, then driving and plugging them to
cancer cells where associated antigens were expressed from around 1 500 molecules [10]. As
of August 2017, the FDA and EMA approved three ADCs for therapeutic use, namely
brentuximab vedotin, trastuzumab emtansine and inotuzumab ozogamicin, while
blinatumomab and catumaxomab are the only two bsAbs approved to date. A less heralded

2
group of antibody based species comprising proteins or peptides fused with an Fc have also
successfully emerged, and a total of 10 Fc-fusion products have been approved to date [4,11].
The plethora of different antibody formats brought some new challenges for their analytical
characterization. Of particular concern, protein aggregates are one of the most important
critical quality attributes (CQA), since they can generate an immune reaction that inhibit the
efficacy of the product [12] or cause hypersensitivity responses [13].
Size-exclusion chromatography (SEC) is the reference method for the quantitative
measurement of high molecular weight species (HMWS) of recombinant therapeutic mAbs
[14]. In SEC, the separation of HMWS or low molecular weight species (LMWS) from the main
chromatographic peak is achieved based on their difference in hydrodynamic volumes and the
filtration of the different species through a column packed with particles having well-defined
pore sizes [15]. Historical SEC is still routinely performed in the biopharmaceutical industry
with relatively large columns of 7.8 x 300 mm, resulting in 30-40 min analysis time. The
introduction of sub-3 µm UHP-SEC columns of 4.6 x 150 mm greatly reduced the run time by
about 5-folds, allowing analysis times comprised between 5 and 10 min that would significantly
increase the method throughput [16]. However, further reduction of column dimensions would
not be desired, due to the predominant impact of the liquid chromatographic system on the
non-retentive separation mechanism in SEC [17]. In SEC, non-specific chemical interactions
between challenging mAbs or ADCs and the silica-based stationary phase can induce
chromatographic peak tailing and/or adsorption, leading to underestimation of the HMWS
amount [18, 19]. In this context, careful selection of the SEC stationary phase optimization of
the mobile phase is required when analyzing challenging hydrophobic samples (such as
ADCs) or relatively basic mAbs (such as rituximab) [19, 20]. To support the growth of mass
spectrometry (MS) in the biopharmaceutical field and disclose the unambiguous identification
of the separated species, volatile mobile phases have been recently suggested in SEC-MS for
the analysis of reduced and denatured mAbs [21] as well as for intact mAbs [22].
The aim of this present study was to evaluate the suitability of an MS compatible mobile phase
for the measurement of HMWS in SEC for 30 FDA and EMA approved mAbs and mAb-related
products. The mAbs covered a wide range of physico-chemical properties with pI values
comprised between 6.1 and 9.4 and significant differences in hydrophobicity. Although diol
phase remains the most predominantly used silica-surface modifier for SEC [23], the bonding
chemistry of the SEC column used in this study was an alternative one (kept proprietary). This
column packed with sub-3 µm particles in the 4.6 x 150 mm format achieved fast separations
of the HMWS or/and LMWS from the main peak within 6.5 min. The performance obtained with
a mobile phase containing 100 mM ammonium acetate in water was compared to the one
attained with a non-volatile mobile phase containing 300 mM of potassium-based salts.
Representative chromatograms of therapeutic mAbs, ADCs, bsAbs and Fc-fusion proteins are

3
provided for both SEC conditions. In addition to the isoelectric points (pI) [24], relative
hydrophobicities of the different mAb products were determined by HIC. These two features,
namely pI and hydrophobicity, allowed us to better understand and explain differences in SEC
chromatographic profiles and measured HMWS. To the best of our knowledge, this is a very
first study, which provides detailed information on the size variants and SEC chromatographic
behavior of therapeutic proteins possessing such a huge diversity in their hydrophobicity and
isoelectric points. In addition, a clear correlation was found between their hydrophobicity and
size exclusion chromatographic elution time (and peak shape) which can be helpful for method
development.

2. Experimental

2.1 Chemical and reagents

LC grade water was obtained from Fisher Scientific (Reinach, Switzerland). BioUltra grade
potassium chloride, potassium phosphate monobasic, potassium phosphate dibasic,
ammonium sulfate and LC-MS grade ammonium acetate were all purchased from Sigma-
Aldrich (Buchs, Switzerland).
FDA and EMA approved therapeutic antibodies were kindly provided by the Center of
Immunology Pierre Fabre (Saint-Julien en Genevois, France). The NIST mAb reference
Material 8671 was purchased from the National Institute of Standards & Technology
(Gaithersburg, MD, USA). All therapeutic proteins products were re-aliquoted in sterile
conditions under a laminar flow hood. The list of biopharmaceutical proteins, their types,
subclasses and target antigens is provided in Table 1. The mAb pIs can be found in the
literature [24].

2.2 Sample preparation

Samples were systematically analysed in triplicates by SEC using both volatile and non-volatile
mobile phases.
Except blinatumomab, which was formulated at only 11.6 µg/mL, therapeutic proteins were
diluted 5 to 70 times down to a final concentration of 1.0 mg/mL. Water was used as sample
diluent for SEC analysis, while HIC mobile phase “A” containing 1 M of ammonium sulfate and
0.1 M of potassium phosphate buffer in water (pH = 6.8) was employed for diluting samples
for HIC.

2.3 Equipment
Experiments were performed on a Waters ACQUITY UPLC™ H-Class system equipped with
a quaternary solvent delivery pump, an auto-sampler and a fluorescence (FL) detector

4
(excitation at 280 nm, emission at 340 nm, 5 Hz, 0.2-s time constant) and ultraviolet (UV)
detector. The Waters ACQUITY UPLC™ H-Class system includes a flow-through needle
(FTN), a 0.5 μL UV detector flow-cell and a 2 μL FL detector flow-cell.

2.3 Chromatographic conditions

2.3.1 Separation materials
The AdvanceBioSEC column (2.7 μm, 150 mm × 4.6 mm, 300 Å) employed in this study was
kindly provided by Agilent Technologies (Wilmington, DE, USA), whereas the mAbPac HIC-10
column (5.0 µm, 100 × 4.6 mm, 1000 Å) was purchased from Thermo Fisher Scientific AG
(Sunnyvale, CA, USA). Based on our previous studies, the AdvanceBioSEC column was
selected as it was one of the two most effective sub-3 µm column to limit non-specific
interactions between proteins and the stationary phase [16].

2.3.2 Separation conditions

For the separation of antibodies related size variants, the AdvanceBioSEC column was
operated at a flow rate of 0.35 mL/min, and at 25°C. The optimised SEC mobile phase was
prepared with 0.05 M of potassium phosphate buffer + 0.25 M of potassium chloride in water
at pH = 6.8. Indeed, it was previously demonstrated that potassium has the ability to shield
residuals silanolates of the stationary phase better than sodium, and at least 0.2 M of
potassium should be added to limit secondary electrostatic interactions [20]. However, it is
important to keep in mind that a limited amount of phosphate buffer salts (typically comprised
between 50 mM and 200 mM) has to be added to the mobile phase as they also have a strong
salting out effect compared to chloride anions [25]. Based on these observations, the MS
compatible SEC mobile phase was therefore composed of only 100 mM of ammonium acetate
in water (pH = 6.9). In all cases, the injection volume was 0.5 µL, except for blinatumomab (5.0
µL), due to its much lower concentration.
To understand the effect of protein hydrophobicity on SEC elution and support our findings,
relative mAb hydrophobicities were also determined by HIC by using mobile phases prepared
at pH 6.8. The 24 mAb products were injected in alphabetical order and the sequence was
repeated 3 times. Monoclonal antibodies were subsequently ranked based on their HIC
retention factor. HIC mobile phase “A” was composed of 0.1 M of potassium phosphate plus
1 M of ammonium sulfate in water, whereas mobile phase “B” was 0.1 M of potassium
phosphate in water. The mobile phase B composition was increased from 0% to 100% in 20
min during the HIC run. Injection volume was equal to 5.0 µL.

2.5 Software

5
Liquid chromatography instrument control and data analysis were performed by Empower Pro
3 software (Waters). Data treatment was done in Excel (Microsoft).

3. Results and Discussion

3.1 Qualitative evaluation of the chromatographic performance in modern SEC
Historically, non-volatile mobile phases containing phosphate buffers in combination with NaCl
or KCl have been widely used for the analysis of proteins under native conditions in SEC. Such
conditions were required to limit non-specific chemical interactions between the analyzed
proteins and the SEC stationary phase. However, column suppliers have recently made some
significant improvements in terms of column chemistry [16]. Indeed, they have not only
introduced some columns packed with sub-3 µm particles resulting in greater peak efficiency,
but they also drastically improved the chemical inertness of the chromatographic support. As
example, trastuzumab emtansine (a highly hydrophobic ADC [26]) can now be successfully
analyzed with purely aqueous mobile phase using the TSKgel UP-SW3000 support, while in
the past, a non-negligible amount of 2-propanol (15%) had to be added to the mobile phase
with the previous generation of column from the same provider (TSK 3000SW XL column)
[16,18]. Based on this observation, our purpose was to evaluate whether the use of non-volatile
salts in the mobile phase is still required when working with the state of the art UHP-SEC
columns.

3.1.1 SEC profiles using volatile and non-volatile mobile phases

Some representative chromatographic profiles using both non-volatile and volatile mobile
phases are provided in Figures 1 and 2. The chromatographic behavior of six different mAbs
is shown in Figure 1, while some other types of antibodies (ADCs, Fc-fusion proteins and the
bsAb) are presented in Figure 2. Except for the two most acidic biopharmaceutical samples,
namely panitumumab (pI = 6.8) and abatacept (pI = 4.5-5.5), the use of a volatile mobile phase
containing 100 mM of ammonium acetate had always a detrimental effect on the
chromatographic peak shape and on the overall performance of the SEC method. A larger
peak tailing for the monomer and stronger adsorption of the aggregates were observed with
the volatile mobile phase containing only 100 mM ammonium acetate. This was true for all the
other biopharmaceutical products having basic pI (pI >7). In particular, a strong peak tailing
was observed for mAbs containing basic patches, such as adalimumab (asymmetry was higher
than 5 and equal to 1.2 with the non-volatile and volatile mobile phase, respectively) [24]. As
shown in Figures 1 and 2, the HMWS (peaks eluted before the monomer) of adalimumab,
denosumab, pembrolizumab, trastuzumab and trastuzumab emtansine could not be observed
anymore with the ammonium acetate mobile phase, because of their probable adsorption onto
the surface of the stationary phase. Indeed, aggregates are more prone to interact with the

6
stationary phase than the main monomeric species, in reason of the ionic interactions due to
their higher number of charges [27]. Based on these observations, it can be stated that the
nature and the ionic strength of the mobile phase has still a critical effect on SEC performance,
even with a current state of the art UHP-SEC column. In particular, potassium ions have been
shown to interact stronger with silica gel columns than ammonium does [28, 29], therefore they
are expected to better reduce secondary electrostatic interactions. Depending on the quality
of the silica and the polymeric coating (density or surface coverage) used for the synthesis of
the bonded phase, the salt concentration can be subsequently adjusted [30] and the use of
300 mM of potassium-based salts versus 100 mM of ammonium acetate used in our study was
likely to increase their benefit.
On the contrary, the hydrophobic interactions that can also potentially occur between
hydrophobic therapeutic proteins such as bevacizumab and the SEC stationary phase, were
less affected by the nature of the two mobile phases tested in this study. This observation was
in line with our expectations, since only the ionic composition of the mobile phase was modified
between the volatile and non-volatile one, while no organic solvent was added. Then, the
hydrophobic interactions, which could be responsible for strong peak distortion in SEC, should
remain relatively similar between the two tested mobile phase conditions and would be driven
almost exclusively by the intrinsic hydrophobicity of the therapeutic antibodies. To verify our
hypothesis, the mAb hydrophobicity was determined by HIC and the apparent HIC retention
factors (k) were compared for the 24 mAbs (Figure 3). Due to the differences in size of the Fc-
fusion proteins and the bsAb, and the high heterogeneity of ADC products, only mAb products
were injected in HIC to study the effect of protein hydrophobicity on the elution time in SEC.
Significant differences in hydrophobicity were evidenced for the 23 FDA and EMA approved
mAb products, since HIC retention factors range from 1 to 6. Fourteen mAbs had k values
comprised between 1 and 3 and were considered as poorly to moderately hydrophobic ones,
while eight mAbs had k values ranging from 3 to 5 and were classified as moderately to highly
hydrophobic. With HIC retention factors greater than 5, atezolizumab and pembrolizumab were
the two most hydrophobic mAbs. Such very high hydrophobicity was expected with the non-
glycosylated atezolizumab as well as for pembrolizumab for which X-ray structures suggested
that Fc glycans appeared to be buried between the two CH2 domains and have limited
accessibility [31].
Then, SEC elution times were plotted against HIC retention factors (k), with both ammonium
acetate and phosphate mobile phases (Figure 4A and Figure4B, respectively). For the mAbs
of limited hydrophobicity, having HIC k values comprised between 1 and 3.5 (yellow dots in
Figure 4), the elution times in SEC remained constant with the non-volatile mobile phase
(variation of less than 0.04 min for the yellow dots in Figure 4A), whatever the analyzed sample.
A higher deviation in SEC elution times was observed with the volatile mobile phase (variation

7
was equal to 0.12 min for the yellow dots in Figure 4B). This larger variation of elution times
confirmed the presence of stronger ionic interactions described in the previous paragraphs in
SEC with the non-volatile mobile phase. However, for the most hydrophobic mAbs having HIC
k values higher than 3.5 (blue dots in Figure 4), secondary hydrophobic interactions between
the mAb samples and SEC material were evidenced by an increase of elution time in SEC.
Interestingly, this trend followed a general linear relationship with R² of 0.96 and 0.95 using
non-volatile and volatile SEC mobile phases, respectively. There were only two outliers in this
representation, namely ofatumumab and atezolizumab. Their different behavior could be
attributed to additional secondary interactions or differences in conformations. Yang et al. have
already observed such increase of elution/retention time with the highly hydrophobic
pembrolizumab (HIC retention factor of 5.4) compared to other mAbs like natalizumab (HIC
retention factor of 2.2) due secondary hydrophobic interactions [32]. Finally, despite a
significant increase of elution time in SEC, due to hydrophobic interactions, the separation of
the HMWS from the main peak was still achieved for bevacizumab and trastuzumab
emtansine, with the non-volatile mobile phase.

3.1.2 Evaluation of peak asymmetry (As)

To further evaluate the suitability of SEC conditions for aggregates measurement, peak
asymmetries (As) of the main peaks were systematically compared using both mobile phases.
According to the European Pharmacopeia (EP), a suitable As should be comprised between
0.8 and 1.5 for the principal peak [33]. Except for four mAbs, (i.e. atezolizumab, belimumab,
ixekizumab and pembrolizumab), and one ADC sample, (i.e. brentuximab vedotin), which
possess As values between 1.5 and 1.8, As was acceptable for the 26 remaining mAbs and
mAb-related products, when using the non-volatile mobile phase. On the other hand, only five
therapeutic antibodies fit the criteria with the volatile mobile phase (see Table 2). In this case,
the peak tailing was more pronounced, with an average As equal to 2.5 and several As values
higher than 5.
For brentuximab vedotin that had an asymmetry value out of the acceptable range (As of 1.7
and 2.7, with the non-volatile and volatile mobile phases, respectively), the hydrophobicity
differences between the various drug-to-antibody ratio (DAR) species are likely to contribute
to the chromatographic peak broadening and tailing, as they could elute at different elution
times based on previous observations (see Figure 4, describing the change in SEC elution
time as a function of mAb hydrophobicity). For atezolizumab, ixekizumab and pembrolizumab
having HIC k values greater than 4 and belimumab (HIC retention factor of 2.8), secondary
hydrophobic interactions and/or partial unfolding of the main species could explain their
atypical behavior with the non-volatile mobile phase, compared to the other mAb products.
With these challenging products, the addition of a small amount of organic modifier in the

8
mobile phase, such as 2-propanol or acetonitrile (0-10%), may reduce the chromatographic
peak tailing, but will also decrease column lifetime (based on our experience) and increase the
risk of product denaturation.

3.1.3 Achieved resolution (Rs) between the main peak and HMWS / LMWS
Finally, a baseline resolution (Rs > 1.5) between HMWS and the main peak, together with an
As for the main peak comprised between 0.8 and 1.5 was obtained for 19 therapeutic proteins
with the non-volatile mobile phase, whereas only five of them meet these two criteria with the
ammonium acetate mobile phase (Table 2). With the MS compatible mobile phase, suitable
As and baseline resolution were achieved only for three acidic mAb (eculizumab,
pantitumumab, reslizumab) and two acidic Fc-fusion proteins (abatacept, belatacept). In
addition to these five acidic antibodies, fourteen basic therapeutic proteins could be analysed
with the mobile phase containing potassium salts, including the hydrophobic ADC trastuzumab
emtansine and the bsAb blinatumomab. Among the twelve remaining antibodies for which
HMWS could not be adequately measured with suitable As and Rs, four proteins with an As >
1.5 were already identified in section 3.1.2, while eight other therapeutic antibodies had a
suitable As but no baseline resolution achieved between the main peak and HMWS. This was
the case for cetuximab, ipilimumab, the NIST mAb, ofatumumab, pembrolizumab,
ramucirumab, rituximab and etanercept. For these mAbs and the Fc-fusion protein, it has to
be noted that not only aggregates can be separated from the main peak, but also in some
cases, some partially denatured monomers [27]. In fact, partially denatured monomers are
generally described as pre-monomeric peaks, due to their low resolution to the main peak. In
another on-going study (data not shown), peaks that eluted before the main peak of
pembrolizumab were indeed identified as denatured monomers by two-dimensional liquid
chromatography hyphenated to MS (2D-LC/MS). Finally, separation of LMWS from the main
peak could only be achieved when using the non-volatile mobile phase (lower asymmetry
compared to the volatile mobile phase), with Rs comprised between 3 and 3.6 for eight mAbs
and equal to 1.9 for nivolumab. Therefore, these two types of fragments (LMWS and HMWS)
can potentially be separated in SEC.

3.2 Quantitation of the aggregates and fragments of mAb products

When the As factor was unsuitable, due to the presence of non-specific interactions, reliable
HMWS quantitation could not be expected. As example, for an As factor of 1.6, only 0.6% of
HMWS was measured for natalizumab (pI = 7.8) with the non-volatile mobile phase vs. 1.4%
with the non-volatile mobile phase (As of 1.2).
Figure 5 shows the mean percentage and minimum, maximum amounts of HMWS relative to
the total peak area estimated from the SEC measurements (triplicate analysis), for the four

9
class of antibodies (i.e. mAbs, fusion proteins, ADCs and bsAbs). The full and open bars
correspond to the case of non-volatile and volatile mobile phases in SEC, respectively. Due to
its poor performance, only five values were provided for the SEC condition involving a volatile
mobile phase. With the non-volatile mobile phase, the SEC measurements were found to be
unreliable for four samples (atezolizumab, belimumab, ixekizumab and brentuximab vedotin),
due to a strong tailing of the main peak (As > 1.5). Therefore, these four values were not
reported in figure 5. It is worth mentioning that non-specific interactions for these four
therapeutic proteins would be probably reduced by adding organic modifiers in the mobile
phase, while ensuring non-denaturing analysis conditions. Finally, an asterisk highlights some
tricky samples, namely therapeutic products having a suitable As value (below 1.5), but not a
baseline resolution between HMWS and the main peak. In accordance with the United States
Pharmacopeia (USP) monograph <129> released in 2016 [14], species eluting before the main
peak were systematically considered as HMWS (main percentages reported in Figure 5), but
partially denatured monomers might also be quantified at the same time [27]. The limit of
quantitation for the HMWS and LMWS relative to the total peak area was equal to 0.1%.
With the mobile phase containing 100 mM of ammonium acetate, HMWS could only be
measured for the three acidic mAbs, namely eculizumab (pI = 6.0), panitumumab (pI = 6.8)
and reslizumab (pI = 7.1) and the levels were similar to those obtained with the phosphate-
based mobile phase (see Figure 5). With the non-volatile mobile phase, the amount of HMWS
for 21 mAb products was estimated between 0.1% and 8.8% (Figure 5). The values were in
good agreement with those already reported in the literature for bevacizumab (2.5% vs 3.0%),
cetuximab (0.3% vs 0.3%), rituximab (0.7% vs 0.9%) and trastuzumab (0.4% vs 0.4%), except
infliximab (0.9% vs 0.1%) [34]. Pembrolizumab (HMWS = 8.8%) was the only mAb with a
relative HMWS amount greater than 5% that would be consistent with the fact that hydrophobic
proteins tend to aggregate [35], but as discussed in section 3.1.3, partially denatured
monomers were also probably quantified.
As for the mAbs, only the two acidic Fc-fusion proteins, namely abatacept (pI = 4.5-5.5) and
belatacept (pI = 4.5-5.5), were suitable for the SEC analysis by using the volatile mobile phase,
and the measured content of HMWS was similar to those one obtained with the non-volatile
mobile phase (see Figure 5). With the non-volatile mobile phase, the levels of aggregates for
six Fc-fusion proteins, ADC and bsAb were comprised between 0.7% and 13.2% (Figure 5).
Critical amount of HMWS greater than 5% were measured for blinatumomab (13.2%) and
etanercept (5.1%). Such high levels were expected with the bsAb blinatumomab, as it is a
divalent dimer that comprised two variable single chain fragments (scFv) [36] that tend to easily
denature and aggregate [37]. Finally, the increase of HMWS content for trastuzumab
emtansine (2.7%) compared to trastuzumab (0.4%) confirmed that this ADC obtained through
the mAb conjugation with the maytansinoid emtansine (DM1) drug is logically more prone to

10
aggregation than the naked mAb, due to enhanced lipophilicity induced by the presence of the
lipophilic DM1 payload [38].
Figure 6 shows the mean of relative percentages and minimum, maximum values (for three
replicates) of LMWS for the tested mAb and mAb-related products. Contrary to the
measurements of HMWS, LMWS can also be quantified by capillary gel electrophoresis (CGE)
[14] or reversed phase chromatography (RPLC). Among the 31 antibody products tested in
this study, fragments were found for only 15 mAb products as well as for three other type of
therapeutic proteins and their levels were comprised between 0.1% and only 0.9%, when using
the non-volatile mobile phase. LMWS were unlikely to be observed with the ammonium acetate
mobile phase, due the strong tailing often observed for the main peak, which masks the
presence of LMWS eluted after the main peak. Except reslizumab, none of the other five acidic
therapeutic antibodies contained LMWS, although best SEC performance are expected with
these proteins, due to the absence of secondary electrostatic interactions. Therefore, it seems
that acidic proteins are probably less prone to fragmentation than basic proteins.

4. Conclusion
The potential of a recent UHP-SEC material packed with sub-3 µm particles was thoroughly
evaluated for a wide range of therapeutic proteins using both non-volatile (phosphate salts)
and volatile (ammonium acetate) mobile phase conditions. These mAb- and related-mAb
products covered a wide range of physico-chemical properties with molecular weights between
54 and 153 kDa, pI values comprised between 6.1 and 9.4 and HIC retention factors ranging
from 1.2 to 6.0. The HMWS separation from the main chromatographic peak was
systematically achieved within only 6.5 min by using a state of the art 4.6 x 150 mm UHP-SEC
column packed with sub-3µm particles.
Only five acidic therapeutic antibodies (pI < 7) could be analyzed in SEC using an ammonium
acetate mobile phase, by obtaining suitable As factor (As < 1.5) and baseline resolution
between HMWS and the main chromatographic peak (Rs > 1.5). The use of a potassium-
based mobile phase was always required when analyzing basic antibody products (pI > 7).
Such mobile phase conditions allowed the successful and reliable measurement of HMWS for
19 therapeutic proteins, including the challenging ADC trastuzumab emtansine, without adding
organic modifier in the mobile phase. This study confirms that potassium phosphate buffer and
KCl are more effective to limit secondary electrostatic interactions than ammonium acetate
buffers and therefore the use of an MS-compatible mobile phase is restricted to handful
therapeutic proteins with pI < 7. Secondary hydrophobic interactions were critical for four
additional antibodies (As > 1.5) and might require the addition of organic modifiers in the mobile
phase for accurately measuring the amount of HMWS and LMWS.

11
Finally, for the eight remaining antibody products having Rs < 1.5 where no non-specific
interactions were evidenced (As < 1.5), multiangle light scattering (MALS) detector or MS
identification would be required to confirm whether species eluting before the main peak
correspond to aggregates or partially denatured monomers.

5. Acknowledgements
The authors wish to thank Liz Bevan and Tony Edge from Agilent Technologies, for providing
the AdvanceBioSEC column used in this study.
Davy Guillarme wishes to thank the Swiss National Science Foundation for support through a
fellowship to Szabolcs Fekete (31003A 159494). Jean-Luc Veuthey from the University of
Geneva is also acknowledged for useful comments and discussions.

12
References
[1] J.E. Schiel, D. L. Davis, O. V. Borisov, (eds.), State-of-the-Art and Emerging Technologies
for Therapeutic Monoclonal Antibody Characterization Volume 2. Biopharmaceutical
Characterization: The NISTmAb Case Study, American Chemical Society (2015)
[2] G. Kohler, C. Milstein, Continuous cultures of fused cells secreting antibody of predefined
specificity, Nature, 256 (1975), pp. 495-497.
[3] M.E. Reff, C. Carner, K.S. Chambers, P.C. Chinn, J.E. Leonard, R. Raab, R.A. Newman,
N. Hanna, D.R. Anderson, Depletion of B cells in vivo by a chimeric mouse human monoclonal
antibody to CD20, Blood, 83 (1994), pp. 435-445.
[4] FDA. CDER therapeutic biologic products. CDER billable biologic product list. Available at
https://www.fda.gov/drugs/developmentapprovalprocess/druginnovation/default.htm
Accessed 7 August 2017.
[5] G.J. Weiner, Building better monoclonal antibody-based therapeutics, Nat. Rev. Cancer 15
(2015) 361–370.
[6] S. E. Strome, E. A. Sausville and D. Mann, A Mechanistic Perspective of Monoclonal
Antibodies in Cancer Therapy Beyond Target-Related Effects, The Oncologist 12(9) (2007)
1084-1095.
[7] A. Beck, L. Goetsch, C. Dumontet & N. Corvaïa, Strategies and challenges for the
next generation of antibody–drug conjugates, Nature Reviews Drug Discovery 16, (2017) 315–
337.
[8] Company quarterly reports. Available at http://investors.merck.com/financials/quarterly-
reports/default.aspx and https://www.bms.com/investors/financial-reporting/quarterly-
results.html. Accessed 30 July 2017.
[9] X. Zhang, Y. Yang, D. Fan, and D. Xiong, The development of bispecific antibodies and
their applications in tumor immune escape, Exp Hematol Oncol 6(12) 2017.
[10] A. Choulika, Humankind vs. Cancer: The Scorecard (2017) Milken Institute. Retrieved
from https://youtu.be/2Ru1VLGDlHk?list=PLwJK8JzK8C_fT1czQseSmwjIGWEeE7TgL
[11] R. Jafari, N.M. Zolbanin, H. Rafatpanah, J. Majidi, T. Kazemi, Fc-fusion Proteins in
Therapy: An Updated View, .Curr Med Chem. 24(12) (2017) 1228-1237.
[12] R. Franco, G. Daniela, M. Fabrizio, G. Ilaria, H. Detlev, Influence of osmolarity and pH
increase to achieve a reduction of monoclonal antibodies aggregates in a production process
Cytotechnology, 29 (1) (1999), 11-25.
[13] A.S. Rosenberg, Effects of protein aggregates: an immunologic perspective, AAPS J., 8
(3) (2006), 501-507.
[14] Monograph <129> Analytical Procedures for Recombinant Therapeutic Monoclonal
antibodies, USP-NF PF39(3).

13
[15] S. Fekete, A. Beck, J.L. Veuthey, D. Guillarme, Theory and practice of size exclusion
chromatography for the analysis of protein aggregates, J. Pharm. Biomed. Anal., 101 (2014),
43-55.
[16] A. Goyon, A. Beck, O. Colas, K. Sandra, D. Guillarme, S. Fekete, Separation of protein
biopharmaceutical aggregates using size exclusion chromatography columns packed with sub-
3 μm particles, J. Chromatogr. A, 1498 (2016), 80-89.
[17] A. Goyon, D. Guillarme, S. Fekete, The importance of system band broadening in modern
size exclusion chromatography, J. Pharm. Biomed. Anal., 135 (2017), 50-60.
[18] A. Wakankar, Y. Chen, Y. Gokarn, F.S. Jacobson, Analytical methods for physicochemical
characterization of antibody drug conjugates, mAbs, 3 (2011), 161-172.
[19] T. Arakawa, D. Ejima, T. Li, J.S. Philo, The critical role of mobile phase composition in
size exclusion chromatography of protein pharmaceuticals, J. Pharm. Sci., 99 (2010), pp.
1674-1692.
[20 A. Goyon, A. Beck, J.L. Veuthey, D. Guillarme, S. Fekete, Comprehensive study on the
effects of sodium and potassium additives in size exclusion chromatographic separations of
protein biopharmaceuticals, J. Pharm. Biomed. Anal. (2016) in press. doi:
10.1016/j.jpba.2016.09.031.
[21] H. Liu, G. Gaza-Bulseco, C. Chumsae, Analysis of Reduced Monoclonal Antibodies Using
Size Exclusion Chromatography Coupled with Mass Spectrometry, J. Am. Soc. Mass
Spectrom. 20 (12) (2009) 2258-2264.
[22] M. Haberger, M. Leiss, A.-K. Heidenreich, O. Pester, G. Hafenmair, M. Hook, Rapid
characterization of biotherapeutic proteins by size-exclusion chromatography coupled to native
mass spectrometry, mAbs, 8(2) (2016) 331-339.
[23] E.S.P.Bouvier, S.M.Koza, Advances in size-exclusion separations of proteins and
polymers by UHPLC, Trends Anal. Chem. 63 (2014) 85-94.
[24] A. Goyon, M. Excoffier, M. C. Janin Bussat, B. Bobaly, S. Fekete, A. Beck & D. Guillarme,
Determination of isoelectric points and relative charge variants of 23 commercial therapeutic
monoclonal antibodies, [submitted].
[25] F. Hofmeister, Zur Lehre von der Wirkung der Salze (about the science of the effect of
salts), Arch. Exp. Pathol. Pharmakol., 24 (1888), pp. 247-260.
[26] A. Beck, G. Terral, F. Debaene, E. Wagner-Rousset, J. Marcoux, M.C. Janin-Bussat, O.
Colas, A. Van Dorsselaer, S. Cianférani, Cutting-edge mass spectrometry methods for the
multi-level structural characterization of antibody-drug conjugates, Expert Rev Proteomics.
13(2) (2016) 157-83.
[27] J.S. Philo, Is any measurement method optimal for all aggregate sizes and types? AAPS
J. 8(3) (2006) 564-571.

14
[28] K. Ohta and K. Tanaka, Ion chromatographic separation of common mono- and divalent
cations on an unmodified silica gel column by elution with oxalic acid containing crown ethers,
Analyst 124 (1999) 505–510.
[29] T. IWACHIDO, T. IKEDA and M. ZENKI, Determination of Ammonium and Other and Rain
Water by Ion Chromatography Ion Exchanger Major Using Cations in River Silica Gel as an
Ion Exchanger, Anal. Sci. 6 (1990) 593-597.
[30] J. Nawrocki, The silanol group and its role in liquid chromatography, J. Chromatogr. A,
779 (1997) 29-71.
[31] G. Scapin, X. Yang, W. W. Prosise, M. McCoy, P. Reichert, J. M Johnston, R. S. Kashi &
C. Strickland, Structure of full-length human anti-PD1 therapeutic IgG4 antibody
pembrolizumab, Nat. Struct. Mol. Biol. 22(12) (2015) 953-960.
[32] X. Yang, Y. Zhang, F. Wang, L.J. Wang, D. Richardson, M. Shameem, A. Ambrogelly ,
Analysis and purification of IgG4 bispecific antibodies by a mixed-mode chromatography, Anal.
Biochem. 484 (2015) 173-179.
[33] Chapter 2.2.46, Chromatographic Separation Techniques, European Pharmacopeia, 6th
ed. (6.0, Vol 1), 2008.
[34] R. Moore, A. Farrell, J. Bones, and K. Cook, Optimizing protein aggregate analysis by size
exclusion chromatography, poster note 72229, Thermo Scientific.
[35] A. L. Fink, Protein aggregation: folding aggregates, inclusion bodies and amyloid, Folding
and Design 3 (1) (1998) R9-R23.
[36] U. Brinkmann & R. E. Kontermann, The making of bispecific antibodies, mAbs 9 (2) 2017
182-212.
[37] A. Wörn, A. Plückthun, Stability engineering of antibody single-chain Fv fragments, J. Mol.
Biol. 305 (5) (2001) 989-1010.
[38] A. A. Wakankar, M. B. Feeney, J. Rivera†, Y. Chen, M. Kim, V. K. Sharma, and Y. J.
Wang, Physicochemical Stability of the Antibody−Drug Conjugate Trastuzumab-DM1:
Changes due to Modification and Conjugation Processes, Bioconjugate Chem. 21(19) (2010)
1588–1595.

15
Figure captions

Figure 1. Size variant profiles obtained in SEC for 6 representative mAbs by using both non-
volatile and volatile mobile phases. Experimental conditions described in section 2.3.

Figure 2. Size variants profiles obtained in SEC for 6 representative mAb-related products
sucha as Fc-fusion proteins, ADCs and bsAb by using both non-volatile and volatile mobile
phases. Experimental conditions described in section 2.3.

Figure 3. Experimental hydrophobicity for 24 mAbs determined by HIC. Data bars filled in
yellow, red, light blue and dark blue correspond to IgG2/4, IgG2, IgG4 and IgG1 subclasses,
respectively.

Figure 4. Plots of the elution time of the main peak with the non-volatile SEC mobile phase (A)
and the volatile mobile phase (B) vs. experimental HIC retention factors. Experimental
conditions described in section 2.3.

Figure 5. Amounts of HMWS (expressed in % of the main peak) determined by SEC with both
non-volatile and volatile mobile phases, for 27 different mAb and mAb-related samples. Data
bars filled in yellow, red, light blue and dark blue correspond to IgG2/4, IgG2, IgG4 and IgG1
mAb subclasses, respectively. Data bars filled in black correspond to the Fc-fusion proteins,
ADCs and bsAb products. Open and full bars correspond to the values obtained with the
volatile and non-volatile mobile phases, respectively.

Figure 6. Comparison of the LMWS amounts measured by SEC with both non-volatile and
volatile mobile phases for mAbs samples (data bars filled in yellow, red, light blue and dark
blue for IgG2/4, IgG2, IgG4 and IgG1 mAb isotypes, respectively) and Fc-fusion proteins,
ADCs and bsAb products (data bars filled in black).

16
Figure 1

Figure 2.

17
18
Figure 3.

Relative standard deviations were lower than 5%.

19
Figure 4.

20
Figure 5.

21
Figure 6.

22
Table captions
Table 1. List of the 31 commercial mAb products, their type, subclass and therapeutic target.
Table 1.
Subclass
Brand name Type Target
/ Format
Adalimumab Humira hu IgG1 TNF
Atezolizumab Tecentriq hz IgG1 PD-L1
Belimumab Benlysta hu IgG1 BLyS
Bevacizumab Avastin hz IgG1 VEGF
Cetuximab Erbitux ch IgG1 EGFR
Denosumab Prolia hu IgG2 RANK-L
Eculizumab Soliris hz IgG2/4 C5
Elotuzumab Empliciti hz IgG1 SLAMF7
Infliximab Remicade ch IgG1 TNF
Ipilimumab Yervoy hu IgG1 CTLA-4
Ixekizumab Taltz hz IgG4 IL-17a
Natalizumab Tysabri hz IgG4 a4 integrin
NISTmab NA hz IgG1 NA
mAbs

Nivolumab Opdivo hu IgG4 PD1

Obinutuzuma Gazyvaro hz IgG1 CD20
b
Ofatumumab Arzerra hu IgG1 CD20
Palivizumab Synagis hz IgG1 RSV
Panitumumab Vectibix hu IgG2 EGFR
Pembrolizum Keytruda hz IgG4 PD1
ab
Pertuzumab Perjeta hz IgG1 HER2
Ramuciruma Cyramza hu IgG1 VEGFR2
b
Reslizumab Cinqair hz IgG4 IL-5
Rituximab Rituxan ch IgG1 CD20
Trastuzumab Herceptin hz IgG1 HER2
Abatacept Orencia hu Fc-prot CD80/CD8
6
Fc-fusion
proteins

Belatacept Nulojix hu Fc-pro CD80/CD8

6
Etanercept Enbrel hu Fc-pro TNF
Ziv- Zaltrap hu Fc-pro VEGF/PIG
aflibercept F
Brentuximab
ADCs

Adcetris ch IgG1 CD30

vedotin
Trastuzumab
Kadcyla hz IgG1 HER2
emtansine
bAb

Blinatumoma scFv- CD19 x

Blincyto mo
b scFv CD3

Type: human (hu), humanized (hz), chimeric (ch), mouse (mo)

Format: Bispecific T-cell engager (BiTE)

23
 Table 2. Comparison of the main peak elution time, Asymmetry factor, resolution between the
main chromatographic peak and the HMWS or/and LMWS obtained by SEC using both non-
volatile and volatile mobile phases.
Table 2.
Non-volatile mobile phase Volatile mobile phase
Rs Rs Rs
Rs
tE (min) As (HMW tE (min) As (HMW (LMW
(LMWS)
S) S) S)
Adalimumab 3.24 1.6 3.31
1.2 (3) 3.6 (7)
(0.0) (21) (0.7)
Atezolizumab 3.52 1.8 3.42 2.6
(0.5) (15) (0.1) (18)
Belimumab 3.28 1.7 3.32 5.2
1.3 (6)
(0.1) (17) (0.2) (13)
Bevacizumab 3.48 3.44 2.1 1.7
1.2 (1) 1.8 (3)
(0.2) (0.1) (15) (17)
Cetuximab 3.22 3.22 1.9
1.2 (0) 1.4 (7) 3.5 (2.9)
(0.1) (0.1) (12)
Denosumab 3.27
1.2 (1) 1.6 (9) 3.3 (0.1) 2.3 (6)
(0.1)
Eculizumab 3.28 1.9 3.25 1.7
1.2 (0) 1.2 (1)
(0.2) (10) (0.3) (13)
Elotuzumab 3.28 3.27
1.2 (0) 1.7 (2) 1.7 (1) 1.7 (1)
(0.1) (0.0)
Infliximab 3.28 3.28 2.2
1.3 (3) 1.9 (9) 2.0 (3)
(0.1) (0.1) (15)
Ipilimumab 3.45 3.44 4.1
1.3 (9)
mAbs

(0.1) (0.5) (23)

Ixekizumab 3.40 1.5 3.44 5.0
(0.0) (12) (0.6) (16)
Natalizumab 3.28 3.29
1.2 (0) 2.0 (3) 1.6 (5) 1.8 (1)
(0.1) (0.4)
NISTmab 3.26 3.34 5.7
1.3 (3)
(0.0) (0.9) (22)
Nivolumab 3.25 3.28 1.9
1.2 (0) 1.9 (1) 1.9 (5.6)
(0.1) (0.9) (12)
Obinutuzuma 3.26 3.28 2.5
1.2 (1) 1.8 (5) 3.3 (5)
b (0.1) (0.3) (13)
Ofatumumab 3.46 3.43 2.2
1.1 (0) 3.2 (0.6)
(0.0) (0.6) (13)
Palivizumab 3.24 3.27 2.8
1.2 (0) 1.7 (2) 3.6 (3)
(0.1) (0.7) (17)
Panitumumab 3.26 3.25
1.2 (0) 1.8 (0) 1.2 (1) 1.8 (2)
(0.1) (0.1)
Pembrolizum 3.83 3.68
ab (0.1) (0.4)
Pertuzumab 3.28 3.29
1.2 (0) 1.8 (2) 3.6 (0.3) 1.9 (7)
(0.0) (0.4)

24
 Ramuciruma 3.27 3.1 3.30 2.7
1.2 (0)
b (0.0) (10.4) (0.6) (16)
Reslizumab 3.27 3.26
1.2 (0) 1.6 (8) 1.4 (8) 1.6 (4)
(0.1) (0.1)
Rituximab 3.36 3.38 3.4
1.3 (2) 3 (10.2)
(0.0) (0.8) (20)
Trastuzumab 3.26 3.28
1.2 (0) 1.8 (0) 2.1 (9)
(0.1) (0.3)
Abatacept 3.38 3.34
1.1 (0) 1.9 (2) 1.1 (0) 1.9 (3)
(0.0) (0.0)
Fc-fusion

Belatacept 3.39 3.35

proteins

1.1 (0) 2.1 (5) 1.1 (1) 1.9 (2)

(0.0) (0.0)
Etanercept 2.79 2.79
1.3 (0) 1.3 (0) 1.5 (7) 1.1 (2)
(0.0) (0.0)
Ziv- 3.12 3.12 2.2
1.4 (0) 1.5 (2) 1.6 (0)
aflibercept (0.0) (0.1) (17)
Brentuximab 3.25 3.22
1.7 (1) 1.4 (1) 2.7 (6) 1.5 (2)
ADCs

vedotin (0.0) (0.1)

Trastuzumab 3.32 3.34 4.6
1.4 (1) 1.5 (2)
emtansine (0.0) (0.3) (15)
bAb

Blinatumoma 3.82 1.2 3.84

1.6 (3) 2.8 (9) 1.6 (6)
b (0.0) (0.5) (0.1)

Elution time (tE), chromatographic peak resolution (Rs), peak symmetry factor (As)
Relative standard deviations in percentage are in bracket

25