Journal of Chromatography A, 1465 (2016) 117–125

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Size distribution analysis of influenza virus particles using size

exclusion chromatography
Judith Vajda a,∗ , Dennis Weber b , Dominik Brekel b , Boris Hundt c , Egbert Müller a
a
Tosoh Bioscience GmbH, Im Leuschnerpark 4, 64347 Griesheim, Germany
b
Karlsruher Institut für Technologie, Institut für Bio- und Lebensmitteltechnik, Engler-Bunte Ring 3, 76131 Karlsruhe, Germany
c
IDT Biologika GmbH, Am Pharmapark, 06861 Dessau-Roßlau, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Size exclusion chromatography is a standard method in quality control of biopharmaceutical proteins.
Received 6 June 2016 In contrast, vaccine analysis is often based on activity assays. The hemagglutination assay is a widely
Received in revised form 23 August 2016 accepted influenza quantification method, providing no insight in the size distribution of virus par-
Accepted 25 August 2016
ticles. Capabilities of size exclusion chromatography to complement the hemagglutination assay are
Available online 26 August 2016
investigated. The presented method is comparatively robust regarding different buffer systems, ionic
strength and additive concentrations. Addition of 200 mM arginine or sodium chloride is necessary to
Keywords:
obtain complete virus particle recovery. 0.5 and 1.0 M arginine increase the hydrodynamic radius of the
Analytical size exclusion chromatography
Pandemic H1N1
whole virus particles by 5 nm. Sodium citrate induces virus particle aggregation. Results are confirmed by
Hemagglutination assay dynamic light scattering. Retention of a H1N1v strain correlates with DNA contents between 5 ng/mL and
Fragmentation 670 ng/mL. Quantitative elution of the virus preparations is verified on basis of hemagglutination activ-
Aggregation ity. Elution of hemagglutination inducing compounds starts at a flow channel diameter of 7000 nm. The
Process analytical techniques universal applicability is demonstrated with three different influenza virus samples, including an indus-
trially produced, pandemic vaccine strain. Size distribution of the pandemic H1N1v 5258, H1N1 PR/8/34,
and H3N2 Aichi/2/68 preparations spreads across inter- and intra-particle volume and extends to the
secondary interaction dominated range. Thus, virus particle debris seems to induce hemagglutination.
Fragments generated by 0.5% TritonTM X-100 treatment increase overall hemagglutination activity.
© 2016 The Author(s). Published by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction Currently, HA quantification mostly relies on two methods: the

Vaccination is an effective cure against the pandemic outbreak
of influenza. Various influenza vaccines from different manufactur-
ers are based on inactivated whole virus particles. Influenza virus
particles have a size of 80–120 nm. Hemagglutinin (HA) and neu-
raminidase are the two most abundant proteins on the surface of
Influenza A virus particles. Currently, 17 different HA and 10 differ-
ent neuraminidases have been characterized and give name to the
different influenza A subtypes [1]. The amount of HA in an influenza
vaccine is a crucial parameter for release of the biopharmaceutical.
Guidelines by the European Pharmacopoeia require 15 ␮g of HA
antigen for each of the present virus strains in a trivalent human
vaccine [2].
∗ Corresponding author.
E-mail addresses: judith.vajda@tosoh.com (J. Vajda),
dennis.weber2@student.kit.edu (D. Weber), uvdbe@student.kit.edu (D. Brekel),
Boris.Hundt@idt-biologika.de (B. Hundt), Egbert.Mueller@tosoh.com (E. Müller).
single radial immunodiffusion (SRID) assay and the HA assay. SRID
is based on antigen binding and is the gold standard for release
of human vaccines [3]. The HA assay utilizes hemagglutination
of erythrocytes [4]. Dilutions of the investigated virus sample are
incubated with a chicken erythrocyte preparation. In case sufficient
virus particles are present to induce hemagglutination, erythro-
cytes will float carpet-like in solution. Virus particle concentrations
are estimated assuming hemagglutination requires one virus par-
ticle per erythrocyte. This assay dates back to the 1940s and is
currently still accepted in research, development, quality con-
trol testing and the release of animal vaccines. Drawbacks of this
method are the lack of standardization, its discrete format, devi-
ations due to erythrocyte aging [5], secondary interactions of pH,
buffer ions [6] and other compounds in complex virus samples.
Virus inactivation agents have been shown to specifically affect HA
activity of different strains [7,8]. Jonges et al. evaluated the use of
different conditions and agents for viral inactivation. Heat and ␤-
propiolactone lead to a decreased HA-titer. Comparatively smaller
losses of the HA-titer can be observed after formalin treatment.

http://dx.doi.org/10.1016/j.chroma.2016.08.056
0021-9673/© 2016 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/
4.0/).
118 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125
TritonTM X-100 concentrations exceeding 0.2% lead to a complete
loss of hemagglutination activity of H3N2. However, the deter-
mined HA-titer of H7N3 exceeds the initial value after treatment
with 1% TritonTM X-100 [8]. The influence on HA activity extends
to inorganic ions typically used during downstream processing.
Calcium ions have been shown to mediate hemagglutination in
presence of amyloid proteins [9].
Both methods, SRID and HA assay, are generally recognized
and are being applied in research and development of purifica-
tion processes. Kalbfuss et al. use the HA assay for mass-balancing
of filtration. They conclude that roughly 40% of the HA activity is
deposited on or inside the membrane, since it could neither be
recovered in the permeate fraction nor in the retentate. They fur-
ther suggest that this may represent cell debris [10]. Changes in
HA activity due to fragmentation of initially spherical virus parti-
cles and aggregation might contribute as well. A recently published
manuscript of our group describes discrepant mass balances, as
well. Occasionally, the HA activity recovered from different anion
exchange resins exceeds HA activity prior to chromatography by
more than 100% [11]. In this case, non-conformity of activity was
shown to be independent from the sample matrix. Thus, compar-
ing the efficiency of different purification schemes based on HA
activity may lead to an incomplete picture. This is in particular
important when comparing data obtained with different virus feed-
streams. Non-closing mass balances and reproducibility issues are
addressed by another manuscript by Kalbfuss et al., where they
suggest the use of a continuous HA assay format [12]. Undoubt-
edly, this assay format leads to a general improvement of the assay.
However, matrix interference remains a problem. Consequently,
buffer exchange steps might be necessary to quantify samples from
process development.
During the last years, enhanced process understanding and
control has been pushed by the authorities [13]. This can be accom-
plished through process analytical technologies that are applicable
to a wide design space of conditions potentially being used during
the manufacturing process. Public institutions, such as the Euro-
pean Medicine Agency (EMA), encourage research in the field to
complement the SRID assay [14]. Investigation of size, content, and
immunogenicity of aggregates in the drug product or substance is
recommended [14]. Numerous other techniques to either quantify
influenza virus particles or the activity have been published. These
include, but are not limited to: electron microscopy [15,16], surface
plasmon resonance [17,18], enzyme linked immunosorbent assay
(ELISA) [19], and reversed phase high pressure liquid chromatog-
raphy [3]. Electron microcopy and surface plasmon resonance are
comparatively expensive. Surface plasmon resonance further suf-
fers from the same constraint as the SRID, since it requires an
appropriate set of antibodies for new influenza strains. This holds
true for most ELISAs. Reversed phase liquid chromatography is
sample destructive and strong interference with ␤-propiolactone
[5] limits its versatility. Except for electron microscopy, none of
the previously listed methods allows to investigate the aggregate
or fragment content of virus preparations. Alternative techniques
to investigate the size distribution of virus samples are dynamic
light scattering (DLS) or differential centrifugal sedimentation [20].
DLS is a mild, non-matrix assisted method, which is a great advan-
tage with regards to potential secondary interactions. However,
it does not allow sampling of different size fractions of the virus
preparations for further activity analysis. In contrast, differential
centrifugal sedimentation is afflicted with shear forces.
Analytical size exclusion chromatography (SEC) is a non-
destructive technique that has been used for decades to analyze
the aggregate and fragment content of therapeutic proteins [21].
Recently, several approaches to apply SEC to the aggregate anal-
ysis of virus and virus like particle samples have been published
[22–25]. Earlier approaches in SEC of viruses did not pay much
attention to this aspect [26]. Availability of SEC columns that
allow for size distribution analysis of large virus particles, such as
influenza, is limited. Due to the size of whole influenza virus parti-
cles, the estimated pore diameter distribution of such media should
range between 300 nm to 1000 nm. Kalashnikova et al. found that
behavior of protein coated nanoparticles in adsorption chromatog-
raphy is mostly determined by the surface proteins [27]. This
may hold true for non-adsorptive chromatography, as well. Hence,
assessing pore volume and pore diameter by inverse SEC of such
media is difficult, due to the limited availability of monodisperse
proteins or protein-coated standards that have an appropriate
hydrodynamic radius. An alternative method is provided by mer-
cury porosimetry. Mercury intrudes at increasing pressure into the
intra-particle and inter-particle space of dry porous particles. Pore
diameters and pore volumes can be calculated from the required
pressure and the corresponding mercury volume [28].
Herein, we present a complimentary analytical SEC method. The
current study includes 3 different influenza A virus strains. The
pandemic H1N1 vaccine strain 5258 is industrially produced in
mammalian cells. Influenza A H1N1 PR/8/34 and H3N2 Aichi/2/68
originate from embryonated chicken cells.
2. Material and methods
2.1. Influenza samples
The pandemic H1N1 vaccine strain (5258) was produced in
adherent Madin Darby Bovine Kidney (MDBK) cells at IDT Biologika
GmbH (Dessau-Roßlau, Germany). A reference process has been
published by Hundt et al. [29]. The H1N1v 5258 virus sample
was inactivated by ␤-propiolactone treatment. The feedstream was
volumetrically concentrated by a factor of 20 to a HA activity of
2000 HAU/100 ␮L. Samples for virus particle fragmentation exper-
iments were pre-purified by preparative SEC. TOYOPEARL HW-65F
(Tosoh Bioscience GmbH, Griesheim, Germany) was packed to a
bed height of 22 cm. 100 mM sodium phosphate, pH 7.0 containing
200 mM sodium chloride was used as mobile phase at a flow rate of
150 cm/h. All chemicals were purchased from Sigma Aldrich/Merck
KGaA (Darmstadt, Germany), unless otherwise stated. Besides this
feedstream and SEC pre-purified feedstream, 15 H1N1v samples
with different host cell DNA contents were available. These sam-
ples originate from method development experiments. HA-activity
of these samples ranges from 6 HAU/100 ␮L to 280 HAU/100 ␮L.
Influenza A H1N1 PR/8/34 and H3N2 Aichi/2/68 were purchased
from Charles River Avian Vaccine Services (North Franklin, USA).
Both viruses were produced in embryonated chicken eggs and inac-
tivated by formalin treatment. The virus samples were purified and
concentrated to 2 mg/mL by density gradient centrifugation.
2.2. Mercury porosimetry
Mercury porosity of the TSKgel G6000PWxl stationary phase
was accomplished with a PoreMaster 60-GT (Quantachrome,
Odelzhausen, Germany). Porosimetry experiments were conducted
according to the instructions of the instrument supplier. The resin
was washed with deionized water and freeze-dried at −45 ◦ C and
0.03 mbar for 24 h with a Zirbus VaCo 2 (Bad Grund, Germany)
lyophilizer prior to the porosimetry experiments. 14.34 mL wet
resin correspond to 1 column volume of TSKgel G6000PWxl
(7.8 mm ID × 30 cm L). Mass and volume of the freeze-dried resin
were determined and used for further calculation. 250 mg of the
resin was filled into a penetrometer and connected to the mercury
porosimetry instrument. The penetrometer was filled with mer-
 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125
cury at atmospheric pressure to ensure settling of the resin in the
penetrometer.
Mercury porosimetry data obtained with freeze-dried poly-
methacrylate resins of lower pore diameter are in well agreement
with iSEC data of these resins. Due to the lack of appropriate stan-
dards, such a comparison is not available for TSKgel G6000PWxl.
Pore diameters were calculated using Eq. (1), where P is the
change in the applied pressure,  Hg the surface tension of mercury,
 the contact angle between mercury and the surface and rpore the
pore radius [28].
2Hg·cos
P =
rpore
2.3. Dynamic light scattering experiments
The hydrodynamic radius of H1N1v in different buffers was
measured using a Beckmann Coulter DelsaMax Pro (Krefeld,
Germany) dynamic light scattering device. Dialysis into different
buffers was accomplished overnight at 4 ◦ C. The investigated
buffer systems correspond to the buffers applied in analytical
size exclusion chromatography: 100 mM sodium phosphate,
100 mM potassium phosphate, 100 mM sodium citrate, 100 mM 2-
[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol
(Bis/Tris), 100 mM 2-(N-morpholino)ethanesulfonic acid (MES)
and 100 mM 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic
acid (HEPES), pH 7.0, all substituted with 200 mM sodium chloride.
Another set of buffers was based on 100 mM sodium phosphate,
pH 7.0 with 0 mM, 200 mM, 500 mM and 1 M sodium chloride
or arginine/HCl, respectively. Each measurement consisted of 30
individual measurements with three repeats. Blanks of all buffers
were included. Interlaced injections with 100% HPLC grade ethanol
were used as negative control.
2.4. Analytical size exclusion chromatography
TSKgel G6000PWxl (7.8 mm ID × 30 cm L) columns were used
for analytical SEC (Tosoh Bioscience GmbH). Low densities of resid-
ual anionic groups on the surface of the 13 ␮m polymethacrylate
beads prevent from ionic interactions with anions. The estimated
exclusion limit of this column for proteins is <20000 kDa.
Void volumes of TSKgel G6000PWxl and the different chro-
matography systems were determined with injections of 20 ␮L
5% acetone in water or 100 ␮L of a 1 M sodium chloride solu-
tion. Acetone or salt were eluted in 100 mM sodium phosphate, pH
7.0 + 200 mM sodium chloride at 1.0 mL/min. Retention of the cor-
responding peaks was used as a reference in further calculations of
the intra- and inter-particle volumes and account for the respective
system void volumes, as well.
The applied flow rate during SEC was 1.0 mL/min. Flow rates
were chosen according to findings reported in a previous publi-
cation of our group [30]. 100 mM sodium phosphate, pH 7.0 with
200 mM sodium chloride was used as a standard SEC buffer. The
impact of different buffer systems on virus particle recovery and
size distribution was investigated with various buffers at pH 7.0:
100 mM potassium phosphate, 100 mM sodium citrate, 100 mM
HEPES, 100 mM Bis/Tris and 100 mM MES. 200 mM sodium chlo-
ride was added to each of these buffers in order to suppress ionic
interactions with the stationary phase and to stabilize the virus par-
ticles. The effect of different concentrations of arginine and sodium
chloride was evaluated using 100 mM sodium phosphate buffer at
pH 7.0. Sodium chloride and arginine hydrochloride were added
in concentrations of 0 mM, 200 mM, 500 mM and 1 M. Recoveries
were determined by the area under the virus particle curves.
UV signals were traced at 214 nm/215 nm or 280 nm. Buffer test-
ing was based on the UV 280 nm signal, since some of the applied
119
buffers absorb UV light at 214 nm/215 nm. Due to the higher sig-
nal to noise ratio in 100 mM sodium phosphate buffer, other SEC
experiments were monitored at these lower wavelengths. 100 ␮L
of the samples were injected in analytical experiments. The col-
umn was overloaded in semi-preparative SEC experiments. 1 mL of
the virus samples was injected. The pre-purified and concentrated
H1N1 PR/8/32 and H3N2 Aichi/2/68 samples were 3-fold diluted
prior to injection. 1 mL fractions were collected. All samples were
analyzed in duplets or triplets.
2.5. Hemagglutination assay
(1)
The hemagglutination activity was determined using a Tecan
Freedom Evo 150. Chicken erythrocyte preparations in Alsever
solution were purchased from preclinics (Potsdam, Germany).
4 mL of the erythrocyte preparation were diluted with 40 mL
phosphate buffered saline (PBS) consisting of 1.54 mM potassium
dihydrogen phosphate, 8.10 mM disodium hydrogen phosphate,
2.68 mM potassium chloride and 137 mM sodium chloride, pH 7.5
at 25 ◦ C. After centrifugation at 500g, pelleted erythrocytes were
resuspended in 40 mL cold PBS. Erythrocyte concentration was
determined with a Fuchs-Rosenthal cytometer and adjusted to
2 × 107 cells per mL. 100 ␮L of the H1N1 containing samples were
transferred into every second row of the first column of U-bottom
96 well plates. A 1:20.5 dilution of the samples was pipetted into
the first well of the remaining rows. Serial 1:2 dilutions of all sam-
ples in phosphate buffered saline were completed. 100 ␮L of the
erythrocyte solution were added to all wells and plates were incu-
bated for one hour at room temperature. Results were evaluated
manually and by absorbance readings at 700 nm.
2.6. Virus particle fragmentation
Fragmentation of pre-purified H1N1v 5258 was accomplished
through treatment with TritonTM X-100 or heat incubation accord-
ing to the protocol from Jonges et al. [8]. 1 mL aliquots of the
pre-purified virus sample were incubated for 1 h with 0.0, 0.2, 0.5,
and 1.0% (v/v) TritonTM X-100 at room temperature. Samples were
dialyzed into 100 mM sodium phosphate buffer, pH 7.0 containing
200 mM sodium chloride at 4 ◦ C to remove the detergent. Alterna-
tively, 1 mL samples were incubated for 30 min at 20, 40, 50, 60,
and 70 ◦ C. Experiments were performed at least in duplets.
2.7. DNA assay
Double stranded DNA contents of the 15 samples from method
development experiments were determined using a Qubit dsDNA
HS assay kit (Thermo Fisher Scientific, Dreieich, Germany). The
assay was conducted according to the instructions of the manu-
facturer. In brief, 20 ␮L of the samples were mixed with 180 ␮L
of the assay buffer. The double stranded DNA concentration was
determined through fluorescence measurements. The lower limit
of detection was 5 ng/mL.
3. Results and discussion
3.1. Pore diameter measurements
14.34 mL packed wet bed correspond to 3.1 g freeze-dried resin
with a non-compressed volume of 17.5 mL. This corresponds to a
methacrylate mass density of 0.22 g/mL in the wet packed bed.
Comparing volumes of the wet packed bed and the freeze-dried
free resin leads to a compression factor of roughly 1.2. This is in
agreement with data obtained for polymethacrylate resins [31,32],
indicating that the applied freeze-drying procedure mostly pre-
serves the spherical shape of the resin particles. The calculated
120 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125

Fig. 1. Mercury porosimetry data for the TSKgel G6000PWxl resin. The mean pore
diameter of the resin is 250 nm. The pore diameter distribution ranges from 20 nm
to 2000 nm. Intra-particle volume merges into inter-particle volume.

Fig. 2. Hydrodynamic radiuses of H1N1v samples in different buffer systems. Hydro-

dynamic radiuses are in the range of 50 nm. Arginine addition or use of citrate buffer
lead to an increase of the hydrodynamic radius. In absence of sodium chloride or
arginine, measured radiuses are comparatively smaller.
volume contribution of methacrylate to the packed column bed is
4.6 mL.
The pore diameter distribution of TSKgel G6000PWxl was deter-
3.2. Dynamic light scattering
mined using mercury porosimetry. Intra- and inter-particle volume
cannot be delimited. The intruded mercury volume is continuously
Hydrodynamic radiuses of H1N1 samples in the different buffer
growing and strives through a minimum at 2000 nm (Fig. 1). Given
systems are shown in Fig. 2. Hydrodynamic radiuses are in the
2000 nm defines the border between intra- and inter-particle vol-
range of 50 nm. This is in accordance with values reported in lit-
umes, corresponding values can be calculated. The resulting pore
erature [34]. 100 mM sodium phosphate buffer +200 mM sodium
diameter distribution ranges from 20 nm to 2000 nm with a mean
chloride lead to a hydrodynamic radius of 59 nm. If this is set as
value of 250 nm. The calculated intra-particle volume of the column
a reference, sodium citrate buffer and sodium phosphate buffer
is 3.9 mL. These results suggest that virus particles are not excluded
substituted with 0.5 or 1.0 M arginine lead to an increase of the
from the pore volume of the column. Available volume of pores or
hydrodynamic radius. Arginine is well-known for its ability to inter-
flow channels with a diameter ranging from 500 nm to 3000 nm is
act with proteins in solution [35,36,21]. The introduced changes in
limited, however. At the low end of calibration, capacity for host
the electric double layer seem to increase the hydrodynamic radius
cell proteins and other low molecular impurities is limited, as well.
of the virus particles.
Virus particle aggregates represent a product related impu-
The increase of the average hydrodynamic radius in citrate
rity with comparatively greater hydrodynamic radius. Capacity for
might be due to hydrophobic interactions, since sodium citrate is a
aggregates that can be separated due to pore diffusion correlates
kosmotropic buffer. Impurities in the feedstream, such as DNA, host
inversely with their hydrodynamic radius. However, separation
cell proteins or cell debris might interact hydrophobically with the
of very large virus aggregates might eventually be supported by
whole virus particles. Aggregation based on hydrophobic interac-
hydrodynamic effects and perfusive flow channels [33]. The calcu-
tions might contribute, as well. Resolution of different size species
lated inter-particle volume of the column is 5.6 mL and directly
in DLS data can be difficult. In many cases an average hydrodynamic
merges with the intra-particle volume. The greatest part of the
radius is provided.
inter-particle volume constitutes of flow channels with diameters
The smallest hydrodynamic radius was obtained in 100 mM
smaller than 10,000 nm. Such flow channels contribute with 4.0 mL
sodium phosphate buffer without sodium chloride or arginine.
to the total volume of the packed bed. Due to the size distribu-
The hydrodynamic radius in this buffer is 27 nm. The lower ionic
tion of the polymethacrylate particles, column packings exhibit a
strength of this solution probably affects the electric double layer
size distribution with regards to the flow channel diameters. Flow-
of the virus particles.
through channels in the particles may contribute to this volume, as
well. Both may further extend the separation range. Linearity of this
extension remains to be proven, since the availability of appropriate 3.3. Size exclusion chromatography in different buffers
calibration standards is limited.
The presented pore diameter data indicates that aggregates, Six different buffer systems at a concentration of 100 mM and pH
monomers and fragments of influenza whole virus particles can 7.0 were applied for SEC of H1N1v. Furthermore, different concen-
be separated, analytically. Hence, the column is used for the sub- trations of arginine and sodium chloride were added as additives
sequent analytical and semi-preparative SEC experiments. In the to 100 mM sodium phosphate. Corresponding chromatograms are
semi-preparative approach, loadings did not exceed 0.7 mg. This shown in Fig. 3A and B. Retention of the whole virus particle peak
corresponds to a loaded mass of <0.05 mg/mL of packed bed. remains constant at 8.4 min. Provided 2000 nm is the threshold
Preparative separations require high capacity for host cell protein, between intra- and inter-particle volume, whole virus particles
while a group separation of virus particles might be satisfying. To hardly access the intra-particle volume. It seems separation of virus
this end, resins with greater pore volume provided by smaller pores particles is mostly based on perfusive flow channels with a diame-
might be more efficient. ter of 3500 nm.
 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125
Fig. 3. SEC chromatograms of H1N1v. A: Separations in 100 mM sodium phosphate,
100 mM sodium citrate, 100 mM HEPES, 100 mM MES, 100 mM Bis/Tris and 100 mM
potassium phosphate, all pH 7.0. All buffers were substituted with 200 mM sodium
chloride to supress ionic secondary interactions. Virus particles elute at 8.4 min in all
buffer systems. Citrate buffer leads to elution of a small peak prior to the whole virus
particle peak. B: Separations in 100 mM sodium phosphate buffer containing 0 mM,
200 mM, 500 mM or 1 M sodium chloride or arginine/HCl, respectively. Increasing
concentrations lead to an increase of the area under the peak of the components
eluting at 10 min or later. The area under the curve of the whole virus particle peak
decreases in absence of additive.
The applied buffer systems and additives hardly affect the elu-
tion volume of virus particles. 100 mM sodium phosphate with an
addition of 1 M arginine represents an exception. In this buffer sys-
tem, retention is shifted by 0.1 min. DLS measurements in presence
of arginine reveal an increase in the hydrodynamic radius by 5 nm
compared to 100 mM sodium phosphate +200 mM sodium chloride
(Fig. 2). This increase may account for the increased retention due
to the addition of 1 M arginine.
Addition of arginine further affects peak shape and height of
the substances eluting at 10 min and later. The peaks decrease and
become broader at increasing arginine concentrations. The feed-
stream samples are a complex mixture of virus particles and cell
debris. The previously discussed arginine-protein interactions are
also known to suppress secondary interactions in SEC [35,36,21].
This may explain why the least eluting impurity elutes compara-
tively earlier. This impurity cannot be eluted within the separation
range of the column in buffer systems without arginine. Addition of
sodium chloride even increases retention further. Hence, this least
peak most probably represents a hydrophobic impurity, such as
121
the M2 membrane channel protein. Changes in peak shape due to
addition of arginine might be based on these effects, as well.
Sodium citrate buffer represents another exception, which is in
agreement to DLS data. A small and broad peak elutes prior to the
whole virus particle peak. This peak might constitute virus particle
aggregates. Aggregation is one factor affecting size distribution of
virus particle preparations. Sodium citrate is a kosmotropic buffer.
The support of hydrophobic interactions may lead to virus particle
aggregation. DLS data indicates an increase in the average hydrody-
namic radius (Fig. 2). Aggregates as a separate scattering compound
cannot be identified. However, DLS of size distributions such as
the virus particle samples may not clearly distinguish between the
different species.
Chromatograms of SEC indicate that a certain ionic strength
similar to physiological conditions supports recovery of whole
virus particles. The virus particle peak in 100 mM sodium phos-
phate buffer without additive is comparatively smaller. The areas
under the curve of the whole virus particle peaks in all buffers are
shown in Fig. 4A and B. 100 mM sodium phosphate without additive
lead to an area of 21 mAU*min, whereas a maximum recovery of
34 mAU*min can be achieved in 100 mM sodium phosphate buffer
with addition of 200 mM sodium chloride. Further increase of the
sodium chloride addition decreases the whole virus particle recov-
ery. This might be due to hydrophobic interactions. According to
these results, there is an optimum ionic strength. Ionic secondary
interactions must be suppressed on the one hand. On the other,
hydrophobic secondary interactions must not be supported and
residual anionic charges on the resin surface must not be saturated
with counter ions.
Whole virus particle recovery in sodium citrate buffer is sim-
ilar to sodium phosphate buffer. Given the peak eluting prior to
the whole virus particle peak in citrate buffer represents virus par-
ticle aggregates, hydrophobic interactions between virus particles
appear more likely than between virus particles and the stationary
phase. The low density residual anionic charges on the surface of
the stationary phase may be hold responsible for this difference.
3.4. Hemagglutinin activity of size exclusion chromatography
fractions
Semi-preparative SEC chromatograms of H1N1v 5258, H1N1
PR/8/34, and H3N2 Aichi/2/68 on TSKgel G6000PWxl are shown
in Fig. 5. The HA activity of the collected fractions is plotted on
the right y-axis. The higher system dead volume of the prepara-
tive system and the great injection volume have a negative impact
on chromatographic performance. Fractions corresponding to the
peak eluting between 8 and 10 min have the highest HA activity.
Fractions eluting after the main virus peak contain HA activity, as
well. Elution of HA activity ranges from 6.5 mL to 19.5 mL. This cor-
responds to pore and channel diameters starting at 7000 nm and
smaller. The lower pore diameter cannot be determined, since elu-
tion of HA activity extends to the range dominated by secondary
interactions.
Roughly 30% of the total HA activity elutes in the intra-particle
volume, representing lower molecular weight components.
Hemagglutination is considered a cross-linking reaction. It appears
reasonable that the hemagglutination inducing molecules must
not have a minimum size, but at least two functional sites. Thus,
virus particle debris most likely mediates hemagglutination. This
means virus particle fragmentation can lead to higher HA activity,
compared to whole virus particles, since the total number of HA
mediating molecules increases. This has important implications on
the use of HA activity as a measure for virus particle quantification.
Virus particle quantification based on activity measurements can
lead to significantly overestimated virus particle contents. Differ-
122 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125

Fig. 4. Recovery determined by the area under the curve of the whole virus particle
peak of H1N1v from SEC. A: Recovery in 100 mM sodium phosphate and 100 mM
sodium citrate are higher compared to 100 mM HEPES, 100 mM MES, 100 mM
Bis/Tris and 100 mM potassium phosphate. B: 100 mM sodium phosophate was sub-
stituted with different concentrations of sodium chloride or arginine/HCl. Highest
recovery is obtained in presence of 200 mM sodium chloride. The applied arginine
concentration has no impact on the area under the curve of the whole virus particle
peak. Fig. 5. Semi-prep SEC chromatograms of H1N1v 5258 (A), H1N1 PR/8/34 (B), H3N2
Aichi/2/68 (C) on TSKgel G6000PWxl. The HA activity of the corresponding SEC frac-
tions is plotted in the graph. HA activity elutes in the inter- and intra-particle volume
ent degrees of virus fragmentation may also contribute to the high and extends to the volume dominated by secondary interactions. Roughly 30% of the
HA activity elutes in the intra-particle volume after the whole virus particle peak.
standard deviation of the HA assay.
Summed HA activities of all fractions are in good agreement
with the initial HA activity of the sample prior to SEC. 3.9 log of assay. Further, the HA assay uses a discrete format. Precision of the
the H1N1v 5852 sample were loaded in Fig. 5A. The HA activity of assay could be enhanced by a greater number of erythrocyte dilu-
all fractions normalized against the volume is 4.1 log HAU. The ini- tions. Thus, the applied buffer conditions can be considered suitable
tially injected HA activity of H1N1 PR/8/34 is 5.4 log HAU (Fig. 5B). to quantitatively recover influenza virus samples.
The recovered HA activity of this sample sums up to 5.2 log HAU. A
comparatively greater, but still acceptable loss appears for the frac- 3.5. Hemagglutination of different virus particle fragments
tions of H3N2 Aichi/2/68 (Fig. 5C). 4.8 log HAU of the loaded 5.4 log
HAU can be recovered from TSKgel G6000PWxl. Considering SEC The previously reported results suggest that virus fragments
is a diluting method, HA activity of the fractions framing the virus increase the total HA activity of a sample. Different virus inacti-
particle peaks might be lower than the limit of detection of the HA vation methods lead to decreasing or increasing HA activity. Jonges
 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125
Fig. 6. Hemagglutination activity of SEC pre-purified H1N1v 5258 in HAU per 100 ␮L
after heat incubation (A) and TritonTM X-100 treatment (B). HA activity of the Triton
reference sample is comparatively lower, due to an additional dialysis step. Tem-
peratures and Triton concentrations are indicated in the corresponding graphs. HA
activity decreases with increasing incubation temperature. 0.2% and 0.5% (v/v) Triton
lead to an increase in HA activity. Treatment with 1% Triton decreases HA activity.
et al. evaluated the impact of different inactivation agents on
hemagglutination activity [8]. The experiments revealed that HA
activity of some influenza strains increases after treatment with
the membrane solubilizing detergent TritonTM X-100. In contrast,
they found a decrease in HA activity in response to heat treatment.
HA activity of H1N1v 5258 after heat incubation or TritonTM X-100
treatment is shown in Fig. 6A & B. Incubation at 50 ◦ C or higher
leads to a loss of HA activity. In contrast, HA activity is increas-
ing after treatment with 0.2 and 0.5% TritonTM X-100, but declines
again after treatment with 1% TritonTM X-100. Jonges et al. found
no such decrease after passing a maximum HA activity. However,
the general trend observed with H1N1v 5258 corresponds to their
study.
Analytical SEC chromatograms of the pre-purified and treated
samples are shown in Fig. 7A & B. All samples are pre-purified and
elute within the inter- and intra-particle volume, except for the 1%
Triton treated sample. This sample contains compounds eluting in
the volume affected by secondary interactions. Increasing concen-
trations of the membrane solubilizing detergent generally increase
the fragment peaks. Detergents improve stability of vaccine formu-
lations [37,34]. Considering vaccines are often quantified through
HA activity measurements, this stabilizing effect might be due to an
increase of the number of hemagglutination-inducing molecules.
123
Fig. 7. SEC chromatograms of H1N1v 5258 after heat incubation (A) or TritonTM
X-100 treatment (B) on TSKgel G6000PWxl. Temperatures and detergent concen-
trations are indicated in the graph. The whole virus particle peak eluting at 8.4 min
decreases with increasing temperature or detergent concentration. Fragment peaks
increase. Except for the sample treated with 1% Triton, all components elute within
the inter- and intra-particle volume.
Aging of HA proteins might be compensated by membrane solubi-
lization.
Increasing the incubation temperature leads to an increase of
the low molecular weight peak, as well. It appears fragments do
not necessarily increase the HA activity of a sample. Activity of the
HA protein seems unaffected by low concentrations of TritonTM
X-100. The enveloping virus membrane is presumably affected,
though. Higher concentrations of the detergent may lead to pro-
tein unfolding and subsequent activity loss. Unfolded proteins are
prone to secondary interactions and may represent the peak eluting
after the intra-particle volume. Critical Triton concentrations may
depend on a particular HA. Inactivation by heat treatment seems
to immediately affect functionality of HA.
The presented data indicates that HA activity and size distribu-
tion of the investigated virus particle preparations do not correlate.
Eventually, this means size distribution analysis by SEC or other
techniques cannot substitute activity based assays of vaccine for-
mulations. However, the size of an antigen is an important trigger
for the immune system [38]. Thus, assessment of the size distri-
bution of vaccine preparations as an additional quality parameter
seems reasonable.
124 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125
Fig. 8. Retention of the whole virus particle peak of 15 H1N1v samples is plotted
against their corresponding DNA content. DNA concentrations range from <5 ng/mL
to 676 ng/mL. Data is widely scattered. Nevertheless, linear regression indicates an
increase of retention at increasing DNA purity.
3.6. DNA purity
Changes in size distribution of whole virus particle prepara-
tions are not limited to aggregation or fragmentation. Adsorption
of host cell derived DNA might increase the average virus parti-
cle size. Virus particle preparations with different host cell DNA
content were analyzed with SEC in 100 mM sodium phosphate, pH
7.0 + 200 mM sodium chloride. Retention increases towards lower
molecular weight for virus samples with higher DNA purity. Cor-
responding data is shown in Fig. 8. Retention of the whole virus
particle peak is plotted against the DNA content determined by
a fluorescence assay. Accuracy of such a calibrated SEC for quan-
tification of DNA purity is certainly lower compared to alternative
techniques. This method cannot replace polymerase chain reaction
or fluorescence assays for quantification of DNA in virus particle
samples. Nevertheless, SEC might still provide an indication on DNA
purity in downstream processing of the same virus strain.
4. Conclusions
Pore diameter distribution of TSKgel 6000PWxl ranges from
20 nm to 2000 nm. Intra-particle volume merges with the inter-
particle volume. Separation of influenza A virus particle monomers,
aggregates and fragments is based on both, intra- and inter-particle
volume. Retention of whole virus particles from the H1N1 vaccine
strain and the two strains propagated in embryonated chicken eggs
is similar. 100 mM sodium phosphate, pH 7.0 with 200 mM sodium
chloride seems to have appropriate ionic strength to prevent from
secondary interactions of the whole virus particles. This eluent pro-
vides full recovery of HA activity. More than 30% of the HA activity
of the samples is attributed to low molecular weight variants. These
virus particle fragments may increase total HA activity, potentially
leading to overestimated hemagglutinin masses. Further, retention
of the whole virus particle peak can be used for estimating the
amount of surface adsorbed DNA.
Hence, SEC adds valuable data regarding the size distribution
of influenza virus samples to the commonly used HA assay for the
quantification of influenza viruses. Alternatively, the method could
be combined with SRID. Future purification processes may benefit
from a more detailed picture of the virus size distribution and how
it is affected during the different steps of downstream processing.
Acknowledgements
This work was supported by funding of the German Federal Min-
istry of Education and Research, Grant 0315640D, and is part of
the research project “Optimization of an industrial process for the
production of cell-culture-based seasonal and pandemic influenza
vaccines”.
References
[1] C.M. Mair, K. Ludwig, A. Herrmann, C. Sieben, Receptor binding and pH
stability—how influenza A virus hemagglutinin affects host-specific virus
infection, Biochim. Biophys. Acta 1838 (2014) 1153–1168, http://dx.doi.org/
10.1016/j.bbamem.2013.10.004.
[2] L. Opitz, J. Salaklang, H. Büttner, U. Reichl, M.W. Wolff, Lectin-affinity
chromatography for downstream processing of MDCK cell culture derived
human influenza A viruses, Vaccine 25 (2007) 939–947, http://dx.doi.org/10.
1016/j.vaccine.2006.08.043.
[3] J. Transfiguracion, A.P. Manceur, E. Petiot, C.M. Thompson, A. a Kamen,
Particle quantification of influenza viruses by high performance liquid
chromatography, Vaccine 33 (2015) 78–84, http://dx.doi.org/10.1016/j.
vaccine.2014.11.027.
[4] G.K. Hirst, The quantitative determination of influenza virus and antibodies
by means of red cell agglutination, J. Exp. Med. 75 (1942) 49–64, http://dx.doi.
org/10.1084/jem.75.1.49.
[5] C.M. Thompson, E. Petiot, A. Lennaertz, O. Henry, A. a Kamen, Analytical
technologies for influenza virus-like particle candidate vaccines: challenges
and emerging approaches, Virol. J. 10 (2013) 141, http://dx.doi.org/10.1186/
1743-422X-10-141.
[6] G.L. Miller, Influence of pH and of certain other conditons on the stability of
the infectivity and red cell agglutinating activity fo influenza virus, J. Exp.
Med. 80 (1944) 507–520.
[7] P. Bonnafous, M.-C.C. Nicolaï, J.-C.C. Taveau, M. Chevalier, F. Barrière, J.
Medina, et al., Treatment of influenza virus with beta-propiolactone alters
viral membrane fusion, Biochim. Biophys. Acta: Biomembr. 1838 (2014)
355–363, http://dx.doi.org/10.1016/j.bbamem.2013.09.021.
[8] M. Jonges, W.M. Liu, E. van der Vries, R. Jacobi, I. Pronk, C. Boog, et al.,
Influenza virus inactivation for studies of antigenicity and phenotypic
neuraminidase inhibitor resistance profiling, J. Clin. Microbiol. 48 (2010)
928–940, http://dx.doi.org/10.1128/JCM.02045-09.
[9] H. Hamazaki, Calcium-mediated hemagglutination by serum amyloid P
component and the inhibition by specific glycosaminoglycans, Biochem.
Biophys. Res. Commun. 150 (1988) 212–218, www.sciencedirect.com/
science/article/pii/0006291X88905074 (accessed: 12.01.16).
[10] B. Kalbfuss, Y. Genzel, M. Wolff, A. Zimmermann, R. Morenweiser, U. Reichl,
Harvesting and concentration of human influenza A virus produced in
serum-free mammalian cell culture for the production of vaccines,
Biotechnol. Bioeng. 97 (2007) 73–85, http://dx.doi.org/10.1002/bit.21139.
[11] J. Vajda, D. Weber, S. Stefaniak, B. Hundt, T. Rathfelder, E. Müller, Mono- and
polyprotic buffer systems in anion exchange chromatography of influenza
virus particles, J. Chromatogr. A 1448 (2016) 73–80, http://dx.doi.org/10.
1016/j.chroma.2016.04.047.
[12] B. Kalbfuss, A. Knöchlein, T. Kröber, U. Reichl, Monitoring influenza virus
content in vaccine production: precise assays for the quantitation of
hemagglutination and neuraminidase activity, Biologicals 36 (2008) 145–161,
http://dx.doi.org/10.1016/j.biologicals.2007.10.002.
[13] J.O. Josefsberg, B. Buckland, Vaccine process technology, Biotechnol. Bioeng.
109 (2012) 1443–1460, http://dx.doi.org/10.1002/bit.24493.
[14] European Medicines Agency, Guideline on Influenza Vaccines—Quality
Module, 2014 http://www.ema.europa.eu/docs/en GB/document library/
Scientific guideline/2014/06/WC500167817.pdf.
[15] T. Noda, Native morphology of influenza virions, Front. Microbiol. 2 (2011)
269, http://www.pubmedcentral.nih.gov/articlerender.
fcgi?artid=3249889&tool=pmcentrez&rendertype=abstract (accessed:
30.01.16).
[16] S.M. Dumitrescu, C.N. Zaharia, I. Samuel, F. Bârnaure, N. Manolescu, Electron
optic investigations on influenza virus A(H1N1) interacted with some
aromatic compounds, Virologie 32 (2016) 289–292, http://www.ncbi.nlm.nih.
gov/pubmed/7331234 (accessed: 10.09.15).
[17] D.K. Takemoto, J.J. Skehel, D.C. Wiley, A surface plasmon resonance assay for
the binding of influenza virus hemagglutinin to its sialic acid receptor,
Virology 217 (1996) 452–528, http://www.sciencedirect.com/science/article/
pii/S0042682296901397 (accessed: 28.02.16).
[18] C.-F. Mandenius, R. Wang, A. Aldén, G. Bergström, S. Thébault, C. Lutsch, et al.,
Monitoring of influenza virus hemagglutinin in process samples using weak
affinity ligands and surface plasmon resonance, Anal. Chim. Acta 623 (2008)
66–75, http://www.sciencedirect.com/science/article/pii/
S0003267008010854 (accessed: 28.02.16).
[19] S. Chun, C. Li, G. Van Domselaar, J. Wang, A. Farnsworth, X. Cui, et al.,
Universal antibodies and their applications to the quantitative determination
of virtually all subtypes of the influenza A viral hemagglutinins, Vaccine 26
(2008) 6068–6076, http://www.ncbi.nlm.nih.gov/pubmed/19007587
(accessed: 28.02.16).
 J. Vajda et al. / J. Chromatogr. A 1465 (2016) 117–125
[20] M.M. Pieler, A.R. Fortuna, U. Reichl, M.W. Wolff, Investigation of virus
aggregation on performance of downstream processing in influenza vaccine
manufacturing, Chem. Ing. Tech. 86 (2014) 1588, http://dx.doi.org/10.1002/
cite.201450558 (accessed: 27.02.16).
[21] P. Hong, S. Koza, E.S.P. Bouvier, Size-exclusion chromatography for the
analysis of protein biotherapeutics and their aggregates, J. Liq. Chromatogr.
Relat. Technol. 35 (2012) 2923–2950, http://dx.doi.org/10.1080/10826076.
2012.743724.
[22] Z. Wei, M. Mcevoy, V. Razinkov, A. Polozova, E. Li, J. Casas-Finet, et al.,
Biophysical characterization of influenza virus subpopulations using field
flow fractionation and multiangle light scattering: correlation of particle
counts, size distribution and infectivity, J. Virol. Methods 144 (2007) 122–132,
http://dx.doi.org/10.1016/j.jviromet.2007.04.008.
[23] Y. Yang, H. Li, Z. Li, Y. Zhang, S. Zhang, Y. Chen, et al., Size-exclusion HPLC
provides a simple, rapid, and versatile alternative method for quality control
of vaccines by characterizing the assembly of antigens, Vaccine 33 (2015)
1143–1150, http://dx.doi.org/10.1016/j.vaccine.2015.01.031.
[24] M. Havlik, M. Marchetti-Deschmann, G. Friedbacher, P. Messner, W. Winkler,
L. Perez-Burgos, et al., Development of a bio-analytical strategy for
characterization of vaccine particles combining SEC and nanoES GEMMA,
Analyst 139 (2014) 1412–1419, http://dx.doi.org/10.1039/c3an01962d.
[25] C. Ladd Effio, S.A. Oelmeier, J. Hubbuch, High-throughput characterization of
virus-like particles by interlaced size-exclusion chromatography, Vaccine 34
(2016) 1259–1267, http://dx.doi.org/10.1016/j.vaccine.2016.01.035.
[26] M. Potschka, Size exclusion chromatography of DNA and viruses: properties
of spherical and asymmetric molecules in porous networks, Macromolecules
24 (1991) 5023–5039, http://dx.doi.org/10.1021/ma00018a008.
[27] I.V. Kalashnikova, N.D. Ivanova, T.G. Evseeva, A.Y. Menshikova, E.G. Vlakh, T.B.
Tennikova, Study of dynamic adsorption behavior of large-size
protein-bearing particles, J. Chromatogr. A 1144 (2007) 40–47, http://dx.doi.
org/10.1016/j.chroma.2006.10.060.
[28] G. Carta, A. Jungbauer, Protein Chromatography: Process Development and
Scale-Up, Wiley-VCH Verlag, Weinheim, 2010.
125
[29] B. Hundt, N. Mölle, S. Stefaniak, R. Dürrwald, J. Weyand, Large pilot scale
cultivation process study of adherent MDBK cells for porcine Influenza A virus
propagation using a novel disposable stirred-tank bioreactor, BMC Proc. 5
(2011) P128, http://dx.doi.org/10.1186/1753-6561-5-S8-P128.
[30] J. Vajda, W. Conze, E. Müller, Kinetic plots in aqueous size exclusion
chromatography of monoclonal antibodies and virus particles, J. Chromatogr.
A. 1426 (2015) 118–125, http://dx.doi.org/10.1016/j.chroma.2015.11.057.
[31] T.R. Mccaw, E.K. Koepf, L. Conley, Evaluation of a novel methacrylate-based
protein a resin for the purification of immunoglobulins and Fc-fusion
proteins, Biotechnol. Prog. 203 (2014), http://dx.doi.org/10.1002/btpr.1951.
[32] Toyopearl instruction manual, Toyopearl Instr. Man. (n.d.), 1–20, http://www.
separations.eu.tosohbioscience.com/NR/rdonlyres/DD2316A4-7228-420B-
908A-35E724C3369F/0/M09P18A TIM.pdf (accessed: 24.08.15).
[33] N.B. Afeyan, N.F. Gordon, I. Mazsaroff, L. Varady, S.P. Fulton, Y.B. Yang, et al.,
Flow-through particles for the high-performance liquid chromatographic
separation of biomolecules: perfusion chromatography, J. Chromatogr. 519
(1990) 1–29, http://dx.doi.org/10.1016/0021-9673(90)85132-F.
[34] T.C. Kon, A. Onu, L. Berbecila, E. Lupulescu, A. Ghiorgisor, G.F. Kersten, et al.,
Influenza vaccine manufacturing: effect of inactivation, splitting and site of
manufacturing. Comparison of influenza vaccine production processes, PLoS
One 11 (2016) e0150700, http://dx.doi.org/10.1371/journal.pone.0150700.
[35] D. Ejima, R. Yumioka, T. Arakawa, K. Tsumoto, Arginine as an effective
additive in gel permeation chromatography, J. Chromatogr. A 1094 (2005)
49–55, http://dx.doi.org/10.1016/j.chroma.2005.07.086.
[36] R. Yumioka, H. Sato, H. Tomizawa, Y. Yamasaki, D. Ejima, Mobile phase
containing arginine provides more reliable SEC condition for aggregation
analysis, J. Pharm. Sci. 99 (2010) 618–620, http://dx.doi.org/10.1002/jps.
21857.
[37] E.J. D’Hondt, H.B. Engelmann, Process for Producing Influenza Vaccine,
US20120219587 A1, 2012, http://www.google.com/patents/US20120219587
(accessed: 12.08.16).
[38] N.K. Jerne, Paper 36 waiting for the end, in: I. Levkovits (Ed.), Portrait Immune
Syst. Sci. Publ. N K Jerne, 2nd edn., World Scientific Pub Co Inc., 1996, p. 480.