Expression of Soluble VEGF Receptor 2 and
Characterization of Its Binding by Surface
Xianming Huang,1Claudia Gottstein, Rolf A. Brekken, and Philip E. Thorpe
Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, Texas 75235
Received October 21, 1998
Vascular endothelial growth factor (VE GF ) is an en-
dothelial cell specific mitogen that induces angiogen-
esis in several pathological conditions. T o block angio-
genesis, soluble VE GF receptor can be used. In this
study, we describe a method for high yield expression
of soluble VE GF receptor 2 (sF lk-1) in a baculovirus
expression system (30 mg purified sF lk-1 per L of in-
sect cell supernatant). We also determined the binding
constants for both human and mouse VE GF to the
recombinant receptor by surface plasmon resonance.
In this cell-free assay, under the given experimental
conditions, the on-rate kawas 0.5-2.2 x 106M?1s?1and
the off-rate kdwas 2-4 x 10?4s?1(KD? 2-6 x 10?10M).
T o our knowledge this is the first study to report on-
and off-rates for the VE GF :sF lk-1 interaction. Heparin
was not required for the binding of VE GF to sF lk-1 in
this assay. T he obtained values will serve as baseline
parameters for the design of improved versions of re-
combinant soluble VE GF receptor.
© 1998 Academic Press
Vascular endothelial growth factor (VEGF) and its
receptors VEGFR1 (Flt-1, fms-like tyrosine kinase)
and VEGFR2 (KDR/Flk-1, kinase insert domain-
containing receptor/fetal liver kinase) have recently
attracted considerable interest because of their in-
volvement in vasculogenesis and angiogenesis (1–9).
Both VEGF and its receptors are overexpressed in a
number of angiogenesis dependent diseases such as
malignant tumors and diabetic retinopathy. The inhi-
bition of VEGF-mediated signals could thereforeoffer a
valuable approach for the therapy of these diseases.
Inhibition can be achieved through neutralizing anti-
bodies and soluble receptor constructs which block the
activation of VEGF receptors by preventing either li-
gand binding or ligand induced dimerization of recep-
tors. Previous reports have demonstrated that both
neutralizing VEGF and anti-KDR antibodies as well
as soluble VEGF receptor constructs are capable of
reducing or inhibiting tumor growth in animal models
It is now necessary to develop methods for express-
ing recombinant soluble VEGF receptor in high yields
for extensive preclinical and clinical testing. In this
report, we describe the expression of sFlk-1 in a bacu-
lovirus system and its purification by affinity chroma-
tography. The yield of highly purified receptor after a
single purification step was 30 mg/L supernatant.
Surface plasmon resonance was used to determine
the binding of the soluble recombinant receptor in real
time to its ligand VEGF. On- and off-rates for the
VEGF:sFlk-1 interaction, and the effect of heparin on
this interaction are reported.
MATERIALS AND METHODS
1. Construction of plasmids.
provided by Dr. Ihor Lemischka (Princeton University). This con-
struct contains the full length mouse Flk-1 cDNA. Double stranded
DNA encoding for the first 762 amino acids of Flk-1 was obtained by
PCR reaction using the primers GCACAGCTGATGGAGAGCAAG-
GCGCTGCTA and CTGCAGCTGCTATTCCAAGTTGGTCTTTTC
with the Flk-1 cDNA construct as a template. The resulting 2.3 kB
PCR fragment was cloned into pVL1393 (Pharmingen) and into
pBlueBacHis (InVitrogen). The DNA sequence was verified by auto-
mated fluorescence sequencing.
A Flk-1 cDNA construct was kindly
2. Expression of soluble Flk-1.
cultures of Spodoptera frugiperda (Sf9) cells were grown in Sf900II-
SFM (Gibco/BRL) supplemented with 50 ?g/ml penicillin and 50
?g/ml streptomycin (Gibco/BRL). Recombinant plasmids pVL1393/
sFlk-1 and pBlueBacHis/sFlk-1 were cotransformed with baculovi-
rus gold DNA (Pharmingen) intoSf9 cells by using lipofectin (Gibco/
BRL) according tothe manufacturer’s instructions. The recombinant
virus was plaque purified as described (18). To confirm the expres-
sion of recombinant protein, aliquots of conditioned medium of in-
fected Sf9 cells were analyzed by Western Blot using the monoclonal
antibody 1A8 directed against sFlk-1 (Brekken et al.; manuscript in
Monolayer and suspension cell
1To whom correspondence should be addressed at present ad-
dress: Maine Medical Center Research Institute, South Portland,
ME. Fax: (207) 761-9782. E-mail: HuangX@mail.mmc.org.
Abbreviations used: sFlk-1, soluble fetal liver kinase 1; VEGF121,
VEGF 165, vascular endothelial growth factor, 121 or 165 amino
acids length (splice variants); VEGFR, vascular endothelial growth
BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS 252, 643–648 (1998)
ARTICLE NO. RC989717
Copyright © 1998 by Academic Press
All rights of reproduction in any form reserved.
preparation). Sf9 cells were infected with sFlk-1 recombinant bacu-
lovirus at a multiplicity of infection (MOI) of 10. After incubation at
room temperature for 1 h the cells were pelleted and resuspended in
fresh culture medium. The cells were grown for various time periods
at 28 °C. The supernatant was removed and concentrated by mem-
brane filtration using a membrane with a 30 kD cut-off size. The
concentrated supernatant was then dialyzed against PBS pH?7.4
for further purification.
3. Optimization of expression conditions.
vector with the better yields, pVL1393/sFlk, we further optimized
the expression conditions. Sf9 cells were infected with sFlk-1-
recombinant virus at MOIs of 1, 2, 5 and 10. Samples of the super-
natants weretaken at 24h, 36h, 48h, 60h, 72h, and 96h post infection
and quantified by ELISA. For the ELISA, 96-well plates were coated
with 1A8 antibody (see above) and blocked with 5% Casein Acid
Hydrolysate (Sigma). Reference standards and samples were added
at various concentrations and incubated at room temperaturefor 1 h.
Affinity purified rabbit anti-Flk-1 polyclonal antibody (produced in
our laboratory) was used to detect sFlk-1.
Using the expression
4. Affinity purification.
raphy was prepared using the monoclonal antibody 1A8 (see above)
and protein G beads (Pharmacia). The antibody was coupled to the
beads as described (19).
Supernatant from virus-infected Sf9 cell cultures was concen-
trated by membrane filtration, dialyzed against PBS and loaded onto
the 1A8 affinity column. The column was washed with PBS pH?7.4
and soluble sFlk-1 was eluted with 0.1 M citric acid, pH?3.5. Absor-
bance at 280 nm was measured and pooled peak fractions were
dialyzed against PBS pH?7.4 and then concentrated by membrane
An affinity column for liquid chromatog-
5. SDS–PAGE analysis.
gradient gels using the Phastsystem (Pharmacia). To assess the
amount of glycosylation, samples were deglycosylated by digestion
with peptide N-glycosidase F (PNGase F, Boehringer Mannheim) for
12 h at 37 °C prior to loading on the gel.
SDS-PAGE was performed on 4-15%
6. Real time binding studies.
performed using the automated surface plasmon resonance based
measuring system Biacore2000 (Biosensor). Recombinant sFlk-1 was
immobilized on CM5 sensorchips (Biosensor) by amine coupling us-
ing an amine coupling kit (Biosensor) or ligand-thiol coupling using
PDEA (2-(2-pyridinyldithio)ethaneamine hydrochloride) as coupling
reagent (Biosensor) according to the manufacturer’s instructions.
For binding studies recombinant human VEGF165 (R&D), recombi-
nant murine VEGF165 (R&D) or recombinant human VEGF121
(R&D) was injected intoa flowcell with immobilized sFlk-1 at various
concentrations. 4 M MgCl2was used to regenerate the chip after
VEGF injection. A blank flowcell served as negative control and the
background signals observed in this flowcell were subtracted from
the signals in the sample flowcell. To verify the specificity of the
binding, sFlk-1 in 40fold, 200fold, and 1000fold molar excess was
coinjected with recombinant human or mouse VEGF.
To estimate the influence of mass transport, kinetic data were
obtained at different immobilization densities and at different flow-
rates (10 ?l/min, 30 ?l/min, 50 ?l/min and 70 ?l/min). An immobili-
zation level of 1300 RU and 30 ?l/min flow rate were used as stan-
To test for rebinding, the soluble receptor sFlk-1 was injected in
the dissociation phase of VEGF-binding, and the off-rate was com-
pared to the off-rate without the competitor sFlk-1.
The influence of heparin on the binding of VEGF to its soluble
receptor was tested by preincubation of heparin at various concen-
trations between 0.1 ng/ml and 10 ?g/ml with VEGF. Binding con-
stants were determined using the curve fitting software Biaevalua-
tion 2.1 (Biosensor).
Real time binding analysis was
1. Expression and Purification of Soluble Flk-1
Western Blot analysis of conditioned medium of Sf9
cells infected with recombinant baculovirus carrying
either pVL1393/sFlk-1 or pBlueBacHis/sFlk-1 revealed
a protein of MW?100 kDa (Figure 1). The signal ob-
tained with pVL1393 as a carrier was much stronger
than that with pBlueBacHis, and therefore this plas-
mid was used in the further experiments.
The results of the expression optimization studies
are illustrated in Figure 2. The expression of sFlk-1
was detected as early as 24 h post infection with max-
imal accumulation and minimal degradation of the
protein occurring at 60 h post infection at various in-
fection doses except for MOI?1, which has a maximum
production at 72 h post infection. We observed a
slightly increased yield at MOI?2, namely 35 mg/L.
The best expression conditions therefore are obtained
using the infection dose MOI?2 and harvesting 60
hours post infection. After affinity chromatography us-
ing immobilized 1A8, 30 mg of purified sFlk-1 could be
isolated per L of conditioned medium.
2. SDS–PAGE Analysis
The purified fractions were analyzed by SDS-PAGE.
As shown in Figure 3, the affinity purification yielded a
nearly homogeneous protein with a mass of 100 kDa.
Based on the slower than predicted mobility of sFlk-1
and the 16 potential N-linked glycosylation sites in the
cDNA derived amino acid sequence, it was reasoned
that sFlk-1 was posttranslationally modified by glyco-
sylation. In Figure 3, an increase in the mobility of
sFlk-1 from 100 kDa to 80 kDa can be seen after
incubation with PNGase F, demonstrating that sFlk-1
produced by the Sf9 cells is heavily glycosylated.
F IG. 1.
ples were separated by SDS-PAGE and analyzed by western blot
with affinity purified rabbit anti-Flk-1 antibody (left) or anti-Flk-1
monoclonal antibody 1A8 (right). Lane 1: purified seap/Flk-1 fusion
protein as a positive control; lane 2: conditioned medium from unin-
fected Sf9 cells; lane 3: medium from pBlueBacHis/sFlk-1 infected
Sf9 cells; lane 4: medium from pVL1393/sFlk-1 infected Sf9 cells.
SFlk-band at 100 kDa. NR: non reduced; R: reduced.
Western Blot analysis of the conditioned Sf9 cells. Sam-
Vol. 252, No. 3, 1998BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
3. Realtime Binding Studies
SFlk-1 was immobilized by either amino- or ligand-
thiol-coupling on the CM5 sensorchips at various den-
sities between 520 and 1900 RU (response units).
Amino-coupling was the preferred method, since Rmax
values (maximal response) were slightly higher and
off-rates slightly lower than with ligand-thiol coupling.
We compared the binding abilities of human and
mouse VEGF to the murine sFlk-1 and observed al-
most identical binding curves (Figure 4). The binding
could be completely inhibited by addition of a 1000-fold
excess of sFlk-1 to the injected VEGF (Figure 5). The
variations of calculated binding parameters were
within 20% (SD) of the mean value, when different
immobilization levels or different concentrations of
VEGF wereused. Variation of flowrates caused slightly
higher variabilities. The values for on- and off-rates
under different experimental conditions are shown in
Table 1. Addition of heparin to the injected VEGF
inhibited the binding of VEGF165 in a dose dependent
fashion (Figure 6). At a concentration of 10 ?g/ml hep-
arin the inhibition was about 50%. This was due to a
decrease in the on-rates, while the off-rates were not
significantly changed. No inhibition was seen at con-
centrations below 0.1 ?g/ml. When injections were pro-
longed to saturation the addition of 0.1 ?g/ml or 0.01
?g/ml slightly increased the maximal binding (up to
20%) despite decreased on-rates. VEGF121 binding
was not affected by the addition of heparin.
When sFlk-1 was injected in the dissociation phase,
off-rates were increased slightly, which indicates, that
rebinding occurs to some extent.
In this study we describe the high yield expression
and affinity purification of soluble VEGF-receptor 2
and the characterization of its binding to VEGF. Flk/
KDR is one of two known VEGF receptors. It has been
shown to be responsible for the transmission of a mi-
togenic signal to endothelial cells (20). Therefore Flk/
KDR is a potential target for angiogenesis inhibition.
We cloned and expressed the murine receptor Flk-1.
Cloning, expression and purification of the human ho-
molog KDR, can be performed in the same manner,
F IG. 2.
fected Sf9 cells. Insect cells were cultured and infected as described
in Materials and Methods. At the indicated times post infection,
samples were analyzed for sFlk-1 by quantitative ELISA.
Time course of sFlk-1 expression in pVL1393/Flk-1 in-
F IG. 3.
tant from pVL1393/sFlk-1 infected Sf9 cells were purified on a 1A8
affinity column with a bed volume of 10 ml (see Materials and
Methods). Samples were analyzed by SDS-PAGE followed by Coo-
massie Blue staining. The purified product was deglycosylated using
PNGase F. Lane 1: sFlk-1 after deglycosylation; lane 2: sFlk-1 before
deglycosylation; lane 3: molecular weight marker. Numbers on the
right indicate molecular mass in kDa.
Affinity purification of sFlk-1. 400 ml culture superna-
F IG. 4.
A: human VEGF165; concentrations are 100nM, 50nM, 10nM, 5nM,
1nM, 0.5nM, 0.1nM; B: mouse VEGF165; concentrations are 100nM,
50nM, 10nM, 5nM, 1nM, 0.1nM. Saturation of binding sites is
achieved between 10nM and 50nM in both cases.
Binding of recombinant VEGF165 toimmobilized sFlk-1.
Vol. 252, No. 3, 1998 BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
since the antibody used for affinity purification, 1A8,
also reacts with KDR (Brekken et al., manuscript in
preparation). In previous studies in which soluble
VEGF-receptor was expressed, tags had tobe added to
the receptor to provide a recognition site for affinity
purification. These tags include SEAP (secreted alka-
line phosphatase) and IgG fragments (21–24). Where
yields were reported, they ranged between 150 ?g/L
and 5 mg/L. Another recently reported tag is 6His
(Histidine), (17) for Ni-NTA affinity purification. In the
baculovirus system, the addition of a His-tag at the
N-terminus of a recombinant protein can inhibit the
secretion of the protein into the supernatant. This
might be a reason for the poor expression of His-
tagged sFlk-1, which we observed when using the
pBlueBacHis transfer vector. Using the pVL1393
transfer plasmid, which contains no tags, we obtained
at least 100-fold higher yields of recombinant protein
as determined by Western Blot (Figure 1). With 1A8, a
monoclonal antibody against Flk-1, we were able to
affinity purify sFlk-1 without using any tags. After a
single purification step we obtained nearly homoge-
nous protein with a yield of 30 mg/L crude superna-
To analyze the binding of the recombinant sFlk-1 to
its ligand VEGF, we used surface plasmon resonance
for two major reasons: First, surface plasmon reso-
nance is a cell free assay. Cell based assays are impor-
tant to understand the biology of VEGF binding and
signalling in its natural environment. The soluble re-
ceptor, however, is designed to capture VEGF in solu-
tion, where nomembrane environment is involved. For
our studies, a binding assay that lacks any cell mem-
brane components was warranted. Other cell free as-
says, where KDR-Fc fusion proteins were captured on
microtiter plates, have been successfully used (25), but
the sFlk-1 without an Fc-component cannot be immo-
bilized on microtiter-plates without impairing its bind-
ing capabilities (data not shown). Second, kinetic data
(on- and off-rates) can be determined as well, although
they are subject to restrictions due to problems intrin-
sic for the Biosensor systems used here (26). This is of
importance, since this method allows to screen im-
proved versions of VEGFR, as obtained by protein en-
gineering, with specific focus on the decrease of off-
rates. In addition, surface plasmon resonance allows
binding reactions to be studied in real time and no
secondary labeling reagents are required.
In a first set of experiments we showed that sFlk-1
binds specifically tohuman or mouseVEGF 165 (Figures
5 and 6). A curve fitting software (BIAevaluation, Bio-
sensor) was used to determine on- and off-rates from
the obtained binding curves. Dissociation constants
(KD) were calculated from the on- and off-rates.
F IG. 5.
excess of soluble sFlk-1. Concentrations of soluble sFlk-1 in solution
are 0, 40fold, 200fold and 1000fold molar excess over VEGF. Con-
centration of VEGF is 5nM. A: human VEGF165; B: mouse
Inhibition of VEGF binding to immobilized sFlk-1 by
Binding of VEGF to Immobilized sFlk-1 under Various Conditions Using Surface Plasmon Resonance
VEGF species and additions
hVEGF165 50nM ? Heparin (1 ?g/ml)
hVEGF165 5nM ? 1000 ? sFlk-1
mVEGF165 20nM ? 1 ?g/ml Heparin
mVEGF165 5nM ? 1000 ? sFlk-1
1.2 ? 106
3.3 ? 105
2.2 ? 106
1.9 ? 106
4.0 ? 105
2.2 ? 106
4.1 ? 10?4
3.3 ? 10?4
2.4 ? 10?4
6.2 ? 10?4
2.7 ? 10?4
3.0 ? 10?4
3.4 ? 10?10
1.0 ? 10?9
1.1 ? 10?10
3.3 ? 10?10
6.8 ? 10?10
1.4 ? 10?10
Note. Immobilization level in this experiment was 1700 response units. Response levels for the given concentration as well as binding
constants are shown.
Vol. 252, No. 3, 1998BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
In our system, the on-rates were 0.5-2.2 x 106
M?1s?1, and the off-rates were 2.0-6.2 x 10?4s?1with
no significant differences between human and mouse
VEGF165. The calculated values for the dissociation
constant KDranged between 1.1 x 10?10M and 6 x
10?10M. VEGF121 showed a lower affinity for sFlk-1,
due to a lower on-rate (data not shown). Previously
reported KDvalues for Flk-1/KDR binding to VEGF
were mostly obtained from cell based assays. The num-
bers range from 1 x 10?8M (for the proposed low affin-
ity interaction)/3 x 10?10M (for the high affinity inter-
action) toas low as 2.5 x 10?12M, depending on the cell
line used. In two studies, the binding of KDR-Fc to
VEGF was assessed in cell free assays (24, 25) with KD
values of 3-4 x 10?11M and 1.5 x 10?10M, respectively.
In a crosslinking assay Kendall (27) reported a KDof
approximately 1 nM for the VEGF:KDR interaction.
Previous data are not available for on- and off-rates of
the VEGF:VEGFR2 interaction.
We were alsointerested in assessing the influence of
heparin on the VEGF:VEGFR binding in our system,
because in a therapeutic setting heparin-like mole-
cules are not necessarily available at the site of
VEGF:sVEGFR binding. Heparin at concentrations
above0.1 ?g/ml dosedependently inhibited thebinding
by reducing the on-rate. This could be because heparin
either bound to the injected VEGF or to sFlk-1 and
impeded the rate of association of VEGF with its re-
ceptor. SFlk-1, when added to the injected VEGF in-
hibited its binding to immobilized sFlk-1 without ad-
dition of heparin. These findings strongly suggest that
heparin is not essential for the binding of VEGF to its
receptor, nor does it cement the sFlk-1:VEGF complex
once formed. Kaplan (24), who studied the KDR-Fc
binding to VEGF in a radioligand binding assay on
agarose beads, also found no requirement for heparin
in VEGF:VEGFR binding. The dissociation constant of
the VEGF:KDR-Fc interaction was slightly higher in
the presence of heparin (1.7 x 10?10M versus 1.0 x
10?10M). Similarly, Keyt (25) found that heparin was
not needed for VEGF:VEGFR binding in another cell
These findings are in contrast with the observa-
tions made in cell based assays (21, 28–30), where
heparinase treatment of VE GFR-bearing cells abol-
ishes their ability to bind VE GF. Also, cells deficient
in heparan sulfate biosynthesis cannot bind VE GF
(30). Addition of exogenous heparin can increase
VE GF binding when applied at an optimal concen-
tration which varies from cell line to cell line. High
concentrations of heparin (10 ?g/ml or above) gener-
ally inhibit the VE GF-binding. Gitay-Goren (28) re-
ported that an increase in binding of VE GF to its
receptor in the presence of heparin correlates with
an increase in the number of receptors, while de-
crease of binding correlates with a decreased recep-
tor number. A possible explanation for these findings
is that the negatively charged heparin molecules on
the surface of cells are required to present the recep-
tor in a form that can bind VE GF. If an excess of
heparin is added, receptor binding sites might be-
come covered and unavailable for VE GF binding.
In summary, we have developed a method to gen-
erate functional soluble VE GFR2 in high yields. We
also describe a nonradioactive method for determi-
nation of binding constants, including on- and off-
rates, for the VE GF receptor:ligand interaction. This
method can be easily applied in routine measure-
ments of improved versions of VE GFR obtained by
We thank Dr. Peter Schuck for helpful discussions.
1. Fong, G-H., Rossant, J ., Gertsenstein, M., and Breitman, M. L.
(1995) Nature 376, 66–70.
2. Shalaby, F., Rossant, J ., Yamaguchi, T. P., et al. (1995) Nature
3. Frank, S., Hubner, G., Breier, G., Longaker, M. T., Greenhalgh,
D. G., and Werner, S. (1995) J . Biol. Chem. 270, 12607–12613.
4. Burke, P. A., Lehmann-Bruinsma, K., and Powell, J . S. (1995)
Biochem. Biophys. Res. Comm. 207, 348–354.
5. Alon, T., Hemo, I., Itin, A., Pe’er, J ., Stone, J ., and Keshet, E.
(1995) Nature Med. 1, 1024–1028.
F IG. 6.
heparin. Human and mouse VEGF165 showed identical inhibition
curves. Shown here are the sensorgrams for mouse VEGF165 at
20nM (A) and 5 nM (B). Concentrations of heparin are 0, 0.1 ?g/ml,
1 ?g/ml, 10 ?g/ml (from top tobottom). Heparin concentrations lower
than 0.01 ?g/ml had no effect.
Inhibition of VEGF165 binding to immobilized sFlk-1 by
Vol. 252, No. 3, 1998BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS