395
Clinical Science (I998) 94,395-404 (Printed in Great Britain)
Vascular endothelial growth factor (VEGF) is released from
platelets during blood clotting: implications for measurement of
circulating VEGF levels in clinical disease
Nicholas J. A. WEBB*t, Martyn J.BOTTOMLEP, Carolyn j. WATSON* and Paul E. C. BRENCHLEY*
*ManChester Renal Research Group, Manchester Royal Infirmary and University Department of Medicine,
Manchester M I 3 OJH, U.K., and tDepartment of Paediatric Nephrology, Royal Manchester Children's
Hospital. Pendlebury, Manchester M 2 7 4HA. U.K.
(Received 26 Auystl3 I October 1997; accepted 5 November 1997)
1. Dysregulated vascular endothelial growth factor
(VEGF) expression has been reported in several
pathological states based upon evidence of elevated
serum VEGF levels. Using two immunoassays for
VEGF, this study determines normal plasma and
serum VEGF ranges, determines which are more
likely to reflect circulating VEGF levels and investigates a potential contribution of VEGF from platelets to VEGF levels detected in serum.
2. The presence of soluble VEGF receptor, sflt-1, at
a molar excess of 7:l significantly reduced measured VEGF levels in both assays. Serum VEGF
levels were higher than plasma levels in children
[(mean fS.E.M.) 306.1 f39.4 versus 107.4 k24.9
pg/ml, P <0.00011 and adults (249.4 f46.4 versus
Serum VEGF
76.1 f10.7 pg/ml, P <0.0001).
increased with clotting time (P= 0.0005 to compared
with 2 h samples); plasma VEGF levels were not
affected by time between sampling and centrifugation.
3. Calcium-induced clotting of platelet-rich but not
platelet-poor plasma induced VEGF release with a
proportional response between platelet count and
VEGF level and isolated platelets released significant quantities of VEGF upon incubation with
thrombin. Reverse transcriptase-PCR studies confirmed that platelets express VEGF121 and VEGFm
mRNA.
4. These data suggest that plasma is the preferred
medium to measure VEGF levels; a significant and
highly variable platelet-mediated secretion of VEGF
during the clotting process invalidates the use of
serum as an indicator of circulating VEGF levels in
disease states.
INTRODUCTION
Vascular endothelial growth factor (VEGF), also
known as vascular permeability factor, is a
disulphide-linked homodimer of 34-42 kDa. It is an
endothelial cell mitogen [ l ] which promotes angiogenesis and also has potent vascular permeability
enhancing properties [2]. Alternate splicing of
mRNA results in the generation of four protein
species of 121, 165, 189 and 206 amino acids. On the
basis of protein structure, particularly the pattern of
conserved cysteines and sequence similarity, VEGF
has been considered to be a member of the plateletderived growth factor protein family [3].
VEGF mediates biological function by interaction
with two specific tyrosine kinase receptors on endothelial cells: flt-1 (fms-like tyrosine kinase) [4] and
KDR (kinase insert domain containing receptor) [5],
also known as VEGFRl and VEGFR2 respectively.
The flt-I gene also undergoes alternate splicing to
produce both membrane-bound and soluble receptor variants. The soluble variant, sflt-1, has been
shown to neutralize the effects of VEGF in vitro [6],
although as yet there is no clear evidence to support
a similar role in vivo.
Dysregulated VEGF expression has been implicated in a number of pathological situations, including tumour angiogenesis/metastasis [7], rheumatoid
arthritis [8-lo], diabetic retinopathy [ l l , 121, glomerular disease [131 and ovarian hyperstimulation
syndrome [14]. Elevated serum VEGF levels have
also been detected in women with pre-eclampsia
[15] and in patients with a variety of cancers [16].
The ability to measure circulating VEGF protein
levels is therefore clearly of importance to clinical
researchers in a wide range of specialities.
A variety of immunoassays using combinations of
polyclonal and monoclonal antibodies have been
described for measurement of VEGF, predominantly in serum, by a number of authors [16-191. It
is unclear how these assays relate to each other in
terms of sensitivity and specificity for different forms
of VEGF and no comparison of different recombinant VEGF (rVEGF) standards has been undertaken. Importantly, there has been no validation of
the use of serum VEGF as a suitable indicator of
Key words: immunoassays. platelets, vascular endothelial growth factor, vascular permeability factor.
:smAb, monoclonal antibody; PBMC. periphd Mood mononuclear cell; RT, m r s e tranariptase;VEGF, vascular endothelialgrowth factor.
Comspdrme: Dr N K M]A
~Webb.
396
N. J. A. Webb et al.
circulating VEGF levels. A recent study showing
that stimulated platelets secrete VEGF [20] raises
the possibility that VEGF detected in serum is an
artefactual consequence of platelet activation during
blood clotting and that only plasma values reflect
the circulating level of VEGF.
The aims of this study were to (i) describe a
soluble receptor-mediated capture assay for VEGF,
compare its performance with that of a monoclonalantibody (mAb)-mediated capture assay and investigate the potential for inhibition of VEGF signal by
the presence of sflt-1 in both assays, (ii) determine
the normal paediatric and adult VEGF ranges in
plasma and serum using these two different assays,
(iii) determine whether plasma or serum measurements are more likely to reflect circulating VEGF
levels and (iv) to investigate a potential contribution
from platelets to measured VEGF levels in serum.
MATERIALS AND METHODS
Recombinant (baculovirus-derived) VEGF16s and
sflt-1 were kindly donated by Dr Don Ogilvie,
Zeneca Pharmaceuticals (Alderley Edge, U.K.).
VEGFm, VEGF ELISA kits, recombinant interleukin-1, 2, 3, 4, 6, 8 and 10, tumour necrosis factorcc and interferon-y were all purchased from R&D
Systems (Abingdon, U.K.). Peroxidase-conjugated
donkey anti-rabbit and goat anti-mouse antibodies
were purchased from Jackson ImmunoResearch
(Luton, U.K.). Mouse monoclonal anti-VEGF,
phorbol 12-myristate 13-acetate and thrombin were
all purchased from Sigma Chemical Co (Poole,
U.K.). Optiprep was purchased from Robins Scientific (Knowle, U.K.). U937 cells were obtained from
the European Collection of Animal Cell Cultures
(Porton Down, U.K.). Taq DNA polymerase, reverse
transcriptase, RNase inhibitor, oligo(dT)12-18,100 bp
DNA ladder and dNTP solutions were purchased
from Life Technologies (Paisley, U.K.).
Production of poiycionai antibodies
Male New Zealand White rabbits were immunized by subcutaneous injection of 50 pg of rVEGF;
booster doses of 50 pg were given at 4 and 8 weeks.
IgG isolation was performed by affinity purification
by passage over a Hitrap column coated with
rVEGF. Bound antibodies were eluted from the
Hitrap-VEGF column with 0.1 mM glycine, pH 2.5,
neutralized with solid Tris to pH7.0 and concentrated to approximately 0.5 mg/ml. All animal care
was in accordance with U.K. Home Office guidelines.
Assay techniques
fIt-I capture assay. sflt-1 (1 ,ug/ml in 0.05 M
carbonate buffer, pH 9.6, 100 pl/well) was adsorbed
on to 96-well plates (MicroFLUOR W, Dynatech
Laboratories U.S.A.) for a minimum of 16 h at 4°C.
The plates were washed 10 times in wash buffer
(0.1 M PBS, pH 7.2, 0.05% Tween 20) before and
after being blocked with assay buffer (5% BSA in
PBS, 150,ul/well) for 1 h on an agitator at room
temperature. The wells were then filled with either
samples, standards or controls (100 plhvell). Plasma
or serum samples were added undiluted. The
standard curve was generated using Zeneca rVEGF
[rVEGF(Zen)] diluted in assay buffer and ranged
from 19 pg/ml to 10 ng/ml in nine serial 2-fold steps.
Background wells contained assay buffer alone. The
culture supernatant from U937 cells (monocytic cell
line, ECACC no. 85011440) stimulated with phorbol
12-myristate 13-acetate, 5 ng/ml for 22 h, acted as a
positive control for ascertainment of inter- and
intra-plate variability. Plates were incubated for 16 h
at 4°C after which they were washed 10 times in
wash buffer. Affinity-purified rabbit-anti-VEGF
polyclonal antibody (dilution 1:500 in assay buffer,
100 pl/well) was then added for 4 h on an agitator at
room temperature followed by a further 10 washes.
The peroxidase-conjugated goat anti-rabbit IgG
(heavy- and light-chain-specific) diluted 1:500 in
assay buffer, lOOplhvell, was then added and the
plate was incubated at room temperature on an agitator for 2 h followed by a further 10 washes. The
plate was then developed using the Amerlite buffer
and tablet packs (Amersham International, U.K.)
according to the manufacturer’s instructions and
read on an Amerlite enhanced luminescence microtitre plate reader (Berthold, U.K.). To assess the
assay specificity, plasma from a healthy adult shown
to have undetectable levels of VEGF and a 5% BSA
solution were spiked with the following cytokines at
a concentration of 5 ng/ml; interleukin-1, 2, 3, 4, 6, 8
and 10, tumour necrosis factor-cc and interferon-y.
Samples were then assayed.
db-mediated capture assay. This was performed
according to the manufacturer’s instructions (R&D
Systems). Briefly, plasma, serum or cell culture
supernatant samples and a rVEGF [rVEGF(R&D)]
standard curve (31.2 pg/ml to 2 ng/ml) were added
in assay diluent to precoated wells and incubated for
2 h at room temperature. After a standard washing
step, anti-VEGF conjugate was added to each well
and incubated for 2 h at room temperature. After a
further washing step, substrate solution was added
to each well for 20 or 25 min, after which stop solution was added and the plate read on a Molecular
Devices ThermoMax plate reader at 450 nm. The
culture supernatant from U937 cells as described
above was also used to ascertain inter- and intraplate variability with this assay. The assay has been
shown to be specific for VEGF with no interference
from a range of cytokines tested at 50 ng/ml.
As different rVEGF sources were being used to
generate the standard curves in these assays
[rVEGF(Zen) with the flt-1 capture assay and
rVEGF(R&D) with the mAb-mediated capture
Platelet release of vascular endothelial growth factor
assay], to allow comparison of the two assays, a
standard curve from 31.2pglml to 2nglml and 12
samples at 500pg/ml of each rVEGF were measured using the mAb-mediated capture assay. To
determine the effect of the addition of sflt-1 on both
assay systems and to help determine whether both
assay systems recognize both VEGF and VEGF
complexed to flt-1, serum (clotting time 2 h) from a
normal control subject known to have a high serum
VEGF level (750 pg/ml) was incubated in 1 ml aliquots for 2 h at room temperature with sflt-1 at concentrations varying between 1.25 and 80 nglml
[molar ratios (flt-1:VEGF) of approx 1: 1 to 50: I].
Samples were then measured in duplicate on both
assay systems.
Determinationof normal paediatric and adult VEGF
ranges; sample collection and storage
Blood samples were collected from healthy adult
volunteers (n = 34) and also from otherwise healthy
children undergoing minor surgical procedures
under general anaesthesia (n = 19). Paediatric
samples were taken immediately after the induction
of anaesthesia before the commencement of surgery.
A separate venepuncture site was chosen rather
than using the intravenous cannula to obtain
samples to avoid any contamination by intravenous
anaesthetic agents. Whole blood was collected into
Becton and Dickinson (B&D) plain glass and EDTA
(0.072ml of 7.5% EDTA per 3 m l sample) blood
tubes at room temperature, and centrifugation of
samples and separation of serum and plasma was
performed within 2 h of venepuncture. Serum and
plasma samples were stored in multiple aliquots at
-80°C before assaying using both assay systems.
The approval of Salford Health Authority Ethical
Committee was obtained for the collection of blood
samples from children.
Determination of whether plasma or serum
measurements are more likely to reflect circulating
VEGF levels
Whole blood samples (2 ml) were collected from
12 healthy adult volunteers into B&D plain glass
tubes. One 2 ml sample was centrifuged immediately
for 10min at 2000rpm and the supernatant
removed and then microfuged to remove any contamination from erythrocytes. The remaining
samples were allowed to clot at room temperature
for 2, 4 and 24 h. Whole blood samples were also
collected from 10 of these 12 adults into B&D
EDTA tubes; again, one sample was centrifuged
immediately and the others left standing at room
temperature for 2, 4 and 24 h. At the various time
points, samples were centrifuged for 10 min at 523g
and the serum or plasma removed and stored as
multiple aliquots at - 80°C before assaying using the
397
mAb-mediated capture assay. To determine whether
there was any effect of temperature, the serum clotting experiment was repeated in six individuals with
whole blood samples left to clot at both room temperature and 37°C.
To assess the stability of VEGF to freeze-thaw
cycles, a serum sample (clotting time 2 h) was separated into seven aliquots which were subjected to
between 0 and 6 freeze-thaw cycles (room temperature to -80°C). Serum VEGF levels were then
measured using the mAb-mediated capture assay.
Investigation of a potential contribution from platelets to
measured VEGF levels in serum
To investigate whether release of intra-cytoplasmically stored or membrane-bound VEGF during the
clotting process contributes to VEGF levels detected
in serum, 3 ml EDTA blood samples were taken
from six adult volunteers. The samples were microfuged at 13000 rpm for 10 min in order to obtain a
cell pellet containing erythrocytes, leucocytes and
platelets. The platelet-poor plasma was then
removed and the cell pellet washed three times in
PBS to remove any remaining plasma or plasma
products. The pellet was then respun and subjected
to five consecutive freeze-thaw cycles from -80°C
to room temperature. After the final thaw, the cells
were resuspended to a total volume of 3 ml in assay
buffer and revolved on a rolling plate for 15 min at
room temperature before respinning and assaying of
the supernatant using the mAb-mediated capture
assay.
To investigate whether platelets contribute to the
release of VEGF during the clotting process, whole
blood samples were taken from six adult volunteers
by free flow into citrated blood tubes via an
indwelling 19 G cannula (thus avoiding the activation of platelets which occurs when blood samples
are drawn into a syringe via a narrow bore needle).
Samples were centrifuged at 170g for 10min and
the platelet-rich plasma removed and then divided
into two portions. One portion was centrifuged
again at 2700g for 15 min to produce platelet-poor
plasma [21]. Coulter counter analysis (Technicon
H3) was performed on these plasma samples to
quantify the platelet content and the degree of
leucocyte contamination. One-millilitre samples of
both platelet-rich and platelet-poor plasma were
then placed in glass tubes and calcium chloride was
added to a final concentration of 4 m M to induce
clotting at room temperature. Unstimulated samples
acted as controls. At time 0 (to) and at 2 h, samples
were centrifuged and the plasma/serum removed
and frozen at -80°C before assaying using the
mAb-mediated capture assay. For each individual,
two parallel 2 ml whole blood samples collected in a
plain glass tube were either centrifuged immediately
or allowed to clot for 2 h before centrifugation and
removal of the serum. In a similar experiment, to
398
N.J. A. Webb et al.
confirm whether a platelet ‘dose-response’ was
present, platelet-rich plasma was serially diluted
with autologous platelet-poor plasma to achieve
final platelet concentrations of between 4 and
150 x 109A and calcium chloride (4 mM final concentration) then similarly added to all samples.
In further experiments, after venepuncture of
healthy adult volunteers (n = 10) using free flow
from an indwelling 19 G cannula into citrated blood
tubes, pure platelet populations were isolated using
OptiprepTM density gradient centrifugation. The
platelet count and purity were established using
Coulter counter analysis. One-millilitre aliquots of
isolated platelets were incubated with thrombin
(10 units/ml) for 2 h at 37”C/5% C02, with unstimulated platelets incubated under identical conditions
acting as a control. Thrombin activity at this concentration was confirmed, the thrombin time being 18 s
in a standard blood clotting assay. Two parallel
whole blood samples collected from each individual
into plain glass tubes were either centrifuged
immediately or allowed to clot for 2 h at room temperature. At time 0 (to) and at 2 h, samples were
centrifuged and the plasma/serum removed and frozen at -80°C before assaying using the mAbmediated capture assay.
To investigate the expression of VEGF mRNA in
platelets, pure platelet populations were isolated as
above from three healthy adult volunteers, platelet
count and purity being established using Coulter
counter analysis. Total RNA was extracted using a
modification of the guanidinium thiocyanate-phenol-chloroform method [22]. cDNA was synthesized
from 5 pg of total RNA using superscript reverse
transcriptasem and oligo(dT)12-18 according to the
manufacturer’s instructions. PCR amplification was
performed using previously described primers known
to amplify all reported VEGF splice variants [23] and
CTAP-III[24], and B-actin primers designed to amplif j cDNA sequence and not related processed pseudogene sequences [forward primer 5’-GCC GTC
TIC CCC TCC ATC-3’ (nucleotides 125-143);
reverse primer 5‘-TAG CAA CGT ACA TGG CTG
GGG-3’ (nucleotides 446-427, Genbank accession
no. X00351), product size 322 bp]. PCR amplification was carried out in a Perkin-Elmer 480 DNA
thermal cycler in 20 pl reactions overlaid with paraffin oil containing 1/25th (4%) of the volume of the
cDNA reaction: 10 pmol of each primer, 0.75 mM of
each dNTP, 10% DMSO, 3.5 mM MgClz, 16.6 mM
(NH4)zS04, 67 mM TrisMCl pH 8.0, 85 pg/ml BSA
and 0.5 units of Tuq DNA polymerase. Thermal cycling conditions consisted of an initial denaturation
step at 94°C for 3min followed by 40 cycles for
VEGF and 35 cycles for CTAP-I11 cDNA amplification at 94°C for 1 min, 55°C for 1min, 72°C for
1 min; and 30 cycles for B-actin cDNA amplification
at 94°C for 1 min, 65°C for 1 min, 72°C for 1 min;
and a final extension step at 72°C for 5 min. PCR
products and a 100 bp DNA ladder (Life Technologies) were electrophoresed through 2% agarose gels,
stained with ethidium bromide and visualized and
photographed under UV illumination.
Statistical analysis
Paired and unpaired t-tests were used where data
were determined to be normally distributed using
the Kolmogorov-Smirnov test. The Wilcoxon matched pairs test was used for paired non-parametric
data. Calculations were performed using GraphPad
PrismTM,GraphPad Software Inc, U.S.A.
RESULTS
Characteristics of the VEGF assays
The concentrations and orientation of the capture
and detection proteins and antibodies for the flt-1
capture assay were determined and optimized in
preliminary experiments. The resultant assay protocol is described in the Materials and Methods section and typical standard curves generated using
rvEGF165 and ~ V E G F I Zare
I shown in Fig. 1. The
usual working range of the flt-1 capture assay was
78 pg/ml-10 ng/ml using VEGFMSfor the standard
curve; when VEGFlzl was used, the standard curve
showed saturation binding above 2.5 ng/ml. The
inter-plate coefficient of variation, calculated from
results of a total of 18 assays, was 11.49% with an
intra-plate coefficient of variation, calculated using
10 measurements on a single plate, of 4.52%. The
assay was found to be specific for VEGF, with cytokine/growth factor-spiked plasma and 5% BSA producing background values only (results not shown).
The mAb-mediated capture assay is reported to
detect both V E G F I ~and
I VEGF165 and to be sensitive to 5-10 pg/ml, although the recommended
standard curves have minimum standards of
31.2pg/ml for use with plasma and serum samples
and 15.6 pg/ml for culture supernatants. The inter-
10
100
1000
10000
VEGF (pg/ml)
Fig. 1. Standard cuwes for Rt-l capture assay. Standard curves for
rMGF16 and ~VEGFIII (39pglrnl-lOnglml) in 5% BSA were measured
using the nt-l capture assay. A,MGFIII; m. M G F I ~ s .
399
Platelet release of vascular endothelial growth factor
0
1
Added flt-l(ng/ml)
VEGF (pg/ml)
Fig. I. Comparison of scandad cuwes using both rVEGFIbs(R&D) and
rVEGFI&en). Standard curves using rMGF16s(R&D) and rMGF~t.~(zen)
(3 I pg/ml-2 nglml) were measured using the &mediated capture assay.
A, rMGF(Zen); m, M G F (MID).
plate coefficient of variation, calculated from results
of a total of 16 assays, was 16.1%.
That both recombinant VEGF standards behaved
identically in the mAb-mediated capture assay is
shown in Fig. 2, with standard curves for both proteins diluting out similarly. The rVEGF(Zen) used
in the flt-1 capture assay, when read on the mAbmediated capture assay, produced higher mean
values than the rVEGF(R&D) (649 & 22.4 pg/ml
versus 523 f 17.2 pg/ml; P = 0.0002, unpaired t-test).
The equivalence ratio for rVEGF(Zen)/rVEGF
(R&D) standards in this assay is therefore 1.24:l.
Furthermore, on measuring plasma samples containing significant quantities of VEGF in the two different assays, measured values were marginally higher
in the mAb-mediated capture assay (r2= 0.646, P =
0.0029) (Fig. 3). The U937 supernatant positive
control produced higher readings in the mAb-mediated
capture assay (1502 f 60.4 pg/ml) than in the flt-1
r(
J u
3001
- 1
0
Iho
2ho
10 20 30 40 50 60 70 80
3hO
VEGF levels in mbmediated
FQ. 4. Elica of a d d i i n of flt-l on measured VEGF kveb in swum
containing VEGF. &-I [1.25-8Onglml (mdar ratios I: l-SO:l)] was
added to serum known to contain significant amounts of M G F (750 pg/ml).
After incubationfor 2 h at room temperature, samples were measured using
both the &mediated
capture assay and the fk-l capture assay. M G F
levels are shown as a percentage of the valw obtained in serum with no
added llt-I. A, llt-I capture assay; m. &mediated capture assay.
capture assay (980.2 k26.6 pg/ml] ( P < 0.0001, unpaired t-test).
The effect of adding flt-1 to serum known to contain significant amounts of VEGF is shown in Fig. 4;
measured VEGF values fell significantly with
increasing amounts of added flt-1 in both the mAbmediated and the flt-1 assays, showing 50% inhibition at a molar ratio of flt-lNEGF of approximately
7 : 1. Therefore, both assays for VEGF are sensitive
to the presence of soluble flt-1 and do not allow
detection of VEGF in the form of a fltl-VEGF
complex.
Normal paediatric and adult plasma and serum VEGF
ranges
Using both the mAb-mediated and flt-1 capture
assays, paired plasma and serum samples were measured in a total of 19 children. The results are
shown in Table 1. Serum VEGF levels were significantly higher than those in paired plasma samples
[P<0.0001, mAb-mediated capture assay; P =
0.0004, flt-1 receptor capture assay (paired t-tests)].
Table 2 shows normal adult plasma and serum (clotting time 2 h) levels from a total of 34 adults.
Twenty of these plasma samples were measured
Table 1. N o d values of VEGF in p a e d i i paired plasma and
serum samples. Values are means (SEM). Statistical significance:
*P<O.OOOI serum versus plasma levels (paired t-test); t P = O.OOO4 serum
versus plasma levels (paired t-test). #All values Mow lower detection limit
for assay.
capture assay (pg/ml)
Fig. 3.
capture and Rt- I assays. Plasma
samples from MKmal adults were measured using both assays. samples
where circulating MGF levels were above the background detection limits
for both assays were compared using linear repsion (rz=0.646,
P = 0.0029).
M G F (pg/ml)
Comparison of & m c d i
&mediated
Plasma
Serum
107.4 (24.9)*
306.1 (39.4).
assay (n = 19)
llt-l captureay(n=15)
<78t#
200.0 (Z6.4)t
400
N. J. A. Webb et al.
Table 2. N d values of VEGF in adult plasma and serum samples.
Values are means (SEM). Statistical significance: *P = 0.32, & WKUS tlt- I
assay (20 paired plasma samples, paired t-test); tPi0.0001 serum versus
plasma (unpaired 1-test).
700
600
z
MGF (pg/ml)
&-mediated
assay (n = 34)
fi-I Capture aSSaY (fl = 20)
76. I(10.7)'t
249.4 (46.4)t
Plasma
Serum
3
400
c*
8
108.3 (I1.5).
300
200
/---
100
I
-
0
using both assays; no significant difference was
found in measured values between the two assays.
Similar to the paediatric findings, serum levels were
significantly higher than plasma levels (P<0.0001
unpaired t-test).
I
B
400
t%
E
300
4
8
820
22
24
Fig. 6. Effect of time from sampling to centrifugation on plasma VEGF
levels in normal adults. EDTA-anticoagulatedwhole Mood samples (2 ml)
from normal adults (n = 10) were collected and either centrifuged immediately (to sample) or allowed to stand at mom temperature for varying time
pints up to 24 h. At the various time points samples were centrifuged and
plasma levels measured Hing the mAb-mediated capture assay. P not significant, to versus 2 h.
Do plasma or serum measurements reflect circulating
8
2
l l m c (hours)
VEGF levels?
The effect of clotting time on serum VEGF levels
in 12 healthy adults is shown in Fig. 5 . A significant
rise in VEGF levels at 2 h compared with to values
(P = 0.0005, Wilcoxon matched pairs test) was
detected, with these elevated values persisting at 4
and 24 h. Two-hour serum VEGF levels were significantly higher in samples left to clot at 37°C
[402.7 pg/ml (interquartile range 256.9-535.8)] compared with those left at room temperature [300.9 pg/
ml (interquartile range 185.9-441.4)] (I' = 0.03, Wilcoxon matched pairs test). Plasma samples from 10
of these 12 adults left for similar time intervals
before centrifugation and the removal of plasma
showed no significant difference between to and 2 h
and later samples (P = 0.11 paired t-test) (Fig. 6).
VEGF in serum samples was stable after up to six
=
(4kr)
-
repeat freeze-thaw cycles with no significant change
in measured values (coefficient of variation 9.3%).
Contribution of platelets to VEGF levels in serum
Supernatant VEGF levels from total blood cell
samples (erythrocytes, leucocytes and platelets) subjected to five freeze-thaw cycles (n = 6 normal
adults) were significantly increased (median 172.7
pg/ml, interquartile range 155.0-198.2 pg/ml) compared with levels in to platelet-poor plasma (median
39.8 pg/ml, interquartile range 36.0-71.5 pg/ml;
P = 0.03, Wilcoxon matched pairs test). The addition
of calcium to platelet-rich plasma (median platelet
count 117 x 109/1, leucocyte contamination <0.1%)
resulted in significantly elevated levels of VEGF
release after clotting compared with platelet-poor
plasma (median platelet count 9 x lo9& leucocyte
contamination t0.1%). VEGF levels in parallel
control serum from whole blood samples are shown
(Table 3). To provide further supportive evidence
for the contribution of platelets to the increased
VEGF levels detected in serum compared with
plasma samples, VEGF levels in platelet-rich plasma
made to clot by the addition of calcium chloride fell
200
Table 3. VEGF levels after calcium chlorideinduced clotting of
platelet-poor and platelet-rich p h a samples (n = 6 sub-).
Values
are medians (interquartie ranges). Statistical significance: *P = 0.03,
Wilcoxon matched pairs test; t P = 0.03, Wilcoxon matched pairs test. #All
values below lower detection limit for assay.
100
0
I
1
I
2
4
6
-1
020
22
24
Time (hours)
Fig. 5. Effect of clotting time on serum VEGF kvek in normal adults.
Whole blood samples (2 ml) from normal adults (n = 12) were collected and
either centrifuged immediately (to sample) or allowed to clot at room temperature for ~ l y l n gtime points up to 24 h. At the various time points
samples were centrifuged and serum levels measured using the mAbmediated capture assay. Statistical significance: P = O.ooOS, to WKUS 2 h
(Wilcoxon matched pairs test).
MGF (pg/ml)
to
2 h control
ZhCaCh(4mM)
Platelet-poor plasma
Platelet-rich plasma
Normal serum
34 (<31-56.5)
<31#
<3 I#
<31t#
I14 (85-179.5)*t
48.5 (<31-81.5)
244 (203-382.5)
<31#*
Platelet release of vascular endothelial growth factor
6ool
O!
0
40 I
a
.
h
I
25
50
1
75
100
125
1
I
150
175
01
0
as the platelet count of the platelet-rich plasma fell
secondary to dilution with platelet-poor plasma (Fig.
7)Isolated platelets (mean platelet count 142 x 10y/l,
mean leucocyte contamination 0.41%) generated
significant quantities of VEGF upon stimulation
with thrombin compared with unstimulated control
cells. VEGF levels in parallel serum from whole
blood samples are shown (Table 4); a highly significant relationship was detected between 2 h serum
VEGF levels and 2 h thrombin-induced platelet
VEGF release (Fig. 8). Reverse transcriptase
(RT)-PCR of mRNA from a pure population
(leucocyte contamination < 0.1%) of freshly isolated
human platelets from three normal adults showed
low but detectable PCR product corresponding to
mRNA for the two soluble VEGF splice variants,
V E G F ~ Zand
I VEGFlbs (Fig. 9). To eliminate contaminating leucocyte mRNA as the source of this
detected PCR product, peripheral blood mononuclear cell (PBMC) cDNA was amplified by PCR
using identical conditions to those used for the
platelet cDNA amplification. PBMCs, which have
previously been shown to express VEGF mRNA
[23] and which were shown to be the predominant
500
1
I
t
750
1000
1250
Serum VEGF (pg/ml)
Platelet count (x 10~11)
Fig. 7. Effect of platelet count on VEGF released after the addition of
cakium chloride t o platelet-rich platma. Platelet-rich plasma (platelet
count I50 x 109/1) was serially diluted with autologous platelet-poor plasma
(platelet count 4 x 109/l). Calcium chloride (final concentration 4 mM) was
added to each sample which was then allowed to stand at room temperature for 2 h. After centrifugation, plasma VEGF levels were measured using
the mAb-mediatedcapture assay.
I
250
Fig. 8. Comparison of serum VEGF kvels and thrombin-induced
VEGF release by platelets in normal adults. Isolated platelets from
normal adults (n = 10) were stimulated with thrombin (10 unitslml) for 2 h
at 3PC/5% CO2. A parallel whole blood sample collected at the same time
from each donor was allowed to clot for 2 h at r w m temperature. Samples
were then centrifuged and the supernatantlserum removed. VEGF levels
were measured using the mAb-mediated capture assay. Statistical significance: r2 = 0.766, P = 0.0009.
contaminating leucocyte in the platelet preparations
by Coulter counter analysis, were isolated from
whole blood (n = 3 normal adults) using standard
Ficoll density gradient centrifugation. Although the
level of contamination of the platelet preparation by
PBMCs was low (<0.1% by Coulter counter analysis), the relatively higher mRNA content of PBMCs
1 2 3 4 5 6 7
VEGF
r( 607bp
4- 535bp
+403bp
CTAP-III
4- 387bp
Table 4. VEGF levels after thrombin-induced clotting of purified
platelet samples (n = 10 sub-).
Values are medians (interquartile
ranges). Statistical significance: *P = 0.0156, Wilcoxon matched pairs test.
#All values below lower detection limit for assay.
VEGF (pg/ml)
10
2 h control
2 h thrombin (IOunits/ml)
Platelets
Normal serum
<31#
<31#*
116.5 (41-192.5)*
32 ( < 3 1-59.5)
203.5 (I54-349.5)
Fig. 9. RT-PCR analysis of VEGF, CTAP-Ill and !-actin expression in
human platelets. RT-PCR products were elecvophoresed through 2%
agarose gels and stained with ethidium bromide. Lanes I and 7 contain I pg
of a 100 bp DNA ladder (Life Technologies); lanes 2, 3 and 4 represent
RT-KR products after amplification of platelet cDNA from three different
donors: lane 5 is a positive control (paediatric kidney) and lane 6 is a negative control (no cDNA template). Bands at 403,535and 607 bp correspond
to VEGFIII, VEGFM and VEGFtm respectively.
N. J. A. Webb et al.
402
could theoretically produce a higher percentage contamination of the resultant cDNA preparation. To
eliminate the possibility that PBMC-derived cDNA
could be responsible for the observed PCR result,
we tested PBMC-derived cDNA (Fig. 10A) and
PBMC-derived cDNA diluted 1:20 and equivalent
to 5% RNA contamination of the platelet preparation (Fig. 10B) under identical PCR conditions. As
shown in Fig. 10B, no PCR product was detected,
thus eliminating contaminating PBMCs as the
source of the VEGF mRNA in Fig. 9.
DISCUSSION
The two assays described here have been shown
to be specific for VEGF and to measure both
soluble secreted VEGF forms, VEGFIZI and
VEGF165. The addition of flt-1 to serum containing
a high concentration of VEGF reduced the detected
VEGF level with a clear dose-response being
observed; these findings show that neither the mAbmediated capture assay nor the soluble receptor
capture assay detect circulating VEGF-sflt-1 complexes. It is likely, therefore, that the anti-VEGF
antibody utilized in the mAb-mediated capture assay
recognizes an epitope situated within or close to the
flt-1 binding site so that the epitope is masked when
VEGF is bound to flt-1. Both assays measured similar VEGF levels in normal human plasma and the
measured difference in rVEGF at 500pg/ml was
1
2
3
4
5
6
7
Fig. 10. RT-PCR analysin of VEGF expression in human PBMCs.
RT-PCR products were electrophoresed through 2% agarm gels and
stained with ethidium bromide. Lanes I and 7 contain Ip g of a I00 bp DNA
ladder (Life Technologies); lanes 2, 3 and 4 represent RT-PCR products
after amplification of PBMC cDNA from three different donors; lane 5 is a
positive control (paediatric kidney) and lane 6 is a negative control (no
cDNA template). A RT-PCR products of PBMC-derived cDW B.
RT-PCR products of PBMC-derived cDNA diluted I :20 and equdent to
5% RNA contamination of the platelet preparation (both performed under
identical reaction conditions). Bands at 403, 535 and 607 bp correspond
with VEGFIII, MGFtss and V E G F I ~respectively.
small. We have previously tested other combinations
of antibodies in an attempt to produce a more sensitive assay, in particular an affinity purified chicken
polyclonal anti-VEGF (raised in-house) as capture
and a monoclonal anti-VEGF antibody (Sigma
V4758) as detection reagent, but we have been
unable to improve significantly on the mAbmediated capture assay described above (N.J.A.
Webb, M.J. Bottomley, C.J. Watson and P.E.C.
Brenchley, unpublished work).
We have reported for the first time the normal
ranges of VEGF in children, and have shown
normal range data in adult plasma samples. Previous
publications have reported normal adult serum
levels, although no comment was made about clotting times [19].
In attempting to establish normal ranges and to
quantify VEGF in clinical samples by immunoassay,
the first requirement is to confirm the validity of
using either plasma or serum values and to determine whether either medium is sensitive to artefactual VEGF production or loss during or after
sample collection. Our results in this study clearly
demonstrate a significant difference in the normal
values of VEGF for an individual depending on
whether plasma or serum VEGF is measured, an
issue which other published assay reports have not
addressed. Serum values are invariably higher by a
factor of approximately 6-fold with an apparent variable generation of VEGF signal between individuals
over time for up to 4 h. This difference between
serum and plasma VEGF levels has been noted with
both the assays reported here and also with an
in-house antibody capture assay using chicken antiVEGF and monoclonal anti-VEGF reagents (results
not shown). In a clinical situation, where blood
samples are taken and left for variable times before
processing, the contribution from the clotting process would effectively rule out the use of serum
measurement. However, even if there was strict uniformity of clotting time for all samples, the large
interpersonal variation in generation of VEGF in
clotted samples makes interpretation of any
observed difference between disease and control
groups very difficult and may invalidate the results.
This study is, therefore, the first to suggest that
plasma is the preferred medium in which to measure
in vivo circulating VEGF levels and to establish
normal range data in plasma for both children and
adults. Previous studies reporting elevated serum
VEGF levels in various disease states compared
with normal controls therefore need to be reviewed
in light of these findings. There are an increasing
number of reports of the use of rVEGF in human
subjects as therapy for ischaemic injury [25, 261
which underlines the importance of accurately measuring circulating levels of the protein, taking into
account the presence of sflt-1 which, if conjugated to
VEGF, will prevent its detection with existing assay
systems.
VEGF could be generated during blood clotting
Platelet release of vascular endothelial growth factor
as a result of several mechanisms. Initially, we
investigated the potential release of a preformed
cytoplasmic or membrane-bound pool of VEGF
from blood cells. Freeze-thawing of blood cell
pellets generated a significant 4-fold increase in
VEGF compared with to plasma levels, suggesting
that a potential reservoir of cytosolic or membranebound VEGF exists. When plasma was made to clot
by the addition of calcium chloride, VEGF was
released into the serum in significant amounts with
platelet-rich but not with platelet-poor plasma, with
a clear dose-response being seen. Purified platelets
isolated from whole blood also released VEGF
upon stimulation with thrombin. These data strongly
support the role of platelets in this observed VEGF
release after clotting.
The findings of previous studies investigating
platelet VEGF mRNA expression are inconsistent; Katoh et al. [27] detected mRNA for
VEGFm only, whereas Charnock-Jones et al. [28]
failed to detect mRNA using nested PCR primers,
although no mRNA positive control lane for platelets was shown. More recently, Mohle et al. [20]
identified VEGF secretion from a megakaryocyte
cell line, ex-vivo-generated megakaryocytes and isolated platelets. In this study, we confirm platelets
as a significant source of VEGF and in addition
identify by RT-PCR mRNA for the two soluble
secreted VEGF splice variants, VEGFm and
VEGF165. As the assay systems described detect
both of the corresponding proteins, it is not possible
to determine which is the predominantly expressed
form.
It is possible that platelets are solely responsible
for the VEGF released upon clotting of whole blood
samples. VEGF levels in platelet-rich plasma and
pure platelet samples were approximately 50% of
those in parallel whole blood samples allowed to
clot for an identical time period, although the
median platelet count of 117 x 109A in the plateletrich plasma and mean count of 142 x 109A in the
pure platelet preparations reflects the incomplete
recovery of platelets from whole blood.
An attractive hypothesis is that platelet activation
during blood clotting induces VEGF secretion as
part of the biological response to injury. As an
immediate response, a pulse of VEGF released
from platelets might alter local capillary permeability, allowing plasma proteins easy access to the
site, provide a chemoattractant signal for monocytes,
neutrophils [29] and endothelial cells [30] and
stimulate endothelial cells into cell division and new
capillary formation.
ACKNOWLEDGMENTS
This work was funded by grants from the North
West Kidney Association, the Wellcome Trust and
the Arthritis and Rheumatism Council, U.K. We
would like to thank Dr Don Ogilvie of Zeneca Phar-
403
maceuticals for the generous gift of recombinant
VEGF and flt-1, and the Department of Haematology, Royal Manchester Children’s Hospital, for
performing the Coulter counter analysis.
REFERENCES
I. Gospodarowicz D, Abraham JA, SchillingJ. Isolation and chatacterisationof a
vascular endothelial cell mitogen produced by pituitaryderived fdliculo stellate
cells. Proc Nad Acad Sci USA 1989; 86: 73 I I- 15.
2. Senger DR Galli SS, Dvorak AM, P w z z i CA,Harvey VS. Dvorak HF. Tumour
cells secrete a vascular permeabilityfactor that promotes accumulationof ascites
fluid. Science (WashingtonDC) 1983; 219 983-5.
3. Conn G, Bayne ML, Soderman DD, et al. Amino acid and cDNA sequences of a
vascular endothelial cell mitogen that is homologous to platelet-derivedgrowth
factor. Proc Nad Fuad Sci USA 1990; 87: 2628-32.
4. de Vries C, ExobedoJA, Ueno H, Houck K, Ferrara N. W\lliams LT. The fins
like tyrosine kinase, a receptor for vascular endothelial growth factor. Science
(WashingtonDC) 1992; 255 989-91.
5. Terman BI, Dougher-VermazenM, Canion ME, et al. Identificationof the KDR
tyrosine kinase as a receptor for vascular endothelial cell growth factor. Biochem
Biophys Res Commun 1992; 187: 1579-86.
6. Kendall Rl. Thomas KA Inhibition of vascular endothelial growth factor aaivity
by an endogenously encoded soluble receptor. Proc Nad Fuad Sci USA 1993;
90: 10705-9.
7. Dvorak HF. Brown LF, Detmar M, Dvorak AM. Vascular permeahlity factor/
vascular endothelial growth factor, microvascularhyperpenneability. and
angiogenesis. Am J Pathol 1995; 146. 1029-39.
8. Fava RA, Olsen NJ.Spencer-Green G, et al. Vascular permeabilityfactor/
endothelial growth factor (VPFMGF): accumulationand expression in human
synovial fluids and rheumatoid synovial tissue. J Exp Med 1994; 180: 34 I -6.
9. Bottomley MJ, Hdt PLJ,Watson CJ,Webb NJA,Brenchky PEC. Vascular
endothelial growth factor mediates synovial inhmation. Arthritis Rheumatism
1996 39(Suppl. 9) 5131.
10. Koch AE,Harlow LA, Haines GK, et al. Vascular endothelial growth factor - A
cytokine modulating endothelial function in rheumatoid arthritis. J lmmunol 1994;
152: 4149-56.
I I. Ferrara N. Vascular endothelial growth factor - The trigger for
neovascularizationin the eye. Lab Invest 1995; 7 2 6 15- 18.
12. Adamis AP. Miller JW,Bemal MT, et al. l n c d vascular endothelial growth
factor levels in the vitreous of eyes with proliferative diabetic retincpathy. Am
J Ophdralmol 1994; I18: 445-50.
13. Brenchky PEC. VPFNEGF: a modulator of microvascularfunction with potential
roles in glomerular pathophysidogy.J Nephml 1996 9 1-8.
14. Neulen J, Yan ZP, Raa& 5, et al. Human chorionic gonadotropin-dependent
expressionof vascular endothelial growth factor/vascular permeahlity factor in
human ganulosa cells: Importance in ovarian hypemimulation syndrome.J Clin
Endocrinol Metab 1995; 80: 1967-71.
IS. Baker PN, Krasnow J. Roberts JM, Ye0 KT. Elevated serum levels of vascular
endothelial growth factor in patients with preeclampsia. Obstetrics Gynecol
1995; 86: 815-21.
16. Kondo S. Atan0 M. Matsuo K, Ohmori I, Suzuki H. Vascular endothdial growth
factor/vaxular penneabrlity factor is detectable in the sera of tumor-bearing
mice and cancer patients. Biochim Biophys Acta 1994; 1221: 21 1-14.
17. Houck KA, Leung OW. Rowland AM, Winer J, Ferrara N. Dual regulation of
vascular enhthelialgrowth factor bioaMilabiliby genetic and proteolyhc
mechanisms. J Bid Chem 1992; 267: 2603 1-7.
18. Ye0 KT, Siwssat TM, Faix ID, Senger DR Ye0 TK. Development of timeresolved immunofluorometric assay of vascular permeability factor. Clin Chem
1992; 38: 71-5.
19. Hanatani M, Tanaka Y, Kondo 5, O h m ’ I, Suzuki H. Sensitive
chemiluminescence enzyme immunoassay for vascular endothdial growth factor/
vascular permeability factor in human serum. Bioxi Biotechnol Biochem 1995;
5 9 1958-9.
20. Mohle R Green D, Moore MAS, Nachman RL, Rafii S. Constitutive production
and thrombin-induced release of vascular endothelial growth factor by human
megakaryocytes and platelets. Proc Nad Acad Sci USA 1997; 94: 663-8.
2 I . British society for Haematdogy; BCSH Hawnostarisand ThrombosisTask Force.
Guidelines on platelet function testing. J Clin Pathd 1988; 41: 1322-30.
22. Chomaynski P. Sacchi N. Single-step method of RNA isolation by acid
guanidiniumthiocyanate-phenol-chloroform extraction. Anal Bochem 1987;
162: 156-9.
404
N. J. A. Webb et al.
23. lijima K, Yoshikawa N, Connolly DT, Nakamura H. Human mesangial cells and
peripheral blood mononuclearcells produce vascular permeabilityfactor. Kidney
Int 1993; 44:959-66.
24. Power CA,ClemetsonJM. Clemetson KJ,Wells TNC. Chemokine and
cherdine receptor mRNA expression in human platelets. Cytokine 1995; 7
479-82.
25. Casxells W. Growth factor therapies for vascular injury and ixhaemia.
Circulation 1995; 91: 2699-702.
26. h e r JM, Pieczek 4 Schainfeld R, et al. Clinical evidence of angiosenesis after
arterial gene transfer of phVEGF165 in patient with ixhaemic limb. Lancet 1996;
348: 370-4.
27. Katoh 0.Tauchi H, Kawaishi K, Kimura4 Satow Y. Expression of the vascular
endothelial growth factor (VEGF) receptor gene, KDR, in haemopoieticcells and
inhibitory effect of VEGF on apoptotic cell death caused by ioniing radiation.
Cancer Res 1995; 55: 5687-92.
28. Charnock-JonesDS,Sharkey AM, fhjput-Williams J, et al. Identificationand
localisationof alternately spliced m W for vascular endodrelialgrowth factor in
Bid
human uterus and emogen regulation in endometrialcarcinomacell li.
Reprod 1993; 48: I 120-8.
29. Bar(e0n B, Sonani S, ulou D,Weich H& M a n m i 4 Marme D. Migration of
human montxytes in response to vascular endothdial growth factor (VEGF) is
mediated by the VEGF receptor tlt-I.Blood 1996 8 7 3336-43.
30. Senger DR Ledbetted SR, Claffey KP, Papadopoulos-Sergbu4 Perruzzi CA
Detmar M. Stimulationof endothelial cell migration by vascular permeability
factorlvascular endothdial growth factor through coqeratiw mechanisms
involving the a&, osteopontin and thrombin. Am J Pathol 1996; 149 293-305.