FactorXIIbindingtoendothelialcellsdependson caveolae
Inger Schousboe
1
, Peter Thomsen
2
and Bo van Deurs
2
1
Department of Medical Biochemistry & Genetics, and
2
Structural Cell Biology Unit, Department of Medical Anatomi,
The Panum Institute, University of Copenhagen, Denmark
It is now generally accepted t hat factorXII (FXII) binds to
cellular surfaces in the vascular s ystem. One of t he suggested
receptors of this binding is the glyco sylphosphatidylinositol-
anchored urokinase-like plasminogen activator (u-PAR)
harbored in caveolae/lipid rafts. However, binding of FXII
to human umbilical vein endothelialcells (HUVEC) has
never been shown to be localized to these specialized mem-
brane structures. Using microscopical techniques, we here
report that FXII binds to specific patches of the HUVEC
plasma membrane with a high density of caveolae. Further
investigations of FXII bindingtocaveolae were performed
by sucrose density-gradient centrifugations. This showed
that the majority of FXII, chemically cross-linked to
HUVEC, could be identified in the same fractions of
the gradient as caveolin-1, a marker of caveolae, while the
majority of u-PAR was identified in noncaveolae lipid rafts.
Accordingly, cholesterol-depleted cells were found to bind
significantly reduced amounts of FXII. These ob servations,
combined with the presence of a minority of u-PAR in
caveolae co ncomitant with F XII binding, indicate that FXII
binding to u-PAR may be secondary and depends upon the
structural elements within caveolae. Thus, FXII bindin g to
HUVEC dependson intact caveolaeon the cellular surface.
Keywords: factor XII; HUVEC; lipid rafts; c aveolae;
u-PAR.
Factor XII (FXII) is a zymogen present in plasma at a
concentration o f 350 n
M
. A t local increases i n the
Zn
2+
concentration above the normal plasma level of
25 l
M
, FXII binds toendothelialcells along with plasma
prekallikrein. The latter becomes attached to the cells via
complex formation with high-molecular-weight kininogen
(HK) [1]. Binding of FXII, as well as plasma prekallik-
rein in co mplex with HK, initiates in vitro areciprocal
activation of FXII and PK, which experimentally can
activate plasminogen, and the coagulation and comple-
ment systems. So far it h as not been possible to s how
that FXII has an impact in any of these systems in vivo,
and its biological role remains elusive. By affinity
chromatography and antibody inhibition, receptors for
HK have been identifi ed as c ytokeratin 1 [2], urokinase-
like plasminogen activator (u-PAR) [3], and the receptor
for the globular head of the subunit C1q in complement
C1 (gC1qR, a lso known a s p 33) [4,5]. T he binding of
FXII to immobilized, purified gC1qR in competition with
HK tent atively suggested that FXII and HK share the
gC1qR as a common receptor [1,5]. Incubation of FXII
with prekallikrein and HK in the presence of gC1qR
leads to an FXII-dependent conversion of prekallikrein to
kallikrein. The observation that the same conversion was
observed when gC1qR was exchanged with cytokeratin 1,
suggested that c ytokeratin-1 could also be a receptor for
FXII [6]. To analyze whether these putative receptors
accounted for the binding of FXII to human umbilical
vein endothelialcells (HUVEC), it was recently shown
that fluorescein isothiocyanate (FITC)-labeled FXII inter-
acts with the multiprotein assemply of not only gC1qR
and cytokeratin 1, but also u-PAR [ 7]. u-PAR is a
glycosylphosphatidylinositol (GPI)-linked glycoprotein,
which plays a central role in the regulation of pericellular
proteolysis and participates in events leading to cell
activation. It is harbored in caveolae/lipid rafts [8]. This
prompted us to analyze, by microscopical techniques and
sucrose gradient centrifugation, whether FXII binding to
HUVEC is dependent on caveolae/lipid rafts.
Materials and methods
Activated FXII (FXIIa) was obtained as a lyophilized
powder from Enzyme Research Laboratories
1
(Leiden, the
Netherlands). It was s tored at 4 °C until dissolved a nd
then stored in aliquots at )80 °C. FXII was purchased
from Haematologic Technologies
2
(Essex Junction, VT,
USA), as a high-concentration solution in 50% (v/v)
glycerol, and stored at )20 °C as recommended by
the manufacturer. It migrated as a single band, with an
M
r
of 80 000, on reducing SDS/PAGE. SDS polyacryl-
amide gels (4–12%) were from Invitrogen (Taa-Stru
¨
p,
Denmark). All dilutions of FXII were performed in
siliconized test tubes and excess dilutions were d iscarded.
Correspondence to I. Schousboe, Department of Medical Biochemistry
& Genetics, The Panum Institute, University of Copenhagen,
Blegdamsvej 3C, DK-2200 Copenhagen N, Denmark.
Fax: + 45 35367980, Tel.: + 45 35327800,
E-mail: schousboe@imbg.ku.dk
Abbreviations:DTSSP,3,3¢-dithio-bis(succinimidylpropionate); FXII,
Factor XII; FXIIa, activated FXII; GPI, glycosylphosphatidylino-
sitol; HK, high-molecular-mass kininogen; HRP, horseradish
peroxidase; HUVEC, human umbilical vein endothelial cells;
MbCD, methyl-b-cyclodextrin; u-PAR, urokinase-like
plasminogen activator receptor.
Enzyme: activated factorXII (EC. 3.4.21.38).
(Received 2 December 2003, revised 5 May 2004,
accepted 27 May 2004)
Eur. J. Biochem. 271, 2998–3005 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04229.x
Poly(vinylidene difluoride) membranes were obtained
from Amersham Biosciences AB, and the chemiluminis-
cence enhancer (SuperSignalÒ West Femto Maximum
Sensitivity Substrate) and the crosslinking reagent 3,3¢-
dithio-bis(succinimidylpropionate) ( DTSSP) were obtained
from Pierce. Methyl-b-cyclodextrin ( MbCD) was from
Sigma. The primary antibodies used were affinity-purified
goat anti-(human FXII) IgG from Affinity Biologicals
Corp.
3
(Ancaster, Ontario, Canada), affinity-purified rab-
bit anti-(caveolin-1) I gG from Transduction Laboratories
4
(San Diego, CA, USA), mouse anti-(human CD31)
immunoglobulin, clone JC70A (M0823), from Dako-
Cytomation
5
(Glostrup, Denmark), and protein G-purified
rabbit anti-(u-PAR) immunoglobulin (399R) from
American Diagnostics
6
(Greenwich, CT, USA). For
immunofluorescence, the secondary antibodies were Alexa
488-conjugated chicken anti-(goat IgG) and TRITC-
conjugated swine anti-rabbit IgG from Transduction
Laboratories. For Western blotting, the s econdary anti-
bodies were horseradish peroxidase (HRP)-conjugated
swine anti-rabbit
7
immunoglobulins (P 0217) and b iotin-
ylated rabbit anti-goat immunoglobulins (E 0466) and
biotinylated goat anti-rabbit immunoglobulins (E 0432)
from DakoCytomation. Stabilized HRP-conjugated
goat anti-(mouse IgG) Ig and Restore
TM
Western blot
stripping buffer were from Pierce. All the biotinylated
antibodies had been solid-phase absorbed to minimize
cross-reactions with human immunoglobulins and fetal
bovine serum. HRP-conjugated streptavidin (P 0397)
was from D akoCytomation and molecular m ass m ark-
ers from Bio-Rad. H-
D
-Pro-Phe-Arg-para-nitroaniline
(S-2302) was from Chromogenix (Milan, Italy)
8
.All
other reagents were of the purest grade commercially
available.
Endothelial cell culture
Cryopreserved primary cultures of HUVEC (Clonetics
9
,San
Diego, CA, USA) were subcultured as described previously
[9]. Seven/eight generation c ells (third passage) were used
throughout the experiments. F or gradient centrifugation the
cells were plated in 75 cm
2
flasks, and for activity measure-
ments the cells were plated in 12-well microtiter plates.
Microscopic analysis w as performed oncells plated on
Labtek 4-chamber slides (Nalgene Nunc International
Corp.) at a d ensity of 10
4
cellsÆcm
)2
. The mediu m was
changed on day 3 and the cells used on days 4 or 5. To
eliminate any influence from FXII present in t he complete
medium, cells were exposed to serum-free medium (medium
lacking fetal bovine s erum)
10
, prior to the experimental
application, when indicated.
Cholesterol depletion and cholesterol determination
Cells were grown in serum-free medium for 3 h and
subsequently in the same m edium containing 1% (w/v)
MbCD for 30 min or the period of time indicated. The
cholesterol concentration in lysates of native and c holes-
terol-depleted cells was determined spectrophotometri-
cally by a cholesterol/peroxidase assay [ 10]. Protein
concentration was determin ed by the m ethod of
Bradford [11].
Binding of FXII to cells
The culture medium was aspirated and the cells were
washed twice over a period of 20 min with Locke’s buffer
(154 m
M
NaCl, 5.6 m
M
KCl, 3.6 m
M
NaHCO
3
,2.3m
M
CaCl
2
,1.0m
M
MgCl
2
,5.6m
M
glucose, 5 m
M
Hepes,
pH 7.4) containing 15 l
M
ZnCl
2
(wash buffer) followed
by a 15 min incubation in 0.1% (w/v) gelatin in wash buffer
(block buffer). As judged visually by microscopic inspection,
the washing procedure did not remove cells from the
surface, or change the morphology of the cells. The block
buffer was aspirated and the c ells incubated at room
temperature with FXII or FXIIa diluted at least 300-fold in
block buffer, giving a final concentration of 100–300 n
M
.
After 60 min of incubation, the cells were placed on ice, the
incubation medium above the cells was aspirated and the
cells were washed continuously for 5 s in a mild stream of
ice-cold wash buffer a nd subsequently handled as described
below.
Amidolytic activity of cell-bound FXIIa
As it has previously been shown that FXIIa binds to the
cells in a manner identical to FXII [9], binding of FXIIa was
used to quantify relatively the amount of FXII bound to
cholesterol-depleted cells. Normal and cholesterol-depleted
cells in 12-well microtiter plates were incubated with FXIIa,
as described above, and analyzed for the amidolytic activity
of cell-bound FXIIa, as previously described [12]. To each
well was added 600 lLof0.8m
M
S-2302 in 50 m
M
Tris,
12 m
M
NaCl, 10 m
M
EDTA, pH 7.8. From previous
experiments it was known that FXIIa cleavage of S-2302
on the surface of HUVEC was linear, with time, for at least
4 h. After this period of time, the reaction was s topped by
acidification and the absorbtion read at 405 n
M
.
Electron microscopy
HUVEC cells were fixed with 2% (v/v) formaldehyde and
0.1% (v/v) glutaraldehyde in 0.1
M
sodium phosphate
buffer, pH 7.2. The cells were washed, scraped off t he flasks,
pelleted, and postfixed with O sO
4
, contrasted en bloc with
1% (w/v) uranyl acetate, dehydrated in a graded series of
ethanols, an d embedded in Epon. Sections were examined
in a Philips CM 100 electron microscope (Philips, Eindh-
oven, the Netherlands).
Immunofluorescence microscopy
For immunofluorescence microscopy of cell-bound FXII
and its colocalization with caveolin-1, the cells were fixed
with )20 °C methanol subsequent to washing after incuba-
tion with FXII. Methanol was chosen as a fixative because
preliminary results had shown that glutaraldehyde fixation
destroyed the immunogenicity of FXII. The fixed cells were
blocked in block buffer [5% (w/v) goat serum (DakoCyto-
mation) in NaCl/P
i
(PBS)] and incubated with primary and
fluorescent secondary antibodies. The primary antibodies
were a mixture of goat anti-(human FXII) IgG (diluted
1 : 50) and rabbit anti-(caveolin-1) IgG (diluted 1 : 400).
The secondary antibodies were a mixture of Alexa 488-
conjugated chicken anti-(goat IgG) Ig (Molecular Probes)
Ó FEBS 2004 FXII bindingtocaveolae (Eur. J. Biochem. 271) 2999
(diluted 1 : 400) a nd TRITC-conjugated pig anti-(rabbit
IgG) Ig (DakoCytomation) (diluted 1 : 50), r espectively.
The cells were mounted with prolong anti-fade medium
from Molecular Probes. Specificity analyses of the anti-
bodies showed no reaction of the secondary antibodies with
cells incubated in the absence of the primary antibodies.
Confocal microscopy
The microscope used was a Zeiss LSM 510 confocal
microscope. The objectives were a plan-neofluor ·20/0.5, a
C-apochromate ·40/1.2 W corr, and a Plan apochromat
·100/1.4 oil Iris lens. The 488 nm line from an Argon laser,
and the 543 and 633 lines from two Helium/Neon lasers,
were used for excitation.
Sucrose gradient centrifugation
Protein separation on a sucrose gradient was performed as
described by Spisni et al. [13]. Briefly, one 75 cm
2
flask
containing a confluent layer of HUVEC was incubated with
FXII, and FXII adhered to the cell surface was crosslinked
to the receptor with D TSSP, as described above. To avoid
disruption of caveolae/lipid rafts during lyses of the cells,
this and the following procedures were performed on ice,
unless stated otherwise. After scraping off and collecting the
cells by centrifugation, the cell pellet was homogenized by
trituration (an 0.80 · 80 mm ne edle) and the cells were
lysed for 30 min in 1 mL of ice-cold 25 m
M
P
i
, 150 m
M
NaCl, 5 m
M
EDTA (PNE)/TX-buffer [PNE buffer
11
containing 1% (v/v) Triton X-100, 0.1 m
M
phenyl-
methanesulfonyl fluoride and 0.22 mgÆmL
)1
leupeptin).
Nuclei and cell debris were removed by centrifugation
(5000 g, 5 min). The supernatant was mixed (1 : 1) with
80% (w/v) sucrose in PNE buffer and placed at the bottom
of an ultracentrifuge tube. A linear sucrose gradient [5–35%
(w/v) in TNE buffer) was layered on top of the lysate. The
gradient was centrifuged for 20 h at 200 000 g (4 °C) in an
SW 40 Ti Beckman rotor. Fractions of 600 lLwere
withdrawn from the bottom of the gradient and mixed 1 : 1
(v/v) with 40% ice-cold trichloroacetic acid. After 2 h of
incubation (at 4 °C), precipitated protein was c ollected by
centrifugation, the supernatant aspirated and the precipitate
washed in 1 mL of ether/ethanol (1 : 1, v /v). Precipitated
proteins were disso lved by boiling for 5 min in reducing
Laemmli buffer.
SDS/PAGE and immunoblotting
For Western blot analysis, proteins were separated on
4–12% SDS/polyacrylamide gels and transferred to
poly(vinylidene difluoride) membranes according to stand-
ard procedures. Standard samples of caveolin-1 and FXII
and molecular mass markers were run simultaneously. After
incubation for 1 h with NaCl/Tris (TBS) block buffer
[50 m
M
Tris, 0.15 m
M
NaCl, pH 8.0, containing 0.1% (v/v)
Tween 20 and 0.1% (w/v) BSA], the membranes were
probed with goat anti-FXII immunoglobulin (diluted
1 : 10 000)/biotinylated rabbit anti-goat
12
immunoglobulins
(diluted 1 : 10 000)/HRP-conjugated streptavidin (diluted
1 : 10 000) and rabbit anti-(caveolin-1) immunoglobulins
(diluted 1 : 10 000)/HRP-conjugated swine anti-rabbit
immunoglobulin (diluted 1 : 10 000), respectively. The blot
probed with antibodies against FXII was subsequently
stripped u sing Restore
TM
Western b lot stripping buffer,
used according to the manufacturer’s instructions, and
probed against rabbit anti-(u-PAR) immunoglobulin
(diluted 1 : 10 000)/HRP-conjugated swine anti-rabbit Ig
(diluted 1 : 10 000) or mAb CD31 (diluted 1 : 1000)/HRP-
conjugated goat anti-(mouse IgG) Ig (diluted 1 : 10 000).
Dilutions of antibodies were performed in 1% nonfat skim
milk in TBS block buffer. Detection was carried out using
the chemiluminescence enhancer, SuperSignalÒ West
Femto Maximum Sensitivity Substrate, as recommended
by th e manufacturer, and the results were monitored on a
Las Chemiluminator. The intensity of the bands was
measured using the Image Gauge, quant menu.
Statistics
Analysis of variance (
ANOVA
) with the post hoc Student’s
t-test was u sed to determine the s tatistical significance of
difference between sample groups.
Results
FXII binds to HUVEC caveolae
Caveolae, 50–100 nm invaginations of the plasma mem-
brane, are a subset of sphingolipid- and cholesterol-enriched
lipid rafts, characterized by the presence of t he protein
caveolin [14]. Caveolae are abundant in endothelial cells
[15]. In agreement with this, HUVEC contained numerous
caveolae, as revealed by electron microscopy (Fig. 1A,B).
Interestingly, the caveolae were not evenly distributed in the
plasma membrane of these cells, but appeared regionally at
very high densities, separated b y caveolae-free membrane
segments. This concentration of caveolae at certain regions
of the HUVEC surface was also visualized by confocal
microscopy using an antibo dy against caveolin-1. Thus, an
intense immunofluorescence signal was obtained in certain
stretches of the plasma membrane (Fig. 2A). Double-
immunofluorescence labeling for caveolin-1 and bound
FXII showed a very high degree of co-localization
(Fig. 2B,C), indicating t hat bound FXII was associated
with caveolae.
The integrity of caveolaedependson a certain level of
plasma membrane cholesterol, and cholesterol depletion
achieved by incubating cells with M bCD makes caveolae
disappear and allows the released caveolin to become
diffusely distributed in the plasma membrane and to be
internalized [16, 17]. To a nalyze further the apparent
association of FXII binding with HUVEC caveolae, cells
were therefore c holesterol-depleted by incubation with
MbCD. Incubation of HUVEC with MbCD [1% (w/v),
60 min] caused a reduction in cellular cholesterol to 20% of
the level found in untreated cells. By electron microscopy it
could be shown t hat this treatment of HUVEC resulted in
an almost complete removal of caveolae. Thus, the patches
with high concentrations of caveolae were never seen in
cholesterol-depleted cells and, only rarely, could single,
caveolae-like structures be identified (Fig. 1 C). This was
further confirmed by confocal microscopy (Fig. 2C,E). The
impact of MbCD treatment o n the binding of FXII was
3000 I. Schousboe et al. (Eur. J. Biochem. 271) Ó FEBS 2004
analyzed a t vary ing MbCD concentration and treatment
periods. By confocal microscopy, only little and diffusely
distributed FXII could be identified on HUVEC after
60 min of incubation with 1% (w/v) MbCD (Fig. 2D). The
ability of MbCD-treated cellsto bind FXII was analyzed
quantitatively by measuring the amidolytic activity of
FXIIa, w hich binds to the cells in a m anner i dentical to
that of FXII [9]. This way it w as shown that progressively
decreasing amounts o f FXIIa adhered to t he cells with
increasing periods of exposure to 1% (w/v) MbCD. The
amidolytic activity of FXII a in the wells exposed for a
longer period of time than 30 min remained constant, but
significantly lower (P<0.001) than the activity measured
in the wells not exposed to MbCD (Fig. 3). No changes in
cell number or in protein content were identified during the
treatment.
FXIIa binding associated with the Triton X-100 insoluble
low-density fraction
A common means of identifying proteins associated with
caveolae is to investigate the detergent i nsolubility at 4 °C.
To examine whether the FXII/receptor complex might be
insoluble in cold Trito n X-100, FXII was crosslinked to the
receptor by t he cleavable disulfide cross-linking reagent,
DTSSP. This prevented dissociation of FXII from the
receptor during the subsequent solubilization and ultracen-
trifugation. To minimize contamination of the surface of the
plastic with nonspecifically bound FXII, the cells were
scraped off before solubilization. Figure 4 shows a Western
blot of an SDS polyacrylamide gel of lysates of native and
MbCD-treated cells incubated with block buffe r i n t he
presence or absence o f FXII and subsequently exposed to
DTSSP. While no visible bands were observed in the lysates
of cells incubated in the absence of FXII, FXII was present
in the reduced lysates of c ells incubated with F XII,
regardless of wh ether these were native or treated with
MbCD. The intensity of t he anti-FXII reacting band in the
lysates f rom MbCD-treated cells was, however, apparently
weaker than that from native cells. To confirm that this was
not caused by analytical variations, the blot was stripped
and probed with anti-CD31. CD31 is a marker protein of
endothelial cells. It appeared as a clear band in reduced
samples and as a smear in nonreduced samples of lysates of
cells treated with DTSSP. M easuring the intensity of the
anti-FXII and the anti-CD31 reacting bands in the reduced
samples on the same blot showed, in three individual
experiments, that the FXII/CD31 in lysates of cells treated
with MbCD was 50 ± 10% of that measured in lysates of
native cells. The invisibility of FXII in nonreduced samples
Fig. 1. Caveolae in human umbilical vein
endothelial cells (HUVEC) cells, detected by
electron microscopy. Caveolae in control cells
are shown in (A) and (B). (A) A section
perpendicular to the plasma membrane,
revealing the typical shape of these structures,
whereas (B) is a tangential section better
showing the high density of caveolae. CP,
clathrin-coated pit. (C) Section through a
cholesterol-depleted cell in which the arrow
indicates a single, caveolae-like structure.
Bar: 200 nm.
Ó FEBS 2004 FXII bindingtocaveolae (Eur. J. Biochem. 271) 3001
indicates a poor anti-FXII immunoglobulin reactivity of
FXII crosslinked to the receptor, or formation of very high
M
r
complexes of FXII.
The solubility, in cold Triton X-100, of the FXII/receptor
in lysates o f c ells was a nalyzed by sucrose gradient
centrifugation. Figure 5A shows a Western b lot of a
reducing SDS polyacrylamide gel of every second fraction
of the gradient consisting of 20 fractions. An 80 kDa
molecular mass anti-FXII immunoglobulin-reacting band
was present in the soluble fractions (representing one of
three fractions), with the density of a 40% sucrose solution,
as well as in the higher density fractions of the floating
fractions w ith sucrose d ensities from 40 to 2 3%. W ith
decreasing intensity throughout the gradient, all of t he
undiluted fractions stained heavily when probed with anti-
caveolin-1 (results not shown). This indicated that caveolae
were distributed throughout the entire gradient. The
intensity of these bands, however, d ecreased with decreasing
density of t he gradient. T his was evidenced by probing for
caveolin in samples d iluted fivefold, which convincingly
showed the presence of t he highes t c oncentration o f
caveolin-1 in the h igher density of the floating fractions.
FXII was thus identified in the same fractions as those
containing the highest concentrations of caveolin-1. This
indicates that FXII originating from the disulfide FXII/
receptor complex was harbored in caveolae. As a possible
receptor for FXII, the localization of u-PAR (M
r
50 000–
60 000) was analyzed. In accordance with the presence of
large vesicles rich in GPI-anchored proteins (lipid rafts),
as well as smaller caveolar vesicles lacking GPI-anchored
proteins in Triton X-100 insoluble membranes [18], u-PAR
could be i dentified mainly in the very-low-density part of the
Fig. 2. Colocalization of cell-bound factor XII
(FXII) and caveolin-1. Human umbilical vein
endothelial cells (HUVEC) were grown in
serum-free medium for 4 h. Then, medium in
half of the wells was exchanged with the same
medium containing methyl-b-cyclodextrin
(MbCD) and the incubation was continued
for another 60 min. The cells were then
washed, incubated with 100 n
M
FXII, and
subsequently fixed with methanol. The fixed
cells were first incubated with a mixture of
goat anti-FXII (1 : 50) and rabbit anti-
(caveolin-1) (1 : 200) and second with a
mixture of Alexa 488-conjugated chicken
anti-(goat IgG) Ig (1 : 400) and TRITC-
labeled swine anti-(rabbit IgG) Ig (1 : 50).
(B) and (D) (green) show FXII, and (A), (C)
and (E) (red) show caveolin-1. The panels are
representative photomicrographs of at least
three independent experiments. Bar: 20 lm.
3002 I. Schousboe et al. (Eur. J. Biochem. 271) Ó FEBS 2004
gradient (Fig . 5A). U pon treatment w ith MbCD, the
density of caveolae, as well as lipid rafts, increased. As a
result of this, the presence of caveolin-1 and FXII d isap-
peared from their upper density fractions, while u-PAR was
skewed towards the higher density fractions (Fig. 5B). The
weakly stained 60- and 70 kDa anti-FXII reacting bands,
noticeable throughout the gradients, were present with
varying intensity in all of the three experiments performed
and were considered to be artifacts.
Discussion
By a combination of microscopy and gradient centrifuga-
tion, the present study shows that HUVEC-bound FXII
localizes to caveolae, and that this localization is dependent
upon the structural integrity of these elements in the plasma
membrane.
It is generally accepted that FXII binds to HUVEC and
that this binding is receptor mediated. Investigation of this
binding, by immunofluorescence microscopy of monolayers
of HUVEC incubated with FXII, indicated, in contrast to a
previous study [7], that FXII does not bind diffusely to the
cells but to specific patches in the plasma membrane
corresponding to caveolae-rich domains. The reasons for
the discrepancy between this and the former study may be
several. In the present study, native FXII was incubated
with morphologically intact HUVEC before fixation, while
in the previous study the cells were fixed before incubation
with FXI I. Furthermore, the FXII bind ing was, in the
former study, visualized by FITC labeling, whereby Lys
residues (of possible importance f or the binding of FXII to
HUVEC) might have been blocked.
Caveolae are highly immobile plasma membrane invag-
inations [17] abundant in the cellular membrane of endo-
thelial cells [14]. They are enriched in glyc olipids and
cholesterol. Depleting the cells for cholesterol by treatment
with MbCD disrupts the integrity of caveolae. Cholesterol-
deprived cells incubated with FXII showed microscopically
only little binding of FXII. This confirms that FXII binds to
a receptor harbored in caveolae. Previous analyses have
shown that MbCD preferentially extracts cholesterol from
the outside and partially solubilizes GPI-anchored and
Fig. 3. Quantification of factorXII (FXII) bin ding to c ells preincubated
with methyl- b-cyclodextrin (MbCD) f or varying p eriods of time. Human
umbilical vein endothelialcells (HUVEC), grown in serum-free med-
ium, were, for varying periods of time, pre-exposed to 1% (w/v)
MbCD and subsequently incubated with 100 n
M
activated FXII
(FXIIa) for 1 h. After washing, the cells were incubated with S-2302.
Theordinateindicatestheabsorbancemeasuredat405 nmafter4 hof
incubation. The concentration of protein was measured in the incu -
bation mixture by the method of Br adford [11]. The activity wa s
measured on three individual setups of cells and determined as t he
mean value ± SD. The significance of variance of FXIIa bound to
HUVEC, pre-exposed for varying periods of time to MbCD, was
calculated by a n alysis of variance (
ANOVA
). Use of an asterisk
14
indicates
a P-value of < 0.001 relative to the control (0 min incubation period),
determined using the post-hoc Student’s t-test.
Fig. 4. Western blots of factorXII (FXII) cross-linked to membranes of
native and cholesterol-depleted cells. Oneofthreecultureflasksof
human u mbilical vein endothelialcells (HUVEC), grown in serum-free
medium for 15 h, was pre-exposed to 1% (v/w) methyl-b-cyclodextrin
(MbCD) for 30 min. This flask (+MbCD +FXII) and a second
(–MbCD +FXII) of the three were subsequently incubated with
100 n
M
FXII, while the third flask (–MbCD –FXII) was incubated
with block buff er. Aft er this incubation, the cells were washed and
incubated with DTSSP. After neutralization of the DTSSP, the cells
were lysed. Western blots of reduced(+)andnonreduced(–)samples
15
were visualized by sequential incub ation with goat anti-FXII, bio tin-
ylated rabbit anti-(goat IgG), horseradish peroxidase (HRP)-conju-
gated streptavidin and Sup erSignalÒ West Femto Maximum
Sensitivity Substrate (upper part of the figure). The blot was then
stripped us ing R esto re
TM
Western blot s tripping bu ffer and in cu bated
sequentially with mouse anti-(human CD31), HRP-conjugated goat
anti-(mouse IgG) and SuperSignalÒ West Femto Maximum Sensi-
tivity Substrate (lower part of the fi gure).
Ó FEBS 2004 FXII bindingtocaveolae (Eur. J. Biochem. 271) 3003
transmembrane p roteins [18]. A s i t has been previously
shown that no differentiation exists between the binding of
FXII and FXIIa, the change in ability of M bCD-treated
HUVEC to bind FXII was quantified by measuring the
amount of bound FXIIa colorimetrically [9]. This showed
that a significantly lower amount of FXII was bound to
MbCD-treated cells than to untreate d cells. However,
treatment with MbCD did not completely prevent the
binding of FXII to the layer of cells in the well. In fact,
50% of the binding persisted after MbCD treatment. This
was confirmed by density measurements of Western blots of
FXII extracted from the cells. As no FXII was observed on
the fluorescen ce microscopy of MbCD-treated cells incu-
bated with FXII, a considerable amount of the cell-bound
FXII might h ave been removed during t he extensive
washing of the fixed cells accompanying the immunofluo-
rescence staining.
Binding of FXII to lipid-rich domains in the c ellular
membrane was f urther verified by sucrose gradient centrif-
ugation. While confocal microscopy and the activity
measurements were performed on intact cells without
crosslinking, the gradient centrifugations were performed
on lysates of cellsto which FXII had been chemically
crosslinked to cellular proteins in its immediate vic inity by
incubation with the cleavable membrane-impermeable
DTSSP. Sucrose gradient centrifugation of cold Triton
X-100 lysates of cells showed that FXII was present in the
same fractions as the marker protein for caveolae, caveolin-
1. Both FXII and the majority of caveolin-1 were
distributed in the soluble fraction, as well as in the higher
density floating fractions of the gradient. As it has
previously been suggested that F XII binds to the G PI-
anchored receptor, u-PAR [7], and this receptor is organized
in caveolae and lipid rafts on the cell surface [19], it was
tempting to investigate this possibility f urther. We showed
that u-PAR, in contrast to FXII, originating from the FXII/
receptor complex, appeared to be present in mainly the
lower density of the floating fractions and thus not only in
the fractions containing the highest concentration of
caveolin I and FXII. In line with this, immunofluorescence
microscopy of the cellular localization of u -PAR indicated
that u-PAR was equally distributed on the plasma mem-
brane surface (results not shown).
The difference in fractional distribution of u-PAR and
FXII in the sucrose gradient indicates that FXII binding to
u-PAR may be secondary to FXII bindingto caveolae. This
is in accordance with previous findings showing that FXII
binding to HUVEC could be inhibited by antibodies to
u-PAR, but not by the much smaller ligands of u-PAR,
urokinase-like
13
plasminogen activator (u-PA), and vitronec-
tin [7]. It may also explain the biphasic nature of FXII
binding to HUVEC [9].
The binding of FXII tocaveolae indicates that FXII
binds to a receptor predominantly localized in these specific
structural elements of the cell membrane containing a subset
of sphingolipids and cholesterol. Moreover, in addition to
GPI-anchored proteins, caveolae harbor GPI-anchored
proteoglycans (glypicans). Glypicans are highly sulfated
by substitution with heparan sulfate and chondroitin sulfate
[20], creating a binding site for po lycationic molecules [21].
Among several glucosaminoglycans, both heparan sulfate
and chondroitin sulfate have been shown to enhance the
rate of FXII activation in solution [22]. Combining the
established knowledge about FXII bindingto negatively
charged surfaces in general, and proteoglycans in particular
[23], with our present findings, we tentatively propose that
Fig. 5. Analysis of Triton X-100 solubility b y sucrose gradient centri-
fugation. Native and methyl-b-cyclodextrin (MbCD)-treated cells, to
which factor X II (FXII) had been c ross-linked, were e xtracted with ice-
cold Triton X-100 and subjected to su crose gradient centrifugation, as
described in the Materials and methods. Every second fraction of the
gradient
16
was analyzed for the distribution of FXII, caveolin-1 and
urokinase-like plasminogen activator receptor (u-PAR) by reducing
SDS/PAGE and Western blotting. The distribution of FXII and
caveolin-1 was probed o n individual blots, while the distribution of
u-PAR was probed on stripped blots (F ig. 4). (A) Results obtained
using native cells; (B) results obtained with MbCD-treated cells. The
distribution is representative of three individual experiments.
3004 I. Schousboe et al. (Eur. J. Biochem. 271) Ó FEBS 2004
FXII binds to a glypican in caveolae. Direct analysis of the
detergent-solubilized FXII/receptor complex is ongoing but
at the moment it must be concluded that the present
investigation was unable to confirm previous studies [7],
suggesting that u-PAR is the most likely receptor of FXII.
Acknowledgements
The expert te chnical assistance of Ms Mette Olsen a nd Birgit Harder is
greatly appreciated. The work was supported by grants from the
Danish Medical Research Council and the Novo Nordisk foundation.
References
1. Reddigari, S.R., Shibayama,Y.,Brunnee,T.&Kaplan,A.P.
(1993) Human Hageman factor (Factor XII) and high molecular
weight kininogen compete for t he sam e binding site on h uman
umbilical vein endothelial cells. J. Biol. Chem. 268, 11982–11987.
2. Hasan, A.A., Zisman, T. & Schmaier, A.H. (1998) Identification
of cyto keratin 1 as a binding protein and presentation rec eptor for
kininogens onendothelial cells. Proc. Natl Acad. Sci. USA 95,
3615–3620.
3. Colman, R.W., Pixley, R.A., Najamunnisa, S., Yan, W., Wang, J.,
Mazar, A. & McCrae, K.R. (1997) Binding of high molecular
weight kininogen to human endothelialcells is mediated via a site
within d omain 2 an d 3 of the urokinase recep tor. J. Clin. Invest.
100, 1481–1487.
4. Herwald, H., Dedio, J., Kelsler, R., Loos, M. & Mu
¨
ller-Esterl, W.
(1996) Isolation and characterization of the kininogen -binding
protein p33 from endothelial c ells. Identity with the gC1q receptor.
J. Biol. Chem. 271, 13040–13047.
5. Joseph, K., Ghebrehiwet, B., Peerschke, E.I., R eid , K.B. &
Kaplan, A.P. (1996) Identification of the zinc-dependent endo-
thelial cell bindin g protein for high m olecular weight kin inogen
and factor XII: i dentity with the receptor that binds to the g lobular
heads of C1q (gC1q-R). Proc. Natl Acad. Sci. USA 93, 8552–8557.
6.Joseph,K.,Shibayama,Y.,Ghebrehiwet,B.&Kaplan,A.P.
(2001) Factor XII-dependent contact activation on endothelial
cells and binding proteins gC1qR and cytokeratin 1. Thromb.
Haemost. 85, 119–124.
7. Mahdi, F., Madar, Z.S ., F igueroa, C.D. & Schmaier, A.H. (2002)
Factor XII interacts with the m ultiprotein assemply of urokinase
plasminogen activator receptor, gC1qR, and cytokeratin 1 on
endothelial cell membranes. Blood 99, 3585–3596.
8. Stahl, A . & Mu
¨
ller, B.M. (1995) The urokinase activator r eceptor,
a GPI-linked protein, is localized in caveolae. J. Cell. Biol. 129,
335–344.
9. Schousboe, I. (2001) Rapid and cooperative binding of fac tor XII
to human umbilical vein endothelial cells. Eur. J. Biochem. 268,
3958–3963.
10.Gamble,W.,Vaughan,M.,Kruth,H.S.&Avignan,J.(1978)
Procedure for determination of free and t otal c holesterol in micro-
or nanogram amounts suitable for studies with cultured cells.
J. Lipid. Res. 19, 1068–1070.
11. Bradford, M.M. (1976) A rapid and sensitive method for the
quantitation of microgram quantities of p rotein utilizing the
principle of protein–dye binding. Anal. Biochem. 72, 248–254.
12. Schousboe, I. (2003) Binding of activated factorXIIto endothelial
cells affects its inactivation by the C1-esterase inhibitor. Eur. J.
Biochem. 270, 111–118.
13. Spisni, E., Griffoni, C., Santi, S., Riccio, M., Marulli, R., Barto-
lini,G.,Toni,M.,Ulrich,V.&Tomasi, V. (2001) Colocalization
prostacyclin (PGI2) synthase-caveolin-1 in endothelialcells and
new roles for PGI2 in angiogenesis. Exp. Cell. Res. 266, 31–34.
14. van Deurs, B., Roepstorff, K., Hommelgaard, A.M. & Sandvig,
K. (2003) Caveolae: anchored, multifunctional platforms in the
lipid ocean. Trends Cell Biol. 13, 92–100.
15. Stan, R.V. (2002) Structure and function of endothelial caveolae.
Microsc. Res. Tech. 57, 350–364.
16. Rodal, S.K., Skretting, G., Garred, O., Vilhard, F., van Deurs, B.
& Sandvig, K. (1999) Extraction of cholesterol with methyl-beta-
cyclodextrin perturbs formation of clathrin-coated endocytic
vesicles. Mol. Biol. Cell. 10 , 961–974.
17. Thomsen, P., Roepstorff, K., Stahlhut, M. & van Deurs, B. (2002)
Caveolae are highly immobile plasma membrane microdomains,
which are not inv olved in constitu tive endocytic t rafficking. Mol.
Biol. Cell 13, 238–250.
18. Ilangumara, S. & Hoessli, D.C. (1998) Effects of cholesterol
depletion by cyclodextrin on the sph ingolipid microdomains of the
plasma membrane. Biochem. J. 15, 433–440.
19. Varma, R. & Mayor, S. (1998) GP I-anchored proteins ar e
organized in sub micron domains on the cell surface . Nature 394,
798–801.
20. Chen,R.L.&Lander,A.D.(2001) Mechanisms underlying pre-
ferential assembly of heparan sulfate o n glypican-1. J. Biol. C hem.
276, 7507–7517.
21. Fransson, L A
˚
. (2003) Glypicans. Int. J. Biochem. Cell Biol. 35,
125–129.
22. Hojima, Y., Cochrane, C.G., Wiggins, R.C., Austen, K.F. &
Stevens, R.L. (1984) In vitro activation of the contact (Hageman
factor) system of plasma by heparin and chondroitin sulfate E.
Blood 63, 1453–1459.
23. Røjkjær, R. & Schousboe, I. (1997) The surface-dependent auto-
activation mechanism of factor XII. Eur. J. Biochem. 243,
160–166.
Ó FEBS 2004 FXII bindingtocaveolae (Eur. J. Biochem. 271) 3005
. fractional distribution of u-PAR and
FXII in the sucrose gradient indicates that FXII binding to
u-PAR may be secondary to FXII binding to caveolae. This
is. u-PAR may be secondary and depends upon the
structural elements within caveolae. Thus, FXII bindin g to
HUVEC depends on intact caveolae on the cellular