(Received for publication, December 27, 1996, and in revised form, March 24, 1997)
From the Departments of Physiology and Obstetrics and Gynaecology, University of Sydney, Human Reproduction Unit, Royal North Shore Hospital, St. Leonards, New South Wales 2065, Australia
This report confirms that human umbilical vein endothelial cells activated by A23187 produce platelet-activating factor (PAF) (22.4 ± 9.9 ng/106 cells/h; mean ± S.E.). A proportion of the PAF produced (56%) was released by cells into the medium. The PAF released, however, was not detected without prior organic extraction, and the method of organic extraction was critical for detection. Extraction with 80% ethanol was not successful, but a modified methanol/chloroform extraction method was. These observations may explain some of the conflicting reports in the literature on release of PAF by activated endothelial cells. The requirements for organic extraction may reflect the nature of cell-released PAF's binding by albumin; it was observed that PAF added to identical media could be detected in a bioassay without the requirement for extraction. Such PAF was also readily degraded by PAF-acetylhydrolase added to media, while PAF released from cells was resistant to such degradation, suggesting that it was released in a "protected" configuration. Stimulation of cells was performed in media with albumin as the only extracellular macromolecule. Limited proteolytic digestion of the albumin with trypsin and pepsin showed that PAF released by cells was located exclusively between amino acids 240 and 386 (domain II), while no synthetic PAF added to media was located on this region. These results are identical to those described for the release of PAF by the early embryo. Albumin exposed to embryos had a higher thiol concentration (0.77 ± 0.04 µmol of thiol/µmol of albumin; mean ± S.E.) than control media to which an equivalent amount of synthetic PAF was added (0.59 ± 0.02 µmol of thiol/µmol of albumin) (measured with Ellman's reagent). Furthermore, albumin from conditioned media was more susceptible to reduction by 10 mM dithiothreitol than control albumin, as assessed by its mobility on PAGE. The protected configuration of released PAF was caused by cell-dependent conformational changes to albumin involving cysteine-cysteine disulfide bonds. Partial reduction with dithiothreitol of albumin exposed to cells resulted in released PAF being able to be detected directly in a bioassay without the requirement for prior organic extraction.
Platelet-activating factor (PAF1; 1-O-alkyl-2-acetyl-sn-glycero-3-phosphocholine) is a biologically active ether phospholipid (1, 2). It has wide tissue distribution and may function in normal physiological processes such as inflammation, neural activity, and reproduction. It may also have a role as a mediator in pathological states such as asthma, ischemia, gastric ulceration, hypertension, atherosclerosis, and shock.
Many studies show that endothelial cells do not produce PAF under basal conditions in vitro, but synthesis can be stimulated by a variety of agonists (see Ref. 3 for review). PAF synthesis occurred in endothelial cells from diverse vascular beds, including aorta, umbilical vein, and pulmonary artery (4). Several reports (5-11) have shown that virtually all PAF synthesized by stimulated endothelial cells remained associated with the cells. By contrast, a number of reports (12-19) show that as much as 25% was released.
A common feature of reports that have failed to detect PAF release from endothelial cells (5-11) is that they have used methods other than a modified chloroform-methanol extraction method (20) for extracting phospholipids prior to assay. A recent study (21) showed that this method of extracting media was required to detect PAF released from preimplantation embryos, due apparently to PAF's exclusive binding to domain II of albumin. Binding of PAF at that site on albumin resulted in PAF being protected from the actions of PAF-acetylhydrolase (PAF:AH) and also resulted in PAF being undetectable in bioassays without prior extraction. It was suggested that binding of PAF to domain II of albumin resulted in it being presented in a protected form (21). PAF added to medium also bound to albumin but was not found on domain II and was not in a protected form. We were interested to determine whether binding of PAF to albumin in this protected configuration occurs for PAF released by cells other than embryos and also to investigate the nature of the interaction which results in the protected configuration.
In this study we show that activation of endothelial cells results in the production of PAF and its release into medium. It is bound to albumin in a protected form, being exclusively located on domain II of albumin. It is also shown that binding of embryo-derived and endothelial cell-derived PAF to albumin was associated with conformational changes to albumin which involved modification of the cysteine-cysteine disulfide bonds, while the the conformation conferred by disulfide bonds was necessary for PAF to remain in its "protected" configuration.
Human umbilical vein endothelial cells (HUVEC) were collected (Maternity Unit, Royal North Shore Hospital of Sydney, NSW, Australia) as described (22). Blood was flushed from the vein, and cells were removed by 0.05% (w/v) collagenase (ICN Biomedicals, Costa Mesa, CA) in Hanks' balanced salt solution, pH 7.2 (calcium/magnesium-free; Sigma). The cells were washed by centrifugation and resuspended in HUVEC growth medium (medium 199, ICN Biomedicals) supplemented with 20% (v/v) fetal calf serum (ICN Biomedicals), 50 µg of heparin/ml (Sigma), and 100 µg of endothelial cell growth supplement/ml (from bovine neural tissue: Sigma).
They were plated onto 75-cm2 flasks (Corning) coated with
gelatin (from porcine skin; Sigma) and re-fed every second day until confluent. Following 4-6 passages, cells were transferred to 35-mm Petri dishes (Lux; Nunc, Naperville, IL) coated with gelatin at a
seeding density of 2 × 105 cells/1.5 ml. At
confluence, dishes were washed three times with medium 199 + 3 mg of
bovine serum albumin/ml (BSA (crystallized, Pentex); Miles, Kankakee,
IL) (basal medium). Cells were then incubated for 60 min with 1 ml of
basal medium, basal medium with 2.5 µM A23187, or an
equivalent volume of vehicle in basal medium (0.25 µl of
Me2SO/ml of basal medium). The endothelial cell-conditioned medium (ECCM) was collected, and the cells (washed free of incubation medium) were resuspended in 1 ml of fresh basal medium. ECCM and cells
were stored at 20 °C.
Two-cell mouse embryos were collected
from superovulated Swiss outbred albino mice (Department of Veterinary
Physiology, University of Sydney, NSW, Australia) as described (21).
Thirty two-cell embryos were cultured for 24 h at 37 °C in 5%
CO2 in air in 1 ml of embryo culture medium (HTFM: Ref. 23)
containing 3 mg BSA/ml. Embryos were removed after 24 h, and the
embryo-conditioned medium (ECM) was stored at 20 °C.
PAF in media or cell samples was extracted by a modified version (20, 24) of the Bligh-Dyer organic extraction method (25), followed by partial purification by TLC (24). Cells were sonicated for 1 s prior to extraction. Quantitative measurement of PAF was performed by either a platelet aggregation bioassay (25) or a radioimmunoassay (RIA) (26).
Preparation of ECM and ECCM PoolsIndividual ECM and ECCM
media samples were assayed. Where PAF was detected, samples were pooled
to give large volumes of identical media with which to carry out
characterization studies. Controls for ECCM were prepared in the same
media used for the stimulation of endothelial cells containing 2.5 µM A23187. Pools of control media were prepared by
supplementing control medium with synthetic PAF (from bovine heart
lecithin; Sigma) to a concentration equivalent to the cell-derived PAF
present in ECM and ECCM. Radiolabeled PAF ([3H]C16:0 PAF:
1-O-[hexadecyl-1,2
-3H]-labeled, 60 Ci/mmol:
NEN Life Science Products) was added to all pools (at 50,000 dpm/ml), to act as a recovery trace. All pools were stored as 1-ml aliquots at
20 °C.
The PAF-receptor antagonist, WEB 2170 (6-(2-chloro-phenyl)8,9-dihydro-1-methyl-8-(4-morpholinylcarbonyl)-4H,7H-cyclopenta[4,5]-thieno[3,2-f][1,2,4]triazolo-[4,3-a][1,4]diazepine; a generous gift of Boehringer Ingelheim KG, Ingelheim am Rhein, Germany) was used to confirm that the bioactive material released from endothelial cells was PAF. Citrated rabbit whole blood was added to a micromixing device. Antagonist or vehicle (5 µl) was added, followed by 45 µl of test material or known PAF standard. The extent of platelet aggregation after 15 min was used to assess PAF activity as described (27).
Assessment of the Susceptibility of Endothelial Cell-derived PAF to Hydrolysis by PAF:AHPAF:AH converts PAF into its biologically inactive metabolite, lysoPAF. The effect of PAF:AH on endothelial cell-derived PAF was compared with its effect on synthetic PAF in equivalent medium. Serum was collected from a healthy volunteer and heat-inactivated (20 min at 56 °C). Half was left untreated and the remainder acid-treated (pH 3.0), with 1 M HCl (BDH Chemicals, Poole, United Kingdom) for 20 min at 37 °C and then returned to pH 7.4 by 1 M NaOH (BDH), which deactivates PAF:AH (28). The PAF:AH specific activity was measured as described previously (21). Untreated or acid-treated sera (100 µl) were added to 900 µl of the ECCM pool or corresponding synthetic PAF pool, for 24 h at 37 °C, in 5% CO2 in air. Medium was then extracted and the PAF concentration measured by bioassay.
Characterization of PAF's Binding Site on Albumin by Limited Proteolytic AnalysisLarge digestion fragments of BSA were prepared by limited proteolysis with pepsin (EC 3.4.23.1, from porcine stomach mucosa; Sigma) or trypsin (EC 3.4.21.4, Type XIII, L-1-tosylamide-2-phenylethylchloromethyl ketone-treated, from bovine pancreas; Sigma). The methods were as described (29-31). Pepsin was prepared in 0.12 M ammonium formate (Sigma), pH 3.7, and trypsin in 0.04 M Tris-HCl (Sigma), pH 8.15. Pepsin digestion was performed at a 1:250 pepsin:BSA ratio (w/w), and trypsin was used at 1:750 (w/w). The size of the major digestion products were estimated by SDS-PAGE using the PhastSystem (Pharmacia Biotech Inc.) with homogeneous 20% gels and visualized by Coomassie Blue/silver stain.
Proteolysis was performed on ECCM and its respective control medium, to which an equivalent concentration of synthetic PAF was added. Following proteolytic digestion or sham digestion, the products were separated on native PAGE gels (15%). The gels were sliced and regions containing fragments of interest were excised, and homogenized in 2 ml of 0.05 M phosphate buffer, pH 6.8, allowing proteins to diffuse out. The protein content of the gel slices was measured and used to estimate the efficiency of the enzyme reaction and protein recovery. Each fragment was subjected to organic extraction and assayed for PAF by bioassay.
Determination of BSA Thiol ConcentrationA
spectrophotometric assay (32, 33) was used to measure the concentration
of thiols present in BSA in an ECM pool compared with BSA in equivalent
medium but not exposed to embryos. This control medium was supplemented
with an equivalent concentration of synthetic PAF. The assay measures
the absorbance of the thionitrobenzoate anion at 412 nm following
reaction of thiol with 5,5-dithionitrobenzoic acid (Ellman's reagent,
Sigma). The absorbance was measured and the molar concentration of
thiols present calculated using Beer's Law and the molar extinction
coefficient of the thionitrobenzoate anion.
Cysteine (Sigma) was used to validate the assay and to generate a
standard curve. The reaction was started by addition of 150 µmol of
5,5-dithionitrobenzoic acid to 1 ml of ECM or control medium,
containing 1 mM EDTA (Sigma). Incubation was for 20 min at
room temperature. The thiol concentrations of BSA in the ECM pool and
synthetic PAF pool were measured at 1/2, 1/5, and 1/10 dilutions with
protein-free HTFM (giving a final concentration of 1.5, 0.6, and 0.3 mg
of BSA/ml, respectively). The results were expressed as µmol of
thiol/µmol of BSA.
The tertiary structure of BSA in ECM and in control medium was indirectly assessed by examining electrophoretic mobility in SDS-PAGE under partially reducing conditions (34). Dithiothreitol (ICN Biomedicals) was added to medium (10 mM final concentration) and incubated at 25 °C with gentle mixing. The reduction was halted at 0, 5, 10, 20, 40, and 60 min by the addition of iodoacetamide (0.1 M final concentration, Calbiochem) and left for 2 h in the dark. The electrophoretic mobility of BSA following treatment was compared with molecular mass standards with SDS-PAGE under non-reducing conditions (PhastSystem with homogeneous 20% gels and combined Coomassie Blue/silver stain). Molecular mass estimations were performed by scanning gels with the PhastImage Gel Analyzer (Pharmacia) at 603 nm and the PhastImage Analysis Program (Version 1.0: Pharmacia).
Recovery of Endothelial Cell-derived PAF Released from BSA by Dithiothreitol TreatmentTo assess whether the conformation conferred on albumin by disulfide bonds affected the ability of PAF to be detected in assays, ECM and ECCM were exposed to a final concentration of 10 mM dithiothreitol for various periods and then the PAF concentration in media tested in the platelet aggregation bioassay.
This strategy was extended to determine whether such reduction resulted in PAF being transferred from albumin to other hydrophobic surfaces. Amberlite resin beads (XAD-2; E. Merck, Darmstadt, Germany) can act as a solid-phase adsorption matrix for PAF (35). Beads were added to polypropylene chromatography columns (Bio-Rad). Dithiothreitol (0.5 mmol) was added to ECCM (0.25 ml) and applied to the XAD-2 matrix after mixing. The slurry was mixed for 2 h while incubating at 65 °C. Medium was removed by centrifugation (50 g) and the beads washed twice with protein-free medium, followed by seven washes with 1 ml of 30% (v/v) methanol. PAF was eluted with 3 ml of acetonitrile:methanol (75:25; BDH). The eluants were reduced to dryness under nitrogen and then immediately resuspended with the diluent used in the PAF RIA (NEN Life Science Products). PAF was allowed to come to solution during a 1-h incubation at 25 °C. The PAF concentration was measured by RIA.
Statistical AnalysisData were analyzed by comparison of means of two populations using Student's t test (Minitab Statistical Package, Version 10, Pennsylvania State University). The differences in thiol concentration were assessed by two-way ANOVA.
Stimulation of endothelial cells with 2.5 µM A23187 induced all cultures to produce PAF (n = 18). Total PAF production was 22.4 ± 9.9 ng of PAF/106 cells/h (mean ± S.E.), of which 56% was released into the medium. The amount of PAF released into the media by 1 × 106 cells was 12.7 ± 1.4 ng (mean ± S.E.). PAF could not be detected in media in which cells were treated with basal media alone or basal media with vehicle (Me2SO).
In another sample, medium conditioned by 5 × 105 endothelial cells released 3.0 ± 0.2 ng of PAF (mean ± S.E.). However, direct assay of the medium (addition to bioassay without prior organic extraction) failed to detect any PAF present. By contrast, synthetic-PAF added to such media was readily detected without prior extraction. Samples of the same medium were also extracted with 80% ethanol. The ethanol was removed under nitrogen and lipids resuspended with HTFM with 3 mg of BSA/ml. PAF was not detected following this extraction procedure, yet it was readily recovered and detected following Bligh-Dyer extraction.
Pharmacological Characterization of PAF Released from Endothelial CellsWEB 2170 (100 ng/ml) completely inhibited platelet
aggregation induced by 3 ng of synthetic PAF/ml (Fig. 1,
inset). This concentration of WEB 2170 also significantly
(p < 0.001, t test) inhibited platelet aggregation induced by endothelial cell-derived PAF, following its
extraction and partial purification by TLC. The antagonism of
endothelial cell-derived PAF was equivalent to that of the same
concentration of synthetic PAF (Fig. 1).
Susceptibility of Endothelial Cell-derived PAF to PAF:AH
Synthetic PAF or ECCM were exposed to 10% (v/v) untreated
or acid-treated serum for 24 h (Fig. 2). PAF was
still present in ECCM after a 24-h exposure to serum containing PAF:AH,
and the PAF concentration was not significantly different from that
found after exposure to acid-treated serum (in which PAF:AH was
deactivated). By comparison, synthetic PAF in similar media exposed to
serum was degraded after 24 h, but synthetic PAF exposed to
acid-treated serum was unaffected.
Limited Proteolysis of BSA with Pepsin and Trypsin
ECCM or
control medium was subjected to limited proteolysis with pepsin or
trypsin in a similar manner to that previously described (21). The
digestion products are referred to as P-44 (pepsin digestion, 44.4 kDa)
and T-23 (trypsin, 36.9 kDa). Amino-terminal sequencing (21) showed
that P-44 corresponded to amino acids 1-386 of albumin and T-23 to
amino acids 240-583. There was a stoichiometric recovery of
endothelial cell-derived PAF on both P-44 and T-23 after digestion
(Fig. 3), suggesting that it was bound to a site that
was common to both albumin fragments: amino acids 240-386, which
correspond to domain II of albumin (29-31). By contrast, PAF added to
control media was not found at this location. This suggests that PAF
added to solution does not readily bind to that site on albumin
accessed by cell-released PAF.
Susceptibility of Albumin to Reduction by Dithiothreitol
The results of the foregoing experiments are identical to those found for the association of embryo-derived PAF with albumin (21). We were interested to learn whether the binding of cell-released PAF to albumin in this apparently protected configuration involved conformational changes to albumin. The conformation of albumin is largely governed by 17 disulfide bonds in its structure (36). The status of these bonds were examined by observing the susceptibility of albumin to reduction with dithiothreitol. The reduction of disulfide bonds causes an increase in the hydrodynamic volume of the protein in SDS. The consequent reduced electrophoretic mobility (giving the appearance of increased molecular weight) is an indirect measure of the susceptibility to reduction.
Using this strategy, ECM and control media (with synthetic PAF added)
were incubated with 10 mM dithiothreitol at 25 °C for increasing time. Albumin exposed to embryos showed a faster increase in
apparent molecular weight than did control media (Fig.
4). After reduction with dithiothreitol for 20 min, the
electrophoretic mobility of BSA that had not been exposed to embryos
had not changed (68 kDa), while the apparent molecular mass of BSA in
ECM had increased to 76 kDa. This difference between BSA in
embryo-conditioned and control medium may suggest that exposure of
albumin to embryos caused it to be more susceptible to reduction with
dithiothreitol than was BSA not exposed to embryos.
Determination of the Thiol Concentration in Albumin
To assess
whether this apparently greater susceptibility was due to a change in
the thiol status of albumin, the thiol concentration was measured. The
spectrophotometric assay was validated using the thiol-containing amino
acid cysteine (Fig. 5, inset). The relationship between calculated thiol concentration (y axis)
and cysteine concentration (x axis) was y = 0.996x + 0.048, r = 1.000. Fig. 5 shows the
thiol concentration of media in which murine embryos had been cultured
for 24 h (ECM pool) compared with control media with synthetic PAF
added. At all dilutions measured, the BSA in the ECM pool had a
significantly higher thiol concentration than that in the synthetic PAF
pool (p < 0.01, ANOVA), with the thiol concentration
being 0.77 ± 0.04 and 0.59 ± 0.02 µmol of thiol/µmol of
BSA in ECM and control media, respectively (mean ± S.E.).
Bioavailability of Cell-released PAF following Reduction of Albumin with Dithiothreitol
To assess whether the
cell-dependent changes in the thiol status of albumin
affects the binding of PAF in its protected configuration, the
bioactivity (platelet aggregation in vitro) of PAF in
embryo- and endothelial cell-conditioned medium, was measured following exposure of media to dithiothreitol for various periods of time (Fig.
6). As previously observed, PAF bioactivity was not
detected in untreated embryo- or endothelial cell-derived media, but
was detected in preparations following modified Bligh-Dyer extraction. PAF activity was detected in media treated with dithiothreitol for 30 min (being approximately 70-80% of that detected following organic
extraction). Following treatment with dithiothreitol for 60 min, the
amount of PAF detected was less than that observed at 30 min, while no
PAF was detected after 90 min of treatment. Synthetic PAF activity
added to control media (which had been treated with 10 mM
dithiothreitol for 0 or 30 min before assay) was unaffected (results
not shown), indicating that dithiothreitol had no effect on the assay
or on PAF bioactivity. The results suggest that reduction of disulfide
bonds in albumin in cell-conditioned medium altered the association of
PAF and albumin in such a way that PAF became detectable in the
bioassay.
The loss of bioactivity after longer periods of exposure to dithiothreitol (60 and 90 min) may have been caused by the reduction of albumin occurring to a degree that it was no longer a suitable carrier for PAF. A consequence of this may be that PAF was lost from solution onto the hydrophobic surfaces of the incubation vessels. This possibility was tested by the use of a strategy designed to allow for the convenient recovery of any PAF that was lost from albumin following its reduction with dithiothreitol. Dithiothreitol was added to ECCM in the presence of a slurry of Amberlite resin XAD-2 chromatography beads. The beads were to provide a high surface area hydrophobic binding site for PAF released from albumin. PAF was recovered by washing the beads with organic solvents and PAF was measured by RIA. In the absence of dithiothreitol, no endothelial cell-derived PAF was recovered from the XAD-2 beads, suggesting that under such conditions albumin had a higher affinity for PAF. After reduction with dithiothreitol for 60 min, approximately 60% of the estimated PAF present in ECCM was recovered from XAD-2 beads (Table I).
|
The results confirm that human umbilical vein endothelial cells activated by A23187 produce PAF. While a significant proportion of the PAF remained associated with the cells, much of it was released into medium. This PAF was not bioactive in media in a platelet aggregation assay, nor was it recovered following ethanolic extraction. This behavior of endothelial cell-derived PAF contrasts with that of synthetic PAF added to media under the same conditions.
Synthetic PAF was readily detected by platelet aggregation without the requirement for extraction, and was also readily extracted by ethanol (80%). We have shown previously (21) that this behavior of synthetic PAF is independent of the method of its preparation. Cell-released PAF was also apparently different from PAF retained by cells, which can be readily extracted by 80% ethanol treatment (11). The reason why the modified Bligh-Dyer method was able to extract PAF from ECCM but 80% ethanol could not is not clear. The relative volume of methanol used was much greater, being 1:19 (media:methanol) compared with only 1:4 (media:ethanol). We have found that the very slow dropwise addition of media to the larger volume of methanol followed by incubation at room temperature is required for successful extraction. The subsequent secondary extraction of methanol with chloroform may also be important in the successful recovery of PAF. Systematic study of the variables is needed to define the minimum requirements for successful extraction of PAF from media released from cells.
The failure of previous reports to detect the release of PAF from endothelial cells may have been due to the measurement of PAF without extraction, or following simple ethanolic extraction procedures, or using various other mixtures of chloroform and methanol (5-12). It can be concluded that extraction of cell-conditioned media with the methods used in this report are required before release of PAF by cells can be excluded. It also shows that recovery of synthetic PAF added to media does not act as a suitable positive control for the recovery of cell-released PAF.
One study (37) showed that stimulated endothelial cells produce both PAF and its sn-1-acyl analogue (acyl-PAF), with acyl-PAF being the predominant product. That study measured the concentration of PAF and acyl-PAF remaining associated with the cells but did not determine whether either was released into medium. It was suggested that acyl-PAF can mimic the biological actions of PAF, but was substantially less potent. Thus measurement of the bioactive material released by endothelial cells does not exclude the possibility that it was induced by acyl-PAF. The antibody used in the RIA in the current study is more selective for PAF (38) than that used in earlier studies (39). The good agreement on the amount of PAF released by cells detected by the bioassay and the RIA is unlikely if the bioactivity was due to acyl-PAF. The amount of released PAF detected by the RIA was similar after treatment of extracted material with phospholipase A1 (results not shown). Since acyl-PAF, but not PAF, is sensitive to digestion with phospholipase A1 (37), we conclude that most of the bioactivity detected from endothelial cells was PAF. This does not exclude the possibility that acyl-PAF may also be released and bind in a similar fashion to PAF. This question could be addressed by mass spectrometric analysis of the released phospholipids.
It has previously been demonstrated that extracellular albumin is an acceptor for PAF released by cells (40), presumably by removing PAF from the lipophilic environment of the membrane. We observed (21) that the release of PAF by preimplantation embryos was exclusively at domain II of albumin, and the current study shows that this is also the binding site for PAF released by endothelial cells.
BSA (583 amino acids) has 80% sequence homology with human serum
albumin (41), and most structural studies have been performed on human
serum albumin. It is a "heart-shaped" globular molecule, similar to
an equilateral triangle. It is proposed that there are three
cylindrical peptide segments (domains I, II, and III) connected by
solvent exposed -helical chains, which are common sites of
proteolytic cleavage. Domain II contains a hydrophobic core, which can
provide a binding region for hydrophobic molecules (42, 43). Clay
et al. (44) provided kinetic evidence that albumin possessed
four binding sites for PAF which had an average dissociation constant
of 0.1 µM. The rabbit PAF receptor has an estimated
kd for PAF of ~0.5 nM (45). Synthetic
PAF added to control culture medium caused platelet aggregation in whole blood down to a concentration of 0.5 ng/ml (~0.9
nM). However, at a concentration of 3.0 ng/ml (~5.6
nM), endothelial cell-derived PAF in untreated culture
medium did not cause platelet activation. This concentration was well
below the stated kd for PAF binding by albumin, yet
above the kd of the platelet receptor. It might
therefore be expected that the kinetics would favor PAF transfer from
albumin to the PAF-receptor, as was the case for synthetic PAF. The
absence of this suggests that PAF released by cells either binds to
sites on albumin that have a much higher affinity than those described
by Clay et al. (44), or that the conformation of albumin
causes PAF bound at domain II to be solvent- or sterically protected in
a way that does not occur for PAF binding to albumin in the absence of
cells.
The secondary and tertiary structure of albumin is highly dependent on the presence of 17 interchain disulfide bonds, which link 34 (of the available 35) cysteine residues. The free cysteine residue is generally at amino acid 34 in domain I of albumin. This should be detected as 1 µmol of thiol/µmol of BSA, yet approximately 0.5 µmol of thiol/µmol of albumin is normally detected (32, 46). This may be the result of dimerization of albumin (47) in solution or be due to the reactive Cys-34 being "solvent-protected" by the helices in some albumin molecules (48). BSA exposed to embryos expressed more reactive thiol residues than untreated medium with PAF added, showing that binding of PAF per se did not induce this increase in thiol concentration. The increase in thiol concentration was therefore apparently due to a cell-dependent process. The observation that albumin was more readily reduced by dithiothreitol after incubation with embryos suggests that the cell-dependent process involved conformational changes to albumin involving cysteine-cysteine disulfide bonds. It has been shown (49) that interaction of albumin with cells or surfaces caused a conformational change resulting in a mixed population of albumin molecules. While the mechanisms of the change are not well understood, it appears to cause a flattening of the molecule giving it a greater surface area when bound to cells and a higher binding affinity. It has been proposed (49) that this conformational change promotes dissociation of passenger fatty acids, facilitating transfer from albumin to the cell. It will be of interest to determine whether a similar mechanism operates in reverse for the removal of PAF from cells.
Exposure of embryo-conditioned media and endothelial cell-conditioned media to dithiothreitol, under conditions expected to cause reduction of albumin, resulted in a large proportion of the expected PAF activity present to be detected in a direct assay of platelet aggregation (without prior organic extraction). Control experiments confirmed that the aggregation was not caused by the dithiothreitol itself, nor did dithiothreitol reduce the sensitivity of the assay to PAF. This result suggests that binding of PAF to domain II of albumin, which results in its protected configuration, involves protein conformation that is dependent upon the disulfide bonds that can be reduced by dithiothreitol.
The observation that PAF activity was lost with prolonged exposure to dithiothreitol may have several possible explanations. One is that as albumin was reduced (and thus changed conformation) it lost its affinity for PAF, resulting in PAF being adsorbed by the surfaces of the holding tube, as is known to occur in protein-free media (50). Using different types of vessels may reduce this loss of activity. Confirmation that the cause of the time-dependent loss of activity was its loss to hydrophobic surfaces, was the ability to recover PAF from XAD-2 chromatography beads after incubation with dithiothreitol treated ECCM. The results indicate that as the conformation of albumin is altered with reduction, PAF becomes "exposed" and hence available to bind to the platelet receptor, causing platelet aggregation in the bioassay. As reduction proceeds, PAF seems to be readily lost from albumin and bound by other hydrophobic surfaces, such as the test tube or XAD-2 beads.
These experiments provide indirect evidence that the protected nature of PAF's binding to albumin involves disulfide bonds between cysteine molecules. The observation that breaking these disulfide bonds is necessary for making PAF accessible in vitro infers that some form of disulfide isomerization or reduction may be required for PAF's "release" from the cell and binding on domain II of albumin. The changed thiol status of albumin following exposure to cells implicates cellular enzymes in this process.
Studies with embryo-conditioned medium show that, while embryo-derived PAF in untreated media is inactive in vitro (21), upon injection into animals it can induce thrombocytopenia (51). Such studies show that cellular-dependent factors must be involved in the creation and processing of cell-released PAF in its protected form and its exposure in vivo to allow it to be bioactive. Further experiments are required to determine the nature of this mechanism. One possibility may be the presence of cell surface protein disulfide isomerases (52, 53), which might cause reduction of some disulfides resulting in solvent exposure of domain II, facilitating the loading of PAF at this site. It might be speculated that a similar process, in reverse, may be required to allow PAF to be made available at target cells.
While there is some controversy regarding the ability of some cell types to release PAF, for some other cell types such as activated basophils, there is general agreement that PAF is released upon cellular activation. This might suggest that PAF released from some cell types, or under some conditions, is not in the protected configuration described in this report. Systematic investigations are required to assess this possibility.
This study used defined culture media with albumin as the only extracellular macromolecule. The next important question to investigate will be to determine whether PAF binds to albumin in this protected form when a complex protein source such as serum is present. Several studies have detected PAF in blood (54, 55). In view of the high PAF:AH levels present in blood, this is a surprising result, inferring that this PAF may be protected from PAF:AH.
In conclusion, PAF released from embryos and endothelial cells binds to albumin at domain II (amino acids 240-386), protecting it from the hydrolytic effects of PAF:AH in vitro. PAF added to solution does not bind to this site on albumin in vitro. This binding makes extracting and measuring cell-released PAF difficult, but may also act to increase the half-life of PAF released from these cell types. Should such cryptic binding also occur in vivo, it may well influence PAF's half-life and hence its potential to act as a circulating mediator. The impact of such binding by albumin on the kinetics of its recognition by the PAF-receptor in vivo requires investigation.
We thank the following people from Royal North Shore Hospital for assistance: E. Gallery for supply of the human umbilical cords, C. Jackson for help with endothelial cell culture, M. Bonafacio for chromatography advice, A. Koch and J. Arkell for expert technical assistance, and K. O'Neill for preparation of the manuscript.