Granulocyte-Macrophage Colony-stimulating Factor Is a Stimulant of Platelet-activating Factor and Superoxide Anion Generation by Human Neutrophils*

Human granulocyte-macrophage colony-stimulating factor (GM-CSF) was studied for its ability to stimulate the synthesis and release of the inflammatory mediator platelet-activating factor (PAF) from human neutrophils as measured by bioassay and incorporation of [‘Hlacetate into PAF. GM-CSF stimulated the synthesis but not the release of PAF from neutrophils. PAF syn- thesis took place in a time- and concentration-depend-ent manner, was dependent on a pertussis toxin-sen- sitive G protein and could be inhibited by antibodies to GM-CSF. On the other hand, pre-incubation of neutro- phils with GM-CSF followed by stimulation with the bacterial tripeptide in

Human granulocyte-macrophage colony-stimulating factor (GM-CSF) was studied for its ability to stimulate the synthesis and release of the inflammatory mediator platelet-activating factor (PAF) from human neutrophils as measured by bioassay and incorporation of ['Hlacetate into PAF. GM-CSF stimulated the synthesis but not the release of PAF from neutrophils. PAF synthesis took place in a time-and concentration-dependent manner, was dependent on a pertussis toxin-sensitive G protein and could be inhibited by antibodies to GM-CSF. On the other hand, pre-incubation of neutrophils with GM-CSF followed by stimulation with the bacterial tripeptide formylmethionylleucylphenylalanine caused PAF synthesis and release. The effect of GM-CSF was qualitative and not simply the result of larger amounts of PAF being synthesized since similar amounts were generated in response to the calcium ionophore A23187 but no released PAF could be detected. In functional studies GM-CSF stimulated superoxide anion generation from neutrophils with a time and dose relationship that paralleled PAF synthesis. In addition, the serine protease inhibitor L-l-tosylamide-2-phenylethyl chloromethyl ketone, which inhibits PAF synthesis, reduced PAF accumulation as well as superoxide generation, raising the possibility of a causal relationship between cell-associated PAF and cell activation. These results identify PAF as a direct product of GM-CSF stimulation in neutrophils where it may play a role in signal transduction and demonstrate that PAF is released only after subsequent neutrophil stimulation. The selective release of PAF may play a role in regulating and amplifying the inflammatory response.
Granulocyte-macrophage colony-stimulating function (GM-CSF)' is a glycoprotein of M, 22,000 initially described as a multilineage hemopoietic growth factor that stimulates the production of neutrophils and other myeloid cell types (1,2 ) . It has become apparent, however, that GM-CSF is poten-*This work was supported by grants from the National Health and Medical Research Council (Australia). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
tially also an important regulator of the inflammation since GM-CSF: (i) is produced by cells present at inflammatory sites such as activated T lymphocytes, tissue macrophages, endothelial cells, and mast cells (3)(4)(5); (ii) enhances the function of mature neutrophils in vitro and in vivo as well as that of monocytes, eosinophils, and basophils (6-12); (iii) prolongs the survival of these cells in vitro (13); and (iv) can be detected at sites of inflammation (14).
Extensive evidence from a number of laboratories indicates that GM-CSF acts as a "priming" agent on neutrophils, enhancing superoxide anion release and leukotriene B4 production in response to second stimuli such as FMLP, C5a, and PAF (8, 9,15). Despite these and other biological effects of GM-CSF in vitro and its incipient clinical use, little is known about effects of GM-CSF which may shed light on the mechanism of action of this cytokine. Investigations of signals stimulated by GM-CSF have yielded mainly negative results with GM-CSF unable to translocate protein kinase C (16,17), activate phospholipase C (16)(17)(18), or elevate calcium levels (16). However, GM-CSF has been reported to stimulate the release of arachidonate from neutrophils pre-loaded with [3H]arachidonic acid (16, 19), a potentially important observation since in human neutrophils arachidonic acid is preferentially mobilized from the sn-2 position of 1-alkyl-linked phosphatidylcholine, a major phospholipid subclass in neutrophils (20). This release leads to the generation of lyso-PAF which may be acetylated at the sn-2 position by a calcium-dependent and -specific acetyltransferase to form PAF (21, 22), one of the most potent inflammatory mediators.
In this study, we show that indeed human GM-CSF stimulates the production of PAF from human neutrophils in a time-and concentration-dependent manner and through a pertussis toxin-sensitive G protein. In addition, GM-CSF is shown to cause superoxide anion generation with a dose response and kinetics paralleling PAF formation. Importantly, the GM-CSF-stimulated PAF remained wholly cellassociated unless neutrophils were further stimulated with FMLP in which case released PAF was readily detected. These results identify PAF as a product of GM-CSF stimulation and raise the possibility of PAF playing a second messenger role. Moreover, the release of PAF from GM-CSF-primed neutrophils may play an important role in inflammatory and immune responses.

Purification of Human Neutrophils
Neutrophils were obtained from the peripheral blood of normal volunteers after dextran sedimentation (Dextran T-500, Pharmacia, Uppsala, Sweden) and density gradient centrifugation at 450 X g on Lymphoprep (Nycomed, Oslo, Norway) followed by hypotonic lysis of erythrocytes using 0.2% sodium chloride solution. This solution was then brought to the correct osmolarity using a 1.6% sodium chloride solution. The cell preparations were resuspended to various concentrations ranging from IO6 to 5 X lO'/ml in RPMI 1640 (pH 7.3) plus 0.1% bovine serum albumin (BSA) Fraction V, fatty acidfree (Boehringer, Sydney, Australia) and 20 mM HEPES (Sigma). The assays were carried out in this medium unless otherwise stated. This method yielded cells which were >99% viable by trypan blue exclusion and >98% identifiable as neutrophils.
Recombinant Human Granulocyte-macrophage Colony-stimulating Factor and Other Reagents Recombinant human GM-CSF was from lot no. 9A01N013 containing 9.3 X IO6 units/mg and was 99% pure. This was generously provided by Dr. S. C. Clark (Genetics Institute, Cambridge, MA).
The endotoxin content of this material and of GM-CSF diluted in medium was <0.002 ng/ml as measured by the limulus amebocyte lysate assay. rh tumor necrosis factor (TNF)-a was from lot no. 3056-55 containing 5 X lo7 units/mg and was 99.8% pure (generous gift from Genentech, South San Francisco, CAI. Anti-GM-CSF and antiinterleukin-3-specific sera were generous gifts from Genetics Institute. Purified pertussis toxin was a gift from Dr. G. Barrit (Flinders Medical Centre, Adelaide, Australia). FMLP was obtained from Sigma, as were the calcium ionophore A23187 and phorbol myristate acetate (PMA).
Assay for PAF PAF Synthesis and Extraction-Purified neutrophils at concentrations which ranged from lo6 to 5 X 107/ml were incubated in assay medium at 37 "C with appropriate stimulus or medium for defined periods of time (see "Results"), after which the cells were rapidly cooled on ice and aliquots dispensed and centrifuged at 4 "C for 5 min at 450 X g. Supernatants were assayed without prior extraction, whereas cells were extracted into ice-cold 80% ethanol overnight.
Following removal of the precipitate by centrifugation at 4 "C for 5 min at 10,000 X g, the extract was evaporated to dryness under reduced pressure. The dried extracts were reconstituted in 250 p1 of Tyrode's solution containing 0.25% BSA and then bioassayed for their ability to induce rabbit platelet aggregation.
PAF Bioassay-PAF was detected by the aggregation of washed rabbit platelets as described previously (21). Aggregation measurements were carried out using a chronolog 540 aggregometer calibrated with the platelet suspension to set minimum light transmission and the resuspension buffer to set maximum light transmission. Aggregations were performed at 37 "C with stirring. Aggregation responses were assayed by the addition of 450 pl of platelet suspension (-1 X lo8 platelets) and 50 p1 of reconstituted sample of supernatant to a siliconized glass cuvette in the absorbance chamber. Responses were allowed to develop until a maximum change in light transmission occurred. The amount of PAF was expressed as femtomoles/107 cells and was quantitated by comparison with a standard curve constructed using known concentrations of hexadecyl PAF (Novachem, Switzerland).
Criteria for PAF Identification-Biologically active material extracted from cells was characterized as PAF on the basis of the following: ability to induce indomethacin-resistant platelet aggregation; inhibition of bioactivity by a selective PAF receptor antagonist WEB2086 (1 p~) (Boehringer Ingelheim, Sydney, Australia) and comigration of extracted samples with authentic tritiated PAF on silica gel thin layer chromatography mobile phase (chloroform/methanol/ HaO/acetic acid, 65:35:6:0.1, v/v).

Labeling of Neutrophils with r3H]Acetate and Agonist Stimulation
Purified neutrophils at 107/ml were pre-incubated with 20 pCi/ml I3H]acetate (Amersham, Bucks, United Kingdom) for 10 min at 37 "C and incubated for 60 min in the presence or absence of GM-CSF and for a further 15 min at 37 "C with FMLP or medium. The reaction was stopped by extraction.
Extraction, Separation, and Quantification of PAF The lipids were extracted from the stimulated neutrophils using the Bligh/Dyer extraction method. Initially, the reaction was terminated by the addition of 2 ml of methanol (1% acetic acid) to the neutrophil suspension (0.9 ml). The precipitate was removed by centrifugation, 1 ml of chloroform was added to the supernatant, and the tubes were vortexed to extract the PAF and other phospholipids. Phase separation was achieved by the addition of 2 ml of chloroform/ H20 (50:50), and the lower phase was evaporated to dryness under reduced pressure. PAF in the chloroform phase was then separated from other lipids on TLC. The samples were visualized with Is vapor and scraped from the TLC plates in a narrow zone based on their comigration with authentic PAF (I3H]PAF was used in extracts of endogenous PAF; synthetic unlabeled PAF was used when separating PAF labeled by I3H]acetate incorporation). The bioactivity of the samples from all zones of the TLC plates was quantified by the platelet bioassay.

Superoxide Anion Generation
Superoxide anion release was measured in a colorimetric assay based on the reduction of ferrocytochrome c. Briefly, 100 pl of purified neutrophils (lo6 cells) was added to 100 p1 of freshly prepared cytochrome c (Sigma, type VI; 12.4 mg/ml). Various concentrations of cytokines were added in volumes of less than 10 pl and the reaction mixtures were made up to 1 ml with medium and incubated in polypropylene tubes that were continuously shaken. The mixtures were then incubated at 37 "C for various periods of time up to 60 min (see "Results"). Under these conditions, >98% neutrophils could be recovered indicating that little or no adherence to the surface of the tubes had occurred. Cells were incubated with FMLP (Sigma) or with PMA (Sigma) for a further 15 min. After incubation, the cells were rapidly cooled and pelleted at 4 "C, and supernatants were transferred to plastic disposable cuvettes (Kartelle Plastics, Adelaide, Australia). Superoxide production was measured by the reduction of cytochrome c monitored at 550 nm using a DU-50 spectrophotometer (Beckman Instruments, Palo Alto, CA). Levels of superoxide were quantitated as described previously (9).

Statistical Analysis
All experiments were performed 2-6 times. Within one representative experiment, means of values were compared by the Student's t test. When evaluating data from several donors, the Wilcoxon's matched pairs test was used.

Characterization of GM-CSF Stimulation of PAF Synthesis
in Human Neutrophils-GM-CSF stimulated PAF synthesis in human neutrophils. In experiments using neutrophils from 16 different individuals, GM-CSF consistently increased the levels of cell-associated PAF (Fig. 1). FMLP also stimulated PAF synthesis; however, most individuals responded more to GM-CSF than to FMLP. In these experiments no PAF was detected in the cell supernatants (limit of detection, 100 fmol/ ml) .
The stimulation of PAF synthesis by GM-CSF was characterized by carrying out time course and dose-response experiments. A time course experiment with 3 X lo-' M GM-CSF representative of four others showed that the optimal time of GM-CSF stimulation was 45 min with a clearly detectable effect a t 10 min (Fig. M). During this time, the levels of PAF in neutrophils incubated with medium alone remained low and relatively constant. A dose-response experiment with GM-CSF representative of six others showed that the maximal stimulation of PAF accumulation could be achieved at a concentration of 3 x 10-l' M, with a 50% Human Granulocyte Activation maximal stimulation at approximately 3 X lo-" M (Fig. 2B).
While GM-CSF or FMLP, used separately, directly stimulated PAF synthesis, GM-CSF and FMLP showed a synergistic effect when used sequentially. In this case, the amount of cell-associated PAF was 5,945 fm01/107 cells (mean of 14 experiments), a value significantly higher than those obtained with GM-CSF (1,086 fm01/107 cells) or FMLP (935 fm01/107 cells) alone ( p < 0.001 by the Wilcoxon's matched pairs test). To establish that PAF accumulation in neutrophils was stimulated by GM-CSF and not a contaminant, two types of controls were carried out. First, endotoxin levels were measured in the medium containing GM-CSF. By the limulus amebocyte assay this preparation contained <0.002 ng/ml endotoxin. Second, the use of a polyclonal anti-GM-CSF antibody inhibited the ability of GM-CSF to directly stimulate PAF synthesis and to prime neutrophils for an enhanced response to FMLP (Table I).
The involvement of G proteins in the stimulation of PAF synthesis by GM-CSF was examined by pre-incubating neutrophils with 500 ng/ml pertussis toxin, a concentration previously shown to be optimal in inhibiting FMLP stimulation (24). Pertussis toxin inhibited PAF accumulation in neutrophils stimulated with GM-CSF, TNF-CY, or FMLP (Fig. 3). This inhibition was due to a direct effect on G proteins and not to a toxic effect on neutrophils since pertussis toxin did not inhibit stimulation of PAF accumulation by PMA, an agonist that directly interacts with protein kinase C.
In order to seek further evidence that GM-CSF stimulates PAF synthesis, a second technique was used. Neutrophils were pre-incubated with [3H]acetate and examined for the amount of [3H]acetate incorporated into a polar lipid that comigrated with authentic PAF on TLC following stimulation. GM-CSF was found to stimulate the incorporation of [3H]acetate into PAF (Table 11).
GM-CSF Primes Neutrophils for PAF Release-Although GM-CSF stimulated PAF synthesis in human neutrophils this remained cell-associated and no PAF could be detected in the cell supernatants under these conditions (Table 111) or using a range of cell numbers (lo6 to 5 x lo7 neutrophils/ml) or of BSA concentrations (0.01-1%) in the medium (data not shown). In addition, we examined the possibility that endogenous inhibitors of PAF-induced aggregation were masking the release of PAF. Supernatants from FMLP-stimulated neutrophils had no effect on the ability of synthetic PAF (6-600 fmol) to induce platelet aggregation.
However, PAF release was detected in neutrophils primed with GM-CSF and stimulated with FMLP (Table 111). As a control, neutrophils were stimulated with the calcium ionophore A23187. In this case equally large amounts of PAF were cell-associated but no PAF release could be detected, indicating that the PAF detected in the supernatants of neutrophils treated with GM-CSF and FMLP is not simply a result of the increased levels of cell-associated PAF but of a qualitative     In another set of experiments, neutrophils from eight different individuals were primed with GM-CSF and stimulated with FMLP for 5 min at 37 "C. In seven of these individuals the amount of cell-associated PAF was 12,321.8 & 4,469.6 (mean +-S.E.) fm01/107 cells and the amount of PAF in the supernatant was 790.7 k 355.4 fm01/107 cells with the cell supernatant uersus cell-associated PAF ratio ranging from 0.02 to 1.5. In neutrophils from a different individual, however, the cell-associated PAF was 490 fm01/107 cells, but no released PAF was detected.
GM-CSF-stimulated PAF Accumulation Parallels Neutrophil Actiuation-To investigate the relationship between the accumulation of PAF and cell function, GM-CSF was studied for its ability to stimulate neutrophil superoxide anion generation. We found that although at early time points GM-CSF stimulated little 0; release, this increased with prolonged incubation (Fig. 4A). This stimulation of 0, release was concentration-dependent with maximal levels of 0, released a t a GM-CSF concentration of 3 X 10"' M and 50% stimulation achieved with a GM-CSF concentration of 3 X lo-" M ( Fig. 4B). Specificity controls showed that the stimulation of 0; by GM-CSF could be prevented by anti-GM-CSF but not by anti-interleukin-3 serum (Fig. 5A). In addition, the stimulation of 0; production by GM-CSF as well as TNF-a and FMLP was inhibited by pertussis toxin-sensitive G protein.
Effect of TPCK on PAF Synthesis and Function in Human Neutrophils-The effect of TPCK, a serine protease inhibitor which blocks the synthesis of PAF (25), was investigated on GM-CSF-stimulated PAF accumulation and function. TPCK reduced both the accumulation of PAF and the production of 0; (Table IV). Parallel measurements of lactodehydrogenase release demonstrated that the inhibition of TPCK was not due to toxic effects on neutrophils.

DISCUSSION
This study shows that human GM-CSF stimulates the synthesis of PAF in human neutrophils, and that PAF is released upon subsequent stimulation with FMLP. The release of PAF from human neutrophils in uivo may amplify the inflammatory response by increasing vasopermeability, enhancing the recruitment of leukocytes, and directly activating and priming granulocyte function.
Our results disagree with a previous report in which GM-CSF was shown not to stimulate PAF formation in neutrophils (26). This discrepancy may be explained by the use of a very sensitive bioassay in the present studies. However, we also detected PAF formation by measuring the incorporation of [3H]acetate into PAF which was the method used in the previous study. Neutrophils were incubated with [3H]acetate throughout the exposure to GM-CSF, whereas in the previous report (26) [3H]acetate was added to the cells only for the last 10 min of GM-CSF incubation. The latter approach may be suboptimal for detection of GM-CSF effects. In addition, the [3H]acetate-labeling technique appears to be a less sensitive index of PAF formation than measuring endogenous PAF levels since in our experiments GM-CSF increased the latter by 2.8-fold, whereas acetate incorporation increased by 1.8fold. This may be related to the bioassay measuring net amounts of PAF regardless of the pathways involved, whereas in the ["]acetate method the label is incorporated into several pools including a PAF precursor pool, and assumes that the specific activity of [3H]acetyl-CoA remains unchanged following stimulation.
Specificity controls demonstrated that the stimulation of PAF synthesis was due to GM-CSF and not to lipopolysaccharide, a bacterial product that induces PAF production in neutrophils a t concentrations above 10 ng/ml (27). First, the experiments described here used highly purified GM-CSF which contained undetectable levels (<0.002 ng/ml) of lipopolysaccharide at the highest concentration used. Second, a polyclonal antibody specific for GM-CSF significantly re-  PAF production required a relatively prolonged exposure of neutrophils to GM-CSF. PAF was detectable by 10 min and plateau levels were reached by 45 min. In contrast, FMLP has been shown to maximally stimulate PAF formation at 5 min (28). In other experiments not shown here, 300 PM GM-CSF and 45 min preincubation were also optimal in priming neutrophils for enhanced PAF production in response to subsequent stimulation with FMLP.

PAF synthesis and release in human neutrophils stimulated bv GM-CSF and FMLP
The synergistic effect of GM-CSF and FMLP in PAF formation may be related to the combined effect of two second messenger pathways involving phospholipase Az and phospholipase C. Thus both GM-CSF and FMLP liberate arachidonic acid (16,19) probably by activation of phospholipase Az which may lead to the accumulation of lyso-PAF and conversion to PAF by the rate-limiting enzyme acetyltransferase. Both phospholipase AP and acetyltransferase can be activated by phosphorylation and calcium (29).
Stimulation of PAF biosynthesis by GM-CSF was dependent on a pertussis toxin-sensitive G protein. Pre-incubation B of neutrophils with pertussis toxin did not significantly reduce either the synthesis of PAF or superoxide anion generation in response to PMA, indicating that the inhibition of GM-CSF-induced PAF synthesis was specific. These results are consistent with the GM-CSF receptor being coupled to a pertussis toxin-sensitive G protein for the stimulation of some neutrophil functions, the others being induction of c-fos mRNA (30), and the uptake of phosphate (31).
An important finding was the demonstration that the synthesis of PAF stimulated by either GM-CSF or FMLP remained cell-associated. PAF, however, could be readily detected in the cell supernatants after sequential stimulation with GM-CSF and FMLP. This did not appear to be simply due to an overspill of the now much increased cell-associated PAF into the supernatants since the calcium ionophore A23187 stimulated similar cell-associated levels of PAF but no detectable release, suggesting that GM-CSF induces qualitative changes in human neutrophils leading to PAF release. This was a consistent finding observed in 10 of 11 individuals examined. Neutrophil heterogeneity in the amount of PAF produced and in the relative amounts secreted were also  The release of PAF has important implications since PAF is a powerful modulator of inflammation, increasing vascular permeability (34), chemotaxis of leukocytes (35), and the release of cytokines from monocytes and macrophages (29). In addition, local levels of neutrophil-derived PAF may act in a n autocrine manner to directly stimulate granulocyte function as shown in vitro with high doses of PAF or to "prime" neutrophils as shown with lower doses of PAF (36). In this respect, PAF would operate as an autocrine factor capable of feeding back on the same cells that release it.
The observation that GM-CSF induced PAF biosynthesis is intriguing and raises the possibility that PAF functions as a second messenger in neutrophils (23,37, 38). In an effort to relate PAF generation to neutrophil function we investigated the effect of GM-CSF in stimulating the neutrophil respiratory burst. We found that GM-CSF caused significant 0; production which was specific, since it could be inhibited by antibodies to GM-CSF, and was mediated through a pertussis toxin-sensitive G protein. Previous experiments by us and others (9,39,40) showed that GM-CSF enhanced 0; production in response to a second stimulus while by itself having little or no effect. The inability to detect stimulation of 0; generation by GM-CSF alone was probably due to the fact that cytochrome c was present only during the second incubation (5 min) with FMLP but was not included in the first incubation with GM-CSF. Superoxide anion, having a half life of the order of milliseconds, would have dismutated spon-taneously before being able to reduce cytochrome c. Similar results have been recently reported with adherent neutrophils where 0; (41) and HzOz (42) were measured. In these cases GM-CSF stimulated 0; and HzOZ production from adherent neutrophils with a similarly prolonged time course. Although the neutrophils used in the present study were not adherent to the incubation tubes we cannot rule out some transient cell-cell or cell-tube contact nor some activation during the purification procedure (43). Thus these experiments cannot unequivocally distinguish between a direct effect of GM-CSF and an enhancing effect on preactivated cells.
The stimulation of 0; production by GM-CSF provides an important functional correlate to PAF production. The temporal correlation between PAF generation and 0; release can also be extended to other functions including surface receptor upregulation (9), adherence (33), and degranulation (44). Inhibition experiments also seem to support the notion of PAF playing a role in signal transduction. Thus, inhibition of GM-CSF stimulated PAF synthesis by the serine protease inhibitor TPCK, a compound previously used to inhibit TNF-astimulated PAF production (25), caused a parallel decrease in superoxide anion release (Table IV) and in adherence to human endothelial cells,' without causing the release of the vital enzyme lactodehydrogenase (Table IV). However, TPCK could affect other pathways and its inhibitory effect suggests but does not prove a causal relationship between PAF synthesis and superoxide production. In a study using rabbit neutrophils, the specific PAF receptor antagonists WEB2086 and CV6209 inhibited FMLP-induced superoxide anion production and leukotriene B4 formation (38), suggesting a role for PAF in the signal transduction process.
The accumulation of GM-CSF-stimulated PAF may be also related to the priming effect of this cytokine. Small amounts of PAF may be bound immediately to specific surface or intramembranous receptors causing priming of the cells. The use of established PAF antagonists as well as developing antagonists that cross the cell membrane should facilitate the testing of these hypotheses.