Platelet Factor 4-Induced Neutrophil-Endothelial Cell Interaction: Involvement of Mechanisms and Functional Consequences Different From Those Elicited by Interleukin-8

IN THE PAST, EVIDENCE has been accumulated that inflammation and thrombosis can no longer be regarded as processes occuring independent of each other, but they have been recognized as highly overlapping and interactive events. Thus, activated platelets not only secrete mediators involved in the regulation of hemostasis, but also various other components modulating wound repair as well as the functional activities of inflammatory cells. After platelet activation, blood leukocytes and the vascular endothelium are the first cellular elements to become exposed to the platelet release products.1,2 A considerable proportion of these consists of 2 α-granule proteins that belong to the CXC-subfamily of chemokines, the platelet factor 4 (PF-4) and the connective tissue-activating peptide III (CTAP-III), which both are found in serum at micromolar concentrations.3-5Although these chemokines are structurally closely related molecules, their functional activities for polymorphonuclear neutrophil granulocytes (PMN) were found to be different in many respects. The CTAP-III represents an inactive precursor of the neutrophil-activating peptide 2 (NAP-2),6 with the latter becoming rapidly generated through limited proteolysis of CTAP-III by a PMN-associated protease.7,8 NAP-2 behaves like a classic CXC-chemokine in that it stimulates neutrophils to undergo chemotactic migration, degranulation, and adherence through interaction with the 2 types of G protein-coupled interleukin-8 (IL-8) receptors (CXCR-1 and CXCR-2) that are expressed on these cells.9,10 Although NAP-2 differs from the prototype CXC-chemokine IL-8 by its preferential binding to CXCR-2, by its capacity to attract neutrophils over a wider range of concentrations,11 and by a lower potency for the induction of other functions, it may be regarded as a first-line mediator within the vasculature.

By contrast, the role of PF-4 is less clear. As we could recently show, this chemokine neither interacts with receptors CXCR-1 or CXCR-2 on neutrophils nor does it induce chemotaxis or degranulation in these cells.12 Likewise, none of the additional classic functions induced by other CXC-chemokines, including intracellular Ca²⁺-flux and adhesion to protein-coated surfaces, could be elicited by PF-4 alone in suspended neutrophils.12 Instead, a more specialized role for the chemokine was indicated by its requirement for a costimulus, ie, tumor necrosis factor-α (TNF-α), to induce a restricted spectrum of effector functions, such as the selective exocytosis of secondary granule contents (but not primary granule contents) and the enhancement of neutrophil adhesion to gelatin and plasma proteins.12 Moreover, investigating PMN for the presence of specific PF-4 binding sites, we obtained evidence that the chemokine's functions are mediated through an integral chondroitin sulfate proteoglycan not related to the G protein-coupled 7-transmembrane-domain receptors for chemokines and not responding to the ligand by a Ca²⁺ signal.13 As a further difference to other chemokines, we observed that binding of PF-4 to its receptor requires tetramerization of the chemokine.13Because this would only occur at relatively high concentrations of the chemokine (from 200 nmol/L PF-4 on), it could be envisaged that the environment in which PF-4 may activate neutrophils in vivo would be restricted to sites of acute platelet activation. In a situation in which platelet activation occurs as a consequence of an ongoing inflammatory process, the costimulus TNF-α would be readily available.

However, this model imposes questions as to the conditions that will allow neutrophil stimulation by PF-4 at the very onset of inflammation, eg, in a situation in which platelet activation occurs as a consequence of exogenous mechanical lesion of the tissue. Because proinflammatory cytokines are virtually absent during this initial stage, alternative costimulatory mechanisms would be required. In our present report, we have dealt with this question by investigating neutrophil activation by PF-4 in a setting more close to in vivo conditions in that we examined the chemokine's stimulatory potential in the presence of unstimulated endothelial cells (EC). We observed that, under these conditions, PF-4 did not require exogenous TNF-α to induce PMN adhesion to an EC layer and that adhesion was mediated by surface molecules different from those activated by IL-8. Furthermore, PF-4–stimulated PMN-endothelial interaction was also involved in providing a costimulatory signal for the induction of secondary granule exocytosis, whereas IL-8–stimulated cells did not respond by exocytosis. Altogether, our results provide further evidence for a special role of PF-4 in neutrophil activation.

Human monocytic recombinant IL-8 (rIL-8; ie, the 72-residue isoform) was obtained from Pepro Tech Inc (Rocky Hill, NJ). Human natural PF-4 was purified in our laboratory from release supernatants of thrombin-stimulated platelets in a 3-step procedure as previously described.12 Briefly, the major contaminant β-thromboglobulin antigen (β-TG Ag) was first removed by immunoaffinity chromatography. PF-4 in the flow-through was then further enriched using a heparin-Sepharose affinity column (Pharmacia/LKB, Freiburg, Germany) and was finally purified to homogeneity by high-performance liquid chromatography (HPLC) on an analytical cyanopropyl column (4.6 × 250 mm, 5 mm, wide pore; Baker Research Products, Phillipsburg, NJ). Eluates and fractions were screened for the presence of potential contamination by β-TG Ag by ELISA as described elsewhere.12 Detection of PF-4 was performed by the use of a quantitative sandwich ELISA. Briefly, wells were coated with 5 μg/mL monoclonal antibody (MoAb) PF1 (see below) specific for PF-4, in 0.1 mol/L bicarbonate, pH 9, overnight at 4°C. After extensive washing, all subsequent incubation steps with antigen samples and immunoreagents were performed in dilution buffer (phosphate-buffered saline [PBS]-Tween/15% bovine serum) at 37°C for 1 hour. A polyclonal rabbit anti–PF-4 serum (Alexis, Grünberg, Germany) was used as detecting antibody, and development was performed as described.14 The final PF-4 preparation exceeded 99% purity, containing no detectable protein contaminants according to analyses in silver-stained sodium dodecyl sulfate-polyacrylamide gels and by automated N-terminal amino acid sequencing (kindly performed by Dr A. Petersen, Department of Clinical Medicine, Forschungszentrum Borstel, Borstel, Germany).

A murine MoAb against PF-4 (clone PF1) was generated in our laboratory after immunization of Balb/c mice with horse myoglobin-conjugated human PF-4, according to standard protocols. The antibody (IgG1 isotype) specifically recognized PF-4 as evidenced by total competition of its binding to solid-phase–coated PF-4 by excess of soluble antigen. The antibody showed cross-reactivity neither with bovine or rabbit PF-4 nor with the related human CXC-chemokines β-TG Ag, IL-8, interferon γ inducible protein-10 (IP-10), or MGSA, as assayed by the same method. MoAbs directed against CD18 (clone MHM23), CD11a (clone MHM24), CD11b (clone 2LPM19C), CD54 (clone 6.5B5), and CD31 (clone HEC7) as well as murine isotype control antibodies IgG2a and IgG2b were all purchased from Dako (Hamburg, Germany). Anti-CD49d and anti-CD11c MoAbs were obtained from T Cell Diagnostics Inc (Woburn, MA), whereas MoAbs directed against CD62L (clone Dreg56) and CD102 (clone BT-1) were from Coulter-Immunotech (Hamburg, Germany). MoAbs against CD106 (clone 1.G11B1) were purchased from Endogen (Woburn, MA). An antibody directed against human IL-2 (clone B0-7)15served as an IgG1-isotype control. Rabbit polyclonal antilactoferrin was purchased from Sigma (Deisenhofen, Germany) and secondary goat-antimouse antibodies (dichlorotriacinylaminofluorescein [DTAF]- or horseradish peroxidase [HRP]-conjugated) as well as donkey-antirabbit (HRP-conjugated) were obtained from Dianova (Hamburg, Germany).

PMN were routinely isolated from citrated blood of healthy single donors by gradient centrifugation on Ficoll-Hypaque to a purity greater than 95% in all events, as previously described.7Viability was examined by trypan-blue exclusion and exceeded 98% in all experiments. Human EC were isolated from umbilical cord veins by collagenase treatment and cultured in dishes coated with fibronectin, as described previously.16 17 The cells were maintained in M199 (Biochrom, Berlin, Germany) supplemented with 1% penicillin/streptomycin, 1% L-glutamine (both from Biochrom), 5% fetal calf serum (FCS), 30 μg/mL EC growth factor (both from Boehringer Mannheim, Mannheim, Germany), and 20 μg/mL heparin (Sigma). Cells were subcultured after trypsinization (0.5% trypsin solution, supplemented with 0.2% EDTA; Biochrom) and used throughout passages 2 to 4. In experiments performed with fixed cells, PMN or EC were treated with 3% paraformaldehyde in PBS for 5 minutes and subsequently washed 4 times with D-PBS/0.1% bovine serum albumin (BSA) (low endotoxin BSA; Serva, Heidelberg, Germany) before use.

EC were grown in microtiter plates for 2 to 4 days to allow formation of confluent layers. Cell were then washed 2 times with warm (37°C) assay buffer (D-PBS/0.1% BSA supplemented with 0.9 mmol/L CaCl₂ and 0.5 mmol/L MgCl₂) directly before use. Then, 150-μL aliquots of PMN (2 × 10⁵ cells) suspended in assay buffer were added to the washed EC together with PF-4 or IL-8 at the concentrations indicated. In some experiments, stimulation with the chemokines described above was performed in the presence of various antibodies (10 μg/mL each) as indicated in the text. Because some of the antibody samples contained sodium-azide as a preservative, the corresponding isotype controls were also supplemented with azide to identical concentrations. No differences in cellular reactivity were seen in comparison with azide-free control samples. In experiments in which PMN or EC were preincubated with PF-4, the respective cells were washed after 20 minutes at 37°C with an excess of assay buffer, and PMN were then added to EC as described above. Neutrophils were allowed to adhere to EC for 20 minutes at 37°C. Nonadherent cells were pelleted to one edge of the wells by centrifuging the plates at an angle of 45° at 150g for 1 minute at room temperature, unless otherwise indicated. Subsequently, 100-μL aliquots of supernatant were recovered for determination of lactoferrin (see below), and the pellet of nonadherent cells was removed by careful aspiration. Cells remaining adherent to EC were lysed in assay buffer containing 0.1% Triton X-100, and the number of adherent PMN was determined by measurement of neutrophil-specific endogenous β-glucoronidase enzymatic activity as described for the neutrophil chemotaxis assay.18 Cell numbers were calculated by means of a standard of lysed cells run in parallel. In experiments performed in the presence of antibodies, no differences between the various isotype controls were observed. Therefore, only data obtained with the IgG1 controls are indicated in the figures. In experiments performed with paraformaldehyde-fixed PMN, adherent neutrophils were resuspended in assay buffer and cell numbers determined in a cell counter (Coulter, Krefeld, Germany). As seen in control assays, cell numbers determined by the latter method did not differ by more then 4.2% from those determined using the endogenous β-glucuronidase method.

As mentioned above, supernatants for the determination of PMN granule marker release were taken immediately after sedimention of nonadherent cells and subsequently monitored for contents of lactoferrin using a quantitative sandwich ELISA as described elsewhere.10Release rates for lactoferrin were expressed as the percentage of total content determined in lysates of detergent-treated PMN prepared in 0.1% hexadecyl-trimethyl-ammoniumbromide. Because some experiments were performed with PMN in the presence of fixed EC, side-effects caused by the potential presence of residual paraformaldehyde had to be excluded. Therefore, supernatants from fixed EC incubated in the absence of PMN were collected and used as assay buffer in an exocytosis assay with PF-4–activated PMN. No difference in exocytosis was seen between samples receiving this medium as compared with samples receiving freshly prepared assay buffer.

Neutrophils were incubated with stimuli and EC for 20 minutes at 37°C as described for the adhesion assay described above, and microtiterplates were subsequently transferred 10 to 15 minutes on ice to allow detachment. After careful resuspension, PMN were washed with cold D-PBS followed by incubation with different antibodies (2 μg/mL each) as indicated in the text for 30 minutes on ice. After labeling with secondary DTAF-conjugated goat-antimouse IgG (15 μg/mL), samples were analyzed in a flow cytometer (model FACStar PLUS; Becton Dickinson, Heidelberg, Germany). To allow comparison of fluorescence intensities obtained in different experiments, instrument settings were calibrated with fluorescein-labeled latex beads (Becton Dickinson).

In a first set of experiments, the ability of PF-4 to induce PMN adhesion to a monolayer of unstimulated EC was examined over a wide range of stimulus concentrations in comparison with that of IL-8. Moreover, the capacity of PMN to respond by exocytosis under these conditions was investigated by determining the amount of secondary granule marker lactoferrin released into the culture supernatants. As shown in Fig 1A, PF-4 at concentrations of 0.4 μmol/L and greater induced PMN adhesion to EC in a dose-dependent manner, with maximum adhesion (∼64% of PMN added) being obtained within a range of 2 to 10 μmol/L of the chemokine. As expected, IL-8 induced PMN adhesion at substantially lower concentrations (in the nanomolar range), but with an efficacy remarkably inferior to that of PF-4 (∼37% of PMN added; Fig 1B). As a further difference, practically no lactoferrin could be detected in cell culture supernatants derived from IL-8–stimulated cells, whereas increasing levels of the granule marker (up to 18% of total content of PMN added) were found with PF-4–stimulated cells (Fig 1A and B). These data provide first evidence that PF-4 has the capacity to induce PMN adhesion to EC without the requirement for an exogenous costimulus and that the presence of EC enables the chemokine to stimulate PMN for secondary granule marker exocytosis. Moreover, the diverging efficacies of PF-4 and IL-8 for induction of cell adhesion, and especially the failure of IL-8 to stimulate granule exocytosis, suggested that neutrophil-EC interaction in response to either chemokine involved different mechanisms.

In further experiments, we addressed the question of whether the diverging efficacies of PF-4 and IL-8 for the induction of PMN adhesion resulted from the circumstance that adhesion in response to either chemokine exhibited different stability under the experimental shear force applied to remove nonadherent cells (ie, the force resulting from centrifugation at 150g at an angle of 45°C). Thus, 2 parallel sets of PMN samples, stimulated with a saturating concentration of either PF-4 (10 μmol/L) or IL-8 (100 nmol/L), were allowed to adhere to EC for 20 minutes, and the percentage of cells remaining attached to the EC monolayer was determined after exposure to different shear forces, as imposed by centrifugation at discrete gravities ranging from 50g up to 400g. As shown in Fig 2, there was no major difference between IL-8– and PF-4–mediated cell adhesion (66% ± 8% and 79% ± 6%, respectively) at a low shear force (ie, at 50g). Under these conditions, background adhesion as represented by unstimulated control cells was relatively high (32% ± 8%), thus obscuring the chemokine-mediated effects. However, increasing the shear force by applying a series of higher centrifugation rates led to a dramatic step-wise decrease in the proportion of cells remaining adherent in response to IL-8 (down to 15% ± 5% at 400g), whereas PF-4–mediated adhesion was not remarkably altered over the entire range (being still 67% ± 11% at 400g), and background levels were always around 8% ± 0.2% at 200g to 400g. These data clearly show that PF-4–induced PMN adhesion to EC is much more stable to shear forces than that elicited by IL-8, which suggests that different cellular adhesion molecules and/or their different functional regulation may be involved in response to either chemokine.

To address the question of whether PF-4 and IL-8–mediated PMN adhesion to EC might involve different adhesion molecules, corresponding assays were performed in the presence of blocking MoAbs directed against various adhesion molecules known to be either selectively expressed on PMN (CD11a, CD11b, CD11c, CD18, and CD62L) and on the endothelium (CD54, CD103, and CD106) or to be coexpressed on both cell types (CD49d, CD31, and CD15). As shown in Fig 3(upper panel), PF-4–dependent neutrophil adhesion was substantially inhibited (by more than 76%) after the addition of either anti-CD11a, anti-CD18, or anti-CD62L, which indicated that at least 2 different PMN-expressed adhesion receptors, the β₂-integrin LFA-1 and L-selectin, were involved in this process. Furthermore, blocking CD54 (intercellular adhesion molecule-1 [ICAM-1]), which is EC-expressed and represents one of the natural ligands for LFA-1, caused inhibition of PMN adhesion to various degrees, depending on the PF-4 concentration used (78% and 46% of inhibition at 2 and 10 μmol/L PF-4, respectively). Because incomplete, ie, only about half-maximal, inhibition obtained at the higher dosage of PF-4 may be due to the circumstance that LFA-1 may interact with ICAM-2 in addition to ICAM-119 effects of anti–ICAM-2 antibodies alone or in combination with anti–ICAM-1 antibodies on PF-4–induced adhesion were investigated. Similar to ICAM-1, anti–ICAM-2 antibodies alone caused only a partial inhibition (72.8% ± 4.3% of inhibition). However, a combination of both antibodies resulted in a total inhibition of PF-4–induced adhesion (90.6% ± 8.2% of inhibition, data not shown), indicating that both counterligands for LFA-1 are involved in this process. None of the other antibodies used, including the isotype controls, showed an effect on PF-4–dependent cell adhesion greater than 5% of inhibition. Parallel control experiments performed with IL-8–stimulated cells showed a substantially different inhibition pattern (Fig 3, lower panel). In agreement with published data,20 21 IL-8–dependent adhesion was sensitive to antibodies directed against MAC-1 (CD11b/CD18), but was completely unaffected in the presence of anti-CD11a or anti-CD62L antibodies. Different from PF-4, adhesion induced by IL-8 could be entirely inhibited by antibodies directed against ICAM-1 (85.7% ± 14.0% of inhibition), whereas anti–ICAM-2 antibodies were without effect in this setting (data not shown). Thus, these data demonstrate that PF-4– and IL-8–driven cell adhesion involves different PMN-associated molecules. Whereas at least 2 adhesion receptors (LFA-1 and L-selectin) become engaged through stimulation by PF-4, the MAC-1 antigen appears to be the only receptor becoming involved in response to IL-8.

Consequently, our next approach was to examine whether PF-4–induced PMN-EC interaction is accompanied by quantitative upregulation of the adhesion molecules involved. These experiments were performed by first allowing PMN to attach to EC in the presence or absence of PF-4 and IL-8, to detach adherent cells by incubation on ice, and then to determine the levels of surface marker expression in the total PMN population after immunofluorescence staining for CD11a, CD11b, CD18, and CD62L. The data obtained by flow cytometric analyses are given in Table 1. In accordance with published data,20 22 IL-8 (at 20 nmol/L) provoked a shift in median fluorescence intensity (MFI) in cells stained with anti-CD11b (from 109 to 209 MFI) as well as an increase in anti-CD18–labeled PMN (from 118 to 198 MFI), as compared with unstimulated control cells, whereas no change in MFI was observed in anti-CD11a–labeled PMN. By contrast, PF-4 (at 5 μmol/L) did not affect the expression of any of these surface markers. Interestingly, the level of surface-expressed CD62L was downregulated by more than 90% after exposure to IL-8, whereas only minor changes (∼17% reduction in MFI) were observed after PF-4 stimulation. These results demonstrate that PF-4, at variance with IL-8, neither upregulates integrin expression nor induces shedding of L-selectin in PMN. Thus, it would appear that PF-4–mediated cell adhesion is rather brought about by affinity modulation of 1 or both of the adhesion molecules involved.

With respect to our previous observation that PF-4 in the absence of an exogenous costimulus is neither able to induce exocytosis in suspended PMN nor to increase their adherence to plasma- or gelatin-coated surfaces,12 we wondered by which means such a costimulatory signal(s) could be provided by EC. First, we sought to get an idea on whether the contribution by EC just consisted in providing a passive matrix (eg, by the constitutive expression of surface molecules interactive with PMN) or whether metabolically active EC were required, retaining the capacity to respond to PF-4–stimulated PMN or PF-4 itself (eg, by the upregulation of membrane-expressed costimulatory molecules or by the release of soluble costimulatory factors). In a first set of experiments, we examined whether PF-4 acted exclusively on PMN to promote adhesion and exocytosis or whether stimulation of EC was also required. Whereas PMN pre-exposed to PF-4 (0.4 to 10 μmol/L) and subsequently washed to remove soluble chemokine still underwent adhesion as well as exocytosis, correspondingly treated EC did not promote any of these functions (data not shown), indicating that PF-4 by itself did not elicit a costimulatory signal in EC. To estimate whether functional activation of EC was required at all, these cells were fixed with paraformaldehyde to render them metabolically inactive. Subsequently, their capacity to promote PMN adhesion as well as granule exocytosis in the presence of PF-4 was assayed as initially described. For comparison, assays using fixed PMN in combination with viable EC and controls using exclusively viable cells were run in parallel. As shown in Fig 4A, the ability of viable PMN to undergo adhesion in response to increasing concentrations of PF-4 was not impaired in the presence of fixed EC, whereas fixation of PMN led to complete abrogation of their ability to adhere to viable EC. Thus, PF-4–dependent PMN adhesion does apparently not require activation of EC, but appears to be controlled by the regulation of PMN-associated mechanisms. However, as evident from the data presented in Fig 4B, this may be somehow different with PF-4–dependent granule marker exocytosis. Here, lactoferrin release by viable PMN cocultured with fixed EC is markedly reduced as compared with PMN stimulated in the presence of viable EC. Not only was an approximately 4-fold higher threshold of PF-4 (0.4 v 0.1 μmol/L) required for the induction of a measurable response, but also in the presence of fixed EC an apparent saturation of the response was already reached at a level amounting to less than 50% of that induced by the same dosages of PF-4 (2 to 10 μmol/L) in the presence of viable EC. As expected, fixed PMN were not able to respond to PF-4. These results demonstrate that, for the induction of a full exocytsosis response by PF-4, at least 1 of the costimulatory signals contributed by EC can only be provided by viable cells, whereas PF-4–mediated adhesion appears to be functional in response to constitutive signal(s) provided by EC serving as a passive matrix.

As demonstrated by the results given above, the signals contributed by EC for the induction of adhesion and exocytosis in PMN may be different. However, these signals appear to be secondary to those that become elicited by PF-4 in PMN, because prestimulation of PMN with PF-4 is sufficient to promote either of these functions, whereas prestimulation of EC is uneffective (see above). Thus, we examined whether the same PMN surface molecules (CD62L and CD11a/CD18) that we found to be involved in PF-4–mediated adhesion to EC would also contribute signal(s) for the induction of exocytosis. For this purpose, adhesion assays were performed in the presence of antibodies directed against CD11a, CD18, and L-selectin (as described above), and the proportion of adherent PMN as well as lactoferrin contents in the cell-free supernatants were determined. As shown in Fig 5 and consistent with the results depicted in Fig 3, PF-4–induced neutrophil adhesion was inhibited by all 3 antibodies. Similarly, in the presence of anti-CD11a and anti-CD18, the levels of lactoferrin in the supernatants was substantially reduced (by more than 65% and 83% at 10 μmol/L PF-4 with anti-CD11a and anti-CD18, respectively). However, no inhibition of exocytosis was obtained in the presence of anti-CD62L. From these data we conclude that PF-4–induced activation of surface-expressed integrin LFA-1 and its subsequent interaction with adhesion molecules on the endothelium may elicit costimulatory signals required for PF-4–induced exocytosis. By contrast, L-selectin appears to have a selective contribution in the PF-4–mediated adhesion processes.

Although neutrophil activation in response to CXC-chemokines has been intensively investigated during the last years, a role for platelet factor 4 as a neutrophil-directed mediator is only beginning to emerge. Early reports suggesting an activity profile for PF-4 essentially identical to that of IL-823-26 could not be confirmed in more recent studies that failed to detect any PMN-stimulating properties of the chemokine.6,27-29 This was most likely due to the availability of more sophisticated purification methods, leading to the exclusion of contaminating mediators such as NAP-2 from natural PF-4 preparations.12 Although it now appears clear that PF-4 by itself is not active on suspended neutrophils, we recently discovered that an appropriate costimulus, such as TNF-α, may confer at least a restricted spectrum of functional activities on PF-4.12 This observation implicated a role for PF-4 as a mediator participating in prolonged inflammatory processes, eg, in chronic diseases involving platelet activation. However, because massive secretion of PF-4 also occurs after platelet activation in response to exogenous tissue lesion, we wondered whether there would exist costimulatory principles other than those conferred by proinflammtory cytokines. In our present report, we demonstrate that unstimulated vascular EC may provide signals facilitating short-term PF-4–mediated neutrophil activation in terms of adhesion to an EC layer and subsequent exocytosis of PMN secondary granule contents.

Adhesion to endothelium is a general phenomenon induced in marginating PMN by CXC-chemokines such as IL-8, and, as we could confirm for the latter chemokine, this occurs via upregulation of CD11b/CD18-integrins (MAC-1), whereas CD26L (L-selectin) becomes shedded from the cell surface and is thus not involved in establishing firm adhesion.20,21,29 By contrast, blocking experiments using antibodies directed to PMN-expressed adhesion molecules showed that PF-4–mediated adhesion follows different principles, involving CD11a/CD18-integrins (LFA-1) as well as L-selectin but not MAC-1. Likewise, a coadhesive effect by α₄β₁-integrin (CD49d), as previously described for IL-8–stimulated PMN,30 was not seen under these conditions. Although, as determined by fluorescence-activated cell sorting (FACS) analyses, PF-4 did not induce quantitative upregulation of integrins, adhesion was much more resistant to shear force than that induced by IL-8, with the number of cells remaining attached to the endothelium being unchanged after centrifugation even at 400g, whereas under identical conditions adhesion of IL-8–stimulated PMN was almost down to background levels. Although it appears likely that affinity modulation of the corresponding surface molecules may also contribute to the enhanced stability of PF-4–mediated adhesion, our results indicate that the pivotal mechanism responsible for this phenomenon consists in the simultaneous engagement of 2 different adhesion receptors. This may be derived from our observation that blocking antibodies to either CD11a or CD62L completely prevented cell adhesion, suggesting that a joint contribution by both molecules is required for bringing about sufficient avidity in cell-cell interaction. Most likely, this interrelationship is supported by affinity-upregulation in LFA-1, because such changes are well-known to occur with integrins,31 whereas no corresponding observations have so far been reported for L-selectin. However, as reported by Gopalan et al,32 L-selectin may not only act as an adhesion receptor but, after its cross-linking with antibodies, may also function as a signalling molecule, as seen by its ability to activate CD11a/CD18-dependent PMN adhesion to ICAM-1 in cells costimulated with a threshold concentration of IL-8. Such a role for L-selection would provide one explanation for our findings that blocking antibodies to CD26L prevented PF-4–mediated PMN adhesion.

Nevertheless, the involvement of LFA-1 instead of MAC-1 in PF-4–mediated PMN adhesion remains an unusual feature, inasmuch as various other PMN-directed stimuli, such as IL-8, NAP-2, FMLP, and TNF-α, have been shown to activate MAC-1, and a selective role for LFA-1 in cell adhesion to endothelium has so far only been demonstrated in lymphocytes.33 As it appears, activation of LFA-1 in PF-4–stimulated PMN is also central to the induction of secondary granule marker exocytosis, because in our experiments, blocking antibodies to CD18 and CD11a, but not to other CD11 chains prevented PF-4–induced lactoferrin release. In contrast, IL-8–stimulated PMN, upregulating MAC-1, did not undergo exocytosis. Interestingly, an anti-CD62L antibody that blocked PF-4–mediated adhesion did not prevent exocytosis, which indicates that L-selectin is not involved in mediating the latter function and, moreover, that firm adhesion of PMN to EC is not a prerequisite for the induction of exocytosis. Thus, simply contact between these cells, facilitating interaction of LFA-1 with its counterreceptors on EC, is likely to be sufficient to provide a costimulatory signal.

Although PF-4 specifically binds to both PMN and EC,13,34 35 our results indicate that the primary signal leading to adhesion as well as exocytosis is elicited in PMN. This was evident from our observation that PMN preincubated with PF-4 still underwent either of these functions, whereas preincubation of EC was without effect. The role of EC in this setting appears to be complex. Obviously, metabolically active EC are not required to promote adhesion of PF-4–stimulated PMN, because paraformaldehyde-fixed EC bound as many PMN as did nonfixed, viable cells. These results suggest that constitutive expression of counterreceptors for LFA-1 and L-selectin (such as ICAMs, selectins, and sialylated Lewis X oligosaccharides, respectively) is sufficient for the establishment of firm adhesion. Although in this respect EC may thus be regarded to function as a passive matrix, their role in providing a costimulus for the induction of exocytosis is less clear. Although PF-4–stimulated PMN exposed to paraformaldehyde-fixed EC were still able to respond by exocytosis, there was a substantial reduction in the amount of lactoferrin liberated. This loss in capacity could not be overcome by increasing the dosage of PF-4. The latter observation suggests that at least 2 different kinds of mechanisms are involved, one that is independent of metabolically active EC and is most likely elicited via interaction of LFA-1 with constitutively expressed counterreceptors and a second one that depends on additional costimulatory signal(s) that can only be provided by viable EC. Whether this kind of signal consists in the upregulation of further endothelial counterreceptors or in the liberation of soluble EC-derived factors remains at present unknown. Regarding this model, we are aware of the possibility that the reduced costimulatory potential of fixed EC also could be caused by the destruction of a fraction of constitutively expressed counterreceptors by paraformaldehyde. This assumption would make it unnecessary to suggest a contribution by viable EC. Nevertheless, it would demonstrate that a population of EC-expressed surface molecules in addition to that involved in adhesion is required to stimulate PMN for unimpaired exocytosis.

Taken together, these results further support our view that PF-4 has a special role in the recruitment of neutrophils to the vascular wall and in the initiation of an inflammatory response. Its ability to promote PMN-adhesion to unstimulated endothelium as well as secondary granule exocytosis without the requirement for a proinflammatory costimulus such as TNF-α indicates a relevant contribution by this chemokine at the very onset of an inflammatory situation. As compared with the functional spectrum of a classic CXC-chemokine such as IL-8, that of PF-4 appears to be more focused on establishing especially firm adhesive PMN/EC interaction that is resistant to high shear stress. Such conditions are likely to occur in larger vessels, in which blood flow remains largely unimpaired after thrombus formation in the microvasculature. After adhesion, the EC-dependent exocytosis of secondary granule contents may represent a further adaptation to the conditions encountered in larger blood vessels. Apart from lactoferrin and adhesion molecules, secondary granules contain releasable matrix-degrading enzymes, including collagenase and gelatinase,36 which could support PMN to digest their way through the more massive vessel walls found within this tissue. The fact that PF-4–mediated exocytosis requires prior contact between PMN and the endothelium to become established points to tight regulation of this process. By these means the release of granule contents will become strictly localized to sites of cell adhesion, and the surrounding tissue will be protected from enzymatic destruction. According to previous studies performed in our laboratory, PF-4 is neither chemotactic for PMN nor does it elicit random migration. It may thus be envisaged that a special role of this chemokine will consist in immobilizing PMN at sites of thrombus formation and to render these cells capable of migrating in response to chemotactically active CXC-chemokines such as IL-8 or the platelet-derived NAP-2. Because the generation of NAP-2 requires proteolytic cleavage of its platelet-secreted precursors by PMN,7 cells immobilized by PF-4 in the vicinity of a thrombus may contribute to enhanced NAP-2 generation and, as a further consequence, to NAP-2–mediated recruitment and activation of further PMN at sites of tissue lesion and inflammation. According to this model, PF-4 could act as a key mediator in the initiation of a cascade of inflammatory processes leading to host defense and tissue repair, but could also be responsible for adverse effects in pathological situations.

The authors thank Drs B. Katzmann and H. Klüter (Institute of Immunology and Transfusion Medicine, Medical University of Lübeck, Lübeck, Germany) for the generous supply of platelet concentrates. We are indebted to Dr A. Petersen for performing sequence analyses of the PF-4 preparations and to Dr H. Moll (District Hospital Bad Segeberg, Bad Segeberg, Germany) for supplying us with umbilical cords. We thank G. Kornrumpf for her expert work of cell culture and especially acknowledge C. Pongratz and I. von Cube for perfect technical assistance.