Differential Regulation of β₁ Integrins by Chemoattractants Regulates Neutrophil Migration through Fibrin

Rhodamine-conjugated polyethylene glycols of 3.5 kD (Rh-PEG 3.5 kD) and 10 kD (Rh-PEG 10 kD) were prepared as described (Loike et al., 1993, 1995). Sources of antibodies and peptides were as follows: mouse anti-β₁ (P4C10) and the peptides GRGDSP and GRGESP were from GIBCO BRL. Mouse anti–human α_x (LeuM5) was from Organon-Teknika Inc. Mouse anti–human α_Mβ₂ (MAC-1) was from Upstate Biotechnology Co. Mouse anti–human α₅ (SAM1), rat anti–human α₆ (GoH3), mouse anti–human α₄ (HP2/1), and mouse anti–human β₃ integrin (SZ21) were from Immunotech. Phycoerythrin-conjugated F(ab)₂ anti–mouse IgG was from Jackson ImmunoResearch. Mouse anti-β₃ (PPM6/13) was from Biosource International. Mouse anti-β₃ (MAB1957z) was from Chemicon International. Alexa 488–conjugated F(ab)₂ anti– mouse IgG was from Molecular Probes. LTB4, fMLP, PMA, thrombin, and Ficoll-Hypaque were from Sigma Chemical Co. Mouse anti–chicken β₁ integrin (CSAT) and mouse monoclonal anti–human β₁ integrin (AiiB2) were generous gifts from Dr. Clayton Buck (University of California, San Francisco, CA). Mouse mAb 15/7, which recognizes an activation epitope on human β₁ integrins (Bohnsack et al., 1995), was from Athena Neurosciences. Mouse mAb IB4, which blocks the ligand-binding domains of human β₂ integrins (Wright et al., 1983), was a generous gift from Dr. Samuel D. Wright (Merck, Rahway, NJ). PPACK was from Calbiochem-Novabiochem, Matrigel from Becton Dickinson, and collagen I from GIBCO BRL. Purified fibronectin was from Vitex International. Fibrinogen was from American Diagnostica Inc. Fibrinogen uncontaminated by Factor XIII, fibronectin, and vitronectin, a generous gift of Dr. Jeffrey Weitz (MacMaster University, Hamilton, Ontario, Canada), was prepared from fibrinogen obtained from Enzyme Research Labs FIBI. It was first adsorbed with gelatin-agarose to remove fibronectin and then passed over an affinity column to remove Factor XIII. The fibrinogen was precipitated with 25% ammonium sulfate, dialyzed against 150 mM NaCl, 20 mM Tris (pH 7.4), adsorbed with an antibody to human vitronectin linked to Affi-gel, and dialyzed. PAGE analysis showed the resulting fibrinogen to be free of fibronectin or Factor XIII. Western blot analysis revealed no vitronectin (data not shown).

Gels, ∼1 mm thick, composed of fibrin, Matrigel, or collagen type IV, were formed in cell culture inserts (pore sizes 3 or 8 μm) from Becton Dickinson as described (Loike et al., 1995). Fibrin gels were gently washed with PBS to remove any residual PPACK.

Fibrin/fibrinogen-coated surfaces were prepared as described (Wright et al., 1988; Loike et al., 1992, 1993, 1995) and PMN adhesion was measured by phase-contrast microscopy. Close apposition of PMN to fibrin/fibrinogen-coated surfaces was defined as exclusion of Rh-PEG 10 kD from zones of contact between PMN and fibrin/fibrinogen measured by fluorescence microscopy as described (Loike et al., 1993).

PMN were prepared as described (Wright et al., 1988) from fresh heparinized blood from healthy adult donors after informed consent. PMN used in these experiments were >95% pure as determined by Wright-Giemsa staining (Wright et al., 1988). 10⁶ PMN in 100 μl of PBS supplemented with 5.5 mM glucose and 0.1% human serum albumin (PBSG-HSA) were placed in the upper compartment of each insert and incubated for 0–6 h at 37°C in a humidified atmosphere containing 95% air/5% CO₂. At the times and concentrations specified, chemoattractants, antibodies, and/or peptides were added to the top and/or bottom compartments in 500 μl of PBSG-HSA. At the end of incubations, chambers were shaken to dislodge PMN from the lower surface of the inserts. The medium in each lower compartment was collected and its content of PMN was determined using a Coulter counter (Loike et al., 1995). Unless otherwise indicated, all values reported are the average of six different samples from at least three independent experiments.

PMN (10⁵ cells/200 μl of PBSG-HSA) were incubated in suspension at 37°C for 30 min in the presence or absence of fMLP (10⁻⁷ M) or LTB4 (10⁻⁷ M), transferred to 96-well polystyrene tissue culture microtiter plates (Corning), incubated for 30 min at 4°C in 200 μl PBSG-HSA containing the indicated primary antibody (2 μg/ml), washed three times with PBSG-HSA at 4°C, further incubated for 30 min at 4°C with either Alexa 488–conjugated or phycoerythrin-conjugated rabbit anti–mouse F(ab′)₂ in 200 μl of PBSG-HSA, washed three times again with PBSG-HSA at 4°C, and resuspended at 4°C in 300 μl PBS containing 2% BSA and 0.3 mg/ml propidium iodide to determine cell viability. The contribution of dead cells (usually <2%) was removed from the final data analysis. The mean fluorescence intensity of 3–5 × 10³ cells was determined using a Becton Dickinson FACSCalibur^®.

PMN chemotax through three-dimensional gels composed of reconstituted basement membrane proteins containing collagen IV, laminin, and fibronectin (Matrigel; Fig. 1), or collagen I (Loike et al., 1995) in response to a gradient of fMLP or LTB4. In contrast, PMN chemotaxis through fibrin gels or plasma clots is dependent upon the specific chemoattractant used. fMLP-stimulated PMN do not migrate through fibrin gels or plasma clots, whereas LTB4-stimulated PMN do (Fig. 2 A; Loike et al., 1995). Checkerboard analyses confirmed that PMN migrate through these gels in response to a chemoattractant gradient (Loike et al., 1995). Placement of equimolar concentrations of both fMLP and LTB4 into the bottom chambers inhibited PMN from migrating through fibrin gels (Fig. 2 A; Loike et al., 1995), confirming that fMLP's effect is dominant over LTB4's effect.

Commercial fibrinogen contains small amounts of fibronectin and vitronectin. To test whether matrix components other than fibrin are responsible for inhibiting migration of fMLP-stimulated PMN through fibrin gels and plasma clots, we performed additional experiments using fibrin gels formed from purified fibrinogen that contained no detectable fibronectin, plasminogen, Factor XIII, or vitronectin. PMN stimulated with LTB4, but not with fMLP, migrated through gels formed from purified fibrinogen (Fig. 2 B). Moreover, collagen I gels (60 μg/ insert) each containing 10 μg of purified fibronectin did not affect the migration of either fMLP- or LTB4-stimulated PMN, whereas the addition of fibrinogen to such gels blocked migration of fMLP-stimulated PMN (data not shown). These results are consistent with reports (Asakura et al., 1997; Farrell and al-Mondhiry, 1997; Suehiro et al., 1997; Miettinen et al., 1998) that fibrin(ogen) contains sequences that are ligands for β₁ integrins, and confirm that fibrin is the matrix component that inhibits migration of fMLP-stimulated PMN.

To examine the roles of β₁ and β₂ integrins in PMN migration through Matrigel (Fig. 1), or fibrin (Figs. 2 A and 3), we added antibodies that block β₁ or β₂ integrins to the upper compartment of Matrigel or fibrin-coated inserts together with PMN and measured the number of PMN that migrated into the lower compartment in response to fMLP or LTB4. As expected, mAb IB4, directed against β₂ integrins (Wright et al., 1983), blocked PMN migration through Matrigel (Fig. 1) or fibrin gels (Fig. 2 A) in response to LTB4. Antibody IB4 also blocked fMLP-stimulated PMN migration through Matrigel (Fig. 1), and did not alter fMLP's inhibitory effect on PMN chemotaxis through fibrin gels (data not shown). These results are consistent with previous reports (Diamond and Springer, 1994; Springer, 1995; Premack and Schall, 1996) that anti– β₂ integrin antibodies block PMN migration through endothelia and through gels formed by a variety of extracellular matrix proteins.

In contrast, mAbs AiiB2 (Bohnsack et al., 1990) and P4C10 (Carter et al., 1990), which block the common β chain of β₁ integrins (CD29), had no effect on fMLP- or LTB4-stimulated chemotaxis through Matrigel (Fig. 1), or on LTB4-stimulated PMN migration through fibrin gels (Fig. 2 A). However, these same anti–β₁ chain antibodies reversed fMLP's inhibitory effect on PMN chemotaxis through fibrin (Fig. 3 A).

Control experiments showed that CSAT (Lallier and Bronner-Fraser, 1991), a mAb that binds to chicken but not human β₁ integrins, SZ21, an antibody against β₃ integrins (Lawson and Maxfield, 1995), and PM6/13, another antibody against β₃ integrins (Patel et al., 1998), did not alter the inhibitory effect of fMLP on PMN migration through fibrin gels (Fig. 3 A and data not shown). These antibodies also did not affect migration of LTB4-stimulated PMN through fibrin gels (data not shown).

Among the antibodies directed against the α chains of β₁ integrins, only those directed against α₅ chains were effective in reversing fMLP's inhibitory effect on PMN migration through fibrin gels (Figs. 2 B and 3 A). Neither antibodies against α₄ chains nor antibodies against α₆ chains of β₁ integrins affected migration of fMLP- or LTB4-stimulated PMN through fibrin gels (Fig. 3 A) or Matrigel (Fig. 3 B and data not shown).

To confirm that β₁ integrins directly interact with fibrin(ogen), we examined the effects of anti–β₁ integrins on the migration of fMLP-stimulated PMN through gels formed of purified fibrinogen, lacking detectable levels of fibronectin, vitronectin, plasminogen, or Factor XIII. Both antibodies directed against β₁ and α₅ chains of β₁ integrins (Fig. 2 B) reversed fMLP's inhibitory effect on chemotaxis through these gels. These results are consistent with reports (Asakura et al., 1997; Farrell and al-Mondhiry, 1997; Suehiro et al., 1997; Miettinen et al., 1998) that fibrin(ogen) contains sequences that are ligands for β₁ integrins.

The peptide GRGDSP blocks the interaction of β₁ integrins with RGD ligands on matrix proteins (Pierschbacher and Ruoslahti, 1987). Like antibodies against β₁ integrins, addition of GRGDSP peptide to the medium allowed fMLP-stimulated PMN to migrate through fibrin gels (Table I). Control experiments showed that GRGESP peptide, which does not block binding of β₁ integrins to fibronectin or other RGD-containing matrix proteins (Pierschbacher and Ruoslahti, 1987), did not reverse the inhibitory effect of fMLP on PMN migration through fibrin gels (Table I). Neither peptide affected the number of PMN that migrated through fibrin in response to LTB4 (Table I) or through Matrigel in response to fMLP (Fig. 3 B).

PMN were incubated in control medium or in medium containing fMLP or LTB4 and allowed to adhere to fibrin-coated 96-well plates. In the absence of chemoattractant <1% PMN adhered to fibrin (data not shown). Over 40% of fMLP-stimulated PMN and ∼50% of LTB4-stimulated PMN adhered to fibrin. mAb IB4, which blocks the ligand-binding sites of three different β₂ integrins (α_Lβ₂, α_Mβ₂, and α_Xβ₂; Wright et al., 1983; Loike et al., 1991), inhibited adhesion of fMLP- or LTB4-stimulated PMN to fibrin by 75–80% (Fig. 4). In contrast, mAb AiiB2, which blocks the ligand-binding sites of β₁ integrins (Bohnsack et al., 1990), had no significant effect on the number of fMLP-stimulated PMN that adhered to fibrin, and enhanced by ∼25% adhesion of LTB4-stimulated PMN to fibrin (Fig. 4). These experiments show that β₂ integrins are the primary PMN surface receptors that mediate adhesion of chemoattractant-stimulated PMN to fibrin.

The findings presented above indicate that β₁ integrins, and specifically α₅β₁ integrins, mediate the qualitatively distinct effects of fMLP and LTB4 on PMN adhesion to, and migration through, fibrin gels. To determine whether fMLP and LTB4 differentially affect the activation of β₁ integrins we used mAb 15/7, which recognizes a conformationally determined epitope on activated β₁ integrins (Bohnsack et al., 1995). PMN incubated for 30 min with fMLP exhibited a 10–22-fold increase in binding of mAb 15/7 (Fig. 5 J), compared with unstimulated PMN (Fig. 5 B), whereas PMN incubated for the same length of time with LTB4 (Fig. 5 F) showed little change over unstimulated PMN (Fig. 5 B) with respect to binding of mAb 15/7. Control experiments showed that surface expression of β₁ integrins was stimulated approximately twofold by LTB4 (Fig. 5 G), and approximately threefold by fMLP (Fig. 5 K), whereas β₂ integrin surface expression was stimulated approximately fivefold by LTB4 (Fig. 5 H) and approximately ninefold by fMLP (Fig. 5 L). Other studies showed that the extent of expression of the epitope for antibody 15/7 on β₁ integrins was dependent upon the dose of fMLP used to stimulate the PMN, and that 5 × 10⁻⁶ M fMLP induced maximal expression of this epitope (not shown). In contrast, LTB4 concentrations 10–50-fold higher (i.e., 10⁻⁶ to 5 × 10⁻⁶ M) than those used in the experiments described in Fig. 5 did not increase expression of the 15/7 epitope on β₁ integrins (data not shown).

We have used exclusion of Rh-PEG 10 kD from zones of contact between chemoattractant-stimulated PMN and fibrin-coated surfaces as a measure of the closeness of apposition of PMN to the underlying substrate (Loike et al., 1995). Previously, we reported an inverse correlation between the formation of zones of close apposition between chemoattractant-stimulated PMN and fibrin gels and the capacity of PMN to migrate through these gels (see Fig. 7 in Loike et al., 1995). In the present experiments we used exclusion of Rh-PEG 10 kD to test whether antibodies and peptides that block β₁ integrins, and that facilitate migration of fMLP-stimulated PMN through fibrin gels (Figs. 2 and 3 A and Table I), affect the closeness of apposition of these cells to fibrin. Antibodies against the β chain of β₁ integrins, or against the α₅ chain of α₅β₁ integrins (not shown), reduced the percentage of fMLP-stimulated PMN that excluded Rh-PEG 10 kD from zones of contact with fibrin from 80% to 20–30% (Fig. 6), and reduced the percentage of LTB4-stimulated PMN that excluded Rh-PEG 10 kD from these contact zones from 20% to <2% (Fig. 6). These experiments together with those shown in Fig. 5 show there is a direct correlation between the capacity of fMLP or LTB4 to activate PMN β₁ integrins and the capacity of these chemoattractants to promote close apposition between PMN and fibrin (as measured by exclusion of Rh-PEG 10 kD), and to inhibit PMN migration through fibrin gels (Fig. 3 A).

Tumor-promoting phorbol esters, like ligands that bind to PMN and macrophage fibronectin receptors, activate α_Mβ₂ (CD11b/CD18) for phagocytosis of C3bi-coated particles (Wright and Silverstein, 1982), and promote formation of zones of close apposition between phorbol ester–stimulated PMN and fibrinogen-coated surfaces (Table II). However, phorbol ester–stimulated PMN do not migrate into fibrin gels, even when treated with antibodies against α₅β₁ integrins (data not shown). These findings suggested that phorbol esters activate α_Mβ₂ integrins for close apposition to fibrin independently of β₁ integrins. To test this prediction, PMN were incubated with or without antibodies against β₁ integrins, allowed to adhere to fibrin- or fibrinogen-coated surfaces in medium containing PMA, and then were incubated with Rh-PEG 10 kD. 77% of PMA-treated PMN formed zones of close apposition on fibrin even when they had been treated with antibodies against β₁ integrins. In contrast, <15% of the PMA-stimulated PMN that adhered to these surfaces formed zones of close apposition when treated with antibodies against β₂ integrins (Table II). This experiment shows that when suitably activated, β₂ integrins are capable of mediating close apposition between PMN and fibrin-coated surfaces in the absence of β₁ integrin ligation.

The different effects of fMLP and LTB4 on PMN adhesion to and chemotaxis through fibrin gels appear to be a consequence of qualitative differences in the effects of these chemoattractants on the activity of β₁ integrins. That is, fMLP activates β₁ integrins (Fig. 5 and Table III), stimulates PMN to adhere closely to fibrin(ogen) (Fig. 6 and Table III; Loike et al., 1995), and inhibits PMN chemotaxis through fibrin gels (Fig. 2 and Table III; Loike et al., 1995). In contrast, LTB4 neither activates β₁ integrins (Fig. 5 and Table III) nor induces PMN to adhere closely to fibrin(ogen) (Fig. 6 and Table III; Loike et al., 1995), and stimulates PMN to migrate through fibrin gels (Fig. 2 and Table III; Loike et al., 1995). To our knowledge, this is the first demonstration that signals initiated by two chemically distinct chemoattractants with their respective seven membrane spanning/heterotrimeric G protein–coupled receptors exert different effects on the activation state of a specific β₁ integrin, and regulate PMN migration.

As shown in Fig. 2, fibrin(ogen) is unique among the matrix and plasma proteins tested in arresting the migration of fMLP-stimulated PMN. This is particularly notable in the case of fibronectin, a well-recognized ligand for α₅β₁ integrins. The failure of fibronectin to induce migration arrest suggests that fibrin(ogen) has heretofore unrecognized properties, independent of its ability to bind α₅β₁ integrins, that are important in its ability to cause migration arrest.

DiMilla et al. (1993) and Palecek et al. (1997) reported that smooth muscle cells migrate optimally on fibronectin-coated surfaces when their integrins bind to these surfaces at intermediate strengths. Weber et al. (1996) reported an inverse correlation between the strength of adhesion of chemokine-stimulated monocytes to surfaces coated with the 120-kD RGD-containing fibronectin fragment and the capacity of these cells to migrate across filters coated with this fibronectin fragment. The findings of Keller et al. (1979) and of Wilkinson et al. (1984), and those reported in Fig. 6, demonstrate an inverse correlation between closeness of apposition of PMN to surfaces coated with proteins that express ligands for PMN receptors and the ability of PMN to migrate on or through matrices containing these proteins. Thus, it seems likely that loose versus close apposition between cells and matrix protein–coated substrates reflects weak versus strong adhesion, respectively, between the cells and the substrate.

Antibodies against β₂ integrins reduced adhesion, inhibited close apposition between PMA-stimulated PMN and fibrin (Table II), and blocked PMN migration through fibrin (data not shown). Antibodies against β₁ integrins had no effect on any of these parameters (Table II and data not shown). These results demonstrate that the interaction of activated β₂ integrins with fibrin is both required and sufficient for PMA-stimulated PMN to form zones of close apposition on fibrin (Table II), and that PMA bypasses the requirement for engagement of activated β₁ integrins by matrix proteins for PMN to form zones of close apposition on fibrin.

Although the signal transduction pathways by which chemoattractants regulate PMN β₁ and β₂ integrins remain to be elucidated, our findings lead us to make three suggestions regarding the organization of these pathways.

First, antibodies that activate β₁ integrins do not promote adhesion of unstimulated PMN to fibrin, or inhibit LTB4-stimulated chemotaxis of PMN through fibrin gels (unpublished data). These results suggest that signals initiated by both fMLP receptors and activated β₁ integrins are required to inhibit chemotaxis of fMLP-stimulated PMN through fibrin gels.

Second, the finding that fMLP and PMA have similar effects on PMN adhesion to and migration through fibrin gels might suggest that the interaction of activated β₁ integrins of fMLP-stimulated PMN with fibrin activates protein kinase C, and that this is the mechanism by which fMLP signals β₂ integrins to bind closely to fibrin. However, Laudanna et al. (1996) reported that calphostin C, a protein kinase C inhibitor, blocks adhesion of PMA-stimulated, but not of fMLP-stimulated, mouse lymphocytes transfected with fMLP receptors, to VCAM-1–coated surfaces. (Adhesion of chemokine-stimulated lymphocytes to VCAM-1 is mediated by activated α₄β₁ integrins.) Laudanna et al. (1996) identified rho as a key participant in fMLP- and IL-8–mediated activation of α₄β₁ integrins in mouse lymphocytes. This finding suggests to us that rho acts downstream of Gα_i in activating β₁ integrins. The report of Caron and Hall (1998) that rho participates in coupling CR3 (CD11b/CD18) to the actin cytoskeleton suggests that rho also affects β₂ integrin–mediated functions. Whether PMN LTB4 receptors activate rho is unknown and should be investigated.

Third, binding of fMLP to its receptor activates Gα_i (Laudanna et al., 1996). The specific Gα activated by LTB4 in PMN has not been reported. Pertussis toxin, which inactivates Gα_i, blocks most effects of LTB4 and of fMLP on human PMN. Thus, the finding that fMLP activates β₁ integrins (Fig. 5 J) while LTB4 does not (Fig. 5 F) suggests that binding of LTB4 to its receptor activates Gα subunits other than, or in addition to, Gα_i and that this difference in Gα subunit utilization is responsible for the divergent effects of fMLP and LTB4 on β₁ integrin activation (Fig. 5), and on closeness of PMN adhesion to fibrin (Fig. 6). Indeed, Arai and Charo (1996) have shown differential utilization of Gα subunits after MCP-1 or IL-8 stimulation of MCP-1 or IL-8 receptor transfected HEK293 cells, and Yokomizo et al. (1997) have demonstrated that pertussis toxin treatment does not ablate Ca²⁺ increases stimulated by LTB4 in LTB4 receptor-bearing CHO cells.

Our studies suggest at least three distinct mechanisms by which fMLP could inhibit PMN migration through fibrin gels. First, the combined strengths of adhesion of activated β₁ and β₂ integrins to fibrin could be sufficient to immobilize PMN on fibrin. Our unpublished finding that antibodies that activate β₁ integrins do not inhibit migration of LTB4-stimulated PMN through fibrin gels casts doubt on this combined-strength-of-adhesion hypothesis as an explanation for the inhibitory effect of fMLP on PMN chemotaxis through fibrin.

Second is the possibility that binding of fMLP or LTB4 to its cognate receptors directly and differentially activates β₂ integrins for strong or weak adhesion, respectively. According to this hypothesis, fMLP-activated β₁ integrins play no role in inhibiting chemotaxis of fMLP-stimulated PMN through fibrin. However, since activated β₁ integrins mediate outside-in signaling, RGD peptides and antibodies against β₁ integrins reverse fMLP's inhibitory effect on PMN migration through fibrin by stimulating β₁ integrins to signal trans-dominant negative (Diaz-Gonzalez et al., 1996) effects on β₂ integrins. Against this hypothesis are the findings that antibodies against the α₆ chains of β₁ integrins (Fig. 3), and antibodies that activate β₁ integrins (unpublished data), do not reverse fMLP's inhibitory effect on PMN chemotaxis through fibrin.

Third, and we think most likely, is that the capacity of fMLP to promote close adhesion to, and to block migration through, fibrin gels is mediated by a cascade of signals (diagrammed in Fig. 7), in which the interaction of activated β₁ integrins with the fibrin matrix causes trans-dominant activation of β₂ integrins. This mechanism is consistent with previous studies (Pommier et al., 1983; Wright et al., 1984; Brown, 1992) showing that interaction of PMN or macrophages with RGD-containing matrix proteins activates α_Mβ₂ integrins for phagocytosis of C3bi-coated particles.

As shown in Fig. 7, we suggest that the interaction of LTB4 or fMLP with their respective PMN receptors generates a “common” signal that activates β₂ integrins for loose adhesion to fibrin (Fig. 7, A and B, and D and E, respectively). In addition, we propose that fMLP receptors (Fig. 7 C) also signal activation of α₅β₁ integrins (Figs. 5 J and 7 E). We further suggest that binding of activated α₅β₁ integrins to fibrin matrices clusters these integrins, thereby generating an outside-in signal that activates β₂ integrins (Fig. 7 F), for close apposition between PMN and fibrin-coated substrates (Fig. 7 G).

Close apposition reflects tight adhesion (Keller et al., 1979; Wilkinson et al., 1984; DiMilla et al., 1993; Palecek et al., 1997), presumably mediated by the coupling of β₂ integrins to the cytoskeleton. We do not know whether tight adhesion causes, or is merely associated with, cessation of migration. In either case, PMN cease migrating (Figs. 2 A and 3 A, and Table III). We propose that antibodies and peptides that block the interaction of activated α₅β₁ integrins with fibrin (Figs. 2 and 3, and Table I) inhibit these outside-in signals, thereby blocking trans-dominant activation of β₂ integrins for close apposition to fibrin (Fig. 6 and Table II) and allowing PMN to migrate through fibrin.

The interaction of LTB4 with its receptor also generates a signal that stimulates β₂ integrins for loose apposition. However, LTB4 does not activate β₁ integrins (Fig. 5 F). Therefore, these β₁ integrins do not bind to the matrix, do not generate outside-in signals, and therefore do not initiate trans-dominant activation of β₂ integrins for close apposition (Fig. 7, A–C), or cessation of migration.

Further work is needed to determine whether cessation of migration is merely a function of strong adhesion between PMN and fibrin or whether it reflects reorganization of the PMN cytoskeleton as observed by Dustin et al. (1997) in antigen-sensitized T lymphocytes. They found that these cells become immobilized when they encounter MHC class II molecules containing a peptide antigen recognized by the T lymphocytes' antigen receptors. They identified changes in microtubule organization of these sessile T lymphocytes that distinguish them from their randomly migrating brethren. We suspect that PMN that adhere to fibrin after fMLP stimulation will exhibit similar changes in cytoskeletal organization.

Our findings suggest that the availability of many different chemoattractants (e.g., fMLP, LTB4, IL-8, C5a, etc.) serves two complementary functions. First, they provide redundancy, thereby assuring that pathogenic microbes are detected rapidly by the innate immune system. Second, they reflect the need to direct PMN to different tissue sites and to prepare them for interactions with many different types of ligands.

Our findings also suggest an alternative to the notion that leukocyte accumulation at a specific anatomic site in vivo requires the presence of a gradient of chemoattractant/ chemokine emanating from that site. While there is no doubt that gradients of chemoattractants/chemokines are formed in vitro (Keller et al., 1979; Wilkinson et al., 1984; Huber et al., 1991; Campbell et al., 1996, 1997; Foxman et al., 1997; Palecek et al., 1997), they may be difficult to maintain in vivo in the face of the perturbing effects of muscular contraction and variations in blood and lymph flow.

Leukocytes in the vascular system begin to enter specific tissue compartments when they encounter a chemoattractant/chemokine. We suggest that once within this tissue compartment leukocytes migrate randomly in response to a relatively uniform concentration of matrix-bound chemoattractant/chemokine. When in the course of this random walk they encounter extracellular matrix proteins or cells that express ligands for a specific activated β₁ integrin, they adhere strongly and become sessile. By regulating activation of specific receptors and adhesive strengths, concentrations of chemoattractants/chemokines well below those required to saturate or desensitize chemoattractant/chemokine receptors can mediate a stochastic process by which leukocytes accumulate at specific anatomic sites and form highly ordered structures (e.g., granulomas, germinal centers). According to this model, leukocytes accumulate at specific anatomic sites by a process that is similar in principle to the accumulation of flies on fly paper.

Foxman et al. (1997) showed that multiple chemoattractants/chemokines can work in combination to elicit migration patterns that cannot be achieved by a single chemoattractant/chemokine. The mechanisms we and they have described are complementary. These mechanisms are likely to be of special importance within tissue compartments where overlapping fields of chemoattractants/chemokines/cytokines surely occur, and where cells migrate in stepwise fashion from one anatomic site to another (e.g., T cell movement in lymph nodes from T cell–rich paracortical zones to germinal centers; Garside et al., 1998), PMN accumulation at foci of bacterial infection, or of immune-complex deposition (Wilkinson et al., 1984). The essential point of the findings reported here is that by endowing leukocytes, and probably all migrating cells, with a modest number of receptors for different chemoattractants, chemokines, and cytokines, nature has made optimal use of instructive and selective mechanisms to achieve a level of organizational specificity that would otherwise require substantially more genetic information.

We thank Drs. A.R. Horwitz and Sally Zigmond for helpful discussions, Eugene Butcher and Ellen Foxman for suggestions about the manuscript, Eric Brown for sharing unpublished data, and one of the reviewers for particularly thoughtful suggestions.

This study was supported by National Institutes of Health grant AI20516, Bristol-Myers Squibb Research, and a generous gift of the late Samuel W. Rover.

C3bi

cleaved b fragment of the 3rd complement component

C5a

a fragment of the cleaved 5th complement component

fMLP

formyl methionyl leucyl phenylalanine

HSA

human serum albumin

IL-8

interleukin 8

LTB4

leukotriene B4

PMN

polymorphonuclear leukocytes

Rh-PEG

rhodamine-conjugated polyethylene glycol