Responses of Neutrophils to Anti-Integrin Antibodies Depends on Costimulation through Low Affinity FcγRs: Full Activation Requires Both Integrin and Nonintegrin Signals¹

Neutrophils perform most of their in vivo functions while adherent to adjacent cells or extracellular matrix proteins. Integrins play a pivotal role in these processes by providing the molecular link between the neutrophils and their environment. The requirement for integrins is exemplified by the neutrophil migration defect and severe bacterial infections in leukocyte adhesion deficiency type I patients (1) who lack CD18, the common β₂ chain of LFA-1 and Mac-1. In vitro, neutrophils can be activated by cytokines (TNF, GM-CSF), chemokines (IL-8), or microbial products (fMLP, LPS) while adherent to surfaces coated with integrin ligands such as extracellular matrix proteins or rICAM-1 (2, 3). This activation requires β₂ integrin-mediated adhesion of the cells (4, 5).

In most biological systems, integrin-mediated adhesion provides an accessory signal to other, adhesion-independent stimuli. Examples include the β₁ integrin-dependent proliferation of fibroblasts upon growth factor stimulation (6), the requirement for the LFA-1 integrin for TCR-induced T cell cytotoxicity (7), or the α_Vβ₃ integrin-dependent fusion of osteoclasts upon stimulation by M-CSF and receptor activator of NF-κB (RANK) ligand (8, 9). Integrin-dependent activation of neutrophils is also mostly attained by coincidental activation of both integrins and other inflammatory receptors (see above). However, plate-bound Abs against the α- or β-chains of neutrophil integrins were found to be able to trigger neutrophil activation without any additional stimulus (10). The amplitude of this response was similar to that upon activation of extracellular matrix-adherent neutrophils by proinflammatory cytokines, suggesting that cross-linking of integrins is sufficient for full activation of neutrophils in the absence of any additional stimulus. This conclusion has been widely accepted and activation by immobilized anti-integrin Abs has been used by a number of groups (including ourselves) for the analysis of neutrophil integrin function (3, 5, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20). However, the possible contribution of other, nonintegrin signals (in particular, ligation of FcRs) to neutrophil activation through anti-integrin mAbs has not yet been explored.

In this study we provide evidence that, in addition to engagement of integrins, neutrophil activation by plate-bound anti-integrin Abs also requires the ligation of low affinity FcγRs (in particular, FcγRIII in case of murine and FcγRIIA in case of human cells). Our results suggest that ligation of integrins alone is not sufficient for full activation of neutrophils unless another inflammatory (e.g., cytokines) or immunological stimulus (e.g., FcR-cross-linking) initiates a second activation signal. The mechanism of neutrophil integrin signaling suggested by these results is in better agreement with the general working principles of the immune system.

Mice lacking the FcR γ-chain (FcRγ), the common signaling chain of multiple FcRs (Fcer1g ^{tm1Rav/tm1Rav}, referred to as FcRγ^−/−) were purchased from Taconic Farms (Germantown, NY). Mice lacking the low affinity FcγRIII (CD16) (Fcgr3 ^{tm1Sjv/tm1Sjv}, referred to as FcγR3^−/−) were purchased from The Jackson Laboratory (Bar Harbor, ME). Both strains were on a pure C57BL/6 genetic background. Wild-type C57BL/6 mice were purchased from Charles River Laboratories (Wilmington, MA) or from the Hungarian National Institute of Oncology (Budapest, Hungary). Mice were maintained under specific pathogen-free conditions at the University of California (San Francisco, CA) or in a conventional animal facility at the Loránd Eötvös University (Budapest, Hungary).

The mAbs used in this project were purified from hybridoma supernatant or purchased from commercial sources (Table I) (21). Hybridoma cells were propagated in Hybrimax medium (Sigma-Aldrich, St. Louis, MO) and the culture supernatant was passed over columns of protein A (Zymed Laboratories, South San Francisco, CA) or protein G (Invitrogen Life Technologies, Carlsbad, CA) beads. After intensive washes, mAbs were eluted by a glycine-HCl buffer (pH 2.7) and dialyzed against PBS. For initial experiments, the C71/16 Ab (and an isotype control, R35-95) was obtained from BD Pharmingen (San Diego, CA) instead of purified from hybridoma supernatant. Identical results were obtained with the Abs from the two different sources.

F(ab′)₂ of the anti-human CD18 (IB4) and isotype control (K9) mAbs were prepared using the ImmunoPure F(ab′)₂ preparation kit (Pierce, Rockford, IL) according to the manufacturer’s instructions. Fab of the IV.3 mAb was prepared by the same commercial kit using an extended digestion time to allow the removal of Cys residues containing the disulphide bonds between the two Fabs (22). In all cases, optimal conditions for mAb digestion were determined by testing the electrophoretic mobility of the proteolytic products on both reducing and nonreducing SDS-PAGE (F(ab′)₂ and Fab can be separated under nonreducing, but not under reducing, conditions).

Mouse bone marrow neutrophils were isolated by Percoll (Amersham Biosciences, Uppsala, Sweden) density gradient centrifugation and hypotonic lysis of RBC as previously described (23). Human neutrophils were isolated from venous blood of healthy volunteers as described (24). Cells were kept in Ca²⁺- and Mg²⁺-free medium until use and preincubated for 10 min at 37°C in the assay medium before activation. Unless otherwise stated, neutrophil assays were performed at 37°C in HBSS supplemented with 20 mM HEPES (pH 7.4).

A total of 10⁶ bone marrow neutrophils were incubated with 2 μg of unlabeled Abs on ice, washed, and incubated with FITC-labeled anti-rat Ab (clone 7/5 from Dr. G. László, Eötvös Loránd University, Budapest, Hungary). The labeled cells were analyzed on a BD Biosciences (Mountain View, CA) FACScan flow cytometer.

For most Ab-mediated cross-linking experiments, mAbs were immobilized by exploiting the highly efficient binding of the carbohydrate chains of Ab Fc portions to ELISA plates (25), as previously described (5). Briefly, mAbs against mouse or human integrins or isotype-matched control Abs were incubated in 96-well Maxisorp (Nalge Nunc International, Naperville, IL) plates at 20 μg/ml in carbonate buffer (pH 9.6), followed by blocking with 10% FCS. An alternative protocol was used to immobilize mAb F(ab′)₂ (which lack a carbohydrate-rich Fc-portion and thus bind less efficiently to ELISA plates) or to coat larger surfaces with mAbs for signaling experiments (because ELISA plates are only available in a 96-well format). In these cases, Abs were covalently bound to tissue culture plates according to a modified protocol of Berton et al. (10). Plates were incubated with 0.1 mg/ml poly-l-lysine (Sigma-Aldrich) for 30 min, washed, then incubated with 2.5% glutaraldehyde (Sigma-Aldrich) for another 15 min and washed again. Thereafter, mAbs or their F(ab′)₂ were added at 20 μg/ml and incubated for 4 h, followed by blocking of nonspecific binding sites. Identical results to those presented in this paper were also obtained when the unmodified immobilization protocol of Berton et al. (10) was used, i.e., when Abs were captured by protein G covalently bound to tissue culture dishes (not shown). It is worth noting that the responses to immobilized anti-CD18 Abs were not abolished when protein G was omitted from the protocol of Berton et al. (10), suggesting that additional mechanisms (e.g., direct binding to free glutaraldehyde groups) were still able to capture the Abs in the absence of protein G (not shown).

For experiments on integrin-ligand surfaces, plates were coated with 150 μg/ml human fibrinogen or 1 μg/ml recombinant murine ICAM-1 (provided by Dr. C. M. Vines, University of New Mexico, Albuquerque, NM) as described (5).

Murine or human neutrophils were added at 4 × 10⁵ or 1 × 10⁵ cells/well, respectively, in the presence or absence of 50 ng/ml murine or 20 ng/ml human TNF-α (PeproTech, Rocky Hill, NJ). Release of superoxide was determined by a cytochrome c reduction test as described (5), using a Labsystems iEMS microplate reader (Helsinki, Finland).

In experiments using blocking Abs, neutrophils were preincubated with 4 μg/10⁶ cells anti-mouse FcγRII/III (2.4G2) or Gr1 (RB6-8C5; isotype control) mAb, 4 μg/10⁶ cells anti-human FcγRIII (3G8) or CD45 (HI30; isotype control) mAb, or 6 μg/10⁶ of the Fab of the anti-human FcγRIIA (IV.3) mAb. The role of the p38 MAPK was tested by preincubating the cells with the indicated concentration of SB203580 (Calbiochem, San Diego, CA) for 15 min before stimulation. Cell viability, as determined by the Erythrosin B dye exclusion test, remained >95% after pretreatment with the maximal concentration (100 μM) of SB203580 or vehicle tested. Vehicle alone had no effect on the respiratory burst of neutrophils (not shown).

Neutrophils were plated on 96-well plates precoated with the indicated mAbs and incubated for 60 min either at 4°C or at 37°C. Although the former temperature allowed us to exclude secondary processes (e.g., spreading) which could alter the primary adhesion of the cells, incubation at 37°C gave a much more robust and consistent signal. After the incubation, the wells were washed three times with a Labsystems electronic multichannel Finnpipette at low speed setting and the number of cells that remained adherent was quantitated by an acid phosphatase assay as described (5).

For analysis of Syk tyrosine phosphorylation, neutrophils were incubated for 15 min at 37°C on 6-cm tissue culture dishes precoated with anti-CD18 mAbs as described above, followed by rapid cooling of the samples on ice. Radioimmunoprecipitation assay (RIPA) lysates of these cells were then prepared and used for immunoprecipitation with a polyclonal anti-Syk Ab (N-19 from Santa Cruz Biotechnology, Santa Cruz, CA) followed by phosphotyrosine immunoblotting as described (5). Overall tyrosine phosphorylation was determined from the same RIPA lysates as described (5). For the analysis of p38 MAPK phosphorylation, samples were lysed by a Triton-based lysis buffer added from a 5× concentrated stock solution (24) and lysates were processed for immunoblotting with phosphospecific (Cell Signaling Technology, Beverly, MA) or non-phosphospecific (C-20 from Santa Cruz Biotechnology) Abs against the p38 MAPK as described (24).

These experiments were performed in the absence of Mg²⁺ salts to avoid secondary integrin-mediated activation. For cross-linking of FcRs, murine neutrophils were incubated for 30 min on ice in the presence of 1 μg/10⁶ cells rat anti-mouse FcγRII/III Ab (2.4G2). Cells were washed, resuspended in the presence or absence of 12.5 μg/10⁶ cells mouse anti-rat IgG (clone 7/5), and incubated for 5 min at 37°C. The reaction was stopped by rapid cooling of the samples. The cells were then lysed by a Triton-based lysis buffer from a 5× concentrated stock solution and the phosphorylation of the p38 MAPK was determined as described (24). The effect of TNF stimulation was tested from similarly prepared lysates of neutrophils incubated for 5 min with 50 ng/ml murine TNF-α.

All presented data are representative of three or more independent experiments. Error bars represent SD of triplicate or quadruplicate readings.

For statistical analysis, the stimulus-induced responses of mutant cells or cells preincubated with blocking Abs were expressed in percent of those of wild-type or control preincubated cells in each experiment. The thus obtained normalized values were then averaged across all similar experiments. The Student t test was then used to determine whether this average (referred to as “remaining response”) was statistically different from 100%. Mean, SD, and p values from these calculations are mentioned in the text.

Adhesion-dependent activation of neutrophils can be induced by the addition of proinflammatory agonists to neutrophils plated on surfaces coated with integrin ligand proteins (2). Neutrophils can also be activated by plate-bound mAbs against CD18, the common β-chain of β₂ integrins, in the absence of an additional stimulus (10). Although both activation systems require CD18 (4, 5), their dependence on divalent cations differs significantly, as shown in Fig. 1, A and B. Consistent with the Mg²⁺-dependent binding of integrins to their natural ligands, superoxide release induced by TNF stimulation of murine neutrophils on fibrinogen-coated surfaces does not occur in the absence of Mg²⁺ salts (Fig. 1,A). However, superoxide release triggered by immobilized anti-CD18 mAbs does not depend on the presence of Mg²⁺ salts in the extracellular space (Fig. 1 B), supposedly because this activation proceeds through an Ag-Ab interaction which, unlike the binding of integrins to their natural ligands, does not require Mg²⁺ ions. Thus, removal of Mg²⁺ salts from the extracellular medium can ensure the activation of neutrophils through anti-integrin mAbs without concomitant ligation by physiological integrin ligands. All further experiments with anti-CD18 mAb stimulation were thus performed in the absence of Mg²⁺ salts.

Next, we tested, using a number of isotype-matched control Abs, whether the specificity of anti-CD18 mAbs against the β₂ integrin chain is important for their stimulatory effect. An isotype-matched control Ab which does not recognize any cell surface Ags was unable to induce any respiratory burst from murine neutrophils and, more importantly, plate-bound Abs against L-selectin or the murine granulocytic maturation marker Gr1 also failed to induce superoxide release from these cells (Fig. 1,C). Binding of the anti-L-selectin and anti-Gr1 mAbs to neutrophils in suspension, as determined by flow cytometry, was even superior to that of the anti-CD18 mAb (Fig. 1,D) and the adhesion of neutrophils to the plate-bound anti-L-selectin or anti-Gr1 mAbs was also similar to or higher than their adhesion to the anti-CD18 mAb, both at 4°C (Fig. 1,E) or 37°C (Fig. 1 F). These results suggest that cross-linking of other, nonintegrin cell surface molecules by immobilized mAbs is not sufficient to induce cell activation, indicating that the integrin binding of immobilized anti-CD18 mAbs is essential for their stimulatory effect.

Neutrophil activation by immobilized anti-integrin mAbs is being viewed as a measure of direct cross-linking of integrins without stimulation of other cell surface molecules. However, despite the fact that intact Abs are used in these experiments, the contribution of FcR activation by the Ab Fc portions has never been explored. As a first attempt to test this possibility, we determined the responses of neutrophils genetically deficient in the FcR γ-chain (FcRγ), the common signaling chain of most FcRs, to plate-bound anti-integrin mAbs (26, 27). To our surprise, FcRγ^−/− neutrophils failed to release superoxide when plated on immobilized anti-CD18 mAb surfaces (Fig. 2,A; 5.2 ± 0.4% remaining response, p = 6.0 × 10⁻⁶, n = 3). A similar defect was seen when neutrophils were activated by plate-bound mAbs against the α-chains of β₂ integrins (CD11a or CD11b; data not shown), or against non-β₂ integrins such as the α₄ (CD49d) chain (which pairs with the β₁ chain; Ref. 16) or the β₃ (CD61) integrins (Fig. 2 B). These results suggest that FcRs are involved in the responses of neutrophils to anti-integrin mAbs. However, because FcRγ is also involved in signaling from a number of non-FcR molecules (see Discussion), further identification of the molecules involved was required before the role of FcRs could be formally proven.

To better define the receptor possibly linking anti-integrin mAb stimulation to the FcR γ-chain in murine neutrophils, we tested the responses of cells lacking the low affinity FcγR, FcγRIII (CD16). As shown in Fig. 3,A, FcγR3^−/− neutrophils failed to release superoxide when plated on immobilized anti-CD18 mAbs (2.3 ± 8.6% remaining response, p = 1.5 × 10⁻¹⁸, n = 17). Importantly, CD18 was expressed at similar levels on the surface of wild-type and FcγR3^−/− neutrophils (Fig. 3,B), and the cells from the two genotypes adhered similarly to the immobilized anti-CD18 mAbs (Fig. 3,C), indicating that the decreased functional response is not due to an integrin surface expression or primary adhesion defect. Respiratory burst in response to cross-linking of the α₄ (CD49d) or β₃ (CD61) integrin chains was also defective in FcγR3^−/− neutrophils (Fig. 3 D) despite normal expression of these molecules on FcγR3^−/− cells and normal adhesion of the mutant neutrophils to the immobilized anti-CD49d or anti-CD61 mAbs (not shown). Similarly, the respiratory burst response was also defective when FcγR3^−/− neutrophils were plated on immobilized mAbs against the α-chains (CD11a, CD11b) of β₂ integrins or the β₁ (CD29) integrin chain (not shown). Thus, functional responses of murine neutrophils to plate-bound anti-integrin mAbs requires FcγRIII.

In addition to triggering an antimicrobial response (respiratory burst), plating murine neutrophils on anti-CD18 mAb surfaces also induced the tyrosine phosphorylation of multiple protein bands (Fig. 4,A). This phosphorylation response was significantly decreased in FcγR3^−/− cells (Fig. 4,A), again indicating the contribution of FcγRIII to anti-integrin mAb-induced responses. Of the multiple bands phosphorylated in wild-type cells, a 65–75 kDa cluster is expected to include the Syk tyrosine kinase which is involved in both FcR (28, 29, 30) and integrin (5, 31, 32) signaling. In fact, Syk became tyrosine phosphorylated upon anti-CD18 stimulation and this response was diminished in FcγR3^−/− neutrophils (Fig. 4,B). Another prominent phosphoprotein migrates at ∼40 kDa, similar to the migration of the p38 MAPK. As shown in Fig. 4 C, p38 MAPK became phosphorylated in wild-type, but not FcγR3^−/−, neutrophils plated on immobilized anti-CD18 mAbs.

Taken together, cross-linking of murine neutrophil integrins by anti-integrin mAbs is not sufficient for the initiation of respiratory burst or the activation of intracellular phosphorylation events; instead, contribution from FcγRIII is required for all these responses.

Human neutrophils can also be activated by plate-bound Abs against CD18, but the requirement for FcRs cannot be tested by genetic modification of these cells. Our results with murine cells suggest, however, that the Fc portion of anti-CD18 mAbs may be functionally important for full activation of neutrophils. To test this possibility, we plated human neutrophils on plate-bound intact anti-human CD18 or isotype-matched control mAbs, or F(ab′)₂ prepared from the same Abs. Similar to the results obtained with murine cells, intact anti-human CD18 mAbs, but not intact isotype-matched control Abs were able to induce a pronounced respiratory burst from human neutrophils (Fig. 5,A). However, no response could be induced when the cells were plated on immobilized F(ab′)₂ of the same anti-CD18 mAbs (Fig. 5,A; 1.5 ± 1.7% remaining response, p = 9.9 × 10⁻⁵, n = 3), even though the adhesion of the cells to the intact mAbs and the F(ab′)₂ was indistinguishable (Fig. 5 B). These results indicate that the Fc portion of the anti-integrin mAbs is essential for their stimulatory effect on neutrophils.

If binding of FcRs to the Ab Fc portions is in fact required for the respiratory burst initiated by anti-integrin mAbs, then preincubation of neutrophils with FcR-blocking Abs would be expected to inhibit this response. As shown in Fig. 6,A, respiratory burst of murine neutrophils plated on anti-CD18 mAbs was significantly inhibited by preincubation of the cells with an FcγRII/III-blocking Ab (52 ± 11% remaining response, p = 4.2 × 10⁻⁵, n = 7), compared with control cells pretreated with an isotype-matched Ab to Gr1. This effect was not due to insufficient binding of the control mAbs to the cell surface, because the binding of the anti-Gr1 mAb was even higher than that of the anti-FcγRII/III mAb (Fig. 6 B). It is unlikely that the anti-murine FcγRII/III blocking mAb exerted its effect on FcγRII, because mice do not possess the activating FcγRIIA, while FcγRIIB primarily mediates inhibitory effects. Thus, these experiments confirm by an independent approach that FcγRIII is required for responses of murine neutrophils to anti-integrin mAbs.

Unlike murine neutrophils which only express a single low affinity activatory FcγR, FcγRIII, human neutrophils have two such proteins, FcγRIIA and FcγRIII (of the two isoforms of the latter protein, human neutrophils only express the GPI-linked FcγRIIIB). We next attempted to determine whether mAbs directed against either of these proteins are able to block the activation of human neutrophils plated on anti-integrin mAb-coated surfaces. Based on our results with FcγR3^−/− murine neutrophils, we first tested the effect of a function-blocking anti-human FcγRIII mAb or an isotype-matched control mAb against human CD45 on responses of human neutrophils. Interestingly, there was no difference between the anti-CD18-induced responses of cells preincubated with the control or the anti-FcγRIII mAb except for a marginal (though consistent) delay in the kinetics of the anti-FcγRIII-preincubated cells (Fig. 6,C; 99 ± 11% remaining response at 60 min, p = 0.99, n = 3). The anti-FcγRIII mAb has likely exerted its blocking effect on FcγRIII because the Ab strongly inhibited the responses of neutrophils to immobilized immune complexes (not shown). Thus, human FcγRIII does not appear to be significantly involved in responses of human neutrophils plated on immobilized anti-integrin mAbs. We next tested whether a blocking mAb against human FcγRIIA is able to attenuate the same responses. Because a suitable isotype-matched control Ab recognizing another neutrophil surface molecule was not available, we have generated Fabs of the anti-FcγRIIA mAb to exclude possible side effects due to the presence of the Fc portion of the blocking mAb. As shown in Fig. 6 D, this anti-human FcγRIIA Fab strongly inhibited the response of human neutrophils plated on immobilized anti-CD18 mAbs (25 ± 15% remaining response, p = 0.013, n = 3), even though the adhesion of the cells to the anti-CD18 coated plate was not affected (not shown). These results again confirm that low affinity activating FcγRs are indispensable for neutrophil responses to anti-integrin mAbs and indicate that in human neutrophils, this costimulatory signal proceeds through FcγRIIA rather than through FcγRIII.

Traditionally, integrin-dependent responses of neutrophils are triggered by stimulation of the cells with soluble inflammatory agonists (cytokines, chemokines, bacterial products) while adherent to integrin ligand surfaces. To test whether low affinity FcγRs are also required for these responses, we tested the effect of the FcγR3^−/− mutation or the FcγR-blocking mAbs on these responses. As shown in Fig. 7,A, TNF-induced respiratory burst was indistinguishable in wild-type and FcγR3^−/− murine neutrophils plated on fibrinogen (105 ± 29% remaining response, p = 0.58, n = 10) or recombinant murine ICAM-1 surfaces. Similarly, pretreatment with the anti-murine FcγRII/III mAb also failed to inhibit these same responses in wild-type murine cells (Fig. 7,B; 117 ± 15% remaining response on fibrinogen surface, p = 0.45, n = 3) and the anti-human FcγRIIA Fab did not affect the respiratory burst of human neutrophils stimulated with TNF while adherent to a fibrinogen-coated surface (Fig. 7 C; 110 ± 35% remaining response, p = 0.67, n = 3). Thus, the low affinity FcγRs involved in anti-integrin mAb-induced respiratory burst are not required for TNF-induced responses of neutrophils adherent to integrin ligand surfaces.

The above experiments suggest that cross-linking of integrins alone is not sufficient for full activation of neutrophils and instead a second signal (either from inflammatory cytokines or FcRs) is essential for optimal stimulation. To further illustrate the compatible and interchangeable nature of second signals originating either from cytokines or FcRs, we attempted to rescue the defective superoxide release of FcγR3^−/− neutrophils on anti-integrin mAbs by an additional TNF stimulus. This experiment was performed in the absence of Mg²⁺ ions and thus integrins can only bind to the substratum through Ag-Ab interaction to the immobilized anti-CD18 mAbs, but not through a conventional receptor-ligand binding mechanism (see Fig. 1, A and B). As shown in Fig. 8, TNF stimulation of wild-type cells plated on isotype control mAbs did not induce a considerable effect. Plating the same cells on anti-CD18 mAbs induced a robust response even in the absence of TNF and this response was only slightly increased upon addition of TNF, supposedly because the nonintegrin costimulatory signal was already maximal in the absence of TNF. In contrast, while anti-CD18 stimulation alone failed to induce any response from FcγR3^−/− cells, addition of TNF to this reaction induced a strong activation of FcγR3^−/− cells which reached 75 ± 12% (n = 4) of the response of wild-type cells plated on anti-CD18-coated surfaces in the absence of TNF (Fig. 8). These results suggest that TNF stimulation can substitute the second, nonintegrin signal lacking in FcγR3^−/− cells and it also indicates that anti-CD18 mAbs can provide the “first”, integrin-mediated signal under these conditions.

The above experiments suggest that adherent activation of neutrophils requires two parallel signals: the first one originating from integrins and the second one coming either from soluble agonists (e.g., TNF) or from ligation of FcRs (in particular, FcγRIII in case of murine or FcγRIIA in case of human cells). It would be tempting to speculate that the two nonintegrin signals converge on a single common pathway, which would then support the other, integrin-mediated signal. One candidate for such a common nonintegrin signaling pathway may be the p38 MAPK cascade. This possibility is supported by the fact that both TNF simulation (Fig. 9,A; 2.0 ± 0.7-fold increase, n = 3) and cross-linking of FcγRII/III (Fig. 9,B; 1.5 ± 0.3-fold increase, n = 4) leads to phosphorylation of p38 MAPK in suspended wild-type murine neutrophils under Mg²⁺-free conditions. Because the ligation of integrins is strongly unfavored under these conditions, it is likely that the activation of p38 MAPK by TNF and FcRs can occur independent of integrin signaling. The role of the p38 MAPK in FcγRIII signaling is further supported by the fact that the phosphorylation of this kinase in response to anti-CD18 is impaired in FcγR3^−/− neutrophils (see Fig. 4 C above).

To further test the requirement for the p38 MAPK during integrin-mediated activation of neutrophils, we determined the effect of SB203580, a p38 MAPK inhibitor, on respiratory burst induced by immobilized anti-integrin mAbs or by TNF stimulation of fibrinogen-adherent neutrophils. SB203580 inhibited, in a dose-dependent manner, the respiratory burst of both murine (Fig. 9,C) or human (Fig. 9 D) neutrophils stimulated either by TNF on a fibrinogen-coated surface or through plate-bound anti-CD18 mAbs. However, the inhibition of the TNF-induced respiratory burst of neutrophils adherent to fibrinogen was more sensitive to SB203580 than the responses to immobilized anti-CD18 mAbs. Though the reason for this difference is presently unclear, it should be noted that in our prior studies (24), the maximal inhibition of the p38 MAPK in fMLP-stimulated human neutrophils was not attained until above 10 μM SB203580, which is closer to the sensitivity of the anti-CD18-induced responses to the drug.

Though the requirement for integrins in a number of neutrophil functions is well established, the relative contribution of integrin and nonintegrin signals to the eventual response of these cells is poorly understood. Integrin-mediated adherent activation of neutrophils is traditionally attained through stimulation of neutrophils, adherent to integrin ligands, by a soluble inflammatory agonist (2). Though the hierarchy of these dual signals was initially unclear, the observation that integrin cross-linking by anti-integrin mAbs was sufficient for full activation of neutrophils (10) suggested that integrins alone are responsible for the initiation of neutrophil responses and the role of soluble agonists is restricted to promoting the binding of integrins to their ligands, e.g., by an inside-out signal increasing the affinity and/or avidity of integrins. In this context, Ab-mediated integrin cross-linking became the standard way to initiate integrin outside-in signaling and to study the mechanisms of integrin signaling without the need to interpret a second, nonintegrin signal. All results obtained so far have conformed to this simple and intrinsically consistent signaling paradigm.

Our results presented in this paper place the above findings in a very different perspective. Neutrophil activation by anti-integrin mAbs also appears to follow the two-signal model whereby a second, nonintegrin signal is again required for full activation of the cells. In this system, however, the second, nonintegrin signal is provided through low affinity FcγRs (FcγRIII in the case of murine, and FcγRIIA in case of human cells) rather than through a soluble inflammatory stimulus (Figs. 3–6). If so, then there is no reason to believe that integrin cross-linking alone is sufficient for full activation of neutrophils. These results also argue against a linear signaling pathway whereby the sole function of a second nonintegrin signal is to increase the binding capacity (affinity or avidity) of integrins to their ligands.

Our results indicate that anti-integrin Abs initiate a cooperative signal between integrins and FcRs. There has been a number of observations suggesting that integrins are functionally coupled to other molecules present on the surface of neutrophils, including Fcγ- and FcαRs, urokinase-type plasminogen activator receptors, tetraspanins, or the CD47 integrin-associated protein (reviewed in Refs. 33 and 34). Neutrophil activation by anti-integrin mAbs may proceed through the same mechanisms as these cooperative signals and thus the experimental approach used in this paper may provide a possible model system for the analysis of neutrophil activation through coligation of integrins and FcRs.

During this project, the first indication for a role of FcRs came from the observation of defective anti-integrin mAb-induced responses of FcRγ^−/− neutrophils, which lack the common signaling chain of multiple FcRs (Fig. 2). However, though FcRγ was traditionally viewed as a signaling chain exclusively linked to FcR function, a number of non-FcR cell surface molecules have recently been shown to signal through FcRγ. These include GpVI, the collagen receptor of platelets (35, 36); the paired Ig-like receptor PIR-A on multiple lineages of the immune system (37, 38, 39); the osteoclast-specific OSCAR protein (40, 41); and the recently described DCAR molecule on dendritic cells (42). In agreement with the role of FcRγ in non-FcR functions, FcRγ^−/− neutrophils also failed to respond to certain stimuli that are clearly not expected to signal through FcRs (A. M. and C. A. L., unpublished observation). Thus, care should be taken when interpreting results obtained exclusively from the use of FcRγ^−/− mice. Our additional experiments with FcγR3^−/− animals (Figs. 3 and 4) and FcγR-blocking mAbs (Fig. 6), as well as the inability of an anti-CD18 F(ab′)₂ to induce cell activation (Fig. 5) have nevertheless provided clear evidence for the role of FcRs in neutrophil responses to anti-integrin mAbs.

The results with the FcγR3^−/− neutrophils could be interpreted either such that FcRs are ligated by the Ig Fc portions of the anti-integrin mAbs, or FcRs (in this case, the murine FcγRIII) may only be required as a signaling or scaffolding intermediate (e.g., by being a component of the integrin receptor-complex) without the need for binding to extracellular Igs. The facts that immobilized F(ab′)₂ of anti-CD18 mAbs are unable to induce a respiratory burst (Fig. 5) and the responses to anti-integrin mAbs can be inhibited by blocking Abs against the ligand-binding domains of FcRs (Fig. 6) support the first scenario, i.e., the Fab and Fc portions of anti-integrin mAbs bind to integrins and FcγRs, respectively.

Our experiments on murine neutrophils identify FcγRIII as the FcR required for activation of these cells by anti-integrin mAbs. Although FcγRIII is the only low affinity activating FcγR in mice, humans express three such receptors: FcγRIIIA (which corresponds to murine FcγRIII), as well as FcγRIIA and FcγRIIIB. These three low affinity receptors use slightly different proximal signaling mechanisms: FcγRIIA contains an intrinsic ITAM whereas FcγRIIIA (similar to the murine FcγRIII) uses the ITAM of the noncovalently linked FcR γ-chain for downstream signaling. The phosphorylation of these ITAMs recruits the Syk tyrosine kinase which will then be responsible for downstream signaling. FcγRIIIB, in contrast, has no transmembrane domain (it is instead linked to the plasma membrane by a GPI anchor) and does not associate with FcRγ. Though the mechanism and functional role of FcγRIIIB activation is still unclear, some investigators believe that it is functionally linked to FcγRIIA and thus delivers an activatory signal (43, 44). Human neutrophils express two of the above mentioned receptors: FcγRIIA and FcγRIIIB. Our experiments with anti-FcγRIII blocking mAbs (which block both FcγRIIIA and FcγRIIIB) indicate that FcγRIIIB is not involved in neutrophil responses to anti-integrin mAb stimulation. Instead, human neutrophils use FcγRIIA for this purpose, as indicated by the inhibitory effect of an anti-FcγRIIA blocking Ab on the activation of human neutrophils plated on immobilized anti-integrin mAbs (Fig. 6,D). This issue also raises the interesting question of whether FcRγ, the signaling chain of several FcRs (including murine FcγRIII), but not of human FcγRIIA (which has an intrinsic ITAM) would also be required for anti-integrin-induced activation of human neutrophils, similar to what was observed in the case of murine cells (Fig. 2). Despite these issues and the possible differences between the human and mouse cells, all our data consistently indicate that anti-integrin mAb stimulation of neutrophils requires low affinity FcγRs in both species.

Activation of neutrophils by anti-integrin mAbs has been used by a number of groups (including ourselves) to investigate the mechanisms responsible for integrin outside-in signaling (3, 5, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20). Those results should be reinterpreted in light of the new findings on the role of FcRs in this assay system. Furthermore, while the use of immobilized anti-integrin mAbs has been the most specific and most widely used approach to cross-link integrins, there have been other, less well-defined attempts to stimulate neutrophil integrins without any additional stimuli. These include the activation of neutrophils through a plate-bound recombinant polyvalent integrin ligand (poly-RGD) (5, 18, 19) and stimulation of adherent cells by Mn²⁺ ions which supposedly induce the high affinity conformation of integrins (45). Results (in particular, the extent of neutrophil activation) obtained in those systems will need to be carefully re-evaluated in light of the findings presented in this paper. In particular, possible caveats arising from insufficient blocking of tissue culture plastic (which may activate yet unknown cell surface receptors) and the widespread and rather nonspecific effects of Mn²⁺ ions will need to be reconsidered.

Of particular interest is the fact that TNF stimulation of FcγR3^−/− neutrophils on anti-CD18 surfaces is able to restore the defective responses of these cells (Fig. 8). On one hand, this observation confirms that anti-CD18 mAbs are capable of delivering an integrin-mediated signal even in the absence of FcγRIII if another signal is provided, e.g., in the form of TNF stimulation. It also suggests that the second, nonintegrin signal can interchangeably arise either from FcRs or from soluble inflammatory stimuli. Our experiments showing that the p38 MAPK is activated by both TNF and FcRs in an apparently integrin-independent manner (Fig. 9, A–B) raises the possibility that the two nonintegrin signals converge on this kinase. The role of the p38 MAPK is further supported by the inhibitory effect of SB203580 on both the TNF-induced responses of neutrophils on fibrinogen-coated surfaces and the responses initiated by immobilized anti-CD18 mAbs (Fig. 9, C and D). It is presently unclear why these two assay systems are differentially sensitive to SB203580. One possible explanation is that TNF and FcRs activate two different isoforms of the p38 MAPK. Neutrophils express at least two such isoforms, p38α and p38δ (46, 47), which are differentially activated by LPS and oxidative stress (46). Of the two isoforms, p38α is significantly more sensitive to inhibition by SB203580 (48).

Taken together, by identifying another group of cell surface receptors involved in anti-integrin mAb stimulation of neutrophils, our results challenge the current view that integrin signaling alone is sufficient for full activation of these cells. The two-signal hypothesis suggested by the results presented in this paper would provide a more controlled release of the highly toxic antimicrobial substances of neutrophils and it would also fit better with the general working mechanism of the immune system where cellular responses are initiated by multiple parallel signaling mechanisms.

We thank Árpád Mikesy, Krisztina Makara, and Yongmei Hu for mouse colony maintenance; Anna Erdei and Péter Csermely for access to the animal facility and other equipment in Budapest; Erzsébet Seres-Horváth, Krisztina Sütő, and Edit Fedina for technical assistance; Cheryl Vines for rICAM-1; Glória László and the Salk Institute for hybridoma stocks and suggestions about mAb production; and Eric Brown and Steven Edwards for helpful thoughts about FcR-blocking experiments.