The GTPase-activating protein ARAP3 regulates chemotaxis and adhesion-dependent processes in neutrophils

Neutrophils are an essential component of the innate immune responses to invading pathogens and tissue damage. In response to inflammatory stimuli, neutrophils are recruited from the bloodstream to infected tissues, where they produce reactive oxygen species (ROS) and phagocytose and kill pathogens.¹  It is vital that the neutrophil response to inflammatory stimuli is well controlled. Leukocyte adhesion deficiency (LAD) patients, in whom the recruitment process is impaired, suffer from repeated, severe bacterial infections. Conversely, excessive neutrophil responsiveness is also detrimental, causing tissue damage and inflammation, as exemplified in autoimmune diseases.

Work aimed at identifying the molecular machinery controlling neutrophils has identified important regulatory players. Rap isoforms regulate adhesion-dependent processes by modulating integrin affinity and avidity.²  Loss of leukocyte Rap1 activation is a cause of LAD-III,^3,4  demonstrating the crucial role of this small GTPase in neutrophil control. Agonist-activated PI3Ks generate on activation the lipid second messengers phosphatidylinositol-(3,4,5)-trisphosphate [PtdIns(3,4,5)P₃] and PtdIns(3,4)P₂. These cause plasma membrane recruitment and activation of PI3K effectors, which regulate multiple enzymes and pathways.⁵  Known PI3K effectors include regulators of Rho and Arf family small GTPases, GTPase-activating proteins (GAPs), and guanine nucleotide exchange factors. Rho family small GTPases are involved in regulating the neutrophil NADPH oxidase and chemotaxis,^6,7  whereas Arf family members regulate phagocytosis, ROS production, and chemotaxis.^8,9 

ARAP3 was identified from porcine neutrophils in a screen for PtdIns(3,4,5)P₃–binding proteins. In vitro, ARAP3 is a PtdIns(3,4,5)P₃–dependent Arf6 GAP and a Rap-activated RhoA GAP. In addition, in vivo, PtdIns(3,4,5)P₃ drives the recruitment of ARAP3 to the plasma membrane, bringing it into the vicinity of its activator, Rap-GTP, and its substrates, RhoA-GTP and Arf6-GTP.^10,11  The remaining ARAP family members (ARAP1/2) share ARAP3's multidomain structure.^12,13  We sought to identify ARAP3's physiologic function in neutrophils. Using neutrophils from a conditional Arap3^−/− mouse model in biologic assays, we demonstrate that ARAP3 regulates adhesion-dependent processes. Loss of ARAP3 causes the preactivation of neutrophil β2 integrins and hyperresponsiveness in adhesion-dependent situations, feeding into several biologic responses such as adhesion-dependent ROS formation, granule release, and chemotaxis. Whereas Rap activity was not affected by the loss of ARAP3, RhoA activation was increased in an adhesion-dependent setting, suggesting a role for ARAP3 downstream of Rap in neutrophils. Our results suggest a modulatory role for ARAP3 in neutrophils, which prevents their activation in inappropriate situations.

Unless otherwise stated, materials were obtained from Sigma-Aldrich.

Generation of the Arap3^fl/fl mouse has been described previously.¹⁴  For the inducible knockout, Arap3^fl/fl mice were crossed with Rosa-ERT2Cre⁺ mice.¹⁵ Cre expression was induced by a single intraperitoneal injection with tamoxifen, as described previously,¹⁶  which caused a complete but transient deletion of ARAP3 in bone marrow–derived neutrophils. Mice were housed in open racks or individually ventilated cages in small animal barrier units at the Babraham Institute or Walter-Brendel-Zentrum and used 11 or 12 days after induction. Animal work carried out at the Babraham Institute and Walter-Brendel-Zentrum was approved by United Kingdom Home Office Project licenses PPL80/1875 and PPL80/2335 and Regierung Oberbayern, Germany (AZ55.2-1-54-2531-134/08), respectively.

Neutrophils were isolated from the bone marrow of sex- and age-matched mice 10-14 weeks of age with discontinuous Percoll gradients (Amersham), as described previously,^17,18  using endotoxin-free reagents throughout.

Tail vein blood was collected in EDTA-coated microvettes (Sarstedt) for analysis using a Vet ABC animal blood cell counter.

ROS production was measured by chemiluminescence using a luminol-based assay in polystyrene 96-well plates (Berthold Technologies), as described previously.¹⁹  Cells (5 × 10⁵) were incubated with luminol (150μM) and HRP (18.75 U/mL) for 10 minutes at 37°C. Where indicated, neutrophils were added manually to wells precoated with fibrinogen (150 μg/mL) or polyRGD (20 μg/mL) with or without murine TNFα (20ng/mL); or to immobilized immune complexes (100μg/mL IgG-BSA). For soluble agonist assays (fMLF [fMet Leu Phe] 10μM final concentration), cells were primed for 1 hour at 37°C in the absence (mock) or presence of mouse TNFα (1000 U/mL) and GM-CSF (100 ng/mL). For all assays, measurements were started immediately and light emission was recorded. Data output was in relative light units (RLUs) per second or total RLUs integrated over the indicated measured periods of time.

For micropipette chemotaxis assays, neutrophils were resuspended in HBSS, 15mM HEPES, pH 7.4, and 0.05% BSA, and allowed to attach to a glass coverslip. Cells were stimulated at 37°C with a point source of fMLF delivered from a microinjection pipette (Femtotip; Eppendorf) using 50 mbar of pressure, as described previously,¹⁸  and monitored by time-lapse imaging for 30 minutes using an inverted Zeiss Axiovert microscope and Axiovision v3.0 software. Chemotaxis assays involving Dunn chambers or EZ-TAXIS chambers were carried out as described previously^18,20  using fMLF as the chemoattractant. Chemotaxis in transwells was assayed using 3-μm-pore polycarbonate filter units (Millipore) as described previously.¹⁸  Chemotaxis in a 3D collagen matrix was as described previously,^21,22  except that neutrophils were prepared by centrifugation over a discontinuous Percoll gradient without prestimulation with TNFα. Dulbecco PBS was used to dilute rat tail collagen (Roche). Cells were tracked using the “manual tracking” plug-in and tracks were analyzed using the “chemotaxis tool” plug-in (Ibidi) in ImageJ v1.37 software.

MAPK, Erk, and PKB activation assays were carried out essentially as described previously,²³  with cells plated onto polyRGD-coated or heat-inactivated FCS (hiFCS)–blocked dishes. For details, see supplemental Methods (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Release of gelatinase granules after plating onto a fibrinogen-coated surface, or as induced by stimulation with fMLF and cytochalasin B, was detected by zymography, as described previously.²⁴ 

Surface integrins were visualized by FACS analysis of stained cell populations. Integrin clustering was analyzed by confocal microscopy of unstimulated, stained cells and used as a readout for β2 integrin avidity. To measure integrin affinity, unstimulated cells were allowed to bind to ICAM1-Fc in solution; bound ICAM1-Fc was measured using quantitative immunofluorescence (see supplemental Methods for details).

Cohorts of C57B/6 mice were lethally irradiated and subsequently reconstituted by tail vein injection with 4 × 10⁶ bone marrow cells from Arap3 model mice or from sex- and age-matched controls.

Rectangular glass capillaries (VitroCom) with a cross-section of 40 μm × 400 μm were coated with rmE-selectin (20 μg/mL), rmCXCL1 (keratinocyte-derived chemokine, 15 μg/mL), and rmICAM-1 (15 μg/mL), and served as ex vivo flow chambers, as described previously²⁵  (see supplemental Methods for details).

Intravital microscopy was performed essentially as described previously²⁶  (see supplemental Methods for details).

Cremaster muscle whole mounts were prepared as described previously.²⁵  Leukocytes were Giemsa stained in fixed tissues and analyzed microscopically (see supplemental Methods for details).

RhoA-GTP was determined by G-LISA assay (Cytoskeleton).²⁷  Arf6-GTP and Rap1-GTP were identified by pull-down assay, as described previously^28,29  (see supplemental Methods for details).

We recently reported that deleting Arap3 in the germline causes embryonic lethality.¹⁴  To analyze the function of ARAP3 in neutrophils, we crossed Arap3^fl/fl mice with mice expressing a tamoxifen-inducible Cre (ERT2Cre). A single Cre induction caused the temporary loss of ARAP3 protein from bone marrow–derived neutrophils of Arap3^fl/flERT2Cre⁺ animals (Figure 1A). ARAP3 protein expression was not affected by injecting tamoxifen into Arap3^fl/fl or into Arap3^+/+ERT2Cre⁺ control mice, or by mock inducing (vehicle control) any of these mice (Figure 1A). ARAP1 and ARAP2, which are expressed only weakly in murine bone marrow–derived neutrophils, were not affected by inducibly deleting Arap3 (data not shown). Similarly, peripheral blood cell counts from tamoxifen- or mock-induced Arap3^fl/flERT2Cre⁺ animals were not affected (Figure 1B).

Neutrophils produce ROS on appropriate stimulation. PI3Ks and Rho, Rap, and Arf family small GTPases are involved in regulating the phagocyte oxidase.^6,9,30  We found no difference in ROS production by neutrophils from tamoxifen- or mock-induced Arap3^fl/flERT2Cre⁺ mice or controls after stimulation with the soluble agonist fMLF (Figure 2A-B and supplemental Figure 1A for controls), nor with the nonphysiologic, soluble stimulus phorbol 12-myristate 13-acetate (data not shown). This indicated that the machinery required for ROS production was intact in neutrophils lacking ARAP3.

Neutrophils attached to several surfaces generate long-lasting, β2 integrin–dependent ROS when costimulated with an additional proinflammatory stimulus (eg, a cytokine or bacterial product).^31,32  In contrast, the synthetic multivalent integrin ligand polyRGD does not require any costimulation.²³  We found that adhesion-dependent ROS production after plating cells onto polyRGD-coated surfaces was increased in tamoxifen-induced Arap3^fl/flERT2Cre⁺ neutrophils compared with controls (Figure 2C-D and supplemental Figure 1B) in the presence or absence of TNFα. We observed the same trend with neutrophils plated onto fibrinogen-coated surfaces, again with or without costimulation with TNFα (Figure 2E-F and supplemental Figure 1C). This suggested that β2 integrins in the ARAP3-deficient neutrophils might be present in a preactivated state, able to produce adhesion-dependent ROS even in the absence of costimulation with a proinflammatory cytokine.

Immune complexes activate neutrophils by signaling through cell-surface Fc receptors. This also induces ROS production,²⁴  a function that is critical for autoimmune inflammatory disorders. Fc receptors and integrins share signaling intermediates.³³  We assessed ROS production after plating tamoxifen-induced Arap3^fl/flERT2Cre⁺ neutrophils (and the relevant controls) in immobilized immune complex–coated wells (BSA–anti-BSA). In agreement with the results described in the previous paragraph, we observed a heightened response when cells lacking ARAP3 were plated onto immune complex–coated surfaces (Figure 2G-H and supplemental Figure 1D). In summary, ROS production due to integrin (or Fc-receptor) engagement, but not to stimulation with a soluble agonist, was up-regulated in neutrophils lacking ARAP3.

Prompted by the results obtained with the ROS assays, we investigated whether ARAP3 might also be important for other events downstream of integrin engagement in neutrophils. We analyzed the phosphorylation status of protein kinase B (PKB, also called Akt) and of p38 MAPK, 2 downstream components of outside-in signaling. We plated neutrophils onto dishes that had been blocked with hiFCS or coated with polyRGD. Both kinases were hyperphosphorylated in tamoxifen-induced Arap3^fl/flERT2Cre⁺ neutrophils, but not in controls, after plating on polyRGD (Figure 3A and supplemental Figure 2A).

In neutrophils, integrin outside-in signaling leads to degranulation, firm adhesion, and spreading. To determine whether ARAP3 is involved in these functions, we plated mock- and tamoxifen-induced Arap3^fl/flERT2Cre⁺ neutrophils in hiFCS-blocked and fibrinogen-coated plastic wells and quantified gelatinase granule release. Gelatinase granule release was increased in tamoxifen-induced Arap3^fl/flERT2Cre⁺ neutrophils (Figure 3B and supplemental Figure 2B), but not in controls, upon plating the cells on fibrinogen. Control stimulation with a powerful soluble stimulus (fMLF in combination with cytochalasin B) did not lead to an increased release of gelatinase granules in ARAP3-deficient cells (Figure 3B and supplemental Figure 2B), indicating that the cellular machinery required for granule release was unaffected.

We analyzed cell adhesion and spreading 10 minutes after plating neutrophils onto dishes coated with integrin ligands and found increased adhesion and spreading of tamoxifen-induced Arap3^fl/flERT2Cre⁺ neutrophils compared with controls (Figure 3C shows adhesion and spreading after plating onto polyRGD and supplemental Figure 2C shows the controls). The same trend was observed when cells were plated onto fibrinogen (data not shown).

We analyzed adhesion of peripheral blood neutrophils to immobilized E-selectin, ICAM1, and CXCL1 ex vivo in a flow chamber system,²⁵  a setup that represents a more accurate reflection of the situation of neutrophils in the bloodstream because of the constant shear forces. ARAP3-deficient cells again adhered better to the coated chamber surface (Figure 3D). Our results show that intracellular events downstream of integrin activation were up-regulated in Arap3^−/− neutrophils (outside-in signaling).

We focused on the 2 major neutrophil β2 integrins, LFA-1 and Mac-1, to identify the reason for the observed increased responses of adhesion-dependent neutrophil functions. The total amount of Mac-1 in neutrophils that did or did not contain ARAP3 was unaltered (supplemental Figure 3A). Similarly, surface LFA-1 and Mac-1 on neutrophils that had or had not been prestimulated with TNFα was nearly identical on tamoxifen- and mock-induced Arap3^fl/flERT2Cre⁺ neutrophils (Figure 4A), suggesting that the differences in biologic activities observed were not due to an increased number of surface integrins.

Integrin ligand binding affinity and avidity are regulated by inside-out signaling. Integrins reversibly change their tertiary structure between low-, intermediate-, and high-affinity status, and they reversibly cluster, modulating their binding avidity.³⁴  To compare β2 integrin affinity of neutrophils from tamoxifen- or mock-induced Arap3^fl/flERT2Cre⁺ mice in the absence of affinity status–specific antibodies for murine β2 integrins, we allowed cells to bind to ICAM1 in solution and analyzed binding using quantitative immunofluorescence on cells that had settled on glass slides after fixation. ARAP3-deficient cells bound to ICAM1 significantly more strongly than controls (Figure 4B), whereas cells isolated from tamoxifen and mock-induced Arap3^+/+ERT2Cre⁺ mice behaved in the same manner (supplemental Figure 3B). We next analyzed the avidity of LFA-1 and Mac-1 by staining neutrophils kept in solution using antibodies for these surface integrins. After fixation, cells were allowed to settle on slides for microscopic analysis. LFA-1 was significantly more clustered in ARAP3-deficient neutrophils than in controls (Figure 4C and supplemental Figure 3C); this trend was also observed for Mac-1, whereas cells from tamoxifen- and mock-induced Arap3^+/+ERT2Cre⁺ mice displayed similar distributions of both LFA-1 and Mac-1 (supplemental Figure 3D). In summary, β2 integrin affinity and avidity were increased in ARAP3-deficient neutrophils, indicating increased inside-out signaling.

We analyzed chemotaxis in neutrophils that did or did not express ARAP3. We initially performed “needle chemotaxis” assays in which a chemoattractant-filled micropipette coupled to a microinjector set to a constant, slow stream was inserted into a glass-bottomed dish containing adherent neutrophils, causing the formation of a steep gradient of chemoattractant. Control neutrophils adopted a polarized shape and migrated consistently toward the tip of the micropipette, but neutrophils from tamoxifen-induced Arap3^fl/flERT2Cre⁺ mice were unable to chemotax toward fMLF in this assay (Figure 5A and supplemental Videos 1-2; controls are shown in supplemental Figure 4A). These neutrophils extended a pseudopod in one direction for a short time, then retracted this pseudopod and generated another one in a different direction and so forth, making little if any net progress in any direction in the process. This indicated that neutrophils lacking ARAP3 had a severe chemotaxis defect, although they clearly reacted to being stimulated with fMLF.

For a more quantifiable response, we performed chemotaxis assays in Dunn chambers. For these assays, cells moved on a “bridge” through a 20-μm-deep gap between 2 glass surfaces in a shallow gradient of chemoattractant.³⁵  Unexpectedly, Arap3^−/− neutrophils were able to move directionally in Dunn chambers (Figure 5B). However, tracking the chemotaxing cells and analyzing the tracks confirmed that neutrophils from tamoxifen-induced Arap3^fl/flERT2Cre⁺mice had a significantly reduced ability to chemotax toward fMLF in terms of Euclidean distance (the shortest distance between the start and end points) and total accumulated distance covered, and also in their speed (Figure 5C-D). As expected, there were no significant differences between neutrophils from tamoxifen- or mock-induced Arap3^+/+ERT2Cre⁺ control mice in the Dunn chamber assays (supplemental Figure 4B).

Neutrophils from tamoxifen- and mock-induced Arap3^fl/flERT2Cre⁺ mice in EZ-TAXIS chambers³⁶  in which cells were lined up at a narrow gap (5-μm depth) through which they squeezed when exposed to a shallow gradient of chemoattractant (fMLF; Figure 5E) exhibited the same trend. Interestingly, the differences between neutrophils from tamoxifen- and mock-induced Arap3^fl/flERT2Cre⁺ mice chemotaxing in EZ-TAXIS chambers were more subtle than in those measured in the Dunn chamber assays (Figure 5F-G).

Several in vitro studies with neutrophils lacking integrins^21,37,38  or in which integrin function had been inhibited using antibodies or chelators³⁹  have shown that integrins are required for migration on 2D but not in 3D substrates (or when “chimneying” between close glass surfaces). To our knowledge, this issue has not been addressed in depth in cells with increased integrin activity. We wondered whether the different results we obtained in the different chemotaxis assays might reflect the fact that we were measuring different mixtures of integrin-dependent and integrin-independent chemotaxis. We reasoned that needle chemotaxis assayed highly integrin-dependent chemotaxis on a 2D glass coverslip, whereas Dunn and EZ-TAXIS chambers, both of which analyze chemotaxis of cells squeezing through more or less narrow gaps, might assay mixtures of 2D and 3D chemotaxis. To determine whether ARAP3 might be most important for integrin-dependent chemotaxis, we performed chemotaxis assays in integrin-independent setups in which cells migrated in 3D substrates. We measured chemotaxis toward fMLF in Boyden chambers (transwells), in which the cells squeeze through 3-μm polycarbonate pores, which has previously been documented as being integrin independent,⁴⁰  and found no significant differences between ARAP3-deficient and control neutrophils (Figure 5H). We also analyzed neutrophil chemotaxis in a 3D collagen matrix (Figure 5J), which has previously been shown to support chemotaxis of entirely integrin-deficient leukocytes.²¹  Interestingly, analysis of tracks from neutrophils from tamoxifen- and mock-induced Arap3^fl/flERT2Cre⁺ mice migrating in the collagen gels indicated that there was no difference in the total accumulated distances migrated (or in the speed, data not shown) in the 2 groups of cells, but Euclidean distances were different (Figure 5K). Hence, ARAP3-deficient cells ended up significantly nearer to their starting points than controls, suggesting a directionality defect in ARAP3-deficient cells. Neutrophils from tamoxifen- or mock-induced Arap3^+/+ERT2Cre⁺ mice behaved identically (supplemental Figure 4C).

In summary, our experiments show that ARAP3-deficient neutrophils displayed a context-dependent chemotaxis defect in vitro. Our data suggest an integrin-dependent component to ARAP3-regulated chemotaxis, which affects the total distance traveled by the cells in a substrate-dependent fashion. They also suggest an additional directional chemotaxis defect in neutrophils lacking ARAP3, which appears to be integrin independent.

During inflammation, the recruitment of neutrophils that circulate in the bloodstream in a quiescent, nonadhesive state follows a well-established cascade of adhesion and activation steps,⁴¹  starting with the capture of free-flowing neutrophils to the inflamed endothelium, followed by rolling along the endothelium. Both capture and rolling are mediated by selectins. During rolling, endothelial-expressed chemokines, together with selectin-mediated signaling events, trigger the activation of neutrophil-expressed β2 integrins (inside-out signaling), which leads to further slowing down and eventually to firm neutrophil arrest on the endothelium. This is followed by steady strengthening of the adhesive contacts and neutrophil spreading along the endothelium, postarrest steps that require interactions between neutrophil integrins and their endothelial ligands (outside-in-signaling). Finally, neutrophils start crawling along the luminal surface of the inflamed endothelium in search of an appropriate exit point out into the tissue.

To address the relevance of our observations in vivo, we analyzed neutrophil recruitment in mouse cremaster muscle venules in the absence of ARAP3. Because our in vitro assays suggested a preactivated state of β2 integrins in ARAP3-deficient neutrophils, we evaluated baseline adhesion and extravasation of neutrophils in unstimulated cremaster muscle whole mounts. To prevent trauma-induced inflammatory changes, tamoxifen- and mock-induced Arap3^fl/flERT2Cre⁺ and Arap3^+/+ERT2Cre⁺ mice were killed before preparation of the cremaster muscle. In agreement with the previous in vitro adhesion and flow chamber assays, intravascular adhesion was increased in tamoxifen-induced Arap3^fl/flERT2Cre⁺ animals compared with controls, which exhibited low numbers of adherent intravascular neutrophils under these conditions (Figure 6A). In addition, the baseline number of perivascular neutrophils was also found to be higher in ARAP-3–deficient animals compared with controls (Figure 6B and supplemental Figure 5). These findings corroborate our in vitro data indicating that deletion of ARAP3 causes preactivation of neutrophils.

To further analyze the functional relevance of ARAP3 in vivo under inflammatory conditions and to exclude a possible effect of ARAP3 in the endothelial compartment, we evaluated neutrophil adhesion and extravasation in TNFα-stimulated cremaster muscle whole mounts of tamoxifen-induced Arap3^fl/flERT2Cre⁺ and Arap3^+/+ERT2Cre⁺ bone marrow chimeras. Stimulation with TNFα resulted in an increased recruitment of neutrophils in both groups compared with baseline conditions. However, the number of ARAP3-deficient, adherent intravascular neutrophils was significantly higher than that of controls. There were also more perivascular ARAP3-deficient cells (Figure 6C). We next directly evaluated intravascular adhesion in postcapillary venules of the cremaster muscle by intravital microscopy. Microvascular and hemodynamic parameters were similar between the groups (supplemental Table 1). These experiments showed an increased adhesion efficiency using the equation (adherent cells/mm²)/systemic WBCs in ARAP3-deficient neutrophils compared with controls (Figure 6D). More adherent ARAP3-deficient neutrophils stayed firmly attached and showed a decreased tendency to crawl intraluminally compared with controls (Figure 6E and supplemental Videos 3-4). In agreement with observations from chemotaxis assays in 3D matrices in vitro, extravascular ARAP3-deficient cells were clearly able to migrate interstitially.

Rap1 has previously been shown to modulate integrin affinity and avidity in several settings.²  We measured Rap1 activity in neutrophils that were kept in suspension or that had been allowed to adhere to polyRGD. We saw a robust activation of Rap1 on plating cells on polyRGD, but there was no significant difference in Rap1 activity between cells lacking ARAP3 and controls (Figure 7A). This indicated that Rap1 activity was not affected by the presence or absence of ARAP3 in neutrophils.

We previously showed that ARAP3 regulates the small GTPases Arf6 and RhoA. We therefore performed activity assays with neutrophils kept in suspension or plated onto polyRGD. Measuring global Arf6-GTP in lysates from these neutrophils, we saw a dramatic activation of Arf6 upon adhesion to polyRGD, but we did not detect any significant differences between the 2 genotypes (Figure 7B). When analyzing RhoA, we saw an increased activity in cells plated onto polyRGD. Although the extent of activation was variable between experiments, the RhoA activation in ARAP3-deficient cells was increased every time (Figure 7C), whereas RhoA activation in neutrophils from tamoxifen- and mock-induced Arap3^+/+ERT2Cre⁺ mice was very similar (supplemental Figure 6). In summary, our data suggest that ARAP3 modulates RhoA in an adhesion-dependent context. These findings are in agreement with our previous work identifying ARAP3 as a Rap-regulated RhoA GAP.¹¹ 

ARAP3, one of 24 PtdIns(3,4,5)P₃–binding proteins identified from porcine neutrophils,¹⁰  is abundant in human and mouse neutrophils. The present study represents the first characterization of ARAP3 function in these highly specialized immune cells. Because germline deletion of Arap3 is embryonically lethal,¹⁴  we used a tamoxifen-inducible, ubiquitous Cre to delete Arap3 conditionally in the adult mouse to efficiently (although transiently) delete Arap3 in bone marrow–derived neutrophils. Our mouse model was unsuitable for any longer-term in vivo experiments such as models of autoimmune diseases. In addition, we observed variable basal peritoneal leukocyte numbers even in intraperitoneally mock-injected control mice, rendering our model unsuitable for assessment of leukocyte recruitment after thioglycollate-induced sterile peritonitis. Given that Cre expression can in some circumstances lead to nonspecific artifacts, tamoxifen induction was kept to a minimum by routinely comparing tamoxifen- and mock-induced Arap3^fl/flERT2Cre⁺ mice with tamoxifen- and mock-induced Arap3^+/+ERT2Cre⁺ mice, controlling for tamoxifen and for Cre expression. None of the phenotypes described for ARAP3-deficient neutrophils were due to tamoxifen or to inducing Cre. However, in agreement with previous studies finding nonspecific effects caused by expression of Cre,^42,43  we observed a significant loss of erythrocytes in the bone marrow of tamoxifen-induced Arap3^fl/flERT2Cre⁺ and Arap3^+/+ERT2Cre⁺ mice 12 days after tamoxifen induction (supplemental Figure 7). This side effect was temporary, because no differences in bone marrow erythrocytes were found 6 weeks after induction (data not shown). Because we observed no differences in circulating blood cells, we are confident that this temporary loss of bone marrow erythrocytes had no effect on any of the measurements we carried out.

We noted increased responses with ARAP3-deficient neutrophils in all adhesion-dependent responses that were tested, including in vitro assays carried out with purified bone marrow–derived neutrophils, ex vivo assays in flow chambers, and in vivo observations in the cremaster muscle system. A response was triggered even in ARAP3-deficient, unstimulated cells both in vitro (eg, increased ROS production in response to plating onto fibrinogen in the absence of TNFα and increased gelatinase granule release in cells plated in heat-inactivated FCS-blocked wells) and in vivo (eg, increased baseline adhesion in unstimulated cremaster muscle venules). We measured no differences in surface Mac-1 and LFA-1, the most abundant, constitutively expressed neutrophil integrins. Instead, both the affinity and the avidity of these β2 integrins were significantly increased in cells lacking ARAP3 (inside-out signaling). Our data support 2 conclusions: (1) both inside-out and outside-in signaling could be altered independently of one another, or (2) the increased responses observed in events downstream of integrin ligation could, at least in part, be a direct consequence of the increased inside-out signaling we measured.

In addition to regulating responses that are classically thought of as downstream of integrin ligation, we observed a striking, context-dependent chemotaxis defect in ARAP3-deficient neutrophils. The literature describes integrin-dependent and integrin-independent modes of cell migration. Integrin-dependent migration involves force generation because of integrin-dependent, transient adhesions to a typically 2D substratum, whereas integrin-independent migration involves amoeboid squeezing of cells through narrow pores of a 3D matrix, propelled by the force of a polarized actomyosin cytoskeleton. Relevant experiments have been carried out either with healthy human neutrophils in which integrin function was inhibited, using chelators or blocking antibodies,³⁹  or with cells taken from LAD patients.³⁷  More recently, analysis of chemotaxis with integrin-deficient murine neutrophils supported very convincingly the conclusions drawn in these earlier studies.^21,38  Fewer studies have addressed the integrin dependency of chemotaxis in cells with hyperactive integrins, as is the case in ARAP3-deficient neutrophils. We conclude that in ARAP3-deficient neutrophils, integrin-dependent chemotaxis (needle chemotaxis, in which cells migrate on a glass coverslip) is defective because the adhesive contacts formed with the glass coverslip are too tight. Chemotaxis in Dunn or EZ-TAXIS chambers represent intermediates between integrin-dependent and integrin-independent chemotaxis, with an increasing amount of chimneying between the glass coverslips forming the bridge performed by cells the tighter the gap between the 2 coverslips. This explains why we observed a smaller defect with ARAP3-deficient neutrophils in EZ-TAXIS chambers with their narrow gaps (which were previously used to assay neutrophil chemotaxis in the presence of chelating agents⁴⁰ ) than in Dunn chambers, in which the gap is significantly wider. Finally, chemotaxis in integrin-independent circumstances (eg, in a 3D collagen gel or a polycarbonate transwell in vitro or in interstitial migration in vivo) was not affected in ARAP3-deficient cells in terms of the total accumulated distance traveled by the neutrophils. Therefore, varying integrin dependency in different chemotaxis assays is the most likely explanation for the reproducible yet seemingly contradictory results we obtained with these different chemotaxis assays. We propose that ARAP3 is required for integrin-dependent chemotaxis, but is dispensable for integrin-independent chemotaxis. In addition, our data hint that there may be an additional defect in directionality of chemotaxis. ARAP3-deficient cells failed to persistently polarize in the orientation of the chemoattractant in needle chemotaxis assays. In addition, despite equal total accumulated distances, ARAP3-deficient neutrophils migrated shorter Euclidean distances in 3D matrices. Future work will be directed at establishing the nature of any directionality defect in chemotaxis of ARAP3-deficient neutrophils.

When analyzing cremaster muscle venules, we observed increased numbers of intravascular, but also perivascular, ARAP3-deficient leukocytes, suggesting increased extravasation. However, given the increased adhesion to the endothelium, it is conceivable that transmigration of a fraction of these cells could have masked a mild transmigration defect. Unfortunately, we were unable to address this possibility experimentally because we could not use the sterile peritonitis model.

Having identified PI3K as major regulator of ARAP3,^10,14  the phenotype of ARAP3-deficient neutrophils was somewhat surprising. We previously showed that ARAP3 is regulated by Rap-GTP in addition to PI3K.¹¹  Interestingly, Rap proteins have been demonstrated to be major modulators of integrin inside-out signaling in lymphocytes, macrophages, neutrophils, and platelets, affecting integrin affinity and avidity (although not surface expression), ligand binding, and cell spreading.^30,44-46  Rap activity was not affected in neutrophils devoid of ARAP3, indicating (in agreement with our previous work¹¹ ) that ARAP3 is not upstream of Rap.

We assessed Arf6-GTP in neutrophils kept in solution or plated onto polyRGD, which led to significant Arf6 activation. We did not notice any differences between ARAP3-deficient neutrophils and controls. Given that ARAP3 is just one of several Arf6 GAPs, and that we analyzed global Arf6 activity, this negative result was not very surprising. It is conceivable that an analysis of localized Arf6 activity would reveal differences that are hidden behind the robust activation of Arf6 on plating neutrophils onto polyRGD.

When assaying RhoA activity, we noticed that the increase measured in ARAP3-deficient cells was subtly increased compared with controls in each experiment (on average, 1.6-fold). RhoA has been shown to regulate leukocyte integrin activation in several situations.^47-50  In thymocytes, RhoA was shown to be required for optimal integrin activation by Rap1,⁵¹  indicating that RhoA can be part of Rap-dependent inside-out activation of integrins. Given that Rap regulates ARAP3 as a Rho GAP and that ARAP3-deficient neutrophils are characterized by heightened β2 integrin activity and hyperresponsiveness in adhesion-dependent processes, we speculate that Rap activates ARAP3 as a Rho GAP to fine-tune integrin activity in neutrophils. Because we know that PI3K activity is required for ARAP3 activation, it is likely to function as a coincidence detector for PtdIns(3,4,5)P₃ and Rap-GTP. It will be interesting to dissect the regulatory inputs of PI3K and Rap into ARAP3 in the future.

The authors thank Simon Walker for help with image analysis, Nick Ktistakis for the anti-βCOP antibody, Michelle Janas for help with procedures, Anthony Green (Cambridge Institute for Medical Research, Cambridge, United Kingdom) for use of a blood cell counter, and John Ferguson, Su Kulkarni, and Michael Sixt (Institute of Science and Technology, Klosterneuberg, Austria), and Klaus Ley (La Jolla Institute for Allergy and Immunology, La Jolla, CA) for helpful discussions.

This work was supported by the Biotechnology and Biological Sciences Research Council and the Medical Research Council (G0700740). S.V. holds a David Phillips Fellowship (BB/C520712).

Contribution: L.G., K.E.A., C.N., M.S., and S.V. designed experiments; L.G., K.E.A., C.N., T.M., L.N., M.S., and S.V. performed experiments; T.L. provided reagents; L.G., C.N., A.S.-P., P.T.H., M.S., L.S., and S.V. analyzed and/or interpreted data; and S.V. and M.S. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Sonja Vermeren (née Krugmann), The Inositide Laboratory, The Babraham Institute, Babraham Research Campus, Cambridge CB22 3AT, United Kingdom; e-mail: sonja.vermeren@bbsrc.ac.uk.