Asthma is a common disease that causes considerable morbidity. Increased numbers of airway eosinophils are a hallmark of asthma. Mechanisms controlling the entry of eosinophils into asthmatic lung have been intensively investigated, but factors regulating migration within the tissue microenvironment are less well understood. We modeled this by studying chemoattractant and growth factor-mediated human eosinophil migration within a three-dimensional collagen matrix. Stimulation with GM-CSF induced dose-dependent, random migration with a maximum of 77 ± 4.7% of cells migrating. In contrast, CCL11 and C5a caused a more modest although significant degree of migration (19 ± 1.8% and 20 ± 2.6%, respectively). Migration to GM-CSF was partially dependent on Ca2+ and αΜβ2 integrins. The Rho family of small GTPases regulates intracellular signaling of cell migration. GM-CSF-induced migration was only partially dependent on Rho kinase/Rho-associated kinase (ROCK) and was independent of RhoA activation. In contrast, CCL11-induced migration was fully dependent on both RhoA and ROCK. Activation of RhoA was therefore neither necessary nor sufficient to cause eosinophil migration in a three-dimensional collagen environment. This study suggests that eosinophil growth factors are likely to be required for eosinophil migration within the bronchial mucosa, and this involves signal transduction pathways distinct from those used by G protein-associated chemoattractants.
Asthma is a common global condition that causes considerable morbidity to both adults and children. A characteristic feature of asthma is the relatively selective accumulation of eosinophils within the bronchial mucosa, and there is substantial circumstantial evidence implicating these cells in causing several aspects of the disease. An extensive amount of research has been undertaken to define the mechanisms controlling eosinophil egress from the circulation into tissue, and the processes involved are now largely understood (1, 2, 3, 4).
The microlocalization of leukocytes within tissue is of considerable importance to their function in health and disease (5, 6). Once in the tissue, the eosinophil interacts with structural cells such as the epithelium or nerves and also migrates into the airway lumen where they undergo apoptosis. The factors regulating the migration and microlocalization of eosinophils within the three-dimensional (3D)3 tissue environment are not known. It is predicted that the fate of the eosinophil will be controlled by the way in which the cell integrates signals from growth factors (such as GM-CSF), adhesion (from matrix and structural cells), and chemoattractants (e.g., CCL11/eotaxin) within a 3D context.
Cell migration is an integral part of leukocyte function and is tightly controlled. The Rho family of small GTPases are key regulators of signaling pathways critical to cytoskeletal rearrangements and migration (7, 8). It is now well established that migration requires protrusion at the leading edge, contraction of the cell body, and detachment of adhesions at the cell rear. During leukocyte migration, RhoA is thought to regulate rear detachment through effector proteins such as Rho kinase/Rho-associated kinase (ROCK) (9, 10). However, recent studies have shown that differential requirements for Rho- and ROCK-dependent migration vary according to cell type and the environment (11, 12, 13). These studies demonstrate that our knowledge of the role of Rho GTPases during migration in a complex environment is unclear and appear to be cell context-dependent.
Many cell types display dramatically different properties, migrating in a 3D environment, such as tissue, rather than in a two-dimensional (2D) environment, like a microscope slide surface (14, 15). Lung remodeling in asthma is characterized by collagen deposition in tissue parenchyma (2, 16, 17). Therefore, in this study, we sought to dissect the molecular signals regulating eosinophil migration using a 3D collagen gel model system to mimic the bronchial tissue. Interestingly, we show that most eosinophils migrated in response to GM-CSF, but relatively few cells migrated in response to the G protein-linked chemoattractants CCL11 and C5a, and no cells migrated in response to lysophosphatydic acid (LPA). Neither RhoA nor ROCK was necessary or sufficient for growth factor-mediated eosinophil motility. This study demonstrates that in a 3D environment, eosinophils have differing requirements for RhoA and ROCK depending upon the stimulus, and that growth factor-dependent migration assumes a central role in mediating cell motility.
Materials and Methods
Materials and reagents
GM-CSF, CCL11, and anti-β1 integrin were obtained from R&D Systems. C3 exoenzyme, anti-RhoA Ab, Cytoskelfix, and Rho activation assay were obtained from Cytoskeleton (tebu-bio). Histopaque and C5a was purchased from Sigma-Aldrich. EDTA and 10× MEM were purchased from Invitrogen. Y-27632 was obtained from Alexis Biochemicals. Collagen was purchased from INAMED. Anti-integrin αΜ and anti-β2 integrin Abs were purchased from Chemicon International. Alexa Fluor 546-phalloidin and DAPI were obtained from Invitrogen: Molecular Probes.
Human blood (100 ml) was taken from normal, mild atopic, and asthmatic volunteers with the approval of the Leicestershire Health Ethics Committee. RBC were sedimented with 6% dextran. Eosinophils were purified by density gradient centrifugation with Histopaque. Cells were further purified by negative immunomagnetic selection using anti-CD16-coated beads and the magnetic cell separation (MACS) system (Miltenyi Biotec). Eosinophils obtained were typically >96% pure. By trypan blue cells were 100% viable after isolation and 96% viable after overnight culture without serum. All studies have been reviewed and approved by the appropriate Institutional Review Committee.
Collagen gels and migration
Three-dimensional collagen gels were constructed as described by Freidl and Brocker (18). Purified eosinophils (3–4 × 106 cells/ml) in 2% FBS were used directly in the migration assays. In all experiments, control cells were unstimulated eosinophils. Cells were mixed with collagen I, 10× MEM, and various chemoattractants. The gel was loaded in a chamber made from a coverslip sealed with wax on a microscope slide and incubated at 37°C for 1 h before videotaping. Eosinophil migration was recorded by time-lapse videomicroscopy using a Zeiss Axiovert 25 microscope with a heated stage at magnification ×20 (numerical aperture of 0.3) and a QImaging Retiga 1300 CCD camera. The movies were recorded and analyzed by commercially available Improvision software (Fig. 1). The x–y-coordinates of cells were acquired by the software over the 20-min period and the pathway (cell track) was automatically calculated from this data. From the migratory pathway data, the software calculates the velocity, distance migrated, and displacement from the point of origin. The percentage of migration was calculated for all cells that were tracked for >3 min and migrated >10 μm from the point of origin. For chemotaxis (gradient) gels, 3D migration chambers were fabricated as above except the chambers were only filled three-fourths of the way, allowed to polymerize, and then filled with fluid chemoattractant, sealed, and filmed. Directionality was calculated by counting the number of cells going toward the chemoattractant as shown on the vector graphs plotted by the software. All data in the paper were generated using the chemokinetic model except for Fig. 3. Additionally, CCL11 was determined to be fully active in the Boyden chamber chemotaxis assay (data not shown).
Rho activation assay
The assay was used according to the manufacturer’s instructions. Briefly, freshly isolated eosinophils were starved of serum for 18 h before the experiment. Cells were used at 2 × 106 per data point and were then stimulated for 3 min with cytokines, lysed, and centrifuged at 10,000 rpm for 2 min at 4°C. Supernatants were checked for protein concentration, and equal amounts of protein were loaded in wells of the Rho-GTP affinity plate. Signal was detected with RhoA Ab and plates were read at OD490. LPA was used to stimulate cells as a positive control for cellular activation of RhoA. Recombinant His-Rho was used as a positive control for the assay.
RhoA-GTP pulldown assays
Freshly isolated eosinophils were serum starved for 18 h. Cells were then stimulated for 3 min with cytokines, lysed, and flash frozen in liquid N2. Thirty micrograms of protein was separated for input control. The lysates were stored at −80°C. On the day of the experiment, 300 μg of protein (2 × 106 cells per data point) was incubated with 50 μg of Rhotekin-RBD beads (Cytoskeleton) for 1 h at 4°C. Levels of Rho-GTP and total Rho were visualized by SDS-PAGE and Western blotting using an anti-RhoA Ab.
Immediately after videotaping, collagen gels were immersed in Cytoskelfix for 4 min at −20°C as per the manufacturer’s protocol. Gels were then washed in PBS and permeabilized in 0.1% Triton X-100 for 10 min. Cells were then blocked with 10% normal goat serum for 1 h and washed with PBS. F-actin was visualized with 1 μM Alexa Fluor 546-phalloidin. Rho was stained at 1/100 with mouse anti-human mAb. Cells were mounted on slides with Prolong. Images were captured using an EC Plan-Neofluar ×63 (numerical aperture of 1.25) oil immersion lens on a Zeiss Axiovert 200M microscope using Openlab software (Improvision) and were edited using Photoshop (Adobe Systems).
Results are expressed as means ± SE. All statistics were paired t tests unless stated otherwise. Significance was attained at p < 0.05. A commercially available software program, GraphPad Prism, was used (Graphpad Software).
Induction of motility by chemoattractants
To investigate how eosinophils migrate in a complex, 3D environment, we tested the ability of GM-CSF, CCL11, and C5a to elicit chemokinetic motility of eosinophils in a dose-dependent manner in 3D collagen matrices. Without stimulation, we consistently observed that <3% of cells migrated. GM-CSF elicited a robust, dose-dependent, chemokinetic response with most cells (∼70–80%) migrating in a random manner to concentrations of 0.05 ng/ml and above (Fig. 2,A). CCL11 and C5a are both potent eosinophil chemoattractants that may be involved in asthma (19, 20, 21). In contrast to GM-CSF, only ∼20% of cells migrated in response to optimal concentrations of these chemoattractants (Fig. 2, B and C). Additionally, LPA has also been shown to stimulate eosinophil migration (22). However, LPA (10 μM) was unable to elicit migration of eosinophils in our 3D collagen matrix (data not shown). We also observed no differences in migration rates between asthmatic and nonasthmatic samples (data not shown). To investigate differences in eosinophil motility induced by a chemoattractant gradient within the 3D matrix, we modified the collagen matrix set-up as described in Materials and Methods to create such a gradient. In a side-by-side comparison, the percentage of eosinophils migrating in response to CCL11 was the same in a nongradient vs a gradient format (Fig. 3,A). In contrast, the percentage of cells migrating in response to GM-CSF was much lower in a gradient compared with a chemokinetic model. However, of the cells that migrated, more cells moved in a directional manner toward the source of CCL11 than toward vehicle control or GM-CSF (Fig. 3 B).
Comparison of chemoattractant-induced motility
The few unstimulated cells that did migrate spontaneously traveled with an average velocity of 7.8 μm/min. This was similar to the velocity of CCL11-stimulated cells (8.4 μm/min) but significantly slower than GM-CSF-stimulated cells (13.02 μm/min) (Fig. 4,A). Peak velocities of 24–30 μm/min were not unusual for either GM-CSF- or CCL11-induced migrating cells. The average distance migrated was approximately the same for GM-CSF and CCL11 at ∼120 μm from origin (Fig. 4,B). This was significantly greater than unstimulated, spontaneously migrating control cells. Treating eosinophils with an optimal concentration of CCL11 and variable concentrations of GM-CSF resulted in an inhibitory effect of CCL11 at low concentrations of GM-CSF, but this was not observed at higher concentrations of GM-CSF (Fig. 4 C).
Chelating Ca2+ or blocking β2 integrins inhibits motility
Many adhesive interactions are calcium-dependent (23, 24, 25). We used EDTA to determine the effect of depleting extracellular Ca2+ on eosinophil migration. EDTA (5 mM) significantly inhibited GM-CSF-mediated migration by ∼50% (Fig. 5,A). To further investigate the molecular mechanisms involved in adhesive interactions in a more complex environment, we examined the role of β integrins in 3D collagen matrices. Blocking β1 integrins, αLβ2 (LFA-1), or αXβ2 (P150,95) had no effect on eosinophil migration (data not shown). However, blocking αMβ2 (Mac-1/CD11b) (Fig. 5,B) or β2 (CD18) (Fig. 5 C) significantly inhibited GM-CSF-induced migration to the same degree as EDTA.
Critical differential requirements for the Rho GTPase pathway
The RhoA/ROCK pathway is important in cell migration. To elucidate the role of RhoA/ROCK in eosinophil migration within the collagen gel, we investigated the effect of Y-27632, a Rho kinase/ROCK inhibitor (26), on cell motility. Treating eosinophils with 10 μM Y-27632 for 30 min before the addition of chemoattractants significantly inhibited CCL11-induced migration (Fig. 6). However, Y-27632 had no effect on cells stimulated with 5 ng/ml GM-CSF, although there was significant inhibition of migration induced by 0.5 ng/ml of GM-CSF. These data indicate that ROCK is necessary for CCL11-stimulated migration, but not for migration to the maximal dose of GM-CSF. Next, we sought to analyze the activation of Rho in eosinophils by CCL11 and GM-CSF. Human eosinophils were stimulated for 3 min and tested for their ability to activate Rho. Eosinophils stimulated with CCL11 and LPA demonstrated significant activation of RhoA protein after 3 min. However, stimulation of eosinophils by GM-CSF failed to activate RhoA above basal levels (Fig. 7,A). To confirm our results with the RhoA activation assay, we measured RhoA activation by pulldown experiments using Rhotekin-Rho-binding domain beads, which preferentially react with active GTP-bound RhoA. Levels of total and active GTP-bound RhoA obtained by Western blotting showed results similar to those of the activation assay. Again, cells were stimulated for 3 min with CCL11 or GM-CSF. Compared with control cells or cells stimulated with 5 ng/ml GM-CSF, CCL11-stimulated cells demonstrated an increase in active GTP-bound RhoA (Fig. 7,B). We also looked at RhoA localization in cytokine-stimulated cells. Eosinophils migrating in response to CCL11 stimulation demonstrated clear RhoA staining compared with GM-CSF-stimulated or unstimulated cells (Fig. 7,C). These data demonstrate that RhoA is in its active GTP-bound form in CCL11-stimulated cells, but not in cells stimulated with GM-CSF. We then tested what effect inhibition of Rho would have on migrating eosinophils. We inhibited Rho activation by using cell-permeable C3 exoenzyme, a selective Rho inhibitor, in our 3D gels. Eosinophils were treated with 4 μg/ml C3 for 2 h at 37°C before the addition of chemoattractants and construction of the gels. Migration of CCL11-stimulated cells, which showed active GTP-bound Rho, was completely abolished by inhibiting Rho (Fig. 7 D). No inhibition of migration was observed with stimulation by GM-CSF at both 5 and 0.5 ng/ml.
Eosinophils are primarily tissue-dwelling cells, and the mechanisms that control their microlocalization in tissue are likely to be of importance to their role in both health and disease. This is the first study to investigate eosinophil migration in a 3D matrix environment that seeks to model the tissue environment. Using this model, we observed that eosinophils are unable to move unless stimulated. Furthermore, using the chemokinetic model, growth factor stimulation causes most cells to move in a random fashion, whereas chemoattractant stimulation only results in a minority of cells migrating. Interestingly, in the gradient version of this 3D model, the percentage of cells migrating to GM-CSF was much lower and similar to CCL11, although more cells migrated in a directional fashion to CCL11. The differences were associated with use of distinct signal transduction pathways in response to GM-CSF compared with CCL11.
We used a 3D collagen matrix because collagen deposition is increased in asthmatic lung (2, 16, 17) and collagen may be the most critical component for structural integrity and load-bearing and therefore critical for homeostasis and cellular responses to injury (27). Freshly isolated eosinophils were introduced into the collagen matrix, which was allowed to polymerize, and then migration was observed during a 20-min period using time-lapse video microscopy. During the course of the 20 min, some cells would migrate into the gel and out of the field of view, so the distance moved is probably an underestimate. The cells moved with variable speeds, but our data are in agreement with previous reports that show leukocytes have faster migration rates than do other cell types, such as fibroblasts or endothelial cells, in collagen matrices (15, 18, 28). As chemoattractants such as CCL11 generally cause cells to migrate in greater numbers in Boyden chamber assays than do the chemokinetic growth factors such as GM-CSF, we expected CCL11 to induce at least as much migration as GM-CSF in the gel (29, 30, 31, 32). However, this was not the case. For a cell to move it needs to adhere at its leading edge and disadhere at the uropod. Inspection of the nonmotile eosinophils appeared to show that the CCL11-stimulated cells had an activated appearance compared with unstimulated cells in that their leading edge was active but that they could not release their uropod. It is possible therefore that within the context of a 3D environment without a chemotactic gradient that CCL11 caused the cells to adhere too avidly, thus preventing detachment. In support of this concept and in contrast to the priming effect of GM-CSF with other chemoattractants in Boyden chamber assays, we observed that CCL11 actually inhibited GM-CSF-induced migration at lower concentrations.
In agreement with previous reports (24, 25), we found that depletion of extracellular Ca2+ with the calcium chelator, EDTA, inhibited eosinophil motility. Although many adhesive interactions are calcium-dependent, one potential mechanism for inhibition of GM-CSF-induced migration by calcium chelation may be due to inactivation of matrix metalloproteinases (MMPs). Degradation of extracellular matrix by MMPs is important to migrating structural cells such as fibroblasts and was also involved in eosinophil migration through matrigel in a Boyden chamber assay (33). However, migrating leukocytes have been shown to use nonproteolytic amoeboid shape change to negotiate movement through collagen lattices (34). Leukocytes such as T cells, monocytes, and dendritic cells crawl along collagen fibrils using contact guidance and squeeze through existing gaps in the lattice as opposed to digestion with proteases. Nonetheless, it remains possible that MMPs are involved in eosinophil migration through collagen gels, and this idea deserves further investigation.
The data on integrin blocking is more complex. β1 integrins are generally viewed as the classical collagen receptors (23) and block leukocyte migration in vitro and in vivo (35, 36, 37). Our data demonstrate that in a more complex, 3D environment, eosinophil motility is β1-independent (data not shown) and this coincides with previous data on migrating leukocytes in a 3D environment (28). However, function dictates that leukocytes are fast-moving cells, and as haptokinetic migration models predict that expression of β1 integrins and adhesive strength may be inversely proportional to migration speed, there may be reason for leukocytes to use other integrins (38). Additionally, β2 integrins are specific to leukocytes and are promiscuous in their ligands including extracellular matrix components. Our data show that blocking β2 and αMβ2 partially inhibits migration.
For leukocytes migrating on a 2D substrate, the prevailing wisdom says that Rho is localized to the cytoplasmic tail and functions through its effector proteins such as Rho kinase/ROCK to retract the tail (9, 10, 39, 40). Furthermore, previous studies show that the Rho pathway is involved in eosinophil migration in vitro (22, 41) including tail retraction (22) and the migratory response to CCL11 (41). However, studies on MDCK and HeLa cells show that Rho/ROCK activity in the tail of migrating cells is context-dependent, as cells migrating at low density on collagen-coated glass coverslips showed high Rho activity in the tail, but cells migrating in a monolayer did not (42). Furthermore, more precise live cell imaging studies using fluorescently tagged RhoA probes show RhoA is active both at the leading edge and the tail of migrating cells (42, 43). Compared with a 2D substrate, migrating cells have continuous morphological changes and much more dynamic interactions with their substrate in a 3D environment. Thus, our imaging data showing a more diffuse distribution of RhoA in CCL11-stimulated motile cells may be due to cell contact with the 3D nature of their environment. This hypothesis is borne out by our data showing diffuse expression of β2 integrins throughout the migrating cell body (data not shown), when on 2D substrates β2 expression is located only at the leading edge of migrating leukocytes (44).
In contrast to the eosinophil response to CCL11, we found that these cells demonstrated robust motility in response to GM-CSF. It is possible that the extreme motility of GM-CSF-stimulated cells is due to the activation of other GTPases. Rac inhibition of RhoA activity has previously been described (43, 45, 46) and down-regulation of RhoA by Rac has been specifically shown in response to growth factor-induced migration (42, 43, 45). It has also been recently demonstrated that the small GTPases RhoE and Gem compete for binding with RhoA for downstream effectors including ROCK (47, 48) and that the expression of RhoE stimulates migration (49). Thus, GM-CSF-stimulated motility may be modulated by other GTPases or other Rho isoforms and this will require further investigation.
Previous studies have shown that expression of RhoA seems to be cell type-dependent. Sahai and Marshall have shown various types of motile tumor cells that do not express RhoA, and invasion was blocked by C3 transferase or Y-27632 in some cases but not others (11). Furthermore, other studies have shown dominant-negative constructs of Rho blocked migration in endothelial cells, but not in fibroblasts or epithelia (12). In agreement with these studies, our data demonstrate that eosinophils are robustly mobile in the presence of GM-CSF but do not show RhoA activation, nor is their motility blocked by C3 transferase. Conversely, while CCL11 stimulates RhoA activation in eosinophils, they are not very motile in response to it. It is probable that multiple signaling pathways are being activated by these two cytokines.
Recent studies have also shown that cells use the Rho pathway differently depending on if they are in a 2D or 3D environment. Several studies have now found that cells treated with Y-27632 behave differently on a 2D substrate than in a 3D environment. Sahai and Marshall have found Y-27632 blocked tumor cell motility on a 2D substrate, but not in 3D (11). Nakayama et al. also found that the effects of Y-27632 can vary by cell type and environment (12). Conversely in some cases, application of Y-27632 stimulates cell migration (42, 50) on a 2D substrate, but no stimulation by Y-27632 was observed in a 3D environment (50). Therefore, use of the Rho pathway is both cell context and environment context-dependent. Consistent with this, we found that inhibition of ROCK abolished CCL11-stimulated motility, but not motility stimulated by the maximal dose of GM-CSF.
In conclusion, the comparison of these two cellular pathways, one that induces robust motility of eosinophils and one that does not, provides us with an excellent tool to further dissect the signaling pathways regulating eosinophil migration. This may be particularly important as there is considerable interest in pharmacological modulation of the RhoA pathway in allergic disease and asthma. For example, the ROCK inhibitor, Y-27632, has been investigated in animal models as a modulator of eosinophilia and airway hyperresponsiveness (51, 52). Because we shown in human eosinophils that use of the Rho/ROCK pathway is cell context and environment context-dependent, this has important implications for the development of drugs targeting the migration of these cells within human tissue.
We thank Dr. Das Mutalithas, Caroline Palmqvist, and our research nurses Sarah Mellor, Sue McKenna, Bev Hargadon, and Maria Shelley for excellent and enthusiastic technical assistance.
The authors have no financial conflicts of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by a grant from Asthma U.K. to A.J.W.
Abbreviations used in this paper: 3D, three dimensional; 2D, two dimensional; LPA, lysophosphatydic acid; MMP, matrix metalloproteinase; ROCK, Rho-associated kinase.