The Chemokine CX3CL1 Reduces Migration and Increases Adhesion of Neurons with Mechanisms Dependent on the β₁ Integrin Subunit¹

The chemokine CX3CL1 and its specific receptor CX3CR1 are abundantly expressed in the CNS where they mediate microglia-neuron interaction under physiological and pathological conditions (1). In contrast with earlier studies, where the expression of CX3CL1 and CX3CR1 was considered to be the exclusive prerogative, respectively, of neuronal and glial cells (2, 3, 4, 5), it is now documented that, in different species and conditions, both neurons and glial cells express CX3CL1 and its receptor (6, 7, 8, 9, 10, 11). CX3CL1 is present on cells as a transmembrane molecule, which is transformed into the soluble form upon extracellular shedding, mediated by the activation of the constitutive or inducible metalloproteinases ADAM10 and ADAM17 (12, 13, 14). Several pieces of evidence describe that toxic insults and nerve injuries induce an increase of CX3CL1 expression and its release from neurons with microglial recruitment (2, 11, 13, 15, 16, 17). The cytokine milieu locally generated in different chronic and acute neuroinflammatory conditions may influence glial expression of CX3CL1 and CX3CR1 (10). The CX3CL1 released plays a direct neuroprotective role, reducing the neuronal damage caused by toxic insults (7, 11, 18, 19), and impairing IFN-γ- and LPS-induced microglial activation (5, 19). Although CX3CR1 is expressed both in neurons and glial cells, its chemotactic activity develops only in microglial cells (20, 21). In different cell types, it is reported that CX3CL1-mediated chemotaxis depends on receptor-activated signal transduction pathways (20, 22, 23), while the adhesive properties have been explained both by intrinsic adhesive function of the CX3CL1/CX3CR1 molecules and the activation of intracellular signaling (24, 25, 26, 27, 28). Another issue of debate is the role of integrins in mediating the adhesive properties of CX3CL1: it has been reported that CX3CL1 regulates integrin avidity, similarly to what has been shown for other chemokines (26), but conflicting data are reported (24, 25).

During development, chemokine expression is functional to correct localization of neural precursors to their final destination; in particular, CXCR4 is essential for proper neural migration in the hippocampus, cerebellum, neocortex (29, 30, 31), and in the spinal cord (32). Furthermore, it has been reported that CX3CL1, together with other chemokines, is able to modulate mesenchymal stem cell migration in the brain (33), where these cells could differentiate toward a neural phenotype (34, 35, 36). CX3CL1 is also expressed by embryonic and adult neural progenitor cells (37), and has trophic effects on neural precursors (38).

In this study, we compare the chemotactic effects of CX3CL1 on neurons and glial cells obtained from either the cerebellum or the hippocampus of newborn rats and report for the first time that soluble CX3CL1 specifically reduces basal neuronal motility and increases neuron adhesion to the extracellular substrate laminin. We demonstrate that these effects are both dependent on receptor signal transduction and are mediated by the activation of the integrin subunits β₁ and α₆, with minor differences observed between neurons obtained from different brain sources. Furthermore, we report that neuron cotreatment with CX3CL1 and CXCL12, a chemokine specific for the CXCR4 receptor, counteracts the effects of a single chemokine on chemotaxis and adhesion. The fact that both CX3CL1 and CXCL12 are constitutively expressed in the brain, together with the observation that both their expression and those of their receptors may increase upon neuroinflammation or traumatic injuries, strongly indicates that the simultaneous stimulation of cells with these chemokines may have both physiological and pathological implications.

Polyclonal Abs to rat (TP501) and human (TP502) CX3CR1 were obtained from Torrey Pines Biolabs; anti-phospho-Akt (Ser⁴⁷³), anti-Akt, anti-phospho-ERK1/2 (Thr²⁰²/Tyr²⁰⁴), and anti-ERK2 Abs were obtained from New England Biolabs. mAbs to α₆ were obtained from Beckman Coulter, anti-β₁ was obtained from Immunological Sciences, and anti-MHC class I was obtained from Serotec. Transwell cell culture inserts were obtained from BD Labware; pertussin toxin (PTX)⁴ and recombinant rat CX3CL1 were obtained from Calbiochem; all culture media were obtained from Invitrogen Life Technologies. LY294002 and PD98059 were obtained from Alexis Italia; laminin (from Engelbreth-Holm-Swarm, murine sarcoma) was obtained from Sigma-Aldrich.

Hippocampal neuronal cultures were prepared from 1- or 2-day-old (p1-p2) Sprague Dawley rats. In brief, after careful dissection from diencephalic structures, the meninges were removed and tissues were chopped and digested in 0.25% trypsin for 15 min at 37°C. Cells were dissociated and either immediately used for chemotaxis and adhesion assays or were plated at a density of 8 × 10⁵ in poly-l-lysine coated 35-mm dishes in MEM with Earl’s salts and GlutaMAX containing 10% dialyzed and heat-inactivated FBS, 100 μg/ml gentamicin, and 25 mM KCl, and maintained at 37°C in 5% CO₂. After 24 h, cytosine-β-d-arabinofuranoside was added at a final concentration of 5 μM to prevent glial cell proliferation. Cells were used for experiments after 7 days. Primary cultures of cerebellar granule cells were obtained from p7-p8 rats, as already described (39). The use of animals in this study follows the protocols approved by the institutional animal care and use committee of the University of Rome La Sapienza.

Mixed glial cell cultures were prepared from hippocampi and cerebella of newborn and p7 rats. In brief, the hippocampi and cerebella were isolated, mechanically dissociated, and plated at low density (3 × 10⁶ cells/90-mm dish) and cultured in basal, supplemented with 10% heat-inactivated FCS and 5 mM KCl, for 20 days. Once confluent, the cells were left for 5–7 days without medium changes to favor microglial proliferation. The mixed glial cells were then gently shaken, centrifuged, and the supernatant, containing an enriched population of microglia, was collected (40).

Cerebellar granule neurons (CGN) and hippocampal cultures, cultured for 7 days, were incubated for 2 h in Locke’s buffer and stimulated with CX3CL1 (100 nM) for different times, from 1 to 15 min. Corresponding cellular lysates were quantified for protein content and ∼20 μg of total proteins were analyzed by SDS-PAGE and Western blot with phospho-Akt and phospho-ERK1/2-specific Abs. Parallel blots loaded with the same samples were analyzed for total Akt and ERK2 content as further control of equal protein loading.

Chemotaxis assay was performed on neurons, microglia, and astrocytes, freshly obtained from rat hippocampi and cerebella. Cells were resuspended in serum-free medium and plated onto laminin- (neurons) or poly-l-lysine- (microglial cells) treated 12-mm transwells (3-μm pore size polycarbonate filters for neurons and 8-μm pore size for microglia and glia cells; 5 × 10⁵ cells/well). The lower chambers contained CX3CL1 alone or together with CXCL12 (at concentrations ranging from 0.05 to 100 nM), prepared in the same medium. For chemokinetic assay, CX3CL1 was present both in the lower and upper chambers. When necessary, cells were preincubated at 37°C for 2 h with PTX (1 μg/ml), anti-rat CX3CR1 (TP501, 3 μg/ml), rabbit preimmune Igs (3 μg/ml), or for 15 min with PD98059 (30 μM) and LY294002 (25 μM), or with blocking mAbs for β₁, α₆, or MHC class I (10 μg/ml, 30 min, on ice) that were also present in the upper chamber during the assay. In some experiments, lower doses of PD98059 (10 μM) and LY294002 (10 μM) were tested. The chambers were incubated for 2 h at 37°C in a moist 5% CO₂ atmosphere. After incubation, cells were treated with 10% trichloroacetic acid on ice for 10 min and the nonmigrating cells adhering to the upper face of the filters were scraped off, while cells on the lower side were stained with a solution containing 50% isopropanol, 1% formic acid, and 0.5% (w/v) brilliant blue R250 and dried on a glass slide. The number of migrating cells was counted in 20 fields with a ×63 objective. The results were expressed as the mean cell number ± SE and the chemotactic index was obtained by the ratio between chemokine-treated vs untreated cells.

For adhesion assay, freshly obtained neurons or microglial cells (3 × 10⁴ cells/well) were plated on laminin- (50 μg/ml) or poly-l-lysine- (100 μg/ml) coated 96-well plates, respectively. When necessary, cell suspension was preincubated with PTX (1 μg/ml, 2 h), TP501 (3 μg/ml, 2 h), PD98059 (30 μM, 15 min), or LY294002 (25 μM, 15 min) at 37°C, or with blocking Abs for β₁, α₆, or MHC class I (10 μg/ml, 30 min, on ice). Cells were seeded in serum-free medium containing 0.1% BSA and allowed to adhere for 30 min at 37°C. Cells were then treated with CXCL12 in the presence or in the absence of CX3CL1 (both 5 nM) or water, as control, for different times (from 2 to 20 min). Treatment with CX3CL1 was performed backwards from 20 to 2 min to stop the adhesion at the same moment in the whole plate. Inverting the plates pulled off nonadhering cells and, after two washes with PBS, adherent cells were treated with lysis buffer and nuclei were counted in a hemocytometer, as described (41). Nonspecific attachment to BSA alone (0.01% in PBS), ranging from 2 to 3 × 10⁴, was subtracted from values obtained for specific adhesion at each time point. Each assay was performed at least in triplicate. In preliminary experiments, the adhesion assay was performed by labeling cells with calcein AM (10 μM, 15 min; Molecular Probes) and by analyzing the fluorescence of adherent cells upon chemokine stimulation. Because similar results were obtained with the two methods, we decided to use the first one for practical reasons.

We decided to investigate the effect of CX3CL1 on neuronal migration because it has been demonstrated that only microglial cells, in the brain, respond to CX3CL1 in the classical chemotactic assay (2, 13, 20). For this purpose, hippocampal and CGN neurons, freshly dissociated from either p2 or p7 rats, respectively, were treated as described in Materials and Methods and analyzed for chemotaxis vs a chemokine gradient. The results reported in Fig. 1 show that CX3CL1 is not chemotactic for neurons, yet, interestingly, it induces a dose-dependent inhibition of neuronal cell movement, both on hippocampal- (A) and cerebellar- (B) derived cells, with maximal effect obtained at 0.1 nM CX3CL1 for both cell types. Inhibition of cell movement was observed both with poly-l-lysine and laminin used as substrate, with similar dose dependence. All successive chemotaxis experiments were performed on laminin, it being the physiological ligand for neuronal-expressed integrins. To verify whether the results obtained could reflect a change in chemokinetic activity, CX3CL1 was simultaneously put in both the upper and lower chambers to eliminate the gradient. Under these conditions, no significant differences in cell movement were observed for both cell types (Table I, n = 5).

The inhibitory effect on cell movement was specific for neurons, because when microglial cells obtained either from hippocampus or cerebellum were tested for chemotaxis in response to CX3CL1 (3 nM), we observed cell migration (Fig. 2, right), while mixed glial populations–mainly containing astrocytes, obtained from the same brain regions–were unresponsive to CX3CL1 (Fig. 2, left).

To verify whether the inhibitory effect of CX3CL1 on neuronal migration was mediated by chemokine interaction with its specific receptor CX3CR1, the same experiments were performed in the presence of the blocking Ab specific for rat CX3CR1, TP501. When hippocampal (Fig. 3,A) or cerebellar (Fig. 3,B) neurons were preincubated with TP501 (3 μg/ml) for 2 h, the effect of CX3CL1 (0.1 nM) on motility was completely abolished; similar neuron treatment with preimmune rabbit Igs (3 μg/ml, 2 h pretreatment) failed to abolish the CX3CL1-mediated effect (data not shown). When both kinds of neurons were preincubated with PTX (1 μg/ml, 2 h) before chemokine treatment, a block of the CX3CL1 effect was observed, indicative of receptor coupling to G_i protein subtypes in both cell types, as already shown in hippocampal neurons (Refs. 7 and 11 , Fig. 3). Because it has been reported that CX3CL1 induces the activation and the phosphorylation of ERK1/2 and of the PI3K substrate Akt in neurons (6, 7, 11) and we have confirmed these data in our cellular systems (11 , Fig. 3), we also tested the involvement of the ERK1/2 and PI3K signaling on the CX3CL1-mediated effect, using their specific pharmacological inhibitors, PD98059 (30 μM, 15 min) and LY294002 (25 μM, 15 min). In previous studies, we already provided evidence that in neuronal primary cultures, these drugs, at the indicated doses, are fully effective in inhibiting, respectively, ERK1/2 and Akt phosphorylation (42, 43). Besides, even if we were aware that these doses were far above the reported IC₅₀ (44), no toxic effects were observed for LY294002 and PD98059 at the time points indicated in several experimental paradigms (11, 43, 45). The results shown in Fig. 3 indicate that the inhibitory effect on migration of hippocampal neurons was independent of ERK1/2 and dependent on the PI3K pathway, while both pathways were involved in the modulation of CGN migration. When, in some experiments, lower doses of PD98059 (10 μM, 15 min) and LY294002 (10 μM, 15 min) were used, we obtained results similar to those reported above (data not shown). Nevertheless, in successive experiments, we preferred to use the higher doses already tested for their efficacy in the inhibition of the specific kinase activities (11, 43, 45).

Both cerebellar and hippocampal neurons express functional CXCR4 receptors and migrate in response to their specific stimulation (29, 43, 46, 47 ; Fig. 4). Because we demonstrated that CX3CR1 activation in these cells reduced cell motility, we analyzed the effect of the simultaneous stimulation of CX3CR1 and CXCR4 receptors in these neurons. The results reported in Fig. 4 demonstrate that, in the presence of CX3CL1, CXCL12 was still able to induce chemotaxis (p = 0.039 and p = 0.0004, respectively, for hippocampal and cerebellar neurons) and, in the presence of CXCL12, CX3CL1 was still able to reduce cell movement (p = 0.025 and 0.00027, respectively, for hippocampal and cerebellar neurons). Nevertheless, the cotreatment with the two chemokines CX3CL1/CXCL12 did not produce any change in cell migration rate compared with untreated neurons, thus abolishing the single chemokine effects reported above.

We next addressed the experiments to see whether the different effects of CX3CL1 on the migration of neuronal and glial cells could be explained by modulation of cell adhesion to molecules of the extracellular matrix. For this reason, experiments of cell adhesion were performed on laminin substrates for neurons, and on poly-l-lysine for microglial cells. For microglial cells, poly-l-lysine was preferred to laminin, because it has been reported that microglial cells require the presence of cytokines for adhesion to laminin (48). The results obtained on neurons are shown in Fig. 5 and indicate that CX3CL1 stimulates a time-dependent increase of hippocampal (A) and cerebellar (B) neuron adhesion to laminin; in each set of experiments, a control with vehicle alone was included at the longer time point and the number of adhering cells was not significantly different from those indicated at time 0. Microglial cells adhesion was reduced by 10-min CX3CL1 treatment (5 nM) to 53 ± 10% of control (n = 4 triplicate experiments). The increased adhesion of neurons was drastically reduced or abolished by TP501 or PTX treatment (Fig. 6, left panels). Preimmune rabbit Igs (3 μg/ml, 2 h pretreatment), used as control for TP501 specificity, were ineffective (data not shown). The CX3CL1-mediated increase of cell adhesion was also completely abolished by PD98059 and LY294002 treatment (Fig. 6, right panels), indicating the involvement of the ERK and PI3K pathways in CX3CL1-mediated neuronal cell adhesion.

To investigate the specificity of the effect of CX3CL1 on cell-laminin interaction, the hippocampal and cerebellar neurons were pretreated with a blocking anti-integrin β₁ mAb, because β₁ integrins are the main targets of laminin together with α₆β₄. The laminin we used for our assay contains as major constituent laminin-1, which is preferentially recognized by α₆β₁ integrin (49, 50). The results shown in Fig. 7 indicate that when hippocampal (A) and cerebellar (B) neurons were preincubated with anti-β₁ mAb (30 min, on ice), the effect of CX3CL1 on cell adhesion was completely prevented. A similar drastic block or reduction of cell adhesion was observed, respectively, on the hippocampal and cerebellar neurons, with an anti-α₆ mAb (Fig. 7). As control, an anti-MHC class I mAb was used under the same experimental conditions: the results shown in Fig. 7 demonstrate that this mAb was ineffective in reducing CX3CL1-mediated cell adhesion, confirming the specificity of anti-β₁ and anti-α₆ blocking mAbs.

When the same mAbs were used to investigate the role of integrin subunits in mediating the CX3CL1 inhibitory effect on cell movement, we observed a similar complete block by anti-β₁ mAb on neurons from the hippocampus (Fig. 8,A) and cerebellum (Fig. 8,B) while anti-α₆ mAb was slightly inhibitory for hippocampal neuron movement and without effect on cerebellar neurons. MHC class I blocking mAb was again ineffective on CX3CL1-mediated reduction of movement on both neuronal cell types (Fig. 8).

Because we observed mutual inhibitory effects on neuronal movement between CXCL12 and CX3CL1, we wondered whether analogous effects were obtained for cell adhesion. For this reason, we decided to investigate whether CXCL12/CX3CL1 cotreatment could modulate CX3CL1-mediated cell adhesion. The results shown in Fig. 9 indicate that CXCL12 did not increase neuron adhesion to laminin for the time points analyzed (from 2 to 20 min), a slight increase being observed for CGN only at 20 min. Interestingly, however, neuron cotreatment with CXCL12/CX3CL1 blocked CX3CL1-mediated cell adhesion at all time points analyzed. Comparable results were obtained on hippocampal and cerebellar neurons (Fig. 9).

The level of CX3CL1 in the CNS has been often found increased in association with traumatic and/or pathological events and it is considered responsible for microglial recruitment to the injured sites (2) and neuronal protection from toxic stimuli (7, 11, 18, 19). In this respect, several reports have described the chemotactic activity of neuron-released CX3CL1 on microglial cells and the increased expression of CX3CL1 and CX3CR1 on astrocytes only after their treatment with inflammatory cytokines (21). In contrast, data of CX3CL1 effects on neuron movement are completely lacking. The movement of neural precursors may have physiological reasons during development and repairing functions upon pathological insults; in this respect, many recent studies have focused on the role of chemokines in regulating the correct migration of neural progenitors both in the developing brain (1, 29, 31, 51) and to sites of neuroinflammation (52). In this study, we show that CX3CL1 specifically reduces neuron movement, being chemotactic for microglial and ineffective on astroglial cells, regardless of the brain sources. This effect on neurons is already maximal at CX3CL1 concentrations of 0.1 nM, a dose that is compatible with other described physiological effects (11).

In addition, we show that CX3CL1 reduces the movement of neurons and increases their adhesion to the extracellular substrate laminin with mechanisms that require the β₁ and in part also the α₆ integrins, the PTX-dependent G proteins, and the activation of the PI3K and ERK pathways. The involvement of intracellular signaling in mediating CX3CL1 inhibition of neuron movement and increase of adhesion indicates that these effects are not caused by CX3CL1/CX3CR1 acting as adhesion molecules, as described in different cell types (24, 25). In particular, the contribution of the ERK1/2 and PI3K pathways to CX3CL1-induced neuron adhesion is comparable to that reported for human monocytes (27), even if the involvement of other more downstream targets has not yet been proved.

The role of α₆β₁ integrin in mediating neuronal cell migration and adhesion has been already partially investigated and it has been shown that both α₆ and β₁^null mice show abnormalities in the migration rate of specific neuron populations during development (53, 54); similar results are also reported with the inhibition of α₃β₁ integrin (55). However, the involvement of integrins in CX3CL1-mediated cell adhesion is somewhat controversial: the full-length membrane form of CX3CL1 is reported to mediate cell-cell adhesion by virtue of its adhesive properties, with mechanisms independent of receptor activation or modulation of integrin avidity for the extracellular matrix substrates (22, 24, 25), but dependent on integrins and receptor signal transduction in different cell types (26, 28). The soluble chemokine domain of CX3CL1 is chemotactic on leukocytes (28, 56, 57), but is also reported to exert adhesive properties on monocytes (27, 58). The general view is that both soluble and membrane-anchored forms of CX3CL1 may have chemotactic action on most leukocytes but exert adhesive properties on monocytes; the reasons for these differences are not clear, but it has been hypothesized that CX3CR1 couples to different G proteins in different cell types (59). This scenario is quite similar to what we describe in the nervous system, where CX3CL1 specifically increases neuron adhesion to the extracellular substrate laminin, being chemotactic for microglial cells, indicating that the effect of CX3CL1 is cell-type specific in different tissues.

It is interesting to note that, at least in the two neuronal systems we used in this study, the simultaneous stimulation of the CX3CR1 and CXCR4 receptors has no net effects with respect to cell adhesion and migration (see Figs. 4 and 9), as if receptor costimulation would result in their mutual functional inactivation, likely due to their reciprocal neutralization. Nevertheless, we cannot exclude that the block of neuron adhesion and migration we observe upon CX3CR1/CXCR4 costimulation is indirectly mediated through the modulation of multiple signal transduction pathways. Among them, of particular interest, are the ERK1/2 and PI3K/Akt pathways, which are activated by both chemokines in neurons (Refs. 6 , 7 , and 43 and this study) and the small G protein rac or the MAPK p38, known to be involved in the regulation of cell movement and adhesion.

Our observation that CX3CL1/CXCL12 cotreatment abolished the net chemotactic activity of CXCL12 on neurons is analogous to the CX3CL1-mediated reduction of MCP-1-stimulated chemotaxis in monocytes (58). These data may have physiological implications, because it is known that CXCL12 has important roles in regulating neural progenitor migration during CNS development (29, 30, 31) and in the peripheral nervous system (32) and we speculate that the simultaneous presence of CX3CL1, or temporally regulated CX3CL1 expression, may contribute to this modulation and is worthy of further investigation.

The authors have no financial conflict of interest.