The chemokine SDF-1 differentially regulates axonal elongation and branching in hippocampal neurons

The development of neuronal circuits is supported by a fine spatial and temporal orchestration of several cellular events, including cell migration, process formation and synaptogenesis. The generation of elaborate dendritic trees and axonal arborization is controlled by a repertoire of extracellular molecules acting in a coordinated manner on guidance, elongation and branching. Accumulating evidence indicates that molecules involved in axon guidance are also implicated in process growth and branching. For instance, semaphorin 3A known to repel axons in the cortex (Bagnard et al., 1998; Polleux et al., 1998; Polleux et al., 2000) has been shown to reduce axon branching (Bagnard et al., 1998; Bagri et al., 2003). Similarly, Slit1, which acts as a chemorepellant for cortical axons, promotes the growth and branching of both axons and dendrites (Nguyen Ba-Charvet et al., 1999; Wang et al., 1999; Whitford et al., 2002). In the same way, recent data suggest that the chemokine SDF-1 is a bifunctional factor regulating both axon pathfinding and elongation (Xiang et al., 2002; Arakawa et al., 2003; Chalasani et al., 2003). SDF-1, originally isolated in the immune system for its role in leukocyte chemoattraction (Tashiro et al., 1993; Nagasawa et al., 1996), has subsequently been shown to play important functions in many organs including the haematopoietic system, cardiovascular system and brain (Nagasawa et al., 1996; Tachibana et al., 1998; Zou et al., 1998). A particularity of SDF-1 among the large family of chemokines is that it binds to only one receptor, CXCR4, a G protein-coupled receptor for which SDF-1 represents to date the only known ligand. Early evidence of the central role of SDF-1 in brain development came from the striking abnormality of the cerebellum of knockout mice lacking either SDF-1 or its receptor CXCR4 (Ma et al., 1998; Zou et al., 1998). Further examination of these mouse lines indicated additional defects in other brain regions, such as the hippocampus and the neocortex (Lu et al., 2002; Stumm et al., 2003). The reported alterations in the brain architecture could originate from dysfunctions in cell migration, axonal guidance or process formation. All of these cellular events are regulated by SDF-1. Thus, SDF-1 has been shown to be involved in migration of hippocampal, cerebellar and cortical neurons (Bagri et al., 2002; Zhu et al., 2002; Stumm et al., 2003), axon guidance (Lu et al., 2001; Xiang et al., 2002; Chalasani et al., 2003) and axon elongation (Arakawa et al., 2003). However, the question whether SDF-1 could differentially influence the development of the axon and the other neurites remains open. In addition, the possibility that SDF-1 could selectively affect elongation versus branching is still unexplored. Moreover, there is no data on the cellular distribution of CXCR4 in developing neurons although this would represent an important step in elucidating the potential functions of SDF-1 at different stages of neuron development. We thus studied the distribution of CXCR4 during the development of cultured hippocampal neurons by immunofluorescence, and tested the effect of SDF-1 in the patterning of the axon and the other neurites. We used a low-density hippocampal neuron culture system because the axon and the other neurites emerge simultaneously from the cell body allowing us to compare the effects of SDF-1 on these two classes of processes. We found that CXCR4 immunoreactivity is strikingly redistributed during development from the tip of nerve processes at early stages to a more widespread localization along the processes, predominantly along axons, as the cells mature. In addition, our data showed that SDF-1 regulates axonal elongation and branching, without affecting the other neurites. This study defines SDF-1 as a new extracellular signal involved in the axonal patterning of hippocampal neurons, which suggests that this chemokine might play a critical role on neuronal connectivity during development.

Rat hippocampal cultures were prepared from 18-day-old rat embryos by previously described methods (Goslin et al., 1998). All the protocols were carried out in accordance with French standard ethical guidelines for laboratory animals (Agreement number 75-669). Briefly, hippocampi were dissected and dissociated by trypsin and trituration and were plated at low density (4000 cells/cm²) on glass coverslips coated with poly-L-lysine in MEM with 10% calf serum. After cells were allowed to attach on coverslips, they were transferred to a dish containing a glial feeder layer cultured from newborn rat forebrain and were maintained for up to 2 weeks in serum-free MEM with N2 supplements. The neurons were used at 1, 2, 4, 7, 10 and 15 days after plating for immunocytochemical staining and morphology analysis.

For pharmacological treatments, neurons were incubated at day 1 with 50 nM SDF-1 (AbCys) or 10 μM bicyclam (custom synthesized by Orgalink, Gif sur Yvette, France), or both, or corresponding vehicle (control cells) for 16 hours and were fixed at day 2 in 4% paraformaldehyde, 4% sucrose in phosphate-buffered saline (PFA) for 15 minutes at room temperature for quantification analysis. Transfection experiments of green fluorescent protein (GFP) and CXCR4-GFP were performed at plating with either GFP or CXCR4-GFP expression vectors using the Lipofectamine 2000 reagents (Invitrogen) essentially according to the manufacturer's instructions. The CXCR4-GFP expression plasmid was a gift of M. Alizon (Institut Cochin, Paris, France). Transfected neurons were plated at 14,000 cells/cm² and processed as for untransfected cells. When neurons were co-cultured with COS cells, SDF-1-transfected COS cells were plated at a density of 10,000 cells/cm² on day 1 cultured neurons previously transfected with CXCR4-GFP and were placed in the incubator for 16-20 hours without the glial feeder layer. Cells were fixed in PFA at day 2 and immunostained for SDF-1.

For western blot analysis of cultured neurons, ∼500,000 cells at day 7-10 were scraped into PBS, pelleted and resuspended in Laemmli buffer. Proteins were separated by SDS-PAGE (12% acrylamide), transferred to a nitrocellulose membrane, blocked in 20 mM Tris-HCl, pH 7.4 containing 0.45 M NaCl, 0.1% Tween 20 (TBST) and 10% dried milk. The membranes were incubated overnight in polyclonal goat anti-CXCR4 antibody (1:2000; Santa Cruz Biotechnology) diluted in TBST containing 8% dehydrated milk, washed with TBST, incubated for 1 hour in HRP-conjugated secondary goat anti-mouse antibody (1:20,000; Jackson ImmunoResearch) and visualized using chemiluminescent substrate (Amersham) and exposure to X-ray film. For western blot analysis of rat brain homogenate, rat forebrains were homogenized using 10 strokes with a motor driven Dounce homogenizer in 10 mM Tris-HCl, pH 7.4 containing 320 mM sucrose, 5 mM EDTA and a cocktail of protease inhibitors. The resulting suspension was centrifuged at 3500 g for 10 minutes. The supernatant was collected, centrifuged at 20,000 g for 20 minutes and the pellet was resuspended in 10 mM Tris-HCl containing 5 mM EDTA. The membrane preparation was solubilized in Laemmli buffer and samples were resolved by SDS-PAGE, and processed as described above.

Neurons were fixed in PFA for 20 minutes and permeabilized for 5 minutes in 0.25% Triton X-100. The cells were blocked for 30 minutes in 10% bovine serum albumin (BSA) in PBS and incubated for 2 hours at 37°C in primary antibodies diluted in PBS containing 3% BSA. The primary antibodies used were as follows: mouse monoclonal anti-MAP2 antibody clone AP20 shown to react with MAP2-A, MAP2-B but not with MAP2-C or MAP1 (1:400; Sigma), rabbit anti-synaptophysin (1:1000, Sigma), mouse anti-SV2 (1:50; Developmental Studies Hybridoma Bank), goat anti-CXCR4 (1:100; Santa Cruz Biotechnology) and goat anti-SDF-1 (1:100; Santa Cruz Biotechnology). Specificity of CXCR4 and SDF-1 antibodies has previously been characterized in non-neuronal as well as neuronal cells (Hasegawa et al., 2001; Banisadr et al., 2002; Banisadr et al., 2003). Cells were then incubated with appropriate Texas Red- or FITC-conjugated secondary antibodies (Jackson Laboratories), and the coverslips were mounted on glass slides for analysis on a BX 61 microscope (Olympus). Pictures were acquired with a ×63/1.4 objective using a digital camera (DP 50, Olympus) driven by Analysis image-acquisition software. For experiments involving transfection, each coverslip was systematically scanned with ×25 lens and images of each transfected cell acquired with a ×63/1.4 objective. Images for presentation were prepared for printing with Adobe Photoshop.

Neurons at day 10 were incubated at 37°C for 1 hour with 5 μM Fura-2/AM in PBS pH 7.4 supplemented with 0.8 mM MgCl₂, 1.3 mM CaCl₂, 20 mM HEPES, 5 mM glucose and 0.2% pluronic F-127. Before analysis, the coverslips were inserted into a temperature-controlled chamber at 35°C and examined with an inverted epifluorescence microscope (Nikon, Japan). Images were captured every 2 seconds and the ratio of fluorescence intensities at 340 and 380 nm were recorded using the Ca²⁺ imaging system Quanticell 700 (Visitech, UK).

COS-7 cells were grown in DMEM containing 10% calf serum. Transfections with CXCR4-GFP and SDF-1 cDNAs were performed using the Lipofectamine kit (Qiagen). Cells transfected with CXCR4-GFP were used for internalization experiments. After 24 hours of transfection, cells were incubated for 2 hours with or without 100 nM SDF-1, fixed in PFA and the distribution of CXCR4-GFP was examined. To evaluate the distance of diffusion of SDF-1 released from a transfected cell, SDF-1 and CXCR4-GFP were expressed separately in COS cells and subsequently co-cultured. The pattern of CXCR4-GFP was analysed in correlation with the distance separating the CXCR4-GFP-expressing cells from the SDF-1-transfected cells. We defined two groups of CXCR4-GFP-transfected cells based on the occurrence of receptor internalization (see Results). Cells exhibiting a pattern of diffuse CXCR4-GFP signal were classified as the control group and were consistently found at least 50 μm from a SDF-1-expressing COS cell. Cells exhibiting a CXCR4-GFP internalization pattern characterized by the presence of intracellular clusters were classified as the `SDF-1-stimulated' group and were consistently found at a distance between 0 and 5 μm from an SDF-1-transfected COS cell. CXCR4-GFP-transfected cells found between 5 and 50 μm showed variation in receptor pattern distribution and were excluded from the analysis. For neuron and COS cell co-culture experiments, COS cells were transfected with SDF-1, collected 24 hours later and plated over a day 1 neuronal culture previously transfected with CXCR4-GFP. Cells were fixed 16 hours later with PFA and the immunostaining procedure for SDF-1 was performed essentially as described above.

Phase-contrast images of 30-40 neurons per coverslip were randomly acquired with a microscope as described above. Two coverslips per condition were analysed from two to three independent experiments. Images projected on the computer screen were traced with a mouse using the Analysis software program and number and length of processes, number of primary branch points, number of growth cones, area of growth cones and cell bodies were scored. For analysis of neurons transfected with CXCR4-GFP and co-cultured with SDF-1-expressing COS cells, the former were classified into two groups depending on the distance separating the transfected neuron from an SDF-1-transfected COS cell (see above). For a distance between 0 and 5 μm, neurons were classified as the `SDF-1 stimulated' group, and for a distance above 50 μm, neurons were pooled into the control group. Neurons between 5 μm and 50 μm from a COS cell were not taken into account. For both groups, length and number of primary processes as well as number of growth cones per process were scored as described above.

Primary cultures of hippocampal neurons were grown at low density to allow the study of CXCR4 cellular distribution in individual neurons and to perform subsequent morphometric analysis. CXCR4 protein expression in this neuronal model was first examined by western blot analysis. The anti-CXCR4 antibody revealed a band at 45 kDa, as observed for rat forebrain homogenate (Fig. 1), and was consistent with the reported CXCR4 molecular mass (Stumm et al., 2002). To study the cellular distribution of CXCR4, neurons cultured for 10-15 days were immunostained by double labelling with the CXCR4 antibody and the microtubule-associated protein 2 (MAP2) antibody, shown to be a dendritic marker in hippocampal neurons as early as day 4 of culture (Mandell and Banker, 1995). CXCR4 immunoreactivity, detected in about 50% of neurons, was mainly observed along MAP2-negative axons and, to a lesser extent, within MAP2-positive soma and dendrites (Fig. 2). Not all processes were immunostained, but a subset of processes originating from one neuron was, suggesting a preferential targeting of CXCR4 to specific branches. Within nerve cell bodies, CXCR4 immunoreactivity predominated at the periphery of the perikarya, sparing the nucleus and most of the cytoplasm. Along the axonal and somatodendritic domains, CXCR4 immunoreactivity was not distributed uniformly, but was present in a pattern of stretches and large puncta (Fig. 2A insets). To determine whether CXCR4 clusters were associated with synapses, neurons were simultaneously labelled for CXCR4 and the synaptic marker synaptophysin. There was no overlap between synaptophysin and CXCR4 clusters, indicating that CXCR4 clusters were localized at extrasynaptic sites (Fig. 2B).

To determine whether CXCR4 was expressed as functional receptors on cultured hippocampal neurons, intracellular Ca²⁺ mobilization was measured by fluorescent Ca²⁺ imaging on live neurons following CXCR4 activation by 20 nM SDF-1. It is important to note that these experiments were carried out in the absence of the glial feeder layer, thus allowing us to measure direct effects of SDF-1 on hippocampal neurons. Indeed, the low-density culture system we used represents a virtually pure neuron culture model (Goslin et al., 1998). About 70% of hippocampal neurons gave rise to an [Ca²⁺]_i increase upon SDF-1 stimulation both in soma and processes (Fig. 3). The increase of [Ca²⁺]_i was detected within 3 seconds on average after SDF-1 exposure, peaked between 7 and 12 seconds, and usually recovered to near basal levels within 1 minute. As CXCR4 immunoreactivity was distributed in both dendrites and axons, we examined whether the SDF-1-induced [Ca²⁺]_i increase was detectable in both processes. Based on different morphological features between axons and dendrites, we distinguished the axons as thin processes often far removed from any cell body and the dendrites as much thicker processes generally forming an elaborate tree around a cell body. SDF-1 stimulation triggered [Ca²⁺]_i rises in both dendrites and axons within the same time frame (Fig. 3). Taken together, these data suggest that CXCR4 was expressed on the neuronal cell membrane of somatodendritic and axonal domains, where it was functionally coupled to a second messenger cascade leading to an increase in [Ca²⁺]_i.

CXCR4 distribution was studied by immunofluorescence during development of cultured hippocampal neurons. At day 2, CXCR4 immunoreactivity was strikingly concentrated at the tip of neuronal processes and branches, independently of cell-cell contact (Fig. 4A). Critical developmental stages of these low-density hippocampal cultures have been well characterized (Dotti et al., 1988) and day 2 has been defined as a selective axon elongation step. Thus, according to published work (Dotti et al., 1988), we defined the longer process as the growing axon and the remaining short processes as minor neurites. CXCR4 accumulation was equally observed at the endings of growing axons and minor neurites and showed no correlation with the presence of growth cones. Some labelling was also seen throughout some processes but was mostly distributed in patches. Nerve cell bodies also exhibited CXCR4 immunostaining localized both at the periphery and throughout the cytoplasm. Day 4 is characterized by continued axonal growth and elongation of the remaining minor processes that will acquire dendritic morphological features. CXCR4 distribution at this stage of culture was similar to that observed at day 2, consisting of a further specific enrichment of CXCR4 immunoreactivity at the growing regions of both dendrites and axons, which, at this stage, often contacted neighbouring neurons (Fig. 4B). By day 7, which corresponds to a stage of continued maturation of axonal and dendritic arborization and synaptogenesis, the pattern of CXCR4 distribution resembled that observed at day 10-15. CXCR4 immunoreactivity was no longer enriched at growing tips, but was more widely distributed along neuronal processes (Fig. 4C). The stronger CXCR4 immunoreactivity in growing regions of the processes observed at early developmental stages could result either from a preferential receptor targeting to these sites or from differences in CXCR4 immunoreactivity towards the antibodies at and out of the growing regions. For instance, biochemical modifications or protein-protein interactions taking place at some specific sites could impede the recognition of CXCR4 receptors by the antibodies. To discriminate between these two possibilities, we analysed the distribution of CXCR4-GFP transfected into neurons, which was then directly visualized without the use of antibodies (Fig. 4D). As observed for endogenous immunolabelled CXCR4, transfected CXCR4-GFP was predominantly localized at the tips of the neurites from neurons cultured for 2 days. By contrast, the GFP signal resulting from the transfection of the parent GFP control vector was diffusely distributed throughout the cell with, however, a higher fluorescence intensity in the cell body probably due to the greater thickness of the soma compared with neuronal processes (Fig. 4E). These data further confirm that at early stages of neuronal development CXCR4 was selectively targeted to the growing regions of neurites.

Based on our finding that CXCR4 receptors were located at strategic sites to regulate neurite outgrowth, we examined whether the CXCR4 agonist SDF-1 could affect neuronal development. Neurons at day 1 were incubated for 16 hours with 50 nM SDF-1, fixed and phase-contrast images of randomly chosen neurons were acquired for quantification of several morphological parameters using a computer-assisted image analysis system. Neurite morphology of SDF-1-treated neurons showed numerous alterations compared to control cells (Fig. 5). The number of primary processes originating from the neuronal soma was slightly increased in SDF-1-treated cells (Fig. 5F). Among the primary processes, we distinguished between the axon, defined as the longest process and the other neurites, which will differentiate into dendrites later on in development (Dotti et al., 1988). Treatment of neurons with the CXCR4 agonist induced a 32% decrease in the axonal length of SDF-1-treated cells compared to control cells, whereas the length of the other neurites remained unchanged (Fig. 5C). The number of branch points per 100 μm of process was selectively increased by 45% along the axon of SDF-1-treated cells compared to control cells, whereas no difference was observed along the other neurites (Fig. 5G). As another index of neuronal development, we quantified the number of growth cones per primary process. Neurons incubated with SDF-1 showed a lower number of growth cones per primary process than that observed in control neurons (Fig. 5H). However, no change either in the growth cone area nor in the soma area were detectable after treatment with SDF-1 (Fig. 5D,E). To confirm that the modifications in neuron morphology induced by SDF-1 treatment were mediated through CXCR4, neurons were treated with 50 nM SDF-1 for 16 hours in the presence of 10 μM bicyclam, a CXCR4 antagonist (Hatse et al., 2002). In the presence of bicyclam alone or mixed with SDF-1, the axonal length, the number of axonal branch points and the number of growth cones per process were all similar to values in the control, suggesting that the effects of SDF-1 on neuronal morphology were indeed mediated through SDF-1 binding to CXCR4 (Fig. 5C,G,H). Importantly, neuronal survival as determined by the density of untransfected as well as CXCR4-GFP-transfected neurons was unaffected by SDF-1 treatment. Thus, in neurons at day 1-2 corresponding to a stage of axonal elongation in this hippocampal culture model, the chemokine SDF-1 affected axonal development by decreasing the elongation of the primary axon, and by promoting the emergence of axonal branches. Moreover, these effects were associated with a reduction in the number of growth cones at the tip of the neurites.

To confirm the regulatory role of SDF-1 in axonal development, we examined whether secreted, instead of exogenously added SDF-1 could act as an extracellular factor controlling axon patterning through CXCR4. The experiments consisted of co-culturing non-neuronal cells expressing high levels of SDF-1 with neurons transfected with CXCR4-GFP to analyse the morphology of the transfected neurons in the vicinity of SDF-1-expressing cells. We first ensured that the GFP-tagged receptor was functional as assessed by Ca²⁺ imaging (data not shown). Although several cell systems transfected with SDF-1 cDNA were shown to produce and release SDF-1 in a biologically active form, we performed initial experiments to demonstrate it in our system. We used the reported property of CXCR4 to internalize upon agonist activation as a marker of SDF-1 activity (Amara et al., 1997; Signoret et al., 1998; Orsini et al., 1999). SDF-1 and CXCR4-GFP were expressed separately in COS-7 cells and subsequently co-cultured. A typical endosomal pattern of internalized CXCR4-GFP receptor was noticed in cells located in close proximity to an SDF-1-transfected COS cell (Fig. 6C). This pattern was similar to that observed in CXCR4-GFP-transfected COS cells incubated for 2 hours with 100 nM SDF-1 (Fig. 6B), conditions previously reported to induce receptor internalization (Amara et al., 1997; Signoret et al., 1998; Orsini et al., 1999). By contrast, in cells located at a greater distance from an SDF-1-expressing cell, CXCR4-GFP was homogeneously distributed throughout the cell (Fig. 6C), as observed in control untreated CXCR4-GFP-expressing cells (Fig. 6A). This suggests that only CXCR4-GFP-expressing cells that were located close to SDF-1-transfected COS cells were activated by secreted SDF-1. We measured the distances between SDF-1- and CXCR4-transfected cells and consistently found a receptor internalization pattern for a distance less than 5 μm, and conversely, a control pattern for a distance greater than 50 μm. Based on these observations, CXCR4-GFP-expressing neurons co-cultured with SDF-1-transfected COS cells were divided into two groups: the SDF-1-stimulated group corresponding to neurons located up to 5 μm from SDF-1-expressing cells and the control group corresponding to neurons located at a distance greater than 50 μm from SDF-1-transfected COS cells (Fig. 6D,E). Neurons 5 to 50 μm distant from a SDF-1-transfected cell were ignored for this analysis. Measurements of the number of primary processes, their length, number of branching points and the density of growth cones were recorded for the two groups (Fig. 7). Although the number of primary processes originating from the transfected cell body was greater in the SDF-1-stimulated group than in the control group, no statistical differences were observed between the two groups (Fig. 7A). The average axonal length was lower for the SDF-1-stimulated groups than for the control group (Fig. 7B). As shown for non-transfected neurons, the reduction in process length was unique to axons and was not observed for the other neurites. The number of branch points per 100 μm of process was increased more than twofold along the axon of SDF-1-stimulated cells compared to control cells, whereas no difference was observed along the other neurites (Fig. 7C). In addition, a marked reduction in the number of growth cones per process was seen in the SDF-1-stimulated group compared to that in the control group (Fig. 7D).

This study reveals several novel features on the localization and function of the chemokine SDF-1 and its receptor CXCR4 in cultured hippocampal neurons. First, CXCR4 immunoreactivity exhibited a striking developmental change in its cellular distribution. Early in development, CXCR4 was particularly concentrated at the leading edge of growing neurites. As the cells matured, staining at the tip of the processes declined and became more broadly distributed along axons and, to a lesser extent, dendrites. Ca²⁺ imaging upon SDF-1 stimulation indicated that both the axonal and somatodendritic CXCR4 responded to SDF-1 and were thus both membrane-associated and functionally coupled to the intracellular signalling cascade. In accordance with the localization of CXCR4 at the growing region of neurites, we showed that SDF-1 triggered several effects on neuronal morphogenesis. At day 2, SDF-1 reduced growth cone number and axonal outgrowth but stimulated axonal branching. These latter two effects were not observed within the other neurites. This study unravels a new role for SDF-1/CXCR4 in specifying hippocampal neuron morphology by regulating axonal patterning at early stage of neuronal development.

We found a marked modification in the pattern of CXCR4 distribution during development of hippocampal neurons in culture. Between day 4 and day 7, CXCR4 redistributed from the tip of neuronal processes to a broader localization along axons and, to a lesser extent, along the somatodendritic domain. This switch in receptor distribution occurred after the initial period of axonal and dendritic differentiation, but preceded the synaptogenesis and the extensive growth and branching of nerve processes generating the elaborate arborization of mature neurons. Previous studies on the developmental profile of the expression of cerebral CXCR4 and SDF-1 proteins and mRNAs have shown a particularly high expression during the embryonic stage and at the early postnatal period followed by a progressive decrease thereafter until adulthood (Westmoreland et al., 1998; McGrath et al., 1999; Tham et al., 2001). Differences in the level of expression of CXCR4 depending on the brain area have also been reported during ontogeny, with high levels in the ventricular zone of cell proliferation at the embryonic stage followed at the early postnatal stage by a progressive increase in other brain areas such as the hippocampus, cerebral cortex, thalamus and cerebellum (Westmoreland et al., 1998; Tissir et al., 2004). Our results extend these previous data by showing the existence at the cellular level of a differential localization of CXCR4 in hippocampal neurons during development. This modification in the compartmentalization of CXCR4 suggests that this chemokine receptor may play different functional roles within the same neuron over the course of its development. Moreover, this observation implies that there are different cellular mechanisms governing CXCR4 targeting in neurons depending on the stage of neuron development. One of the factors responsible for the redistribution of CXCR4 could be the level of SDF-1 stimulation. There are a large number of studies demonstrating the activity dependency of the cellular localization of several receptors, including synaptic (Craig, 1998; Bredt and Nicoll, 2003) and non-synaptic (Boudin et al., 2000; Dumartin et al., 2000; Csaba et al., 2001) receptors. As the expression of neuronal and glial SDF-1 was strongly regulated during development (Tham et al., 2001), variations in the intensity of CXCR4 stimulation by its endogenous ligand could be an important factor in the differential targeting of CXCR4 during development. Alternatively, selective interactions during development between CXCR4 and protein partners implicated in anchoring, transport and signalling could account for differential CXCR4 targeting.

The transient enrichment of CXCR4 at growing regions of neuronal processes in immature neurons led us to investigate the role of SDF-1/CXCR4 in neuronal development. We found that exogenous as well as cell-produced SDF-1 influenced neuronal development by acting on axonal patterning. These actions were probably mediated by neuronal CXCR4 and not by an indirect event through glial-released molecules. Although CXCR4 has been described in astrocytes (Dorf et al., 2000), microglia (Lavi et al., 1997; Albright et al., 1999) and neurons (Bajetto et al., 1999; Banisadr et al., 2002; Stumm et al., 2002), several lines of evidence suggest that SDF-1 acts directly on neuronal CXCR4 receptors to modulate axonal morphology rather than indirectly via glial cells. First, the concentration of CXCR4 receptors at growing regions of neuronal processes makes them the probable molecular targets of SDF-1 in mediating effects on neuronal development. Second, when hippocampal neurons were cultured in the absence of the glial feeder layer, the inhibitory action of SDF-1 on axonal outgrowth was still observed. Given that the low-density culture system we used represents a virtually pure neuron culture model (Goslin et al., 1998), our data suggest that the CXCR4 receptors mediating these effects were located on neurons. These observations do not exclude the additional contribution of glial CXCR4, but strongly suggest a role for the neuronal CXCR4 receptor in regulating neuronal morphology.

Our findings show that SDF-1 influences several parameters of hippocampal neuron development in different ways. Whereas the growth cone number and axon elongation were both reduced upon SDF-1 treatment, the number of processes emerging from the cell body and that of primary axonal branches were increased. Process elaboration involves a coordinated and complex repertoire of cellular systems. A highly controlled balance between elongation and branching is critical to achieve the ultimate shape of a neuron, which underlies its ability to connect to multiple targets properly. Elongation and branching of neurites may operate through similar mechanisms as reported for tumour necrosis factor-α, which affects both events through a RhoA-mediated pathway (Neumann et al., 2002), or through distinct albeit related mechanisms as reported for δ-catenin (Martinez et al., 2003). For the latter, inhibition of Src family kinases enables δ-catenin to promote neurite outgrowth whereas inhibition of the RhoA pathway associated with δ-catenin expression leads to the formation of branched secondary neurites. Alternatively, some molecules, such as netrin-1 and semaphorin-3A, independently affect elongation and branching events (Dent et al., 2004). Our results showing that SDF-1 reduces axon elongation but promotes axon branching suggest that different mechanisms underlie each of these effects. Interestingly, it has been shown that different effectors of Ras signalling play distinct roles in axonal patterning. Activated Raf-1 causes axon lengthening, whereas active Akt increases axon branching (Markus et al., 2002). Along these lines, it has been reported that SDF-1-induced CXCR4 stimulation elicits activation of the Ras-Akt pathway in several neuronal models (Floridi et al., 2003; Peng et al., 2004), whereas no study has reported the activation of Raf-1 by SDF-1. One can hypothesize that SDF-1-induced axonal branching might involve the Ras-Akt pathway and that of axon growth inhibition could involve the Rho/ROCK pathway as previously reported in cerebellar granule cells (Arakawa et al., 2003).

Our data and previous studies point to the critical role of SDF-1 and its receptor CXCR4 in several aspects of neuron and brain development. Concerning patterning of neuronal processes, SDF-1 has been previously shown to differentially affect axon elongation in cerebellar granule cells according to the concentration range (Arakawa et al., 2003). A low concentration of SDF-1 stimulated axon elongation whereas a higher concentration (from 30 to 125 nM) repressed axon formation. Our study extends these data by demonstrating that SDF-1 not only regulates axon elongation, but also regulates branching and that these effects were selectively observed with axons and not detected within the other neurites, at least with the SDF-1 concentration used in the present study. In addition, SDF-1 acts as a chemotactic molecule that influences axonal guidance (Chalasani et al., 2003) and migration of cerebellar, hippocampal and cortical neurons (Lu et al., 2001; Bagri et al., 2002; Lu et al., 2002; Zhu et al., 2002; Stumm et al., 2003). The ability of SDF-1 to affect multiple aspects of neuron development appears to be shared by other factors such as some members of the slit and semaphorin families (Polleux et al., 1998; Nguyen Ba-Charvet et al., 1999; Wang et al., 1999; Whitford et al., 2002; Bagri et al., 2003).

Although CXCR4 was detected at the leading edge of both the axons and the other minor neurites, SDF-1 modelling activity affected only the former and not the latter. One possibility is that the apparent selective action of SDF-1 on axonal patterning might simply result from the concentration of SDF-1 used in the present study, and that higher doses would affect axons as well as the other neurites. Another explanation relies on the fact that axons and dendrites are molecularly distinct compartments, with a different intracellular machinery and cytoskeletal organization, which may contribute to the differential effect of SDF-1 on axons and other neurites (Craig and Banker, 1994). Alternatively, an effect on dendrite patterning might not be detected as early as day 2, but might be evidenced at later stage of development, especially at the time of dendrite extension.

The data described in the present study add to recent advances in the identification of signals that regulate axonal development and further characterize the function of SDF-1/CXCR4 interactions in neurons of the central nervous system. Our data identify SDF-1 as a new extracellular signal involved in the axonal patterning of hippocampal neurons, which suggests that this chemokine might play a critical role in the establishment of neuronal connectivity during development.

We thank M. Alizon and C. Labbé-Jullié for the gift of CXCR4-GFP cDNA, A. Lombet for help with Ca²⁺ imaging and A. Nicot and P. Mechighel for help with western blot experiments. This work was supported by INSERM.