Dynamics and function of CXCR4 in formation of the granule cell layer during hippocampal development

In the developing hippocampus, granule cell progenitors (GCPs) arising in the ventricular zone (VZ) migrate to the subpial region, and form the granule cell layer (GCL) of the dentate gyrus (DG). To understand the mechanism of GCL formation, we investigated the dynamics and function of CXCR4 which is expressed by the GCPs and is a receptor of the CXCL12 chemokine secreted by cells surrounding the DG. In the VZ, CXCR4 was expressed on the plasma membrane of the GCPs. During their migration and in the DG, CXCR4 was internalized and accumulated as puncta close to the centrosomes, Golgi apparatus, and lysosomes. Phosphatase analysis suggested that both phosphorylated and dephosphorylated CXCR4 exist on the plasma membrane, whereas CXCR4 in intracellular puncta was mainly dephosphorylated. Intraventricular administration of the CXCR4 antagonist AMD3100 resulted in the disappearance of CXCR4 expression from the intracellular puncta, and its appearance on the plasma membranes. Furthermore, AMD3100 treatment resulted in precocious differentiation, delayed migration, and ectopic GCPs. Taken together, these results suggest that during the development and migration of GCPs, CXCR4 on the plasma membrane is phosphorylated, internalized, sorted to the centrosomes, Golgi apparatus, and lysosomes, and functionally regulates GCP differentiation, migration and positioning.

expressed by GCPs, we used Gfap-GFP transgenic mice, because we previously demonstrated that GCPs express GFAP from the beginning of dentate development, and Gfap-GFP expressing cells accurately represent the GCPs and their immediate neural progeny 13 . In the VZ at E18.5, CXCR4 was expressed on the plasma membrane of Gfap-GFP-positive (Gfap-GFP+) cells ( Fig. 2A,B,C,a1,b1, and C1). In the DMS and DG, CXCR4 immunoreactivity was observed as punctate structures at the base of processes of Gfap-GFP+ fusiform-shaped cells (Fig. 2 a2,a3,b2,b3,c2, and c3). However, we also found punctate CXCR4 immunoreactivity in Gfap-GFP-negative (Gfap-GFP−) areas where Gfap-GFP-negative migrating neuronal precursors could be present. To confirm this, we labeled migrating cells by in utero electroporation, in which the pCAGGS-BFP plasmid was electroporated into the ventricular surface of the hippocampal region. We found CXCR4 puncta in BFP+/Gfap-GFP− cells with a nucleus positive for NeuroD, an immature neuronal marker, suggesting the existence of punctate CXCR4 in neuronal precursors in which Gfap promoter activity is downregulated (Supplemental Fig. 1). Additionally, to demonstrate the early distribution pattern of CXCR4 in Gfap-GFP cells, we analyzed the prospective hippocampus at E14.5 (Supplemental Fig. 2). The distribution pattern of CXCR4 in the E14.5 prospective hippocampus was similar to that in the E18.5 hippocampus. In the VZ, CXCR4 was expressed on the plasma membrane of Gfap-GFP cells. In the DMS and prospective dentate gyrus, CXCR4 was found as puncta in the Gfap-GFP cells. These results suggest that during migration of granule progenitors, CXCR4 distribution is altered from the plasma membrane to intracellular puncta.

Intracellular distribution of CXCR4+ puncta in GCPs.
Because in the immune system, CXCR4 is internalized and sorted to organelles such as the lysosome 24 , we next investigated whether CXCR4+ puncta in the DMS and DG are associated with any organelles. To detect the Golgi apparatus, lysosome and centrosome, we used antibodies for GM130, LAMP1, and γ-tubulin, respectively. In the DMS, 40.5%, 54.6% and 73.6% of CXCR4-positive puncta were in contact with structures immunoreactive for GM130, LAMP1 and γ-tubulin, respectively ( Fig. 3A,C,E, and G). In the DG, 46.3%, 56.4%, and 74.7% of CXCR4-positive puncta were in contact with GM130, LAMP1, and γ-tubulin-immunoreactive structures, respectively (Fig. 3B,D,F, and G, n = 3, 9 sections in 3 embryos). Furthermore, immunoelectron microscopy demonstrated that CXCR4 immunoreactivity is often localized near the centrosome at the base of a process (Fig. 3H). These results suggest the possibility that CXCR4 may be sorted to the Golgi apparatus, lysosomes, and centrosomes, and may hence function with these organelles.
Inhibition of CXCL12/CXCR4 signaling alters the intracellular localization of CXCR4. We next investigated whether the formation of CXCR4-positive puncta in the DMS and DG is regulated by CXCL12/ CXCR4 signaling, because it is widely accepted that CXCR4 is internalized by CXCL12 stimulation [34][35][36][37][38] . To this end, the CXCR4 antagonist AMD3100 29-31 was injected into the telencephalic lateral ventricle of E15.5 embryos in utero, and embryos were left to develop until E18.5. In the whole of the hippocampus, no CXCR4+ puncta as seen in control were detected in the AMD3100-treated mice, and instead all CXCR4 staining was found in the plasma membrane (Fig. 4). Dephosphorylation by λPP did not significantly alter the number of cells with CXCR4 on their plasma membrane in the VZ-SFR (-PP, 2290 ± 2384 cells vs. +PP, 4320 ± 1428 cells, P = 0.134, n = 6, 6 sections in 6 embryos, Student t-test) and DMS (-PP, 5408 ± 2706 cells vs. +PP 8912 ± 3604 cells, P = 0.113, n = 6, 6 sections in 6 embryos, Student t-test) in AMD3100-treated embryos, suggesting that the majority of migrating cells have dephosphorylated CXCR4 in AMD3100-treated embryos, probably because of the absence of CXCL12 stimulation (Supplemental Fig. 4). In the VZ, the expression pattern of CXCR4 was basically similar between control and AMD3100-treated mice (Figs 4A1, 2, B1, and B2). In the DMS and DG, strong CXCR4 expression was observed on the plasma membrane ( Fig. 4A3, A4, B3, and B4). Additionally, CXCR4 cells appeared to be accumulated around the hippocampal fissure. These results suggested that CXCL12/CXCR4 signaling controls the formation of CXCR4+ puncta.

Inhibition of CXCL12/CXCR4 signaling induces precocious neural differentiation in the VZ.
To understand the properties of cells in the VZ after AMD3100 administration, we examined the expression of the proliferating cell marker Ki67, and the neural progenitor marker NeuroD. Initially we counted the number of Hoechst-labeled nuclei in the VZ at a distance of 250 μm dorsal from the dentate notch, because it corresponds approximately to the GCP-generating region. In the present study, we will refer to this as the GCP-generating ventricular zone (GCP-VZ). In the GCP-VZ ( Fig. 5A-E), we found no differences (P = 0.231, Student t-test) in the number of Hoechst-labeled cells between the control (921 ± 101 cells in the GCP-VZ, n = 3, 10 sections from each of 3 embryos) and AMD3100-treated embryos (801 ± 125 cells in the GCP-VZ, n = 4, 10 sections from each of 4 embryos). Immunohistochemical analysis showed that in the control, Ki67+ cells were densely packed in the VZ, but in the AMD3100-treated mice, they were sparsely distributed ( Inhibition of CXCL12/CXCR4 signaling induces delayed migration and precocious differentiation of GCPs. To investigate whether CXCL12/CXCR4 signaling regulates the final differentiation of granule cells, we analyzed the expression of the granule cell marker Prox1. AMD3100 was injected into the telencephalic lateral ventricle of E15.5 embryos in utero, and embryos were left to develop until E18.5. In control embryos, cells strongly positive for Prox1 were mainly localized in the DG (Fig. 6A), but in AMD3100 treated-embryos, cells with strong ectopic expression of Prox1 were detected in the VZ and DMS (Fig. 6B). The average number of cells with strong Prox1 staining in the VZ and DMS significantly increased (P < 0.05; Student t-test) in the AMD3100-treated mice (207 ± 87 cells, n = 6, 10 sections from each of 6 embryos) compared with control mice (37 ± 14 cells, n = 5, 10 sections from each of 5 embryos, Fig. 6C). However, no difference was observed in the total number of Prox1-positive cells in the VZ + DMS and DG between the two conditions ( Fig. 6C; control, 1,696 ± 235 cells vs. AMD3100, 1,675 ± 156 cells, P = 0.875, n = 5 and 6, 10 sections from each of 5 or 6 embryos, respectively, Student t-test), suggesting that inhibition of CXCL12/CXCR4 signaling does not alter the production Inhibition of CXCL12/CXCR4 signaling results in ectopic Gfap-GFP+ cells surrounding the hippocampal fissure. As mentioned above, in AMD3100-treated embryos, CXCR4+ cells appeared to be accumulated in a region surrounding the hippocampal fissure that was occupied with p73+ cells (Fig. 7). In the present study, we call this area the hippocampal fissure-surrounding region (HFSR). To investigate the properties of the accumulated cells, we first examined the localization of Gfap-GFP+ cells of control and AMD3100-treated Gfap-GFP mouse embryos. In the control embryos, a few Gfap-GFP+ cells were present in the HFSR (Fig. 7A2 and A3) and the processes from Gfap-GFP+ cells in the DG invaded into the HFSR (Fig. 7A2  and A3). In the AMD3100-injected mice, many ectopic Gfap-GFP+ cells accumulated in the HFSR (Fig. 7B2  and B3). The proportion of Gfap-GFP+ cells in the HFSR to total GFP+ cells in the DG and HFSR significantly increased (P < 0.05, Student t-test) in the AMD3100-treated mice (19.2% ± 1.7%, n = 4, 10 sections from each of 4 embryos) compared with control mice (15.3% ± 2.3%, n = 3, 10 sections from each of 3 embryos; Fig. 7C), although no difference was detected in the number of p73-positive cells in the HFSR between the mice ( Fig. 7D; control, 1,093 ± 38 cells vs. AMD3100, 1,120 ± 107 cells, P = 0.741, n = 3 and 4, 10 sections from each of 3 or 4 embryos, respectively).

Discussion
Previous studies using CXCR4-deficient or CXCL12-deficient mice indicated that CXCL12/CXCR4 signaling regulates the differentiation and migration of GCPs, as well as organization of the GCL 14,19,20 . However, the dynamics of the CXCR4 protein during GCP development have remained unclear. Our present study using a combination of an antibody specific for non-phosphorylated CXCR4, a λPP and an inhibitor of CXCL12/CXCR4 signaling, AMD3100 suggested that CXCL12/CXCR4 signaling induces the phosphorylation and internalization of CXCR4 as well as the intracellular punctate accumulation near the centrosomes, Golgi apparatus and lysosomes. Furthermore, functionally, this signal acts to maintain the stem cell-like properties of the GCPs, and regulates the differentiation, migration and final positioning of the GCPs. Because in the immune system and other organs, the dynamics of CXCR4 induced by CXCL12 signaling is involved in the function of CXCR4 28, 39-42 , it is probable that the phosphorylation and internalization of CXCR4 in GCPs correlate with the differentiation and migration of GCPs.
We found that in the VZ, CXCR4 is mainly present on the plasma membrane, in either the phosphorylated or non-phosphorylated form. Additionally, the number of cells with CXCR4+ intracellular puncta was low. It is well known that the binding of CXCL12 to CXCR4 activates two processes: one is a number of signaling cascades mediated via heterotrimeric G-proteins that lead to proliferation and migration, and the other is phosphorylation-dependent endocytosis of the CXCL12/CXCR4 complex followed by lysosomal degradation and desensitization of the receptor signal 24,43 . In our present study, the existence of non-phosphorylated CXCR4 on the plasma membrane and a small number of intracellular CXCR4+ puncta imply that CXCR4 is largely available on the plasma membrane, and is not significantly down regulated. Furthermore, inhibition of CXCL12/CXCR4 signaling by the CXCR4 antagonist, AMD3100 caused a decrease in the number of proliferating progenitor cells and an increase in the number of differentiated neuronal precursors. These results suggest the precocious maturation of GCPs in the VZ. A similar type of precocious differentiation of progenitor cells has been reported in the developing cerebellum. Analysis using CXCR4-null mice showed that CXCL12/CXCR4 signaling is required for anchoring progenitors to granule cells within the external granule layer (EGL), which possesses an environment that maintains progenitor proliferation. The loss of CXCR4 leads to the premature migration of the progenitors away from the EGL 44,45 . Taken together, it is most likely that CXCR4-mediated signaling functions in maintaining stem cell-like features of the GCPs in the VZ and preventing their precocious differentiation.
In the DMS, CXCR4 was found both on the plasma membrane and in the intracellular puncta, the latter of which were present close to the centrosome, Golgi apparatus and lysosomes. CXCR4 molecules in the plasma membrane were primarily phosphorylated. When CXCL12/CXCR4 signaling was inhibited, most CXCR4 molecules were confined to the plasma membrane and were dephosphorylated. These results suggest that the formation of CXCR4+ intracellular puncta and phosphorylation of CXCR4 are regulated by CXCL12/CXCR4 signaling. In the immune system, phosphorylation-dependent internalization of CXCR4 is well documented. CXCR4 on the plasma membrane of T-cells, after exposure to CXCL12, is rapidly internalized, and then trafficked to the Golgi apparatus to regulate T-cell migration 41 , or is sorted to lysosomes 46 to be degraded or recycled [39][40][41] . It has been proposed that the internalization of CXCR4 receptors from the cell surface and their degradation by lysosomes cause a decrease in the number of CXCR4 on the cell surface, and regulate the availability of CXCR4 receptors, as well as the magnitude and duration of signaling. Studies of hematopoietic stem cells indicate that a high level of CXCL12 induces desensitization and receptor internalization, whereas a low level of CXCL12 promotes cell motility, proliferation, and migration [47][48][49] . In the adult hippocampus and in primary embryonic hippocampal culture, the level of CXCR4 in CXCL12-stimulated neurons is reported to be under dual control by the recycling and degradation of internalized CXCR4 50 . Taken together, it is probable that the binding of CXCL12 to CXCR4 on the plasma membrane of the GCPs induces the phosphorylation and internalization of CXCR4, and the internalized CXCR4 is sorted to the centrosome, Golgi apparatus, and lysosomes.
A few studies have reported the intracellular punctate expression of CXCR4 in the developing nervous system, but the function of these CXCR4+ punctate structures remains unclear. In cultured neuronal progenitors from the E15 rat brain, CXCR4 was found as punctate aggregates in the perinuclear region 51 . In developing CXCR7-null mice in which CXCL12 levels are increased, intracellular clustering of CXCR4 is detected in migrating GnRH-positive neurons, although similar clusters are not found in wild-type mice 52 . This implies that excess CXCL12/CXCR4 signaling causes the intracellular clustering of CXCR4. In the adult hippocampus, CXCR4 expression is found in the plasma membrane and intracellular puncta of neural progenitors/precursors in the subgranular zone, and inhibition of CXCL12/CXCR4 signaling leads to an increase in the number of cells with a CXCR4+ membrane 50 . In addition to the intracellular punctate expression of CXCR4, we further found the association of CXCR4+ puncta with the centrosome, Golgi apparatus, and lysosomes. As mentioned above, CXCR4 that is sorted to lysosomes may be degraded or recycled. However, the roles played by CXCR4 that is trafficked to   A1, B1, and E-L; 20 µm in A2, A3, B2, and B3.
the centrosome and Golgi apparatus remain unclear. In this respect, it has been reported that the centrosome and Golgi apparatus are associated with cell migration 41,53,54 . It is hence possible that CXCR4 in association with the centrosome and Golgi apparatus play a role in the migration of GCPs.
In the present study, we found that the inhibition of CXCL12/CXCR4 signaling causes a reduction in the number of Prox1+ cells in the DG, suggesting the delayed migration of GCPs. Furthermore, in this condition, the GCPs underwent precocious differentiation before arriving at the DG. Similar delayed migration has been shown by studies using CXCR4-deficient and CXCL12-deficient mice. These studies indicated that CXCL12/CXCR4 signaling regulates the migration of CR cells 55,56 , GnRH neurons in the forebrain 52 , and dopamine neurons in the midbrain 57 . In the hippocampus of CXCR4-deficient mice, the delayed migration and premature differentiation of GCPs has been reported 14,19,20 . Taken together, these studies suggest that in the DMS, CXCR4 is involved in the migration and differentiation of GCPs.
Normally, GCPs migrate and settle in the DG where they differentiate into granule cells or continue to proliferate to produce granule cells. However, in AMD3100-treated mice, many GCPs were ectopically detected in the HFSR, suggesting that inhibition of CXCL12/CXCR4 signaling disrupts the final positioning. It is not clear how GCPs normally stop and settle in the DG, and how the inhibitor disturbs the positioning. In this respect, in zebrafish, the internalization of CXCR4 was reported to be essential for the precise arrival of primordial germ cells at the target in CXCL12-guided cell migration 58 . Consistently, we found that in the DG, CXCR4 was mainly present in intracellular punctate structures. Thus, one possibility is that the internalization of CXCR4 in GCPs is involved in their precise positioning within the DG, and the inhibition of internalization disrupts this positioning. In the immune system, it has also been reported that immune cells are attracted by low concentrations of CXCL12, but are repulsed by high concentrations of CXCL12 59 . In the developing hippocampus, the concentration of CXCL12 is thought to be very high in the HFSR, because the HFSR contains many CR cells that secrete CXCL12. Therefore, another possibility is that the high CXCL12 concentration serves as a repellent to the GCPs, and prevents invasion of the GCPs into the HFSR.
The intracellular distribution pattern of CXCR4 and the AMD3100-induced alteration in the GCPs depended on the hippocampal region, such as the VZ, DMS, and DG. In the hippocampus, CXCL12 is secreted by CR cells located in the HFSR. Concentrations of the CXCL12 signal may decrease from the DG to the VZ. CXCL12 stimulation is well known to cause the phosphorylation and subsequent internalization of CXCR4. The concentration of CXCL12 could define the extent of phosphorylation and internalization of CXCR4. During the migration of GCPs from the VZ to DG via the DMS, CXCR4 can be gradually phosphorylated in response to CXCL12, and then progressively internalized and accumulated as the GCPs approach the DG. It is thus possible that the different dynamics and functions of CXCR4 within these three regions are induced by the different intensity of CXCL12 signals.
On the basis of this concept, we propose a hypothetical model for the dynamics and function of CXCR4 in the formation of the GCL during hippocampal development (Fig. 8). In the VZ, which is located far from the HFSR, CXCL12 levels are very low, which induces the partial phosphorylation of CXCR4. In this situation, the amount Figure 8. A hypothetical model of the dynamics and function of CXCR4 in the formation of the GCL during hippocampal development. CXCL12 is secreted by Cajal-Retzius cells located in the hippocampal fissure-surrounding region (HFSR). A gradient of CXCL12 molecule concentration is thought to exist in the hippocampus, decreasing from the DG to the VZ. CXCL12 stimulation is well known to cause the phosphorylation and subsequent internalization of CXCR4. Thus, the proposed hypothesis is as follows. In the VZ (low level of CXCL12), CXCR4 is mainly present on the plasma membrane, and is partially phosphorylated. CXCR4 signaling may function in maintaining the stemness of ventricular cells. In the DMS (medium level of CXCL12), CXCR4 on the plasma membrane is mostly phosphorylated and partially internalized. The internalized CXCR4 traffics to the centrosomes, Golgi apparatus or lysosomes and is dephosphorylated where it may play a role in cell migration. In the DG (high level of CXCL12), CXCR4 is mostly internalized, and accumulates as intracellular punctate structures close to the centrosomes, Golgi apparatus, or lysosomes, and also becomes dephosphorylated. CXCR4 may thus regulate the final positioning of granule cell progenitors. of internalized CXCR4 is very low. As a result, CXCR4 is mainly present on the plasma membrane, in either its phosphorylated or non-phosphorylated form. When CXCR4 is located on the plasma membrane of cells in the VZ, CXCR4 may be involved in maintaining stem cell-like features of the GCPs in the VZ, regulating neural differentiation, and preventing precocious maturation. In the DMS, which is the mid-point between the VZ and DG, the CXCL12 levels are moderate. The moderate levels of CXCL12/CXCR4 signaling result in the complete phosphorylation of CXCR4 and the partial internalization of CXCR4. The internalized CXCR4 accumulates as punctate aggregates close to the centrosomes, Golgi apparatus, and lysosomes. This internalized CXCR4 can be recycled or degraded when it is sorted to the lysosomes, or might function in cell migration with microtubules when it traffics to the centrosomes and Golgi apparatus. In the DG near the HFSR, CXCL12 levels are high, and CXCR4 is mostly internalized, and accumulates as intracellular punctate structures close to the centrosomes, Golgi apparatus, and lysosomes. This condition may regulate the final positioning of the GCPs. This hypothesis provides new insight into the association between the dynamics and function of morphogens and their receptors in neural development.

Methods
Animals. The generation of mouse Gfap-GFP mice 60 has been described previously. Mice were maintained under standard conditions (12-hour light/dark cycle) in the animal care center of Tokyo Medical University. All experiments were performed in accordance with institutional, science community, and national guidelines for animal experimentation and approved by the Institutional Animal Care and Use Committee (IACUC) of Tokyo Medical University and also conform to the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH publication no. 80-23), revised in 1996. All efforts were made to minimize the number of animals used and their suffering. For timed mating, the day of vaginal plug detection was counted as embryonic day 0.5 (E0.5). The day of birth was counted as postnatal day 0. Forty-two embryos were used in this study.

In utero electroporation and intraventricular injection.
In utero electroporation of mouse embryos was performed as previously described 61 with minor modifications 61 . Briefly, pregnant wild-type mice at E15.5 were anesthetized with sodium pentobarbital. After cleaning the abdomen with 70% ethanol, a midline incision of approximately 3 cm was made. The uterus was exposed, and the lateral ventricle of the embryos was identified under transillumination. The pCAGGS-BFP plasmid constructed from a vector developed by Niwa et al. 62 was diluted with PBS (final DNA concentration: 0.5 µg/ml), and 0.5 µL of the solution was injected into the lateral ventricle with a glass capillary pipette connected to an electric injector (BJ-110, BEX Co. Ltd.). Square pulses (30 V, 50 milliseconds, 4 times at 1-second intervals) were delivered to the neocortex with tweezer-type electrodes, which consisted of a pair of round platinum plates that were 3 mm in diameter (LF650P3, BEX, Tokyo, Japan), and an electroporator (CUY21, Nepa Gene, Chiba, Japan). The electroporated embryos were subsequently fixed at E18.5.
For the blocking of CXCL12/CXCR4 signaling, surgery was performed in the same manner as described above. One µL of AMD3100 solution (12.6 mM in PBS; Sigma-Aldrich, St. Lous, MO) or PBS was injected into the telencephalic lateral ventricle of E15.5 embryos in utero, and embryos were left to develop until E18.5.
Tissue preparation. For embryonic time points, pregnant mice were deeply anesthetized with sodium pentobarbital and embryos were removed via laparotomy. The brains were removed from the skull and immersion-fixed overnight in 4% paraformaldehyde in 0.1 M PB at 4 °C. The fixed brains were washed with PBS and the forebrains were embedded in OCT compound and stored at −80 °C. The forebrains were sectioned serially and coronally into 20-µm slices using a cryostat (Fig. 1), or the medial cerebral cortices including the hippocampus were cut serially and perpendicularly to the septo-temporal axis in the same manner (Figs 2-7). Every four sections were mounted on slides and were subjected to immunohistochemistry and quantification.
To detect total CXCR4 (both phosphorylated and non-phosphorylated forms), sections of the hippocampus were treated with λPP according to the method previously described by Yang et al. 57 , because anti-CXCR4 UMB2