CXC Chemokine Receptor 7 (CXCR7) Affects the Migration of GnRH Neurons by Regulating CXCL12 Availability

Gonadotropin-releasing hormone (GnRH) neurons are neuroendocrine cells, located in the hypothalamus, that play an essential role in mammalian reproduction. These neurons originate in the nasal placode and migrate during embryonic development, in association with olfactory/vomeronasal nerves, first in the nose, then through the cribriform plate to enter the forebrain, before settling in the hypothalamus. One of the molecules required for their early migration in the nose is the chemokine CXCL12, which is expressed in the embryonic nasal mesenchyme in an increasing ventral to dorsal gradient, presumably guiding GnRH neurons toward the forebrain. Mice lacking CXCR4, the receptor for CXCL12, exhibit defective GnRH cell movement and a significant reduction in their number, suggesting that CXCL12/CXCR4 signaling is important in the migration and survival of these neurons. Here, we investigated the role of the more recently identified second CXCL12 receptor, CXCR7, in GnRH neuron development. We demonstrate that CXCR7 is expressed along the migratory path of GnRH neurons in the nasal cavity and, although not expressed by GnRH neurons, it affects their migration as indicated by the ectopic accumulation of these cells in the nasal compartment in CXCR7−/− mice. Absence of CXCR7 caused abnormal accumulation of CXCL12-RFP at CXCR4-positive sites in the nasal area of CXCL12-RFP-transgenic mice and excessive CXCL12-dependent intracellular clustering of CXCR4 in GnRH neurons, suggesting internalization. These findings imply that CXCR7 regulates CXCL12 availability by acting as a scavenger along the migratory path of GnRH neurons and, thus, influences the migration of these cells in a noncell-autonomous manner.


Introduction
Gonadotropin-releasing hormone (GnRH) neurons are a small group of hypothalamic cells that control mammalian reproduction by secreting GnRH decapeptide into the portal vessels of the pituitary. There, GnRH stimulates the production of gonadotropins, which control gonadal function (Merchenthaler et al., 1984). During development, GnRH neurons originate in the nasal placode and migrate in association with olfactory/vomeronasal nerve axons to the hypothalamus (Schwanzel-Fukuda and Pfaff, 1989;. In humans, failure of these neurons to reach the hypothalamus results in infertility (Hardelin, 2001). The mechanisms that guide the migration of GnRH neurons are complex, as they encounter different molecular environments during their migratory journey. Thus, their migration is affected by a wide range of factors (Cariboni et al., 2007b;Wierman et al., 2011;Giacobini and Prevot, 2013) including the chemokine CXCL12 (Schwarting et al., 2006;Toba et al., 2008;Casoni et al., 2012), which has an established role in directed cell migration (Tiveron and Cremer, 2008;Lewellis and Knaut, 2012). CXCL12 has been observed in the nasal mesenchyme (NM), whereas its receptor, CXCR4, has been localized in migrating GnRH neurons and olfactory/vomeronasal nerve axons. Further, CXCR4-deficient mice exhibit a loss of GnRH neurons and impaired migration, suggesting the importance of CXCL12/CXCR4 signaling in the development of this system (Schwarting et al., 2006;Toba et al., 2008).
Recent work established CXCR7 as the second CXCL12 receptor (Balabanian et al., 2005;Burns et al., 2006). Unlike CXCR4, however, CXCR7 fails to signal through G␣ i proteins and does not facilitate cell migration in response to CXCL12, suggesting that it is an atypical chemokine receptor (Burns et al., 2006;Boldajipour et al., 2008;Rajagopal et al., 2010). CXCR7 has a 10-fold higher binding affinity to CXCL12 than CXCR4, and its property to rapidly internalize CXCL12 and to degrade large amounts of extracellular chemokine has led to the concept that it acts as a scavenger to control the levels of the chemokine Naumann et al., 2010;Hoffmann et al., 2012). In this context, studies in zebrafish (Boldajipour et al., 2008) and mice (Sánchez-Alcaniz et al., 2011) have demonstrated that CXCR7 indeed plays a role in cell migration by regulating CXCL12 pro-tein levels available to CXCR4-expressing cells in the process of migration.
Here we investigated the contribution of CXCR7 in CXCL12 signaling in GnRH neuron migration. We found that CXCR7 is expressed within the NM, especially in its ventral aspect, but not by GnRH neurons or their guiding axons. Mice deficient of CXCR7 exhibit impaired migration and ectopic GnRH neuron distribution in the nasal area, suggesting that this receptor affects their directed migration in an indirect manner, possibly by regulating CXCL12 availability. Indeed, a CXCL12/RFP fusion protein (CXCL12-RFP) accumulated at different sites in the nasal area of CXCL12-RFP-transgenic mice lacking functional CXCR7 than of CXCL12-RFP-transgenic controls. Furthermore, CXCR7null mice exhibited severely reduced CXCR4 immunoreactivity and excessive intracellular clustering of CXCR4 in the entire nasal area including GnRH neurons, which suggests receptor hyperactivation due to CXCL12 overstimulation. In support of this hypothesis, CXCL12 Ϫ/Ϫ ;CXCR7 Ϫ/Ϫ mice displayed lack of CXCR4 internalization and increased CXCR4 levels, pointing to a role for CXCR7 in controlling CXCL12 availability in this system.
For quantification of GnRH neurons, embryonic heads (E12.5-E16.5) were fixed in PFA, made in PBS, for 4 -8 h at RT. After fixation, tissue was cryoprotected in 30% sucrose in PBS, embedded and frozen in Tissue-Tek OCT (Sakura Finetek), and sectioned sagittally at 20 m using a cryostat (Bright Instruments). Sections were washed in PBS and incubated with H 2 O 2 to quench endogenous peroxidase activity. They were then blocked in PBS, containing 5% NGS (Sigma-Aldrich) (v/v) and 0.1% Triton X-100 (v/v) (Sigma-Aldrich), at RT for 2 h. They were subsequently incubated in anti-GnRH (1:1000) antibody at RT overnight, followed by washes in PBS, and incubated in biotinylated antirabbit (1:200; Vector Laboratories) for 2 h. Immunoreactivity was visualized with the ABC kit (Vector Laboratories) followed by DAB (Sigma-Aldrich) development. For each genotype, we analyzed at least three embryos at each age, and all GnRH-positive cells were counted in every section of each head under a 40ϫ microscope objective. For all experiments, data were expressed as mean Ϯ SEM. To determine statistical significance, we used unpaired Student's t test, with p Ͻ 0.05 considered to be statistically significant.
CXCR4-immunoreactivity was quantified in UMB-2-stained coronal head sections of CXCR7 Ϫ/Ϫ , CXCL12 Ϫ/Ϫ ;CXCR7 Ϫ/Ϫ , and homozygous or heterozygous control littermates. For quantification, the nose region was photographed under identical imaging conditions with an Axio Imager A1 microscope (Zeiss) connected to a ProgRes C5 camera (Jenoptik). The sensory epithelium of the vomeronasal organ (VNO) and the olfactory epithelium (OE) were defined as areas of interest using ImageJ software. The proportion of the stained area at the region of interest was measured after setting a uniform threshold level. Measurements were expressed as percentage of the control group.

CXCR7 is expressed in the nasal area
We first analyzed the spatiotemporal profile of CXCR7 expression in the developing nasal area from early to late embryonic life, and compared its pattern of expression to that of CXCR4 and their agonist, CXCL12. Analysis was performed in embryos from E12.5 to E16.5, a period that covers the early stage of migration (E12.5) of GnRH neurons, following their emergence from the nasal placode at E10.5 (Schwanzel-Fukuda and Pfaff, 1989), and the migration between the nasal compartment and the basal forebrain (E14.5-E16.5).
As we observed only minor differences in the overall patterns of the three genes between E12.5 and E14.5, we shall only provide details of their expression at E14.5 ( Fig. 1). At this age, CXCR7 was strongly expressed in the VNO and the surrounding area ( Fig. 1 A, AЈ). Specifically, CXCR7 mRNA was detected in the sensory epithelium of the VNO (Fig. 1AЈ, arrow), the surrounding cartilaginous capsule, the cavernous body adjacent to the VNO, and the NM dorsomedial to the VNO (Fig. 1AЈ). CXCL12 expression was found to overlap with CXCR7 in the apical zone of the VNO (Fig. 1AЈ,BЈ). CXCR4 was present throughout the VNO with the most prominent signal being localized in the basal zone (Fig. 1CЈ). Cells in the nasal area that appeared to be migrating were also CXCR4 positive (Fig. 1C,CЈ, arrows) and resembled GnRH neurons as previously reported by Schwarting et al. (2006). The respiratory epithelium (RE), which is located lateral to the VNO in the ventral part of the nasal cavity, strongly expressed CXCR7 ( Fig. 1 A, AЈ), but contained only few CXCR4-positive cells (Fig.  1C,CЈ) and no CXCL12 signal ( Fig.  1 B, BЈ). More dorsally, in the OE, both CXCR7 and CXCR4 were detected (Fig.  1AЉ,CЉ). CXCL12 was absent in the OE except a small area in the dorsal-most part of the nasal cavity (Fig. 1BЉ).
In agreement with a previous report (Schwarting et al., 2006), CXCL12 was prominently expressed in the NM (Fig. 1B-BЉ). Dual labeling for CXCL12 and GnRH showed that the CXCL12-positive NM corresponds to the migration corridor of GnRH neurons ( Fig. 1D-DЉ, arrows). Interestingly, CXCR7 was also detected in the NM ( Fig. 1 A, AЈ) but, unlike CXCL12, it was concentrated ventrally (dorsomedial to the VNO) and expressed weakly in the dorsal NM.

CXCR7 is not expressed in migrating GnRH neurons and their guiding axons
It has been established that CXCR7 expressed in non-migrating cells along CXCL12-positive migration paths can modulate CXCL12/CXCR4-guided cell migration (Boldajipour et al., 2008). Recent reports suggested that CXCR7 expressed in migrating neurons can directly influence the migration process (Sánchez-Alcaniz et al., 2011;Wang et al., 2011). We thus wondered if CXCR7 modulates GnRH neuron migration in a direct or indirect manner and sought to determine the identity of CXCR7-and CXCR4-expressing cells in the nasal area. In these experiments, we used BAC transgenic mice carrying eGFP under the control of CXCR7 and CXCR4 promoters in double immunofluorescence for GnRH/eGFP. We found that the overall expression patterns of CXCR7-and CXCR4-eGFP (Fig. 2 A, D-G) matched the distribution of the respective mRNAs ( Fig. 1 A, C). However, in the NM, strong CXCR7-eGFP signal was present along the dorsoventral extent ( Fig. 2A), whereas strong CXCR7 mRNA signal was restricted to the ventral mesenchyme surrounding the VNO (Fig. 1A). This difference could be due to stability of eGFP, which permits accumulation of the reporter protein despite low gene expression. Nevertheless, GnRH neurons showed no colocalization of eGFP in CXCR7-eGFP mice at the three ages examined (E12.5, E14.5, and E16.5; Fig. 2C-CЉ), indicating that GnRH neurons do not express CXCR7 during their migration. Since olfactory/vomeronasal axons serve as guides to migrating GnRH neurons, we tested whether CXCR7 is present in this axonal scaffold by staining sections through the nasal compartment of mice at E12.5-E16.5 for peripherin. These experiments indicated that peripherin-labeled axons were also negative for CXCR7-eGFP (Fig. 2B, arrowheads). In similar experiments, using CXCR4-eGFP mice, we confirmed that, in con-trast to CXCR7, the majority of GnRH neurons (Schwarting et al., 2006;Fig. 2I-IЉ) and their guiding axons (Toba et al., 2008;Fig. 2H, arrowheads) express CXCR4 at all ages examined.

CXCR7 regulates CXCL12 availability and CXCR4 activation in the nasal area
If CXCR7 is not present in GnRH neurons and their scaffold, could it exert an indirect effect on their CXCL12/CXCR4-guided migration? To approach this question, we sought to establish whether CXCR7 influences CXCL12 availability and CXCR4 activation in the nasal area. Since chemokine receptors internalize and accumulate their bound ligands in intracellular vesicles, identification of RFP-labeled cells in transgenic mice expressing RFP-labeled chemokine proved useful to reveal sites of chemokine/chemokine receptor interaction . We examined the nasal area of BAC-transgenic mice expressing an RFP-fused version of CXCL12 (CXCL12-RFP) under the control of the CXCL12 promoter (Bhattacharyya et al., 2008), and found that RFP transcripts (data not shown) exactly recapitulated the pattern of CXCL12 mRNA in wild-types ( Fig. 1B-BЉ), whereas CXCL12-RFP protein (Fig. 3A) did not. Specifically, the NM dorsomedial to the VNO showed matching strong CXCL12 mRNA and robust RFP expression (Figs. 1BЈ, 3 A, D). The VNO, however, which showed moderate expression of CXCL12 mRNA (Fig.  1BЈ), contained low RFP signal (Fig. 3D). The most striking mismatch between CXCL12 expression and RFP was observed in the RE, which showed very strong RFP signal (Fig. 3 A, E) despite being CXCL12 mRNA negative (Fig. 1B). Most of the OE was sparsely RFP positive ( Fig. 3 A, F ) and CXCL12 mRNA negative ( Fig. 1 B, BЉ; a small dorsal area moderately expressed CXCL12 and contained sparse RFP). In CXCR7-eGFP;CXCL12-RFP bitransgenic mice, we observed that the strongly RFP-positive NM and the RE (data not shown) were eGFP positive. High-power confocal analysis confirmed that RFP-signal was often contained within CXCR7-eGFP-positive cells (Fig. 3C, arrows), suggesting that the signal could, at least in part, result from CXCR7mediated CXCL12-RFP accumulation.
We then generated CXCL12-RFP-transgenic CXCR7deficient (CXCL12-RFP;CXCR7 Ϫ/Ϫ ) mice to test this assumption, and found that the pattern of CXCL12-RFP was severely altered in the absence of CXCR7. The most obvious difference was noted in the RE (Fig. 3 E, H ), which normally expresses CXCR7 and contains strong RFP signal. In CXCL12-RFP; CXCR7 Ϫ/Ϫ animals, this structure was virtually RFP negative (Fig. 3H ). Since the RE does not express CXCL12 and contains only few CXCR4-positive cells, this observation suggests that CXCR7 normally accumulates/sequesters RFP in this tissue. Also in the NM dorsomedial to the VNO, the RFP signal was severely reduced in the absence of CXCR7 (Fig. 3 D, G). The residual RFP signal most likely reflects the strong CXCL12-RFP expression. Interestingly, VNO (Fig. 3 D, G) and OE (Fig. 3 F, I ) showed increased RFP labeling in CXCL12-RFP/CXCR7 Ϫ/Ϫ mice. Given that both structures highly express CXCR4 (Fig. 1C-CЉ), the increased RFP labeling might be due to enhanced CXCL12-RFP/ CXCR4 interactions. Collectively, our analysis indicates that CXCL12-RFP preferentially accumulates at sites expressing CXCR7. In the absence of CXCR7, CXCL12-RFP accumulates in CXCR4-expressing structures.
If absence of CXCR7 does lead to increased availability of CXCL12, it would increase CXCL12/CXCR4 interaction, promote CXCR4 internalization, and, thus, alter the subcellular distribution of CXCR4. Careful analysis of the CXCR4 staining in the VNO (Fig. 4 D, E) and OE (Fig. 4K ) in CXCR7 Ϫ/Ϫ mice re- vealed the intracellular accumulation of reaction product (dots), suggesting increased internalization or allocation of the receptor in the cytoplasm of cells. In wild-types, most of CXCR4 was present at the surface and only a fraction appeared internalized in cells of the VNO (Fig. 4B) and OE (Fig. 4J ). Quantitative analysis of the CXCR4-immunoreactive area in OE and VNO confirmed that CXCR4 staining was severely reduced in CXCR7 Ϫ/Ϫ mice (Fig. 4M, middle). Double immunohistochemistry with CXCR4 and ␤III-tubulin, which labels both neuronal somata and processes, revealed absence of CXCR4 from the olfactory/vomeronasal axons in CXCR7 Ϫ/Ϫ mice (Fig. 4 D, E, arrowheads). Having shown that lack of CXCR7 affects the spatial distribution of CXCL12-RFP and CXCR4 throughout the nasal area, we turned to GnRH neurons and found that CXCR4 immunoreactivity was also reduced or even absent when compared with controls ( Fig.  4C-CЉ,F-FЉ). The possibility, that the altered pattern of CXCR4 in the nasal area is due to altered CXCR4 mRNA expression, was ruled out by in situ hybridization for CXCR4, which gave similar results in CXCR7 Ϫ/Ϫ mice and control littermates (data not shown).
To assess whether altered subcellular distribution of CXCR4 and reduced CXCR4 staining in the nasal area of CXCR7 Ϫ/Ϫ mice depends on CXCL12, we generated CXCL12 Ϫ/Ϫ ;CXCR7 Ϫ/Ϫ mice. We found that absence of CXCL12 resulted in upregulation of CXCR4 on the plasma membrane and increased CXCR4 immunoreactivity in cells of the VNO and OE (Fig. 4G, H, L,M, right) as compared with controls. Double immunofluorescence for CXCR4/␤III-tubulin and CXCR4/GnRH revealed high levels of CXCR4 in vomeronasal axons (Fig. 4G,H, arrowheads) and targeting of CXCR4 to the plasma membrane of GnRH neurons (Fig. 4I-IЉ). Thus, loss of CXCR4 from the surface of cells in the nasal area observed in CXCR7deficient mice, a typical response to high levels of CXCL12, was reversed in an experiment that combined removal of both CXCL12 and CXCR7.

CXCR7-deficient mice exhibit impaired GnRH neuron migration
Given our findings suggest that CXCR7 controls CXCL12 availability in the nasal area and CXCL12/CXCR4 interactions in GnRH neurons, we investigated whether its loss might impact on GnRH neuron migration. For this purpose, the number and distribution of GnRH neurons in sagittal sections through the nose and forebrain of CXCR7 Ϫ/Ϫ mice were compared with control littermates at three developmental stages (E12.5, E14.5, and E16.5). Regional analysis revealed an increase in the number of GnRH neurons in the nasal compartment and a concomitant decrease in the forebrain of the CXCR7 Ϫ/Ϫ mice compared with controls at all ages examined (Fig. 5C, F, I ). In detail, at E12.5, there was no significant difference between the total GnRH neuron population of CXCR7 Ϫ/Ϫ mice and controls (knockout (KO): 1346 Ϯ 64 as compared with control: 1289 Ϯ 92, p ϭ 0.64; n ϭ 3; Fig.  5C), implying that the production of these neurons is not affected by the absence of the receptor. Nonetheless, CXCR7 Ϫ/Ϫ mice exhibited a significant increase in GnRH cell numbers in the nasal compartment (with cells accumulating around the VNO; Fig. 5B, arrowhead) as compared with controls ( Fig. 5A, arrow) (KO: 1017 Ϯ 42, control: 754 Ϯ 68, *p ϭ 0.031 Ͻ 0.05; Fig. 5C) and a decrease in the forebrain (Fig. 5B) (KO: 329 Ϯ 23, control: 535 Ϯ 29, **p ϭ 0.0052 Ͻ 0.01; Fig. 5C). This finding suggests a role for CXCR7 in the early steps of GnRH neuron migration, when these cells exit the VNO and start migrating toward the forebrain. At E14.5, when Ͼ50% of GnRH neurons in control animals have already passed the cribriform plate and are directed toward the forebrain (Fig. 5D), the equivalent proportion of neurons in CXCR7-deficient mice is 34% (KO: 368 Ϯ 37 as compared with control: 714 Ϯ 50, **p ϭ 0.001, n ϭ 6; Fig. 5F ) with many GnRH cells remaining in the nasal compartment (Fig.  5E, arrow) (KO: 793 Ϯ 24, control: 637 Ϯ 33, **p ϭ 0.009 Ͻ 0.01; Fig. 5F ). Strikingly, GnRH neurons in CXCR7-deficient mice often formed large multicellular clusters, similar to those previously reported in GNR23 mice carrying a deletion in Epha5 gene (Gamble et al., 2005; Fig. 5M ), whereas they migrated in chainlike formations in wild-types (Fig. 5G). Further, many cells were found abnormally in the nasal epithelium (Fig. 5 H, L, arrowheads) and a small number appeared in the VNO lumen (Fig. 5L,  arrow), where they eventually die (Fig. 5N, arrowheads), indicating misguided migration. Still at this stage (E14.5), the total GnRH cell number did not differ significantly between the two groups of animals, although there was a slight reduction in the mutants (KO: 1161 Ϯ 32, control: 1351 Ϯ 69, p ϭ 0.069 Ͼ 0.05; Fig. 5F ). To determine the fate of the cells that accumulate in the nose, we analyzed CXCR7-null mice of E16.5. By this age, the majority of GnRH neurons in control mice have migrated to the brain, whereas in CXCR7-null animals most GnRH neurons remained in the nasal area (KO: 500 Ϯ 34, control: 298 Ϯ 22, ***p ϭ 0.0001 Ͻ 0.001; n ϭ 3; Fig. 5I ) and only a small percentage (ϳ37%), similar to that observed at E14.5, had passed the cribriform plate (KO: 289 Ϯ 26 as compared with control: 837 Ϯ 21, ***p ϭ 3.21E-09 Ͻ 0.001; Fig. 5I ). Interestingly, the mean total number of GnRH neurons in CXCR7 Ϫ/Ϫ mice of this later age appeared significantly reduced when compared with control animals (KO: 789 Ϯ 57, control: 1135 Ϯ 30, ***p ϭ 0.0004 Ͻ 0.001; Fig. 5I ), implying that abnormally accumulating GnRH neurons in the nose could have been eliminated by programmed cell death.
We next examined the patterning of olfactory/vomeronasal axons stained for peripherin and confirmed that, in CXCR7null mice, the axonal substrates used by migrating GnRH neurons were normally fasciculated and properly positioned (Fig.  5 J, K ), suggesting that CXCR7 is not involved in the patterning/targeting of these axons.

Comparative analysis of CXCL12-, CXCR4-, and CXCR7-null mice
Previous studies have demonstrated that CXCR4-deficient mice display decreased GnRH cell number and impaired migration (Schwarting et al., 2006;Toba et al., 2008). However, comparable analysis of mice lacking the ligand, CXCL12, has not been performed to date. After CXCR7 was identified as a second CXCL12 receptor (Balabanian et al., 2005), we reasoned that examination of CXCL12-null mice in comparison with animals lacking CXCR4 or CXCR7 would shed light on the role of each of these molecules in the development of the GnRH neuron system and enhance our knowledge of the mechanisms that regulate this receptor system. Thus, we analyzed CXCL12 defective mice at E14.5 and found a gross decrease in the total number of GnRH neurons when compared with control littermates (KO: 480 Ϯ 25, control: 1187 Ϯ 16; ***p ϭ 1.98E-05 Ͻ 0.0001; n ϭ 3; Fig. 6K ). We then re-examined CXCR4 Ϫ/Ϫ mice at the same age and, in agreement with previous studies (Schwarting et al., 2006), found reduced total GnRH cell number (KO: 628 Ϯ 12, control: 1227 Ϯ 25; ***p ϭ 2.96E-05; n ϭ 3; Fig. 6L).
As CXCR7 expression was shown to be absent from GnRH neurons, at least within the time frame of our analysis (E12.5-16.5; KO mice die in late gestation), we considered CXCR4 to be the only receptor mediating CXCL12 signaling on GnRH neurons. The similarity between the two phenotypes (significant decrease in the total number of GnRH neurons in CXCR4 Ϫ/Ϫ and CXCL12 Ϫ/Ϫ mice), together with our finding that CXCR7 Ϫ/Ϫ mice do not display such defect (Fig. 6M ), provide credence to this hypothesis.
In addition to the defect in total cell number in both CXCR4 and CXCL12-null mice, the regional distribution of GnRH neurons differed from that of control animals. Specifically, similar to embryos lacking CXCR7 (Fig. 6 D, I,M ), GnRH neurons in CXCL12-deficient mice were found accumulated in the nasal compartment (area between the VNO and nasal/forebrain junction) and not migrating to the brain (Fig. 6 B, G). Thus, at E14.5, there were twice as many neurons in the nose compared with the brain when, in control mice, most of GnRH neurons had already crossed the nasal/forebrain junction and were directed toward the hypothalamus. Counts of cells residing in the forebrain revealed that control animals had five times as many GnRH neurons as CXCL12 Ϫ/Ϫ mice (KO: 148 Ϯ 3, control: 701 Ϯ 9; ***p Ͻ 0.0001; Fig. 6 F, G,K ). Similarly, in CXCR4-lacking mice, the majority (60%) of the cells were still present in the nasal compartment as compared with 42% in control littermates. Concomitantly, the number of GnRH neurons in the brain differed significantly from control mice (KO: 254 Ϯ 17, control: 710 Ϯ 36; ***p ϭ 0.0003; Fig. 6C, H, L).
To gain greater insight on why GnRH neurons remain in the nasal compartment, we carefully examined their appearance and spatial distribution in this area. As depicted in Fig. 6N, considerably more cells in these animals were found abnormally in the OE (Fig. 6 B, D, arrowheads). Specifically, when we analyzed the positions of GnRH neurons in the three groups (CXCL12, CXCR4, and CXCR7 ) of KO mice, we observed that in CXCL12-and CXCR7-deficient mice, 22 and 28% of the total GnRH cell population present in the nose was aberrantly located in the epithelium; in the CXCR4 Ϫ/Ϫ mutant, that percentage was 8.5%. In comparison, hardly any (0 -2%) cells were identified in the OE in all control littermates. These observations strongly suggest that this chemokine/receptor system keeps GnRH neurons to their migratory path. We should add that we found no apparent defect in the peripherin-stained vomeronasal axons of CXCL12 Ϫ/Ϫ and CXCR4 Ϫ/Ϫ mice (data not shown).

Discussion
The migration of GnRH neurons represents a unique example of neuronal migration as these cells are born outside the brain and follow a long and tortuous route to their destinations in the forebrain. Chemokines expressed along migration corridors have been reported to serve as guidance cues for migrating neurons (Borrell and Marín, 2006;Stumm et al., 2007). Accordingly, CXCL12 has been found to be expressed in the NM and to guide migrating GnRH neurons toward the forebrain (Schwarting et al., 2006;Toba et al., 2008). CXCL12 exerts its functions by binding to CXCR4. In the developing brain, CXCL12 and its receptor CXCR4 are highly expressed in the neocortex, cerebellum, and hippocampus, and regulate cell migration and axonal pathfinding (Ma et al., 1998;Chalasani et al., 2003;Tran and Miller, 2003). Although the CXCR4 gene is downregulated in most brain regions postnatally, persistent expression in the hypothalamus and CXCL12/CXCR4-mediated modulation of hypothalamo-pituitary systems points to a role of this receptor in neuroendocrine regulation Li and Ransohoff, 2008;Rostène et al., 2011).
GnRH neurons express CXCR4 early in their development, as they leave the VNO to begin their migration to the brain. CXCR4-deficient mice exhibit impaired GnRH neuron migration and a significant decrease in their total number (Schwarting et al., 2006;Toba et al., 2008), pointing to the importance of CXCL12 signaling in this system. CXCR4 was long considered the only receptor for CXCL12, until it was shown recently that it binds with higher affinity to an additional receptor, CXCR7 (Balabanian et al., 2005). While signaling properties of CXCR7 are presently controversial (Thelen and Thelen, 2008;Rajagopal et al., 2010;Hoffmann et al., 2012;Kumar et al., 2012;Odemis et al., 2012), it appears to act as a CXCL12 scavenger controlling chemokine levels in the extracellular environment by ligand sequestration (Boldajipour et al., 2008; Naumann et al., 2010; Hoffmann et al., Figure 5. Migration of GnRH neurons is altered in CXCR7-null mice. Sagittal sections (A, B) through the nose at E12.5, stained for GnRH, show that cells tend to remain in the area of the VNO (B, arrowhead) in CXCR7 Ϫ/Ϫ mice whereas, in the corresponding controls, they appear to cross the cribriform plate (arrow) and migrate toward the forebrain (FB). Black dots separate the nasal area from the brain. D, E, A similar picture is evident at 14.5, with most GnRH neurons still migrating in the nasal area in CXCR7-null mice (E, arrow). Additionally, some cells are still present in the VNO and, abnormally, in the OE (E, arrowheads). The relative absence of GnRH neurons in the forebrain is indicated with a triangle. G, H, Coronal sections stained for GnRH reveal that, in CXCR7 Ϫ/Ϫ mice, migrating cells deviate from their normal chain-like formation (G) and, instead, appear in the OE and RE (arrowheads in H and L). Further, they form abnormal multicellular clusters in the NM (rectangle in M and at higher magnification in the inset in M ) and, sometimes, in the lumen of the VNO (L, arrow). N, GnRH cells in the lumen of the VNO (arrowhead) show cleaved caspace-3 in a contiguous section, suggesting they are destined to die. J, K, However, guiding axons stained for peripherin appear similar in both genotypes. C, F, I, Compartmental quantification of GnRH cells in the three ages examined showed significant increase of neurons in the nose and a concomitant decrease in the brain of CXCR7 Ϫ/Ϫ mice compared with WT. Moreover, while the total number of cells was similar in the two groups at E12.5 and E14.5, it declined in the KO mice at E16.5 (I ). *p Ͻ 0.05, **p Ͻ 0.01, ***p Ͻ 0.001; Scale bars: (in A) A, B, 150 m; D, E, G, H 2012). In this context, loss of CXCR7 leads to extracellular accumulation and perturbed gradients of the chemokine (Boldajipour et al., 2008;Sánchez-Alcaniz et al., 2011). Similar to CXCL12 and CXCR4, CXCR7 is expressed in the developing hematopoietic system, cardiovascular system, and brain, and CXCR7-deficient mice die perinatally with cardiovascular defects (Sierro et al., 2007).
Here, we provide evidence that CXCR7 is important in the early steps of GnRH neuron migration, as lack of this receptor results in aberrant accumulation of these cells in the nasal cavity. The defects are associated with severe reduction and perturbed subcellular distribution of CXCR4 protein in cells throughout the nasal area, including GnRH neurons.

CXCR7 expression in the nasal area
The first indication that CXCR7 may be involved in GnRH neuron migration came from its expression at the start and along the well defined migratory path of these cells in the nose. Specifically, we identified CXCR7 expression within and around the VNO, where GnRH neurons emerge. Its expression in the more dorsal parts of the NM was significantly reduced up to the cribriform plate, where GnRH neurons enter the brain. Given that CXCR7 sequesters CXCL12 from the extracellular environment (Boldajipour et al., 2008;Naumann et al., 2010;Hoffmann et al., 2012), we may suggest that its pattern of expression may help form a ventrodorsally increasing chemokine gradient. An additional mechanism that could form a CXCL12 gradient is the ventrodorsally increasing CXCL12 expression within the nasal area (Schwarting et al., 2006). Our observation that GnRH neurons as well as the supporting axons of the olfactory/vomeronasal nerve do not express CXCR7, is consistent with the established role of this receptor in shaping CXCL12 gradients. Interestingly, CXCR7 was also expressed in the RE and OE, tissues that are normally avoided by migrating GnRH neurons. Although the RE does not express CXCL12 mRNA, it contained strong RFP signal in transgenic mice expressing CXCL12-RFP fusion protein under the Figure 6. GnRH neuron migration in CXCL12 Ϫ/Ϫ , CXCR4 Ϫ/Ϫ , and CXCR7 Ϫ/Ϫ mice at E14.5. A-D, In the nose, GnRH neurons migrate in a chain-like formation in WT animals but, in KOs, they form clusters (B, arrowheads), and are often found abnormally in the lumen of the VNO (C, arrowhead) or in the epithelium (D, arrowhead). F-I, GnRH staining of coronal sections at the level of the MPOA shows the distribution of these cells in WT animals (F ) and the relative reduction in numbers in the CXCL12 Ϫ/Ϫ (G), CXCR4 Ϫ/Ϫ (H ), and CXCR7 Ϫ/Ϫ (I ) mice. K-M, Compartmental quantification of GnRH neurons shows an overall decrease in CXCL12 Ϫ/Ϫ and CXCR4 Ϫ/Ϫ , but not in CXCR7 Ϫ/Ϫ animals, and a significant reduction in both nose and brain when compared with control animals. N, Spread/distribution analysis of GnRH neurons in the nasal area, divided in two compartments, epithelium and NM, show that 22% of all cells in the nasal area are abnormally localized in the epithelium of CXCL12 Ϫ/Ϫ animals. The respective percentage is 28% in CXCR7 Ϫ/Ϫ and 8.5% in CXCR4 Ϫ/Ϫ mice. No GnRH neurons are found in the epithelium in control animals. E, J, Schematics show the nasal compartment (NC) and the medial preoptic area (MPOA) in the forebrain (FB) as photographed in A-D and F-I, respectively. K-N, *p Ͻ 0.05, **p Ͻ 0.01,***p Ͻ 0.001. Scale bars: (in A) A-D, F-I, 100 m. FB, forebrain; LV, lateral ventricle.