Slit2 and Robo3 modulate the migration of GnRH-secreting neurons

Reproduction in mammals is regulated by hypothalamic gonadotropin-releasing hormone (GnRH) neurons. In the adult rodent brain, these cells are dispersed between the olfactory bulbs and the medial preoptic area (MPOA) and project to the median eminence (ME), where they secrete GnRH into the portal vessels of the pituitary to induce the release of gonadotropins into the circulation (Merchenthaler et al., 1984). Deficiency in GnRH is the cause of hypogonadotropic hypogonadism and Kallmann syndrome, which are characterised by delayed or absent puberty and infertility (Hardelin, 2001).

In rodents, GnRH neurons originate in the nasal placode, which gives rise to the olfactory epithelium (OE) and the vomeronasal organ (VNO), and migrate towards the forebrain apposed to olfactory (OLF) and vomeronasal (VN) axons. Once in the forebrain, they make a turn towards the developing hypothalamus, guided by the caudal branch of the VN nerve. Throughout their migratory journey, which in mice starts at embryonic day (E) 10.5 and is complete by birth, many factors function as guidance cues. Thus, we found that Sema3A is important for the guidance of VN axons and, therefore, the migration of GnRH neurons into the brain, whereas reelin, present in the extracellular matrix of the olfactory bulbs deflects the course of migrating GnRH neurons into the ventral forebrain (Cariboni et al., 2011; Cariboni et al., 2005). Other authors have reported the importance of guidance molecules such as Sema4D, Sema7A, Hgf, netrin and SDF-1 (Cxcl12) in this process (Giacobini et al., 2007; Giacobini et al., 2008; Messina et al., 2011; Schwarting et al., 2004; Schwarting et al., 2006) by acting predominantly as directional cues in the nasal compartment (NC). By contrast, little is known about the factors that act at the final stages of migration in the MPOA (for reviews, see Cariboni et al., 2007a; Wray, 2010).

The secreted Slit glycoproteins (Slit1, Slit2 and Slit3) and their Roundabout receptors (Robo1, Robo2, Robo3) were originally identified as important axon guidance molecules (Ypsilanti et al., 2010; Kidd et al., 1998; Sundaresan et al., 1998; Yuan et al., 1999). During assembly of the nervous system, Slits function mainly as repulsive cues by preventing axons from growing to inappropriate locations. In addition, Slit-Robo interactions modulate neuronal migration, cell death and angiogenesis (Chédotal, 2007). Previous studies have established roles for Slit-Robo in the development of the olfactory system, but no function in the migration of GnRH neurons has been proposed, despite their intimate association with the OLF and VN tracts. Accordingly, in situ hybridisation studies in rat embryos have shown that olfactory receptor neurons in the OE and migrating cells resembling GnRH neurons express high levels of Robo2 at E12-13 and E13-15, respectively (Marillat et al., 2002). Furthermore, genetic studies in mice have implicated Slits and Robos in the targeting of OLF (Cho et al., 2007; Nguyen-Ba-Charvet et al., 2008) and VN (Cloutier et al., 2004; Prince et al., 2009) axons throughout development.

Using genetically altered mouse models, we provide evidence for a direct role for Slit2 and Robo3 in the migration of GnRH neurons. Specifically, we found that mice lacking Slit2 contained fewer GnRH neurons in the forebrain compared with wild-type controls, resulting in much reduced innervation of the ME. However, these animals did not show defects in the fasciculation/targeting of their OLF/VN axons. Unexpectedly, we found that this defect is independent of Robo1 and Robo2, but that it closely phenocopies Robo3 null mice, suggesting that Slit2 might function through Robo3 to modulate the migration of GnRH neurons.

All animal procedures were performed in accordance with institutional and UK Home Office guidelines. Slit- and Robo-deficient mice (all in C57BL/6 inbred strain) were described previously (Plump et al., 2002; Grieshammer et al., 2004; Long et al., 2004; Sabatier et al., 2004).

Immunofluorescence on paraformaldehyde-fixed cryosections was as described previously (Cariboni et al., 2007b). Primary antibodies used were: rabbit anti-peripherin (1:500; Merck Millipore), anti-GnRH (1:400; Immunostar), anti-Robo1 (1:150; gift from Prof. F. Murakami, Osaka University, Japan), anti-Robo3 (R&D Systems) and chicken anti-GFP (1:500; Aves Laboratories), followed by Alexa Fluor-conjugated secondary antibodies (Invitrogen). For immunoperoxidase staining, sections were processed as described previously (Cariboni et al., 2011). The total number of GnRH neurons per head or forebrain was determined by counting all GnRH-positive cells in all sections through each embryo head/brain. We analysed at least three samples for each genotype. The density of GnRH neuron projections to the ME was calculated with ImageJ (NIH).

Single-stranded cDNA was synthesized with AMV reverse transcriptase and random hexamers (Promega) and subjected to PCR with the following primers (5′-3′): Slit2, GCGCGTCTGGTGTGAATGAA and CACAGT - GGCACCAGGAGCAT (product size 210 bp); Robo1, TGAAGCTT - CCGCTACTTTGAC and GTTACTGATGTGATTGCAGACC (338 bp); Robo2, CAGGGCCGGACAGTGACATTC and CTTGGGGGTTGA - TCGCTCTGA (823 bp); Robo3, GTGTGGCTCGCAACTACCTG and AGGATCACCATACGTGGAGG (229 bp); Gapdh, TGGCATTGTGGA - AGGGCTCATGAC and ATGCCAGTGAGCTTCCCGTTCAGC (188 bp).

GN11 and COS-7 cells were cultured as previously described (Cariboni et al., 2007b). Preparation of conditioned media from Slit2-expressing COS-7 cells and chemotactic assays were performed as described (Hernández-Miranda et al., 2011; Cariboni et al., 2007b). Rabbit anti-Robo antibodies (R&D Systems) were added to the GN11 cell suspension before the assay (1:100, for 1 hour at 37°C).

In all studies, we calculated the mean of at least three independent experiments; data are expressed as mean ± s.e.m. To determine statistical significance, we used a paired t-test or, for multiple comparisons, one-way ANOVA and Bonferroni post-hoc test. Statistical analysis was performed using Prism4 (GraphPad Software).

Using transgenic mice carrying GFP in the deleted Slit1 or Slit2 allele (Plump et al., 2002), we studied Slit expression during GnRH neuron development. Sagittal sections of heterozygous Slit1 and Slit2 mouse embryos were stained with GFP, Tuj1 (Tubb3), GnRH (Gnrh1) and peripherin antibodies to reveal the expression of Slit proteins in relation to GnRH neurons and OLF/VN axons. We chose embryos at E14.5, as GnRH neuron migration is at its peak at this stage, and these cells and their associated axons can be visualised in the NC, nasal-forebrain junction and forebrain (FB). Double immunofluorescence revealed that Slit2 is co-expressed in nasal axons and FB neurons with Tuj1, a marker for newborn neurons and axons (Fig. 1A-C). To better characterise these cells, we performed double immunofluorescence for Slit2/GnRH and Slit2/peripherin, which are markers of GnRH neurons and OLF/VN axons, respectively, and found that Slit2 colabels with GnRH in neurons in the NC (Fig. 1D-F′) and with peripherin in OLF/VN axons (Fig. 1J-L), but not with GnRH in FB neurons (Fig. 1G-I) or with peripherin in the caudal branch of the VN nerve (Fig. 1M-O). Slit1 presents a similar pattern of expression, but is absent from GnRH neurons at E14.5 (supplementary material Fig. S1A-I).

To study the role of Slit1 and Slit2 in the development of GnRH neurons, we analysed Slit1 and Slit2 single mutants and compound null mice at different embryonic stages. Sagittal sections taken from heads of E14.5 Slit1, Slit2, Slit1/Slit2 null and wild-type littermates were stained for GnRH. No statistically significant differences in the total number of GnRH neurons were observed between the groups, thus excluding a role for Slit1 and Slit2 in the proliferation/survival of these cells (wild type, 1203±49.15; Slit1^–/–, 1120±18.34; Slit2^–/–, 1255±47.79; Slit1^–/– Slit2^–/–, 1250±32.69; n=3; P>0.05, one-way ANOVA). A similar result was obtained following analysis of mice at E12.5 (wild type, 1042±25.41; Slit1^–/–, 1023±26.39; Slit2^–/–, 1001±9.866; P>0.05, one-way ANOVA). Because of high expression of Slit1 and Slit2 on the axons that guide migrating GnRH neurons to the FB (Fig. 1; supplementary material Fig. S1), we also analysed the appearance of OLF/VN axons. Examination of peripherin-stained sagittal or coronal sections revealed no differences between genotypes in their patterning/fasciculation at either E12.5 or E14.5 (supplementary material Fig. S2).

We then counted GnRH neurons separately in the NC and FB, and observed differences in their distribution in Slit2^–/– mice. Specifically, we found an accumulation of cells in the NC compared with wild-type animals at E12.5 (wild type, 733.0±38.19; Slit2^–/–, 910.0±9.713; n=3; P<0.05, t-test), resulting in a significantly reduced number of cells in the FB (wild type, 237.0±27.43; Slit2^–/–, 94.00±3.786; n=3; P<0.01, t-test) (Fig. 2A-D). Slit2 expression in GnRH neurons was also found at E12.5 (Fig. 3A-C), whereas Slit1 was expressed only by a small number of cells at this stage (Fig. 3D-F), suggesting a possible autocrine role for Slit2. GnRH neuron accumulation in the NC and reduction in the FB of Slit2^–/– mice were also observed at E14.5 (NC: wild type, 578.0±14.22; Slit2^–/–, 3912.3±10.11; n=3; P<0.01, t-test; FB: wild type, 675.0±90; Slit2^–/–, 317.7±18.81; n=3; P<0.05, t-test) (Fig. 2E-K). No differences were found in Slit1^–/– null mice at E14.5 (supplementary material Fig. S3A,B), suggesting that Slit1 does not play a role in GnRH neuron migration.

A significantly reduced number of GnRH neurons was also observed in the MPOA of Slit2^–/– mice at E18.5 (wild type, 855.3±25.86; Slit2^–/–, 398.3±36.06; n=3; P<0.05, t-test) (Fig. 2L-O) accompanied by markedly reduced GnRH innervation of the ME (Fig. 2P-S). Analysis of Slit1/Slit2 compound null mice revealed a similar reduction in the number of cells in the FB at E14.5 (wild type, 587.0±47.35; Slit1^–/– Slit2^–/–, 311.7±22.42; n=3; P<0.01, t-test) and E18.5 (wild type, 682.0±45.71; Slit1^–/– Slit2^–/–, 344±20.55; n=3; P<0.001, t-test), confirming that Slit1 is dispensable for the migration of GnRH neurons.

We also analysed mice lacking Slit3 at E14.5 (n=3), although expression of this molecule in the rodent olfactory system has not been reported (Marillat et al., 2002). This analysis revealed no statistically significant differences in the number or position of GnRH neurons compared with wild-type controls (GnRH⁺ neurons/head: wild type, 1320±35.79; Slit3^–/–, 1290±75.21; n=3; GnRH⁺ neurons/FB: wild type, 699.0±34.22; Slit3^–/–, 696.0±81.09, n=3) (supplementary material Fig. S3A-C,F), excluding its involvement in this system. Accordingly, a similar distribution of GnRH neurons in the MPOA and innervation of the ME were noted in the two groups of mice at E18.5 (supplementary material Fig. S4D,E,G,H).

Robo1 and Robo2 are the predominant Slit receptors in the CNS (Sundaresan et al., 2004), including the olfactory system (Marillat et al., 2002), but no binding preference has been described between the different Slits and these receptors (Mambetisaeva et al., 2005). We performed RT-PCR on RNA isolated from FACS-sorted mouse GFP-GnRH neurons at E14.5 as described previously (Cariboni et al., 2007b) and found expression of Slit2, as well as of Robo1 and Robo2, in these cells (Fig. 3G). Expression of Robo2 mRNA in neurons leaving the OE has been reported in E15 rats (Marillat et al., 2002), and here we confirmed the expression of Robo1 in GnRH neurons by immunostaining coronal sections from E15.5 GFP-GnRH rat embryos with GFP and Robo1 antibodies (Fig. 3H).

We subsequently assessed the number and distribution of GnRH neurons in Robo1^–/–Robo2^–/– mice at E16.5. Counts of labelled cells in coronal sections taken through the FB showed no difference between genotypes (wild type, 890±18.02; Robo1^–/– Robo2^–/–, 921±36.47; n=3; P>0.05, t-test) (Fig. 3I-K), suggesting that Slit2 acts independently of these receptors.

A previous study reported that Robo3 is expressed by populations of tangentially migrating neurons, including GnRH cells (Marillat et al., 2004). We confirmed this finding both by RT-PCR of RNA from E14.5 GFP-GnRH neurons and by immunofluorescence (Fig. 4A,B). We then analysed Robo3^–/– mice and found, similar to Slit2^–/– animals, a significantly reduced number of GnRH neurons in the FB and a concomitant accumulation of cells in the NC compared with wild-type littermates at E14.5 (wild type, 658.0±31.90; Robo3^–/–, 343.7±46.36; n=3; P<0.001, t-test) (Fig. 4C-F). Also, as in Slit2^–/– mice, there were no apparent defects in the fasciculation/patterning of the OLF/VN nerves (supplementary material Fig. S3C-D′).

To confirm that Slit2 acts through Robo3 to guide GnRH neurons in the NC, we performed in vitro chemotactic assays using immortalised GnRH neurons (GN11 cell line). We have previously reported that GN11 cells express Robo1-3 and respond to Slit1 (Hernández-Miranda et al., 2011), and may thus be used to test the effects of Slit2. As shown in Fig. 4, the chemomigration of GN11 cells is reduced in the presence of Slit2-conditioned media (CM) (Fig. 4H) compared with mock CM from COS-7 cells (Fig. 4G). In addition, this effect was blocked by pre-incubation with Robo3 antibody (Fig. 4J,K), but not by Robo1 and Robo2 antibodies (Fig. 4I), suggesting that Slit2 exerts its effects specifically through Robo3.

Taken together, our data suggest that Slit2 and Robo3 are important in the development of the GnRH neuronal system, as lack of these molecules alters the migration of GnRH cells, resulting in their abnormal accumulation in the NC. The similarity of the phenotypes observed in the Slit2^–/– and Robo3^–/– mice, together with the results of the chemotactic assays, suggest that this ligand-receptor pair modulates the migration of GnRH neurons, as reported for cancer cells (Bauer et al., 2011). However, the possibility of interaction with other signalling pathways cannot be excluded. For example, genetic data from C. elegans indicate that Slits might function through DCC (Deleted in colorectal cancer) (Yu et al., 2002), a molecule found to be important in the migration of many neuronal cell types, including GnRH neurons. Mice lacking Dcc or its ligand netrin 1 show a defect in the migration of GnRH cells similar to that reported here (Schwarting et al., 2001; Schwarting et al., 2004), supporting the possibility of cross-talk between Slit2 and netrin 1 signalling in this system. In conclusion, our results highlight the importance of Slit2 and Robo3 in the migration of GnRH neurons in a cell-autonomous manner, and might shed light on the pathogenesis of idiopathic forms of hypogonadotropic hypogonadism and Kallmann syndrome.

We thank Drs M. Tessier-Lavigne and D. Ornitz for supplying Slit mutant mice; Roberta Noseda and Alessandro Fantin for technical assistance; and Drs Sonja Rakic and Nadia Liapi for discussion of results.