Neural crest Notch/Rbpj signaling regulates olfactory gliogenesis and neuronal migration†

Summary The neural crest‐derived ensheathing glial cells of the olfactory nerve (OECs) are unique in spanning both the peripheral and central nervous systems: they ensheathe bundles of axons projecting from olfactory receptor neurons in the nasal epithelium to their targets in the olfactory bulb. OECs are clinically relevant as a promising autologous cell transplantation therapy for promoting central nervous system repair. They are also important for fertility, being required for the migration of embryonic gonadotropin‐releasing hormone (GnRH) neurons from the olfactory placode along terminal nerve axons to the medial forebrain, which they enter caudal to the olfactory bulbs. Like Schwann cell precursors, OEC precursors associated with the developing olfactory nerve express the glial marker myelin protein zero and the key peripheral glial transcription factor Sox10. The transition from Schwann cell precursors to immature Schwann cells is accelerated by canonical Notch signaling via the Rbpj transcription factor. Here, we aimed to test the role of Notch/Rbpj signaling in developing OECs by blocking the pathway in both chicken and mouse. Our results suggest that Notch/Rbpj signaling prevents the cranial neural crest cells that colonize the olfactory nerve from differentiating as neurons, and at later stages contributes to the guidance of GnRH neurons.

Loss-of-function SOX10 mutations are found in roughly one-third of cases of Kallmann's syndrome (combined anosmia and hypogonadotropic hypogonadism) with deafness (Pingault et al., 2013).
Apart from the requirement for Sox10 for normal OEC differentiation (Barraud et al., 2013;Pingault et al., 2013), the molecular mechanisms underlying OEC development have been relatively little studied, especially in comparison with those underlying the development of Schwann cells, the glia of all other peripheral nerves (reviewed by *This paper is dedicated to the memory of Dr. Sophie R. Miller. Jacob, 2015;Jessen, Mirsky, & Lloyd, 2015;Kastriti & Adameyko, 2017). Like all other peripheral glial cells (i.e., Schwann cells and the satellite glia of peripheral ganglia), whose differentiation also requires Sox10 (Britsch et al., 2001), OECs are derived from the embryonic neural crest (Barraud et al., 2010;Forni, Taylor-Burds, Melvin, Williams, & Wray, 2011). The Sox10-expressing 'Schwann cell precursors' associated with embryonic peripheral nerves can be distinguished from their neural crest progenitors by the expression of early glial markers such as myelin protein zero (Mpz, P0) and fatty acid-binding protein 7 (brain lipid-binding protein; brain fatty acid-binding protein) (see Jacob, 2015;Jessen et al., 2015;Kastriti & Adameyko, 2017). Similarly, cells in the OEC lineage can first be identified in the chicken embryo via the onset of immunoreactivity for Mpz at embryonic day (E)3.5-4 (Drapkin & Silverman, 1999;Norgren, Ratner, & Brackenbury, 1992) and also by expression of Sox10 (Barraud et al., 2010), in cells closely associated with the migrating neurons and axons of the olfactory nerve (Drapkin & Silverman, 1999;Norgren et al., 1992). In the mouse, developing OECs can first be identified at E10.5, as Sox10-expressing cells associated with the 'migratory mass' of neurons and olfactory axons (Barraud et al., 2013;Forni et al., 2011).
Immature Schwann cells are both molecularly and phenotypically distinct from Schwann cell precursors: they express, for example, S100, glial fibrillary acidic protein (Gfap) and Desert hedgehog (Dhh), and are able to support their own survival via autocrine signaling (Dong et al., 1999;Meier et al., 1999), whereas Schwann cell precursors depend for their survival on axon-associated neuregulin 1 type III (Dong et al., 1995). The canonical Notch/Rbpj signaling pathway (reviewed by Bray, 2016;Kovall, Gebelein, Sprinzak, & Kopan, 2017) promotes the transition from Schwann cell precursors to immature Schwann cells: the conditional deletion of either Rbpj or Notch1 in Schwann cell precursors using a Dhh-Cre line (Jaegle et al., 2003) delays this transition, while driving expression of the Notch intracellular domain to activate Notch signaling accelerates the transition (Woodhoo et al., 2009).
Here, we aimed to test the role of Notch/Rbpj signaling in developing OECs by using the Tol2 transposase/'Tet-on' electroporation system (Sato et al., 2007;Watanabe et al., 2007) to insert a doxycyclineinducible dominant negative Rbpj construct (Kohyama et al., 2005;Sato et al., 2008) into the genome of chicken cranial neural crest cells, and by using an Mpz-Cre driver line (Feltri et al., 1999(Feltri et al., , 2002 combined with an Rbpj flox/flox line (Tanigaki et al., 2002) to delete Rbpj in mouse OECs. Our chicken experiments showed that Notch/Rbpj signaling is required in the cranial neural crest-derived cells that colonize the olfactory nerve to prevent them from differentiating as neurons, while our mouse experiments revealed defects in GnRH neuron localization when Notch/Rbpj signaling was blocked in OECs.

| R E S U LTS
2.1 | Blocking Notch/Rbpj signaling in cranial neural crest-derived cells with temporal control In order to block Notch/Rbpj signaling in cranial neural crest-derived cells with temporal control, we used the Tol2 transposase/"Tet-on" electroporation system (Sato et al., 2007;Watanabe et al., 2007) to integrate doxycycline-inducible constructs into the genome of cranial neural crest cells. To achieve maximum electroporation efficiency, we performed electroporation ex ovo and then grafted targeted midbrain and caudal forebrain neural folds (containing the premigratory neural crest cells that contribute to the frontonasal mesenchyme and OECs; Barraud et al., 2010) into untargeted host embryos.
The electroporated embryos were cultured according to the Early Chick (EC) culture method (Chapman, Collignon, Schoenwolf, & Lumsden, 2001) for 9-10 hours until cranial neural folds had developed (roughly 2-5 somites; HH71 to 81), but before neural crest emigration had started (midbrain-level neural crest cells are the first to emerge, from the 6-somite stage; Tosney, 1982) (Figure 1a-a 2 ). Midbrain and caudal forebrain neural folds were grafted bilaterally from electroporated donor embryos into wild-type hosts in ovo (Figure 1b-b 2 ) to ensure that, upon final analysis, only donor cranial neural fold-derived cells, including neural crest cells, expressed the Tol2-integrated constructs. To initiate expression of DN-Rbpj/EGFP (and thus block Notch/ Rbpj signaling), the host eggs were injected with doxycycline at embryonic day (E)3.5, when chicken OEC precursors can be detected via the onset of immunoreactivity on the olfactory nerve for the early glial marker myelin protein zero (Mpz, P0) (Drapkin & Silverman, 1999). 2.2 | When Notch/Rbpj signaling is blocked from E3.5 in cranial neural crest-derived cells, most targeted cells on the olfactory nerve form neurons Two days after doxycycline injection, at E5.5 (n 5 2), most control EGFP-targeted neural crest-derived cells on the olfactory nerve expressed the OEC marker Sox10, as expected (Barraud et al., 2010) ( Figure 2a-b 3 ). [The presence of Sox10-negative EGFP-targeted cells could reflect the formation of non-glial neural crest derivatives, such as endoneurial fibroblasts (Joseph et al., 2004) and/or perivascular cells, which are neural crest-derived in the region of the forebrain (Etchevers, Vincent, Le Douarin, & Couly, 2001); indeed, the perivascular cell marker Pdgfrb is detectable in some cells on the olfactory nerve at E6.5 (Miller, Perera, & Baker, 2017).] Immunostaining for the neuronal RNA- Overall, these data show that blocking Notch/Rbpj signaling in cranial neural crest-derived cells for two days from E3.5 (when Mpzpositive cells in the OEC lineage can first be identified on the olfactory nerve; Drapkin & Silverman, 1999) resulted in a majority of the targeted cells on the olfactory nerve adopting a neuronal fate (distinct from migrating GnRH neurons, which are olfactory placode-derived; Barraud et al., 2010;Sabado, Barraud, Baker, & Streit, 2012). This suggests that Notch/Rbpj signaling is required to maintain the glial identity of developing OECs, at least from E3.5 to E5.5.
FIG URE 1 Ex ovo electroporation and neural fold grafting to target cranial neural crest precursors. (a-a 2 ) An EGFP-targeted embryo that was electroporated ex ovo at HH5 (head process stage) and cultured in EC culture (Chapman et al., 2001) to reach the 4 somite-stage (HH8). Almost the entire cranial ectoderm expresses EGFP, including the cranial neural folds containing premigratory neural crest cells. Inset shows a schematic of the ex ovo electroporation procedure [modified from "Electroporation for early chick embryos using New's culture (gastrula), Application note 1"; http://www.sonidel.com/sonidel/in-ovo-electroporation/]. For a detailed description, see Section 4. (b-b 2 ) A 5 somite-stage (HH81) embryo in ovo, photographed immediately after receiving a bilateral neural fold graft (at the level of the caudal forebrain and midbrain) from an EGFP-targeted donor embryo. (c-c 2 ) Ventral view of an embryo that had been similarly grafted with DN-Rbpj/ EGFP-targeted donor neural folds and allowed to develop until E5.5, two days after doxycycline injection at E3.5. EGFP-positive neural crest-derived cells are present in the frontonasal process, including near the nasal pits. E, embryonic day; fb, forebrain; mb, midbrain; nf, neural folds; np, nasal pits; ss, somite-stage MILLER ET AL. | 3 of 12 FIG URE 2 Blocking Notch/Rbpj signaling in cranial neural crest-derived cells from E3.5 results in neurogenesis on the olfactory nerve. Parasagittal sections of chicken embryos at E5.5 that had received grafts of midbrain and caudal forebrain neural folds from ex ovo-electroporated donor embryos at E1.5, with doxycycline injection at E3.5. (a-a 3 ) Control EGFP-targeted embryo, showing ISH for Sox10 and immunostaining for EGFP and Tubb3 (neuronal class III beta-tubulin). EGFP-targeted cells are seen in the neural folds, frontonasal mass and on the olfactory nerve. 2.3 | When Notch/Rbpj signaling is blocked in mouse OECs, more GnRH neurons are found in the lateral regions of the olfactory bulbs We also investigated the effect of blocking Notch/Rbpj signaling on mouse OEC development, by analyzing E16.5 mouse embryos from crosses between the transgenic Mpz-Cre (myelin protein zero-Cre; P0-Cre) driver line (Feltri et al., 1999(Feltri et al., , 2002 and homozygous Rbpj flox/flox mice (Tanigaki et al., 2002). Rat OECs express Mpz from E13 through to adulthood (Lee et al., 2001). We did not find any evidence for ectopic neurogenesis on the olfactory nerve in Mpz-Cre;Rbpj flox/flox mutant mouse embryos at E16.5 (data not shown). However, Cre recombination in the Schwann cell lineage in Mpz-Cre mouse embryos does not occur until between E13.5-E14.5 (Yu et al., 2005;Woodhoo et al., 2009), when Schwann cell precursors are transitioning to immature Schwann cells (Dong et al., 1999). This is considerably later than the onset of Mpz expression in Schwann cell precursors (Jessen et al., 2015). This could explain why we did not detect the neurogenesis phe- showed this difference to be highly statistically significant (p < 0.0001; Figure 4i).
Overall, our data suggest that at early stages of OEC development, Notch/Rbpj signaling is required to block adoption of a neuronal fate, while at later stages, activation of this pathway contributes to the guidance of GnRH neurons. showing the percentage per embryo of all centrally located Gnrh1-positive cells that were found laterally at E16.5, in control Rbpj flox/flox embryos (mean 1.9 6 0.5%; n 5 4 embryos from 2 litters; 176-230 central GnRH neurons counted per embryo; 3-6 of these per embryo were located laterally) versus Mpz-Cre;Rbpj flox/flox mutant embryos (mean 10.6 6 1.2%; n 5 4 embryos from 2 litters; 160-258 central GnRH neurons counted per embryo; 17-23 of these per embryo were located laterally). The mean is 5.5-fold higher for Mpz-Cre;Rbpj flox/ flox mutants (p < 0.0001; unpaired two-tailed Student's t-test; t 5 13.44, 6 degrees of freedom). cp, cribriform plate; E, embryonic day; ns, nasal septum; ob, olfactory bulb; oe, olfactory epithelium; on, olfactory nerve; onl, olfactory nerve layer. Scale-bars: 100 lm in all panels except insets in e-g: 20lm These results suggest that Notch/Rbpj signaling in the cranial neural crest-derived frontonasal mesenchyme cells that colonize the olfactory nerve is required, at least between E3.5 and E5.5, to prevent them from adopting a neuronal fate on the olfactory nerve. Neural crest cells in the frontonasal mass originate from the rostral midbrain and caudal forebrain; these neural crest cells migrate rostrally to populate the frontonasal mass (Noden, 1975). This population of neural crest cells does not normally form neurons (Narayanan & Narayanan, 1978;Noden, 1975), but when grafted adjacent to the rostral hindbrain, will contribute neurons to the trigeminal ganglion (Baker, Bronner-Fraser, Le Douarin, & Teillet, 1997;Noden, 1975). Taken together, these results suggest that Notch/Rbpj signaling is required to prevent neuronal differentiation by the neural crest-derived cells that colonize the olfactory nerve.
In contrast to the results obtained in chicken, we did not detect ectopic neurogenesis on the mouse olfactory nerve at E16.5 when

| OEC precursors are multipotent
Previously, we used the Tol2 transposase/'Tet-on' in ovo electroporation system (Sato et al., 2007;Watanabe et al., 2007) to drive NotchDE, encoding a constitutively active form of mouse Notch1 (Kopan, Schroeter, Weintraub, & Nye, 1996;Sato et al., 2008), in cranial neural crestderived cells from E4 (Miller et al., 2017). This proved to be sufficient to convert both frontonasal mesenchyme cells, and perhaps also developing OECs, to a different cranial neural crest cell fate, namely Pdgfrbpositive perivascular cells (Miller et al., 2017). Here, we found that most targeted cranial neural crest-derived cells on the olfactory nerve formed neurons when Notch/Rbpj signaling was blocked from E3.5.
The competence of OEC precursors to adopt different fates in response to Notch signaling activation or repression is at least partly reminiscent of the multipotent 'Schwann cell precursors' found on other peripheral nerves: during normal development, these form not only immature Schwann cells but also endoneurial fibroblasts (Joseph et al., 2004), melanocytes (Adameyko et al., 2009)

| Notch/Rbpj signaling in embryonic OECs contributes to GnRH neuron guidance
In rodents, hypothalamic GnRH neurons migrate along terminal nerve axons, which segregate as a ventro-caudally oriented branch from the vomeronasal nerve that projects along the medial olfactory bulbs and enters the medial forebrain caudal to the olfactory bulbs, terminating in septal and preoptic areas (Geller et al., 2013;Schwanzel-Fukuda & Pfaff, 1989;Schwanzel-Fukuda, 1999;Taroc et al., 2017;Yoshida et al., 1995). The terminal nerve is a plexiform, ganglionated nerve containing distinct subpopulations of neurons, including GnRH neurons and neuropeptide Y-immunoreactive neurons (for reviews, see e.g. Demski, 1993;Larsell, 1950;Ma, Fleischer, Breer, & Eisthen, 2015;Schwanzel-Fukuda, 1999;von Bartheld, 2004;Wirsig-Wiechmann, Wiechmann, & Eisthen, 2002). The terminal nerve is thought to be neuromodulatory on the olfactory epithelium and, at least in teleost fish, on the retina (for a comprehensive recent review of functional studies, see Ma et al., 2015). GnRH neurons persist into adulthood along the course of the terminal nerve, including in the largest of the terminal nerve ganglia, the 'ganglion terminale' located on the medio-caudal aspect of the olfactory bulb (Schwanzel-Fukuda, 1999).
In the mouse embryo, OECs are intimately associated with GnRH neurons throughout their migration (Geller et al., 2013(Geller et al., , 2017. We and others previously showed that defective OEC differentiation in Sox10null mice results in a significant reduction in the proportion of GnRH neurons entering the forebrain, as well as defects in olfactory axon targeting (Barraud et al., 2013;Pingault et al., 2013). Here, we found that deleting Rbpj in OECs using an Mpz-Cre driver line (Feltri et al., 1999(Feltri et al., , 2002Tanigaki et al., 2002) had no effect on the proportion of GnRH neurons entering the forebrain, but led to a roughly five-fold increase in the mean percentage of GnRH neurons located in lateral rather than medial regions of the olfactory nerve layer and olfactory bulbs at E16.5 (from 1.9 6 0.5% to 10.6 6 1.2% of all centrally located GnRH neurons counted).
A recent study using GnRH immunostaining and 3D imaging of solvent-cleared organs (3DISCO; Ert€ urk et al., 2012) described a singlecell-thick "ring" of GnRH neurons around the olfactory bulbs in both human and E16 mouse embryos (Casoni et al., 2016). In adult mice, terminal nerve GnRH neurons were previously described as including "an arborizing network of cell bodies at the level of the central portion of the olfactory bulb", in connection with the ganglion terminale (Jennes, 1986).
It is plausible that the few Gnrh1-positive cells we identified on sections in the lateral olfactory bulbs of control embryos at E16.5 are part of this "ring". The roughly five-fold increase in the number of Gnrh1-positive cells located laterally in Mpz-Cre;Rbpj flox/flox embryos may suggest that Notch/Rbpj signaling in OECs somehow helps to prevent more GnRH neurons from migrating laterally around the olfactory bulbs to join this "ring". In this context, it is interesting that terminal nerve axons and olfactory/vomeronasal axons express different guidance cue receptors (Taroc et al., 2017), and that OECs located in different regions of the main olfactory system (peripheral olfactory nerve, outer olfactory nerve layer, inner olfactory nerve layer) show at least some molecular and phenotypic heterogeneity (reviewed by Ekberg et al., 2012;Ekberg & St John, 2015). It is possible, therefore, that molecular differences between terminal nerve OECs and olfactory/vomeronasal nerve OECs -potentially involving Notch/Rbpj signaling -somehow contribute to embryonic GnRH neuron guidance, restricting most GnRH neurons to the medial pathway. Testing this speculative hypothesis must await a more detailed spatiotemporal characterization of OECs at the molecular level.

| Conclusions
Overall, we conclude that Notch/Rbpj signaling plays different roles during OEC development. Our chicken electroporation data suggest that Notch/Rbpj signaling prevents neuronal differentiation by the cranial neural crest-derived cells that colonize the olfactory nerve, while our conditional mouse mutant analysis suggests that Notch/Rbpj signaling in OECs later helps to restrict most migrating GnRH neurons to medially coursing terminal nerve axons.

| Embryos
Fertilized chicken (Gallus gallus domesticus) eggs were obtained from Winter Egg Farm (Royston, Hertfordshire, UK). All work with chicken embryos was conducted in accordance with the UK Animals (Scientific Procedures) Act 1986. Experiments using Rbpj flox/flox mice (Tanigaki et al., 2002) and Mpz-Cre (P0-Cre) mice (Feltri et al., 1999(Feltri et al., , 2002 were conducted in accordance with the UK Animals (Scientific Procedures) Act 1986, with appropriate project and personal licenses in place.

| Chicken embryo ex ovo electroporation, in ovo grafting, fixation and histology
Fertilized chicken eggs were incubated in a humidified atmosphere at 388C for 22-24 hours until HH5-6 (head process to head fold stages).
Half were set aside to be the host embryos for neural fold grafts. The remaining embryos were explanted using a Whatman filter paper ring as described (Chapman et al., 2001) and placed in an electroporation chamber (Voiculescu, Papanayotou, & Stern, 2008) containing simple saline solution, with the positive electrode forming the base of the chamber. A 1:1:1 mix of pCAGGS-T2TP, pT2K-CAGGS-rtTA2 S M2 and either pT2K-DN-Rbpj-BI-EGFP or control pT2K-CAGGS-EGFP, at a final concentration of 1.4 lg/ll each, mixed with Fast Green to a final dilution of 2% and sucrose to a final concentration of 8%, was micropipetted over the cranial ectoderm and a "plate"-type negative electrode positioned over the embryo (Figure 1a inset). A Pulse Generator CUY21EX electroporator (BEX Co., Ltd., Japan) was used to apply across the entire blastoderm an initial burst of 50V and five subsequent pulses of 8 V for 75 milliseconds at 50 millisecond intervals, leading to highly efficient targeting of the ectoderm (Figure 1a 1, a 2 ). Each electroporated embryo was placed onto an agar-albumen plate according to the EC culture method (Chapman et al., 2001) and returned to the incubator, together with the remaining unopened eggs, for a further 9-10 hours to reach 2-5 somites. The remaining eggs were opened and 1% black ink (Fount India, Pelikan) in filtered phosphate-buffered saline (PBS) was injected underneath the blastoderm to visualize the embryo.
Electroporated donor and host embryos were not always at precisely the same stage. The neural folds encompassing the caudal forebrain and midbrain were dissected bilaterally from a wild-type host embryo using a pulled glass needle and replaced with the equivalent region from an electroporated donor embryo (Figure 1b-b 2 ). The window was sealed with Parafilm and the egg returned to the incubator. At E3.5, the Parafilm was removed and 500 ml of 100 lg/ll doxycycline (Clontech) injected under the embryo (as described by Sato et al., 2007). The eggs were sealed with Parafilm and returned to the incubator for 2 days before collection and analysis (Figure 1c-c 2 ). Surviving embryos were fixed in modified Carnoy's fixative (6 volumes ethanol, 3 volumes 37% formaldehyde, 1 volume glacial acetic acid), dehydrated into ethanol, cleared in Histosol (National Diagnostics) and embedded in paraffin wax for sectioning at 6 lm on a rotary microtome (Microm).

| Mouse embryo fixation and histology
Embryos were decapitated and immersion-fixed overnight in 4% paraformaldehyde in PBS at 48C. They were cryoprotected by incubating overnight at 48C in 30% sucrose in diethylpyrocarbonate-treated PBS, embedded in O.C.T. compound (Tissue Tek) and flash-frozen in isopentane on dry ice. Ten micrometer sections were taken on a rotary cryostat (Bright Instrument Company).
Digoxigenin-labeled antisense riboprobes were generated as described (Henrique et al., 1995) and in situ hybridization on sections, followed by immunohistochemistry, performed as described (Miller et al., 2017).

| Image capture and processing
Images were captured using a QImaging Retiga 2000R camera and an RGB pancake (QImaging) on a Zeiss AxioSkop 2 MOT compound microscope with QCapture Pro 6.0 software. Images were processed using Adobe Photoshop CS5.1 or CS6.

| Statistical analysis
Initial data analysis was performed using Microsoft Excel. Scatter plots were generated using GraphPad Prism 7 (GraphPad Software, La Jolla, CA), which was also used to check all datasets for normality using the Shapiro-Wilk test, and to compare variances using an F test. Means were compared in GraphPad Prism 7 using an unpaired two-tailed Student's t-test. Rbpj flox/flox mouse embryos. All authors except S.R.M., to whose memory the manuscript is dedicated, read and commented on the manuscript.