Heparan Sulfate Sulfation by Hs2st Restricts Astroglial Precursor Somal Translocation in Developing Mouse Forebrain by a Non-Cell-Autonomous Mechanism

Heparan sulfate (HS) is a cell surface and extracellular matrix carbohydrate extensively modified by differential sulfation. HS interacts physically with canonical fibroblast growth factor (FGF) proteins that signal through the extracellular signal regulated kinase (ERK)/mitogen activated protein kinase (MAPK) pathway. At the embryonic mouse telencephalic midline, FGF/ERK signaling drives astroglial precursor somal translocation from the ventricular zone of the corticoseptal boundary (CSB) to the induseum griseum (IG), producing a focus of Slit2-expressing astroglial guidepost cells essential for interhemispheric corpus callosum (CC) axon navigation. Here, we investigated the cell and molecular function of a specific form of HS sulfation, 2-O HS sulfation catalyzed by the enzyme Hs2st, in midline astroglial development and in regulating FGF protein levels and interaction with HS. Hs2st−/− embryos of either sex exhibit a grossly enlarged IG due to precocious astroglial translocation and conditional Hs2st mutagenesis and ex vivo culture experiments show that Hs2st is not required cell autonomously by CC axons or by the IG astroglial cell lineage, but rather acts non-cell autonomously to suppress the transmission of translocation signals to astroglial precursors. Rescue of the Hs2st−/− astroglial translocation phenotype by pharmacologically inhibiting FGF signaling shows that the normal role of Hs2st is to suppress FGF-mediated astroglial translocation. We demonstrate a selective action of Hs2st on FGF protein by showing that Hs2st (but not Hs6st1) normally suppresses the levels of Fgf17 protein in the CSB region in vivo and use a biochemical assay to show that Hs2st (but not Hs6st1) facilitates a physical interaction between the Fgf17 protein and HS. SIGNIFICANCE STATEMENT We report a novel non-cell-autonomous mechanism regulating cell signaling in developing brain. Using the developing mouse telencephalic midline as an exemplar, we show that the specific sulfation modification of the cell surface and extracellular carbohydrate heparan sulfate (HS) performed by Hs2st suppresses the supply of translocation signals to astroglial precursors by a non-cell-autonomous mechanism. We further show that Hs2st modification selectively facilitates a physical interaction between Fgf17 and HS and suppresses Fgf17 protein levels in vivo, strongly suggesting that Hs2st acts selectively on Fgf17 signaling. HS interacts with many signaling proteins potentially encoding numerous selective interactions important in development and disease, so this class of mechanism may apply more broadly to other biological systems.


Introduction
The corpus callosum (CC) axon tract connects the cerebral hemispheres through the corticoseptal boundary (CSB) in mice and humans and CC malformation is associated with cognitive and neurological conditions in humans (Donahoo and Richards, 2009). Precisely controlled radial glial cell (RGC) somal translo-cation from the ventricular zone (VZ) of the CSB to its pial surface generates midline zipper (MZ) and indusium griseum (IG) astroglial populations required for cerebral hemisphere fusion and subsequent CC axon navigation (Shu and Richards, 2001;Inatani et al., 2003;Shu et al., 2003;Smith et al., 2006;Moldrich et al., 2010;Clegg et al., 2014;Gobius et al., 2016). The movement of RGC astroglial precursors from the glial wedge (GW) to the IG (GW¡IG translocation) forms an astroglial guidepost population that secretes Slit2 to guide CC axons across the telencephalic midline.
Fibroblast growth factors (FGFs) are an evolutionarily ancient family comprising 23 genes in mice and humans of which 15 Fgf20, and Fgf22 in mice) encode "canonical" FGFs that function as paracrine signaling molecules and bind promiscuously to cell surface FGF receptors (FGFRs encoded by Fgfr1-4 in mice) to elicit an extracellular signal regulated kinase (ERK)/mitogen activated kinase (MAPK) response via activating phosphorylation of ERK¡phospo-ERK (pERK). Canonical FGFs are further subdivided into five subfamilies based on phylogeny and Fgf8 subfamily members Fgf8 and Fgf17 are transcribed in the developing CSB in close spatiotemporal proximity posing the question of how they are coordinated (Guillemot and Zimmer, 2011;Ornitz and Itoh, 2015). Under normal conditions, GW¡IG translocation is primarily attributed to Fgf8 and needs to be tightly regulated to ensure that correct numbers of RGCs leave the GW and reach the IG. Deviation above (or below) normal FGF/ERK signaling levels induces too many (or too few) RGCs to translocate with consequent disruption to CC development (Smith et al., 2006;Wang et al., 2012;Clegg et al., 2014;Gobius et al., 2016). Although Fgf17 plays a role in patterning the developing telencephalon, its importance for CC development is less clear and no CC phenotype has been reported in Fgf17 Ϫ / Ϫ embryos (Cholfin andRubenstein, 2007, 2008). Because Fgf8 and Fgf17 are the principal Fgfs transcribed in vicinity of the GW and both activate ERK, mechanisms must exist to keep the total amount of Fgf protein (Fgf8 protein ϩ Fgf17 protein) at the correct level to generate the correct levels of ERK activation for astroglial precursor RGCs to translocate in appropriate numbers.
Heparan sulfate (HS), the carbohydrate component of cell surface and extracellular matrix (ECM) HS proteoglycans, is a negatively charged sulfated polysaccharide that binds canonical FGFs in the ECM to regulate their movement and half-life and also functions as an obligate FGF coreceptor in FGF:FGFR:HS ternary signaling complexes on the cell surface (Guillemot and Zimmer, 2011;Balasubramanian and Zhang, 2016). HS biosynthesis is in two stages, Ext enzymes polymerize uronic acid -glucosamine disaccharides making linear [uronic acid -glucosamine] n HS polymers which are then modified by the enzymatic addition (by HS sulfotransferases, HSTs) or removal (by HS sulfatases, Sulfs) of sulfate groups at specific positions on the disaccharide residues. There are four classes of HST enzymes, Hs2st, Hs3st, Hs6st, and Ndst, each adding sulfate to a specific position; for example, Hs2st only adds sulfate to the carbon atom in position 2 of uronic acid, generating 2-O HS sulfation. Although work in a variety of systems shows that HS itself can play roles both in the transmission of FGF signals through the ECM (non-cell-autonomous role) and the cellular response to FGF (cell-autonomous role), the potential for specific forms of HS sulfation to selectively regulate FGFs by regulating the physical interaction between HS and FGF proteins is much less well understood (Allen et al., 2001;Loo et al., 2001 Qu et al., 2011Qu et al., , 2012Toyoda et al., 2010;Christian, 2012;Zhang H et al., 2012;Ramsbottom et al., 2014;Chan et al., 2015Chan et al., , 2017Balasubramanian and Zhang, 2016).
The HS code hypothesis states that different forms of HS sulfation can encode specific instructions Kreuger et al., 2006). In this study, we discover that 2-O HS sulfation catalyzed by Hs2st functions non-cell autonomously at the developing telencephalic midline to suppress FGF/ERK signaling that drives the somal translocation of astroglial precursors required for normal CC development. We also present evidence that Hs2st plays a selective role by modulating the physical interaction between Fgf17 protein and HS and selectively suppressing Fgf17 protein levels at the CSB.

Materials and Methods
Animals. All mice were bred in-house according to Home Office UK legislation and licenses approved by the University of Edinburgh Ethical Review Committees and Home Office. Embryos analyzed in this study were of either sex. Animal husbandry was in accordance with UK Animals (Scientific Procedures) Act of 1986 regulations. The Hs2st LacZ (Hs2st Ϫ )-null allele comprised a LacZ gene trap vector integrated into the Hs2st locus, the Hs6st1 LacZiresPLAP (Hs6st1 Ϫ )-null allele comprised a LacZiresPLAP gene trap vector integrated in the Hs6st1 locus, and both were genotyped by PCR as described previously (Bullock et al., 1998;Pratt et al., 2006;. For some ex vivo experiments, Hs2st Ϫ /ϩ mice were crossed with mice carrying the TP6.3 tau ()-GFP fusion transgene to generate Hs2st Ϫ / Ϫ and Hs2st ϩ/ϩ embryos with GFP ϩ axons (Pratt et al., 2000). For conditional mutagenesis, floxed Ext1 (Ext1 fl ) or Hs2st (Hs2st fl ) alleles were combined with either Zic4 Cre (septal deletion) or Emx1 CreER (cortical deletion) driver alleles (Inatani et al., 2003;Kessaris et al., 2006;Rubin et al., 2010;Stanford et al., 2010). CreER activity was induced at embryonic day 9.5 (E9.5) by administering tamoxifen (dissolved in corn oil using a sonicator) to pregnant dams by intraperitoneal injection (120 mg/kg dose). Lineages of cells in which Cre was active were visualized using a Rosa26R-floxedstop-EGFP reporter allele (Sousa et al., 2009).
Ex vivo assays. Ex vivo culture experiments were performed essentially as described previously (Niquille et al., 2009) Explants were cultured on nucleopore polycarbonate membranes (Whatman) floating on 1 ml of neurobasal medium (Life Technologies) supplemented with L-glutamine, glucose, and penicillin/streptomycin) at 37°C with 5% CO 2 in a humidified incubator. Brains were dissected from embryos in oxygenated Earle's balanced salt solution (Life Technologies), embedded in lowmelting-point agarose, sliced using a vibratome (Leica VTS-1000), and transferred to modified Eagle's medium (MEM; Life Technologies) with 5% fetal bovine serum for 1 h. For CC axon navigation assays, 400-mthick E17.5 coronal slices incorporating the CC axon tract were prepared and frontal cortex explants from -GFP ϩ slices were transplanted into the equivalent region in -GFP Ϫ slices before culturing in neurobasal medium for 72 h, fixation in 4% paraformaldehyde (PFA), and GFP immunofluorescence. For glial translocation experiments, 10 mg/ml BrdU dissolved in PBS was injected intraperitoneally into pregnant dams with E14.5 litters, which were killed 1 h later and 350 m coronal slices incorporating the CSB prepared for culture. In Fgf17 bead experiments, Affi-Gel blue gel (Bio-Rad) beads presoaked in 100 g/ml recombinant Fgf17 protein (R&D systems) or 5 mg/ml BSA (Sigma-Aldrich) overnight at 4°C were implanted into the slice, one Fgf17 and one BSA bead on either side of the midline just below the GW, and the MEM was replaced with neurobasal medium. For the FGFi culture, MEM was replaced with neurobasal medium containing either 25 M SU5402, 0.1% DMSO (FGFi) or 0.1% DMSO (control). Slices were cultured for 2 or 48 h, fixed in 4% PFA, and 10 m frozen sections were prepared for immunodetection or in situ hybridization. Glial migration out of the VZ toward the pial surface was quantified from BrdU/Sox9 immunofluorescence micrographs by demarcating the basal edge of the VZ (easily identified by Sox9 staining) with a line and counting the number of Sox9 ϩ ;BrdU ϩ cells that had crossed this line. This allowed us to count glial (Sox9 ϩ ) cells that had incorporated BrdU (BrdU ϩ ) when they were in the VZ before the start of the culture and subsequently exited the VZ and migrated toward the midline over the 2 d culture period when the cultures were exposed to experimental substances (SU5402, DMSO, Fgf17 protein, or BSA). Four or six sections were quantified per slice moving rostrally from the most caudal section in which the GW could be identified on both sides of the section.
Imaging. Fluorescence-labeled sections were imaged using either a Leica AF6000 epifluorescence microscope coupled to a Leica DFC360 digital camera or a Nikon Ti: E Inverted confocal microscope. DAB stained and in situ hybridized sections were imaged using a Leica DLMB microscope coupled to a Leica DFC480 color digital camera.
Fgf17 protein quantification. Fgf17 fluorescence was quantified from E14.5 Hs2st ϩ/ϩ ;Hs6st1 ϩ/ϩ , Hs2st Ϫ / Ϫ , and Hs6st1 Ϫ / Ϫ coronal sections that had been processed for Fgf17 immunofluorescence in parallel and imaged under identical conditions in parallel using the same method as described previously for Fgf8 protein quantification (Chan et al., 2017). For each section, ImageJ was used to measure mean fluorescence intensity in a 100 ϫ 150 m box drawn at the CSB encompassing the Fgf17 expression domain. For each embryo, quantification was performed for three sections along the rostrocaudal axis and averaged.
IG Sox9 ϩ cell quantification. Quantification of Sox9 ϩ cells in the IG region of E18.5 Hs2st fl/fl ;Zic4 Cre , Hs2st ϩ/ϩ ;Zic4 Cre , Hs2st fl/fl ;Emx1 CreER , and Hs2st ϩ/ϩ ;Emx1 CreER embryos was performed as described previously (Clegg et al., 2014). A counting box measuring 200 ϫ 200 m was placed on images of coronal sections at the midline with the top edge at the dorsal extent of Sox9 ϩ cells at the IG and the numbers of Sox9 ϩ cells in the box counted. For each embryo, quantification was performed for three sections along the rostrocaudal axis and averaged.
Ligand and carbohydrate engagement (LACE) assay. The LACE assay was performed as described previously (Allen et al., 2001;Allen and Rapraeger, 2003;Chan et al., 2015). Briefly, frozen sections were incubated in 0.05% NABH 4 /PBS for 15 min. After several washes in PBS, sections were incubated in 0.1 M glycine at 4°C overnight. Some sections were incubated with Heparitinase I (Seikagaku) before proceeding. All Fgf and Fgfr-Fc proteins were purchased from R&D Systems. Sections were then treated with 1% BSA/TBS solution for 10 min before incubation with 3 M recombinant mouse Fgf17 and 9 M recombinant human Fgfr1a(IIIc)-Fc or 30 nM recombinant mouse Fgf8b and 100 nM recombinant human Fgfr3 (IIIc)-Fc at 4°C overnight. Fgf17 or Fgf8 were omitted from some assays. Fluorescent LACE signal was generated by incubation with 1/200 anti-human IgG (Fc-specific) Cy3 (Sigma-Aldrich) in 1% BSA/TBS. Hs2st ϩ/ϩ ;Hs6st1 ϩ/ϩ , Hs2st Ϫ / Ϫ , and Hs6st1 Ϫ / Ϫ material that had been processed for each LACE assay condition in parallel was imaged under identical conditions in parallel. For each section, ImageJ was used to measure mean fluorescence intensity in a 100 ϫ 150 m box drawn encompassing the CSB. Background signal was quantified from control LACE experiments from which the FGF ligand was omitted and these values were used for background subtraction. For each embryo, quantification was performed for three sections along the rostrocaudal axis and averaged.
Data analysis and statistics. Results are expressed as mean Ϯ SEM. The statistical test and sample size (n) for each experiment are specified in the figure legends. Statistical comparison between two groups was performed with a t test. Statistical comparison between more than two groups was performed with ANOVA followed by post hoc t test. p Ͻ 0.05 was considered significant.

Hs2st protein is widely expressed in the developing cerebral cortex and at the telencephalic midline
To establish potential sites of action of Hs2st in CC development, we first examined the distribution of cells expressing Hs2st protein and contributing to developing CC structures using Hs2st immunohistochemistry at E14.5 ( Fig. 1A-D) and E18.5 ( Fig.  1E-M ) spanning the period of CC axon tract development. Macroscopically, Hs2st protein distribution closely resembles the Hs2st-LacZ reporter staining previously reported, with widespread Hs2st expression in the developing cerebral cortex and at the CSB at both E14.5 and E18.5 ( Fig. 1 A, E, boxed areas indicate regions shown at higher magnifications in B-D and F-M ) . Subcellularly, the Hs2st signal is punctate consistent with the expected localization of Hs2st in the Golgi apparatus (arrows point to Hs2st ϩ puncta in higher-magnification insets in Fig. 1 B, F,P). At E14.5, there was a high density of Hs2st ϩ puncta at the CSB in the GW region where IG astroglial RGC prescursors reside ( Fig. 1B, with boxed area shown as higher-magnification inset with arrows indicating Hs2st ϩ puncta), with the density falling toward the pial surface although Hs2st ϩ puncta were visible. There were many Hs2st ϩ puncta in the VZ of the cerebral cortex ( Fig. 1C) and also in the cortical plate ( Fig. 1D), indicating that many cortical progenitors and postmitotic neurons express Hs2st. At E18.5, Hs2st is expressed by many cells in the IG (Fig. 1F ) and at the apical surface of the VZ at the GW (Fig. 1G), septum (Fig. 1H ), and ventral telencephalon (Fig. 1I ), with the number of Hs2st expressing VZ cells diminishing as distance from the ventricle increases. In the cerebral cortex, Hs2st is expressed by many cells close to the apical surface of the VZ (Fig. 1J ). Large numbers of postmitotic cortical neurons outside of the VZ express Hs2st and, moving toward the pial surface, the density of Hs2st ϩ puncta varies with laminar position [cf. Fig. 1K-M, showing relatively high Hs2st ϩ puncta density in cortical layers adjacent to the pial membrane ( M) and in the intermediate zone ( K) and lower density in the intervening region ( L)]. We validated the Hs2st antibody by demonstrating absence of the punctate Hs2st ϩ immunostaining in Hs2st Ϫ / Ϫ embryonic material (cf. Fig. 1 N , P and O,Q; note that the more diffuse staining persists in Hs2st Ϫ / Ϫ tissue and we discounted this as nonspecific background) and Western blot showing that the predicted 42 kDa Hs2st protein band was present in Hs2st ϩ/ϩ and absent from Hs2st Ϫ / Ϫ telencephalic protein extracts (Fig.  1R). To conclude, Hs2st protein is present in developing cerebral cortex, the source of CC axons, as well as in progenitor and post-mitotic cells of the CSB region constituting the environment through which midline crossing CC axons navigate. Hs2st expression analysis suggests multiple potential sites of action for 2-O HS sulfation in CC development.
Slit2-expressing IG is expanded in Hs2st ؊ / ؊ embryos We previously reported that increased numbers of astroglia at the pial surface of the Hs2st Ϫ / Ϫ CSB stemmed from precocious glial translocation and found no evidence that changes in cell prolif- eration or death contributed to this phenotype Clegg et al., 2014). To determine whether there is an expansion of the IG in Hs2st Ϫ / Ϫ embryos, we compared the expression of Slit2 mRNA, a marker of GW and IG glia but not MZ glia, between Hs2st ϩ/ϩ and Hs2st Ϫ / Ϫ embryos at E16.5 (Shu and Richards, 2001;Shu et al., 2003). In Hs2st ϩ/ϩ embryos Slit2 ϩ cells form a compact focus at the IG that increases in size moving caudally ( Fig. 2 A, C, E, Slit2 expression domain at IG indicated by brackets). In Hs2st Ϫ / Ϫ embryos, the Slit2 expression domain is greatly expanded at the pial surface along the rostrocaudal axis ( Fig. 2 B, D,F, expanded Slit2 expression domain indicated by brackets). We conclude that an expansion of the Slit2 ϩ IG astroglial population makes a major contribution to the Hs2st Ϫ / Ϫ phenotype.

Cell autonomy of HS and 2-O HS sulfation in astroglial precursor somal translocation and CC development
We next exploited conditional mutagenesis of Hs2st or Ext1 to experimentally uncouple specific functions of 2-O sulfation from more general functions of HS in astroglial precursor translocation and CC development. Widespread expression of HS and 2-O HS sulfation leaves open the possibility that each regulates GW¡IG astroglial precursor somal translocation cell autonomously by modulating the response to signals, non-cell autonomously by regulating the supply of signals, or both. To resolve this, we identified two Cre alleles, Zic4 Cre and Emx1 CreER , that drive LoxP-mediated mutagenesis in the astroglial lineage or in their cellular environment, respectively, and used them to conditionally ablate either HS (Ext1 LoxP mutagenesis) or 2-O HS (Hs2st LoxP mutagenesis) sulfation to test for cell-autonomous or non-cell-autonomous functions. We refer to these as "Zic4 lineage" and "Emx1 lineage," respectively, and next present their characterization using a floxed-stop GFP reporter that turns on GFP expression in Cre-expressing cells and their descendants before describing experiments where they are used to conditionally generate loss-of-function mutations in Ext1 Fl or Hs2st Fl alleles (Inatani et al., 2003;Kessaris et al., 2006;Rubin et al., 2010;Sousa et al., 2009;Stanford et al., 2010).

Characterization of Zic4 and Emx1 lineages
The septum is of Zic4 lineage, as shown by strong expression of the GFP reporter (Fig. 3A). The GFP signal in the intermediate zone of the cerebral cortex (asterisks in Fig. 3A) is due to GFP ϩ thalamocortical axons that project from Zic4-lineage cells in the thalamus and cells of subcortical origin as described previously (Rubin et al., 2010). At the midline, GFP ϩ cells of the Zic4 lineage are predominantly located ventral to the CSB (dashed lines in Fig.  3B), but there is also GFP expression in the IG (boxed area "D" in Fig. 3B). Sox9 is a transcription factor that marks the nuclei of RGCs in the VZ and differentiated astroglia in the IG and MZG and we previously showed that the positioning of Sox9 ϩ cells is of critical importance for the development of the CC (Clegg et al., 2014). Combining GFP with Sox9 immunostaining reveals the contribution of the Zic4 lineage to the CSB astroglial populations. There is a sharp boundary (dashed line in Fig Fig. 3D,E), indicating that these cells are of the Zic4 lineage. These data show that the Zic4 lineage contributes Sox9 ϩ cells to the septal VZ and, strikingly, is the sole source of Sox9 ϩ astroglia in the IG (Fig. 3F).
To mark the Emx1 lineage, tamoxifen was administered to Emx1 CreER embryos harboring the floxed-stop GFP reporter at E9.5 so that early Emx1 expressing cerebral cortex progenitors and their descendants were rendered GFP ϩ . Examination of the expression of the GFP reporter shows that, as expected, the developing cerebral cortex and CC axons are of Emx1 lineage (Fig. 3G) and that, at the midline, GFP expression is predominantly located dorsal to the CSB (dashed lines in Fig. 3H ). Higher magnification shows that there is a sharp boundary between GFP ϩ and GFP Ϫ cells at the VZ of the CSB (dashed line in Fig. 3I ). Combining Sox9 and GFP immunostaining reveals the contribution of the Emx1 lineage to Sox9 ϩ cells. Sox9 ϩ ;GFP ϩ cells (arrowheads in Fig. 3I ) populate the VZ on the cortical side of the boundary with Sox9 ϩ ;GFP Ϫ cells on the septal side (arrows in  Fig. 3 J, K ), indicating that the Emx1 lineage does not contribute Sox9 ϩ cells to the IG. These data show that the Emx1 lineage contributes Sox9 ϩ cells to the cortical VZ, but no cells of this lineage contribute Sox9 ϩ astroglia to the IG (schema in Fig. 3L).

Ext1 is required by both Emx1-and Zic4-lineage cells for CC development
To determine the cellular requirement for HS, we deleted Ext1, which is essential for HS synthesis, in the Zic4 or Emx1 lineages. In control embryos, L1 immunostaining labels axons in the U-shaped CC, whereas GFAP staining labels midline astroglial structures ( Fig. 4 D, F ) and more GFAP at the GW than in controls (arrows in Fig. 4I; cf. Fig. 4G,I ), suggesting that, in these embryos, astroglial precursors translocate less efficiently to the IG and instead remain in the GW. The cerebral cortex of Zic4 Cre ;Ext1 Fl/Fl brains was thinned and the ventricles were enlarged (cf. Fig. 4 A, C) and this hydrocephalus-like phenotype is intriguing because the cerebral cortex is not of the Zic4 lineage, indicating a non-cellautonomous mechanism by which HS regulates cerebral cortex development. The FGFR1/FGF2 LACE assay detects endogenous HS on tissue sections by forming ternary complexes with exogenously added FGF2 and FGFR1 (red LACE signal in Fig. 4J-O) (Allen et al., 2001;Chan et al., 2015). HS is ubiquitously expressed in both cortical and septal compartments of control telencephalon (Fig. 4J, higher magnification of CSB in M ) and, as intended, HS synthesis is blocked in the cortex and cortical axons of Emx1 CreER ;Ext1 Fl/Fl embryos (Fig. 4K, CSB shown at higher magnification in N, with arrows indicating HS-deficient cortical . HS expression is required within both Emx1 and Zic4-lineage cells for CC formation. A-I, Immunofluorescence for L1 (red) at E18.5 labels the CC, whereas GFAP (green) labels glia. In control embryos, the U-shaped CC has formed and is flanked by glia at the IG and GW (A, D, G). In Ext1 fl/fl Emx1 creER embryos, CC axons do not cross the midline, whereas glia at the IG and GW appear largely unaffected (B, E, H ). In Ext1 fl/fl Zic4 cre embryos, CC axons do not cross the midline, whereas glia appear depleted at the IG and form abnormal bundles at the GW (C, F, I ). J-O, FGFR1/FGF2 LACE assay is used to detect the presence of HS. In control embryos, the LACE signal can be seen throughout the telencephalon and is of similar intensity within both the cortex and the septum (J, M ). In Ext1 fl/fl Emx1 creER embryos, the LACE signal is significantly reduced within the cortex (K, N ). In region) and in the septum of Zic4 Cre ;Ext1 Fl/Fl embryos (Fig. 4L, higher magnification of CSB in O, with arrows indicating HSdeficient septum). Predigesting tissue sections with heparitinase eliminated the LACE signal (data not shown), confirming the specificity of this assay for detecting HS. The salient conclusions from the Ext1 conditional mutagenesis for the current study are that HS is indispensable from both the Zic4 and the Emx1 lineages for CC development and that removing HS from the Zic4 lineage inhibits Zic4 lineage astroglia from reaching the IG region.

Hs2st is required by Emx1-lineage but not Zic4-lineage cells for CC development
Having established that both Zic4 and Emx1 lineages need to synthesize HS for normal CC development, we next investigated whether 2-O HS sulfation of the HS is required in either lineage. Hs2st is the sole enzyme capable of imparting 2-O HS sulfation onto HS, so to determine the cellular requirement for 2-O HS sulfation, we deleted Hs2st in the Zic4 or Emx1 lineages. Control Hs2st ϩ/ϩ genotypes (Hs2st ϩ/ϩ ;Emx1 CreER and Hs2st ϩ/ϩ ;Zic4 Cre ) displayed neither CC agenesis nor midline astroglial disorganization; the control embryo shown in Figure 5 ( A, D,G,J; D and G are reproduced from Fig. 3 H, J ) is of the Hs2st ϩ/ϩ ;Emx1 CreER genotype. The CC and the midline astroglial structures form normally in Hs2st fl/fl ;Zic4 Cre conditional mutants and the organization of L1 ϩ axons and GFAP ϩ astroglia are indistinguishable from control embryos (6/6 embryos) (cf. Fig. 5 B, A). The organization of GFP ϩ Zic4-lineage cells is the same in Hs2st fl/fl ;Zic4 Cre embryos as in Hs2st ϩ/ϩ ;Zic4 Cre embryos (cf. Figs. 5 E, H and 3 B, D) and IG Figure 5. Hs2st expression is required within Emx1-lineage cells but not Zic4-lineage cells for CC formation. A-C, Immunofluorescence for L1 and GFAP at E18.5. In control embryos, the U-shaped CC has formed and the IG can be observed above the CC (A). In Hs2st fl/fl Zic4 cre embryos, the CC and IG form normally (B). In approximately half of Hs2st fl/fl ;Emx1 creER embryos, the CC fails to form, IG glia also extend ventrally (asterisks, C). In the remaining Hs2st fl/fl ;Emx1 creER embryos, the CC forms normally (C). D-I, Immunofluorescence for Sox9 labels progenitor cells at the VZ and mature glia at the IG, GFP labels cells in which cre is active. D, G, In control (Hs2st ϩ/ϩ ;Emx1 creER ) embryos, IG glia do not express GFP. E, H, In Hs2st fl/fl ;Zic4 cre embryos, IG glia do express GFP and adopt their normal position. F, I,InHs2st fl/fl ;Emx1 creER embryos, GFP is expressed by cortical neurons and axons but not by abnormally positioned IG glia. J-L, Immunohistochemistry for Hs2st shows expression of Hs2st in the IG. J, In control embryos, punctate Hs2st staining can be seen within IG cells. In Hs2st fl/fl Zic4 cre embryos, Hs2st is not expressed by IG glia (K ). In Hs2st fl/fl ;Emx1 creER embryos, Hs2st is expressed by displaced glial cells (L). Hs2st immunohistochemistry in J-L was performed on adjacent tissue sections to those in D-I. M, Quantification of Sox9 expressing cell number at the IG in control (blue bar, n ϭ 4 embryos, 2 Hs2st ϩ/ϩ ;Zic4 cre ϩ 2 Hs2st ϩ/ϩ ;Emx1 creER ), affected Hs2st fl/fl ;Emx1 creER (orange bar, n ϭ 4 embryos), and Hs2st fl/fl ;Zic4 cre (purple bar, n ϭ 3 embryos). Sox9 ϩ numbers are significantly increased compared with control in Hs2st fl/fl ;Emx1 creER embryos (*p Ͻ 0.05; F (2,7) ϭ 42.16, p ϭ 0.00013, ANOVA, post hoc t tests: control vs Hs2st fl/fl ;Emx1 creER ; Sox9 ϩ cell counts confirm that the numbers of Sox9 ϩ cells in the IG are not significantly different from control Hs2st ϩ/ϩ embryos (Fig. 5M, cf. blue and orange bars), indicating no cellautonomous requirement for Hs2st in the Zic4-lineage Sox9 ϩ IG astroglia. To exclude the possibility of a compensatory mechanism by which the Hs2st fl/fl ;Zic4 Cre IG is populated by Hs2st ϩ/ϩ cells from a different lineage, we performed Hs2st immunohistochemistry and confirmed that Hs2st expression is indeed absent from all cells in the IG (Fig. 5K, note this is an adjacent section from the same embryo to the one shown in H ). In Hs2st fl/fl ; Emx1 CreER embryos, the CC fails to form in ϳ50% of cases and embryos either had a severe phenotype (Fig. 5C, 5/9 embryos) or appeared completely unaffected (Fig. 5CЈ, 4/9 embryos). CC axons form Probst bundles on either side of the telencephalic midline, and the GFAP ϩ IG is expanded (asterisks in Fig. 5C). The anatomy and incomplete penetrance of the CC phenotype in Hs2st fl/fl ;Emx1 CreER embryos closely resemble constitutive null Hs2st Ϫ / Ϫ embryos, indicating that Hs2st function within the Emx1 lineage is sufficient for normal CC development Clegg et al., 2014). As in control Hs2st ϩ/ϩ ;Emx1 CreER embryos (Fig. 5 D, G), the GFP and Sox9 signals did not overlap in the IG region of control or Hs2st fl/fl ;Emx1 CreER embryos (Fig. 5F, boxed area shown at higher magnification in I ) and the Sox9 ϩ cells in the IG of control embryos and the expanded IG of Hs2st fl/fl ; Emx1 CreER embryos were GFP Ϫ (arrows in Fig. 5I indicate Sox9 ϩ ;GFP Ϫ cells). Counts of Sox9 ϩ cells confirmed a significant increase in the IG of affected Hs2st fl/fl ;Emx1 CreER embryos compared with controls (Fig. 5M, cf. blue and purple bars). Immunostaining for Hs2st on adjacent sections confirmed that IG cells in Hs2st fl/fl ;Emx1 CreER embryos retain Hs2st protein expression (Fig. 5L). Because Sox9 ϩ IG astroglia do not belong to the Emx1 lineage, their ectopic position in Hs2st fl/fl ;Emx1 CreER embryos despite retaining Hs2st function allows us to conclude a non-cell-autonomous requirement for Hs2st in the translocation of astroglial precursors to the IG.
The salient conclusions from these conditional mutagenesis experiments are that, although the Zic4-lineage astroglia do require Ext1 to form midline astroglial structures, they do not require Hs2st, strongly suggesting that, whereas these Zic4-lineage cells require HS on their cell surface to respond to translocation signals, there is no need for the HS to be 2-O sulfated. In contrast, Hs2st is absolutely required in the surrounding Emx1-lineage cells, indicating a non-cell-autonomous mechanism by which 2-O HS sulfation controls the transmission of translocation signals to the Zic4-lineage astroglial precursors.
Hs2st is not required cell autonomously by CC axons to navigate the midline The conditional mutagenesis experiments showed that Hs2st has a non-cell-autonomous role in GW¡IG somal translocation, but because Hs2st is expressed throughout the cerebral cortex, did not resolve whether there is an additional cell autonomous requirement in CC axon navigation. To solve this, we performed ex vivo transplantation experiments in which cerebral cortical tissue from transgenic mice ubiquitously expressing GFP, which efficiently labels axons of GFP ϩ cells, was transplanted into GFP Ϫ telencephalic slices containing the CC axon pathway and CSB structures (Pratt et al., 2000;Niquille et al., 2009). When WT E17.5 GFP ϩ cortical explants are transplanted into age-matched GFP Ϫ WT cortical slices, GFP ϩ axons extend across the telencephalic midline, forming the characteristic U shape of the CC and reaching the cortex of the opposite hemisphere (n ϭ 3/3 cultures, arrows in Fig. 6A,D point to crossing axons). When Hs2st Ϫ / Ϫ GFP ϩ cortical explants are transplanted into GFP Ϫ WT slices, axons are able to cross the midline to reach the opposite hemisphere in a manner indistinguishable from that seen in the WT¡WT transplants (n ϭ 4/4 cultures, arrows in Fig. 6B,E point to crossing axons). In contrast, when GFP ϩ WT cortical explants are transplanted into GFP Ϫ Hs2st Ϫ / Ϫ slices, axons are unable to reach the opposite cortical hemisphere and instead remain within the cingulate cortex or invade the septum (n ϭ 6/6 cultures, Fig. 6 C,F, arrowheads point to axons growing into the septum), resembling the in vivo CC phenotype observed in Hs2st Ϫ / Ϫ embryos Clegg et al., 2014). In all cultures, a few axons grew into the septum (arrowheads in Fig. 6D-F). Schematics summarizing these experiments are shown in Figure 6, G-I. These data show that 2-O HS sulfation is not required cell autonomously by CC projection neurons for axon guidance across the midline, strongly suggesting that disorganization of midline guidepost astroglial cells is the primary cause of the Hs2st Ϫ / Ϫ CC agenesis phenotype.
Abnormally high FGF/ERK signaling causes the Hs2st ؊ / ؊ precocious astroglial translocation phenotype We previously reported a correlation between hyperactive ERK signaling at the CSB and precocious somal translocation of astroglia to the midline in Hs2st Ϫ / Ϫ embryos, but we did not formally establish that this stemmed from hyperactive FGF/ERK signaling (Clegg et al., 2014;Chan et al., 2017). To address this, we used an ex vivo assay in which coronal WT or Hs2st Ϫ / Ϫ telencephalic slices incorporating the CSB were cultured on floating membranes for long enough to allow somal translocation to the midline and attempted to rescue the Hs2st Ϫ / Ϫ phenotype by pharmacological abrogation of FGF/ERK signaling. WT or Hs2st Ϫ / Ϫ E14.5 slices were cultured in the presence of the Fgfr1 inhibitor SU5402 dissolved in DMSO (FGFi treatment) to inhibit FGF/ERK signaling or in DMSO alone (untreated control) for 48 h (Fig. 7A). To aid subsequent identification of translocating cells, a subpopulation RGCs undergoing S-phase in the VZ at E14.5 were labeled just before culturing with a single pulse of BrdU. Immunohistochemistry for pErk (brown stain in Fig.  7 B, D, F, H ) confirms inhibition of FGF/ERK signaling in both FGFi treated WT and Hs2st Ϫ / Ϫ cultures (Fig. 7 D, H ) compared with untreated cultures (Fig. 7 B, F ), showing that FGF signaling through Fgfr1 accounts for ERK phosphorylation in both genotypes. This demonstrates that ERK hyperactivation in Hs2st Ϫ / Ϫ embryos does not stem from an FGF-independent mechanism for ERK activation (Clegg et al., 2014;Chan et al., 2017). After 48 h, some Sox9 ϩ cells (red) had left the VZ and translocated to the midline in untreated WT cultures (arrow in Fig. 7C), with many more populating the midline in untreated Hs2st Ϫ / Ϫ cultures (arrow in Fig. 7G), validating that our ex vivo assay replicates the in vivo Hs2st Ϫ / Ϫ phenotype. Consistent with our hypothesis, FGFi treatment of both WT and Hs2st Ϫ / Ϫ cultures resulted in a large decrease in Sox9 ϩ cells reaching the midline (cf. Fig. 7 E, I and C,G). We quantified glial translocation by counting the numbers of Sox9 ϩ cells born in the VZ at E14.5 (Sox9 ϩ ;BrdU ϩ cells, yellow; inset in Fig. 7C, E, G,I shows higher magnification) that had exited the VZ toward the midline (VZ demarcated by dotted line in Fig. 7C, E, G,I ) after 2 d in culture. Counts of BrdU ϩ ;Sox9 ϩ cells showed that glial translocation was significantly greater in Hs2st Ϫ / Ϫ compared with WT cultures along the rostrocaudal axis (dark purple and green lines in Fig.  7J ) and in both cases was almost completely suppressed by FGFi treatment (pale purple and green lines in Fig. 7J ).
We conclude that the precocious glial translocation phenotype in Hs2st Ϫ / Ϫ embryos is caused by hyperactive FGF/ERK signaling from E14.5 on. Together with our Hs2st conditional mutagenesis experiments demonstrating a non-cell-autonomous role for Hs2st in astroglial precursor translocation, we hypothesize that Hs2st normally suppresses the supply of FGF proteins to translocation competent astroglial precursors in the GW.

Hs2st suppresses Fgf17 protein levels
We next sought to identify an FGF protein that is targeted by Hs2st. Despite its well known role in CC development, Fgf8 protein levels are not significantly increased at the CSB of Hs2st Ϫ / Ϫ embryos, forcing us to consider other FGFs (Clegg et al., 2014;Chan et al., 2017). A promising candidate is Fgf17, a member of the Fgf8 subfamily transcribed at the CSB in a similar pattern to Fgf8 (Cholfin and Rubenstein, 2008;Zhang X et al., 2012). Fgf17 is a canonical FGF that binds to HS, so it is potentially regulated via its interaction with HS and is known to play a role in patterning the telencephalon, although its role in CC development has not been fully characterized (Cholfin and Rubenstein, 2007; Hoch et al., 2015; Li and Kusche-Gullberg, 2016). We hypothe-sized that Hs2st normally suppresses Fgf17 protein and predicted that Fgf17 protein levels would be increased at the Hs2st Ϫ / Ϫ CSB. We compared the expression of Fgf17 protein in the developing CSB of WT and Hs2st Ϫ / Ϫ embryos at three developmental stages, E12.5, E14.5, and E16.5, spanning the interval of midline glial translocation. At E12.5, telencephalic Fgf17 protein is restricted to the CSB region with no obvious difference between WT and Hs2st Ϫ / Ϫ (cf. Fig. 8A1,B1 and A2,B2). By E14.5, there is an expanded Fgf17 protein domain at the CSB of Hs2st Ϫ / Ϫ embryos (cf. Fig. 8D1,E1 and D2,E2; *E2 marks the expanded Fgf17 protein domain). Quantification of Fgf17 immunofluorescence shows a significant ϳ2-fold increase in Fgf17 protein levels in this region of Hs2st Ϫ / Ϫ CSB (Fig. 8W, cf. blue and green bars). At E16.5, Fgf17 protein is much closer to detection threshold than at the earlier stages in both genotypes (Fig. 8G1,G2,H1,H2) although the increased protein spread in the mutant persists (* in Fig. 8H2), indicating that the Hs2st Ϫ / Ϫ CSB is exposed to a prolonged overdose of Fgf17 protein spanning E14.5-E16.5. We next examined Fgf17 mRNA at the CSB to determine whether the Figure 6. Hs2st is not required by CC axons to cross the telencephalic midline. A, D, After homotypic transplantation of E17.5 cortical explants from GFP ϩ control tissue into the cortex of GFP Ϫ control brain slices, GFP ϩ CC axons are able to project across the midline (arrows, D). B, E, After transplantation of GFP ϩ Hs2st Ϫ / Ϫ cortical explants into GFP Ϫ control brain slices, GFP ϩ CC axons are able to project across the midline (arrows, E). C, F, After transplantation of cortical explants from GFP ϩ control tissue into the cortex of GFP Ϫ Hs2st Ϫ / Ϫ brain slices, GFP ϩ CC axons are unable to project across the midline and invade the septum. Arrowheads indicate axons navigating into the septum in all conditions. G-I, Schematic of transplant experiments shown in A-C. D-F are higher-magnification images of the boxed region in A-C, respectively. Scale bars, 200 m in all panels. increase in Fgf17 protein in Hs2st Ϫ / Ϫ CSB was underpinned by altered Fgf17 gene expression. There was no evidence for this at E12.5 or E14.5, when the Fgf17 mRNA expression pattern remained similar between Hs2st ϩ/ϩ and Hs2st Ϫ / Ϫ embryos (cf. Fig. 8C1,C2 and F1,F2); however, the expression domain of Fgf17 mRNA is increased in E16.5 Hs2st Ϫ / Ϫ CSB (cf. Fig. 8I1and I2 , * in I2 marks expanded Fgf17 mRNA domain). This subsequent increase in Fgf17 mRNA in the E16.5 Hs2st Ϫ / Ϫ CSB indicates that the Hs2st Ϫ / Ϫ phenotype has a transcriptional component or that there are more cells expressing Fgf17 mRNA in the expanded Hs2st Ϫ / Ϫ IG, although this cannot be the primary event because it is not apparent at E14.5, the stage at which we previously showed that precocious astroglial precursor translocation was well under way in Hs2st Ϫ / Ϫ embryos (Clegg et al., 2014).
Mosaic analysis (Fig. 5) indicated that Hs2st function in the Emx1 lineage negatively regulates a signal promoting GW¡IG translocation of Zic4-lineage glial cells by a non-cell-autonomous mechanism and Fgf17 expression analysis (Fig. 8) makes Fgf17 a strong candidate for the signal. Based on this, we hypothesized that Fgf17 is expressed in cells surrounding the Zic4-lineage cells and performed detection of Fgf17 mRNA or protein in E14.5 WT embryos in which the Zic4 lineage is labeled GFP ϩ . Fgf17 mRNA is expressed at the GW and the IG (Fig. 8J ) and higher-power magnification shows that, in the VZ, GFP ϩ cells express little if any Fgf17 mRNA and, conversely, cells expressing the highest levels of Fgf17 mRNA are GFP Ϫ (Fig. 8K1-K3, arrows indicate GFP ϩ cell location). This complementarity between Fgf17 mRNA-expressing and Zic4-lineage cells is preserved at the IG ( Fig. 8L1-L3, arrows indicate GFP ϩ cell location). Fgf17 protein predominates at the IG (Fig. 8M ) and higher-power magnification shows that, although Fgf17 protein is barely detectable at the GW (Fig. 8N1-N3), there are a number of much more highly Fgf17-expressing cells at the IG and these cells are GFP Ϫ , confirming that they do not belong to the Zic4 lineage ( Fig. 8O1-O3, arrows indicate GFP ϩ cell location). Interestingly, although cells in the GW and IG express comparable levels of Fgf17 mRNA (cf. Fig. 8K1,L1), the expression of Fgf17 protein is much higher in the IG (cf. Fig. 8N1,O1), suggesting a posttranscriptional repression selectively at the GW. Our identification of Hs2st as a repressor of Fgf17 protein levels at this stage makes Hs2st a strong candidate; indeed, closer examination of Hs2st expression using the Hs2st-LacZ reporter shows that Hs2st is expressed in a GW High -IG Low pattern (Fig. 8P, also apparent in the Hs2st immunohistochemistry; Fig. 1B) complementary to the GW Low -IG High Fgf17 protein distribution. Together, these data bolster the idea that Hs2st acts to suppress Fgf17 protein supply to Zic4lineage cells via a posttranscriptional mechanism.
Exogenously applied Fgf17 phenocopies the Hs2st ؊ / ؊ astroglial translocation phenotype Our data suggest that the Hs2st Ϫ / Ϫ phenotype stems from abnormally high levels of Fgf17 protein at the CSB, causing FGF/ ERK hyperactivation and precocious somal translocation to the IG. This requires that Hs2st Ϫ / Ϫ CSB cells are competent to respond to Fgf17 protein and that application of ectopic Fgf17 triggers precocious glial translocation, neither of which has been previously established. We redeployed the CSB ex vivo culture assay (Fig. 7) with the modification that beads soaked in either recombinant Fgf17 protein (Fgf17 treatment) or in BSA (control) were implanted into coronal slices of CSB region on either side of the midline (Fig. 9A). WT or Hs2st Ϫ / Ϫ slices implanted with Fgf17 and BSA beads were cultured for 2 h before processing for Fgf17 (green signal) and pErk (red signal) double immunofluorescence (Fig. 9B). In both WT and Hs2st Ϫ / Ϫ cultures, Fgf17 protein was detectable adjacent to the edge of the bead (green signal) and this activated ERK phosphorylation in a similar pattern (red signal) with no obvious differences between WT and Hs2st Ϫ / Ϫ , indicating that Hs2st Ϫ / Ϫ CSB tissue is competent to respond to Fgf17 (Fig. 9B, top row). The lack of Fgf17 or pERK signal in the BSA control (Fig. 9B, bottom row) confirms Fgf17 antibody specificity and that pERK activation was specifically induced by exogenously applied Fgf17. We performed Sox9/ BrdU analysis (exactly as described above for the FGFi experiments; Fig. 7) to assess the impact of experimentally introduced Fgf17 on astroglial translocation to the midline in WT CSB slices after 48 h in culture and the results were dramatic. The side with the Fgf17-bead showed many more Sox9 ϩ (red) cells in the IG region (large arrow on right side of Fig. 9C) than the side with the BSA bead (smaller arrow on left side of Fig. 9C). Quantification of Sox9 ϩ ;BrdU ϩ (yellow) cells (Fig. 9D shows higher magnification of IG region) confirmed a significant increase in astroglial translocation to the midline along the rostrocaudal axis on the side exposed to Fgf17 [Fig. 9E, cf. green (Fgf17) with black (control) lines]. An important function of IG glia is to secrete Slit2 and repulsively guide CC axons in the correct trajectory across the midline. At E16.5, Slit2 mRNA is normally expressed in the IG region and, in Hs2st Ϫ / Ϫ embryos, the midline Slit2 expression domain is expanded (cf. Fig. 2C,E and D, F; Slit2 expression domain is bracketed). Experimentally introduced Fgf17 is sufficient to phenocopy this aspect of the Hs2st Ϫ / Ϫ phenotype in our ex vivo assay because the side exposed to the Fgf17 bead has a much larger Slit2 domain than the BSA-treated side [cf. left (control) and right (Fgf17) bracketed areas in Fig. 9F], consistent with precocious translocation of excessive numbers of Slit2 ϩ IG glia.
We conclude that Hs2st Ϫ / Ϫ CSB tissue is competent to respond to Fgf17 protein and abnormally high levels of Fgf17 protein are sufficient to phenocopy the Hs2st Ϫ / Ϫ astroglial translocation phenotype consistent with the model presented in Figure 10.

Hs2st selectively facilitates physical interaction between Fgf17 protein and HS
Our in vivo data show that Hs2st suppresses the levels of Fgf17 protein and that this represents a selective interaction between Hs2st-mediated 2-O HS sulfation and Fgf17 protein levels in vivo because Hs2st does not suppress the levels of the closely related Fgf8 protein, and Hs6st1, which catalyzes 6-O HS sulfation, does not suppress Fgf17 protein levels (Clegg et al., 2014;Chan et al., 2017) (Fig. 8). However, these in vivo experiments do not resolve whether differential sulfation has a correspondingly direct selective effect on the physical interaction between HS and Fgf17. To test the hypothesis that Hs2st has a selective effect on the binding of Fgf17 protein to HS molecules, we turned to a biochemical assay, the LACE assay, which probes the physical interaction between HS and FGF proteins by quantifying the ability of endogenous HS in tissue sections to form Fgf:Fgfr:HS complexes with exogenously added Fgf protein and Fgfr ectodomain fused to an Fc tag for immunofluorescent detection (Allen et al., 2001;Chan et al., 2015). We used the Fgf17:Fgfr1 LACE assay to compare the binding of Fgf17 protein to HS in WT, Hs2st Ϫ / Ϫ , and Hs6st1 Ϫ / Ϫ CSB tissue at E14.5 and E16.5 to test the hypothesis that Fgf17:HS physical interaction is selectively sensitive to loss of 2-O HS sulfation in Hs2st Ϫ / Ϫ tissue ( Fig. 11A-J, O). We used the Fgf8:Fgfr3 LACE assay to compare the binding of Fgf8 protein to HS in WT and Hs2st Ϫ / Ϫ CSB tissue at E14.5 to test the hypothesis that the Fgf8:HS physical interaction is insensitive to loss of 2-O HS sulfation in Hs2st Ϫ / Ϫ tissue (Fig. 11K-N, P).

Discussion
Embryonic CC development involves multiple cell and molecular events that ultimately guide callosal axons across the telencephalic midline to connect with their synaptic targets in the contralateral hemisphere. Three subpopulations of midline astroglia play pivotal roles in guiding callosal axons across the telence- Figure 9. Fgf17-bead experiment. A, Experimental outline of Fgf17 protein bead experiment in WT embryos. Pregnant females were injected at E14.5 with a BrdU pulse and CSB slices collected. One Fgf17 and one BSA bead were added to each side of the midline. B, Fgf17 protein and pErk after 2 h in culture. In both WT and Hs2st Ϫ / Ϫ CSB slices, Fgf17 and pErk are seen in tissue surrounding the Fgf17 bead. Staining for either is absent around the BSA bead (demarcated by dotted circle). C, Immunofluorescence for BrdU and Sox9 was performed on WT slices after 48 h in culture; curved dotted lines indicate the basal edge of the VZ and straight dotted line indicates the midline. D, Higher magnification of the arrowed regions in C. E, Quantification of Sox9 ϩ /BrdU ϩ 4 double-labeled cells in WT CSB slice cultures with Fgf17 or BSA bead. The Fgf17 bead significantly increased the number of Sox9 ϩ /BrdU ϩ cells that exit the VZ and moved toward the IG (significant differences indicated on graph as **p Ͻ 0.05, ***p Ͻ 0.001) at the four caudal-most positions (n ϭ 5 embryos; F (1,48) ϭ 65.63, p ϭ 0.000000000155, two-way ANOVA, followed by t test with Sidak's correction for multiple comparisons for Fgf17-bead vs BSA-bead at each rostralcaudal position: position 1, t (48) ϭ 1.45, p ϭ 0.63, t test; position 2, t (48) ϭ 2.36, p ϭ 0.13, t test; position 3, t (48) ϭ 3.65, p ϭ 0.0039, t test; position 4, t (48) ϭ 4.47, p ϭ 0.0003, t test; position 5, t (48) ϭ 4.03, p ϭ 0.0012, t test; and position 6, phalic midline. MZ glia facilitate fusion of the cerebral hemispheres and provide a substrate for crossing callosal axons, whereas Slit2 ϩ IG and GW astroglia channel crossing axons into the correct path by Robo/Slit-mediated chemorepulsion (Shu and Richards, 2001;Bagri et al., 2002;Shu et al., 2003;Gobius et al., 2016). These astroglial populations originate from RGCs born in the VZ of the septal midline and either remain in the VZ at the GW or translocate in response to FGF signals, of which Fgf8 appears to be particularly important, to the pial surface of the telencephalic midline (MZ and IG astroglia) (Smith et al., 2006;Moldrich et al., 2010;Clegg et al., 2014;Gobius et al., 2016). Both Slit2 ϩ IG and Slit2 Ϫ MZ astroglia are essential for CC development and both of these astroglial populations originate from the septal VZ Zic4 lineage, so the lack of an overt CC phenotype in Hs2st Fl/Fl ;Zic4 Cre embryos following conditional knock-out of Hs2st in the Zic4 lineage indicates that neither MZ or IG astroglial precursors have a cell-autonomous requirement for Hs2st to translocate in appropriate numbers. In Hs2st Ϫ / Ϫ embryos, there is an expansion of the Slit2 expression domain at the CSB pial surface coinciding with increased Sox9 ϩ glial cells and this is phenocopied by application of exogenous Fgf17 to Hs2st ϩ/ϩ CSB ex vivo, strongly suggesting that increased numbers of Slit2 ϩ glial cells at the midline reflect excessive GW¡IG somal translocation enlarging the IG (current study). We cannot rule out the possibility that disrupted MZ glial translocation also contributes to the Hs2st Ϫ / Ϫ phenotype, although this would not alter our conclusion that Hs2st plays a non-cell-autonomous role in the Zic4-lineage astroglial translocation phenotype. Our model (Fig. 10) posits that ectopic Slit2 ϩ astroglia at the midline block the transit of CC axons. In principle, this could be tested by rescuing the CC axon midline crossing in Hs2st Ϫ / Ϫ ;Slit2 Ϫ / Ϫ embryos (along similar lines to the Slit2 genetic rescue of the Hs6st1 Ϫ / Ϫ phenotype that we reported in Hs6st1 Ϫ / Ϫ ;Slit2 Ϫ / Ϫ embryos; . However, in contrast to the fully penetrant (100%) Hs6st1 Ϫ / Ϫ CC phenotype, the partial penetrance (ϳ50%) of the Hs2st Ϫ / Ϫ CC phenotype introduces a confounding factor of distinguishing "rescued" from "unaffected" Hs2st Ϫ / Ϫ embryos, a problem that would be compounded if only a proportion of embryos destined to be affected were rescued Clegg et al., 2014), so a prohibitively large number of animals would be required to demonstrate a statistically significant rescue.
Eliminating HS (Ext1 mutagenesis) compared with 2-O HS sulfation (Hs2st mutagenesis) from the same cell lineages allowed us to distinguish physiological functions generally attributable to HS from those specifically requiring 2-O HS sulfation by comparing the Ext1 and Hs2st phenotypes. We found that, although Zic4-lineage cells were unable to support CC development when they lacked HS (Zic4 Cre ;Ext1 Fl/Fl embryos), there was no similar requirement for 2-O HS sulfation in the Zic4 lineage (Zic4 Cre ; Hs2st Fl/Fl embryos), indicating that Zic4-lineage cells require HS but that 2-O HS sulfation is dispensable for their contribution to CC development, specifically the ability of astroglial precursors to cell autonomously sense translocation signals. We found that HS and 2-O HS sulfation are both required in the Emx1 lineage (Emx1 CreER ;Ext1 Fl/Fl and Emx1 CreER ;Hs2st Fl/Fl embryos), although the axonal and astroglial phenotypes were not identical. Somewhat counter-intuitively, removing HS completely from the Emx1 lineage (Emx1 CreER ;Ext1 Fl/Fl embryos) has a less severe effect on the distribution of GFAP ϩ midline astroglia than preserving HS but specifically blocking it's 2-O sulphation (Emx1 CreER ;Hs2st Fl/Fl embryos) with the more subtle disruption to HS causing the more pronounced accumulation of astroglia at the pial surface. We speculate that completely removing HS from the Emx1 lineage results in a general destabilization of FGF protein gradients thus mitigating precocious somal translocation by Zic4 lineage astroglial precursors (Qu et al., 2011(Qu et al., , 2012Shimokawa et al., 2011;Chan et al., 2017). The relatively normal midline astroglial organization in Emx1 CreER ;Ext1 Fl/Fl embryos poses the question of whether glial disorganization is a major contributor to their CC agenesis phenotype. In Emx1 CreER ; Hs2st Fl/Fl embryos, the Probst bundles form right next to the midline, consistent with our hypothesis that ectopic Slit2 ϩ astroglia at the midline are repelling CC axons from crossing the midline (Conway et al., 2011 and current study). In contrast, the Probst bundles in Emx1 CreER ;Ext1 Fl/Fl embryos form much more lateral to the midline at some distance from the IG, indicating that CC axons are misrouted at an earlier stage of their navigation than in Emx1 CreER ;Hs2st Fl/Fl embryos. HS is required cell autonomously for navigating axons to respond to axon guidance molecules, including Netrin1 and Slit2 (Piper et al., 2006;Matsumoto et al., 2007). A plausible explanation is that, in Emx1 CreER ; Ext1 Fl/Fl embryos, the Emx1-lineage HS-deficient CC axons cannot respond appropriately to guidance cues that would normally guide them toward the midline and are already misrouted before they come under the influence of the midline astroglia. In contrast, Hs2st Ϫ / Ϫ CC axons express HS lacking 2-O sulfation, which does not affect their ability to respond to guidance cues (current study), so they reach the midline but are prevented from crossing by the ectopic Slit2 ϩ glia in the expanded IG.
Biochemical (LACE) data show that physical interaction between Fgf17 and HS is facilitated by Hs2st (but not Hs6st1) and that Hs2st facilitates physical interaction between HS and Fgf17 (but not Fgf8), suggesting a molecular mechanism underpinning that in turn suppresses the levels of Fgf17 protein, but not Fgf8 protein, by an unknown mechanism at the CSB. Zic4 lineage astroglial precursors respond to Fgf8 and Fgf17 protein by activating FGF/ERK signaling and translocating (black arrows) to the midline. This generates appropriate positioning of Slit2 ϩ astroglia to guide CC axons across the midline. Loss of 2-O HS from the Emx1 lineage selectively desuppresses Fgf17 protein levels leaving Fgf8 protein unaffected. This causes hyperactive FGF/ERK signaling and more Zic4 lineage astroglial precursors translocate than normal with consequent blocking of CC axon midline crossing by the ectopic midline Slit2 ϩ astroglia. Zic4 lineage astroglial precursor cells do not need to express 2-O HS to respond to FGF signaling proteins and translocate to the midline. Hs2st selectively suppressing levels of Fgf17 in vivo (Allen and Rapraeger, 2003;Clegg et al., 2014;Chan et al., 2015Chan et al., , 2017current study). We speculate that Hs2st exerts its selective effect on Fgf17 protein levels because HS lacking 2-O HS sulfation has reduced affinity for Fgf17 (but not Fgf8), so increasing the halflife of Fgf17 (but not Fgf8) in the ECM by selectively reducing the rate that Fgf17 protein is cleared by HS-mediated receptormediated endocytostis of canonical FGFs while leaving Fgf8 unaffected (Yu et al., 2009). Our conditional mutagenesis experiments clearly demonstrate there is non-cell-autonomous requirement for Hs2st in astroglial precursor translocation in Emx1 CreER ;Hs2st Fl/Fl embryos; however, the reduced efficiency of HS:Fgf17:Fgfr1 complex formation in the Hs2st Ϫ/Ϫ LACE assay implies that Hs2st might also play a cell-autonomous role in the response to Fgf17 protein. We speculate that, even if Hs2st Ϫ / Ϫ astroglial precursors are less sensitive to Fgf17 than their WT counterparts, their translocation to the midline is primarily driven by Fgf8, so it is not significantly affected in Zic4 Cre ; Hs2st Fl/Fl embryos. A putative reduced sensitivity of Hs2st Ϫ / Ϫ astroglial precursor cells to Fgf17 also begs the question of how elevated Fgf17 could trigger precocious glial translocation in Hs2st Ϫ / Ϫ embryos. The Fgf17 bead assay experiment shows that Hs2st Ϫ / Ϫ cells retain competence to respond to Fgf17 by phosphorylating ERK and LACE data show that HS devoid of 2-O HS sulfation still interacts with Fgf17, albeit with reduced efficiency. Therefore, the explanation that best fits our experimental data is that increased Fgf17 protein levels in Hs2st Ϫ / Ϫ embryos override any reduction in competency of Hs2st Ϫ / Ϫ cells to respond to Fgf17 protein and the net effect is elevated FGF/ERK signaling and consequent precocious astroglial translocation.
This study makes two major novel contributions to our understanding of the cell and molecular roles of differential HS sulfation in the regulation of forebrain development. First, a primary cellular role of 2-O HS sulfation in vivo is not to modulate the competence of astroglial precursor cells to respond to translocation signals by a cell-autonomous mechanism (as would be predicted by the classic role for HS in modulating the formation of the FGF:FGFR:HS receptor complex on the surface of responding cells), but instead to regulate the supply of translocation signals to astroglial precursors by a non-cell-autonomous mechanism. Second, the interaction between 2-O-sulfated HS and Fgf17 protein is selective because it does not apply to the closely related Fgf8 protein or to 6-O HS sulfation catalyzed by Hs6st1. The most parsimonious explanation linking these cell and molecular events is that higher than normal levels of Fgf17 protein at the CSB of Hs2st Ϫ / Ϫ embryos causes the precocious astroglial precursor translocation phenotype and subsequent misrouting of CC axons (Fig. 10). Our rescue of the Hs2st Ϫ / Ϫ precocious astroglial precursor translocation phenotype ex vivo by generic pharmacological inhibition of FGF signaling with SU5402 directly supports the hypothesis that hyperactive FGF/ ERK signaling causes the phenotype. Given the well known role of FGF/ERK signaling in triggering astroglial precursor translocation to the IG, our findings that exogenously applied Fgf17 protein is sufficient to phenocopy the Hs2st Ϫ / Ϫ astroglial precursor translocation phenotype and that Hs2st Ϫ / Ϫ CSB cells activate ERK in response to Fgf17 protein, it is extremely unlikely that increased Fgf17 protein levels in vivo would not result in ERK hyperactivation and consequent precocious astroglial precursor translocation in Hs2st Ϫ / Ϫ embryos. However, the current study does not provide formal proof that the elevated levels of Fgf17 protein are solely responsible for the FGF/ERK hyperactivation or precocious astroglial precursor translocation phenotypes in Hs2st Ϫ / Ϫ embryos and we were unable to design an experiment that could further discriminate between the functions of Fgf17 and Fgf8 and directly test functional selectivity of Hs2st for Fgf17 in this context. We considered using a classic rescue experiment strategy by genetically reducing Fgf17 dosage in Hs2st Ϫ / Ϫ embryos (Fgf17 Ϫ / Ϫ ;Hs2st Ϫ / Ϫ rescue), but elected not to because, at best, it would provide equivocal evidence either for or against the hypothesis that 2-O-sulfated HS interacts selectively with Fgf17 protein. FGF/ERK hyperactivation caused by overexpression of a particular FGF protein can be rescued by any experimental manipulation that restores ERK signaling to normal levels and not uniquely by restoring the levels of the FGF protein that underpins the phenotype. Specifically, reducing Fgf17 dosage could elicit a rescue of ERK hyperactivation and collateral phenotypes at the Hs2st Ϫ / Ϫ CSB by reducing FGF/ERK signaling output regardless of whether abnormally high Fgf17 bioavailability was the primary cause. Analogously, we interpret rescue of the Hs6st1 Ϫ / Ϫ precocious astroglial precursor translocation phenotype in Hs6st1 Ϫ / Ϫ ; Fgf8 neo/neo embryos as evidence that Hs6st1 normally acts to keep FGF/ERK signaling in check rather than as evidence for a selective genetic interaction between Fgf8 and Hs6st1 (Clegg et al., 2014). Conversely, failure to rescue the Hs2st Ϫ / Ϫ phenotype in Fgf17 Ϫ / Ϫ ; Hs2st Ϫ / Ϫ embryos (or using other methods to reduce Fgf17 protein levels or functionality) would not falsify the hypothesis that increased Fgf17 bioavailability caused the Hs2st Ϫ / Ϫ phenotype because there are several alternative explanations. When we used a similar strategy in a similar context to rescue the Hs6st1 Ϫ / Ϫ astroglial precursor precocious translocation phenotype by genetically reducing Fgf8 dosage, the rescue was only successful in a minority of isogenic Hs6st1 Ϫ / Ϫ ;Fgf8 neo/neo embryos and a likely explanation is that compensatory mechanisms act when Fgf gene dosage is manipulated (Clegg et al., 2014). Such compensation will generate false-negative results, making it unsafe to interpret unrescued Fgf17 Ϫ / Ϫ ;Hs2st Ϫ / Ϫ embryos as falsifying the hypothesis that the phenotype is underpinned by excess Fgf17 protein.
There are additional technical confounds that could lead to false negatives because a rescue likely requires precise restoration of normal Fgf17 protein levels (so no rescue could reflect technical failure to restore Fgf17 protein levels to normal) and, in any case, the CC phenotype of Fgf17 Ϫ / Ϫ embryos has not been thoroughly characterized, so Hs2st Ϫ / Ϫ ;Fgf17 Ϫ / Ϫ phenotypes may be problematic to interpret (Cholfin andRubenstein, 2007, 2008). In addition to not being decisive for or against selectivity, demonstrating genetic interaction between Hs2st and Fgf17 would not provide insight into whether the interaction was molecularly direct, in contrast to the biochemical LACE data that we present in the current study.
The closely related Fgf8 subfamily members Fgf17 and Fgf8 are both transcribed by cells in the CSB region, yet have different roles in forebrain development with available evidence, although this does not rule out a role for Fgf17, suggesting that Fgf8 is the primary driver of astroglial precursor translocation required for CC development (Cholfin andRubenstein, 2007, 2008;Moldrich et al., 2010;Toyoda et al., 2010;Gobius et al., 2016). The independent suppression of Fgf17 and Fgf8 protein levels by HS modified by Hs2st and Hs6st1, respectively, may have facilitated the evolution of this system by providing a mechanism to tilt the Fgf17:Fgf8 protein balance to give Fgf8 the more dominant role in regulating astroglial precursor translocation (Clegg et al., 2014;Chan et al., 2017, current study). In this sense, there are parallels to other negative regulatory strategies, such as microRNAs that function by protecting cells from the expression of particular proteins that would be detrimental if expressed.