Sortilin regulates sorting and secretion of Sonic hedgehog Free

Sonic Hedgehog (Shh) is a secreted morphogen that activates a highly conserved signal transduction pathway to regulate patterning and proliferation in developing and adult tissues (reviewed in Jiang and Hui, 2008). Mutations in Shh that affect processing or secretion can cause congenital diseases in humans (Maity et al., 2005; Roessler et al., 1996; Schell-Apacik et al., 2003), highlighting the importance of understanding the mechanisms that control trafficking of this morphogen.

Production of biologically active Shh protein requires a complex sequence of post-translational modifications, including cleavage of the Shh signal peptide upon entry into the endoplasmic reticulum (ER) (Lee et al., 1994; Porter et al., 1995), followed by endoproteolysis in the ER that is catalyzed by the Shh C-terminal (ShhC) intein domain, which separates itself from the Shh N-terminal (ShhN) domain and stimulates the addition of a cholesterol molecule to the C-terminus of ShhN (Aikin et al., 2012; Chen et al., 2011; Porter et al., 1996). ShhN is further modified by the addition of a palmitate moiety at the N-terminus (Pepinsky et al., 1998), yielding the mature ligand (ShhNp). Although the majority of full-length Shh is processed to ShhN and ShhC in the ER (Chen et al., 2011; Huang et al., 2013), there is evidence from Drosophila studies that a pool of full-length Shh escapes processing and is targeted for secretion (Tokhunts et al., 2010).

After processing, ShhNp is trafficked to the plasma membrane, where the lipid modifications promote association of ShhNp with lipid rafts (Karpen et al., 2001). Various mechanisms exist to facilitate trafficking and the release of Hedgehog (Hh) proteins, including solubilization of monomeric and multimeric Hh mature ligands, incorporation into lipoprotein complexes and exosomes, and cytoneme-mediated transport (reviewed in Briscoe and Thérond, 2013; Guerrero and Kornberg, 2014). Evidently, Shh processing and secretion are highly complex and regulated events, with numerous context-dependent mechanisms existing to deliver ShhNp to its target cells. In order to identify new Shh-interacting proteins that have the potential to regulate Shh trafficking, we performed a GST-affinity screen using Shh as bait and rat brain microsomal fractions as prey. We identified sortilin (Sort1) as a candidate Shh-interacting protein.

Sort1 is a multifunctional sorting receptor of the VPS10P-domain receptor family, which includes family members Sort1, SorLA (also known as SORL1), SorCS1, SorCS2 and SorCS3 (Hampe et al., 2000; Hermey et al., 2003; Jacobsen et al., 1996; Petersen et al., 1997; Rezgaoui et al., 2001). Sort1, the prototypic family member, is involved in targeting ligands, including neurotrophins and neurotrophin receptors, to various intracellular compartments, including the endosomal and regulated secretory pathway (RSP) (Chen et al., 2005; Evans et al., 2011; Vaegter et al., 2011; Yang et al., 2011), and lysosomes (Canuel et al., 2008; Evans et al., 2011; Hassan et al., 2004; Lefrancois et al., 2003; Ni and Morales, 2006; Yang et al., 2013; Zeng et al., 2009). Sort1 also functions as a cell surface receptor for neurotrophins and other growth factors (Nykjaer et al., 2004; Teng et al., 2005). Sort1 is produced as a precursor polypeptide, with a short cytoplasmic tail containing sorting domains (Nielsen et al., 2001); a transmembrane region; the VPS10P ligand-binding domain (Marcusson et al., 1994); and a short N-terminal pro-peptide, which promotes proper folding of the immature peptide and inhibits premature binding of the majority of Sort1 ligands before cleavage by furin in the late trans-Golgi network (TGN) (Hermey et al., 2003; Munck Petersen et al., 1999). The contextually diverse functions of Sort1 in regulating intracellular trafficking of ligands make it an intriguing new receptor for Shh. We probed this interaction using in vitro expression of Shh along with various perturbations of Sort1 function in cell lines and primary neuron cultures by examining genetic interactions using Sort1-deficient animals in a sensitized Shh system, as well as by manipulating Sort1 function in Shh-dependent tissue patterning in vivo. Our results support a model in which Sort1 functions as a trafficking receptor for Shh at the level of the Golgi, directing it away from pathways that promote Shh secretion. This inhibitory role for Sort1 on the secretion of Shh is supported by in vivo evidence showing that the Sort1 levels are inversely associated with the levels of Shh-dependent patterning and proliferation.

To identify new Shh-interacting proteins, we performed a GST-affinity purification screen using ShhN–GST and ShhC–GST as bait and rat brain microsomal fraction, in two different detergent preparations, as prey. Eluted proteins were resolved by performing SDS-PAGE, and unique bands were excised and analyzed by using mass spectrometry analysis. Candidates were filtered based on peptide abundance, and common known sepharose bead contaminants (Kocks et al., 2003; Shevchenko et al., 2002; Trinkle-Mulcahy et al., 2008) and cytosolic proteins were excluded from further analysis. We identified a number of new interacting candidates (Table S1) that were involved in diverse cellular functions, including intracellular sorting, receipt of extracellular ligands and protein maturation. The efficacy of the screen was confirmed by the identification of several known Shh interacting proteins, including low-density-lipoprotein receptor-related protein-1 (Capurro et al., 2012) and glypican-5 (Gpc5) (Li et al., 2011). Interestingly, we identified two members of the sortilin family, Sort1 and SorLA, in ShhC and ShhN pulldowns, respectively (Table S1). We prioritized Sort1 for further studies because of its known function in neuropeptide and receptor anterograde trafficking in neurons (Chen et al., 2005; Evans et al., 2011; Vaegter et al., 2011; Yang et al., 2011), and also because of the reported role for the C-terminus of Hh proteins in axonal trafficking (Huang and Kunes, 1996).

The interaction between Sort1 and Shh was confirmed in co-immunoprecipitation experiments in cells expressing Sort1–Myc–His, epitope-tagged ShhN (ShhN-Fc and ShhN-AP) (Fig. 1A), the N- or C- fragment of Shh tagged with fluorescent protein (ShhEYFP-N or ShhECFP-C, respectively) (Fig. 1B), and full-length Shh (ShhFL), which undergoes processing to generate ShhN and ShhC (Fig. 1C). Sort1–Myc–His interacted with both Shh domains and ShhFL (Fig. 1A–C). These data demonstrate the potential of Sort1 to interact with ShhN and the precursor protein ShhFL. To further validate the interaction between Shh and Sort1 in cells, we compared the subcellular distribution of both proteins in co-transfection experiments. Because the C-terminal epitope tag on full-length Sort1 could affect normal trafficking of the protein, an untagged version of full-length Sort1 was used in these and all subsequent experiments. We transfected COS1 cells with wild-type ShhFL and Sort1 or a C-terminal truncated form of Sort1 (tSort), which lacks trafficking domains and accumulates in the TGN (Chen et al., 2005; Nielsen et al., 2001). Overexpression of Sort1 altered the distribution pattern of Shh from a diffuse reticular pattern to a more compact perinuclear distribution that colocalized with staining of Sort1 (Fig. 1D). Expression of tSort resulted in a more striking perinuclear accumulation of Shh (Fig. 1D), which overlapped with that of a TGN marker (Fig. 1G) but not with that of an ER marker (Fig. 1H), indicating a predominant interaction of tSort with the biologically active ShhN fragment. To address whether the colocalization of Sort1 and Shh was caused by close physical proximity of the proteins, we used an in situ proximity ligation assay (PLA). Cells were co-transfected with Shh and Sort1, stained with specific primary antibodies, which was followed by staining with species-specific secondary antibodies conjugated with complementary oligonucleotides (Soderberg et al., 2003, 2008). If the secondary antibodies are within 30 nm, the oligonucleotides can be ligated, amplified and visualized as fluorescent puncta. PLA product was detected in cells that co-expressed Shh and Sort1 (Fig. 1E), and some of the product was localized to the vicinity of the Golgi, based on the distribution of a co-transfected Golgi marker (Fig. 1F). This PLA signal represents physical proximity of the proteins rather than random interactions in the secretory pathway because no PLA signal was detected in cells that had been co-transfected with Shh and BDNF (Fig. 1E,I) or Frizzled4 (Fzd4) and controls (Fig. 1I). These findings indicate that Shh and Sort1 can colocalize to the Golgi and that their interaction is, possibly, direct.

In addition to intracellular protein sorting, a small pool of Sort1 localizes to the plasma membrane where it functions in neurotrophin signaling (Nykjaer et al., 2004; Teng et al., 2005). Thus, we also investigated whether Sort1 expression in Hh-receiving cells would modulate Shh pathway activation. We transfected 3T3 cells with Sort1 or Gpc5, a Shh-binding protein that potentiates Hh signaling (Li et al., 2011), and monitored activation of a Gli-luciferase Hh reporter in response to treatment with recombinant Shh. Sort1 expression had no effect on Shh signaling compared with controls, while expression of Gpc5 strongly enhanced Shh signaling (Fig. 1J). This result suggests that Sort1 interactions with Shh are more relevant to Shh trafficking rather than the response of cells to Shh ligand.

To identify the intracellular compartments where both Shh and Sort1 interact, we applied subcellular fractionation followed by western blot analysis to compare the distribution of Shh and Sort1 across various subcellular compartments. Briefly, CHO cells, wild-type or stably transfected with Sort1 (CHO-sortilin), were transiently transfected with Shh, and individual fractions were blotted for Shh, Sort1 and compartment-specific markers (Fig. 2). In wild-type CHO and CHO-sortilin cells, similar levels of ShhFL and ShhN were detected in fractions corresponding to the ER [marked by glucose-related protein 78 (GRP-78); fractions 4–6] (Fig. 2A). In contrast, in CHO-sortilin cells, there was enrichment of ShhN, and to a lesser extent of ShhFL, in the Golgi [marked by Golgi SNARE 28 (GS28) and the protein ‘vesicle transport through interaction with t-SNAREs homolog 1B’ (Vti1b); fractions 7–15]. We do not think that detection of ShhFL in the Golgi is an artifact of overexpression as we achieved comparable levels of Shh protein expression to those observed in other studies (Chen et al., 2011), and previous studies show that full-length Hh can be targeted for secretion in Drosophila (Tokhunts et al., 2010). The increase in ShhN in CHO-sortilin cells was particularly strong in later Golgi fractions (fractions 13–14), which overlapped with Sort1. This Golgi retention of ShhN in CHO-sortilin cells was also associated with reduced secretion of ShhN in the culture medium (Fig. 2B). Therefore, because elevated levels of Sort1 retain Shh proteins in the Golgi and reduce Shh secretion, we suggest that Sort1 functions as a sorting receptor that binds to ShhN in the Golgi and directs ShhN away from pathways that destine Shh for secretion.

Germline Sort1-knockout (Sort1^−/−) is not associated with developmental defects, including phenotypes associated with altered Hh signaling (Jansen et al., 2007; Zeng et al., 2009). Therefore, to examine whether Sort1 function impacts Shh processing in vivo, we examined its effect in a sensitized system using a processing-deficient Shh-mutant mouse model, Shh::GFP (Chamberlain et al., 2008). The Shh::GFP knockin allele is processed inefficiently owing to the insertion of a fluorescent tag in the N-terminal domain of the protein (Fig. 3A) (Chamberlain et al., 2008). Consistent with this processing defect, homozygosity for the Shh::GFP allele (Shh^GFP/GFP) is embryonic lethal, and is characterized by craniofacial deformities and cyclopia due to loss of Shh function (Chamberlain et al., 2008). Previous work has shown that eye-field defects observed in Shh-deficient mice, including the Shh::GFP model, can be rescued by increasing Hh signaling by reducing the gene dosage of Hh pathway antagonists (Chamberlain et al., 2008; Litingtung and Chiang, 2000). Because Sort1 reduces Shh secretion in vitro, and Sort1 is expressed in the ventral midline at the stage of Shh-mediated patterning (Hermans-Borgmeyer et al., 1999), we reasoned that loss of function for Sort1 would ameliorate some of the Hh-associated defects in Shh^GFP/GFP mice. We first addressed whether Sort1 knockdown altered the secretion of ShhEYFP-N in vitro. We used small hairpin (sh)RNA constructs to generate stable PC12 cell lines that had Sort1 knockdown (Fig. S1), transfected them with ShhEYFP-FL and measured the concentration of ShhEYFP-N in cell supernatants by western blot and ELISA (Fig. 3A). Here, we found that levels of secreted ShhEYFP-N were increased in Sort1-knockdown cells compared to those in control PC12 cells (Fig. 3A). Next, we generated Shh^GFP/GFP animals on wild-type or Sort1-mutant backgrounds (Sort1^+/+ versus Sort1^−/−) and compared the width of the eye field at embryonic day (E)14.5, an age at which bilateral separation of the eye field is readily apparent in wild-type animals (Fig. 3Bi). Remarkably, the optic-field width [as confirmed by neural retina marker Vsx2 expression (Fig. 3Bii)] was significantly increased in Shh^GFP/GFP;Sort1^−/− mice compared with Shh^GFP/GFP;Sort1^+/+ controls (Fig. 3Biii, Biv and B′). However, we did not observe full bilateral separation of the eye field that was comparable to that in Shh^+/GFP mice (Fig. 3Bi and B′), indicating that there was not full rescue of Hh signaling in the compound mutants, which is also consistent with the failure to restore Hh target gene expression in the ventral midline in the compound mutant animals (data not shown). These data are consistent with the interpretation that Sort1 exerts an inhibitory effect on Shh secretion in vivo such that Sort1 deficiency results in a partial rescue of midline defects associated with defective Shh secretion (summarized in Fig. 3C).

In some projection neurons of the central nervous system (CNS), Shh processing and lipid modification is required for targeting to the RSP (Beug et al., 2011; Petralia et al., 2011) and for activity-dependent secretion (Beug et al., 2011). Because Sort1 affects Shh trafficking and secretion in non-neuronal cell lines, we next investigated how Sort1 affects Shh targeting to the RSP in primary neurons. In primary rat retinal ganglion cells (RGCs), which express endogenous Shh and Sort1 (Fig. 4A), Shh and Sort1 exhibited low levels of colocalization in the soma and the axon (Fig. 4A and A′). We observed a similar distribution and colocalization pattern between endogenous Sort1 and transfected Shh in primary cortical neurons, confirming the efficacy of these neurons in subsequent Sort1 perturbation studies (Fig. S2A). We next examined the effect of Sort1 knockdown on Shh axonal targeting by examining two different pools of axonal Shh – extracellular (where it accumulates on lipid rafts on the plasma membrane) and intracellular (where it is associated with regulated secretory vesicles), which both require proper Shh trafficking (Beug et al., 2011). When we compared the distribution of Shh in cortical neurons as a function of Sort1 expression, we found an inverse relationship between Sort1 expression and the levels of Shh in the axon (Fig. 4; Fig. S3). RNA interference (RNAi)-mediated knockdown of Sort1 with two shRNA constructs in cortical neurons (shSort1A and shSort1B, validation shown in Fig. S1) increased the colocalization of Shh with synaptic vesicle glycoprotein 2A (SV2)+ RSP vesicles and extracellular Shh (Fig. 4C) in axons. We observed similar effects on Shh distribution in cortical neurons from Sort1^−/− mice (Fig. 4D; Fig. S3). Re-introduction of Sort1 expression in Sort1^−/− neurons significantly reduced Shh levels in the axons (Fig. 4E; Fig. S3), and reduced, albeit not significantly, colocalization of Shh with SV2+ vesicles (Fig. 4E; Fig. S3), suggesting that the effect was specific to loss of Sort1. The effect of Sort1 on Shh distribution in the axon was specific to Shh, as Sort1 knockdown had the opposite effect on hemagglutinin (HA)-tagged BDNF distribution, abolishing axonal trafficking (Fig. S4A), consistent with Sort1 directing BDNF–HA to the axon (Chen et al., 2005). There was no apparent redirection of Shh from other subcellular compartments because colocalization of Shh with lysosomal or endosomal markers was unchanged in Sort1^−/− neurons (Fig. S4B). Conversely, increased Sort1 expression in wild-type cortical neurons reduced the extent of the Shh–SV2 colocalization in the axon and the level of extracellular Shh on the axon (Fig. S2B,C). Sort1 overexpression did not alter the total level of SV2+ vesicles in the axons (Fig. S4C), indicating that the effect of Sort1 on axonal Shh was not secondary to a general perturbation of biogenesis or trafficking of SV2+ vesicles in the axon. These results are consistent with the notion that Sort1 inhibits Shh targeting to axons and SV2+ vesicles, which is independent of trafficking to the lysosome or endosome, and is independent from its regulation of BDNF trafficking in axons.

We next investigated the requirement for Sort1 trafficking in neuron-derived Shh signaling in vivo. We focused on RGCs, the axons of which project from the retina through the optic nerve to targets in the brain. In addition to its short-range effects in the regulation of proliferation of adjacent neural progenitor cells in the retina (Wang et al., 2005), Shh is also transported down RGC axons, where it is released in the optic nerve to stimulate astrocyte proliferation and oligodendrocyte precursor cell migration (Dakubo et al., 2003, 2008; Soukkarieh et al., 2007; Wallace and Raff, 1999). To perturb Sort1 trafficking, we used an overexpression approach in vivo. In control experiments, we determined that tSort expression in cortical neurons caused severe impairment of Shh axonal targeting (Fig. 5A,A′). Consistent with inhibiting RSP targeting, tSort expression in differentiated PC12 cells [which differentiate into a neuronal phenotype that has neuron-like processes and activity-dependent RSP secretion of neuropeptides (Greene and Tischler, 1976)] resulted in a decrease in activity-dependent Shh secretion relative to control conditions (Fig. 5B), consistent with an antagonistic role for tSort in stimulated Shh secretion.

We next asked whether tSort expression in embryonic RGCs in vivo affects Shh-dependent proliferation in the optic nerve and retina. Control or tSort expression vectors were co-transfected with YFP into the eye at E13.5 by performing in utero electroporation, and the tissue was harvested at E15.5 after a 2-h pulse with 5-ethynyl-2′-deoxyuridine (EdU) to label cells in S-phase (schematic in Fig. 6A). Transfection of RGCs was confirmed by the presence of YFP in the cell bodies located in the retina and axons in the optic nerve (Fig. 6B). At this stage, the majority of the cells in the optic nerve are Pax2+ astrocyte precursor cells (Fig. 6C), which are not electroporated in this procedure. Remarkably, the proportion of EdU+ nuclei in the optic nerve was significantly reduced in samples into which tSort had been electroporated intraocularly relative to that in the control (Fig. 6Bi–Bviii and B′). The proportion of EdU+ cells in the retina of tSort-electroporated animals was also reduced relative to control conditions, albeit not significantly (Fig. 6Bix–Bxvi and B″). This effect of tSort expression on proliferation was restricted to glial cells in the optic nerve, as proliferation of mesenchymal cells located adjacent to the nerve was unaffected (Fig. 6B″). Because we observed YFP+ axons and Pax2+ glial cells in the optic nerve (Fig. 6B,C), we ruled out RGC cell death and impaired astrocyte development as alternative interpretations for the tSort-dependent reduction in astrocyte proliferation. Taken together, these findings indicate that tSort expression in the retina exerts a distal non-cell autonomous effect in the optic nerve. Because Shh is the only reported anterograde signal that controls astrocyte proliferation in the optic nerve (reviewed in Tao and Zhang, 2014), our findings are also consistent with the interpretation that tSort interferes with Shh signaling in the axon.

Our study provides the first evidence of a role for Sort1 in modulating Shh trafficking in cell lines and primary neurons. Sort1 can interact with unprocessed, as well as processed, ShhN and ShhC peptides, and overexpression of Sort1 alters the distribution of Shh, retaining it in the Golgi compartment and inhibiting its secretion. Taken together, these data are consistent with a model in which Sort1 acts as a sorting receptor for Shh in the Golgi to direct it away from pathways that promote Shh secretion. Consistent with this model, loss of Sort1 partially rescued Hh-dependent eye-field defects in an Shh-processing-deficient mutant in vivo. Furthermore, Sort1 levels were inversely correlated with polarized trafficking and secretion of Shh in vitro, and Sort1 trafficking was required for axon-dependent effects on glial proliferation in vivo.

Shh-mediated tissue patterning begins with the production and processing of the full-length precursor protein (Lee et al., 1994; Pepinsky et al., 1998; Porter et al., 1995, 1996), whereupon mature ShhNp is secreted in a short and long range manner through a number of mechanisms (Briscoe and Thérond, 2013; Callejo et al., 2011; D'Angelo et al., 2015; Gradilla et al., 2014; Guerrero and Kornberg, 2014; Matusek et al., 2014; Parchure et al., 2015; Vyas et al., 2014). In neurons, Shh utilizes the RSP as a mode of anterograde transport in axons (Beug et al., 2011); therefore, our identification of Sort1 from a GST-affinity screen provided a candidate sorting receptor for Shh. Our data suggest that the interaction between Shh and Sort1 occurs largely in the TGN, based on colocalization and cellular fractionation studies. This is an interesting new concept that supports Sort1 function in regulating the intracellular fate of ShhN. Identifying other Sort1-mediated trafficking pathways for Shh, including a possible role in the recycling of the morphogen, will be an interesting direction for future studies.

The newly identified role for Sort1 as a regulator of Shh secretion led us to predict that Sort1-knockout animals would exhibit Shh gain-of-function phenotypes, because in the absence of Sort1, more Shh would be directed to secretion pathways. However, germline knockout of Sort1 is not associated with gross phenotypic changes in Hh-dependent tissues. This is not unanticipated because Sort1 knockout is associated with subtle phenotypes that do not impact development (Vaegter et al., 2011); for example, Sort1-knockout mice have no overt neurotrophic phenotypes despite its known roles in neurotrophin receptor trafficking. However, Sort1 knockout aggravates neurotrophic phenotypes when crossed on to a sensitized background – p75 neurotrophin receptor knockout (Vaegter et al., 2011). Moreover, endogenous Shh-responsive systems are sensitive to low concentrations of the ligand, as evidenced by the viability of the Shh heterozygous knockout (Chiang et al., 1996). Finally, our screen identified SorLA as another potential new interacting partner of Shh; therefore, there could be compensatory effects mediated by SorLA upon loss of Sort1. Thus, it is possible that any Hh-like phenotypes in Sort1-knockout animals are masked by a combination of compensatory activity from sortilin-family members and the sensitivity of the system to a range of Shh ligand concentrations.

We have therefore investigated genetic interactions between Sort1 and the Hh pathway using sensitized in vivo models of Hh signaling. Loss of Sort1 has no impact on the incidence or latency of medulloblastoma in Ptch^+/− mice (Goodrich et al., 1997), which is dependent on Shh pathway over-activation (C.C. and V.W., unpublished observations). However, reducing Sort1 levels in the context of Shh::GFP homozygosity, characterized by midline defects due to deficient Shh processing, improved eye-field patterning, an Shh-regulated process. The rescue was incomplete, as shown by the lack of expression of midline target genes that require the highest dose of Shh (C.C. and V.W., unpublished observations), indicating that Sort1 knockout does not fully restore Shh signaling.

We have previously reported that Shh is trafficked to the RSP in primary neurons (Beug et al., 2011). Our observation that Sort1 levels negatively correlate with those of Shh in the axon is consistent with the hypothesis that Sort1 also functions in directing Shh away from the RSP pathway. This is notably different from Sort1-dependent anterograde trafficking of BDNF and TRK receptor to axons and to the RSP (Chen et al., 2005; Evans et al., 2011; Vaegter et al., 2011; Yang et al., 2011). This suggests the possibility that Sort1 is the RSP receptor for BDNF and TRK receptors but there exists an additional, as yet unidentified, receptor that likewise regulates Shh targeting to the RSP.

Optic nerve axons have been shown to promote survival of oligodendrocytes and the proliferation of oligodendrocyte precursor cells and astrocytes (Barres et al., 1993; Barres and Raff, 1993; Burne and Raff, 1997). Shh is the only signal that has been identified to date that mediates the effects of axons on astrocyte proliferation (Dakubo et al., 2008; Wallace and Raff, 1999). We showed that ectopic expression of tSort in RGCs induces a non-cell autonomous effect on astrocyte proliferation in the optic nerve. Thus, in addition to its role in regulating RGC survival in development and in injury models (Lebrun-Julien et al., 2010; Nakamura et al., 2007), our study identifies a second function for Sort1 in axon–glial-cell communication. Our findings also raise the interesting possibility that Sort1 function extends to other aspects of optic nerve biology, including oligodendrocyte development and myelination.

Taken together, we have newly identified an interaction between Sort1 and Shh, wherein Sort1 regulates trafficking of ShhN to the secretory pathway and thereby influences the bioavailability of Shh. This interaction has implications for Shh signaling in developing tissues, including neuroepithelial cells at the CNS midline, and in polarized trafficking of the protein in projection neurons. Considering the ubiquitous expression of Sort1 in known Shh-producing cells, it is possible that Sort1 could regulate Shh-mediated patterning in additional tissues not studied here.

Shh-FL, ShhEYFP-FL, ShhEYFP-N and ShhECFP-C constructs were prepared as previously described (Beug et al., 2011). Full-length Shh cDNA (Genbank accession no. NM_009170) was used to generate GST fusion constructs. Shh-N (amino acids 25–198) and Shh-C (amino acids 199–437) were amplified by using PCR and subcloned into the GST recombinant vector pGEX-4T-1 (GE Healthcare) to generate GST–ShhN and GST–ShhC, respectively. pEF, pEF-Sort1 cDNAs were gifts from Dr Peder Madsen (Aarhus University, Aarhus, Denmark). Sort1 was obtained from the IMAGE Consortium and subcloned into pcDNA3.1myc-His(-)C by performing PCR amplification. tSort–Myc was a gift from Dr Carlos Morales (McGill University, Montreal, Québec) and was subcloned into the pCAGGS vector for the electroporation studies. Glypican-5-FLAG was tagged with 3×FLAG upstream of the cleavage site, and subcloned into pcDNA3. BDNF–HA was a gift from Dr Francis Lee (Cornell University, Ithaca, NY). Short hairpin constructs targeting Sort1 were obtained from OriGene. Renilla–luciferase was a gift from Dr Alan Mears (Children's Hospital of Eastern Ontario, Ottawa, ON), Gli-responsive luciferase reporter construct was a gift from Dr Hiroshi Sasaki (Osaka University, Osaka, Japan).

Microsomal fractions were obtained from rats at postnatal days 2–3 as follows. Brains were Dounce homogenized in SIM buffer (250 mM sucrose, 5 mM imidazole, 1 mM MgCl₂, pH 7.4), centrifuged at 10,000 g for 30 min, and the supernatant was centrifuged at 100,000 g for 60 min. The pellet was resuspended in SIM buffer and centrifuged at 100,000 g for 60 min. The resulting pellet represented the microsomal fraction and was solubilized with either 0.5% CHAPS or NP-40 in 20 mM HEPES-KOH pH 7.4, 150 mM NaCl, 1 mM MgCl₂ and clarified by centrifugation at 20,000 g for 30 min. Solubilized microsomes were precleared through incubation with glutathione-coupled Sepharose beads (GE Healthcare). The precleared fraction was incubated with equimolar amounts of GST, GST–ShhN or GST–ShhC coupled to glutathione-bound beads. Beads were washed sequentially with buffer (20 mM HEPES-KOH, pH 7.4), high-salt buffer (20 mM HEPES-KOH pH 7.4, 300 mM NaCl), detergent buffer (20 mM HEPES-KOH pH 7.4, 300 mM NaCl and 0.1% CHAPS or NP-40) and PBS. Bound proteins were eluted by cleavage with thrombin (GE Healthcare) and resolved on precast 8–16% SDS-PAGE gels (Bio-Rad). Gels were silver stained, and bands were excised and identified by performing mass spectrometry using Qstar and LTQ mass spectrometers (Ottawa Institute of Systems Biology, Ottawa, ON).

COS-1 (American Type Culture Collection) and 3T3 (American Type Culture Collection) fibroblasts were maintained in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies) supplemented with 10% fetal bovine serum (FBS). PC-12 cells (American Type Culture Collection) were maintained in RPMI-1640 (Life Technologies), supplemented with 10% horse serum, 5% FBS, 2 mM L-glutamine, and 1× penicillin-streptomycin. For differentiation, PC-12 cells were plated on collagen-IV-coated (Sigma) plates (20 μg/cm²) in RPMI-1640 supplemented with 1% horse serum, 0.5% FBS, 2 mM L-glutamine, 1× penicillin–streptomycin and 50 ng/ml NGF2.5S (Millipore), and were incubated for 5–7 days. Primary RGCs were isolated from rat pups at postnatal day 4, as previously described (Barres et al., 1988). Purified RGCs were plated onto laminin- and poly-D-lysine-coated coverslips, and cultured in differentiation medium, comprising Neurobasal-A medium (Life Technologies) supplemented with 1× B-27 (Life Technologies), 1× Sato medium, 5 µg/ml insulin (Sigma), 1 mM sodium pyruvate (ThermoFisher), 1× T3 (Sigma), 1× N-acetyl cysteine (Sigma), 0.5 mM L-glutamine and gentamicin (Life Technologies). Primary cortical neurons were isolated by trypsinization (Life Technologies) of E14.5-mouse cortices and plated onto poly-D-lysine-coated (Sigma) coverslips (Zeiss) in Neurobasal-A medium (Life Technologies) supplemented with 1× B-27 (Life Technologies), 1× Sato medium, 0.5 mM L-glutamine and gentamicin (Life Technologies). RGCs were prepared and cultured as previously described (Barres et al., 1988). All cells were maintained at 37°C under 5% CO₂. COS-1 and PC12 cells, and primary neurons were transfected with Lipofectamine 2000 (Invitrogen) and 3T3 cells with Trans-IT (Mirus), as per the manufacturer's instructions. Stable transfectants were selected over two weeks in media containing either 1 μg/ml puromycin (Sigma) or 400 μg/ml Geneticin (Life Technologies). For Sort1 knockdown, four functional short hairpins and one scrambled control were tested, and efficiency of Sort1 knockdown was determined by western blotting. In cortical neurons, shRNA-expressing neurons were identified through GFP expression, and Sort1 knockdown was confirmed through a lack of Sort1 staining (immunocytochemistry).

Five 15-cm culture dishes of CHO and CHO-sortilin cells were transiently transfected with an expression construct for murine Shh using Lipofectamine 2000 from Invitrogen. After 48 h, cells were harvested and resuspended in 3 ml of homogenization buffer containing 10 mM HEPES, pH 7.4, 1 mM EDTA and 0.25 mM sucrose plus protease inhibitor cocktail Cømplete from Roche. Thereafter, cells were fractionated by performing a discontinuous iodixanol density gradient protocol, as described previously (Chang et al., 2003), and fractions were blotted for organelle-specific markers, Sort1 and Shh.

Cells were lysed in ice-cold RIPA buffer with protease inhibitors (Roche), and protein concentration was determined using a standard Bradford assay. Lysates were subjected to SDS-PAGE, and proteins were transferred on to Hybond-C Extra membranes (Millipore). Membranes were incubated overnight at 4°C with primary antibodies (Table S2), followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies (Millipore) and were developed with an enhanced chemiluminescence (ECL) detection kit (Millipore). Densitometry was performed on bands using the blot analyzer function in ImageJ. For co-immunoprecipitation, transfected cells were lysed in ice-cold buffer containing 10 mM HEPES-NaOH pH7.4, 100 mM NaCl, 1% NP-40, 10% glycerol, and protease inhibitors (Roche). Clarified lysates were pre-cleared with Protein-A beads (Invitrogen) and then incubated with primary antibodies (Table S2) or species-matched nonspecific IgG, then precipitated using fresh Protein-A beads. Protein complexes were eluted from beads using SDS loading buffer, and analyzed by performing SDS-PAGE.

PC12 cells cultured in differentiation-inducing conditions were transfected with ShhFL, and subsequently incubated for 5–7 days. Cells were washed twice with pre-warmed basal medium [5.6 mM KCl, 145 mM NaCl, 2.2 mM CaCl₂, 0.5 mM MgCl₂, 5.6 mM glucose, 15 mM HEPES-KOH (pH 7.4) and 0.1 mg/ml BSA], and incubated for 60 min with basal medium or stimulating medium [56 mM KCl, 95 mM NaCl, 2.2 mM CaCl₂, 0.5 mM MgCl₂, 5.6 mM glucose, 15 mM HEPES-KOH (pH 7.4) and 0.1 mg/ml BSA] with 1 mg/ml heparin. Cell culture supernatants were centrifuged at 10,000 g to remove cellular debris, and were analyzed by performing an ELISA, as per the manufacturer's instructions (R&D Systems). For luciferase assays, 3T3 cells were transiently transfected with Renilla–Luciferase and Gli–luciferase along with the indicated plasmids, serum-starved for 24 h and incubated with recombinant Shh (R&D Systems) for 24 h. Luciferase and Renilla activity was then measured, as per manufacturer instructions (Promega).

COS1 cells (10,000) were plated on poly-D-lysine+laminin-coated 16-well chamber slides (Nunc, Sigma) for 24 h. A serial dilution of plasmid concentrations (50, 100, 250, 500 and 1000 ng) was co-transfected (SortFL+Shh; BDNF–HA+Shh; and Fzd4–Myc) using Lipofectamine 3000 as per the manufacturer's instructions (ThermoFisher). Co-transfection with pMAX-GFP in preliminary studies was used as a transfection control. Single-plasmid transfections, no primary antibody and no plus-probe conditions were used as negative controls for PLA product synthesis. Cells were briefly washed in PBS and fixed for 10 min in 4% paraformaldehyde (Electron Microscopy Systems). Proximity ligation was performed as per manufacturer's instructions (Duolink Proximity Ligation Assay kit; Sigma). Optimal combinations of plasmid (100 ng per plasmid per well) and antibody (mouse anti-5E1, Developmental Studies Hybridoma Bank, 1:20,000; rabbit anti-sortilin, Abcam ab16640, 1:5000; rabbit anti-HA, Santa Cruz, Y-11, 1:5000; mouse anti-Myc, Sigma, PLA0001, 1:20,000) dilutions were determined in preliminary experiments. Co-transfection with the CellLight^® Golgi-GFP, BacMam 2.0 (ThermoFisher) reporter (5 µl per well) was used as a trans-Golgi counterstain. Images (2048×2048) were acquired on a Zeiss LSM780 confocal microscope using a 40× water immersion objective. Pinhole diameter was set to 1.0 AU, and all laser and detector settings maintained across experimental groups. z-stacks were converted to 8-bit grayscale and imported into Blobfinder (Centre for Image Analysis, Uppsala University). Thresholding parameters used were as follows: minimum nucleus size in pixels squared=2800; cytoplasm size (radius)=100 pixels; blob size=3×3; blob threshold=50. Following Levene's test for homogeneity of variance tests, data were analyzed using a one-way ANOVA and Scheffé's post hoc tests in SPSS 21 software.

Animal husbandry was performed in accordance with the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research, and the University of Ottawa Animal Care Committee. Wild-type CD1 mice (Charles River) were used for the isolation of primary cortical neurons. Sort1-mutant (Sort1^−/−) mice (genotyped by performing PCR using the following primers: F - 5′-CTCAGGAATGGCATTCTCAG-3′, R - 5′-AGCCTTTACCTGGTGTCATC-3′) (Zeng et al., 2009) and Shh::GFP mice (Jackson Labs) [genotyped as described previously (Chamberlain et al., 2008)] were maintained on a C57/Bl6J background (Charles River). F2 mice from Sort1^+/−×Shh^+/GFP matings were mated to generate Sort1;Shh::GFP double mutants and littermate controls at E11.5 and E14.5. Mice were mated in the afternoon, and the presence of a vaginal plug the next morning was considered E0.5. In utero electroporation was performed as previously described (Garcia-Frigola et al., 2007) at E13.5, and cells in S-phase were labelled in vivo through an intra-peritoneal injection of the dam with EdU (Life Technologies) 1 h before mice were killed at E15.5.

Cells cultured on poly-D-lysine-coated coverslips were fixed in 4% PFA for 5–20 min and permeabilized with 0.1% Triton-X100 (as indicated). Fixed cells were incubated overnight at 4°C with primary antibodies (Table S2), followed by incubation with fluorophore-conjugated secondary antibodies (1:1000, Life Technologies). Nuclei were stained with Hoechst 33342 (1:25,000; Invitrogen). Coverslips were mounted on permafrost slides (Fisher) with fluorescent mounting medium (Dako). Embryonic tissue was fixed overnight in 4% PFA, cryoprotected in sucrose and frozen in 50:50 30% sucrose:OCT. In situ hybridization was performed on frozen sections using a digoxigenin-labeled antisense probe against Vsx2. For immunohistochemistry, frozen sections were stained with rabbit anti-GFP and goat anti-Pax2 (1:200; Covance) antibodies (Table S2), and nuclei were stained with Hoechst 3342, and EdU was detected using the 543-nm fluorophore click-IT reaction kit (Life Technologies). For hematoxylin and eosin staining, frozen sections were rehydrated with PBS, stained with hematoxylin, washed and dehydrated using an ethanol series, stained with eosin and rehydrated before mounting.

Immunohistochemistry and immunocytochemistry samples were imaged using a Zeiss LSM510 META confocal microscope with 63×1.4 oil Plan-Appochromat or 20×0.5 EC Plan-Neofluar objectives and the Zen 2009 software; or a Zeiss Axio Imager M2 instrument with 63×1.4 oil Plan-Apochromat or 20×0.8 Ph2 Plan-Apochromat objectives and AxioVision Rel. 4.8 software. Only cells with an intact nucleus and low-to-medium expression of transfected constructs were imaged. In neuron images, axons were identified by length (the longest process was at least ten times the diameter of the neuron body in length), protrusion of few branches and by positive staining for Tau. For distribution analysis, 1-µm optical sections of individual neurons were imaged using a constant gain, and the pixel intensity was measured using ImageJ, with the mean ratio of signal in the axon compared to the soma calculated from measurements of the indicated number of neurons. Colocalization with organelle markers was quantified in 1-µm sections of individual neurons using the Intensity Correlation Analysis plug-in in ImageJ, and mean Pearson's correlation coefficients were compared between conditions.

We thank Dr Francis Lee for the BDNF–HA construct, and Dr Peter Madsen for sortilin constructs. We would also like to thank Sherry Thurig, Katy Morin and Sheila Smiley for performing animal husbandry; Melanie Liekweg and Christine Kruse for expert technical assistance; and Dr Alan Mears for work on sortilin genotyping. Finally, we thank Drs Carlos Morales and Dr Frederic Charron for their insightful comments and suggestions.