A novel proteolytic event controls Hedgehog intracellular sorting and distribution to receptive fields

ABSTRACT The patterning activity of a morphogen depends on secretion and dispersal mechanisms that shape its distribution to the cells of a receptive field. In the case of the protein Hedgehog (Hh), these mechanisms of secretion and transmission remain unclear. In the developing Drosophila visual system, Hh is partitioned for release at opposite poles of photoreceptor neurons. Release into the retina regulates the progression of eye development; axon transport and release at axon termini trigger the development of postsynaptic neurons in the brain. Here we show that this binary targeting decision is controlled by a C-terminal proteolysis. Hh with an intact C-terminus undergoes axonal transport, whereas a C-terminal proteolysis enables Hh to remain in the retina, creating a balance between eye and brain development. Thus, we define a novel mechanism for the apical/basal targeting of this developmentally important protein and posit that similar post-translational regulation could underlie the polarity of related ligands.


INTRODUCTION
The Hedgehog (Hh) family encodes secreted morphogens with roles in patterning, stem cell maintenance and neoplastic disease (Jiang and Hui, 2008;Scales and de Sauvage, 2009;Varjosalo and Taipale, 2008). A unifying and unresolved question concerning these activities is how they are shaped by the secretion and transport mechanisms that deliver Hh to receptive cells. A number of recent studies have documented the important role of secretory targeting in Hh activity (Briscoe and Thérond, 2013). Hh is released apically or basally in large multimeric or small monomeric forms, which are believed to act as long-and short-range signals, respectively (Ayers et al., 2010;Gallet et al., 2003Gallet et al., , 2006Panáková et al., 2005). The interplay between apical and basal release mechanisms can be complex and interdependent (Callejo et al., 2011). Moreover, it has become clear that patterning previously thought to rely on diffusion in extracellular space might instead involve actin-based cellular extensions (e.g. cytonemes) that transport Hh over many cell diameters prior to release (Rojas-Ríos et al., 2012;and reviewed in Kornberg, 2011).
The central role of secretory mechanisms in Hh activity is illustrated by its segregation between two receptive fields in the developing Drosophila compound eye. Hh is synthesized by differentiating photoreceptor neurons and released both apically into the retina, where it propagates a developmental wave of retinal differentiation, and basally, after transport along photoreceptor axons, into the brain, where it induces differentiation of the photoreceptor's postsynaptic target neurons (Fig. 1A and B;Huang and Kunes, 1996;Roignant and Treisman, 2009). Partitioning Hh for release at opposite poles of the photoreceptor neuron is a critical feature of establishing the coordinated development of synaptic partner neurons and their assembly into a precise neural circuit.
How might Hh be partitioned for release at opposite poles of the photoreceptor neuron? Hh is composed of N-terminal and C-terminal domains that dissociate in a self-catalyzed proteolytic cleavage reaction (Lee et al., 1994). The N-terminal product HhNp, modified by cholesterol during self-cleavage, harbors all known Hh signaling activities (Porter et al., 1996). When synthesized in the absence of the C-terminal domain (and hence lacking cholesterol modification), the N-terminal domain is aberrantly targeted and released selectively into the retina (Chu et al., 2006). We previously described a conserved amino acid signal on the C-terminal domain that can override retinal localization, sending both self-cleavage products down the photoreceptor axons for release into the brain (Chu et al., 2006). The question remains, however, how the C-terminal domain, dissociated by self-cleavage, could control secretory targeting, especially of the N-terminal domain, HhNp. The expected products of Hh self-cleavage include the 24 kDa C-terminal domain, HhC 24 (Lee et al., 1994), which harbors the axonal targeting motif near its carboxyl terminus (Chu et al., 2006;Lee et al., 1994). We observed that a significant fraction of HhC in photoreceptor neurons is in the form of a 16 kDa polypeptide (HhC 16 ; Fig. 1C), an isoform that has been previously observed (Lee et al., 1994;Mastronardi et al., 2003). Here we show that this shortened HhC isoform lacks the axonal targeting motif and that HhC cleavage controls the distribution of Hh between the developing eye and brain. We show that this binary targeting decision involves a pathway choice. Hh with an intact C-terminus enters the axon and is secreted from growth cone tips into the brain. HhNp associated with the shortened isoform, HhC 16 , takes an apical pathway and is responsible for the progression of retinal development. Thus C-terminal proteolysis allows Hh to remain in the retina, creating a balance between eye and brain development.

RESULTS
Selective axon transport of HhC 24 , a long-form of the Hh selfcleavage product The biosynthetic maturation of Hh includes proteolysis and lipid modification coupled to movement through the secretory pathway. Upon translocation into the endoplasmic reticulum (ER), the N-terminal secretion signal sequence is removed to yield the 46 kDa polypeptide Hh-Uncleaved (HhU; Fig. 1C). HhU undergoes an intramolecular self-cleavage reaction that yields the 19 kDa cholesterol-modified N-terminal HhNp and a 24 kDa C-terminal fragment, HhC 24 that harbors the self-cleavage catalytic domain. Interestingly, an antibody specific to the C-terminal product (Lee et al., 1994) also recognized a shorter fragment of ∼16 kDa (HhC 16 ; Fig. 1C, and see Hh processing diagram in Fig. S1). The two polypeptides, HhC 16 and HhC 24 , were observed with an additional anti-Hh antibody (Tabata and Kornberg, 1994) and after the expression of Hh isoforms tagged by hemaglutanin antigen (HA) insertion carboxyl-terminal to the self-cleavage site (see below). Moreover, both HhC polypeptides were observed in developing visual system extracts, where only native Hh is expressed (Fig. S2A). A ∼16 kDA polypeptide evidently derived from HhC has been noted previously in Drosophila embryo extracts (Lee et al., 1994).
To determine which Hh polypeptides transit photoreceptor axons for secretion into the brain, the developing visual system was examined by immunohistochemistry and western blot analysis. For the latter, the developing eye-brain complex, purified from third larval instar animals, was separated into eye and brain fractions by cutting the optic stalk that serves as the portal for photoreceptor axons to enter the brain (Fig. 1A). As previously reported (Chu et al., 2006), HhU was observed only in the retinal fraction ( Fig. 1C), indicating that self-cleavage precedes axon transport. The retinal fraction also contained HhNp and both C-terminal species, HhC 24 and HhC 16 . In contrast, only HhC 24 and HhNp were concentrated in the brain fraction; HhC 16 was virtually absent. To localize HhC 16 and HhC 24 in intact tissue, immunohistochemistry was used to detect an epitope-tagged Hh isoform [Hh CHA3 ] that was expressed specifically in the retina. In the construct hh CHA3 , an HAtag inserted at amino acid 267, in HhC, detects HhC 24 but not HhC 16 (see diagram in Fig. S1, 'Construct #13' in Table S1, and Fig. S2B and C). In these animals, punctate anti-HA labeling was concentrated in distal axons and growth cones in the brain (Fig. 1B, top right panel). HhC 24 -positive puncta were also found in the basal region of photoreceptor cell bodies and axons in the eye imaginal disc (Fig. 1D,F″ and H′), and absent from the apical membrane of cell bodies (Fig. 1F″,H). In contrast, an anti-HhC antibody that detects both HhC 16 and HhC 24 (Fig. 1C, bottom panels) (Lee et al., 1994) revealed puncta strongly concentrated in the apical membrane of photoreceptor cell bodies (Fig. 1E′,F′,G and H), in addition to axons and axon termini (Fig. 1B,lower right panel). The signaling domain, HhNp was co-localized with HhC in puncta in both the apical and basal regions, overlapping the HhC 24 -positive puncta in axons and at axon termini, and the presumptive HhC 16 -positive puncta at the apical tips of the photoreceptor cell bodies (Fig. 1E″,G). Hence, the long and short HhC isoforms displayed subcellular localization of opposing polarity, but were nonetheless co-localized with HhNp in both cases.

C-terminal cleavage follows Hh autoprocessing in the ER
The two HhC isoforms, HhC 16 and HhC 24 , display differential targeting to axons and transport into the brain (Fig. 1). We thus considered the possibility that these Hh isoforms might enter distinct intracellular trafficking pathways.
Upon translocation into the ER, Hh's N-terminal secretory signal is removed to yield the 46 kDa HhU (Fig. 1C). HhU self-cleavage to yield HhNp and HhC 24 is thought to occur in the ER (Campbell et al., 2016;Chen et al., 2011). To determine how HhC 24 and HhC 16 are processed, we employed Hh-expressing cell lines which enabled us to observe the generation and degradation order of Hh products, the secretion of each fragment, and the organellar compartments in which these events occur. These approaches defined the ER as the site of HhC 16 generation and demonstrate two distinct secretory outcomes for the HhC fragments We first examined Hh processing in a Drosophila larval CNSderived cell culture system that recapitulates proteolysis yielding the three products: HhNp, HhC 24 , and HhC 16 ( Fig. 2A). These products were also observed in an eye-antennal disc-derived cell line and the S2 line (data not shown). A short pulse of hh NHA expression induced from a heat-shock cassette (hsp 70 -hh NHA ) resulted in the appearance of HhU after a 20-min heat shock. The self-cleavage products HhNp (HA-tagged) and HhC 24 were coincidently detected in significant amounts at 15 min after heat-shock induction ( Fig. 2A). However, another hour passed before HhC 16 was detected ( Fig. 2A). This indication that HhC cleavage follows self-cleavage is consistent with the observation that self-cleavage mutants hh C258A and hh 441STOP did not produce shortened isoforms consistent with C-terminal cleavage in the absence of self-cleavage (Fig. S3A). (Top) Photoreceptor neurons differentiate temporally with the posterior-toanterior progression (right to left) of the morphogenetic furrow across the eye disc. These neurons project their axons (R1-R8) into the brain through the optic stalk, where they spread to dorsal and ventral retinotopic positions (dorsal is up). (Bottom) The R1-R6 axons terminate in the lamina (lam), while R7 and R8 axons terminate in the deeper medulla ganglion. Hh secreted from developing photoreceptor neurons is required for both eye and lamina development. (B) Micrographs showing the distribution of Hh protein (Hh NHA or Hh CHA3 ) expressed with the eye-specific driver GMR-GAL4. Scale bar: 20 µm. (Left panels) HhNp visualized with α-HA antibody staining from the lateral (top left) and horizontal (bottom left) perspectives, as described in A. HhNp (derived from Hh NHA ) is sequestered in puncta in the retinal cell bodies (ed), axons in the optic stalk (os), and growth cones (gc) in the lamina. (Right panels) HhC visualized using either anti-HA antibodies that recognize only full length HhC 24 (derived from Hh CHA3 ; top right) or anti-HhC antibodies (bottom right), which recognize all HhC polypeptides (HhC 16 and HhC 24 ; see C). HhC 24 signal is highly concentrated at the growth cones (top right) while the HhC 24 /HhC 16 staining is evenly distributed from cell body to axon termini (bottom right). (C) Western blot analysis was performed on protein extracts from adult heads (lane 1) or the eye/brain complex of third instar larvae (larval; lanes 2-4) expressing the Hh NHA polypeptide with the GMR-GAL4 driver. In lanes 3 and 4, the eye/brain complex was dissected to separate the eye disc (ed, lane 3) from the optic stalk and brain (bn, lane 4). Only a small fraction of material is detected as uncleaved precursor product (lane 3, top); nearly all Hh in the brain is self-cleavage product (lane 4, top panel). Uncleaved Hh (HhU), HhC 24 , and HhC 16 are detected in the eye disc (lane 3, bottom), nearly all HhC in the brain isolate is HhC 24 (lane 4, bottom). (D) Horizontal schematic of the eye disc showing photoreceptor cell bodies (cb, gray) with their apical tips highlighted (blue). Differentiation proceeds in a posterior ( pos) to anterior (ant) wave (direction indicated by arrow), with the onset of ommatidal development on the left at the morphogenetic furrow (mf ), and more advanced ommatidia at the posterior (right). (E,F) Horizontal views of the late third instar visual system in animals expressing UAS-hh CHA3 (E) or UAS-hh NHA (F) under the control of the pan-neural driver elav-GAL4. Larval brains were stained with anti-HhC (to visualize all HhC isoforms) or anti-HA antibodies to visualize HhN (E) or HhC 24 (F). Anti-HRP stains all neuronal membranes and is concentrated at the apical tips of the photoreceptors (see blue color in D). HhC 24 is concentrated in basal puncta (F and F″) while combined anti-HhC staining reveals strong apical labeling (E′ and F′) from HhC 16 . Puncta labeled by HhNp (E″, anti-HA staining) overlap all HhC (HhC16 and HhC24) stained puncta (see boxed areas in E and F). (G,H) High magnification images of the boxed areas in E (in G,G′) and F (in H,H′). HhNp colabels HhC-positive puncta at the apical tips of photoreceptors (top, G) and axons extending towards the optic stalk (bottom, G′). HhC 24 (H,H′) is absent from the apical tips (H), unlike combined HhC staining. HhC 24 is concentrated in axons extending towards the optic stalk (H′). Scale bars: 5 µm. A diagram (Fig. S1) and table (Table S1) of all Hh constructs can be found in the Supplementary materials. All micrographs are representative of three biological replicates. Fig. 2. Differential formation and export of HhC isoforms. (A) Cells transfected with a heat-shock inducible hh NHA construct, hsp70-hh NHA , were subject to a brief heat pulse, after which cell lysates were prepared at the indicated times. 'n/a' indicates no heat pulse and 'Time 0' lysate was taken at the conclusion of the heat pulse. Western blots were probed with the indicated antibodies. Bands are marked as follows: C24, HhC 24 ; C16, HhC 16 . A loading control was visualized with α-Tubulin (anti-Tub) antibody. (B) The translational inhibitor cycloheximide (CHX) was added to cells stably expressing tub a1 -GAL4>UAS-hh NHA at time 0 [in hours, 'Hr(s)']. Cell lysates were prepared at the indicated time points after CHX addition. Western blot analysis and notation for Hh polypeptides are as in A. The levels of HhNp and HhC 24 steadily decline while HhC 16 remains relatively unchanged. (C) Cells expressing tub a1 -GAL4>UAS-hh NHA were washed with fresh media and incubated for 3 days. Cell pellets and equivalent amounts of total protein from media were analyzed for HhNp (left panels) and HhC isoforms (right). Note that HhC 16 is absent from the media fraction. (D) Cells were stably transfected with tub a1 -GAL4>UAS-hh NHA (top panel) or tub a1 -GAL4>UAS-hh CHA2 (bottom panel) and treated with CHX media, as in B. Media was collected after CHX addition at the indicated times, after which the HA-tagged HhNp or HhC was concentrated by immunoprecipitation and visualized by western blot. Quantification by densitometry is shown as the ratio of Hh species in the media relative to media after 3 days incubation with transfected cells (without CHX addition, leftmost lane). HhNp and HhC 24 appear in the media with similar kinetics to their depletion in CHX-treated cells (B). (E) Brefeldin A (BFA; 20 µM) was added to the culture media of hsp70-hh NHA transfected cells, after which the cells were treated to a heat pulse. After 3 h, the cells were lysed to prepare extracts for western blot analysis. α-HA staining was used to visualize HhNp. α-Tubulin (anti-Tub) level was measured as a loading control. BiP was examined (anti-BiP) to measure induction of the Unfolded Protein Response by either the heat pulse or BFA treatment. (F) Export of HhNp and HhC 24 into media was assessed as in D, using immunoprecipitation to concentrate Hh polypeptides from culture media. For BFA addition, BFA was added to tub a1 -GAL4>UAS-hh NHA expressing cells for 1 h. Cells were then washed and fresh media with BFA was added. Media was collected after 8 h and examined by immunoprecipitation for HhNp (anti-HA) and HhC (anti-HhC) by western blot. Densitometry is displayed for band intensity relative to the '(+) control' band, for which media was collected after 3 days exposure to tub a1 -GAL4>UAS-hh NHA expressing cells. HhNp and HhC 24 in the media were reduced in the presence of BFA. All blots are representative of three biological replicates.
Self-cleavage may thus precede and indeed be required for HhC 16 formation.
The signaling domain, HhNp, is secreted and, when expressed in cultured cells, accumulates in the media (Fig. 2C) (Lee et al., 1994;Maity et al., 2005). Consistent with export, pulse induction of hh + expression ( Fig. 2A) resulted in transient accumulation of intracellular HhNp that peaked at 45 min post-induction. The intracellular level of HhC 24 displayed similar kinetics ( Fig. 2A). The conversion of HhC 24 to HhC 16 could account for the reduction in HhC 24 level at later time points. Surprisingly however, HhC 24 , like HhNp, accumulated in the media of cells transiently expressing tub α1 -GAL4>UAS-hh + (Fig. 2C); in contrast, HhC 16 was not detected in the media. To further resolve the kinetics of Hh processing and secretion, translation in tub α1 -GAL4>UAS-hh + transfected cells was blocked with cycloheximide addition to the media (Fig. 2B). By the first time point after cycloheximide addition (1.25 h), HhU was nearly undetectable (>10-fold reduction). The intracellular levels of HhC 24 and HhNp declined more slowly to ∼50% by 2.5 h after cycloheximide addition. In contrast, the intracellular HhC 16 level was constant for at least 6 h. HhC 24 and HhNp coincidently appeared in the media (Fig. 2C), where their concentrations increased at rates inversely corresponding to their diminishing intracellular levels (Fig. 2D). The control nuclear protein Elav was found only in the cell lysate, indicating that HhC 24 release was not a consequence of cell rupture or death (data not shown). Thus, HhC 24 and HhNp were released from cells in a temporally coincident and quantitatively similar manner, while HhC 16 was stably contained within the cells.
To place self-cleavage and HhC cleavage into a subcellular context, we first examined the formation of the proteolytic products in the presence of the toxin Brefeldin A (BFA), which disrupts COPI-mediated ER to Golgi transport and Golgi to ER recycling (Lippincott-Schwartz et al., 1989). Western blot analysis revealed that the level of HhC 16 was unchanged when BFA was added prior to the induction of hh expression from a heat-shock cassette (Fig. 2E). However, the intracellular levels of HhC 24 and HhNp both increased (HhC 24 , 4.1-fold; HhNp, 1.8-fold). Notably, neither heat-shock nor the addition of BFA increased levels of the ER chaperone BiP, a standard marker for induction of the unfolded protein response ( Fig. 2E and data not shown) (Ryoo et al., 2007; and reviewed by Walter and Ron, 2011). In the absence of BFA, transient hh + expression resulted in contemporaneous accumulation of HhNp and HhC 24 in the media, while HhC 16 and HhU remained in the cells (Fig. 2C,F). In the presence of BFA, the export of HhC 24 and HhNp was greatly diminished (Fig. 2F). Hence, neither selfcleavage nor C-terminal cleavage required COPI-mediated transport to the Golgi. However, cellular export of both HhNp and HhC 24 were COPI-dependent.
To further clarify which Hh products enter the Golgi apparatus, we engineered a Hh isoform with insertion of an N-linked glycosylation site. The isoform was examined for proteolytic processing and Golgi-specific modification that rendered attached carbohydrate moieties resistant to trimming by endoglycosidase H. Cell lysates obtained from Drosophila cell culture expressing a hh gene bearing such a site created by a Lys 340 to Asn substitution were treated with endoglycosidase H (EndoH; Fig. S3B). HhU was entirely EndoH-sensitive (Fig. S3B) which is consistent with its self-cleavage being independent of COPI-mediated ER to Golgi transport (Fig. 2E). In contrast, approximately 75% of HhC 16 was EndoH-sensitive ( Fig. S3B), consistent with its production in the ER. Surprisingly, HhC 24 was entirely EndoH-sensitive ( Fig. S3B) even though Hh is believed to traverse the Golgi. While the absolute EndoH sensitivity of HhC 24 has been reported previously (Bumcrot et al., 1995), it is possible that the protein's secondary structure might prevent Golgi-specific modification and thus, would lead to these results.
In summary, these experiments indicate that HhC 16 and HhC 24 move apicially and basally, respectively, in larval photoreceptors after their formation in the ER.
Proteolytic cleavage at the Hh C-terminus Hh's 9 kDa C-terminal 'tail' is thought to be structurally disordered and sensitive to proteolytic attack (Hall et al., 1997). If HhC 24 were shortened to HhC 16 by the removal of its tail, it would lack the axonal targeting motif (G*HWY) (see Fig. 3D) (Chu et al., 2006). The loss of this motif would account for the absence of HhC 16 from photoreceptor axons and the brain (Fig. 1C), consistent with the lack of Hh axon transport in transgenic and genomic mutants deleting a similar region of the Hh C-terminus (Chu et al., 2006). To determine if HhC 16 is a C-terminally shortened form of HhC 24 , we mapped the cleavage site with maleimide-PEG (mal-PEG) targeted addition (Vitu et al., 2010) and performed size comparison to engineered Hh C-terminal truncations. These approaches defined the span between amino acids 410 and 413 as the site where cleavage yields HhC 16 . The HhC 16 product would thus lack the axonal targeting motif.
In the mal-PEG method, a 1.0 kDa mal-PEG moiety is added to extracted polypeptide at Cysteine (Cys) residues, and then sizeresolved by western blot analysis. The self-cleavage product HhC 24 has two native Cys residues (Cys 258 and Cys 400 ); if these two residues were present in HhC 16 , mal-PEG addition at either or both would increase the molecular weight of HhC 16 by 1.0 or 2.0 kDa, respectively. Mal-PEG additions were considerably stronger following the addition of TCEP, which reduces the disulfide bridge between Cys 258 and Cys 400 , and indicated that most native HhC 16 contains this bond (Fig. 3B). These two novel bands were indeed observed with the expression of a wild-type hh + transgene in both cell culture and the adult eye ( Fig. 3B; data not shown), indicating that both Cys 258 and Cys 400 are contained within HhC 16 . The product with two mal-PEG moieties was however underrepresented, likely due to inefficient addition in the basic environment created by adjacent Tyr 401 and Cys 400 .
For further detailed mapping, a series of constructs was created with single Cys substitutions for amino acids to either side of Cys 400 . These were expressed in a Drosophila cell culture system in which HhC 16 was efficiently produced from a full length Hh transgene (Fig. 3A). Cys substitutions at Ala 398 , Asn 405 , Ser 408 , and Ala 410 resulted in HhC 16 species that were modified at the novel Cys residue (Fig. 3A). Moreover, Ala 410 resulted in HhC 16 species that were modified at the novel Cys residue when expressed in the retina of transgenic animals (Fig. 3B). In contrast, Cys substitution at Gly 413 or Ser 421 did not introduce a novel mal-PEG modifiable site into HhC 16 (Fig. 3A). To confirm localization of the cleavage between Ala 410 and Gly 413 , we engineered a series of HhC truncations by inserting a start codon at the self-cleavage site (Cys 258 ) and stop codon at various carboxyl-terminal sites expected to produce a polypeptide of 16-17 kDa. The truncated polypeptide produced by a stop codon at Leu 414 had slightly slower gel mobility than HhC 16 , while other nearby truncations created products with larger differences in mobility (Fig. S4A). These results indicate that HhC 16 is generated by cleavage between residues Ala 410 and Gly 413 . Notably, Cys substitution at Ala 410 strongly reduced HhC 16 formation in cultured Drosophila cells (∼53% reduction; Fig. 3A).
When the same mutant Hh protein was expressed in the retina of transgenic animals, self-cleavage to yield HhNp occurred normally (Fig. 3C), but the C-terminal fragment accumulated as HhC 24 , while HhC 16 was barely detectable (90% reduction; hh NHA-A410C ). This also indicates that self-cleavage does not require cleavage at this second cleavage site in order to produce mature HhNp. Alignment of Hh from diverse species revealed that the amino acid sequence surrounding Ala 410 is well conserved (Fig. 3D). Interestingly, the HhC cleavage site is adjacent to a hydrophobic amphipathic helix ( Fig. 3D; Fig. S4B and C), which suggests possible association of this domain with hydrophobic membranes and substrates and a potential remodeling/refolding of this hydrophobic stretch when C-terminal cleavage occurs (Hall et al., 1997). Further, such regions are common among proteins associated with apolipoprotein particles, the reported vehicle of Hh transport (Smolenaars et al., 2007;Panáková et al., 2005).

C-terminal cleavage controls Hh spatial localization and targeted signaling activity
If proteolytic loss of the axonal targeting motif is a determinant of Hh localization in photoreceptor neurons, we would expect a C-terminal cleavage site mutation to shift Hh localization from the developing eye to the brain. Moreover, with visual system development under the control of such a mutant, the induction of lamina cells might increase at the expense of photoreceptor cells. To test these predictions, we quantified immunofluorescence from developing photoreceptor cells, comparing the wild-type localization of HhNp and HhC to two Hh mutants that lack Cterminal cleavage. In the wild type (see also Fig. 1B), HhC isoforms and HhNp are present in both the retina and photoreceptor axons (Fig. 4A,C; Fig. S5A). In contrast, HhNp and HhC derived from Hh A410C were shifted to axon termini (Fig. 4A,C; Fig. S5A). Both HhNp and HhC co-labeled puncta were absent from the apical  ), on lysates from adult Drosophila heads expressing the proteins Hh NHA or Hh A410C under control of the eye-specific driver GMR-GAL4. Addition of TCEP reduces the disulfide bridge between Cys 258 and Cys 400 . The considerably stronger MAL-PEG addition after TCEP treatment indicates that most native HhC 16 bears this disulfide. Substitution of Cys at Ala 410 creates a novel site for MAL-PEG addition, increasing the intensity of the MAL-PEG addition bands and decreasing the level of native HhC 16 . (C) Western blot analysis of Hh NHA and Hh A410C mutant in adult transgenic animals, with eye-specific expression driven by GMR-GAL4. Selfcleavage is evidently normal in the mutant (hh A410C ), as indicated by the relatively normal levels of HhU and HhNp (α-HA, middle). The level of HhC 16 is strongly reduced, while the alternative isoform, HhC 24 , is increased. (D) Hedgehog family members from several species were aligned using ClustalW. The residue Ala 410 in Drosophila melanogaster (Dmel) was conserved in all cases ('*' at top). Differing degrees of conservation to either side were classified as fully conserved (yellow, '*'), a 'strong' association group (blue, ':'), or a 'weak' association group (green, '.'). In cartoon (bottom) labeled domains were identified either previously (e.g. 'Intein-like Domain' and 'Axonal Targeting Motif') or by hydrophobicity/amphipathic helix prediction (see Fig. S4B and C). The axonal targeting motif was defined by mutation at Tyr 452 (see Chu et al., 2006). All blots are representative of three biological replicates. membranes of photoreceptor cell bodies (Fig. 4B, bottom left panels), where they are normally found in the wild type ( Fig. 4B and C, right panel, ∼threefold decrease) in clusters surrounding the apical cell marker Bazooka (Djiane et al., 2005). Conversely, the number of HhNp and HhC co-labeled puncta in photoreceptor axons was markedly increased in the Hh A410C mutant (Fig. 4B, bottom right panels). An HA-tag insertion at Ala 358 (hh CHA2 ) also resulted in a C-terminal cleavage mutant phenotype; it displayed normal self-cleavage without forming HhC 16 (Fig. S5B). As with hh A410C , the distribution of HhNp and HhC derived from Hh CHA2 was shifted to axon termini (data not shown).
We have shown that eye and lamina development are controlled by the release of Hh from opposite ends of the photoreceptor neuron (Chu et al., 2006). A mutation in the C-terminal axonal targeting motif resulted in HhNp retention in the retina and a deficit in lamina development (Chu et al., 2006). The genomic mutation, hh 2 , deleted the axonal targeting motif and displayed a similar lamina phenotype. We reasoned that, if more HhNp is released by each photoreceptor axon that arrives in the brain, HhC cleavage mutants might favor lamina development at the expense of eye development. To address this question, the numbers of lamina precursor cells and ommatidia were quantified when HhCcleavage mutant transgenes were used to rescue visual system development in a visual system-specific hh 1 genetic background. The mutant transgenes UAS-hh A410C and UAS-hh CHA2 were expressed specifically and at a low level in the developing eye with the eyeless 116 -GAL4 (ey-Gal4) driver or the strong retinaspecific driver GMR-GAL4 in the presence of the regulatory subunit encoded by tub α1 -GAL80 ts (McGuire et al., 2003) to suppress GAL4 activity. Thus, transgenic hh was supplied in limiting amounts [compare GMR>UAS-hh + or ey>hh CHA3 to the wild type (WT) in Fig. 4D]. We examined the effect on eye and lamina development in late third instar larval animals, before apoptosis eliminated lamina precursor cells that failed to interact with an ommatidial axon fascicle (Huang et al., 1998). In this context, we observed that shifting the polarity of Hh secretion altered the ratio of lamina to retinal development (Fig. 4D,E).
Hh induces the formation of lamina precursor cells, which express the marker Dachshund (Huang and Kunes, 1996;Mardon et al., 1994). We quantified photoreceptor neurons via their expression of the neuronal markers Elav and HRP. The number of lamina precursor cells was quantified in complete Z-stack reconstructions of the brains of late third instar larvae (Fig. 4D,E; data not shown). For each specimen, the corresponding retina was examined to quantify the anterior progression of eye development (Fig. 4D, left panels). Notably, with reduced eye and lamina development in GMR-GAL4, tub α1 -GAL80 ts animals, the ratio of lamina precursor cells to ommatidial columns was increased in Hh A410C , relative to wild type Hh (Fig. 4E). Similarly, when either Hh A410C or Hh CHA2 was expressed with the weak driver ey-GAL4, there were more lamina precursor cells and fewer ommatidia than in the hh + control (Fig. 4D,E). Thus, converse to deletion or mutation that removes the axonal targeting motif, the loss of HhC cleavage favors lamina development at the expense of retinal development.

DISCUSSION
The activity of a morphogen depends on the mechanisms of secretion and dispersal that shape its access to cells of a receptive field. This is the case for Hh, whose secretion and transmission is complex and remains unresolved. One view of Hh transmission posits its diffusion in extracellular space as monomeric protein, multimeric complex or in lipoprotein particles. Another view rests on long cellular extensions, filopodia or cytonemes, over which Hh may be carried for many cell diameters. These modes of transmission are not mutually exclusive and indeed may coexist and cooperate to create the spatial shape of the Hh signaling gradient. Resolving the secretory pathways that emit Hh from its cells of origin is key to understanding these modes of transport.
There is ample evidence that one of the determinants of Hh dispersal is polarized secretion (reviewed by Kornberg, 2011;Therond, 2012). A number of models have based the differential range of Hh on selective export from either the apical or basal poles of the cell. For example, work in the developing Drosophila wing indicates that apically secreted Hh is reabsorbed and redirected to basal cytonemes, which then transmit Hh in a long-range signaling gradient (Callejo et al., 2011). We have defined a system in which polarized secretion accounts for coordinated developmental programs in the Drosophila eye and brain. Apical Hh secretion propagates the temporal wave of ommatidial development in the eye, while basal targeting to photoreceptor axons induces the differentiation of post-synaptic lamina neurons in the brain (see Fig. 1A) (Huang and Kunes, 1996; reviewed by Roignant and Treisman, 2009). We previously defined a small region of the Hh C-terminus that is necessary and sufficient for basal secretion (Chu et al., 2006). Hh lacking this axonal targeting motif is mostly secreted apically, possibly due to an apical targeting signal(s) near Fig. 4. C-terminal cleavage controls the polarity of Hh localization and the balance of eye and brain development. (A) Horizontal perspective of the developing eye, brain complex at late third instar stage, comparing wild-type hh + (left) protein localization to the HhC cleavage mutant, hh A410C (right). Transgenes were expressed with the pan-neural driver elav-GAL4. HhNp localization (α-HA staining) is reduced in the apical retina (ed) and enhanced in the optic stalk (os) and at photoreceptor R1-6 growth cones (gc) in the hh A410C mutant. Scale bar: 20 µm. (B) Higher magnification view (than in A) comparing apical, basal localization of HhNp and HhC in the wild-type (hh + ) and hh A410C mutant. The coalesced apical tips of photoreceptor cells in an ommatidium were marked with Bazooka::GFP (Baz, blue color). Note that both HhNp (anti-HA staining, red color) and HhC (green color) are strongly reduced in the apical region of animals expressing Hh A410C (right middle panel). Green and magenta bars indicate the apical and basal regions, respectively, examined in higher magnification views in the bottom panels. The apical region of the Hh A410C mutant has much less HhNp and HhC staining ( panels demarcated by green bars), while the basal region composed of photoreceptor axons has greater HhNp and HhC staining, as co-labeled puncta in the Hh A410C mutant ( panels demarcated by magenta bars). Scale bar: 20 µm. (C) Quantitative analysis of Hh distribution in the wild type and C-terminal cleavage mutant. (Left) Quantification of HhNp in the retina (ed), optic stalk (os), and growth cone (gc; as described in Chu et al., 2006) based on average pixel intensity measurements (see Materials and Methods). The ratios of pixel intensity measurements were calculated, as indicated. Error bars indicate s.e.m. *P<0.05, ***P<0.001 by two-tailed t-test. (Right) The average number of HhNppositive apical puncta was quantified per unit area after expression of the wild type (hh + ) and mutant (hh A410C ) transgenes. Plots are representative from three biological replicates. Data collected from: hh NHA n=13 and hh A410C n=12 specimens. Error bars indicate s.e.m. ***P<0.001 by two-tailed t-test.
(D) Rescue of eye and lamina development by eye-specific transgene expression in the hh 1 genetic background. Approximately 11 ommatidial columns are formed in the visual system-specific regulatory mutant hh 1 . Lamina induction, measured by the formation of Dachshund (Dac)-positive lamina precursor cells (α-Dac, green color) and Elav-positive lamina neurons (α-Elav, red color) is completely absent in the mutant (hh A410C ) (not shown; Huang and Kunes, 1996). Representative late third instar specimens are shown, with corresponding eye and brain micrographs (lateral perspective). Ommatidial columns (left panels) were revealed by α-HRP staining (grayscale). Regions of ommatidial development are marked by vertical yellow bars at the bottom of each image (left panels). With the strong driver GMR-GAL4, GAL4 activity was attenuated with the temperature-sensitive tub α1 -GAL80 ts inhibitor (McGuire et al., 2003) employed at a semi-permissive temperature (25°C, as shown). Under these conditions, eye and lamina development with the wild-type transgene (GMR>hh + ) is reduced from hh + background (top panels). Rescue with the hh A410C transgene (GMR>hh A410C ) yields fewer ommatidial columns and more lamina precursor cells and lamina neurons. With the weak eye-specific driver eyeless-GAL4, a transgene with normal HhC cleavage yields rescue with normal ommatidial development and reduced lamina development (ey>hh CHA3 ). With a mutant transgene that lacks HhC cleavage, lamina development is rescued, while eye development is reduced (ey>hh CHA2 ). Scale bar: 20 µm. (E) Quantitative analysis of ommatidial development and lamina induction in specimens from experiments shown in D. The average ratio of lamina neurons (Lam. N's) to ommatidial columns (Omm Cols) was determined in 3D reconstructions of confocal micrographs (see Materials and Methods). Significance scores above bars are shown relative to each hh construct with 'normal processing'. *P<0.05, ***P<0.001 by two-tailed t-test. Plot is representative from three biological replicates. Data collected from: ey>hh + n=8; ey>hh A410C n=12; GMR>hh CHA3 n=7; GMR>hh CHA2 n=9; GMR>hh + n=7; and GMR>hh A410C n=7 specimens. All micrographs are representative of three biological replicates.
the N-terminus (T.C. and S.K., unpublished observations). Here we show that the distribution of Hh between the eye and the brain is controlled by proteolytic cleavage at a site in the Hh C-terminal domain.
The proteolytically shortened HhC 16 , which lacks the axonal targeting motif (Figs 1G,H and 4B) was preferentially localized at the apical tips of photoreceptor neurons in puncta containing HhNp, the developmental signaling domain (Fig. 1C,F). This is consistent with the prior observation that Hh remains in the retina in C-terminal deletion and point mutants that lack the axonal targeting motif (Chu et al., 2006). In contrast, HhC 24 , was found in basally localized particles with HhNp localized in photoreceptor axons and growth cones (Fig. 1B,C,E and G′). HhC 24 may be released from growth cones, as it is from cultured cells (Fig. 2), though it has no known signaling activity (Roelink et al., 1994(Roelink et al., , 1995. The shortened isoform HhC 16 appears to be retained in the cell, at least in culture (Fig. 2), despite entering the Golgi (Fig. S3B). This binary decision evidently controls the distribution of Hh between the developing eye and brain, as the distribution of HhNp was shifted to the brain when C-terminal cleavage was blocked by mutation (Figs 3 and 4). Under conditions of limited Hh synthesis, the shift in the polarity of secretion was matched by enhanced induction of lamina precursor cells in the brain and reduced ommatidial development (Fig. 4). It is possible, then, that in normal development the control of C-terminal cleavage balances Hh's activities between the retina and brain. In this regard, we have identified a regulator of Hh C-terminal cleavage that controls Drosophila eye development in a hh-dependent manner (J.R.D. and S.K., unpublished data).
Of note, the HhC 24 cleavage site is adjacent to a hydrophobic amphipathic helix ( Fig. 3D; Fig. S4B and C); such regions are common among proteins associated with apolipoprotein particles (Smolenaars et al., 2007). The ER is a likely source of HhC containing particles since it appears to be the locale where Hh cleavage products are formed (Fig. 2). While it has been shown previously that COPI is not necessary for Hh self-cleavage, far less is known about COPI and COPII dependence on Hh secretion (Fig. 2E,F) (Aikin et al., 2012;Chen et al., 2011). In a genome-wide screen for Hh secretion, COPI inhibition appeared to block most Hh export while COPII knock-down had only a modest effect (Aikin et al., 2012). Thus, HhNp and HhC 24 may be captured in the same particle in this ER-localized process, which leads to their basal targeting and axon transport (Fig. 5). This association between Hh termini is likely mediated in part by the lipophilic moieties on HhNp and the amphipathic tail on HhC24. Cholesterol modification of the mature HhNp ligand, for instance, enables its interaction with lipid raft proteins such as Caveolin (Reggie1) and the putative proton transporter Dispatched (Aikin et al., 2012;Burke et al., 1999;Callejo et al., 2011;Katanaev et al., 2008). In contrast, HhC 16 , possibly associated with HhNp, lacks the axonal targeting motif, and may have its amphipathic helix disrupted by proximity to its novel C-terminus, which results in its apical targeting (Fig. 5). This model is plausible as the stabilization and 'solubilization' of the hydrophobic Dmel HhC 24 upon C-terminal proteolysis (of the last ∼9 kD) has been reported previously (Hall et al., 1997). Such a change in the structure of this C-terminal tail could allow HhNp to instead associate with other binding partners and thus, would influence its apical/basal targeting.
Finally, the HhC cleavage site between residues Ala 410 and Gly 413 , is conserved in diverse Hedgehog family members (Fig. 3D). While it is not clear whether C-terminal cleavage is common, it has been reported for human SHH (Mastronardi et al., 2003). Furthermore, a mutation at this site yields a moderate form of holoprosencephaly (Roessler et al., 2009). It is generally not understood how HhC region mutations yield Hh loss-of-function phenotypes; clearly, defects in secretory targeting are one possibility. Thus we have defined a novel mechanism for the apical/basal targeting of a developmentally important ligand and due to its conservation in humans, it is possible that this same process might underlie the targeting of other posttranslationally modified ligands.

Strains and reagents
The UAS-hh NHA (Burke et al., 1999), UAS-hh CHA2 and UAS-hh CHA3 (Chu et al., 2006) transgenic animals were described previously. The following stocks were obtained from the Bloomington Drosophila Stock Center (Bloomington, IN, USA): y,w, GMR-GAL4/CyO; y,w, eyeless-GAL4 Fig. 5. Hh cleavage products depend on a C-terminal cleavage to determine Hh axonal transport. Sorting of HhC 24 and HhC 16 presumably occurs in the ER where both autocleavage and C-terminal cleavage take place and Hh particles are assembled. 'HhNp' is labeled in blue. 'HhC24' is labeled in orange and contains the 'growth cone targeting sequence' (green circle). HhC16 is also orange but does not possess the targeting sequence. Particles containing HhNp and HhC 24 bud from the ER and travel down the axon. Particles containing HhNp and HhC16 (which lacks the 'growth cone targeting sequence'), however, will remain in the photoreceptor cell body and HhNp will be released apically.
The GMR-GAL4 and eyeless-GAL4 drivers were introduced into a hh 1 background to either create recombinant chromosomes (e.g. GMR-GAL4>UAS-hh NHA ) or to perform genetic eye rescues. Transgenic Drosophila (UAS-hh NHA-A410C ) was made by Best Gene (Chino Hills, CA, USA).

Molecular biology
To construct HhC truncations, primers were made to flank the C258 codon (first codon of HhC) and the last codon of each truncation (e.g. S408, G413). The forward primer substituted an 'ATG' start codon for C258 and the reverse primer placed a 'TAG' stop codon after the truncation site. Truncations were cloned into pUAST with EcoRI and BglII.

Cell culture experiments
ML-DmBG3-c2 cells (DGRC) were cultured in a humidified, 23°C incubator, as previously described (Ui et al., 1994). Qiagen Effectene (Qiagen, Hilden, DE) was used for transfections; 1 mg of DNA per plasmid. Time course: cells transfected 60 h previously with pCasSper-hs-hh NHA were heat-shocked for 20 min at 37°C (control plate at 23°C) and then placed at 23°C. Cells were lysed at specified times following heat-shock and prepared for western blotting. ER to Golgi block: cells transfected 60 h previously with either pCasSper-hs (empty vector) or pCasSper-hs-hh NHA had Brefeldin A (BFA, Sigma Aldrich, St. Louis, MO, USA) added 1 h before heat-shock (final, 20 µM). One 'no BFA' control plate stayed at 23°C. Two 'heat-shock' plates, with or without BFA, spent 20 min at 37°C. Plates were then placed for 3 h at 23°C. Then, cells were scraped off plates and prepared for western blots. Immunoprecipitation (IP): cells were transfected 60 h previously with tub α1 -GAL4 and UAS-hh NHA or UAShh CHA2 had fresh media added. Cycloheximide (CHX, Sigma Aldrich) was added to plates (final, 50 mg ml −1 ). Media aliquots were collected at specified time points. A modified Dynabeads Protein A (Invitrogen, Carlsbad, CA, USA) IP protocol was used to pull HhN::HA or HhC::HA from media. Briefly, 0.9 ml of cell culture media (spun 1000×g, 5 min) was mixed with an equal volume of chilled, Non-Denaturing Lysis Buffer [NDLB, pH 8; 20 mM Tris HCl pH 8, 137 mM NaCl, 1% NP-40, 2 mM EDTA ( pH 8), Protease Inhibitor]. Rabbit α-HA (Y-11 sc-805, Santa Cruz) was then added (final, 0.6 ng ml −1 ) and tubes were nutated overnight at 4°C. In the morning, 75 µl of Protein A beads, washed 3× with chilled NDLB, were added to the media/NDLB/antibody mixture and nutated for 1 h at 4°C. Beads were precipitated with a DYNA I MPC-S magnet, washed 2× with chilled NDLB, and then on the third wash 75 µl Laemmli Buffer (BioRad) (+βME+Protease Inhibitor Cocktail) was added and samples were prepared for western blotting. For the BFA immunoprecipitation experiment, control media was taken from resting transfected cells, after which one set of cells was incubated with BFA (final, 20 µM) for 1 h. After preincubation, cells were washed 2× with fresh media, BFA was reapplied, and cells placed for 8 h at 23°C. Media was then collected from BFA-treated and untreated cells. Following steps were identical to the IP above.

Microscopy and data analysis
Specimens were viewed with constant acquisition settings on a Zeiss LSM700 Inverted confocal microscope. All methodology and statistics (including choice of sample size, exclusion criteria, double blind test, randomization, and choice of statistical test) were performed as previously described and according to standard procedures for this type of Drosophila data as described previously (Chu et al., 2006); e.g. quantification of growth cone, optic stalk and eye disc fluorescence. To count larval lamina neurons, stacks were normalized to the same threshold on a dark background using ImageJ. A smooth function eliminated scattered pixels. The area containing the lamina neuropil was highlighted and the 'Analyze Particles' function was used to give a count. ClustalW (EMBL-EBI) was used to align HhC from different species. Densitometry was performed using ImageJ. Protein molecular weight was estimated using the 'Compute pI/Mw' program (ExPASy).