Cell fate determination during development often requires morphogen transport from producing to distant responding cells. Hedgehog (Hh) morphogens present a challenge to this concept, as all Hhs are synthesized as terminally lipidated molecules that form insoluble clusters at the surface of producing cells. While several proposed Hh transport modes tie directly into these unusual properties, the crucial step of Hh relay from producing cells to receptors on remote responding cells remains unresolved. Using wing development in Drosophila melanogaster as a model, we show that Hh relay and direct patterning of the 3–4 intervein region strictly depend on proteolytic removal of lipidated N-terminal membrane anchors. Site-directed modification of the N-terminal Hh processing site selectively eliminated the entire 3–4 intervein region, and additional targeted removal of N-palmitate restored its formation. Hence, palmitoylated membrane anchors restrict morphogen spread until site-specific processing switches membrane-bound Hh into bioactive forms with specific patterning functions.https://doi.org/10.7554/eLife.33033.001
Each cell in a developing embryo receives information that determines what type of body structure it will form. In fruit flies, this information is partly given by a protein called Hedgehog. In the embryo cells that receive it, Hedgehog can trigger a series of events which activate certain genes and thereby regulate structure formation.
The Hedgehog proteins are produced by a different organizing group of cells: from there they transport within the embryo, creating a gradient. Depending on where a responding cell is in the embryo, it receives a different amount of Hedgehog, which gives the cell its identity. For example, Hedgehog proteins form a gradient across a fruit fly’s developing wing, which creates a visible vein pattern. How Hedgehog proteins form gradients is enigmatic, however, because once produced, they cling to the cells that created them.
The reason for this unusual behavior is that the two ends of the Hedgehog protein are attached to a different fat molecule. In particular, one extremity is linked to a fat molecule called palmitate. These ends’ fatty additions anchor Hedgehog to the cells that produced them. Then, the tethered proteins gather together to form chain-like clusters where they inactivate each other: the extremity with the palmitate ‘hides’ the portion of the neighboring protein that binds to the receiving cells. It is still unclear how Hedgehog can be activated and released to reach these faraway cells.
One hypothesis is that an enzyme comes to the clusters and frees the proteins by cutting both of Hedgehog’s fatty anchors. Thanks to how the palmitate tethers Hedgehog to the cell, the protein is positioned in such a way that when the enzyme makes its snip, the binding site on the neighboring Hedgehog gets exposed: this protein is activated and, when also cut by the enzyme, released.
Here, Schürmann et al. create an array of mutant Hedgehog proteins – for example some without palmitate, some with palmitate that cannot be removed by the enzyme – and study how they affect the development of the wing’s pattern in the fruit fly. Coupled with the imaging of the clusters, these experiments support the hypothesis that the palmitate anchor is necessary so that Hedgehog proteins can be turned on before diffusing away.
The Hedgehog family of proteins is also present in humans, where it presides over the development of the embryo but is also involved in cancer. Understanding how Hedgehog works in the fruit fly could lead to new discoveries in humans too.https://doi.org/10.7554/eLife.33033.002
Hedgehog (Hh) morphogens are dually lipidated 19 kDa proteins that are firmly anchored to the cell membrane of producing cells. Production of all active Hhs begins with autocatalytic cleavage of a precursor molecule by its C-terminal cholesterol transferase domain (Porter et al., 1996b). This results in cholesteroylated vertebrate Sonic hedgehog (Shh) and Drosophila Hh. Next, Hh acyltransferase (Hhat, also designated Skinny hedgehog or Raspberry) attaches a palmitoyl group to a conserved N-terminal cysteine that becomes exposed after signal peptide cleavage (Chamoun et al., 2001; Lee and Treisman, 2001; Micchelli et al., 2002). Hh palmitoylation is critical for later signaling, demonstrated by mutation of the N-terminal cysteine to serine or alanine (C25 > A/S in ShhC25A/S, C85 >A/S in Drosophila HhC85A/S) which abolishes palmitoylation and results in morphogen inactivity (Chamoun et al., 2001; Chen et al., 2004; Dawber et al., 2005; Goetz et al., 2006; Kohtz et al., 2001; Lee et al., 2001; Pepinsky et al., 1998). However, why N-palmitoylation is required for Hh signaling in vivo is still unclear.
Another unusual feature of all Hhs is their multimerization at the surface of producing cells which requires binding to the long, unbranched heparan sulfate (HS) chains of cell surface HS proteoglycans (HSPGs) called glypicans (Chang et al., 2011; Ortmann et al., 2015; Vyas et al., 2008). The Hh cholesterol modification is sufficient to drive this process (Feng et al., 2004; Gallet et al., 2006; Koleva et al., 2015; Ohlig et al., 2011). Despite membrane anchorage and cell-surface HS association, the multimeric Hhs initiate the Hh response in distant cells that express the Hh receptor Patched (Ptc). The question of how dual-lipidated Hh clusters manage to travel and signal to remote target cells is intensely investigated. The most current models propose lipidated Hh transport on filopodia called cytonemes (Bischoff et al., 2013; Sanders et al., 2013) or on secreted vesicles called exosomes (Gradilla et al., 2014) to bridge the distance between Hh-producing and receiving cells.
Hh release through cell-surface-associated proteases, called sheddases, has also been suggested. In vitro, membrane-proximal shedding not only releases Hh ectodomains from their lipidated N-terminal peptides (Dierker et al., 2009; Ohlig et al., 2011) but also activates Hh clusters. This is because N-terminal lipidated peptides block adjacent Hh-binding sites for the receptor Ptc and, thereby, render Hh at the cell membrane inactive. By cleaving these inhibitory peptides during release, sheddases unmask Ptc binding sites of solubilized clusters and thereby couple Hh solubilization with its bioactivation. In this model, the N-palmitate plays two indirect roles for Hh biofunction: first, it ensures reliable membrane-proximal positioning of inhibitory N-terminal peptides as a prerequisite for their efficient proteolytic processing, and second, by its continued association with the cell membrane, it ensures that only fully processed (=activated) Hh clusters are released. This model therefore predicts that inhibition of N-palmitoylation will result in release of inactive soluble proteins with masked Ptc-binding sites (Jakobs et al., 2014; Jakobs et al., 2016; Ohlig et al., 2011; Ohlig et al., 2012). It also predicts that impaired or delayed processing of dual-lipidated Hh will strongly reduce its release and bioactivity in vivo.
By uncovering a dominant negative, cell-autonomous function of non-palmitoylated HhC85S in endogenous Hh, we here support the first prediction. By using a series of transgenic Drosophila melanogaster lines that express untagged Hh, biologically inactive HhC85S, or N-truncated variants thereof in posterior and anterior wing disc compartments, we provide strong evidence that Hh clusters form by direct protein-protein contact and that unprocessed N-terminal peptides block Ptc binding of adjacent endogenous Hhs. As a consequence, we suggest that, due to their reduced activity, soluble clusters with masked Ptc-binding sites impair direct patterning of the 3–4 intervein region of the wing. Supporting this mechanism, targeted deletion on non-palmitoylated inhibitory peptides restores 3–4 intervein formation. We also show that impaired or delayed processing of lipidated Hh strongly reduces its solubilization, and hence its bioactivity, in vivo. We demonstrate that the HS-binding Cardin-Weintraub (CW) motif serves as the preferred N-terminal Hh processing site in vivo, and that impaired processing of this site completely abolishes direct 3–4 intervein wing patterning. Additional targeted deletion of N-palmitate restores wing patterning, demonstrating that one role of palmitoylated membrane anchors is to prevent the release of un- or incompletely processed Hh clusters in vivo. These genetic data are supported by the nano-structure of Hh clusters as revealed by immunoelectron microscopy (IEM) and provide new insights into how Hh relay from the producing cell membrane or between membranes could be achieved.
A first step in decoding possible Hh solubilization modes is to characterize the composition and organization of Hh substrates. It has been previously shown that Hh forms light microscopically visible clusters at the surface of producing cells (Chen et al., 2004; Gallet et al., 2006; Ortmann et al., 2015; Sanders et al., 2013; Vyas et al., 2008). However, the nanoscale structure of these heteroprotein complexes has not been determined. We therefore expressed Shh together with Hh acyltransferase in HEK293-derived Bosc23 cells to produce authentic cell surface Hh clusters for IEM analysis. To this end, we used several different α-Shh antibodies and secondary antibodies conjugated to 5 nm or 10 nm gold particles. Three α-Shh antibodies detected Shh in variably sized cell surface clusters, with the largest complexes exceeding sizes of 100 nm. Notably, as shown in Figure 1a–h and Figure 1—figure supplement 1, many clusters consisted of linear arrangements (Figure 1a–d) or contained linear arrays of closely packed gold particles (Figure 1f,g, arrowheads). Nearest-neighbour analysis of the angular distribution between the three most proximal gold particles (Figure 1i) confirmed that most arrangements were rectangular (90°) or linear (180°), the latter being consistent with Hh multimerization using linear HS chains of glypican HSPGs as templates (Chang et al., 2011; Vyas et al., 2008; Schuermann et al., 2018). Hh linearization during cell-surface multimerization is further consistent with previous structural and biochemical data which suggest a zig-zag arrangement of Hh monomers (Figure 1j) and variably sized Drosophila Hh and vertebrate Shh, ranging from 80 kDa to 600 kDa (Chang et al., 2011; Chen et al., 2004; Jakobs et al., 2014; Ohlig et al., 2011; Ohlig et al., 2012). We therefore next aimed to genetically confirm direct Hh clustering in vivo by using Drosophila melanogaster wing development as a model.
The fly wing develops from the imaginal wing disc (Figure 2a, bottom). The wing primordium at the center of the wing disc differentiates into the wing blade proper, which shows a characteristic pattern of five longitudinal veins (L1-5), an anterior cross vein (connecting L3 and L4) and a posterior cross vein (connecting L4 and L5) (Figure 2a, top) (Hartl and Scott, 2014). The anterior/posterior (a/p) boundary is located slightly anterior to the position of L4 in the adult wing (Figure 2a, red dashed line).
Hh is produced in the posterior wing disc compartment under the control of the transcription factor Engrailed (en) (Tabata et al., 1992; Zecca et al., 1995), which acts indirectly on Hh expression through the repression of the transcriptional Hh repressor Cubitus interruptus (Ci) (Bejarano and Milán, 2009). Hh then moves across the a/p boundary into the anterior compartment, where it binds to Ptc (Ingham et al., 1991). During its movement, Hh forms a gradient of decreasing concentration with increasing distance from the a/p border which corresponds to differential activation of different Hh target genes. Up to ten cell diameters from the a/p boundary, high Hh levels directly pattern the central L3-L4 region of the wing (Mullor et al., 1997; Strigini and Cohen, 1997) by stabilizing Ci155. More distal regions, up to 12–15 cell diameters from the a/p border, depend on Dpp, which is secreted in a stripe just anterior to the a/p boundary in response to low Hh levels. Hh thus plays a role in Drosophila wing patterning by controlling the spatially defined expression of target genes at the a/p border.
We exploited the Hh-regulated wing patterning response as a simple and reliable in vivo assay to test the functional consequences of proteolytic Hh processing. Specifically, we addressed the formation and positioning of longitudinal L3-L4 veins, and investigated whether Hh proteolytic processing in cells of the posterior compartment is a prerequisite for its signaling activity in cells of the anterior compartment (Crozatier et al., 2004). To this end, comparable amounts of Hh and Hh variants (Figure 2b, Supplementary file 1) were expressed from one specific attP 51C landing site on the second chromosome (Bateman et al., 2006) using the Gal4/UAS system (Brand and Perrimon, 1993). In the posterior compartment, Hh was expressed under en-Gal4 control, which is referred to as en >Hh, while in the anterior compartment, Hh was expressed in a stripe of cells under ptc-Gal4 control, referred to as ptc >Hh. As previously shown (Crozatier et al., 2004; Lee et al., 2001; Mullor et al., 1997; Strigini and Cohen, 1997), en >Hh expanded the L3-L4 intervein area and, as a concomitant effect, reduced the L2-L3 intervein space (Figure 2c). By contrast, en-regulated overexpression of non-palmitoylated, biologically inactive HhC85S (en >HhC85S) resulted in L3-L4 veins being proximally apposed and the formation of ectopic anterior cross veins (Figure 2d) (Crozatier et al., 2004; Lee et al., 2001), suggesting that HhC85S competes with bioactive wild-type Hh (Lee et al., 2001). This phenotype is consistent with the higher Hh concentrations required for activation of the target genes, ptc and collier, and L3-L4 development, than those required for the activation of dpp, which patterns the remainder of the wing (Hooper, 2003; Méthot and Basler, 1999; Mohler et al., 2000; Strigini and Cohen, 1997; Vervoort et al., 1999). Wing phenotypes were quantified by dividing the proximal L3-L4 areas by the L2-L3 areas (Figure 2e,f). This revealed significant Hh gain of function upon Hh overexpression in the posterior compartment or loss of function upon HhC85S overexpression (en >GFP served as a normal control: L3-L4/L2-L3 = 0.074 ± 0.002; en >Hh = 0.108 ± 0.005 (+46%), p<0.0001; en >HhCC85S=0.032 ± 0.002 (-57%), p<0.0001).
To investigate the molecular basis of the dominant-negative HhC85S activity in wing disc tissues, we spatially disconnected HhC85S expression from endogenous Hh expression by using ptc >HhC85S. In the event that biologically inactive HhC85S would impair the response to Hh in a non-cell autonomous manner, for example, by binding to and blocking the receptor Ptc, ptc >HhC85S wing phenotypes should be comparable, or even more severe than those observed in en >HhC85S wings. Alternatively, if unprocessed N-terminal HhC85S peptides directly inhibit Ptc binding of associated endogenous Hh produced in the same compartment, we expected ptc >HhC85S wing phenotypes to be less severe than those observed in en >HhC85S wings. Indeed, HhC85S expression under ptc control had little effect on wing development (Figure 2g,h: ptc >GFP: 0.084 ± 0.001, ptc >HhC85S: 0.078 ± 0.002 (-7%), p=0.0026), suggesting that HhC85S cell-autonomously interferes with endogenous Hh, possibly by the random mixing of inactive HhC85S and wild-type Hh at the cell surface (Figure 2i, left). In this mixed association, unprocessed HhC85S N-terminal peptides block wild-type Hh-receptor-binding sites in trans. By contrast, ptc>HhC85S expression in the anterior compartment prevents mixed cluster formation and therefore does not affect the controlled secretion and signaling of endogenous Hh (Figure 2i, right).
To independently confirm that HhC85S dominant-negative function requires direct Hh/HhC85S association with the same clusters, we expressed unlipidated monomeric HhNC85S (Porter et al., 1996a) in vitro and in vivo (Figure 3a). We observed that the expression of soluble HhNC85S under en-control did not affect endogenous Hh function in vivo (Figure 3b), as expected from its exclusion from lipidated Hh clusters at the cell surface. As shown in Figure 3c, relative L3-L4/L2-L3 ratios obtained from three independent HhNC85S lines (1-3) revealed that wing development was not significantly affected (en >GFP: 0.074 ± 0.002, en >HhNC85S (1): 0.07 ± 0.001 (-5%) (p=0.0707), en >HhNC85S (2): 0.074 ± 0.001 (±0%) (p=0.9419), en >HhNC85S (3): 0.075 ± 0.002 (+1%) (p=0.6050), 20 wings were quantified in each line). Notably, HhNC85S expression under ptc control resulted in a small, yet significant gain-of-function phenotype (Figure 3d,e, ptc >GFP: 0.084 ± 0.001, ptc >HhNC85S (1): 0.091 ± 0.002 (+8%) (p=0.0013), ptc >HhNC85S (2): 0.089 ± 0.002 (+6%) (p=0.0348), ptc >HhNC85S (3): 0.093 ± 0.001 (+11%) (p<0.0001), 20 wings were quantified in each line). This is consistent with the concept that Hh inactivation by adjacent unprocessed N-terminal peptides in trans is restricted to clustered, but not unclustered proteins (Ohlig et al., 2011; Ohlig et al., 2012). We conclude that the lack of Hh inhibition by monomeric HhNC85S (Figure 3f), even if expressed in the same cells, is consistent with required direct association of palmitoylated and non-palmitoylated morphogens for dominant-negative HhC85S function.
We next determined the molecular basis of the cell-autonomous inhibitory activities of HhC85S. In vitro, N-terminal peptides block Ptc-receptor-binding sites of adjacent Hh molecules in the cluster in trans (Figure 4—figure supplement 1a–c) (Ohlig et al., 2011). Thus, we predicted that N-terminal truncation of HhC85S should restore Hh biofunction in mixed clusters. To test this idea and to mimick Hh processing observed in L3 Drosophila larvae (Figure 4—figure supplement 1d,e), we consecutively deleted N-terminal amino acids 86–91 (HhC85S;Δ86-91) to 86–100 (HhC85S;Δ86-100) (Figure 4a) and confirmed unimpaired protein expression (Figure 4b) and multimerization (Figure 4—figure supplement 2). All ten constructs were then inserted into the attP-51C landing site on the second chromosome to ensure comparable expression. At least three independent transgenic fly lines were derived from each construct and crossed with the en-Gal4 driver line. We observed unchanged or moderately changed L3-L4/L2-L3 intervein ratios between en >HhC85S and en >HhC85S;Δ86-91 to en >HhC85S;Δ86-97 adult wings (Figure 4c–e,i and Figure 4—figure supplement 3a). However, protein truncation beyond residue R97 gradually restored the biological activity of mixed clusters: en >HhC85S;Δ86-98 and en >HhC85S;Δ86-99 fly wings showed partially restored wing patterning (Figure 4f,g,i) and, strikingly, the posterior expression of HhC85S;Δ86-100 fully restored normal wing patterning (Figure 4h,i and Figure 4—figure supplement 3a). Cell-autonomous inhibitory activities of HhC85S and restored wing patterning upon targeted coexpression of en >HhC85S;Δ86-100 were confirmed with the independent en-driver lines en(2)-Gal4 and hh-Gal4, both controlling transgene expression in the wing disc (Figure 4—figure supplements 4 and 5). These results are consistent with the assembly of Hh clusters by direct protein-protein contacts as a prerequisite for the inhibitory activity of unprocessed N-terminal peptides.
We also observed that wing phenotypes varied between and within fly lines (labeled 1–4 in Figure 4—figure supplement 3b). This variability can be explained by slightly different expression levels or by small stochastic changes in Hh/HhC85S cluster composition with increasing relative amounts of HhC85S, resulting in stronger dominant negative phenotypes. Indeed, temperature-dependent Gal4-regulated transgene amounts (Duffy, 2002) affected dominant-negative wing phenotypes: At 29°C, increased amounts of HhC85S relative to (fixed) endogenous Hh inhibited Hh function more strongly, whereas reduced transgene expression at 18°C inhibited Hh function less strongly (Figure 4—figure supplement 6).
Taken together, we conclude that N-palmitate serves to ensure reliable membrane-association of inhibitory N-termini, making quantitative peptide processing a prerequisite for the solubilization of fully activated clusters (Figure 5a). As a consequence, Hh concentrations at any position in the gradient will strictly correlate with their biological activities (i.e. their Ptc-binding capacities). Impaired N-palmitoylation in this scenario reduces Hh bioactivity to variable degrees, depending on the relative number of unprocessed N-terminal peptides in soluble clusters (Figure 5b). This essential ‘cleavage/activation control’ function is confirmed by fully restored Hh biofunction upon targeted coexpression of en >HhC85S;Δ86-100 (Figure 5c).
In our model, N-palmitate tethers incompletely processed Hh clusters to the cell membrane to prevent their release. To test this hypothesis, we utilized a cell culture model employing Bosc23 cells. To achieve quantitative Hh N-palmitoylation in vitro, we used bicistronic mRNA constructs to couple Shh (the vertebrate Hh ortholog) and Hh acyltransferase expression in the same cells. We then compared the release of fully lipidated Shh, non-palmitoylated ShhC25S, and variants carrying the extended C-terminal membrane anchor N190SVAAKSG-YPYDVPDYA-G198 (G198 represents the cholesterol-modified glycine; italicized underlined letters represent the tag, Figure 6) (Jakobs et al., 2014). Proteins were detected by polyclonal α-Shh antibodies and monoclonal α-HA antibodies on the same (stripped) blots. Grayscale blots were inverted, colored (green: α-Shh signal, blue: α-HA signal) and merged to identify proteins bound by both antibodies (yielding bright blue/cyan signals) and proteins bound by only α-Shh antibodies (green signals).
As shown in Figure 6a, dual-lipidated Shh and ShhHA yielded strong cellular signals but were absent from media, indicating impaired release. By contrast, non-palmitoylated ShhC25S;HA was effectively converted into a C-terminally truncated soluble morphogen, as indicated by an electrophoretic size shift and lack of α-HA antibody reactivity (compare the cellular (c) material in each lane 1 with corresponding media in each lane 3). Three independent quantifications of dual-lipidated Shh, cholesterylated ShhC25S, and non-lipidated ShhNC25S in cells and media (Figure 6b) confirmed that N-palmitoylation controls protein solubilization in vitro (1 hr release: ShhC25A 94 ± 3 arbitrary units (a.u.), ShhNC25A 265 ± 10 a.u., p<0.0001, n = 3; 4 hr release: Shh 18 ± 2 a.u., ShhC25A 238 ± 6 a.u., p>0.0001, n = 3, ShhNC25A 500 ± 58 a.u., p>0.001, n = 2; values express ratios between solubilized/cell-associated proteins). Accordingly, coexpression of dual-lipidated Shh and ShhC25A in the same Bosc23 cells resulted in mixed clusters and thereby a four-fold reduction in ShhC25A release (Shh+ShhC25A: 25.7 ± 5%, ShhC25A alone was set to 100%, p<0.0001, n = 7) (Figure 6c). Importantly, we further observed that dual-lipidated, N-terminally HA-tagged HAShh was not released (Figure 6d). In this construct, the HA tag was inserted at the position of the membrane-proximal CW motif, shifting this previously identified sheddase target site (Ohlig et al., 2012) distally while not affecting its HS-binding capacity.
To test whether the same modification would also impair release of fly Hh, we inserted an HA tag between corresponding Hh amino acids L91 and G92, resulting in the N-terminal HAHh sequence C85GPGRGL91-YPYDVPDYAG92-RHRARN (bold letters represent the CW motif that is shifted nine amino acids away from the preferred membrane proximal site of sheddase activity). We also used Hh, non-palmitoylated HhC85S and HhC85S;Δ86-100 as controls. HAHh was expressed in S2 cells, its unimpaired multimerization confirmed (Figure 6—figure supplement 1), and cellular and soluble proteins compared by SDS-PAGE and immunoblotting (Figure 6e). We observed that all proteins were produced in S2 cells, as indicated by strong α-Hh antibody binding to all cellular forms. In contrast, only low levels of HAHh that retained the tag were solubilized, suggesting that N-terminal processing was impaired in S2 cells. From these experiments, we conclude that N-palmitoylation controls Hh release from the cell surface and restricts possible modes of Hh solubilization to shedding (Figure 6f,g).
We next generated transgenic flies expressing HAHh (Figure 7a) in the posterior compartment. The HA-tagged protein, due to its invariable association with the membrane (Figure 6e) and direct association with endogenous Hh in mixed clusters, was expected to impair endogenous Hh release and to lead to severe dominant-negative mis-patterning phenotypes. Indeed, HAHh expression in the posterior compartment at 25°C largely arrested fly development at the pupal and pharate stages, leading to defective head development characteristic of Hh loss of function (Torroja et al., 2004). Of 230 pharates counted, only three imagos hatched with smaller wings lacking anterior structures (Figure 7b), again characteristic of Hh loss of function (Bejarano et al., 2012). Reduced transgene expression at 18°C largely reversed pharate lethality: 77% of en> HAHh pharates hatched (293 flies from 393 pupae) but wing development was still impaired with all analyzed wings lacking all or most of the L3-L4 intervein area (Figure 7c). This phenotype resembles wing phenotypes of flies expressing non-diffusible HhCD2 (Strigini and Cohen, 1997) or with impaired activity of Hh signaling components such as Fused or Collier (Col) (Ascano and Robbins, 2004; Vervoort et al., 1999), while the distal ‘widening’ of L3 (Figure 7c) is consistent with impaired Hh repression of Iroquois-regulated L3 formation (Crozatier et al., 2004). Notably, the observation that HAHh expression under ptc-control showed only minor effects (Figure 7h,i) confirms cell-autonomous Hh repression by direct HAHh contacts in mixed clusters, and suggests that palmitoylated HAHh peptides restrain these mixed clusters at the cell membrane. We therefore expected that additional C > S mutagenesis, by removing the membrane anchor (Figure 7a), would revert the observed severe mis-patterning phenotypes due to impaired cluster release into milder forms caused by partially impaired Hh binding to Ptc, as described earlier. Indeed, additional mutagenesis of the palmitate acceptor cysteine in en >HAHhC85S flies fully reversed pharate lethality at 25°C and led to wing phenotypes comparable to those of en >HhC85S flies (compare Figure 7d with 7 f and Figure 7e with 7 g).
The Hh gradient emanating from the posterior compartment activates the Hh target genes engrailed (en), collier (col), patched (ptc) and decapentaplegic (dpp) in stripes of anterior cells adjacent to the a/p border. Via Hh-responsive accumulation and nuclear access of Ci155, en and col are induced in a 5- to 7 cell wide anterior stripe, ptc in a 10 cell wide stripe and dpp in a 12–15 cell wide stripe, and the presence and width of these stripes of target gene expression is differentially sensitive to Hh dose (Chen and Struhl, 1996; Strigini and Cohen, 1997). Far from the Hh source, Ci155 is depleted to form the repressor CiR, and Hh target genes are repressed. Cells receiving minimal amounts of Hh activate dpp transcription, cells receiving an intermediate amount of Hh activate the expression of col and ptc in addition to that of dpp, and Hh-dependent anterior en transcription (but not posterior, Hh-independent en transcription) is located closest to the a/p border (Figure 8a,b). Col in the high and intermediate zones down-regulates Dpp responses: This results in the future L3-L4 intervein (Mohler et al., 2000; Vervoort et al., 1999). We used this system to investigate the impact of our mutant forms of Hh on the expression of en, ptc and dpp. We confirmed that posterior Hh overexpression expanded dpp-LacZ expression anteriorly (Figure 8c) and, consistent with established en >Hh expansion of the L3-L4 intervein area (Figure 2c), we confirmed that posterior Hh overexpression expanded ptc-LacZ expression in the presumptive L3-L4 region (Figure 8d). En-controlled expression of the HA-tagged protein at 18°C, in contrast, did not much affect dpp-LacZ expression in the anterior compartment (Figure 8e), but abolished all ptc-LacZ reporter expression and restricted en-expression posteriorly (Figure 8f). This confirms that the complete loss of L3-L4 intervein tissue in adult en >HAHh wings is caused by insufficient Hh levels at the a/p border, and supports the idea that coexpressed HAHh impaired Hh release from the posterior compartment of the wing disc. We note that the observed expansion of dpp expression can be best explained by abrogated Ptc-mediated Hh internalization that normally restricts the Hh gradient (Chen and Struhl, 1996). Consistent with our concept of N-palmitate serving as a membrane anchor to prevent unregulated Hh release, HAHhC85S coexpression restored ptc-LacZ and dpp-LacZ expression (Figure 8g,h) to levels comparable to those detected in HhC85S expressing wing discs (Figure 8i,j), and additional deletion of the unpalmitoylated N-terminal peptide reverted the expanded area of dpp-LacZ expression to wild-type range (Figure 8k,l). Together, these experiments confirm that N-terminal Hh processing converts the insoluble Hh cluster into truncated, bioactive morphogen, and that the palmitate anchor controls completion of this process.
To confirm that impaired processing of palmitoylated Hh variants prevents their solubilization, while impaired processing of unpalmitoylated Hh N-termini affects Ptc receptor binding of soluble clusters (merely reducing their bioactivity), we macroscopically analyzed wings of single and compound transgenic fly lines expressing Hh from the attP 51C landing site on chromosome 2 and HAHh or HAHhC85S from one specific attP2 landing site on chromosome 3. As shown earlier, if expressed under en-Gal4 control, Hh and HAHhC85S strongly affected wing development: En-controlled HAHhC85S reduced the formation of L3-L4 intervein tissue (Figure 9a), and en >Hh expanded the L3-L4 intervein area (Figure 9b). Targeted coexpression of both proteins fully reverted dominant-negative HAHhC85S function in 80% of wings and expanded this area in the remaining 20% of wings (56 wings were analyzed, Figure 9c,d). This indicates that increased Hh amounts ‘titer out’ dominant-negative HAHhC85S function. In contrast, en >Hh did not significantly correct dominant-negative HAHh mis-patterning phenotypes. As previously shown, HAHh expression in the posterior compartment arrested fly development at pupal and pharate stages. All wings of about 4% imagos that hatched (n = 12/320) lacked most anterior structures (Figure 9d,e). Hh coexpression partially reversed pharate lethality (22% imagos hatched, n = 41/182), but all analyzed wings still lacked the complete L3-L4 intervein area (Figure 9d,f). The most likely explanation for this is that increased relative Hh amounts ‘dilute’ the average number of permanent membrane anchors of any given mixed cluster, with only a limited compensatory effect on Hh release and activity due to the remaining tethers.
We confirmed cell-autonomous Hh repression by using another Gal4-line (34B-Gal4) that drives transgene expression in cells that form the most proximal parts of the wing where hh is normally not expressed (Figure 9g)(Brand and Perrimon, 1993). 34B-Gal4-controlled Hh expression in these cells results in phenotypes resembling a natural hh gain-of-function allele, hhMoonrat (Figure 9g´, arrow) (Tabata and Kornberg, 1994). Phenotypes resulting from ectopic hhMoonrat expression are usually mild, varying between overgrowth of anterior wing tissue to slight disorganization of the wing margin and the addition of extra vein material. We observed that 34B > HAHhC85S and 34B > HAHh expression did not affect wing development, confirming spatial disconnection of 34B-directed transgene expression from posterior endogenous Hh production and biological inactivity of both proteins (Figure 9h,i). In compound 34B > Hh;HAHhC85S wings, the activity of mixed clusters was reduced (Figure 9j, arrowhead), while it was completely abolished in 34B > Hh;HAHh wings (Figure 9k,l). This is expected from impaired Ptc-binding of Hh;HAHhC85S clusters in the former situation versus blocked release of Hh;HAHh clusters in the latter situation.
Finally, we investigated the expression of the Hh target gene ptc in flies expressing Hh and HAHhC85S alone and in combination. As shown earlier, posterior Hh overexpression expanded ptc-LacZ expression (compare Figure 10a with Figure 10b), and en-controlled expression of the HA-tagged non-palmitoylated protein strongly reduced ptc-LacZ reporter expression (Figure 10c). Consistent with the restored formation of L3-L4 intervein tissue in adult en >Hh;HAHhC85S wings (Figure 9c), and occasionally gain-of-function in these wings, ptc-LacZ expression in the anterior compartment of the wing disc was expanded (Figure 10d). This shows that coexpressed Hh fully restored dominant-negative HAHhC85S function by expanding ptc-LacZ target gene expression in the presumptive L3-L4 region in the anterior compartment and demonstrates that receiving cells respond to the morphogen.
Together, these experiments confirm a functional link between Hh lipidation, formation of linear cell surface clusters and proteolytic processing of lipidated N-terminal peptides in vivo. Processing serves to convert the lipidated morphogen cluster at the cell surface into the active form (Figure 11). Therefore, the N-palmitate membrane anchor and membrane-proximal CW-residues are functionally linked since the palmitoylation ensures quantitative CW-cleavage as a prerequisite for full Hh activation in vivo.
It is well established that cell-surface HS chains assist in Hh multimerization as a prerequisite for the generation of light microscopically visible aggregates at the cell surface (Ortmann et al., 2015; Vyas et al., 2008). Here, we provide ultrastructural data showing that a significant fraction of Hh assembles into extended linear arrays, consistent with the long unbranched HS-chain structure that scaffolds Hh clusters (Vyas et al., 2008), observed crystal lattice interactions of the vertebrate Shh ortholog (Pepinsky et al., 1998) and functional in vitro data (Ohlig et al., 2011; Ohlig et al., 2012). By exploiting the Drosophila wing development model which is dependent on differential Hh signaling for the formation of distinct wing structures, we further show that N-terminal peptides can block Ptc-receptor-binding of Hh clusters in vivo. Consistent with this, expression of N-truncated Hh mutants in Drosophila revealed that inhibitory peptide removal unmasks Ptc-binding sites and mediates direct, high threshold tissue patterning (Strigini and Cohen, 1997). Yet, contrary to previous observations on N-truncated Shh (Ohlig et al., 2011), we note that all artificially truncated HhC85S;Δ variants were functionally inert. We explain this inactivity by HhC85S;Δ misfolding due to possible intramolecular chaperone function of the 84 amino acid N-terminal Hh pre-peptide (Eder and Fersht, 1995) or unproductive Ptc binding of artificially truncated proteins as described for monomeric ShhN (Williams et al., 1999). We currently investigate these possibilities by insertion of a tobacco etch virus (TEV) protease recognition site into the putative Hh target site to allow for sequence-specific HhC85S cleavage and activation after controlled TEV protease expression in the fly (Harder et al., 2008).
We previously showed that proteolytic conversion targets the N-terminal CW-site in vitro (Dierker et al., 2009; Ohlig et al., 2012). We show here that insertion of HA peptides, which displaces this cleavage site distally without affecting Hh N-palmitoylation (Hardy and Resh, 2012) and HS-dependent multimerization, is sufficient to impair endogenous and transgenic Hh high threshold biofunction in vivo, apparently without affecting low threshold Dpp-mediated Hh activity. Rescue of Hh biofunction by the additional mutation of the cysteine acceptor shows that N-palmitate anchors the unprocessed peptide to the cell membrane to safeguard Hh release. These findings are consistent with enhanced Drosophila Hh release upon RNAi-mediated knockdown of Hh acyltransferase activity (Chamoun et al., 2001) and increased Shh tethering to cell membranes by palmitate (Konitsiotis et al., 2014; Levental et al., 2010). Importantly, while S- and O-linked palmitate moieties are susceptible to enzymatic deacylation by palmitoyl-protein thioesterases (Kakugawa et al., 2015), amide-linked Hh palmitate is thioesterase resistant. This suggests that Hh relay from posterior subcellular structures – at least at some point – requires proteolytic processing of sheddase-accessible, membrane-proximal terminal target peptides. Support for this idea comes from the published replacement of the C-terminal Hh target peptide with transmembrane-CD2 (Strigini and Cohen, 1997). Resulting Hh-CD2 fusion proteins remain permanently membrane associated and generate wings with one single central vein in the region normally occupied by veins L3 and L4, while leaving Dpp-mediated anterior and posterior wing patterning intact. We note that this phenotype is strikingly similar to the en >HAHh phenotype described here. Moreover, required Hh transfer between protruding cell-cell contact structures emanating from the Hh-sending and Hh-receiving compartments, called cytonemes, was recently indicated by impaired Ptc signaling and internalization at contact sites with Hh-CD2 (González-Méndez et al., 2017). While the exact mechanism by which Hh is liberated from the posterior cytoneme membrane was not addressed, proteolytic Hh relay and reception at cytoneme contact sites was suggested by the authors, and is supported by the results shown in our work (Figure 1e).
In addition to cytoneme contact sites, other subcellular structures of P-compartment cells release Hh from the membrane (Figure 12). It has been suggested that the Hh gradient in Drosophila wing imaginal discs consists of apical and basolateral secreted pools formed as a consequence of initial apical Hh secretion, subsequent reinternalization, and apical (D'Angelo et al., 2015) or basolateral (Callejo et al., 2011) resecretion, both depending on the endosomal sorting complex required for transport (ESCRT). Pools of Hh and ESCRT proteins are then secreted together into the extracellular space (Gradilla et al., 2014; Matusek et al., 2014; Vyas et al., 2014), Hh being transported on (Bischoff et al., 2013) or inside of (Chen et al., 2017) basolateral cytonemes, or apically released to promote Hh long-range activity (Ayers et al., 2010) (Figure 12). While Hh shedding may target several of these pools, timely and reliable paracrine Hh function through proteolytic release and extracellular apical diffusion alone (Figure 12b’’) is difficult to envision for two reasons. First, patterning of folded epithelia, such as the Drosophila imaginal disc, poses a problem if spreading were to occur out of the plane of the epithelial cell layer through diffusion or flow, as this would result in morphogen loss into the peripodial space and loss of long-range Hh function. The second limitation is that it normally takes much time for diffusing molecules to travel long distances away from the source because the timescale of diffusion increases with the square of the distance (Berg, 1993; Müller and Schier, 2011). Cytoneme- or exosome-mediated basolateral transport, followed by proteolytic Hh relay over short distances at membrane contact sites (González-Méndez et al., 2017), would effectively solve both problems, as would the idea of heparan sulfate proteoglycan ‘restricted’ Hh transport at the apical cell surface (Han et al., 2004). Our future aim is to distinguish between these possibilities. We also aim to characterize the Hh release factor Shifted (Glise et al., 2005), a soluble protein with structural similarities to vertebrate Scube2 sheddase enhancers (Jakobs et al., 2014; Jakobs et al., 2016; Jakobs et al., 2017), to identify the elusive ‘Hh sheddase’. Finally, we are currently investigating the important question of whether C-terminal Hh processing contributes to its in vivo biofunction in the wing disc and in other developing tissues requiring Hh signaling over shorter ranges, such as in the developing eye (Ma et al., 1993).
In conclusion, we propose that palmitate-controlled quantitative Hh shedding from the cell surface constitutes an essential step in Hh transmission and high-threshold tissue patterning in vivo. While we have used Drosophila wing development to elucidate this molecular process, we expect our results to also be relevant to other Hh-dependent developmental programs and to Hh ligand-dependent cancer induction and progression (Amakye et al., 2013).
The following fly lines were used: Ptc-Gal4 (ptc>): w[*]; P(w[+mW.hs]=GawB)ptc[559.1], Bloomington stock #2017; En-Gal4e16E (En>): P(en2.4-GAL4)e16E, FlyBaseID FBrf0098595; Hh-Gal4 (hh>): w[*];; P(w[+mC]=Gal4)hh[Gal4], Bloomington stock #67046; en(2)-Gal4 (en(2)>): w1118;; P(GMR94D09-Gal4), Bloomington stock #48011; 34B-Gal4 (34B>): y1w[*];; P(w[+mW.hs]=GawB)34B, Bloomington stock #1967. These lines were crossed with flies homozygous for UAS-hh or variants thereof. All Hh cDNAs cloned into pUAST-attP were first expressed in Drosophila S2 cells to confirm correct protein processing and secretion. Transgenic flies were generated by using the landing site 51 C1 by BestGene or in-house by using strain PhiC31(X); attPVK37; attP2 that possesses the landing sites VK37 and attP2. Cassette exchange was mediated by germ-line-specific phiC31 integrase (Bateman et al., 2006). Ptc-LacZ reporter flies were kindly provided by Jianhang Jia, Markey Cancer Center, and Department of Molecular and Cellular Biochemistry, University of Kentucky College of Medicine, Lexington, USA.
Wing discs were fixed, permeabilized and stained with anti-β-galactosidase antibodies (Cappel, MP Biomedicals) and Cy3-conjugated goat-α-rabbit antibodies (Jackson Immuno Research). Posterior, Hh-producing cells were detected with monoclonal antibodies directed against engrailed (en 4D9, DSHB) and Alexa488-conjugated donkey-α-mouse antibodies (Thermo Fisher). Images were taken on a LSM 700 Zeiss confocal microscope using ZEN software. Maximum intensity projections are shown.
Hh cDNA (nucleotides 1–1416, corresponding to amino acids 1–471 of D. melanogaster Hh) and HhN cDNA (nucleotides 1–771, corresponding to amino acids 1–257) were inserted into pENTR, sequenced, and cloned into pUAST for protein expression in S2 cells or the generation of transgenic flies. Mutations were introduced by QuickChange Lightning site-directed mutagenesis (Stratagene). Primer sequences and sequence information is shown in Supplementary file 1. S2 cells (RRID: CVCL_Z232) were cultured in Schneider’s medium (Invitrogen) supplemented with 10% fetal calf serum (FCS) and 100 μg/ml penicillin/streptomycin. The cells were obtained from C. Klämbt, University of Münster, Germany, and tested negative for mycoplasma. S2 cells were transfected with constructs encoding Hh and HhN variants together with a vector encoding an actin-Gal4 driver by using Effectene (Qiagen) and cultured for 48 hr in Schneider’s medium before protein was harvested from the supernatant. Shh constructs were generated from murine cDNA (NM_009170) by PCR (primers are listed in Supplementary file 1). Hh acyltransferase cDNA (NM_018194) was obtained from ImaGenes and cloned into pIRES (ClonTech) for bicistronic Shh/Hh acyltransferase coexpression in the same transfected cells. This resulted in N-palmitoylated, C-cholesterylated proteins. Bosc23 cells (RRID: CVCL_4401) were cultured in Dulbecco's modified Eagle’s medium (Lonza) supplemented with 10% FCS and 100 µg/ml penicillin-streptomycin. The cells were obtained from D. Robbins, University of Miami, USA, authenticated via by PCR-single-locus-technology (Eurofins Forensics), and tested negative for mycoplasma. Bosc23 cells were transfected with PolyFect (Quiagen) and cultured for 48 hr, the medium was changed, and Shh was secreted into serum-free medium for the indicated times. All media were ultracentrifuged for 30 min at 125,000 g, and the proteins were TCA precipitated and analyzed by 15% SDS-PAGE and western blotting with polyvinylidene difluoride membranes. Blotted proteins were detected by α-HA antibodies (mouse IgG; Sigma), α-Shh antibodies (goat IgG; R and D Systems), or α-Hh (rabbit IgG, Santa Cruz Biotechnology). Incubation with peroxidase-conjugated donkey-α-goat/rabbit/mouse IgG (Dianova) was followed by chemiluminescent detection (Pierce). Photoshop was used to convert grayscale blots into merged RGB pictures for improved visualization of terminal peptide processing.
Drosophila third-instar larvae were collected and transferred into a microcentrifuge to which 1 ml lysis buffer was added (PBS containing 1% (v/v) Triton X-100). Larvae were homogenized with a micropestle and the solution was cleared at 15,000 rpm for 15 min at 4°C. The supernatant was sterile-filtered (45 μm) and transferred into a fresh microcentrifuge tube for gel filtration analysis. All processings were conducted at 4°C.
Gel filtration analysis was performed on an Äkta protein purifier (GE Healthcare) on a Superdex200 10/300 GL column (Pharmacia) equilibrated with PBS at 4°C. Eluted fractions were TCA precipitated and analyzed by SDS-PAGE as described earlier. Signals were quantified by using ImageJ.
Sequence analysis was conducted on the CFSSP secondary structure prediction server (http://www.biogem.org/tool/chou-fasman/). All statistical analyses were performed in GraphPad Prism by using the Student’s t test (two-tailed, unpaired, confidence interval 95%). For wing quantifications, 10 male and 10 female wings were analyzed for each data set and ratios between L3-L4 intervein areas and L2-L3 intervein areas determined. All error estimates are standard errors of the mean (SEM).
Shh-expressing Bosc23 cells were fixed overnight at 4°C in 4% paraformaldehyde/glutaraldehyde, washed in PIPES, and dehydrated in a graded ethanol series (30% EtOH, 4°C, 45 min; 50% EtOH, −20°C, 1 hr; 70% EtOH, −20°C, 1 hr; 90% EtOH, −20°C, 1.5 hr; 100% EtOH, −20°C, 1.5 hr; 100% EtOH, −20°C, 1.5 hr). Dehydrated cells were embedded in Lowicryl K4M embedding medium by using the Lowicryl K4M Polar Kit (Polysciences). Cells were then embedded in gelatin capsules, centrifuged twice for 15 min at 1500 rpm, and incubated overnight at −35°C. For polymerization, the resin was UV irradiated for 2 days at −35°C. The embedded samples were cut into 60 nm sections, washed in PBS containing 5% BSA (pH 7.4), and incubated for 2 hr in PBS-BSA containing primary antibodies (α-Shh antibodies from R and D, GeneTex, and Cell Signaling at 1:20 dilution). Samples were washed five times in PBS-BSA and once in Tris-BSA. Secondary antibodies conjugated to 5 nm and 10 nm gold nanoparticles were diluted in Tris-BSA buffer and incubated with the cell sections for 1 hr. Afterwards, samples were washed five times in Tris-BSA and once in dH2O. Contrasting was done with 2% uranyl acetate (15 min) and Reynold’s lead citrate (3 min). Finally, immunogold-labeled cell sections were analyzed by using a transmission electron microscope (CM10, Philips Electron Optics).
The transgenic fly lines Hh-CW (lacking the putative N-terminal Hh processing site), Hh-CW/HA (a variant having this site replaced with a hemagglutinin tag) and HhHS (carrying a C-terminally inserted HA-tag) generated in the course of this study that support the phenotypes described in the manuscript are available upon request from the corresponding author (KG). We plan to publish these new lines separately in the future.
Unraveling the therapeutic potential of the Hedgehog pathway in cancerNature Medicine 19:1410–1422.https://doi.org/10.1038/nm.3389
An intramolecular association between two domains of the protein kinase Fused is necessary for Hedgehog signalingMolecular and Cellular Biology 24:10397–10405.https://doi.org/10.1128/MCB.24.23.10397-10405.2004
Genetic and epigenetic mechanisms regulating hedgehog expression in the Drosophila wingDevelopmental Biology 327:508–515.https://doi.org/10.1016/j.ydbio.2009.01.006
Random Walks in BiologyPrinceton, NJ: Princeton University Press.
The structure of SHH in complex with HHIP reveals a recognition role for the Shh pseudo active site in signalingNature Structural & Molecular Biology 16:691–697.https://doi.org/10.1038/nsmb.1632
Two distinct sites in sonic Hedgehog combine for heparan sulfate interactions and cell signaling functionsJournal of Biological Chemistry 286:44391–44402.https://doi.org/10.1074/jbc.M111.285361
Patterns in evolution: veins of the Drosophila wingTrends in Genetics 20:498–505.https://doi.org/10.1016/j.tig.2004.07.013
Differential range and activity of various forms of the Hedgehog proteinBMC Developmental Biology 5:21.https://doi.org/10.1186/1471-213X-5-21
Heparan sulfate-modulated, metalloprotease-mediated sonic hedgehog release from producing cellsJournal of Biological Chemistry 284:8013–8022.https://doi.org/10.1074/jbc.M806838200
Pro-sequence-assisted protein foldingMolecular Microbiology 16:609–614.https://doi.org/10.1111/j.1365-2958.1995.tb02423.x
A highly conserved amino-terminal region of sonic hedgehog is required for the formation of its freely diffusible multimeric formJournal of Biological Chemistry 281:4087–4093.https://doi.org/10.1074/jbc.M511427200
Exosomes as Hedgehog carriers in cytoneme-mediated transport and secretionNature Communications 5:5649.https://doi.org/10.1038/ncomms6649
Identification of N-terminal residues of Sonic Hedgehog important for palmitoylation by Hedgehog acyltransferaseJournal of Biological Chemistry 287:42881–42889.https://doi.org/10.1074/jbc.M112.426833
Smoothened translates Hedgehog levels into distinct responsesDevelopment 130:3951–3963.https://doi.org/10.1242/dev.00594
Scube2 enhances proteolytic Shh processing from the surface of Shh-producing cellsJournal of Cell Science 127:1726–1737.https://doi.org/10.1242/jcs.137695
Calcium coordination controls sonic hedgehog structure and Scube2-cubulin domain regulated releaseJournal of cell science.
Hedgehog pathway antagonist 5E1 binds hedgehog at the pseudo-active siteJournal of Biological Chemistry 285:26570–26580.https://doi.org/10.1074/jbc.M110.112284
An emerging role of Sonic hedgehog shedding as a modulator of heparan sulfate interactionsJournal of Biological Chemistry 287:43708–43719.https://doi.org/10.1074/jbc.M112.356667
Sonic hedgehog processing and release are regulated by glypican heparan sulfate proteoglycansJournal of Cell Science 128:2374–2385.https://doi.org/10.1242/jcs.170670
Identification of a palmitic acid-modified form of human Sonic hedgehogJournal of Biological Chemistry 273:14037–14045.https://doi.org/10.1074/jbc.273.22.14037
Matthew FreemanReviewing Editor; University of Oxford, United Kingdom
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]
Thank you for submitting your work entitled "Proteolytic processing of palmitoylated Hedgehog peptides specifies the 3-4 intervein region of the Drosophila wing" for consideration by eLife. Your article has been reviewed by a Senior Editor and three reviewers, one of whom is a member of our Board of Reviewing Editors.
Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.
All the reviewers found the paper to be potentially interesting: it is well written and deals with an important topic. But after discussion the consensus view was that the conclusions are not yet well enough supported by the experimental data to justify publication in eLife. In addition to the need for the experiments described to be more rigorously performed, the proposed model makes several predictions that could be tested to provide further evidence.
The reviewers were keen to point out that although eLife has a policy of rejecting papers where it is judged that more than 2 months will be needed for revisions, were this work to develop into a more mature story, there is no barrier to resubmitting it in the future.
This is an interesting paper that uses a combination of microscopy, cell biology and Drosophila genetics to carefully dissect the mechanisms that underlie Hh release and intercellular trafficking. The aim is to resolve the mystery of how a dual lipidated extracellular protein, which oligomerizes and tightly associates with the cell surface of the producing cell, is nevertheless transported across an epithelial field of cells to act on cells at a distance. Several models have been proposed, including trafficking by exosomes or cytonemes, but the conclusion of this paper is that proteolytic shedding of the palmitoylated N-terminal of Hh is the key step.
This is a thought-provoking and carefully designed set of experiments that lead to some plausible and generally well supported conclusions. It's also an important topic, being a central issue about the regulation of a primary signaling pathway of relevance to development and disease. In general, I think it is appropriate for a journal like eLife.
It is, however, a complicated story, and one which still leaves quite a few questions: it will not be the last word on this topic.
Although the experiments are well planned and quite elegant, I found some are hard to follow. I'd suggest more use of diagrams to outline the logic of individual experiments. An example would be the use of Ptc>HhC85S to discriminate between autonomous and non-autonomous effects. People not used to fly developmental biology will find this difficult. Another case would be the Bosc23 cells expressing the differently tagged versions.
The tone of many of the results is quite definitive and I am not sure this is wholly justified. The results are clear and the interpretations plausible, but I am not convinced that in each case their interpretation is the only way of interpreting the data. The paper would be stronger if the current results were more clearly set in the context of other models, with more discussion about how these results challenge or support different views. Since there are already quite well discussed models of Hh trafficking, the current authors need to leave the reader with as clear as possible a view of what this new paper changes and what are the still outstanding questions.
One gap I detected was any discussion about the C-terminal cleavage. If I understand their model correctly, this is also essential for release of active Hh. Is the view that the clustering is also necessary for that? And do they envisage that the N- and C-terminal cleavages are done by the same or different proteases? Moreover, what is known about the identity of the proteases should also be discussed.
The maturation of Sonic Hedgehog (Shh) is very atypical as the final mature protein is dually lipid modified, by an N-term palmitic acid and a covalent link with a molecule of cholesterol on the C-terminus. How it is extracted from the plasma membrane bilayer and is transported to long-range targets is still enigmatic.
The laboratory of Kay Grobe has previously published that SHH is secreted as a multimer in which the lipidated N-term peptide inhibits each adjacent SHH molecule by masking its receptor binding site. The lab has proposed that SHH dual lipids are shed at the surface of producing cells, allowing SHH propagation and activation. The cleavage of these inhibitory peptides therefore unmasks receptor-binding sites of all SHH molecules present in the multimer and permits its propagation and activation of signaling (Ohlig et al., 2011). This last study was performed using in most cases the human kidney cell line HEK293 and its derivative the Bosc23 cell line, which is a model for cancer research. In this new manuscript, the authors use Drosophila transgenics to confirm their model. As appropriate controls are missing, I am overall not convinced about the accuracy of the conclusions proposed here. I highlight below my main critics:
Overexpression of the non-palmitoyled HH variant in wild type HH producing cells of the Drosophila wing discs reveals a dominant negative activity on the final adult wing pattern. The overexpression of this variant in HH receiving cells had only a weak effect. From this difference the authors came to the conclusion that the non-palmitoyled HH variant display a dominant negative effect only when expressed in the same cells with the wild type protein and forming a mixed cluster. This is the basis of the model but I believe several other hypothesis can be proposed (in addition, the fact that not HH target genes are presented in this study weaken the entire picture). For example, the authors do not show that the two drivers express the HH variant at the same level which could explain a differential dominant negative effect. It is also possible that upon forced expression of HH in anterior cells, the protein is not secreted at the correct pole (apical/basal) of the cell and thus cannot bind properly to its receptor. No staining of extracellular HH is provided here. Conventional and extracellular staining of the different HH variants is imperative to get a clearer picture of the behavior of these variants in vivo. One should note that overexpression of this variant has been published previously by several other groups in 2006 and showed an opposite result, with expansion of target genes in both embryos and wing imaginal discs. This is not commented by the authors.
The authors showed that N-terminal truncation of HH reversed the dominant negative activity of the non-palmitoyled HH variant when expressed in flies. These new variants are not forming multimere but monomere. From this, the authors proposed that the inhibitory effect of unprocessed N-terminal peptides is linked to its presence in the multimere. But it is also possible that these new variants cannot bind properly Hh receptor and thus are just "neutral" with no activity whereas the non-palmitoyled variant binds the receptor but do not activate signaling properly. There are numerous in vitro assays that could be performed to test the author's hypothesis, for example, why not analyze binding affinity of HH variants to Ptc with synthesized peptides?
Altogether, the authors propose a model in which Hh is secreted to the plasma membrane as dually lipidated molecule, assembled as multimer at the cell surface, and then cleaved for propagation. What is surprising with the author's model is that both lipids need to be cleaved to release Hh functional cluster from producing cells. This is intriguing as numerous labs have shown that expression of a HH variant without lipid is secreted as monomere with no activity. So, the authors would need to resolve two important questions to confirm their model. First, how the cleaved HH molecule is kept assembled as multimer, and second, how the putative multimer of non-lipidated HH proposed here is able to provide signaling activation when a monomeric non-lipidated HH cannot.
In conclusion, I did not find that the new data provide convincing data and progress for a firm evidence of the model proposed here. I also advise this team to take into more consideration many of the in vivo data already published by other labs, which in several cases appears to be contradictory to the author's own observation.
The manuscript by Grobe and colleagues use the Drosophila wing as the principal model system to analyse the functional role of protein palmitoylation in the N-terminal cysteine of Hedgehog (Hh) signalling molecule. Authors present evidence that mutant forms of Drosophila Hh unable to be palmitoylated (Hh-C85S) or proteolytically processed (HA-Hh) gives rise to a Hh-loss of function adult wing phenotype when overexpressed in Hh-expressing but not responding cells, and that truncation of the N-terminal region of Hh-C85S or a mutation of the N-terminal cysteine in HA-Hh restore these phenotypes. Authors propose that palmitoylation of Hh contributes to the clustering of Hh molecules at the cell membrane of Hh-expressing cells and to the tightly controlled-proteolytic processing and release of an active form of Hh able to bind to Ptc in nearby cells.
The paper is excellently written and the proposed working model is consistent with the results. Authors have done an excellent and thorough exercise by using the Drosophila wing together with EM data and cell culture experiments and propose a mechanistic model of the biological function of Hh-palmitoylation.
Authors conclude that the adult wing phenotype caused by expression of Hh-C85S or HA-Hh in the P compartment is caused by a dominant negative effect on the endogenous Hh molecule (as nicely depicted in their cartoons). Authors should present evidence (1) that the wing phenotype can be rescued by co-expression of a wild type form of Hh, (2) that release of a wild type form of Hh into the adjacent compartment is compromised by Hh-C85S or HA-Hh co-expression (by using available UAS-Hh-GFP transgenic flies), and (3) that anterior cells abutting the P compartment are still sensing Hh signalling (similarly to a tethered form of Hh). Thus, expression of a wild type form of Hh and Hh target genes should be analysed in wing primordia of all genotypes (en-gal4; UAS-Hh-C85S +/- UAS-Hh-GFP and en-gal4; UAS-Hh-A +/- UAS-Hh-GFP) to confirm their proposal. These experiments are feasible.
1) Were male of female wings used in their L3-L4/L2-L3 region measurements?
2) Authors should notice that en-gal4 is also expressed in the A compartment, although at lower levels and later in development. The ability of Hh-C85S and HA-Hh to cause a wing phenotype should be analysed also with the hh-gal4 driver (again available in the fly community).
4) How certain are authors that veins in Figure 6B are L4 and L5 and not L2 and L5. The latter would be expected as these are the ones that depend on Dpp signalling.
[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]
Thank you for resubmitting your work entitled "Proteolytic processing of palmitoylated Hedgehog peptides specifies the 3-4 intervein region of the Drosophila wing" for further consideration at eLife. Your revised article has been favorably evaluated by Naama Barkai (Senior Editor), a Reviewing Editor, and three reviewers.
The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below.
The editors and reviewers have discussed your paper extensively. Our position is that your work is interesting, provocative and of potential broad significance. On the other hand, there is a clear consensus that you do not currently provide a clear enough description of the controversy and uncertainty that surrounds your interpretation. Your paper does not yet allow a definitive conclusion about the model.
eLife is an appropriate journal for this kind of work, but only if the authors are able clearly and explicitly to describe their work in the wider context, allowing a reader to form a balanced picture of where the field stands. We would therefore like to give you one more opportunity to respond to remaining concerns but also – and even more importantly – to rewrite the Introduction and Discussion in a way that gives that broader context and describes to a non-expert reader the uncertainties that remain about your model. What would be a definitive experiment that might prove or refute it? Which areas are you least confident about? Do you have a plan about how to address them? How do you reconcile your ideas with previous apparently contradictory data?
Science is full of uncertainty and journals like eLife should not encourage this to be hidden behind unrealistic confidence or advocacy. If you can amend your paper to provide a fair and balanced description of how the field stands, and how your new data move it forward, but also where the uncertainties still lie, we would be happy to publish it. If you feel that this is not in your best interest and you prefer to argue for your model more directly, our decision would be not to accept it.
In addition to these general points there are still some remaining specific issues in the reviews below. We would encourage you to address these as fully as possible. Where necessary and appropriate, please acknowledge if they provide significant challenges to your model.
As it is the second time I review this work, I will directly follow the main comments I have put together for the first manuscript.
My previous comments: "Overexpression of the non-palmitoyled HH variant in wild type HH producing cells of the Drosophila wing discs reveals a dominant negative activity on the final adult wing pattern. The overexpression of this variant in HH receiving cells had only a weak effect. From this difference the authors came to the conclusion that the non-palmitoyled HH variant display a dominant negative effect only when expressed in the same cells with the wild type protein and forming a mixed cluster. This is the basis of the model but I believe several other hypothesis can be proposed (in addition, the fact that not HH target genes are presented in this study weaken the entire picture).”
The authors made some effort (by providing target gene expression) to answer this first comment. Nevertheless, there are numerous examples in the literature showing that expression of the non-palmitoyled HH variant in the anterior cells of the wing disc leads to a strong dominant negative effect (see for example, Figure D in Lee et al., 2001, Treissman lab). In the same paper, the authors mentioned that overexpression of the non-palmitoyled SHH variant in the posterior region of the mouse limb have no inhibitory effect on endogenous Shh activity, which contradict the model proposed here.
So, the first proposal presented here, that, in producing cells, non-palmitoyled HH proteins associate in cluster with other HH protein in order to act as a dominant negative variant does not fit with published data. In addition, if this model was correct, one would expect a stoichiometric dominant negative effect of the non-palmitoyled HH variant, which again is not supported by the study of Williams et al., 1999.
Also, I still do not understand the model based on a "model linear Drosophila Hh cluster", which I assume is based on the published crystallographic structure of vertebrate Shh. Surprisingly, a non-palmitoyled SHH does not form multimers, whereas a non-palmitoyled Drosophila HH does so (Figure 7 in Chen et al., 2004, Pao-Tien Chuang lab). So it is not clear to me how one can use information from Shh and extrapolate to Hh behavior, knowing that these two proteins behave differently.
My previous comments: "No staining of extracellular HH is provided here. Conventional and extracellular staining of the different HH variants is imperative to get a clearer picture of the behavior of these variants in vivo."
Authors: “As has been pointed out by Art Landers group in Irvine, reliable fluorescence microscopic visualization of free extracellular protein transport is close to impossible:
Using Dpp as an example, it was estimated that free extracellular protein is unlikely to account for more than 3% of total morphogen, and less than 1% of what is normally visualized by fluorescence imaging (Zhou et al., 2012). […] For all of these reasons, we refrained from trying to visualize extracellular Hh in vivo by fluorescence microscopy, and decided to use clear unambiguous genetic assays instead.”
I still believe that, a non-detergent staining at 4 °C with Hh and/or HA antibodies will provide information such as: are these Hh variants accumulating at the cell surface of producing cells, how do they distribute in anterior cells compared to wt Hh etc.? There are also tricks (with for example the expression of a Ptc variant which is not internalized) to increase and visualize accumulation of Hh at the cell surface of receiving cells. As previous studies demonstrated that Ptc-binding restricts Hh spreading, one would expect that HhC25S-containing Hh multimers would spread more anteriorly because they bind Ptc less well compared with wild type Hh as suggested by the present model. Also, there are Hh variants presented in this study which are likely not binding to Ptc, such as the HHC85S;delta86-100. Indeed, a deletion of the first 20 NH2 terminal residues of SHH leads to a protein that do not bind to Ptc anymore (Williams et al., 1999).
My previous comment: “One should note that overexpression of this variant has been published previously by several other groups in 2006 and showed an opposite result, with expansion of target genes in both embryos and wing imaginal discs. This is not commented by the authors.”
Authors: “We wonder which papers and which variant the reviewer does refer to. If it is HhC85S, we are not aware of any publication claiming that this form has any bioactivity; the general consensus is that non-palmitoylated invertebrate Hh is always completely inactive.”
Well, both studies described in Gallet et al., 2003 (Figure 2) and Callejo et al., 2006 (Figure 6) showed that embryonic wg, and larval dpp and iro expressions are expanded under overexpression of HhC85S.
Another issue which needs clarification regards the putative multimer of non-lipidated HH which is proposed in the model here (Figure 5A). How is this multimer able to provide signaling activation whereas a monomeric non-lipidated HH cannot? And how the Hh proteins is kept assembled in the multimer without lipids? This is difficult to understand for me. The authors might be right but their model of cell-surface shedding relies mainly on in vitro experiment in which Shh is expressed at a non-physiological level in cultured cells. So far, if I am correct, in all their "Hh shedding" studies, the authors used the human kidney cell line HEK293 and its derivative the Bosc23 cell line, which is a model for cancer research and which might not mimic physiological conditions.
Interestingly, the acyl transferase that catalyzes the transfer of a palmitoyl moiety to Hh has been identified in flies. When its homolog is mutated in mice (Chen et al., 2004), digit 3 to 5 (which depends on long range activity of Shh) still form, suggesting that, if Shh assembles in multimere in this mutant, the non-palmitoyled SHH is still able to bind Ptc and activate the pathway. How do the authors reconcile their data with this observation?
Furthermore, a wide variety of hydrophobic modifications have been shown to increase the potency of Shh when added at the N-terminus of the protein, ranging from long-chain fatty acids to hydrophobic amino acids (Taylor et al., 2001). I do not see how this fits with the authors model.
Overall, this is a story that is ultimately still not convincing for this reviewer. The model proposed here cannot explain many of the in vivo observations obtained by other labs, some of which are listed above.
Schuermann et al. provide evidence that Hh activity and spread depend on the proteolytic removal of the palmitoylated N-terminal peptide. The paper contains a huge amount of data but there is no "killer experiment" that definitively proves the proposed model. For example, most of the conclusions rest on the use of ectopic expression of mutant versions in a wild-type background instead of replacement of the endogenous hh gene. Moreover, no in vivo localization data is provided. I therefore encourage the authors to raise the possibility that there might be other interpretations of the data and discuss the need for follow up studies of Hh versions expressed from the endogenous locus and visualized in vivo. Despite these concerns, the findings are interesting, provocative and of broad interest.
1) Is the packing seen in Figure 1 really due to Hh or could it be caused by the confirmation and size of the antibodies?
2) The evidence that the mutant versions are expressed at similar levels as wild-type Hh is sparse.
3) The authors try to build models on complex regulatory loops (e.g. ptc feedback in Figure 2) but the outcomes of such loops are very difficult to predict.
4) Subsection “N-palmitate controls Hh release from the cell surface in vitro”, last paragraph: "compelled"?
Authors have successfully addressed all my concerns, so I believe the manuscript is ready for eLife.https://doi.org/10.7554/eLife.33033.029
- Kay Grobe
- Kay Grobe
- Kay Grobe
- Milos Galic
- Kay Grobe
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
The excellent technical and organizational assistance of S Kupich and R Schulz is gratefully acknowledged. The authors thank Marius Mählen for contributing to this work. This work was financed by DFG (German Research Council) GRK1549/1, GR1748/4-1, GR1748/5-1, and CiM FF-2015–02 support.
- Matthew Freeman, Reviewing Editor, University of Oxford, United Kingdom
- Received: October 23, 2017
- Accepted: February 9, 2018
- Version of Record published: March 9, 2018 (version 1)
© 2018, Schürmann et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.