Hedgehog is relayed through dynamic heparan sulfate interactions to shape its gradient

Abstract

Cellular differentiation is directly determined by concentration gradients of morphogens. As a central model for gradient formation during development, Hedgehog (Hh) morphogens spread away from their source to direct growth and pattern formation in Drosophila wing and eye discs. What is not known is how extracellular Hh spread is achieved and how it translates into precise gradients. Here we show that two separate binding areas located on opposite sides of the Hh molecule can interact directly and simultaneously with two heparan sulfate (HS) chains to temporarily cross-link the chains. Mutated Hh lacking one fully functional binding site still binds HS but shows reduced HS cross-linking. This, in turn, impairs Hhs ability to switch between both chains in vitro and results in striking Hh gradient hypomorphs in vivo. The speed and propensity of direct Hh switching between HS therefore shapes the Hh gradient, revealing a scalable design principle in morphogen-patterned tissues.

Fig. 1. Two HS-binding sites in Shh and modeled Drosophila melanogaster Hh.

a–d The first established HS-binding site (the Cardin-Weintraub (CW) motif) of Shh (a, b, pdb: 3m1n) or of D. melanogaster Hh (c, d) is located at the N-terminal extension of one monomer (left in a, c and top in b, d), whereas the second site comprises residues on the globular domain. HS-binding basic residues are in different shades of blue and residues that also bind to Ptc receptors are in violet. Both HS-binding sites are far from each other in the monomeric structures, with a median distance of ~2.7 nm. See Supplementary Tables 1, 2 for details. e In silico analysis of the effect of mutations in wild-type Drosophila HhNΔ251–257 and human ShhNΔ191–197 on the electrostatic contribution to the binding affinity against a heparin disaccharide. ΔΔG^elec values express the difference of ΔG^elec between each mutant protein and its corresponding wild-type protein. In Hh, both the CW domain and globular domain contribute to heparin binding. In both Shh and Hh, the globular glutamate variants have a lower affinity to heparin than the alanine variants. Annotations M and S distinguish results based on different template structures of protein (Electrostatic calculations in Methods, Source data are provided as a Source Data file). f Remaining fraction of Hh (black) and Hh^R/A (red) after 0, 3, 6, and 12 h. Crosses correspond to measured data with small shifts along the time axis used to separate points for clarity. Dots and uncertainty bar mark mean fractions estimated with a Bayesian model (see supplementary information), with the dot marking the median estimate and the uncertainty bars covering 5 to 95% quantiles. As the uncertainty bars have large overlaps, we cannot distinguish the decay dynamics of Hh and Hh^R/A with confidence. Raw data and 5 to 95% quantiles are provided as Source Data file. g A₄₀₅ measurements for each of the three protein variants (color legend) at three different concentrations are shown as dots. A random jitter in the horizontal direction was added to avoid overplotting. The three lines correspond to linear models fitted with the data of the respective proteins. 95%-confidence ribbons (gray areas) for the linear models largely overlap, indicating that courses of data for different proteins cannot be distinguished with confidence. Raw data and 95%-confidence ribbons are provided as Source Data file. h Hh, Hh^R/A, and Hh^R/E induce Hh-specific differentiation of C3H10T1/2 reporter cells to similar degrees, confirming that the arginines do not contribute to Ptc receptor binding. Inset: similar amounts of Hh variants were used in this experiment. One-way ANOVA, Dunnett’s multiple comparisons test, n = 12 for each set. Error bars represent standard deviation. F = 0.098. p = 0.9. Source data, means and p values (n.s.: p > 0.05) are provided as Source Data file.

Fig. 2. Opposing phenotypes of flies expressing hh versus hh^R/A or hh^R/E under endogenous promotor control.

a Quantification of the numbers of ommatidia per eye in female flies. Hh biofunction in eye disks of flies homozygous for transgenic hh (in hh[KO;hh]/hh[KO;hh] flies) is increased over that in w¹¹¹⁸ or heterozygous controls. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparison test. n = 10 for each genotype, F = 13.6. Error bars represent standard deviation. ***p ≤ 0.0004. Source data are provided as a Source Data file. b Inset: Hh produced in the posterior compartment of the wing disc directly patterns the L3-L4 intervein area (orange) and indirectly patterns the L2-L3 intervein area (green). Quantification of L3-L4/L2-L3 intervein areas revealed significantly increased patterning activities of transgenic Hh. Statistical significance was determined by unpaired t-test (two-tailed). n = 13 wings for w¹¹¹⁸ and n = 27 for hh[KO;hh]/hh[KO;hh], ****p < 0.0001. L1-5 indicate the five longitudinal Drosophila wing veins. Source data are provided as a Source Data file. c Segment phenotype changes in hh[KO;hh^R/A]/hh[KO;hh^R/A] larvae resemble those in sfl/sfl larvae that are deficient in HS biosynthesis, and hh[KO;hh^R/E]/hh[KO;hh^R/E] larvae resemble hh^AC/hh^AC larvae that lack hh expression. Segment phenotypes were consistently observed in three independent crosses. Scale bar: 100 μm. c′ Schematic of the self-reinforcing Hh/Wg signaling loop at one of the 14 parasegment boundaries in Drosophila embryos. Due to Hh/Wg interdependency to maintain the loop, depletion of either morphogen will result in embryos that lack a naked cuticle. d Transgenic hh and both hh variants restore eye development in flies homozygous for impaired endogenous hh expression specifically in the eye (hh^bar3/hh^bar3). This demonstrates the largely retained bioactivity of both Hh variants in this system. Scale bar: 100 μm. e Quantification of the numbers of ommatidia per eye of female flies as shown in (d). Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparison test. n = 8 for hh^bar3/hh^bar3, n = 10 for hh^bar3/+ and hh[KO;hh]/hh^bar3, and n = 30 for hh[KO;hh^R/A]/hh^bar3 or hh[KO;hh^R/E]/hh^bar3. Error bars represent standard deviation. F = 357. ****p ≤ 0.0001, ns: p = 0.99. Source data are provided as a Source Data file.

Fig. 3. Eye-disc-specific hh variant expression from clones in moving and static sources.

a The cartoon shows reiterated cell-to-cell Hh signaling from a clonal source (Supplementary Fig. 5) moving at a particular pace to drive photoreceptor differentiation across the eye primordium. Red arrows denote Hh spread, blue arrows denote the movement of the morphogenetic furrow (MF), t₀-t₊₂ indicates early and later stages. The asterisk and dashed area indicate the ocellar field (as schematically shown in b). Clones homozygous for hh[KO] do not express hh, which severely impairs eye development, but hh or hh variant cDNA expression from both alleles restores eye development. Scale bar: 100 μm. Despite the observed restoration of eye development, quantification of ommatidia numbers per eye of female flies reveals significantly impaired activities of both hh variants. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparison test. Error bars represent standard deviation. n = 6 (hh), n = 32 (hh[KO]), n = 19 (hh[KO;hh]), n = 18 (hh[KO;hh^R/A]), and n = 22 (hh[KO;hh^R/E]), F = 1474, ****p ≤ 0.0001. Ctrl: FRT82B donor line. Source data are provided as a Source Data file. b More severely affected development of another eye disc area (as shown in the cartoon adapted from) that uses the same Hh signaling components, but differs from the system described in (a) in that it depends on Hh signaling over a relatively long distance (tens of micrometers) from a static source (indicated by red bent arrows). The asterisk in (a) marks this ocellar region in the disc. Clones homozygous for hh[KO] do not express hh in this area and strongly impair the development of one or more of the three camera-type eyes called anterior and posterior ocelli (aOC and pOC, red circle, the hh expression domain will become the interocellar region (iOR), Supplementary Fig. 5). Clones homozygous for hh[KO;hh^R/E] also impaired their development. The average number of ocelli per hh[KO] fly: 0.44 ± 0.6 ocelli, n = 16. hh[KO;hh^R/E]: 1 ± 0.8 ocelli, n = 24. The normal number of three ocelli/fly was observed for all other genotypes. Error bars represent standard deviation. No sex bias was observed in any assay. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparison test, F = 99. n = 6 (hh), n = 16 (hh[KO]), n = 20 (hh[KO;hh]), n = 18 (hh[KO;hh^R/A]), and n = 24 (hh[KO;hh^R/E]). ****p ≤ 0.0001, ns: p > 0.99. Scale bar: 100 μm. Source data are provided as a Source Data file. Bottom: Fly genotype used in these experiments, the X denotes no hh (hh[KO]), hh (hh[KO;hh]), or hh variants.

Fig. 4. Substitution of HS-binding Hh amino acids affects Drosophila wing patterning.

a Hh produced in clones of the posterior wing disc compartment (as shown in the cartoon, Hh is produced in the posterior compartment of the disc and moves over a significant distance to the anterior receiving compartment, Supplementary Fig. 5) directly patterns the central domain of the adult wing (orange). Wing patterning beyond the central domain (the green area and the anterior top space that is not colored) depends on Dpp expression. Note that wing-mispatterning phenotypes are similar if due to the lack of Hh production in wing disc clones homozygous for hh[KO] or if homozygous for hh[KO;hh^R/E]. Clonal hh expression, in contrast, completely restored wing patterning. b Quantification of wing phenotypes as shown in (a). Wings of female flies developing at 25 °C were analyzed, and Hh function during development was expressed as L3-L4/L2-L3 intervein area ratios. Error bars represent standard deviation. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparison test, F = 152. n = 10 (hh), n = 26 (hh[KO]), n = 39 (hh[KO;hh]), n = 34 (hh[KO;hh^R/A]), and n = 43 (hh[KO;hh^R/E]). ****p ≤ 0.0001, **p = 0.0019, ns: p = 0.96. Source data are provided as a Source Data file. No sex bias was observed in any assay. Ctrl: FRT82B donor line.

Fig. 5. QCM-D analysis of Shh interactions with heparin.

a–a″ Schematic representation and real-time analysis of thrombin binding to the functionalized QCM-D sensor surface. a′ represents the HS/heparin model matrix assembled on a silica surface. The supported lipid bilayer (SLB) exposes 5% biotinylated lipids that bind to streptavidin (SAv) linked to biotinylated heparin. Green arrows indicate free rotation of the coupled heparin chain, black arrows lateral movement on the SLB, and gray double-headed arrows indicate softness of the matrix (as sensed by increased dissipation, ΔD). a Thrombin binding decreases ΔF during the protein incubation step and increases ΔD correspondingly, demonstrating protein binding and retention of a relatively soft heparin/thrombin layer (a″). b, b′ Similar protein loading of the functionalized matrix (-ΔF about −40 Hz in both cases) by Shh is achieved more quickly (as indicated by faster ΔF decrease) and is associated with a negative ΔD, indicating that Shh rigidifies the heparin layer. The film rigidification is due to heparin cross-linking (as confirmed by complementary FRAP assays, see Supplementary Fig. 6c). Representative graphs of three independent experiments are shown.

Fig. 6. The second HS-binding site in Shh facilitates cross-linking and switching between heparins.

a–c QCM-D binding assays analogous to Fig. 5, for Shh (a), Shh^K/A (b), and Shh^K/E (c), with an added step of competition with soluble heparin. d Overlays of frequency responses as shown in (a–c) demonstrate a reduced binding rate and increased unbinding rate in the wash buffer of Shh^K/E, and reduced unbinding of Shh^K/A and Shh^K/E upon competition with soluble heparin, compared to Shh. e Overlay of associated dissipation shifts as shown in (a–c) evidences reduced rigidification of the heparin layer by Shh^K/A, and even more so by Shh^K/E, compared to Shh, demonstrating that the second HS-binding site is essential for HS cross-linking. f Overlay of frequency responses (relative to full protein elution) from the start of soluble heparin-mediated protein elution (t = 0) until t = 40 min after elution start evidences an overall reduced elution rate for Shh^K/A and Shh^K/E compared to Shh, thus demonstrating that the second HS-binding site of Shh promotes rapid HS switching.

Fig. 7. Shh fails to cross-link and switch from heparin to lower sulfated soluble variants.

a In stark contrast to soluble heparin added to the wash buffer A, Shh does not transfer to selectively desulfated heparins with a less negative net charge. b Shh does not switch to low-sulfated chondroitin sulfate (CS, a related type of glycosaminoglycan consisting of repeating sulfated galactosamine-glucuronic/iduronic disaccharides instead of sulfated glucosamine-glucuronic/iduronic disaccharides for HS), even at high CS concentrations. c Model of direct repeated Hh switching. Two binding sites for HS facilitate Shh diffusion in an HS matrix through competition and direct repeated switching between neighboring HS chains (bent arrows). d Functional impairment of one binding site reduces the propensity for direct competition, and thus the rate of switching between HS chains, apparently slowing Hh/Shh spread down. This is expected to result in a shorter and steeper gradient. e High-affinity HS association restricts Hh/Shh movement to the cellular surfaces where suitably sulfated HS resides (the “diffusible zone”). Hh/Shh movement is effectively confined within the zone in two dimensions, with a gradient forming from posterior producing cells (right) to anterior receiving cells (left). Undersulfated HS are poor acceptors for Shh if already bound to higher sulfated HS, which explains that clones deficient in HSPG expression block Hh transport from HS-expressing cells in the Drosophila wing disc. In the absence of properly sulfated (sfl) anterior HS, or without HS (ttv), Hh clusters remain confined to cell surfaces with appropriately sulfated HS (reversed arrows)—representing the constricted diffusible zone. Fully functional GPI-linked Glp-HS chains at the plasma membrane are shown in dark blue; non-functional variants lacking HS chains or linked to undersulfated HS in light blue. P posterior, A anterior.