Inhibition of tetrameric Patched1 by Sonic Hedgehog through an asymmetric paradigm

The Hedgehog (Hh) pathway controls embryonic development and postnatal tissue maintenance and regeneration. Inhibition of Hh receptor Patched (Ptch) by the Hh ligands relieves suppression of signaling cascades. Here, we report the cryo-EM structure of tetrameric Ptch1 in complex with the palmitoylated N-terminal signaling domain of human Sonic hedgehog (ShhNp) at a 4:2 stoichiometric ratio. The structure shows that four Ptch1 protomers are organized as a loose dimer of dimers. Each dimer binds to one ShhNp through two distinct inhibitory interfaces, one mainly through the N-terminal peptide and the palmitoyl moiety of ShhNp and the other through the Ca2+-mediated interface on ShhNp. Map comparison reveals that the cholesteryl moiety of native ShhN occupies a recently identified extracellular steroid binding pocket in Ptch1. Our structure elucidates the tetrameric assembly of Ptch1 and suggests an asymmetric mode of action of the Hh ligands for inhibiting the potential cholesterol transport activity of Ptch1.


Introduction
The Hedgehog (Hh) precursor undergoes autocatalytic processing, and the resulting Nterminal signaling domain is modified with N-terminal palmitoylation and C-terminal  [4][5][6] . Deficient Hh signaling can lead to birth defects, whereas abnormal Hh signaling activation is associated with various human cancers [7][8][9] . Small molecule modulators have been identified to target this pathway, and inhibitors of the Hh signaling have been explored as potential therapeutics for cancer treatment 10,11 . The physiological importance of Ptch is underscored by its association with basal cell carcinoma, the most common cancer in humans 17,18 . The structure of Ptch1 was recently reported by several groups including ours [19][20][21] . The 1447 amino acids in human Ptch1 fold to a tripartite architecture containing two extracellular domains (ECD1 and ECD2), a 12-transmembrane helix (TM) transmembrane domain (TMD) that exhibits a two-fold pseudo-symmetry, and intracellular domains that were unresolved in all the reported structures.

Lacking direct interaction between the Hh receptor Ptch and the downstream Smo
Distinct from most bacterial RND transporters whose structures were resolved as trimers 22 , all the physiologically relevant Ptch1 structures were in monomeric form, despite Ptch1 was shown to be an oligomer in physiological condition 23 . The oligomerization of Ptch1 is mediated by the intracellular middle-loop domain (MLD) and C-terminal domain (CTD) 23 . It was reported that the CTD of Drosophila Ptc formed a trimer 24 . In contrast, another mammalian RND homolog human Niemann-Pick type C1 (NPC1) appears to be a monomer 25,26 . The oligomeric state of Ptch1 therefore requires further investigation.
In our cryo-EM structure of the monomeric human Ptch1 (the C-terminal half of CTD truncated) in complex with an unmodified N-terminal domain of Sonic Hh (ShhN), ShhN engages its concave side to bind to Ptch1 through extensive polar interactions.
Ensuing biochemical and structural characterizations reveal that formation of this interface requires binding of steroid to both the sterol-sensing domain (SSD) and an extracellular steroid binding site (ESBS) enclosed by ECD1 and ECD2 19 . We suggested that ShhN may relieve Smo inhibition by preventing conformational changes of Ptch1 that are required for its transport activity. Two weeks later, Qi et al published the structure of native lipid-modified ShhN (hereafter designated ShhNn) bound to a monomeric mutant Ptch1* (both MLD and CTD truncated) 21 . While the N-terminal palmitoyl moiety and the ensuing fragment of ShhNn bind to a pocket enclosed by ECD and TMD of Ptch1*, the globular protein domain of l ShhNn only has very limited contact Ptch1* through its convex side. Considering the oligomerization of Ptch1, binding and inhibition of oligomeric Ptch1 by ShhN is thereby more complex and remains to be investigated.

Complex of Ptch1 and palmitoylated ShhN obtained by co-expression
We have obtained an optimal human Ptch1 construct (residues 1-1305) that had markedly improved overexpression level and solution behavior compared to the full-length Ptch1 19 .
The major species of this Ptch1 construct existed in an oligomeric form upon sizeexclusion chromatography (SEC), although the minor monomeric form was used for cryo-EM analysis in our previous study. Cryo-samples made from the oligomeric peaks were highly heterogeneous, impeding structural determination to high resolution 19 . To overcome the heterogeneity of oligomeric Ptch1, we made several attempts, including screening of detergents and buffer conditions, engineering of protein with distinct boundaries and chemical cross-linking. Among all the endeavors, when glycol-diosgenin (GDN, Anatrace) was used for protein extraction and purification, the cryo-sample became amenable for cryo-EM analysis.
The details for protein generation can be found in Methods. Briefly, the human Ptch1 (residues 1-1305) with N-terminal FLAG tag and C-terminal His10 tag was coexpressed with untagged human ShhN (residues 1-197) in human embryonic kidney (HEK) 293F cells (Fig. 1a). After tandem affinity purification, the complex was eluted from SEC mainly in an oligomeric form with the elution volume similar to Ptch1 alone, consistent with the blue native PAGE (BN-PAGE) analysis (Fig. 1b). The signal peptide of ShhN expressed in HEK 293F cells was expected to be removed, and the resulting segment containing residues 24-197 should be palmitoylated but without C-terminal cholesterylation 3 . Mass spectrometric analysis confirmed the palmitoylation at Cys24 of ShhN in the co-expressed complex. We will refer to this protein as ShhNp.

Structural determination of tetrameric Ptch1 in complex with ShhNp
Protocols for grid preparation, cryo-EM data acquisition, and structural determination of oligomeric Ptch1 in complex with ShhNp are described in details in Fig. 1c The two 3D reconstructions of Ptch1-ShhNp complex revealed a similar 4:2 assembly (Fig. 1d). The four Ptch1 molecules are organized in two slightly different tetrameric forms, although both are dimer of dimers. One ShhNp molecule is positioned in the middle of each dimer and simultaneously interacts with the extracellular domains from two Ptch1 protomers (Fig. 1d). The two dimers have a relative rotation along an axis perpendicular to the membrane in the two distinct tetrameric assembles ( Fig. 1d and Extended Data Fig. 2d). The two dimers loosely contact each other through ECD2 in both tetrameric assemblies without any interaction in the TMDs ( Fig. 1d and Extended Data   Fig. 2c). The weak interactions may lead to structural flexibilities that limited the resolution of the cryo-EM reconstruction of the tetramer.

Structure of the 2:1 Ptch1-ShhNp complex
Considering that Ptch1 dimer with one ShhNp appears to be the basic unit within each 4:2 complex, we further applied adapted mask on one dimer to reduce the heterogeneity  Table 1). The intracellular segments remained poorly resolved likely due to their flexibility. Consistent with the previous report that the intracellular domains are responsible for the oligomerization of Ptch1 23 , the two Ptch1 protomers have no contact in the resolved structure (Fig. 2c).
One ShhNp simultaneously recognizes two Ptch1 through distinct interfaces ( Fig. 2c). To facilitate illustration, we named the two Ptch1 protomers A and B. Ptch1-A interacts with the pseudo-active site groove (the Ca 2+ -mediated interface) of ShhNp corresponding to the complex resolved by us 19 . Ptch1-B mainly accommodates the Nterminal fragment and palmitoyl moiety with limited contact with the globular domain of ShhNp, corresponding to the one in the Ptch1*-ShhNn complex 21 (Fig. 2d). The transmembrane domains of the two Ptch1 protomers stand in parallel within the membrane plane with the SSDs sandwiched inside (Fig. 2e).

The asymmetric binding of one ShhNp with two Ptch1
Superimposing the structures of the 2:1 Ptch1-ShhNp with our previous 1:1 Ptch1-ShhN (PDB code 6DMY) relative to Ptch1-A reveals nearly no change in Ptch1 (Fig. 3a). The interface between ShhNp and Ptch1-A remains nearly identical to what we have described previously 19 , hence will no longer be illustrated here.
Structural superimposition of our 2:1 complex with Ptch1*-ShhNn (PDB code 6D4J) revealed nearly identical structures for Ptch1, while ShhNp moved slightly away from Ptch1-B likely due to its interaction with Ptch1-A (Extended Data Fig. 4).
Examination of the deposited EM map for Ptch1*-ShhNn (EMDB code  showed that the local resolution for the interface mediated by the N-terminal segment of ShhNn and Ptch1* was ~ 5 Å, unable to support reliable analysis for detailed interactions. The local resolution in our current EM map enabled detailed analysis of the interactions between Ptch1-B and the N-terminal segment of ShhNp (Fig. 3b).
The N-terminal peptide (N15: 24CGPGRGFGKRRHPKK38) and the palmitoyl moiety of ShhNp inserts into a tunnel above TMD that is enclosed by ECD1 and ECD2 of Ptch1-B (Figs. 3b, 3c). The N15 segment mainly interacts with ECD1 of Ptch1-B through extensive polar interactions (Fig. 3b). The positively charged Cardin-Weintraub (CW) motif (residues 32-38) of ShhNp, which is important for binding to heparan sulfate proteoglycans (HSPGs) and Hh transport 27 , dominates the polar interactions with Ptch1 ( Fig. 3b). The palmitate binding pocket is formed by a number of hydrophobic residues from the TMD-ECD connecting elements, including the Linker1, Neck helix 2, Linker7 and Neck helix 8, and TMs 4/6/10/12 (Fig. 3c). The limited contact between the convex side of ShhNp and Ptch1-B involves the E loop and ensuing helix α3 on ECD1, in part overlapping with the interface between Ptch1-A and ShhNp (Fig. 3d).

The cholesteryl moiety of native ShhN inserts into the ESBS of Ptch1*-B
During the preparation of this manuscript, a structure of monomeric Ptch1* in complex with ShhNn at 2:1 stoichiometric ratio was published 28 . Although Ptch1* is a monomeric mutant, one ShhNn can still bind to two Ptch1*. Structural comparison of the two 2:1 complexes reveals nearly identical architecture, except the two Ptch1* molecules move towards each other relative to the ones in our Ptch1-ShhNp complex (Extended Data Fig.   5), possibly owing to the lack of stabilization by the intracellular domains in Ptch1*.
When the maps of the two 2:1 complexes were scrutinized, an extra EM density was founded in that of the Ptch1*-ShhNn complex (EMDB code EMD-8955) (Fig. 4a).
This density, which was not structurally assigned by the authors, is contiguous with the C-terminus of ShhNn and protrudes into a binding cavity corresponding to the ESBS of Ptch1*-B (Fig. 4a). We tentatively built a structural model for this C-terminal segment (C7: 191VAAKSGG197) and it became evident that the end density likely belongs to the cholesteryl moiety of ShhNn (Fig. 4b). The cholesteryl moiety is embedded in a cavity mainly enclosed by ECD1, where nearly 20 hydrophobic residues of ECD1 form the contour of the pocket (Fig. 4c). In support of cholesteryl moiety binding to the ESBS, two cholesterol-like densities were also observed in the similar positions of two ESBS in our structure (Extended Data Fig. 3e)

The different oligomeric organizations of Ptch1 and bacterial RND transporters
The twelve TMs of Ptch1 protomer exhibit identical fold with those in the bacterial RND transporters, although their oligomeric organizations are different (Extended Data Fig. 6).
Both the TMDs and ECDs of trimeric AcrB, the best characterized bacterial RND transporter, have extensive inter-protomer interactions to support trimer formation 29 .
Such organization provides structural basis for the coupled rotating mechanism among the three protomers during substrate transport 30,31 . In contrast, tetramerization of Ptch1 is mediated by the intracellular domains lacking interaction among the TMDs and ECDs.
Such architectural differences suggest that Ptch1 may not undergo the similar rotating mechanism as the bacterial RND transporters and the four Ptch1 protomers likely operate independently of each other. Supporting this notion, deletion of the intracellular domains led to monomerization of Ptch1, which nevertheless remained functional in cultured cell signaling assays 23 . ShhN binding can result in both Ptch1 inhibition and internalization from the primary cilia 14 . The tetrameric assembly of Ptch1 may influence the internalization efficiency, a caveat to be investigated. protomers and an extra one between SSD and ESBS in Ptch1-A, the corresponding site in Ptch1-B is occupied by the palmitoyl moiety (Extended Data Fig. 7a). A tunnel connecting the SSD and ESBS was observed in our previous structure 19 . In the present 2:1 complex, a similar tunnel stretches throughout the ECDs of Ptch1-A, but not Ptch1-B, that may represent the cholesterol transport path (Extended Data Fig. 7b). Nevertheless, there still lacks killer experiment to directly demonstrate the cholesterol transport activity by Ptch1. In addition, an unidentified extracellular cholesterol carrier may be required to take the cholesterol expelled by Ptch1.

Discussion
The asymmetric binding interfaces between one ShhNp and two Ptch1 reveal two distinct inhibitory mechanisms for the two Ptch1 protomers by one ShhNp. In both cases, ShhN binding would block the cholesterol transport activity of Ptch1. Inhibition of Ptch1-A was mediated by the globular domain of ShhNp through a Ca 2+ -mediated interface. We previously proposed that ShhN binding to ECD1 and ECD2 of Ptch1 could prevent the conformational changes of Ptch1 that are required for its transport activity 19 . Inhibition of Ptch1-B was mainly mediated by the palmitoylated N15 peptide, which blocks the tunnel connecting SSD and ESBS (Extended Data Fig. 7b). The structure provides a nice explanation for the recent study that a palmitoylated N-terminal 22-residue peptide could partially activate Hh signaling by binding to Ptch1 36 . An unexpected discovery is the insertion of the cholesteryl moiety into ESBS of Ptch1-B, which will naturally block binding of cholesterol to this site, further block the transport of cholesterol along this path.
No substantial conformational changes were observed between the TMDs of Ptch1 alone and the two Ptch1 molecules in Ptch1-ShhNp complex (Extended Data Fig. 7c), suggesting that ShhNp association do not affect cholesterol binding to TMD.
The palmitoyl and cholesteryl moieties render the Hh ligands as hydrophobic morphogens and several factors have been described as Hh chaperons to ensure its solubility, such as Hh itself to form soluble multimeric Hh 37,38 , lipoprotein particles 39,40 , and Scube proteins 41,42 . The structures suggest that Ptch1 binding could shield both the hydrophobic lipid moieties of ShhN from the aqueous environment (Fig. 5).
In sum, the structure of tetrameric Ptch1 in complex with palmitoylated ShhN at a 4:2 stoichiometric ratio further completes the molecular picture for the interplay between Ptch1 and Shh, and sets an important framework for future investigation of Hh signaling (Fig. 5). It is noted that the asymmetric binding between the surface receptor and the ligand at a 2:1 stoichiometric ratio has been discovered in several cases, such as growth factor receptor 43 , erythropoietin receptor 44 and netrin-1 receptor 45 . Such asymmetric signaling mechanism may be essential or more efficient in the signal transduction from ligands to the receptors. Considering the common dimeric or high-order oligomeric form of surface receptors, the substoichiometric ratio and asymmetric modes of actions between ligands and receptors may represent a paradigm that is more general than currently known.

Protein expression and purification
The cDNA of human Ptch1 (Uniprot: Q13635) (residues 1-1305) was cloned into the pCAG vector with an amino-terminal FLAG tag and a carboxy-terminal His10 tag. The The peak fractions for the oligomeric and monomeric Ptch1 or its complex with ShhN were separately collected.

Blue Native PAGE (BN-PAGE)
Chromatographically purified Ptch1 and Ptch1-ShhNp samples were mixed with 4 × loading buffer and Coomassie G-250 additive and then subjected to 3-12% NativePAGE TM Novex Bis-Tris gel (Invitrogen) for native electrophoresis at 4℃. The electrophoresis was conducted at 150V constant for 60 minutes, and then increase voltage to 250V constant for another 75 minutes. After electrophoresis, the gel was transferred to a container for fixation in fixing solution (40% methanol, 10% acetic acid), and then for staining (0.02% Coomassie R-250 in 30% methanol and 10% acetic acid) and destaining (8% acetic acid). All the procedures were performed according to the manufacturer's protocol.

Cryo-EM sample preparation and data collection
The cryo grids were prepared using Thermo Fisher Vitrobot Mark IV. The Quantifoil R1.2/1.3 Cu grids were firstly glow-discharged with air for 40 s at medium level in Plasma Cleaner (HARRICK PLASMA, PDC-32G-2). Then aliquots of 3.5 µl purified Ptch1-ShhN complex (concentrated to approximately 15 mg/ml) were applied to glowdischarged grids. After being blotted with filter paper for 3.5 s, the grids were plunged

Cryo-EM data processing
A total of 1,226,114 particles were automatically picked with RELION 2.0 49 . After 2D classification, a total of 448,682 particles were selected and subject to a guided multireference classification procedure. The references, two good and two bad, were generated with limited particles in advance. A total of 266,572 particles selected from multireferences 3D classification were subjected to local search 3D classification with adapted mask on the flexible half (to obtain a complete map) and performed five parallel runs at the same time. Then, a total of 107,265 particles were selected from good classes and subjected to seven parallel runs of local search 3D classification without adapted mask.
Two distinct classes were selected, Class I with 39,503 particles and Class II with 25,510 particles, yielding 3D reconstitutions with 6.8 Å and 6.5 Å, respectively. Lastly, ECD masks were applied to increase the local resolution to 4.6 Å and 4.3 Å for these two classes, respectively.
To increase the resolution of stable half, the 266,572 particles selected from multi-references 3D classification were subjected to a global angular search 3D classification with one class and 40 iteration. The outputs of the 30th-40th iterations were subjected to local angular search 3D classification with three classes separately. A total of 171,590 particles were selected by combining the good classes of the local angular search 3D classification, yielding a 3D reconstruction with an overall resolution of 3.6 Å after 3D auto-refinement with an adapted mask on the stable half.
All 2D classification, 3D classification, and 3D auto-refinement were performed with RELION 2.0. Resolutions were estimated with the gold-standard Fourier shell correlation 0.143 criterion 50 with high-resolution noise substitution.

Model building and refinement
Firstly, the map at 3.6 Å was used to build a 2:1 complex structure of Ptch1 and ShhN.
Two previously reported 1:1 complex structures (PDB code 6DMY and 6D4J) served as initial models to be docked into the map with Chimera, following by manual adjustment in Coot to generate the final structure. Then, two 2:1 structures were fitted into the maps of Class I or Class II to generate two complex structures at 4:2 stoichiometry.
All structure refinements were carried out by PHENIX 51 in real space with secondary structure and geometry restraints. Overfitting of the models was monitored by refining the model against one of the two independent half maps and testing the refined