Smoothened Goes Molecular: New Pieces in the Hedgehog Signaling Puzzle*

A general aim of studies of signal transduction is to identify mediators of specific signals, order them into pathways, and understand the nature of interactions between individual components and how these interactions alter pathway behavior. Despite years of intensive study and its central importance to animal development and human health, our understanding of the Hedgehog (Hh) signaling pathway remains riddled with gaps, question marks, assumptions, and poorly understood connections. In particular, understanding how interactions between Hh and Patched (Ptc), a 12-pass integral membrane protein, lead to modulation of the function of Smoothened (Smo), a 7-pass integral membrane protein, has defied standard biochemical characterization. Recent structural and biochemical characterizations of Smoothened domains have begun to unlock this riddle, however, and lay the groundwork for improved cancer therapies.

Members of the Hedgehog (Hh) 3 family of secreted signaling proteins are present in most metazoans and owe their name to the effects that loss of Hh function has on Drosophila embryos, which lose their normal segmented pattern and develop a uniform coat of bristles reminiscent of the coats of hedgehogs (1). As presaged by this phenotype, Hh proteins mediate essential tissue patterning events during many stages of animal development (2), and abnormal Hh function is associated with birth defects and cancer (3). Hh proteins are also involved in tissue maintenance and wound repair in adult animals (4). Hh proteins achieve their patterning effects by functioning as classical morphogens (5). That is, Hh proteins form gradients of decreasing concentration from sites of secretion and induce concentration-dependent differentiation of distinct cell types (6,7). As befits a morphogen, Hh expression, release, diffusion, and signal reception are tightly regulated by multiple factors (8).
Classical and modern genetic techniques have identified several cell-surface proteins and glycans involved in receiving or modifying Hh signals (9). The core components of this process, conserved in all organisms known to have active Hh signaling, are Patched (Ptc) and Smoothened (Smo) (Fig. 1) (10 -13). Ptc functions upstream of Smo and has been genetically and biochemically defined as a primary component of the Hh receptor (14,15). Ptc is a 12-pass integral membrane protein with distant homology to bacterial resistance-nodulation-cell division (RND) transporters (16,17). Transmembrane helices 2-6 of Ptc are also homologous to sterol-sensing domains, which are found in diverse integral membrane proteins and regulate activity in response to levels of free cellular sterols (18). Smo is a member of the Frizzled family (class F) of G-protein coupled receptors (GPCRs) (19), and contains an N-terminal, ϳ14-kDa extracellular cysteine-rich domain (CRD) connected via a linker to 7 membrane-spanning helices (7TM) and an extended (ϳ200 amino acids, human; ϳ450 amino acids, Drosophila) C-terminal tail.
Hh signaling responses are modulated by additional cell-surface components including CDO, BOC, Gas1, Hedgehog-interacting protein, and glypicans in vertebrates and Ihog, BOI, and the glypican Dally-like protein in flies (20 -29). These factors either lack intracellular regions because of glycophosphatidylinositol anchors (Gas1, glypicans) or have intracellular regions that are not implicated in Hh signaling and do not appear to transmit Hh signals across the cell membrane directly (14). Instead, transmission of Hh signals across the membrane appears to be mediated by Smo, the most downstream cellsurface component of the Hh signaling pathway. Consistent with this role, the cytoplasmic tail of Smo becomes heavily phosphorylated and likely changes disposition when the Hh pathway is active (30 -32). These changes are coupled to intracellular signaling events that ultimately converge on members of the Gli family of transcription factors, active forms of which translocate to the nucleus and up-regulate expression of target genes (33).
Recent discovery of the importance of Ptc and Smo localization for normal Hh signaling has added additional complexity to Hh pathway regulation. In vertebrates, Sonic Hh (Shh) and Hh pathway agonists result in movement of Smo from the plasma membrane to the primary cilium, a nonmotile flagellar-like organelle present on most cells, and dispersal of Ptc from its previous localization at the base of the primary cilium (34). Although movement of Smo to the primary cilium appears essential for normal Hh signaling in vertebrates (35), this movement is neither sufficient for signaling (36) nor conserved in flies (37), and a core signaling capacity that is independent of ciliary localization must be present in Smo. This minireview will focus on recent advances in structural and biochemical characterization of Smo, and readers are encouraged to consult other sources for background on additional Hh pathway components.

Patched and Smoothened
In unstimulated Hh-responsive cells, Ptc functions upstream of Smo to inhibit its activity (2). Hh triggers signaling responses by interacting with Ptc to relieve this inhibition, but both how Ptc inhibits Smo and how Hh relieves this inhibition remain unclear. As a small amount of Ptc is sufficient to inhibit a large stoichiometric excess of Smo (16), Ptc does not appear to inhibit Smo through a direct interaction. Rather, the homology of Ptc to transporters and the ability of Smo activity to be modulated by small molecules have led to the widespread belief that Ptc controls Smo through transport of a small molecule intermediary (16). Indeed, the ability of Smo to bind and be inhibited by the plant sterol cyclopamine led to the development of compounds targeting the cyclopamine-binding site for the treatment of cancers with abnormally active Hh signaling (38,39). As some Smo-binding compounds function as Hh pathway agonists, it has been tempting to speculate that an endogenous cyclopamine-like compound modulates Smo activity (40). Indeed, the sterol vitamin D3 has been proposed to function as a Ptc-dependent inhibitor of Smo (41), although this observation awaits confirmation.

Smoothened: 7TM Region
The absence of knowledge of the physiological factors responsible for Smo activation (or inhibition) has presented a frustrating barrier to understanding Hh pathway regulation, but several recent results have begun to clarify this issue. Firstly, Stevens and colleagues (42)(43)(44) have determined atomic resolution crystal structures of the 7TM region of human Smo complexed with five different small molecules, including cyclopamine. These landmark structures show that, despite sharing less than 10% sequence identity with class A GPCRs such as rhodopsin and the ␤ 2 -adrenergic receptor (␤ 2 AR), the Smo7 TM region adopts an overall confor-mation very similar to that of inactive class A GPCRs ( Fig. 2A) (45). As discussed in more detail below (46), this structural homology couples with the observation that activating mutations in Smo occur at sites that appear to stabilize the inactive state of class A GPCRs to suggest that the 7TM region of Smo is likely to undergo GPCR-like conformational changes during its activity cycle (45). Such a conformational cycle would also be consistent with the ability of Smo to signal through G-proteins in certain circumstances (47)(48)(49)(50)(51)(52).
Although the overall fold of its 7TM bundle is conserved with other GPCRs, Smo has additional features including an extension of extracellular loops (ECLs). All of the co-crystallized compounds bind Smo in a long narrow pocket formed by the ECL extensions and upper portions of the transmembrane domains ( Fig. 2) (42)(43)(44). The drug-binding pocket is exposed to the extracellular space, suggesting that drugs and any endogenous ligands access the pocket from the extracellular surface. This extracellular accessibility contrasts with a class A GPCR lipid receptor where the extracellular loops form a closed cage and ligand is thought to access its binding site from within the membrane (53). Although the CRD was deleted from the crystallized Smo 7TM domain, the majority of the residues of the extracellular linker between the CRD and the first TM domain are present and adopt an ordered structure. Disulfide bonds both within the linker and between the linker and the second extracellular loop appear to stabilize the linker structure ( Fig. 2A), and disruption of these disulfides results in increased Smoothened activity (54). In addition, the extracellular linker interacts with the extended extracellular loop connecting TMs VI and VII (ECL3), which forms a cap over the drug-binding pocket. This ordered linker region suggests that the CRD may be directly coupled to the 7TM region and influence its conformation. The five compounds crystallized in the Smo-binding pocket include an agonist (SAG1.5) and four antagonists (LY2940680, SANT1, ANTA XV, and cyclopamine) (see Fig. 4A). All ligands bind in the pocket with their long axes perpendicular to the plane of the membrane but vary in their depth relative to the extracellular outlet (Fig. 2B). At the extremes, cyclopamine interacts predominantly with the extracellular loops, whereas another antagonist, SANT-1, binds deep within the pocket, which spans 28 Å from the top of cyclopamine to the bottom of SANT1. Asp-473, a residue that when mutated to histidine confers resistance to the anti-cancer agent Vismodegib (GDC-0449) (55,56), lines the drug-binding pocket but interacts differently with different antagonists and does not confer universal drug resistance (43). Asp-473 does not directly contact LY2940680, for example, and the D473H substitution does not affect the activity of LY2940680 (57). The variable susceptibility of individual drugs to resistance mutations suggests that second generation drugs or combination therapies may prolong the time to development of resistance.
LY2940680, cyclopamine, ANTA XV, and the agonist SAG1.5 contact the Smoothened extracellular loops lining the top of the ligand-binding cavity, but SANT1 binds more deeply in the pocket and only contacts ECL2, which is positioned within the 7TM region. In contrast to cyclopamine, which binds more tightly to   Variable specificity for 20(S)-OHC among Smo CRDs is perhaps not surprising given that the absence of a cellular sterol hydroxylase known to produce it makes it unlikely to be an endogenous ligand (61). Assuming that endogenous ligands for Smo CRDs exist, the question naturally arises of what that ligand is. A survey of oxysterols for Smo modulatory activity found that 7-keto-27-OHC and 7-keto-25-OHC, both metabolites of 7-ketocholesterol, are able to stimulate Hh signaling in a manner that depends on the presence of the CRD (61). Compounds that bind the CRD and inhibit (azasterols, e.g. 22-azacholesterol) (Fig. 4B) or partially agonize (20(R)-yne, 20-keto-yne) Smo activity validate the Smo CRD as a potential drug target and raise the possibility that an endogenous ligand for the Smo CRD may be an inhibitor rather than an activator (62,63). More work is needed to identify and validate potential CRD ligands, but it seems likely that such ligands exist, and their discovery and characterization will take our understanding of Hh pathway regulation in new directions.

Smoothened: Cysteine-rich Domain
Two immediately exciting prospects stimulated by the discovery of a specific and functionally important sterol-binding site on the Smo CRD were that it might be the route by which Ptc modulates Smo activity or that it might rationalize why cholesterol depletion reduces Hh signaling (71). Defying Occam's razor, however, oxysterol binding by the Smo CRD cannot fully account for either of these processes. Deletion of the CRD from Smo (⌬CRDSmo) alters but does not abolish Shh-mediated pathway activation (61)(62)(63). Although varying levels of responsiveness of ⌬CRDSmo to Shh have been reported, this is likely due to varying tags and expression systems. Rohatgi and colleagues (63) showed that oxysterol-binding mutants of Smo retain negative regulation by Ptc and respond to Shh. Beachy and colleagues (61) showed that deletion of the CRD increases basal Smo activity, but this activity can be reduced by co-expression with Ptc and restored by addition of Shh, indicating that Ptc can exert its effects on Smo independent of the CRD. Higher basal activity and Shh responsiveness of ⌬CRDSmo were also reduced by cyclodextrin depletion, which reduces wild-type Smo activity (71), suggesting that the CRD is also not essential for this process but rather that cholesterol within the cell membrane is needed for normal Smo function. Indeed, a specific role for membrane-localized cholesterol in Smo modulation has been suggested (61), although no ordered cholesterol molecules were identified in the Smo crystal structures. Modulation of Smo activity independent of the CRD or cyclopamine-binding pocket is not unprecedented as Itraconazole (Fig. 4C) acts on Smo at a site distinct from both the canonical 7TM pocket and the CRD to inhibit Hh pathway activity (72).
Cholesterol binding to the 7TM region of GPCRs is also not unprecedented. The structure of ␤ 2 AR bound to cholesterol and the partial inverse agonist timolol led Stevens and colleagues (73)

Targeting Smoothened in the Clinic
Hh pathway-activating mutations in the gene encoding Ptc, and less commonly the gene encoding Smo, are found in subsets of several cancers, most notably basal cell carcinoma (BCC) and pediatric medulloblastomas (46,75). Constitutively active mutants of Smo found in sporadic BCC (W535L 7.55 "SmoM2") and more recently in meningiomas and ameloblastomas (W535L 7.55 , L412F 5.51 ) are resistant to Vismodegib treatment (46, 76 -78). Superscripts refer to Ballesteros-Weinstein numbering. Trp-535 7.55 is absolutely conserved in class F GPCRs and maps to the intracellular tip of TM VII, a region structurally homologous to the NPXXY motif in class A GPCRs (79,80). Trp-535 7.55 overlays with the Tyr 7.53 of the NPXXY motif, which undergoes rearrangement in inactive versus active structures of class A GPCRs (60,81,82), Leu-412 5.51 is highly conserved across class F GCPRs and also appears in a conformationally labile region of GPCRs. In class A GPCRs, residue 5.51 is one of a group of conserved hydrophobic and aromatic residues (3.40, 5.51, 6.44, 6.48) thought to constitute a "transmission switch" that rearranges when agonist binds (45,83). Collectively, these constitutively active mutants bolster the notion that Smo cycles through canonical GPCR inactive-active states.
Vismodegib is a Smo inhibitor that binds to the 7TM pocket (Fig. 4) and has been approved for the treatment of advanced BCC. Resistance to Vismodegib usually appears within a few months, however (84). Cancers with active Hh signaling are often driven by inactivating Ptc mutations, but resistance mutations often appear in Smo, the target of the drug. The Vismodegib resistance mutation originally found in medulloblastoma, D473H (55) 3.32 is further buried at the base of the binding pocket and 5.8 Å from SANT1 at its closest point. It is not known whether these mutations function to disrupt binding of Vismodegib to Smo or to activate Smo, but its position in the Smo structure suggests that W281L is more likely to interfere with ligand binding than V321M. Given the rapid resistance to drugs targeting the Smo 7TM pocket, antagonists that bind the Smo CRD hold out the hope that drugs targeting the CRD may prove more effective or less susceptible to resistance when used either alone or in combination with compounds targeting the Smo 7TM pocket (62,63).
Any discussion of the Smo 7TM and CRD regions naturally leads to questions concerning how these components interact and how their interplay affects the Smo C-terminal tail. Little is known about the structure of the Smo C-tail alone or with the Smo 7TM bundle, but its low complexity and high hydrophilicity suggest that it does not adopt a rigid globular structure. The Smo C-tail is phosphorylated in response to pathway activation, although the identities of the kinases responsible for phosphorylation differ between vertebrates and invertebrates (31,87). A conformational change of the Drosophila Smo C-tail has been proposed to stem from C-tail phosphorylation altering interactions between positively charged clusters of Arg residues and negatively charged clusters of Asp residues (32), but the verte- brate Smo C-tail does not possess the Arg clusters. A C-tail conformational change in vertebrates has also been proposed, however (88).

Conclusion
Multiple inputs (oxysterol binding to the CRD, small molecule binding to the 7TM pocket, and sterols within the cell membrane) are all capable of modulating Smo activity and presumably conformation. Sorting out what the endogenous inputs are, which of these inputs are important in specific instances, how multiple inputs are integrated, how best to exploit various ways of modulating Smo for anticancer therapies, and the role of Ptc in modulating these inputs present exciting challenges. Recent results have helped clarify the nature and sites of these inputs, however, and provided a framework for understanding how each of the parts fit together to achieve remarkable biological results.